summaryrefslogtreecommitdiffstats
path: root/freebsd/sys/dev/bce/if_bce.c
diff options
context:
space:
mode:
authorSebastian Huber <sebastian.huber@embedded-brains.de>2013-10-09 22:42:09 +0200
committerSebastian Huber <sebastian.huber@embedded-brains.de>2013-10-10 09:06:58 +0200
commitbceabc95c1c85d793200446fa85f1ddc6313ea29 (patch)
tree973c8bd8deca9fd69913f2895cc91e0e6114d46c /freebsd/sys/dev/bce/if_bce.c
parentAdd FreeBSD sources as a submodule (diff)
downloadrtems-libbsd-bceabc95c1c85d793200446fa85f1ddc6313ea29.tar.bz2
Move files to match FreeBSD layout
Diffstat (limited to 'freebsd/sys/dev/bce/if_bce.c')
-rw-r--r--freebsd/sys/dev/bce/if_bce.c10847
1 files changed, 10847 insertions, 0 deletions
diff --git a/freebsd/sys/dev/bce/if_bce.c b/freebsd/sys/dev/bce/if_bce.c
new file mode 100644
index 00000000..646e6cbf
--- /dev/null
+++ b/freebsd/sys/dev/bce/if_bce.c
@@ -0,0 +1,10847 @@
+#include <freebsd/machine/rtems-bsd-config.h>
+
+/*-
+ * Copyright (c) 2006-2010 Broadcom Corporation
+ * David Christensen <davidch@broadcom.com>. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of Broadcom Corporation nor the name of its contributors
+ * may be used to endorse or promote products derived from this software
+ * without specific prior written consent.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS'
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+ * THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <freebsd/sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+/*
+ * The following controllers are supported by this driver:
+ * BCM5706C A2, A3
+ * BCM5706S A2, A3
+ * BCM5708C B1, B2
+ * BCM5708S B1, B2
+ * BCM5709C A1, C0
+ * BCM5709S A1, C0
+ * BCM5716C C0
+ * BCM5716S C0
+ *
+ * The following controllers are not supported by this driver:
+ * BCM5706C A0, A1 (pre-production)
+ * BCM5706S A0, A1 (pre-production)
+ * BCM5708C A0, B0 (pre-production)
+ * BCM5708S A0, B0 (pre-production)
+ * BCM5709C A0 B0, B1, B2 (pre-production)
+ * BCM5709S A0, B0, B1, B2 (pre-production)
+ */
+
+#include <freebsd/local/opt_bce.h>
+
+#include <freebsd/dev/bce/if_bcereg.h>
+#include <freebsd/dev/bce/if_bcefw.h>
+
+/****************************************************************************/
+/* BCE Debug Options */
+/****************************************************************************/
+#ifdef BCE_DEBUG
+ u32 bce_debug = BCE_WARN;
+
+ /* 0 = Never */
+ /* 1 = 1 in 2,147,483,648 */
+ /* 256 = 1 in 8,388,608 */
+ /* 2048 = 1 in 1,048,576 */
+ /* 65536 = 1 in 32,768 */
+ /* 1048576 = 1 in 2,048 */
+ /* 268435456 = 1 in 8 */
+ /* 536870912 = 1 in 4 */
+ /* 1073741824 = 1 in 2 */
+
+ /* Controls how often the l2_fhdr frame error check will fail. */
+ int l2fhdr_error_sim_control = 0;
+
+ /* Controls how often the unexpected attention check will fail. */
+ int unexpected_attention_sim_control = 0;
+
+ /* Controls how often to simulate an mbuf allocation failure. */
+ int mbuf_alloc_failed_sim_control = 0;
+
+ /* Controls how often to simulate a DMA mapping failure. */
+ int dma_map_addr_failed_sim_control = 0;
+
+ /* Controls how often to simulate a bootcode failure. */
+ int bootcode_running_failure_sim_control = 0;
+#endif
+
+/****************************************************************************/
+/* BCE Build Time Options */
+/****************************************************************************/
+/* #define BCE_NVRAM_WRITE_SUPPORT 1 */
+
+
+/****************************************************************************/
+/* PCI Device ID Table */
+/* */
+/* Used by bce_probe() to identify the devices supported by this driver. */
+/****************************************************************************/
+#define BCE_DEVDESC_MAX 64
+
+static struct bce_type bce_devs[] = {
+ /* BCM5706C Controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706, HP_VENDORID, 0x3101,
+ "HP NC370T Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706, HP_VENDORID, 0x3106,
+ "HP NC370i Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706, HP_VENDORID, 0x3070,
+ "HP NC380T PCIe DP Multifunc Gig Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706, HP_VENDORID, 0x1709,
+ "HP NC371i Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5706 1000Base-T" },
+
+ /* BCM5706S controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706S, HP_VENDORID, 0x3102,
+ "HP NC370F Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5706S, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5706 1000Base-SX" },
+
+ /* BCM5708C controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708, HP_VENDORID, 0x7037,
+ "HP NC373T PCIe Multifunction Gig Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708, HP_VENDORID, 0x7038,
+ "HP NC373i Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708, HP_VENDORID, 0x7045,
+ "HP NC374m PCIe Multifunction Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5708 1000Base-T" },
+
+ /* BCM5708S controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708S, HP_VENDORID, 0x1706,
+ "HP NC373m Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708S, HP_VENDORID, 0x703b,
+ "HP NC373i Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708S, HP_VENDORID, 0x703d,
+ "HP NC373F PCIe Multifunc Giga Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5708S, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5708 1000Base-SX" },
+
+ /* BCM5709C controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5709, HP_VENDORID, 0x7055,
+ "HP NC382i DP Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5709, HP_VENDORID, 0x7059,
+ "HP NC382T PCIe DP Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5709, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5709 1000Base-T" },
+
+ /* BCM5709S controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5709S, HP_VENDORID, 0x171d,
+ "HP NC382m DP 1GbE Multifunction BL-c Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5709S, HP_VENDORID, 0x7056,
+ "HP NC382i DP Multifunction Gigabit Server Adapter" },
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5709S, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5709 1000Base-SX" },
+
+ /* BCM5716 controllers and OEM boards. */
+ { BRCM_VENDORID, BRCM_DEVICEID_BCM5716, PCI_ANY_ID, PCI_ANY_ID,
+ "Broadcom NetXtreme II BCM5716 1000Base-T" },
+
+ { 0, 0, 0, 0, NULL }
+};
+
+
+/****************************************************************************/
+/* Supported Flash NVRAM device data. */
+/****************************************************************************/
+static struct flash_spec flash_table[] =
+{
+#define BUFFERED_FLAGS (BCE_NV_BUFFERED | BCE_NV_TRANSLATE)
+#define NONBUFFERED_FLAGS (BCE_NV_WREN)
+
+ /* Slow EEPROM */
+ {0x00000000, 0x40830380, 0x009f0081, 0xa184a053, 0xaf000400,
+ BUFFERED_FLAGS, SEEPROM_PAGE_BITS, SEEPROM_PAGE_SIZE,
+ SEEPROM_BYTE_ADDR_MASK, SEEPROM_TOTAL_SIZE,
+ "EEPROM - slow"},
+ /* Expansion entry 0001 */
+ {0x08000002, 0x4b808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 0001"},
+ /* Saifun SA25F010 (non-buffered flash) */
+ /* strap, cfg1, & write1 need updates */
+ {0x04000001, 0x47808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, SAIFUN_FLASH_BASE_TOTAL_SIZE*2,
+ "Non-buffered flash (128kB)"},
+ /* Saifun SA25F020 (non-buffered flash) */
+ /* strap, cfg1, & write1 need updates */
+ {0x0c000003, 0x4f808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, SAIFUN_FLASH_BASE_TOTAL_SIZE*4,
+ "Non-buffered flash (256kB)"},
+ /* Expansion entry 0100 */
+ {0x11000000, 0x53808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 0100"},
+ /* Entry 0101: ST M45PE10 (non-buffered flash, TetonII B0) */
+ {0x19000002, 0x5b808201, 0x000500db, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, ST_MICRO_FLASH_PAGE_BITS, ST_MICRO_FLASH_PAGE_SIZE,
+ ST_MICRO_FLASH_BYTE_ADDR_MASK, ST_MICRO_FLASH_BASE_TOTAL_SIZE*2,
+ "Entry 0101: ST M45PE10 (128kB non-bufferred)"},
+ /* Entry 0110: ST M45PE20 (non-buffered flash)*/
+ {0x15000001, 0x57808201, 0x000500db, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, ST_MICRO_FLASH_PAGE_BITS, ST_MICRO_FLASH_PAGE_SIZE,
+ ST_MICRO_FLASH_BYTE_ADDR_MASK, ST_MICRO_FLASH_BASE_TOTAL_SIZE*4,
+ "Entry 0110: ST M45PE20 (256kB non-bufferred)"},
+ /* Saifun SA25F005 (non-buffered flash) */
+ /* strap, cfg1, & write1 need updates */
+ {0x1d000003, 0x5f808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, SAIFUN_FLASH_BASE_TOTAL_SIZE,
+ "Non-buffered flash (64kB)"},
+ /* Fast EEPROM */
+ {0x22000000, 0x62808380, 0x009f0081, 0xa184a053, 0xaf000400,
+ BUFFERED_FLAGS, SEEPROM_PAGE_BITS, SEEPROM_PAGE_SIZE,
+ SEEPROM_BYTE_ADDR_MASK, SEEPROM_TOTAL_SIZE,
+ "EEPROM - fast"},
+ /* Expansion entry 1001 */
+ {0x2a000002, 0x6b808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 1001"},
+ /* Expansion entry 1010 */
+ {0x26000001, 0x67808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 1010"},
+ /* ATMEL AT45DB011B (buffered flash) */
+ {0x2e000003, 0x6e808273, 0x00570081, 0x68848353, 0xaf000400,
+ BUFFERED_FLAGS, BUFFERED_FLASH_PAGE_BITS, BUFFERED_FLASH_PAGE_SIZE,
+ BUFFERED_FLASH_BYTE_ADDR_MASK, BUFFERED_FLASH_TOTAL_SIZE,
+ "Buffered flash (128kB)"},
+ /* Expansion entry 1100 */
+ {0x33000000, 0x73808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 1100"},
+ /* Expansion entry 1101 */
+ {0x3b000002, 0x7b808201, 0x00050081, 0x03840253, 0xaf020406,
+ NONBUFFERED_FLAGS, SAIFUN_FLASH_PAGE_BITS, SAIFUN_FLASH_PAGE_SIZE,
+ SAIFUN_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 1101"},
+ /* Ateml Expansion entry 1110 */
+ {0x37000001, 0x76808273, 0x00570081, 0x68848353, 0xaf000400,
+ BUFFERED_FLAGS, BUFFERED_FLASH_PAGE_BITS, BUFFERED_FLASH_PAGE_SIZE,
+ BUFFERED_FLASH_BYTE_ADDR_MASK, 0,
+ "Entry 1110 (Atmel)"},
+ /* ATMEL AT45DB021B (buffered flash) */
+ {0x3f000003, 0x7e808273, 0x00570081, 0x68848353, 0xaf000400,
+ BUFFERED_FLAGS, BUFFERED_FLASH_PAGE_BITS, BUFFERED_FLASH_PAGE_SIZE,
+ BUFFERED_FLASH_BYTE_ADDR_MASK, BUFFERED_FLASH_TOTAL_SIZE*2,
+ "Buffered flash (256kB)"},
+};
+
+/*
+ * The BCM5709 controllers transparently handle the
+ * differences between Atmel 264 byte pages and all
+ * flash devices which use 256 byte pages, so no
+ * logical-to-physical mapping is required in the
+ * driver.
+ */
+static struct flash_spec flash_5709 = {
+ .flags = BCE_NV_BUFFERED,
+ .page_bits = BCM5709_FLASH_PAGE_BITS,
+ .page_size = BCM5709_FLASH_PAGE_SIZE,
+ .addr_mask = BCM5709_FLASH_BYTE_ADDR_MASK,
+ .total_size = BUFFERED_FLASH_TOTAL_SIZE * 2,
+ .name = "5709/5716 buffered flash (256kB)",
+};
+
+
+/****************************************************************************/
+/* FreeBSD device entry points. */
+/****************************************************************************/
+static int bce_probe (device_t);
+static int bce_attach (device_t);
+static int bce_detach (device_t);
+static int bce_shutdown (device_t);
+
+
+/****************************************************************************/
+/* BCE Debug Data Structure Dump Routines */
+/****************************************************************************/
+#ifdef BCE_DEBUG
+static u32 bce_reg_rd (struct bce_softc *, u32);
+static void bce_reg_wr (struct bce_softc *, u32, u32);
+static void bce_reg_wr16 (struct bce_softc *, u32, u16);
+static u32 bce_ctx_rd (struct bce_softc *, u32, u32);
+static void bce_dump_enet (struct bce_softc *, struct mbuf *);
+static void bce_dump_mbuf (struct bce_softc *, struct mbuf *);
+static void bce_dump_tx_mbuf_chain (struct bce_softc *, u16, int);
+static void bce_dump_rx_mbuf_chain (struct bce_softc *, u16, int);
+#ifdef BCE_JUMBO_HDRSPLIT
+static void bce_dump_pg_mbuf_chain (struct bce_softc *, u16, int);
+#endif
+static void bce_dump_txbd (struct bce_softc *,
+ int, struct tx_bd *);
+static void bce_dump_rxbd (struct bce_softc *,
+ int, struct rx_bd *);
+#ifdef BCE_JUMBO_HDRSPLIT
+static void bce_dump_pgbd (struct bce_softc *,
+ int, struct rx_bd *);
+#endif
+static void bce_dump_l2fhdr (struct bce_softc *,
+ int, struct l2_fhdr *);
+static void bce_dump_ctx (struct bce_softc *, u16);
+static void bce_dump_ftqs (struct bce_softc *);
+static void bce_dump_tx_chain (struct bce_softc *, u16, int);
+static void bce_dump_rx_bd_chain (struct bce_softc *, u16, int);
+#ifdef BCE_JUMBO_HDRSPLIT
+static void bce_dump_pg_chain (struct bce_softc *, u16, int);
+#endif
+static void bce_dump_status_block (struct bce_softc *);
+static void bce_dump_stats_block (struct bce_softc *);
+static void bce_dump_driver_state (struct bce_softc *);
+static void bce_dump_hw_state (struct bce_softc *);
+static void bce_dump_mq_regs (struct bce_softc *);
+static void bce_dump_bc_state (struct bce_softc *);
+static void bce_dump_txp_state (struct bce_softc *, int);
+static void bce_dump_rxp_state (struct bce_softc *, int);
+static void bce_dump_tpat_state (struct bce_softc *, int);
+static void bce_dump_cp_state (struct bce_softc *, int);
+static void bce_dump_com_state (struct bce_softc *, int);
+static void bce_dump_rv2p_state (struct bce_softc *);
+static void bce_breakpoint (struct bce_softc *);
+#endif
+
+
+/****************************************************************************/
+/* BCE Register/Memory Access Routines */
+/****************************************************************************/
+static u32 bce_reg_rd_ind (struct bce_softc *, u32);
+static void bce_reg_wr_ind (struct bce_softc *, u32, u32);
+static void bce_shmem_wr (struct bce_softc *, u32, u32);
+static u32 bce_shmem_rd (struct bce_softc *, u32);
+static void bce_ctx_wr (struct bce_softc *, u32, u32, u32);
+static int bce_miibus_read_reg (device_t, int, int);
+static int bce_miibus_write_reg (device_t, int, int, int);
+static void bce_miibus_statchg (device_t);
+
+
+/****************************************************************************/
+/* BCE NVRAM Access Routines */
+/****************************************************************************/
+static int bce_acquire_nvram_lock (struct bce_softc *);
+static int bce_release_nvram_lock (struct bce_softc *);
+static void bce_enable_nvram_access (struct bce_softc *);
+static void bce_disable_nvram_access (struct bce_softc *);
+static int bce_nvram_read_dword (struct bce_softc *, u32, u8 *, u32);
+static int bce_init_nvram (struct bce_softc *);
+static int bce_nvram_read (struct bce_softc *, u32, u8 *, int);
+static int bce_nvram_test (struct bce_softc *);
+#ifdef BCE_NVRAM_WRITE_SUPPORT
+static int bce_enable_nvram_write (struct bce_softc *);
+static void bce_disable_nvram_write (struct bce_softc *);
+static int bce_nvram_erase_page (struct bce_softc *, u32);
+static int bce_nvram_write_dword (struct bce_softc *, u32, u8 *, u32);
+static int bce_nvram_write (struct bce_softc *, u32, u8 *, int);
+#endif
+
+/****************************************************************************/
+/* */
+/****************************************************************************/
+static void bce_get_media (struct bce_softc *);
+static void bce_init_media (struct bce_softc *);
+static void bce_dma_map_addr (void *,
+ bus_dma_segment_t *, int, int);
+static int bce_dma_alloc (device_t);
+static void bce_dma_free (struct bce_softc *);
+static void bce_release_resources (struct bce_softc *);
+
+/****************************************************************************/
+/* BCE Firmware Synchronization and Load */
+/****************************************************************************/
+static int bce_fw_sync (struct bce_softc *, u32);
+static void bce_load_rv2p_fw (struct bce_softc *, u32 *, u32, u32);
+static void bce_load_cpu_fw (struct bce_softc *,
+ struct cpu_reg *, struct fw_info *);
+static void bce_start_cpu (struct bce_softc *, struct cpu_reg *);
+static void bce_halt_cpu (struct bce_softc *, struct cpu_reg *);
+static void bce_start_rxp_cpu (struct bce_softc *);
+static void bce_init_rxp_cpu (struct bce_softc *);
+static void bce_init_txp_cpu (struct bce_softc *);
+static void bce_init_tpat_cpu (struct bce_softc *);
+static void bce_init_cp_cpu (struct bce_softc *);
+static void bce_init_com_cpu (struct bce_softc *);
+static void bce_init_cpus (struct bce_softc *);
+
+static void bce_print_adapter_info (struct bce_softc *);
+static void bce_probe_pci_caps (device_t, struct bce_softc *);
+static void bce_stop (struct bce_softc *);
+static int bce_reset (struct bce_softc *, u32);
+static int bce_chipinit (struct bce_softc *);
+static int bce_blockinit (struct bce_softc *);
+
+static int bce_init_tx_chain (struct bce_softc *);
+static void bce_free_tx_chain (struct bce_softc *);
+
+static int bce_get_rx_buf (struct bce_softc *,
+ struct mbuf *, u16 *, u16 *, u32 *);
+static int bce_init_rx_chain (struct bce_softc *);
+static void bce_fill_rx_chain (struct bce_softc *);
+static void bce_free_rx_chain (struct bce_softc *);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+static int bce_get_pg_buf (struct bce_softc *,
+ struct mbuf *, u16 *, u16 *);
+static int bce_init_pg_chain (struct bce_softc *);
+static void bce_fill_pg_chain (struct bce_softc *);
+static void bce_free_pg_chain (struct bce_softc *);
+#endif
+
+static struct mbuf *bce_tso_setup (struct bce_softc *,
+ struct mbuf **, u16 *);
+static int bce_tx_encap (struct bce_softc *, struct mbuf **);
+static void bce_start_locked (struct ifnet *);
+static void bce_start (struct ifnet *);
+static int bce_ioctl (struct ifnet *, u_long, caddr_t);
+static void bce_watchdog (struct bce_softc *);
+static int bce_ifmedia_upd (struct ifnet *);
+static int bce_ifmedia_upd_locked (struct ifnet *);
+static void bce_ifmedia_sts (struct ifnet *, struct ifmediareq *);
+static void bce_init_locked (struct bce_softc *);
+static void bce_init (void *);
+static void bce_mgmt_init_locked (struct bce_softc *sc);
+
+static int bce_init_ctx (struct bce_softc *);
+static void bce_get_mac_addr (struct bce_softc *);
+static void bce_set_mac_addr (struct bce_softc *);
+static void bce_phy_intr (struct bce_softc *);
+static inline u16 bce_get_hw_rx_cons (struct bce_softc *);
+static void bce_rx_intr (struct bce_softc *);
+static void bce_tx_intr (struct bce_softc *);
+static void bce_disable_intr (struct bce_softc *);
+static void bce_enable_intr (struct bce_softc *, int);
+
+static void bce_intr (void *);
+static void bce_set_rx_mode (struct bce_softc *);
+static void bce_stats_update (struct bce_softc *);
+static void bce_tick (void *);
+static void bce_pulse (void *);
+static void bce_add_sysctls (struct bce_softc *);
+
+
+/****************************************************************************/
+/* FreeBSD device dispatch table. */
+/****************************************************************************/
+static device_method_t bce_methods[] = {
+ /* Device interface (device_if.h) */
+ DEVMETHOD(device_probe, bce_probe),
+ DEVMETHOD(device_attach, bce_attach),
+ DEVMETHOD(device_detach, bce_detach),
+ DEVMETHOD(device_shutdown, bce_shutdown),
+/* Supported by device interface but not used here. */
+/* DEVMETHOD(device_identify, bce_identify), */
+/* DEVMETHOD(device_suspend, bce_suspend), */
+/* DEVMETHOD(device_resume, bce_resume), */
+/* DEVMETHOD(device_quiesce, bce_quiesce), */
+
+ /* Bus interface (bus_if.h) */
+ DEVMETHOD(bus_print_child, bus_generic_print_child),
+ DEVMETHOD(bus_driver_added, bus_generic_driver_added),
+
+ /* MII interface (miibus_if.h) */
+ DEVMETHOD(miibus_readreg, bce_miibus_read_reg),
+ DEVMETHOD(miibus_writereg, bce_miibus_write_reg),
+ DEVMETHOD(miibus_statchg, bce_miibus_statchg),
+/* Supported by MII interface but not used here. */
+/* DEVMETHOD(miibus_linkchg, bce_miibus_linkchg), */
+/* DEVMETHOD(miibus_mediainit, bce_miibus_mediainit), */
+
+ { 0, 0 }
+};
+
+static driver_t bce_driver = {
+ "bce",
+ bce_methods,
+ sizeof(struct bce_softc)
+};
+
+static devclass_t bce_devclass;
+
+MODULE_DEPEND(bce, pci, 1, 1, 1);
+MODULE_DEPEND(bce, ether, 1, 1, 1);
+MODULE_DEPEND(bce, miibus, 1, 1, 1);
+
+DRIVER_MODULE(bce, pci, bce_driver, bce_devclass, 0, 0);
+DRIVER_MODULE(miibus, bce, miibus_driver, miibus_devclass, 0, 0);
+
+
+/****************************************************************************/
+/* Tunable device values */
+/****************************************************************************/
+SYSCTL_NODE(_hw, OID_AUTO, bce, CTLFLAG_RD, 0, "bce driver parameters");
+
+/* Allowable values are TRUE or FALSE */
+static int bce_tso_enable = TRUE;
+TUNABLE_INT("hw.bce.tso_enable", &bce_tso_enable);
+SYSCTL_UINT(_hw_bce, OID_AUTO, tso_enable, CTLFLAG_RDTUN, &bce_tso_enable, 0,
+"TSO Enable/Disable");
+
+/* Allowable values are 0 (IRQ), 1 (MSI/IRQ), and 2 (MSI-X/MSI/IRQ) */
+/* ToDo: Add MSI-X support. */
+static int bce_msi_enable = 1;
+TUNABLE_INT("hw.bce.msi_enable", &bce_msi_enable);
+SYSCTL_UINT(_hw_bce, OID_AUTO, msi_enable, CTLFLAG_RDTUN, &bce_msi_enable, 0,
+"MSI-X|MSI|INTx selector");
+
+/* ToDo: Add tunable to enable/disable strict MTU handling. */
+/* Currently allows "loose" RX MTU checking (i.e. sets the */
+/* H/W RX MTU to the size of the largest receive buffer, or */
+/* 2048 bytes). This will cause a UNH failure but is more */
+/* desireable from a functional perspective. */
+
+
+/****************************************************************************/
+/* Device probe function. */
+/* */
+/* Compares the device to the driver's list of supported devices and */
+/* reports back to the OS whether this is the right driver for the device. */
+/* */
+/* Returns: */
+/* BUS_PROBE_DEFAULT on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_probe(device_t dev)
+{
+ struct bce_type *t;
+ struct bce_softc *sc;
+ char *descbuf;
+ u16 vid = 0, did = 0, svid = 0, sdid = 0;
+
+ t = bce_devs;
+
+ sc = device_get_softc(dev);
+ bzero(sc, sizeof(struct bce_softc));
+ sc->bce_unit = device_get_unit(dev);
+ sc->bce_dev = dev;
+
+ /* Get the data for the device to be probed. */
+ vid = pci_get_vendor(dev);
+ did = pci_get_device(dev);
+ svid = pci_get_subvendor(dev);
+ sdid = pci_get_subdevice(dev);
+
+ DBPRINT(sc, BCE_EXTREME_LOAD,
+ "%s(); VID = 0x%04X, DID = 0x%04X, SVID = 0x%04X, "
+ "SDID = 0x%04X\n", __FUNCTION__, vid, did, svid, sdid);
+
+ /* Look through the list of known devices for a match. */
+ while(t->bce_name != NULL) {
+
+ if ((vid == t->bce_vid) && (did == t->bce_did) &&
+ ((svid == t->bce_svid) || (t->bce_svid == PCI_ANY_ID)) &&
+ ((sdid == t->bce_sdid) || (t->bce_sdid == PCI_ANY_ID))) {
+
+ descbuf = malloc(BCE_DEVDESC_MAX, M_TEMP, M_NOWAIT);
+
+ if (descbuf == NULL)
+ return(ENOMEM);
+
+ /* Print out the device identity. */
+ snprintf(descbuf, BCE_DEVDESC_MAX, "%s (%c%d)",
+ t->bce_name, (((pci_read_config(dev,
+ PCIR_REVID, 4) & 0xf0) >> 4) + 'A'),
+ (pci_read_config(dev, PCIR_REVID, 4) & 0xf));
+
+ device_set_desc_copy(dev, descbuf);
+ free(descbuf, M_TEMP);
+ return(BUS_PROBE_DEFAULT);
+ }
+ t++;
+ }
+
+ return(ENXIO);
+}
+
+
+/****************************************************************************/
+/* PCI Capabilities Probe Function. */
+/* */
+/* Walks the PCI capabiites list for the device to find what features are */
+/* supported. */
+/* */
+/* Returns: */
+/* None. */
+/****************************************************************************/
+static void
+bce_print_adapter_info(struct bce_softc *sc)
+{
+ int i = 0;
+
+ DBENTER(BCE_VERBOSE_LOAD);
+
+ if (bootverbose) {
+ BCE_PRINTF("ASIC (0x%08X); ", sc->bce_chipid);
+ printf("Rev (%c%d); ", ((BCE_CHIP_ID(sc) & 0xf000) >>
+ 12) + 'A', ((BCE_CHIP_ID(sc) & 0x0ff0) >> 4));
+
+
+ /* Bus info. */
+ if (sc->bce_flags & BCE_PCIE_FLAG) {
+ printf("Bus (PCIe x%d, ", sc->link_width);
+ switch (sc->link_speed) {
+ case 1: printf("2.5Gbps); "); break;
+ case 2: printf("5Gbps); "); break;
+ default: printf("Unknown link speed); ");
+ }
+ } else {
+ printf("Bus (PCI%s, %s, %dMHz); ",
+ ((sc->bce_flags & BCE_PCIX_FLAG) ? "-X" : ""),
+ ((sc->bce_flags & BCE_PCI_32BIT_FLAG) ?
+ "32-bit" : "64-bit"), sc->bus_speed_mhz);
+ }
+
+ /* Firmware version and device features. */
+ printf("B/C (%s); Flags (", sc->bce_bc_ver);
+
+ #ifdef BCE_JUMBO_HDRSPLIT
+ printf("SPLT");
+ i++;
+ #endif
+
+ if (sc->bce_flags & BCE_USING_MSI_FLAG) {
+ if (i > 0) printf("|");
+ printf("MSI"); i++;
+ }
+
+ if (sc->bce_flags & BCE_USING_MSIX_FLAG) {
+ if (i > 0) printf("|");
+ printf("MSI-X"); i++;
+ }
+
+ if (sc->bce_phy_flags & BCE_PHY_2_5G_CAPABLE_FLAG) {
+ if (i > 0) printf("|");
+ printf("2.5G"); i++;
+ }
+
+ if (sc->bce_flags & BCE_MFW_ENABLE_FLAG) {
+ if (i > 0) printf("|");
+ printf("MFW); MFW (%s)\n", sc->bce_mfw_ver);
+ } else {
+ printf(")\n");
+ }
+ }
+
+ DBEXIT(BCE_VERBOSE_LOAD);
+}
+
+
+/****************************************************************************/
+/* PCI Capabilities Probe Function. */
+/* */
+/* Walks the PCI capabiites list for the device to find what features are */
+/* supported. */
+/* */
+/* Returns: */
+/* None. */
+/****************************************************************************/
+static void
+bce_probe_pci_caps(device_t dev, struct bce_softc *sc)
+{
+ u32 reg;
+
+ DBENTER(BCE_VERBOSE_LOAD);
+
+ /* Check if PCI-X capability is enabled. */
+ if (pci_find_extcap(dev, PCIY_PCIX, &reg) == 0) {
+ if (reg != 0)
+ sc->bce_cap_flags |= BCE_PCIX_CAPABLE_FLAG;
+ }
+
+ /* Check if PCIe capability is enabled. */
+ if (pci_find_extcap(dev, PCIY_EXPRESS, &reg) == 0) {
+ if (reg != 0) {
+ u16 link_status = pci_read_config(dev, reg + 0x12, 2);
+ DBPRINT(sc, BCE_INFO_LOAD, "PCIe link_status = "
+ "0x%08X\n", link_status);
+ sc->link_speed = link_status & 0xf;
+ sc->link_width = (link_status >> 4) & 0x3f;
+ sc->bce_cap_flags |= BCE_PCIE_CAPABLE_FLAG;
+ sc->bce_flags |= BCE_PCIE_FLAG;
+ }
+ }
+
+ /* Check if MSI capability is enabled. */
+ if (pci_find_extcap(dev, PCIY_MSI, &reg) == 0) {
+ if (reg != 0)
+ sc->bce_cap_flags |= BCE_MSI_CAPABLE_FLAG;
+ }
+
+ /* Check if MSI-X capability is enabled. */
+ if (pci_find_extcap(dev, PCIY_MSIX, &reg) == 0) {
+ if (reg != 0)
+ sc->bce_cap_flags |= BCE_MSIX_CAPABLE_FLAG;
+ }
+
+ DBEXIT(BCE_VERBOSE_LOAD);
+}
+
+
+/****************************************************************************/
+/* Device attach function. */
+/* */
+/* Allocates device resources, performs secondary chip identification, */
+/* resets and initializes the hardware, and initializes driver instance */
+/* variables. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_attach(device_t dev)
+{
+ struct bce_softc *sc;
+ struct ifnet *ifp;
+ u32 val;
+ int error, rid, rc = 0;
+
+ sc = device_get_softc(dev);
+ sc->bce_dev = dev;
+
+ DBENTER(BCE_VERBOSE_LOAD | BCE_VERBOSE_RESET);
+
+ sc->bce_unit = device_get_unit(dev);
+
+ /* Set initial device and PHY flags */
+ sc->bce_flags = 0;
+ sc->bce_phy_flags = 0;
+
+ pci_enable_busmaster(dev);
+
+ /* Allocate PCI memory resources. */
+ rid = PCIR_BAR(0);
+ sc->bce_res_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
+ &rid, RF_ACTIVE);
+
+ if (sc->bce_res_mem == NULL) {
+ BCE_PRINTF("%s(%d): PCI memory allocation failed\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Get various resource handles. */
+ sc->bce_btag = rman_get_bustag(sc->bce_res_mem);
+ sc->bce_bhandle = rman_get_bushandle(sc->bce_res_mem);
+ sc->bce_vhandle = (vm_offset_t) rman_get_virtual(sc->bce_res_mem);
+
+ bce_probe_pci_caps(dev, sc);
+
+ rid = 1;
+#if 0
+ /* Try allocating MSI-X interrupts. */
+ if ((sc->bce_cap_flags & BCE_MSIX_CAPABLE_FLAG) &&
+ (bce_msi_enable >= 2) &&
+ ((sc->bce_res_irq = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
+ &rid, RF_ACTIVE)) != NULL)) {
+
+ msi_needed = sc->bce_msi_count = 1;
+
+ if (((error = pci_alloc_msix(dev, &sc->bce_msi_count)) != 0) ||
+ (sc->bce_msi_count != msi_needed)) {
+ BCE_PRINTF("%s(%d): MSI-X allocation failed! Requested = %d,"
+ "Received = %d, error = %d\n", __FILE__, __LINE__,
+ msi_needed, sc->bce_msi_count, error);
+ sc->bce_msi_count = 0;
+ pci_release_msi(dev);
+ bus_release_resource(dev, SYS_RES_MEMORY, rid,
+ sc->bce_res_irq);
+ sc->bce_res_irq = NULL;
+ } else {
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): Using MSI-X interrupt.\n",
+ __FUNCTION__);
+ sc->bce_flags |= BCE_USING_MSIX_FLAG;
+ sc->bce_intr = bce_intr;
+ }
+ }
+#endif
+
+ /* Try allocating a MSI interrupt. */
+ if ((sc->bce_cap_flags & BCE_MSI_CAPABLE_FLAG) &&
+ (bce_msi_enable >= 1) && (sc->bce_msi_count == 0)) {
+ sc->bce_msi_count = 1;
+ if ((error = pci_alloc_msi(dev, &sc->bce_msi_count)) != 0) {
+ BCE_PRINTF("%s(%d): MSI allocation failed! "
+ "error = %d\n", __FILE__, __LINE__, error);
+ sc->bce_msi_count = 0;
+ pci_release_msi(dev);
+ } else {
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): Using MSI "
+ "interrupt.\n", __FUNCTION__);
+ sc->bce_flags |= BCE_USING_MSI_FLAG;
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716))
+ sc->bce_flags |= BCE_ONE_SHOT_MSI_FLAG;
+ sc->bce_irq_rid = 1;
+ sc->bce_intr = bce_intr;
+ }
+ }
+
+ /* Try allocating a legacy interrupt. */
+ if (sc->bce_msi_count == 0) {
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): Using INTx interrupt.\n",
+ __FUNCTION__);
+ rid = 0;
+ sc->bce_intr = bce_intr;
+ }
+
+ sc->bce_res_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ,
+ &rid, RF_SHAREABLE | RF_ACTIVE);
+
+ sc->bce_irq_rid = rid;
+
+ /* Report any IRQ allocation errors. */
+ if (sc->bce_res_irq == NULL) {
+ BCE_PRINTF("%s(%d): PCI map interrupt failed!\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Initialize mutex for the current device instance. */
+ BCE_LOCK_INIT(sc, device_get_nameunit(dev));
+
+ /*
+ * Configure byte swap and enable indirect register access.
+ * Rely on CPU to do target byte swapping on big endian systems.
+ * Access to registers outside of PCI configurtion space are not
+ * valid until this is done.
+ */
+ pci_write_config(dev, BCE_PCICFG_MISC_CONFIG,
+ BCE_PCICFG_MISC_CONFIG_REG_WINDOW_ENA |
+ BCE_PCICFG_MISC_CONFIG_TARGET_MB_WORD_SWAP, 4);
+
+ /* Save ASIC revsion info. */
+ sc->bce_chipid = REG_RD(sc, BCE_MISC_ID);
+
+ /* Weed out any non-production controller revisions. */
+ switch(BCE_CHIP_ID(sc)) {
+ case BCE_CHIP_ID_5706_A0:
+ case BCE_CHIP_ID_5706_A1:
+ case BCE_CHIP_ID_5708_A0:
+ case BCE_CHIP_ID_5708_B0:
+ case BCE_CHIP_ID_5709_A0:
+ case BCE_CHIP_ID_5709_B0:
+ case BCE_CHIP_ID_5709_B1:
+ case BCE_CHIP_ID_5709_B2:
+ BCE_PRINTF("%s(%d): Unsupported controller "
+ "revision (%c%d)!\n", __FILE__, __LINE__,
+ (((pci_read_config(dev, PCIR_REVID, 4) &
+ 0xf0) >> 4) + 'A'), (pci_read_config(dev,
+ PCIR_REVID, 4) & 0xf));
+ rc = ENODEV;
+ goto bce_attach_fail;
+ }
+
+ /*
+ * The embedded PCIe to PCI-X bridge (EPB)
+ * in the 5708 cannot address memory above
+ * 40 bits (E7_5708CB1_23043 & E6_5708SB1_23043).
+ */
+ if (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5708)
+ sc->max_bus_addr = BCE_BUS_SPACE_MAXADDR;
+ else
+ sc->max_bus_addr = BUS_SPACE_MAXADDR;
+
+ /*
+ * Find the base address for shared memory access.
+ * Newer versions of bootcode use a signature and offset
+ * while older versions use a fixed address.
+ */
+ val = REG_RD_IND(sc, BCE_SHM_HDR_SIGNATURE);
+ if ((val & BCE_SHM_HDR_SIGNATURE_SIG_MASK) == BCE_SHM_HDR_SIGNATURE_SIG)
+ /* Multi-port devices use different offsets in shared memory. */
+ sc->bce_shmem_base = REG_RD_IND(sc, BCE_SHM_HDR_ADDR_0 +
+ (pci_get_function(sc->bce_dev) << 2));
+ else
+ sc->bce_shmem_base = HOST_VIEW_SHMEM_BASE;
+
+ DBPRINT(sc, BCE_VERBOSE_FIRMWARE, "%s(): bce_shmem_base = 0x%08X\n",
+ __FUNCTION__, sc->bce_shmem_base);
+
+ /* Fetch the bootcode revision. */
+ val = bce_shmem_rd(sc, BCE_DEV_INFO_BC_REV);
+ for (int i = 0, j = 0; i < 3; i++) {
+ u8 num;
+
+ num = (u8) (val >> (24 - (i * 8)));
+ for (int k = 100, skip0 = 1; k >= 1; num %= k, k /= 10) {
+ if (num >= k || !skip0 || k == 1) {
+ sc->bce_bc_ver[j++] = (num / k) + '0';
+ skip0 = 0;
+ }
+ }
+
+ if (i != 2)
+ sc->bce_bc_ver[j++] = '.';
+ }
+
+ /* Check if any management firwmare is enabled. */
+ val = bce_shmem_rd(sc, BCE_PORT_FEATURE);
+ if (val & BCE_PORT_FEATURE_ASF_ENABLED) {
+ sc->bce_flags |= BCE_MFW_ENABLE_FLAG;
+
+ /* Allow time for firmware to enter the running state. */
+ for (int i = 0; i < 30; i++) {
+ val = bce_shmem_rd(sc, BCE_BC_STATE_CONDITION);
+ if (val & BCE_CONDITION_MFW_RUN_MASK)
+ break;
+ DELAY(10000);
+ }
+
+ /* Check if management firmware is running. */
+ val = bce_shmem_rd(sc, BCE_BC_STATE_CONDITION);
+ val &= BCE_CONDITION_MFW_RUN_MASK;
+ if ((val != BCE_CONDITION_MFW_RUN_UNKNOWN) &&
+ (val != BCE_CONDITION_MFW_RUN_NONE)) {
+ u32 addr = bce_shmem_rd(sc, BCE_MFW_VER_PTR);
+ int i = 0;
+
+ /* Read the management firmware version string. */
+ for (int j = 0; j < 3; j++) {
+ val = bce_reg_rd_ind(sc, addr + j * 4);
+ val = bswap32(val);
+ memcpy(&sc->bce_mfw_ver[i], &val, 4);
+ i += 4;
+ }
+ } else {
+ /* May cause firmware synchronization timeouts. */
+ BCE_PRINTF("%s(%d): Management firmware enabled "
+ "but not running!\n", __FILE__, __LINE__);
+ strcpy(sc->bce_mfw_ver, "NOT RUNNING!");
+
+ /* ToDo: Any action the driver should take? */
+ }
+ }
+
+ /* Get PCI bus information (speed and type). */
+ val = REG_RD(sc, BCE_PCICFG_MISC_STATUS);
+ if (val & BCE_PCICFG_MISC_STATUS_PCIX_DET) {
+ u32 clkreg;
+
+ sc->bce_flags |= BCE_PCIX_FLAG;
+
+ clkreg = REG_RD(sc, BCE_PCICFG_PCI_CLOCK_CONTROL_BITS);
+
+ clkreg &= BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET;
+ switch (clkreg) {
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_133MHZ:
+ sc->bus_speed_mhz = 133;
+ break;
+
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_95MHZ:
+ sc->bus_speed_mhz = 100;
+ break;
+
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_66MHZ:
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_80MHZ:
+ sc->bus_speed_mhz = 66;
+ break;
+
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_48MHZ:
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_55MHZ:
+ sc->bus_speed_mhz = 50;
+ break;
+
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_LOW:
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_32MHZ:
+ case BCE_PCICFG_PCI_CLOCK_CONTROL_BITS_PCI_CLK_SPD_DET_38MHZ:
+ sc->bus_speed_mhz = 33;
+ break;
+ }
+ } else {
+ if (val & BCE_PCICFG_MISC_STATUS_M66EN)
+ sc->bus_speed_mhz = 66;
+ else
+ sc->bus_speed_mhz = 33;
+ }
+
+ if (val & BCE_PCICFG_MISC_STATUS_32BIT_DET)
+ sc->bce_flags |= BCE_PCI_32BIT_FLAG;
+
+ /* Reset controller and announce to bootcode that driver is present. */
+ if (bce_reset(sc, BCE_DRV_MSG_CODE_RESET)) {
+ BCE_PRINTF("%s(%d): Controller reset failed!\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Initialize the controller. */
+ if (bce_chipinit(sc)) {
+ BCE_PRINTF("%s(%d): Controller initialization failed!\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Perform NVRAM test. */
+ if (bce_nvram_test(sc)) {
+ BCE_PRINTF("%s(%d): NVRAM test failed!\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Fetch the permanent Ethernet MAC address. */
+ bce_get_mac_addr(sc);
+
+ /*
+ * Trip points control how many BDs
+ * should be ready before generating an
+ * interrupt while ticks control how long
+ * a BD can sit in the chain before
+ * generating an interrupt. Set the default
+ * values for the RX and TX chains.
+ */
+
+#ifdef BCE_DEBUG
+ /* Force more frequent interrupts. */
+ sc->bce_tx_quick_cons_trip_int = 1;
+ sc->bce_tx_quick_cons_trip = 1;
+ sc->bce_tx_ticks_int = 0;
+ sc->bce_tx_ticks = 0;
+
+ sc->bce_rx_quick_cons_trip_int = 1;
+ sc->bce_rx_quick_cons_trip = 1;
+ sc->bce_rx_ticks_int = 0;
+ sc->bce_rx_ticks = 0;
+#else
+ /* Improve throughput at the expense of increased latency. */
+ sc->bce_tx_quick_cons_trip_int = 20;
+ sc->bce_tx_quick_cons_trip = 20;
+ sc->bce_tx_ticks_int = 80;
+ sc->bce_tx_ticks = 80;
+
+ sc->bce_rx_quick_cons_trip_int = 6;
+ sc->bce_rx_quick_cons_trip = 6;
+ sc->bce_rx_ticks_int = 18;
+ sc->bce_rx_ticks = 18;
+#endif
+
+ /* Not used for L2. */
+ sc->bce_comp_prod_trip_int = 0;
+ sc->bce_comp_prod_trip = 0;
+ sc->bce_com_ticks_int = 0;
+ sc->bce_com_ticks = 0;
+ sc->bce_cmd_ticks_int = 0;
+ sc->bce_cmd_ticks = 0;
+
+ /* Update statistics once every second. */
+ sc->bce_stats_ticks = 1000000 & 0xffff00;
+
+ /* Find the media type for the adapter. */
+ bce_get_media(sc);
+
+ /* Store data needed by PHY driver for backplane applications */
+ sc->bce_shared_hw_cfg = bce_shmem_rd(sc, BCE_SHARED_HW_CFG_CONFIG);
+ sc->bce_port_hw_cfg = bce_shmem_rd(sc, BCE_PORT_HW_CFG_CONFIG);
+
+ /* Allocate DMA memory resources. */
+ if (bce_dma_alloc(dev)) {
+ BCE_PRINTF("%s(%d): DMA resource allocation failed!\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Allocate an ifnet structure. */
+ ifp = sc->bce_ifp = if_alloc(IFT_ETHER);
+ if (ifp == NULL) {
+ BCE_PRINTF("%s(%d): Interface allocation failed!\n",
+ __FILE__, __LINE__);
+ rc = ENXIO;
+ goto bce_attach_fail;
+ }
+
+ /* Initialize the ifnet interface. */
+ ifp->if_softc = sc;
+ if_initname(ifp, device_get_name(dev), device_get_unit(dev));
+ ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
+ ifp->if_ioctl = bce_ioctl;
+ ifp->if_start = bce_start;
+ ifp->if_init = bce_init;
+ ifp->if_mtu = ETHERMTU;
+
+ if (bce_tso_enable) {
+ ifp->if_hwassist = BCE_IF_HWASSIST | CSUM_TSO;
+ ifp->if_capabilities = BCE_IF_CAPABILITIES | IFCAP_TSO4 |
+ IFCAP_VLAN_HWTSO;
+ } else {
+ ifp->if_hwassist = BCE_IF_HWASSIST;
+ ifp->if_capabilities = BCE_IF_CAPABILITIES;
+ }
+
+ ifp->if_capenable = ifp->if_capabilities;
+
+ /*
+ * Assume standard mbuf sizes for buffer allocation.
+ * This may change later if the MTU size is set to
+ * something other than 1500.
+ */
+#ifdef BCE_JUMBO_HDRSPLIT
+ sc->rx_bd_mbuf_alloc_size = MHLEN;
+ /* Make sure offset is 16 byte aligned for hardware. */
+ sc->rx_bd_mbuf_align_pad =
+ roundup2((MSIZE - MHLEN), 16) - (MSIZE - MHLEN);
+ sc->rx_bd_mbuf_data_len = sc->rx_bd_mbuf_alloc_size -
+ sc->rx_bd_mbuf_align_pad;
+ sc->pg_bd_mbuf_alloc_size = MCLBYTES;
+#else
+ sc->rx_bd_mbuf_alloc_size = MCLBYTES;
+ sc->rx_bd_mbuf_align_pad =
+ roundup2(MCLBYTES, 16) - MCLBYTES;
+ sc->rx_bd_mbuf_data_len = sc->rx_bd_mbuf_alloc_size -
+ sc->rx_bd_mbuf_align_pad;
+#endif
+
+ ifp->if_snd.ifq_drv_maxlen = USABLE_TX_BD;
+ IFQ_SET_MAXLEN(&ifp->if_snd, ifp->if_snd.ifq_drv_maxlen);
+ IFQ_SET_READY(&ifp->if_snd);
+
+ if (sc->bce_phy_flags & BCE_PHY_2_5G_CAPABLE_FLAG)
+ ifp->if_baudrate = IF_Mbps(2500ULL);
+ else
+ ifp->if_baudrate = IF_Mbps(1000);
+
+ /* Handle any special PHY initialization for SerDes PHYs. */
+ bce_init_media(sc);
+
+ /* MII child bus by attaching the PHY. */
+ rc = mii_attach(dev, &sc->bce_miibus, ifp, bce_ifmedia_upd,
+ bce_ifmedia_sts, BMSR_DEFCAPMASK, sc->bce_phy_addr,
+ MII_OFFSET_ANY, MIIF_DOPAUSE | MIIF_FORCEPAUSE);
+ if (rc != 0) {
+ BCE_PRINTF("%s(%d): attaching PHYs failed\n", __FILE__,
+ __LINE__);
+ goto bce_attach_fail;
+ }
+
+ /* Attach to the Ethernet interface list. */
+ ether_ifattach(ifp, sc->eaddr);
+
+#if __FreeBSD_version < 500000
+ callout_init(&sc->bce_tick_callout);
+ callout_init(&sc->bce_pulse_callout);
+#else
+ callout_init_mtx(&sc->bce_tick_callout, &sc->bce_mtx, 0);
+ callout_init_mtx(&sc->bce_pulse_callout, &sc->bce_mtx, 0);
+#endif
+
+ /* Hookup IRQ last. */
+ rc = bus_setup_intr(dev, sc->bce_res_irq, INTR_TYPE_NET | INTR_MPSAFE,
+ NULL, bce_intr, sc, &sc->bce_intrhand);
+
+ if (rc) {
+ BCE_PRINTF("%s(%d): Failed to setup IRQ!\n",
+ __FILE__, __LINE__);
+ bce_detach(dev);
+ goto bce_attach_exit;
+ }
+
+ /*
+ * At this point we've acquired all the resources
+ * we need to run so there's no turning back, we're
+ * cleared for launch.
+ */
+
+ /* Print some important debugging info. */
+ DBRUNMSG(BCE_INFO, bce_dump_driver_state(sc));
+
+ /* Add the supported sysctls to the kernel. */
+ bce_add_sysctls(sc);
+
+ BCE_LOCK(sc);
+
+ /*
+ * The chip reset earlier notified the bootcode that
+ * a driver is present. We now need to start our pulse
+ * routine so that the bootcode is reminded that we're
+ * still running.
+ */
+ bce_pulse(sc);
+
+ bce_mgmt_init_locked(sc);
+ BCE_UNLOCK(sc);
+
+ /* Finally, print some useful adapter info */
+ bce_print_adapter_info(sc);
+ DBPRINT(sc, BCE_FATAL, "%s(): sc = %p\n",
+ __FUNCTION__, sc);
+
+ goto bce_attach_exit;
+
+bce_attach_fail:
+ bce_release_resources(sc);
+
+bce_attach_exit:
+
+ DBEXIT(BCE_VERBOSE_LOAD | BCE_VERBOSE_RESET);
+
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Device detach function. */
+/* */
+/* Stops the controller, resets the controller, and releases resources. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_detach(device_t dev)
+{
+ struct bce_softc *sc = device_get_softc(dev);
+ struct ifnet *ifp;
+ u32 msg;
+
+ DBENTER(BCE_VERBOSE_UNLOAD | BCE_VERBOSE_RESET);
+
+ ifp = sc->bce_ifp;
+
+ /* Stop and reset the controller. */
+ BCE_LOCK(sc);
+
+ /* Stop the pulse so the bootcode can go to driver absent state. */
+ callout_stop(&sc->bce_pulse_callout);
+
+ bce_stop(sc);
+ if (sc->bce_flags & BCE_NO_WOL_FLAG)
+ msg = BCE_DRV_MSG_CODE_UNLOAD_LNK_DN;
+ else
+ msg = BCE_DRV_MSG_CODE_UNLOAD;
+ bce_reset(sc, msg);
+
+ BCE_UNLOCK(sc);
+
+ ether_ifdetach(ifp);
+
+ /* If we have a child device on the MII bus remove it too. */
+ bus_generic_detach(dev);
+ device_delete_child(dev, sc->bce_miibus);
+
+ /* Release all remaining resources. */
+ bce_release_resources(sc);
+
+ DBEXIT(BCE_VERBOSE_UNLOAD | BCE_VERBOSE_RESET);
+
+ return(0);
+}
+
+
+/****************************************************************************/
+/* Device shutdown function. */
+/* */
+/* Stops and resets the controller. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_shutdown(device_t dev)
+{
+ struct bce_softc *sc = device_get_softc(dev);
+ u32 msg;
+
+ DBENTER(BCE_VERBOSE);
+
+ BCE_LOCK(sc);
+ bce_stop(sc);
+ if (sc->bce_flags & BCE_NO_WOL_FLAG)
+ msg = BCE_DRV_MSG_CODE_UNLOAD_LNK_DN;
+ else
+ msg = BCE_DRV_MSG_CODE_UNLOAD;
+ bce_reset(sc, msg);
+ BCE_UNLOCK(sc);
+
+ DBEXIT(BCE_VERBOSE);
+
+ return (0);
+}
+
+
+#ifdef BCE_DEBUG
+/****************************************************************************/
+/* Register read. */
+/* */
+/* Returns: */
+/* The value of the register. */
+/****************************************************************************/
+static u32
+bce_reg_rd(struct bce_softc *sc, u32 offset)
+{
+ u32 val = bus_space_read_4(sc->bce_btag, sc->bce_bhandle, offset);
+ DBPRINT(sc, BCE_INSANE_REG, "%s(); offset = 0x%08X, val = 0x%08X\n",
+ __FUNCTION__, offset, val);
+ return val;
+}
+
+
+/****************************************************************************/
+/* Register write (16 bit). */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_reg_wr16(struct bce_softc *sc, u32 offset, u16 val)
+{
+ DBPRINT(sc, BCE_INSANE_REG, "%s(); offset = 0x%08X, val = 0x%04X\n",
+ __FUNCTION__, offset, val);
+ bus_space_write_2(sc->bce_btag, sc->bce_bhandle, offset, val);
+}
+
+
+/****************************************************************************/
+/* Register write. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_reg_wr(struct bce_softc *sc, u32 offset, u32 val)
+{
+ DBPRINT(sc, BCE_INSANE_REG, "%s(); offset = 0x%08X, val = 0x%08X\n",
+ __FUNCTION__, offset, val);
+ bus_space_write_4(sc->bce_btag, sc->bce_bhandle, offset, val);
+}
+#endif
+
+/****************************************************************************/
+/* Indirect register read. */
+/* */
+/* Reads NetXtreme II registers using an index/data register pair in PCI */
+/* configuration space. Using this mechanism avoids issues with posted */
+/* reads but is much slower than memory-mapped I/O. */
+/* */
+/* Returns: */
+/* The value of the register. */
+/****************************************************************************/
+static u32
+bce_reg_rd_ind(struct bce_softc *sc, u32 offset)
+{
+ device_t dev;
+ dev = sc->bce_dev;
+
+ pci_write_config(dev, BCE_PCICFG_REG_WINDOW_ADDRESS, offset, 4);
+#ifdef BCE_DEBUG
+ {
+ u32 val;
+ val = pci_read_config(dev, BCE_PCICFG_REG_WINDOW, 4);
+ DBPRINT(sc, BCE_INSANE_REG, "%s(); offset = 0x%08X, val = 0x%08X\n",
+ __FUNCTION__, offset, val);
+ return val;
+ }
+#else
+ return pci_read_config(dev, BCE_PCICFG_REG_WINDOW, 4);
+#endif
+}
+
+
+/****************************************************************************/
+/* Indirect register write. */
+/* */
+/* Writes NetXtreme II registers using an index/data register pair in PCI */
+/* configuration space. Using this mechanism avoids issues with posted */
+/* writes but is muchh slower than memory-mapped I/O. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_reg_wr_ind(struct bce_softc *sc, u32 offset, u32 val)
+{
+ device_t dev;
+ dev = sc->bce_dev;
+
+ DBPRINT(sc, BCE_INSANE_REG, "%s(); offset = 0x%08X, val = 0x%08X\n",
+ __FUNCTION__, offset, val);
+
+ pci_write_config(dev, BCE_PCICFG_REG_WINDOW_ADDRESS, offset, 4);
+ pci_write_config(dev, BCE_PCICFG_REG_WINDOW, val, 4);
+}
+
+
+/****************************************************************************/
+/* Shared memory write. */
+/* */
+/* Writes NetXtreme II shared memory region. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_shmem_wr(struct bce_softc *sc, u32 offset, u32 val)
+{
+ DBPRINT(sc, BCE_VERBOSE_FIRMWARE, "%s(): Writing 0x%08X to "
+ "0x%08X\n", __FUNCTION__, val, offset);
+
+ bce_reg_wr_ind(sc, sc->bce_shmem_base + offset, val);
+}
+
+
+/****************************************************************************/
+/* Shared memory read. */
+/* */
+/* Reads NetXtreme II shared memory region. */
+/* */
+/* Returns: */
+/* The 32 bit value read. */
+/****************************************************************************/
+static u32
+bce_shmem_rd(struct bce_softc *sc, u32 offset)
+{
+ u32 val = bce_reg_rd_ind(sc, sc->bce_shmem_base + offset);
+
+ DBPRINT(sc, BCE_VERBOSE_FIRMWARE, "%s(): Reading 0x%08X from "
+ "0x%08X\n", __FUNCTION__, val, offset);
+
+ return val;
+}
+
+
+#ifdef BCE_DEBUG
+/****************************************************************************/
+/* Context memory read. */
+/* */
+/* The NetXtreme II controller uses context memory to track connection */
+/* information for L2 and higher network protocols. */
+/* */
+/* Returns: */
+/* The requested 32 bit value of context memory. */
+/****************************************************************************/
+static u32
+bce_ctx_rd(struct bce_softc *sc, u32 cid_addr, u32 ctx_offset)
+{
+ u32 idx, offset, retry_cnt = 5, val;
+
+ DBRUNIF((cid_addr > MAX_CID_ADDR || ctx_offset & 0x3 ||
+ cid_addr & CTX_MASK), BCE_PRINTF("%s(): Invalid CID "
+ "address: 0x%08X.\n", __FUNCTION__, cid_addr));
+
+ offset = ctx_offset + cid_addr;
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+
+ REG_WR(sc, BCE_CTX_CTX_CTRL, (offset | BCE_CTX_CTX_CTRL_READ_REQ));
+
+ for (idx = 0; idx < retry_cnt; idx++) {
+ val = REG_RD(sc, BCE_CTX_CTX_CTRL);
+ if ((val & BCE_CTX_CTX_CTRL_READ_REQ) == 0)
+ break;
+ DELAY(5);
+ }
+
+ if (val & BCE_CTX_CTX_CTRL_READ_REQ)
+ BCE_PRINTF("%s(%d); Unable to read CTX memory: "
+ "cid_addr = 0x%08X, offset = 0x%08X!\n",
+ __FILE__, __LINE__, cid_addr, ctx_offset);
+
+ val = REG_RD(sc, BCE_CTX_CTX_DATA);
+ } else {
+ REG_WR(sc, BCE_CTX_DATA_ADR, offset);
+ val = REG_RD(sc, BCE_CTX_DATA);
+ }
+
+ DBPRINT(sc, BCE_EXTREME_CTX, "%s(); cid_addr = 0x%08X, offset = 0x%08X, "
+ "val = 0x%08X\n", __FUNCTION__, cid_addr, ctx_offset, val);
+
+ return(val);
+}
+#endif
+
+
+/****************************************************************************/
+/* Context memory write. */
+/* */
+/* The NetXtreme II controller uses context memory to track connection */
+/* information for L2 and higher network protocols. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_ctx_wr(struct bce_softc *sc, u32 cid_addr, u32 ctx_offset, u32 ctx_val)
+{
+ u32 idx, offset = ctx_offset + cid_addr;
+ u32 val, retry_cnt = 5;
+
+ DBPRINT(sc, BCE_EXTREME_CTX, "%s(); cid_addr = 0x%08X, offset = 0x%08X, "
+ "val = 0x%08X\n", __FUNCTION__, cid_addr, ctx_offset, ctx_val);
+
+ DBRUNIF((cid_addr > MAX_CID_ADDR || ctx_offset & 0x3 || cid_addr & CTX_MASK),
+ BCE_PRINTF("%s(): Invalid CID address: 0x%08X.\n",
+ __FUNCTION__, cid_addr));
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+
+ REG_WR(sc, BCE_CTX_CTX_DATA, ctx_val);
+ REG_WR(sc, BCE_CTX_CTX_CTRL, (offset | BCE_CTX_CTX_CTRL_WRITE_REQ));
+
+ for (idx = 0; idx < retry_cnt; idx++) {
+ val = REG_RD(sc, BCE_CTX_CTX_CTRL);
+ if ((val & BCE_CTX_CTX_CTRL_WRITE_REQ) == 0)
+ break;
+ DELAY(5);
+ }
+
+ if (val & BCE_CTX_CTX_CTRL_WRITE_REQ)
+ BCE_PRINTF("%s(%d); Unable to write CTX memory: "
+ "cid_addr = 0x%08X, offset = 0x%08X!\n",
+ __FILE__, __LINE__, cid_addr, ctx_offset);
+
+ } else {
+ REG_WR(sc, BCE_CTX_DATA_ADR, offset);
+ REG_WR(sc, BCE_CTX_DATA, ctx_val);
+ }
+}
+
+
+/****************************************************************************/
+/* PHY register read. */
+/* */
+/* Implements register reads on the MII bus. */
+/* */
+/* Returns: */
+/* The value of the register. */
+/****************************************************************************/
+static int
+bce_miibus_read_reg(device_t dev, int phy, int reg)
+{
+ struct bce_softc *sc;
+ u32 val;
+ int i;
+
+ sc = device_get_softc(dev);
+
+ /* Make sure we are accessing the correct PHY address. */
+ if (phy != sc->bce_phy_addr) {
+ DBPRINT(sc, BCE_INSANE_PHY, "Invalid PHY address %d "
+ "for PHY read!\n", phy);
+ return(0);
+ }
+
+ /*
+ * The 5709S PHY is an IEEE Clause 45 PHY
+ * with special mappings to work with IEEE
+ * Clause 22 register accesses.
+ */
+ if ((sc->bce_phy_flags & BCE_PHY_IEEE_CLAUSE_45_FLAG) != 0) {
+ if (reg >= MII_BMCR && reg <= MII_ANLPRNP)
+ reg += 0x10;
+ }
+
+ if (sc->bce_phy_flags & BCE_PHY_INT_MODE_AUTO_POLLING_FLAG) {
+ val = REG_RD(sc, BCE_EMAC_MDIO_MODE);
+ val &= ~BCE_EMAC_MDIO_MODE_AUTO_POLL;
+
+ REG_WR(sc, BCE_EMAC_MDIO_MODE, val);
+ REG_RD(sc, BCE_EMAC_MDIO_MODE);
+
+ DELAY(40);
+ }
+
+
+ val = BCE_MIPHY(phy) | BCE_MIREG(reg) |
+ BCE_EMAC_MDIO_COMM_COMMAND_READ | BCE_EMAC_MDIO_COMM_DISEXT |
+ BCE_EMAC_MDIO_COMM_START_BUSY;
+ REG_WR(sc, BCE_EMAC_MDIO_COMM, val);
+
+ for (i = 0; i < BCE_PHY_TIMEOUT; i++) {
+ DELAY(10);
+
+ val = REG_RD(sc, BCE_EMAC_MDIO_COMM);
+ if (!(val & BCE_EMAC_MDIO_COMM_START_BUSY)) {
+ DELAY(5);
+
+ val = REG_RD(sc, BCE_EMAC_MDIO_COMM);
+ val &= BCE_EMAC_MDIO_COMM_DATA;
+
+ break;
+ }
+ }
+
+ if (val & BCE_EMAC_MDIO_COMM_START_BUSY) {
+ BCE_PRINTF("%s(%d): Error: PHY read timeout! phy = %d, "
+ "reg = 0x%04X\n", __FILE__, __LINE__, phy, reg);
+ val = 0x0;
+ } else {
+ val = REG_RD(sc, BCE_EMAC_MDIO_COMM);
+ }
+
+
+ if (sc->bce_phy_flags & BCE_PHY_INT_MODE_AUTO_POLLING_FLAG) {
+ val = REG_RD(sc, BCE_EMAC_MDIO_MODE);
+ val |= BCE_EMAC_MDIO_MODE_AUTO_POLL;
+
+ REG_WR(sc, BCE_EMAC_MDIO_MODE, val);
+ REG_RD(sc, BCE_EMAC_MDIO_MODE);
+
+ DELAY(40);
+ }
+
+ DB_PRINT_PHY_REG(reg, val);
+ return (val & 0xffff);
+
+}
+
+
+/****************************************************************************/
+/* PHY register write. */
+/* */
+/* Implements register writes on the MII bus. */
+/* */
+/* Returns: */
+/* The value of the register. */
+/****************************************************************************/
+static int
+bce_miibus_write_reg(device_t dev, int phy, int reg, int val)
+{
+ struct bce_softc *sc;
+ u32 val1;
+ int i;
+
+ sc = device_get_softc(dev);
+
+ /* Make sure we are accessing the correct PHY address. */
+ if (phy != sc->bce_phy_addr) {
+ DBPRINT(sc, BCE_INSANE_PHY, "Invalid PHY address %d "
+ "for PHY write!\n", phy);
+ return(0);
+ }
+
+ DB_PRINT_PHY_REG(reg, val);
+
+ /*
+ * The 5709S PHY is an IEEE Clause 45 PHY
+ * with special mappings to work with IEEE
+ * Clause 22 register accesses.
+ */
+ if ((sc->bce_phy_flags & BCE_PHY_IEEE_CLAUSE_45_FLAG) != 0) {
+ if (reg >= MII_BMCR && reg <= MII_ANLPRNP)
+ reg += 0x10;
+ }
+
+ if (sc->bce_phy_flags & BCE_PHY_INT_MODE_AUTO_POLLING_FLAG) {
+ val1 = REG_RD(sc, BCE_EMAC_MDIO_MODE);
+ val1 &= ~BCE_EMAC_MDIO_MODE_AUTO_POLL;
+
+ REG_WR(sc, BCE_EMAC_MDIO_MODE, val1);
+ REG_RD(sc, BCE_EMAC_MDIO_MODE);
+
+ DELAY(40);
+ }
+
+ val1 = BCE_MIPHY(phy) | BCE_MIREG(reg) | val |
+ BCE_EMAC_MDIO_COMM_COMMAND_WRITE |
+ BCE_EMAC_MDIO_COMM_START_BUSY | BCE_EMAC_MDIO_COMM_DISEXT;
+ REG_WR(sc, BCE_EMAC_MDIO_COMM, val1);
+
+ for (i = 0; i < BCE_PHY_TIMEOUT; i++) {
+ DELAY(10);
+
+ val1 = REG_RD(sc, BCE_EMAC_MDIO_COMM);
+ if (!(val1 & BCE_EMAC_MDIO_COMM_START_BUSY)) {
+ DELAY(5);
+ break;
+ }
+ }
+
+ if (val1 & BCE_EMAC_MDIO_COMM_START_BUSY)
+ BCE_PRINTF("%s(%d): PHY write timeout!\n",
+ __FILE__, __LINE__);
+
+ if (sc->bce_phy_flags & BCE_PHY_INT_MODE_AUTO_POLLING_FLAG) {
+ val1 = REG_RD(sc, BCE_EMAC_MDIO_MODE);
+ val1 |= BCE_EMAC_MDIO_MODE_AUTO_POLL;
+
+ REG_WR(sc, BCE_EMAC_MDIO_MODE, val1);
+ REG_RD(sc, BCE_EMAC_MDIO_MODE);
+
+ DELAY(40);
+ }
+
+ return 0;
+}
+
+
+/****************************************************************************/
+/* MII bus status change. */
+/* */
+/* Called by the MII bus driver when the PHY establishes link to set the */
+/* MAC interface registers. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_miibus_statchg(device_t dev)
+{
+ struct bce_softc *sc;
+ struct mii_data *mii;
+ int val;
+
+ sc = device_get_softc(dev);
+
+ DBENTER(BCE_VERBOSE_PHY);
+
+ mii = device_get_softc(sc->bce_miibus);
+
+ val = REG_RD(sc, BCE_EMAC_MODE);
+ val &= ~(BCE_EMAC_MODE_PORT | BCE_EMAC_MODE_HALF_DUPLEX |
+ BCE_EMAC_MODE_MAC_LOOP | BCE_EMAC_MODE_FORCE_LINK |
+ BCE_EMAC_MODE_25G);
+
+ /* Set MII or GMII interface based on the PHY speed. */
+ switch (IFM_SUBTYPE(mii->mii_media_active)) {
+ case IFM_10_T:
+ if (BCE_CHIP_NUM(sc) != BCE_CHIP_NUM_5706) {
+ DBPRINT(sc, BCE_INFO_PHY,
+ "Enabling 10Mb interface.\n");
+ val |= BCE_EMAC_MODE_PORT_MII_10;
+ break;
+ }
+ /* fall-through */
+ case IFM_100_TX:
+ DBPRINT(sc, BCE_INFO_PHY, "Enabling MII interface.\n");
+ val |= BCE_EMAC_MODE_PORT_MII;
+ break;
+ case IFM_2500_SX:
+ DBPRINT(sc, BCE_INFO_PHY, "Enabling 2.5G MAC mode.\n");
+ val |= BCE_EMAC_MODE_25G;
+ /* fall-through */
+ case IFM_1000_T:
+ case IFM_1000_SX:
+ DBPRINT(sc, BCE_INFO_PHY, "Enabling GMII interface.\n");
+ val |= BCE_EMAC_MODE_PORT_GMII;
+ break;
+ default:
+ DBPRINT(sc, BCE_INFO_PHY, "Unknown link speed, enabling "
+ "default GMII interface.\n");
+ val |= BCE_EMAC_MODE_PORT_GMII;
+ }
+
+ /* Set half or full duplex based on PHY settings. */
+ if ((mii->mii_media_active & IFM_GMASK) == IFM_HDX) {
+ DBPRINT(sc, BCE_INFO_PHY,
+ "Setting Half-Duplex interface.\n");
+ val |= BCE_EMAC_MODE_HALF_DUPLEX;
+ } else
+ DBPRINT(sc, BCE_INFO_PHY,
+ "Setting Full-Duplex interface.\n");
+
+ REG_WR(sc, BCE_EMAC_MODE, val);
+
+ if ((mii->mii_media_active & IFM_ETH_RXPAUSE) != 0) {
+ DBPRINT(sc, BCE_INFO_PHY,
+ "%s(): Enabling RX flow control.\n", __FUNCTION__);
+ BCE_SETBIT(sc, BCE_EMAC_RX_MODE, BCE_EMAC_RX_MODE_FLOW_EN);
+ } else {
+ DBPRINT(sc, BCE_INFO_PHY,
+ "%s(): Disabling RX flow control.\n", __FUNCTION__);
+ BCE_CLRBIT(sc, BCE_EMAC_RX_MODE, BCE_EMAC_RX_MODE_FLOW_EN);
+ }
+
+ if ((mii->mii_media_active & IFM_ETH_TXPAUSE) != 0) {
+ DBPRINT(sc, BCE_INFO_PHY,
+ "%s(): Enabling TX flow control.\n", __FUNCTION__);
+ BCE_SETBIT(sc, BCE_EMAC_TX_MODE, BCE_EMAC_TX_MODE_FLOW_EN);
+ sc->bce_flags |= BCE_USING_TX_FLOW_CONTROL;
+ } else {
+ DBPRINT(sc, BCE_INFO_PHY,
+ "%s(): Disabling TX flow control.\n", __FUNCTION__);
+ BCE_CLRBIT(sc, BCE_EMAC_TX_MODE, BCE_EMAC_TX_MODE_FLOW_EN);
+ sc->bce_flags &= ~BCE_USING_TX_FLOW_CONTROL;
+ }
+
+ /* ToDo: Update watermarks in bce_init_rx_context(). */
+
+ DBEXIT(BCE_VERBOSE_PHY);
+}
+
+
+/****************************************************************************/
+/* Acquire NVRAM lock. */
+/* */
+/* Before the NVRAM can be accessed the caller must acquire an NVRAM lock. */
+/* Locks 0 and 2 are reserved, lock 1 is used by firmware and lock 2 is */
+/* for use by the driver. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_acquire_nvram_lock(struct bce_softc *sc)
+{
+ u32 val;
+ int j, rc = 0;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ /* Request access to the flash interface. */
+ REG_WR(sc, BCE_NVM_SW_ARB, BCE_NVM_SW_ARB_ARB_REQ_SET2);
+ for (j = 0; j < NVRAM_TIMEOUT_COUNT; j++) {
+ val = REG_RD(sc, BCE_NVM_SW_ARB);
+ if (val & BCE_NVM_SW_ARB_ARB_ARB2)
+ break;
+
+ DELAY(5);
+ }
+
+ if (j >= NVRAM_TIMEOUT_COUNT) {
+ DBPRINT(sc, BCE_WARN, "Timeout acquiring NVRAM lock!\n");
+ rc = EBUSY;
+ }
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return (rc);
+}
+
+
+/****************************************************************************/
+/* Release NVRAM lock. */
+/* */
+/* When the caller is finished accessing NVRAM the lock must be released. */
+/* Locks 0 and 2 are reserved, lock 1 is used by firmware and lock 2 is */
+/* for use by the driver. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_release_nvram_lock(struct bce_softc *sc)
+{
+ u32 val;
+ int j, rc = 0;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ /*
+ * Relinquish nvram interface.
+ */
+ REG_WR(sc, BCE_NVM_SW_ARB, BCE_NVM_SW_ARB_ARB_REQ_CLR2);
+
+ for (j = 0; j < NVRAM_TIMEOUT_COUNT; j++) {
+ val = REG_RD(sc, BCE_NVM_SW_ARB);
+ if (!(val & BCE_NVM_SW_ARB_ARB_ARB2))
+ break;
+
+ DELAY(5);
+ }
+
+ if (j >= NVRAM_TIMEOUT_COUNT) {
+ DBPRINT(sc, BCE_WARN, "Timeout releasing NVRAM lock!\n");
+ rc = EBUSY;
+ }
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return (rc);
+}
+
+
+#ifdef BCE_NVRAM_WRITE_SUPPORT
+/****************************************************************************/
+/* Enable NVRAM write access. */
+/* */
+/* Before writing to NVRAM the caller must enable NVRAM writes. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_enable_nvram_write(struct bce_softc *sc)
+{
+ u32 val;
+ int rc = 0;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ val = REG_RD(sc, BCE_MISC_CFG);
+ REG_WR(sc, BCE_MISC_CFG, val | BCE_MISC_CFG_NVM_WR_EN_PCI);
+
+ if (!(sc->bce_flash_info->flags & BCE_NV_BUFFERED)) {
+ int j;
+
+ REG_WR(sc, BCE_NVM_COMMAND, BCE_NVM_COMMAND_DONE);
+ REG_WR(sc, BCE_NVM_COMMAND, BCE_NVM_COMMAND_WREN | BCE_NVM_COMMAND_DOIT);
+
+ for (j = 0; j < NVRAM_TIMEOUT_COUNT; j++) {
+ DELAY(5);
+
+ val = REG_RD(sc, BCE_NVM_COMMAND);
+ if (val & BCE_NVM_COMMAND_DONE)
+ break;
+ }
+
+ if (j >= NVRAM_TIMEOUT_COUNT) {
+ DBPRINT(sc, BCE_WARN, "Timeout writing NVRAM!\n");
+ rc = EBUSY;
+ }
+ }
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+ return (rc);
+}
+
+
+/****************************************************************************/
+/* Disable NVRAM write access. */
+/* */
+/* When the caller is finished writing to NVRAM write access must be */
+/* disabled. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_disable_nvram_write(struct bce_softc *sc)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ val = REG_RD(sc, BCE_MISC_CFG);
+ REG_WR(sc, BCE_MISC_CFG, val & ~BCE_MISC_CFG_NVM_WR_EN);
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+
+}
+#endif
+
+
+/****************************************************************************/
+/* Enable NVRAM access. */
+/* */
+/* Before accessing NVRAM for read or write operations the caller must */
+/* enabled NVRAM access. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_enable_nvram_access(struct bce_softc *sc)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ val = REG_RD(sc, BCE_NVM_ACCESS_ENABLE);
+ /* Enable both bits, even on read. */
+ REG_WR(sc, BCE_NVM_ACCESS_ENABLE, val |
+ BCE_NVM_ACCESS_ENABLE_EN | BCE_NVM_ACCESS_ENABLE_WR_EN);
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+}
+
+
+/****************************************************************************/
+/* Disable NVRAM access. */
+/* */
+/* When the caller is finished accessing NVRAM access must be disabled. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_disable_nvram_access(struct bce_softc *sc)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ val = REG_RD(sc, BCE_NVM_ACCESS_ENABLE);
+
+ /* Disable both bits, even after read. */
+ REG_WR(sc, BCE_NVM_ACCESS_ENABLE, val &
+ ~(BCE_NVM_ACCESS_ENABLE_EN | BCE_NVM_ACCESS_ENABLE_WR_EN));
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+}
+
+
+#ifdef BCE_NVRAM_WRITE_SUPPORT
+/****************************************************************************/
+/* Erase NVRAM page before writing. */
+/* */
+/* Non-buffered flash parts require that a page be erased before it is */
+/* written. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_nvram_erase_page(struct bce_softc *sc, u32 offset)
+{
+ u32 cmd;
+ int j, rc = 0;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ /* Buffered flash doesn't require an erase. */
+ if (sc->bce_flash_info->flags & BCE_NV_BUFFERED)
+ goto bce_nvram_erase_page_exit;
+
+ /* Build an erase command. */
+ cmd = BCE_NVM_COMMAND_ERASE | BCE_NVM_COMMAND_WR |
+ BCE_NVM_COMMAND_DOIT;
+
+ /*
+ * Clear the DONE bit separately, set the NVRAM adress to erase,
+ * and issue the erase command.
+ */
+ REG_WR(sc, BCE_NVM_COMMAND, BCE_NVM_COMMAND_DONE);
+ REG_WR(sc, BCE_NVM_ADDR, offset & BCE_NVM_ADDR_NVM_ADDR_VALUE);
+ REG_WR(sc, BCE_NVM_COMMAND, cmd);
+
+ /* Wait for completion. */
+ for (j = 0; j < NVRAM_TIMEOUT_COUNT; j++) {
+ u32 val;
+
+ DELAY(5);
+
+ val = REG_RD(sc, BCE_NVM_COMMAND);
+ if (val & BCE_NVM_COMMAND_DONE)
+ break;
+ }
+
+ if (j >= NVRAM_TIMEOUT_COUNT) {
+ DBPRINT(sc, BCE_WARN, "Timeout erasing NVRAM.\n");
+ rc = EBUSY;
+ }
+
+bce_nvram_erase_page_exit:
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return (rc);
+}
+#endif /* BCE_NVRAM_WRITE_SUPPORT */
+
+
+/****************************************************************************/
+/* Read a dword (32 bits) from NVRAM. */
+/* */
+/* Read a 32 bit word from NVRAM. The caller is assumed to have already */
+/* obtained the NVRAM lock and enabled the controller for NVRAM access. */
+/* */
+/* Returns: */
+/* 0 on success and the 32 bit value read, positive value on failure. */
+/****************************************************************************/
+static int
+bce_nvram_read_dword(struct bce_softc *sc,
+ u32 offset, u8 *ret_val, u32 cmd_flags)
+{
+ u32 cmd;
+ int i, rc = 0;
+
+ DBENTER(BCE_EXTREME_NVRAM);
+
+ /* Build the command word. */
+ cmd = BCE_NVM_COMMAND_DOIT | cmd_flags;
+
+ /* Calculate the offset for buffered flash if translation is used. */
+ if (sc->bce_flash_info->flags & BCE_NV_TRANSLATE) {
+ offset = ((offset / sc->bce_flash_info->page_size) <<
+ sc->bce_flash_info->page_bits) +
+ (offset % sc->bce_flash_info->page_size);
+ }
+
+ /*
+ * Clear the DONE bit separately, set the address to read,
+ * and issue the read.
+ */
+ REG_WR(sc, BCE_NVM_COMMAND, BCE_NVM_COMMAND_DONE);
+ REG_WR(sc, BCE_NVM_ADDR, offset & BCE_NVM_ADDR_NVM_ADDR_VALUE);
+ REG_WR(sc, BCE_NVM_COMMAND, cmd);
+
+ /* Wait for completion. */
+ for (i = 0; i < NVRAM_TIMEOUT_COUNT; i++) {
+ u32 val;
+
+ DELAY(5);
+
+ val = REG_RD(sc, BCE_NVM_COMMAND);
+ if (val & BCE_NVM_COMMAND_DONE) {
+ val = REG_RD(sc, BCE_NVM_READ);
+
+ val = bce_be32toh(val);
+ memcpy(ret_val, &val, 4);
+ break;
+ }
+ }
+
+ /* Check for errors. */
+ if (i >= NVRAM_TIMEOUT_COUNT) {
+ BCE_PRINTF("%s(%d): Timeout error reading NVRAM at "
+ "offset 0x%08X!\n", __FILE__, __LINE__, offset);
+ rc = EBUSY;
+ }
+
+ DBEXIT(BCE_EXTREME_NVRAM);
+ return(rc);
+}
+
+
+#ifdef BCE_NVRAM_WRITE_SUPPORT
+/****************************************************************************/
+/* Write a dword (32 bits) to NVRAM. */
+/* */
+/* Write a 32 bit word to NVRAM. The caller is assumed to have already */
+/* obtained the NVRAM lock, enabled the controller for NVRAM access, and */
+/* enabled NVRAM write access. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_nvram_write_dword(struct bce_softc *sc, u32 offset, u8 *val,
+ u32 cmd_flags)
+{
+ u32 cmd, val32;
+ int j, rc = 0;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ /* Build the command word. */
+ cmd = BCE_NVM_COMMAND_DOIT | BCE_NVM_COMMAND_WR | cmd_flags;
+
+ /* Calculate the offset for buffered flash if translation is used. */
+ if (sc->bce_flash_info->flags & BCE_NV_TRANSLATE) {
+ offset = ((offset / sc->bce_flash_info->page_size) <<
+ sc->bce_flash_info->page_bits) +
+ (offset % sc->bce_flash_info->page_size);
+ }
+
+ /*
+ * Clear the DONE bit separately, convert NVRAM data to big-endian,
+ * set the NVRAM address to write, and issue the write command
+ */
+ REG_WR(sc, BCE_NVM_COMMAND, BCE_NVM_COMMAND_DONE);
+ memcpy(&val32, val, 4);
+ val32 = htobe32(val32);
+ REG_WR(sc, BCE_NVM_WRITE, val32);
+ REG_WR(sc, BCE_NVM_ADDR, offset & BCE_NVM_ADDR_NVM_ADDR_VALUE);
+ REG_WR(sc, BCE_NVM_COMMAND, cmd);
+
+ /* Wait for completion. */
+ for (j = 0; j < NVRAM_TIMEOUT_COUNT; j++) {
+ DELAY(5);
+
+ if (REG_RD(sc, BCE_NVM_COMMAND) & BCE_NVM_COMMAND_DONE)
+ break;
+ }
+ if (j >= NVRAM_TIMEOUT_COUNT) {
+ BCE_PRINTF("%s(%d): Timeout error writing NVRAM at "
+ "offset 0x%08X\n", __FILE__, __LINE__, offset);
+ rc = EBUSY;
+ }
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return (rc);
+}
+#endif /* BCE_NVRAM_WRITE_SUPPORT */
+
+
+/****************************************************************************/
+/* Initialize NVRAM access. */
+/* */
+/* Identify the NVRAM device in use and prepare the NVRAM interface to */
+/* access that device. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_init_nvram(struct bce_softc *sc)
+{
+ u32 val;
+ int j, entry_count, rc = 0;
+ struct flash_spec *flash;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ sc->bce_flash_info = &flash_5709;
+ goto bce_init_nvram_get_flash_size;
+ }
+
+ /* Determine the selected interface. */
+ val = REG_RD(sc, BCE_NVM_CFG1);
+
+ entry_count = sizeof(flash_table) / sizeof(struct flash_spec);
+
+ /*
+ * Flash reconfiguration is required to support additional
+ * NVRAM devices not directly supported in hardware.
+ * Check if the flash interface was reconfigured
+ * by the bootcode.
+ */
+
+ if (val & 0x40000000) {
+ /* Flash interface reconfigured by bootcode. */
+
+ DBPRINT(sc,BCE_INFO_LOAD,
+ "bce_init_nvram(): Flash WAS reconfigured.\n");
+
+ for (j = 0, flash = &flash_table[0]; j < entry_count;
+ j++, flash++) {
+ if ((val & FLASH_BACKUP_STRAP_MASK) ==
+ (flash->config1 & FLASH_BACKUP_STRAP_MASK)) {
+ sc->bce_flash_info = flash;
+ break;
+ }
+ }
+ } else {
+ /* Flash interface not yet reconfigured. */
+ u32 mask;
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): Flash was NOT reconfigured.\n",
+ __FUNCTION__);
+
+ if (val & (1 << 23))
+ mask = FLASH_BACKUP_STRAP_MASK;
+ else
+ mask = FLASH_STRAP_MASK;
+
+ /* Look for the matching NVRAM device configuration data. */
+ for (j = 0, flash = &flash_table[0]; j < entry_count; j++, flash++) {
+
+ /* Check if the device matches any of the known devices. */
+ if ((val & mask) == (flash->strapping & mask)) {
+ /* Found a device match. */
+ sc->bce_flash_info = flash;
+
+ /* Request access to the flash interface. */
+ if ((rc = bce_acquire_nvram_lock(sc)) != 0)
+ return rc;
+
+ /* Reconfigure the flash interface. */
+ bce_enable_nvram_access(sc);
+ REG_WR(sc, BCE_NVM_CFG1, flash->config1);
+ REG_WR(sc, BCE_NVM_CFG2, flash->config2);
+ REG_WR(sc, BCE_NVM_CFG3, flash->config3);
+ REG_WR(sc, BCE_NVM_WRITE1, flash->write1);
+ bce_disable_nvram_access(sc);
+ bce_release_nvram_lock(sc);
+
+ break;
+ }
+ }
+ }
+
+ /* Check if a matching device was found. */
+ if (j == entry_count) {
+ sc->bce_flash_info = NULL;
+ BCE_PRINTF("%s(%d): Unknown Flash NVRAM found!\n",
+ __FILE__, __LINE__);
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return (ENODEV);
+ }
+
+bce_init_nvram_get_flash_size:
+ /* Write the flash config data to the shared memory interface. */
+ val = bce_shmem_rd(sc, BCE_SHARED_HW_CFG_CONFIG2);
+ val &= BCE_SHARED_HW_CFG2_NVM_SIZE_MASK;
+ if (val)
+ sc->bce_flash_size = val;
+ else
+ sc->bce_flash_size = sc->bce_flash_info->total_size;
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): Found %s, size = 0x%08X\n",
+ __FUNCTION__, sc->bce_flash_info->name,
+ sc->bce_flash_info->total_size);
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return rc;
+}
+
+
+/****************************************************************************/
+/* Read an arbitrary range of data from NVRAM. */
+/* */
+/* Prepares the NVRAM interface for access and reads the requested data */
+/* into the supplied buffer. */
+/* */
+/* Returns: */
+/* 0 on success and the data read, positive value on failure. */
+/****************************************************************************/
+static int
+bce_nvram_read(struct bce_softc *sc, u32 offset, u8 *ret_buf,
+ int buf_size)
+{
+ int rc = 0;
+ u32 cmd_flags, offset32, len32, extra;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ if (buf_size == 0)
+ goto bce_nvram_read_exit;
+
+ /* Request access to the flash interface. */
+ if ((rc = bce_acquire_nvram_lock(sc)) != 0)
+ goto bce_nvram_read_exit;
+
+ /* Enable access to flash interface */
+ bce_enable_nvram_access(sc);
+
+ len32 = buf_size;
+ offset32 = offset;
+ extra = 0;
+
+ cmd_flags = 0;
+
+ if (offset32 & 3) {
+ u8 buf[4];
+ u32 pre_len;
+
+ offset32 &= ~3;
+ pre_len = 4 - (offset & 3);
+
+ if (pre_len >= len32) {
+ pre_len = len32;
+ cmd_flags = BCE_NVM_COMMAND_FIRST | BCE_NVM_COMMAND_LAST;
+ }
+ else {
+ cmd_flags = BCE_NVM_COMMAND_FIRST;
+ }
+
+ rc = bce_nvram_read_dword(sc, offset32, buf, cmd_flags);
+
+ if (rc)
+ return rc;
+
+ memcpy(ret_buf, buf + (offset & 3), pre_len);
+
+ offset32 += 4;
+ ret_buf += pre_len;
+ len32 -= pre_len;
+ }
+
+ if (len32 & 3) {
+ extra = 4 - (len32 & 3);
+ len32 = (len32 + 4) & ~3;
+ }
+
+ if (len32 == 4) {
+ u8 buf[4];
+
+ if (cmd_flags)
+ cmd_flags = BCE_NVM_COMMAND_LAST;
+ else
+ cmd_flags = BCE_NVM_COMMAND_FIRST |
+ BCE_NVM_COMMAND_LAST;
+
+ rc = bce_nvram_read_dword(sc, offset32, buf, cmd_flags);
+
+ memcpy(ret_buf, buf, 4 - extra);
+ }
+ else if (len32 > 0) {
+ u8 buf[4];
+
+ /* Read the first word. */
+ if (cmd_flags)
+ cmd_flags = 0;
+ else
+ cmd_flags = BCE_NVM_COMMAND_FIRST;
+
+ rc = bce_nvram_read_dword(sc, offset32, ret_buf, cmd_flags);
+
+ /* Advance to the next dword. */
+ offset32 += 4;
+ ret_buf += 4;
+ len32 -= 4;
+
+ while (len32 > 4 && rc == 0) {
+ rc = bce_nvram_read_dword(sc, offset32, ret_buf, 0);
+
+ /* Advance to the next dword. */
+ offset32 += 4;
+ ret_buf += 4;
+ len32 -= 4;
+ }
+
+ if (rc)
+ goto bce_nvram_read_locked_exit;
+
+ cmd_flags = BCE_NVM_COMMAND_LAST;
+ rc = bce_nvram_read_dword(sc, offset32, buf, cmd_flags);
+
+ memcpy(ret_buf, buf, 4 - extra);
+ }
+
+bce_nvram_read_locked_exit:
+ /* Disable access to flash interface and release the lock. */
+ bce_disable_nvram_access(sc);
+ bce_release_nvram_lock(sc);
+
+bce_nvram_read_exit:
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return rc;
+}
+
+
+#ifdef BCE_NVRAM_WRITE_SUPPORT
+/****************************************************************************/
+/* Write an arbitrary range of data from NVRAM. */
+/* */
+/* Prepares the NVRAM interface for write access and writes the requested */
+/* data from the supplied buffer. The caller is responsible for */
+/* calculating any appropriate CRCs. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_nvram_write(struct bce_softc *sc, u32 offset, u8 *data_buf,
+ int buf_size)
+{
+ u32 written, offset32, len32;
+ u8 *buf, start[4], end[4];
+ int rc = 0;
+ int align_start, align_end;
+
+ DBENTER(BCE_VERBOSE_NVRAM);
+
+ buf = data_buf;
+ offset32 = offset;
+ len32 = buf_size;
+ align_start = align_end = 0;
+
+ if ((align_start = (offset32 & 3))) {
+ offset32 &= ~3;
+ len32 += align_start;
+ if ((rc = bce_nvram_read(sc, offset32, start, 4)))
+ goto bce_nvram_write_exit;
+ }
+
+ if (len32 & 3) {
+ if ((len32 > 4) || !align_start) {
+ align_end = 4 - (len32 & 3);
+ len32 += align_end;
+ if ((rc = bce_nvram_read(sc, offset32 + len32 - 4,
+ end, 4))) {
+ goto bce_nvram_write_exit;
+ }
+ }
+ }
+
+ if (align_start || align_end) {
+ buf = malloc(len32, M_DEVBUF, M_NOWAIT);
+ if (buf == 0) {
+ rc = ENOMEM;
+ goto bce_nvram_write_exit;
+ }
+
+ if (align_start) {
+ memcpy(buf, start, 4);
+ }
+
+ if (align_end) {
+ memcpy(buf + len32 - 4, end, 4);
+ }
+ memcpy(buf + align_start, data_buf, buf_size);
+ }
+
+ written = 0;
+ while ((written < len32) && (rc == 0)) {
+ u32 page_start, page_end, data_start, data_end;
+ u32 addr, cmd_flags;
+ int i;
+ u8 flash_buffer[264];
+
+ /* Find the page_start addr */
+ page_start = offset32 + written;
+ page_start -= (page_start % sc->bce_flash_info->page_size);
+ /* Find the page_end addr */
+ page_end = page_start + sc->bce_flash_info->page_size;
+ /* Find the data_start addr */
+ data_start = (written == 0) ? offset32 : page_start;
+ /* Find the data_end addr */
+ data_end = (page_end > offset32 + len32) ?
+ (offset32 + len32) : page_end;
+
+ /* Request access to the flash interface. */
+ if ((rc = bce_acquire_nvram_lock(sc)) != 0)
+ goto bce_nvram_write_exit;
+
+ /* Enable access to flash interface */
+ bce_enable_nvram_access(sc);
+
+ cmd_flags = BCE_NVM_COMMAND_FIRST;
+ if (!(sc->bce_flash_info->flags & BCE_NV_BUFFERED)) {
+ int j;
+
+ /* Read the whole page into the buffer
+ * (non-buffer flash only) */
+ for (j = 0; j < sc->bce_flash_info->page_size; j += 4) {
+ if (j == (sc->bce_flash_info->page_size - 4)) {
+ cmd_flags |= BCE_NVM_COMMAND_LAST;
+ }
+ rc = bce_nvram_read_dword(sc,
+ page_start + j,
+ &flash_buffer[j],
+ cmd_flags);
+
+ if (rc)
+ goto bce_nvram_write_locked_exit;
+
+ cmd_flags = 0;
+ }
+ }
+
+ /* Enable writes to flash interface (unlock write-protect) */
+ if ((rc = bce_enable_nvram_write(sc)) != 0)
+ goto bce_nvram_write_locked_exit;
+
+ /* Erase the page */
+ if ((rc = bce_nvram_erase_page(sc, page_start)) != 0)
+ goto bce_nvram_write_locked_exit;
+
+ /* Re-enable the write again for the actual write */
+ bce_enable_nvram_write(sc);
+
+ /* Loop to write back the buffer data from page_start to
+ * data_start */
+ i = 0;
+ if (!(sc->bce_flash_info->flags & BCE_NV_BUFFERED)) {
+ for (addr = page_start; addr < data_start;
+ addr += 4, i += 4) {
+
+ rc = bce_nvram_write_dword(sc, addr,
+ &flash_buffer[i], cmd_flags);
+
+ if (rc != 0)
+ goto bce_nvram_write_locked_exit;
+
+ cmd_flags = 0;
+ }
+ }
+
+ /* Loop to write the new data from data_start to data_end */
+ for (addr = data_start; addr < data_end; addr += 4, i++) {
+ if ((addr == page_end - 4) ||
+ ((sc->bce_flash_info->flags & BCE_NV_BUFFERED) &&
+ (addr == data_end - 4))) {
+
+ cmd_flags |= BCE_NVM_COMMAND_LAST;
+ }
+ rc = bce_nvram_write_dword(sc, addr, buf,
+ cmd_flags);
+
+ if (rc != 0)
+ goto bce_nvram_write_locked_exit;
+
+ cmd_flags = 0;
+ buf += 4;
+ }
+
+ /* Loop to write back the buffer data from data_end
+ * to page_end */
+ if (!(sc->bce_flash_info->flags & BCE_NV_BUFFERED)) {
+ for (addr = data_end; addr < page_end;
+ addr += 4, i += 4) {
+
+ if (addr == page_end-4) {
+ cmd_flags = BCE_NVM_COMMAND_LAST;
+ }
+ rc = bce_nvram_write_dword(sc, addr,
+ &flash_buffer[i], cmd_flags);
+
+ if (rc != 0)
+ goto bce_nvram_write_locked_exit;
+
+ cmd_flags = 0;
+ }
+ }
+
+ /* Disable writes to flash interface (lock write-protect) */
+ bce_disable_nvram_write(sc);
+
+ /* Disable access to flash interface */
+ bce_disable_nvram_access(sc);
+ bce_release_nvram_lock(sc);
+
+ /* Increment written */
+ written += data_end - data_start;
+ }
+
+ goto bce_nvram_write_exit;
+
+bce_nvram_write_locked_exit:
+ bce_disable_nvram_write(sc);
+ bce_disable_nvram_access(sc);
+ bce_release_nvram_lock(sc);
+
+bce_nvram_write_exit:
+ if (align_start || align_end)
+ free(buf, M_DEVBUF);
+
+ DBEXIT(BCE_VERBOSE_NVRAM);
+ return (rc);
+}
+#endif /* BCE_NVRAM_WRITE_SUPPORT */
+
+
+/****************************************************************************/
+/* Verifies that NVRAM is accessible and contains valid data. */
+/* */
+/* Reads the configuration data from NVRAM and verifies that the CRC is */
+/* correct. */
+/* */
+/* Returns: */
+/* 0 on success, positive value on failure. */
+/****************************************************************************/
+static int
+bce_nvram_test(struct bce_softc *sc)
+{
+ u32 buf[BCE_NVRAM_SIZE / 4];
+ u8 *data = (u8 *) buf;
+ int rc = 0;
+ u32 magic, csum;
+
+ DBENTER(BCE_VERBOSE_NVRAM | BCE_VERBOSE_LOAD | BCE_VERBOSE_RESET);
+
+ /*
+ * Check that the device NVRAM is valid by reading
+ * the magic value at offset 0.
+ */
+ if ((rc = bce_nvram_read(sc, 0, data, 4)) != 0) {
+ BCE_PRINTF("%s(%d): Unable to read NVRAM!\n",
+ __FILE__, __LINE__);
+ goto bce_nvram_test_exit;
+ }
+
+ /*
+ * Verify that offset 0 of the NVRAM contains
+ * a valid magic number.
+ */
+ magic = bce_be32toh(buf[0]);
+ if (magic != BCE_NVRAM_MAGIC) {
+ rc = ENODEV;
+ BCE_PRINTF("%s(%d): Invalid NVRAM magic value! "
+ "Expected: 0x%08X, Found: 0x%08X\n",
+ __FILE__, __LINE__, BCE_NVRAM_MAGIC, magic);
+ goto bce_nvram_test_exit;
+ }
+
+ /*
+ * Verify that the device NVRAM includes valid
+ * configuration data.
+ */
+ if ((rc = bce_nvram_read(sc, 0x100, data, BCE_NVRAM_SIZE)) != 0) {
+ BCE_PRINTF("%s(%d): Unable to read manufacturing "
+ "Information from NVRAM!\n", __FILE__, __LINE__);
+ goto bce_nvram_test_exit;
+ }
+
+ csum = ether_crc32_le(data, 0x100);
+ if (csum != BCE_CRC32_RESIDUAL) {
+ rc = ENODEV;
+ BCE_PRINTF("%s(%d): Invalid manufacturing information "
+ "NVRAM CRC! Expected: 0x%08X, Found: 0x%08X\n",
+ __FILE__, __LINE__, BCE_CRC32_RESIDUAL, csum);
+ goto bce_nvram_test_exit;
+ }
+
+ csum = ether_crc32_le(data + 0x100, 0x100);
+ if (csum != BCE_CRC32_RESIDUAL) {
+ rc = ENODEV;
+ BCE_PRINTF("%s(%d): Invalid feature configuration "
+ "information NVRAM CRC! Expected: 0x%08X, "
+ "Found: 08%08X\n", __FILE__, __LINE__,
+ BCE_CRC32_RESIDUAL, csum);
+ }
+
+bce_nvram_test_exit:
+ DBEXIT(BCE_VERBOSE_NVRAM | BCE_VERBOSE_LOAD | BCE_VERBOSE_RESET);
+ return rc;
+}
+
+
+/****************************************************************************/
+/* Identifies the current media type of the controller and sets the PHY */
+/* address. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_get_media(struct bce_softc *sc)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_PHY);
+
+ /* Assume PHY address for copper controllers. */
+ sc->bce_phy_addr = 1;
+
+ if (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) {
+ u32 val = REG_RD(sc, BCE_MISC_DUAL_MEDIA_CTRL);
+ u32 bond_id = val & BCE_MISC_DUAL_MEDIA_CTRL_BOND_ID;
+ u32 strap;
+
+ /*
+ * The BCM5709S is software configurable
+ * for Copper or SerDes operation.
+ */
+ if (bond_id == BCE_MISC_DUAL_MEDIA_CTRL_BOND_ID_C) {
+ DBPRINT(sc, BCE_INFO_LOAD, "5709 bonded "
+ "for copper.\n");
+ goto bce_get_media_exit;
+ } else if (bond_id == BCE_MISC_DUAL_MEDIA_CTRL_BOND_ID_S) {
+ DBPRINT(sc, BCE_INFO_LOAD, "5709 bonded "
+ "for dual media.\n");
+ sc->bce_phy_flags |= BCE_PHY_SERDES_FLAG;
+ goto bce_get_media_exit;
+ }
+
+ if (val & BCE_MISC_DUAL_MEDIA_CTRL_STRAP_OVERRIDE)
+ strap = (val &
+ BCE_MISC_DUAL_MEDIA_CTRL_PHY_CTRL) >> 21;
+ else
+ strap = (val &
+ BCE_MISC_DUAL_MEDIA_CTRL_PHY_CTRL_STRAP) >> 8;
+
+ if (pci_get_function(sc->bce_dev) == 0) {
+ switch (strap) {
+ case 0x4:
+ case 0x5:
+ case 0x6:
+ DBPRINT(sc, BCE_INFO_LOAD,
+ "BCM5709 s/w configured for SerDes.\n");
+ sc->bce_phy_flags |= BCE_PHY_SERDES_FLAG;
+ break;
+ default:
+ DBPRINT(sc, BCE_INFO_LOAD,
+ "BCM5709 s/w configured for Copper.\n");
+ break;
+ }
+ } else {
+ switch (strap) {
+ case 0x1:
+ case 0x2:
+ case 0x4:
+ DBPRINT(sc, BCE_INFO_LOAD,
+ "BCM5709 s/w configured for SerDes.\n");
+ sc->bce_phy_flags |= BCE_PHY_SERDES_FLAG;
+ break;
+ default:
+ DBPRINT(sc, BCE_INFO_LOAD,
+ "BCM5709 s/w configured for Copper.\n");
+ break;
+ }
+ }
+
+ } else if (BCE_CHIP_BOND_ID(sc) & BCE_CHIP_BOND_ID_SERDES_BIT)
+ sc->bce_phy_flags |= BCE_PHY_SERDES_FLAG;
+
+ if (sc->bce_phy_flags & BCE_PHY_SERDES_FLAG) {
+
+ sc->bce_flags |= BCE_NO_WOL_FLAG;
+
+ if (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709)
+ sc->bce_phy_flags |= BCE_PHY_IEEE_CLAUSE_45_FLAG;
+
+ if (BCE_CHIP_NUM(sc) != BCE_CHIP_NUM_5706) {
+ /* 5708S/09S/16S use a separate PHY for SerDes. */
+ sc->bce_phy_addr = 2;
+
+ val = bce_shmem_rd(sc, BCE_SHARED_HW_CFG_CONFIG);
+ if (val & BCE_SHARED_HW_CFG_PHY_2_5G) {
+ sc->bce_phy_flags |=
+ BCE_PHY_2_5G_CAPABLE_FLAG;
+ DBPRINT(sc, BCE_INFO_LOAD, "Found 2.5Gb "
+ "capable adapter\n");
+ }
+ }
+ } else if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5706) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5708))
+ sc->bce_phy_flags |= BCE_PHY_CRC_FIX_FLAG;
+
+bce_get_media_exit:
+ DBPRINT(sc, (BCE_INFO_LOAD | BCE_INFO_PHY),
+ "Using PHY address %d.\n", sc->bce_phy_addr);
+
+ DBEXIT(BCE_VERBOSE_PHY);
+}
+
+
+/****************************************************************************/
+/* Performs PHY initialization required before MII drivers access the */
+/* device. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_media(struct bce_softc *sc)
+{
+ if ((sc->bce_phy_flags & BCE_PHY_IEEE_CLAUSE_45_FLAG) != 0) {
+ /*
+ * Configure 5709S/5716S PHYs to use traditional IEEE
+ * Clause 22 method. Otherwise we have no way to attach
+ * the PHY in mii(4) layer. PHY specific configuration
+ * is done in mii layer.
+ */
+
+ /* Select auto-negotiation MMD of the PHY. */
+ bce_miibus_write_reg(sc->bce_dev, sc->bce_phy_addr,
+ BRGPHY_BLOCK_ADDR, BRGPHY_BLOCK_ADDR_ADDR_EXT);
+ bce_miibus_write_reg(sc->bce_dev, sc->bce_phy_addr,
+ BRGPHY_ADDR_EXT, BRGPHY_ADDR_EXT_AN_MMD);
+
+ /* Set IEEE0 block of AN MMD (assumed in brgphy(4) code). */
+ bce_miibus_write_reg(sc->bce_dev, sc->bce_phy_addr,
+ BRGPHY_BLOCK_ADDR, BRGPHY_BLOCK_ADDR_COMBO_IEEE0);
+ }
+}
+
+
+/****************************************************************************/
+/* Free any DMA memory owned by the driver. */
+/* */
+/* Scans through each data structre that requires DMA memory and frees */
+/* the memory if allocated. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_dma_free(struct bce_softc *sc)
+{
+ int i;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_UNLOAD | BCE_VERBOSE_CTX);
+
+ /* Free, unmap, and destroy the status block. */
+ if (sc->status_block != NULL) {
+ bus_dmamem_free(
+ sc->status_tag,
+ sc->status_block,
+ sc->status_map);
+ sc->status_block = NULL;
+ }
+
+ if (sc->status_map != NULL) {
+ bus_dmamap_unload(
+ sc->status_tag,
+ sc->status_map);
+ bus_dmamap_destroy(sc->status_tag,
+ sc->status_map);
+ sc->status_map = NULL;
+ }
+
+ if (sc->status_tag != NULL) {
+ bus_dma_tag_destroy(sc->status_tag);
+ sc->status_tag = NULL;
+ }
+
+
+ /* Free, unmap, and destroy the statistics block. */
+ if (sc->stats_block != NULL) {
+ bus_dmamem_free(
+ sc->stats_tag,
+ sc->stats_block,
+ sc->stats_map);
+ sc->stats_block = NULL;
+ }
+
+ if (sc->stats_map != NULL) {
+ bus_dmamap_unload(
+ sc->stats_tag,
+ sc->stats_map);
+ bus_dmamap_destroy(sc->stats_tag,
+ sc->stats_map);
+ sc->stats_map = NULL;
+ }
+
+ if (sc->stats_tag != NULL) {
+ bus_dma_tag_destroy(sc->stats_tag);
+ sc->stats_tag = NULL;
+ }
+
+
+ /* Free, unmap and destroy all context memory pages. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ for (i = 0; i < sc->ctx_pages; i++ ) {
+ if (sc->ctx_block[i] != NULL) {
+ bus_dmamem_free(
+ sc->ctx_tag,
+ sc->ctx_block[i],
+ sc->ctx_map[i]);
+ sc->ctx_block[i] = NULL;
+ }
+
+ if (sc->ctx_map[i] != NULL) {
+ bus_dmamap_unload(
+ sc->ctx_tag,
+ sc->ctx_map[i]);
+ bus_dmamap_destroy(
+ sc->ctx_tag,
+ sc->ctx_map[i]);
+ sc->ctx_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the context memory tag. */
+ if (sc->ctx_tag != NULL) {
+ bus_dma_tag_destroy(sc->ctx_tag);
+ sc->ctx_tag = NULL;
+ }
+ }
+
+
+ /* Free, unmap and destroy all TX buffer descriptor chain pages. */
+ for (i = 0; i < TX_PAGES; i++ ) {
+ if (sc->tx_bd_chain[i] != NULL) {
+ bus_dmamem_free(
+ sc->tx_bd_chain_tag,
+ sc->tx_bd_chain[i],
+ sc->tx_bd_chain_map[i]);
+ sc->tx_bd_chain[i] = NULL;
+ }
+
+ if (sc->tx_bd_chain_map[i] != NULL) {
+ bus_dmamap_unload(
+ sc->tx_bd_chain_tag,
+ sc->tx_bd_chain_map[i]);
+ bus_dmamap_destroy(
+ sc->tx_bd_chain_tag,
+ sc->tx_bd_chain_map[i]);
+ sc->tx_bd_chain_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the TX buffer descriptor tag. */
+ if (sc->tx_bd_chain_tag != NULL) {
+ bus_dma_tag_destroy(sc->tx_bd_chain_tag);
+ sc->tx_bd_chain_tag = NULL;
+ }
+
+
+ /* Free, unmap and destroy all RX buffer descriptor chain pages. */
+ for (i = 0; i < RX_PAGES; i++ ) {
+ if (sc->rx_bd_chain[i] != NULL) {
+ bus_dmamem_free(
+ sc->rx_bd_chain_tag,
+ sc->rx_bd_chain[i],
+ sc->rx_bd_chain_map[i]);
+ sc->rx_bd_chain[i] = NULL;
+ }
+
+ if (sc->rx_bd_chain_map[i] != NULL) {
+ bus_dmamap_unload(
+ sc->rx_bd_chain_tag,
+ sc->rx_bd_chain_map[i]);
+ bus_dmamap_destroy(
+ sc->rx_bd_chain_tag,
+ sc->rx_bd_chain_map[i]);
+ sc->rx_bd_chain_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the RX buffer descriptor tag. */
+ if (sc->rx_bd_chain_tag != NULL) {
+ bus_dma_tag_destroy(sc->rx_bd_chain_tag);
+ sc->rx_bd_chain_tag = NULL;
+ }
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ /* Free, unmap and destroy all page buffer descriptor chain pages. */
+ for (i = 0; i < PG_PAGES; i++ ) {
+ if (sc->pg_bd_chain[i] != NULL) {
+ bus_dmamem_free(
+ sc->pg_bd_chain_tag,
+ sc->pg_bd_chain[i],
+ sc->pg_bd_chain_map[i]);
+ sc->pg_bd_chain[i] = NULL;
+ }
+
+ if (sc->pg_bd_chain_map[i] != NULL) {
+ bus_dmamap_unload(
+ sc->pg_bd_chain_tag,
+ sc->pg_bd_chain_map[i]);
+ bus_dmamap_destroy(
+ sc->pg_bd_chain_tag,
+ sc->pg_bd_chain_map[i]);
+ sc->pg_bd_chain_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the page buffer descriptor tag. */
+ if (sc->pg_bd_chain_tag != NULL) {
+ bus_dma_tag_destroy(sc->pg_bd_chain_tag);
+ sc->pg_bd_chain_tag = NULL;
+ }
+#endif
+
+
+ /* Unload and destroy the TX mbuf maps. */
+ for (i = 0; i < TOTAL_TX_BD; i++) {
+ if (sc->tx_mbuf_map[i] != NULL) {
+ bus_dmamap_unload(sc->tx_mbuf_tag,
+ sc->tx_mbuf_map[i]);
+ bus_dmamap_destroy(sc->tx_mbuf_tag,
+ sc->tx_mbuf_map[i]);
+ sc->tx_mbuf_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the TX mbuf tag. */
+ if (sc->tx_mbuf_tag != NULL) {
+ bus_dma_tag_destroy(sc->tx_mbuf_tag);
+ sc->tx_mbuf_tag = NULL;
+ }
+
+ /* Unload and destroy the RX mbuf maps. */
+ for (i = 0; i < TOTAL_RX_BD; i++) {
+ if (sc->rx_mbuf_map[i] != NULL) {
+ bus_dmamap_unload(sc->rx_mbuf_tag,
+ sc->rx_mbuf_map[i]);
+ bus_dmamap_destroy(sc->rx_mbuf_tag,
+ sc->rx_mbuf_map[i]);
+ sc->rx_mbuf_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the RX mbuf tag. */
+ if (sc->rx_mbuf_tag != NULL) {
+ bus_dma_tag_destroy(sc->rx_mbuf_tag);
+ sc->rx_mbuf_tag = NULL;
+ }
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ /* Unload and destroy the page mbuf maps. */
+ for (i = 0; i < TOTAL_PG_BD; i++) {
+ if (sc->pg_mbuf_map[i] != NULL) {
+ bus_dmamap_unload(sc->pg_mbuf_tag,
+ sc->pg_mbuf_map[i]);
+ bus_dmamap_destroy(sc->pg_mbuf_tag,
+ sc->pg_mbuf_map[i]);
+ sc->pg_mbuf_map[i] = NULL;
+ }
+ }
+
+ /* Destroy the page mbuf tag. */
+ if (sc->pg_mbuf_tag != NULL) {
+ bus_dma_tag_destroy(sc->pg_mbuf_tag);
+ sc->pg_mbuf_tag = NULL;
+ }
+#endif
+
+ /* Destroy the parent tag */
+ if (sc->parent_tag != NULL) {
+ bus_dma_tag_destroy(sc->parent_tag);
+ sc->parent_tag = NULL;
+ }
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_UNLOAD | BCE_VERBOSE_CTX);
+}
+
+
+/****************************************************************************/
+/* Get DMA memory from the OS. */
+/* */
+/* Validates that the OS has provided DMA buffers in response to a */
+/* bus_dmamap_load() call and saves the physical address of those buffers. */
+/* When the callback is used the OS will return 0 for the mapping function */
+/* (bus_dmamap_load()) so we use the value of map_arg->maxsegs to pass any */
+/* failures back to the caller. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
+{
+ bus_addr_t *busaddr = arg;
+
+ KASSERT(nseg == 1, ("%s(): Too many segments returned (%d)!",
+ __FUNCTION__, nseg));
+ /* Simulate a mapping failure. */
+ DBRUNIF(DB_RANDOMTRUE(dma_map_addr_failed_sim_control),
+ error = ENOMEM);
+
+ /* ToDo: How to increment debug sim_count variable here? */
+
+ /* Check for an error and signal the caller that an error occurred. */
+ if (error) {
+ *busaddr = 0;
+ } else {
+ *busaddr = segs->ds_addr;
+ }
+
+ return;
+}
+
+
+/****************************************************************************/
+/* Allocate any DMA memory needed by the driver. */
+/* */
+/* Allocates DMA memory needed for the various global structures needed by */
+/* hardware. */
+/* */
+/* Memory alignment requirements: */
+/* +-----------------+----------+----------+----------+----------+ */
+/* | | 5706 | 5708 | 5709 | 5716 | */
+/* +-----------------+----------+----------+----------+----------+ */
+/* |Status Block | 8 bytes | 8 bytes | 16 bytes | 16 bytes | */
+/* |Statistics Block | 8 bytes | 8 bytes | 16 bytes | 16 bytes | */
+/* |RX Buffers | 16 bytes | 16 bytes | 16 bytes | 16 bytes | */
+/* |PG Buffers | none | none | none | none | */
+/* |TX Buffers | none | none | none | none | */
+/* |Chain Pages(1) | 4KiB | 4KiB | 4KiB | 4KiB | */
+/* |Context Memory | | | | | */
+/* +-----------------+----------+----------+----------+----------+ */
+/* */
+/* (1) Must align with CPU page size (BCM_PAGE_SZIE). */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_dma_alloc(device_t dev)
+{
+ struct bce_softc *sc;
+ int i, error, rc = 0;
+ bus_size_t max_size, max_seg_size;
+ int max_segments;
+
+ sc = device_get_softc(dev);
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_CTX);
+
+ /*
+ * Allocate the parent bus DMA tag appropriate for PCI.
+ */
+ if (bus_dma_tag_create(bus_get_dma_tag(dev), 1, BCE_DMA_BOUNDARY,
+ sc->max_bus_addr, BUS_SPACE_MAXADDR, NULL, NULL,
+ BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0, NULL, NULL,
+ &sc->parent_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate parent DMA tag!\n",
+ __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ /*
+ * Create a DMA tag for the status block, allocate and clear the
+ * memory, map the memory into DMA space, and fetch the physical
+ * address of the block.
+ */
+ if (bus_dma_tag_create(sc->parent_tag, BCE_DMA_ALIGN,
+ BCE_DMA_BOUNDARY, sc->max_bus_addr, BUS_SPACE_MAXADDR,
+ NULL, NULL, BCE_STATUS_BLK_SZ, 1, BCE_STATUS_BLK_SZ,
+ 0, NULL, NULL, &sc->status_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate status block "
+ "DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ if(bus_dmamem_alloc(sc->status_tag, (void **)&sc->status_block,
+ BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT,
+ &sc->status_map)) {
+ BCE_PRINTF("%s(%d): Could not allocate status block "
+ "DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ error = bus_dmamap_load(sc->status_tag, sc->status_map,
+ sc->status_block, BCE_STATUS_BLK_SZ, bce_dma_map_addr,
+ &sc->status_block_paddr, BUS_DMA_NOWAIT);
+
+ if (error) {
+ BCE_PRINTF("%s(%d): Could not map status block "
+ "DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): status_block_paddr = 0x%jX\n",
+ __FUNCTION__, (uintmax_t) sc->status_block_paddr);
+
+ /*
+ * Create a DMA tag for the statistics block, allocate and clear the
+ * memory, map the memory into DMA space, and fetch the physical
+ * address of the block.
+ */
+ if (bus_dma_tag_create(sc->parent_tag, BCE_DMA_ALIGN,
+ BCE_DMA_BOUNDARY, sc->max_bus_addr, BUS_SPACE_MAXADDR,
+ NULL, NULL, BCE_STATS_BLK_SZ, 1, BCE_STATS_BLK_SZ,
+ 0, NULL, NULL, &sc->stats_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate statistics block "
+ "DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ if (bus_dmamem_alloc(sc->stats_tag, (void **)&sc->stats_block,
+ BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT, &sc->stats_map)) {
+ BCE_PRINTF("%s(%d): Could not allocate statistics block "
+ "DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ error = bus_dmamap_load(sc->stats_tag, sc->stats_map,
+ sc->stats_block, BCE_STATS_BLK_SZ, bce_dma_map_addr,
+ &sc->stats_block_paddr, BUS_DMA_NOWAIT);
+
+ if(error) {
+ BCE_PRINTF("%s(%d): Could not map statistics block "
+ "DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): stats_block_paddr = 0x%jX\n",
+ __FUNCTION__, (uintmax_t) sc->stats_block_paddr);
+
+ /* BCM5709 uses host memory as cache for context memory. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ sc->ctx_pages = 0x2000 / BCM_PAGE_SIZE;
+ if (sc->ctx_pages == 0)
+ sc->ctx_pages = 1;
+
+ DBRUNIF((sc->ctx_pages > 512),
+ BCE_PRINTF("%s(%d): Too many CTX pages! %d > 512\n",
+ __FILE__, __LINE__, sc->ctx_pages));
+
+ /*
+ * Create a DMA tag for the context pages,
+ * allocate and clear the memory, map the
+ * memory into DMA space, and fetch the
+ * physical address of the block.
+ */
+ if(bus_dma_tag_create(sc->parent_tag, BCM_PAGE_SIZE,
+ BCE_DMA_BOUNDARY, sc->max_bus_addr, BUS_SPACE_MAXADDR,
+ NULL, NULL, BCM_PAGE_SIZE, 1, BCM_PAGE_SIZE,
+ 0, NULL, NULL, &sc->ctx_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate CTX "
+ "DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ for (i = 0; i < sc->ctx_pages; i++) {
+
+ if(bus_dmamem_alloc(sc->ctx_tag,
+ (void **)&sc->ctx_block[i],
+ BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT,
+ &sc->ctx_map[i])) {
+ BCE_PRINTF("%s(%d): Could not allocate CTX "
+ "DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ error = bus_dmamap_load(sc->ctx_tag, sc->ctx_map[i],
+ sc->ctx_block[i], BCM_PAGE_SIZE, bce_dma_map_addr,
+ &sc->ctx_paddr[i], BUS_DMA_NOWAIT);
+
+ if (error) {
+ BCE_PRINTF("%s(%d): Could not map CTX "
+ "DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): ctx_paddr[%d] "
+ "= 0x%jX\n", __FUNCTION__, i,
+ (uintmax_t) sc->ctx_paddr[i]);
+ }
+ }
+
+ /*
+ * Create a DMA tag for the TX buffer descriptor chain,
+ * allocate and clear the memory, and fetch the
+ * physical address of the block.
+ */
+ if(bus_dma_tag_create(sc->parent_tag, BCM_PAGE_SIZE, BCE_DMA_BOUNDARY,
+ sc->max_bus_addr, BUS_SPACE_MAXADDR, NULL, NULL,
+ BCE_TX_CHAIN_PAGE_SZ, 1, BCE_TX_CHAIN_PAGE_SZ, 0,
+ NULL, NULL, &sc->tx_bd_chain_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate TX descriptor "
+ "chain DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ for (i = 0; i < TX_PAGES; i++) {
+
+ if(bus_dmamem_alloc(sc->tx_bd_chain_tag,
+ (void **)&sc->tx_bd_chain[i],
+ BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT,
+ &sc->tx_bd_chain_map[i])) {
+ BCE_PRINTF("%s(%d): Could not allocate TX descriptor "
+ "chain DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ error = bus_dmamap_load(sc->tx_bd_chain_tag,
+ sc->tx_bd_chain_map[i], sc->tx_bd_chain[i],
+ BCE_TX_CHAIN_PAGE_SZ, bce_dma_map_addr,
+ &sc->tx_bd_chain_paddr[i], BUS_DMA_NOWAIT);
+
+ if (error) {
+ BCE_PRINTF("%s(%d): Could not map TX descriptor "
+ "chain DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): tx_bd_chain_paddr[%d] = "
+ "0x%jX\n", __FUNCTION__, i,
+ (uintmax_t) sc->tx_bd_chain_paddr[i]);
+ }
+
+ /* Check the required size before mapping to conserve resources. */
+ if (bce_tso_enable) {
+ max_size = BCE_TSO_MAX_SIZE;
+ max_segments = BCE_MAX_SEGMENTS;
+ max_seg_size = BCE_TSO_MAX_SEG_SIZE;
+ } else {
+ max_size = MCLBYTES * BCE_MAX_SEGMENTS;
+ max_segments = BCE_MAX_SEGMENTS;
+ max_seg_size = MCLBYTES;
+ }
+
+ /* Create a DMA tag for TX mbufs. */
+ if (bus_dma_tag_create(sc->parent_tag, 1, BCE_DMA_BOUNDARY,
+ sc->max_bus_addr, BUS_SPACE_MAXADDR, NULL, NULL, max_size,
+ max_segments, max_seg_size, 0, NULL, NULL, &sc->tx_mbuf_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate TX mbuf DMA tag!\n",
+ __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ /* Create DMA maps for the TX mbufs clusters. */
+ for (i = 0; i < TOTAL_TX_BD; i++) {
+ if (bus_dmamap_create(sc->tx_mbuf_tag, BUS_DMA_NOWAIT,
+ &sc->tx_mbuf_map[i])) {
+ BCE_PRINTF("%s(%d): Unable to create TX mbuf DMA "
+ "map!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+ }
+
+ /*
+ * Create a DMA tag for the RX buffer descriptor chain,
+ * allocate and clear the memory, and fetch the physical
+ * address of the blocks.
+ */
+ if (bus_dma_tag_create(sc->parent_tag, BCM_PAGE_SIZE,
+ BCE_DMA_BOUNDARY, BUS_SPACE_MAXADDR,
+ sc->max_bus_addr, NULL, NULL,
+ BCE_RX_CHAIN_PAGE_SZ, 1, BCE_RX_CHAIN_PAGE_SZ,
+ 0, NULL, NULL, &sc->rx_bd_chain_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate RX descriptor chain "
+ "DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ for (i = 0; i < RX_PAGES; i++) {
+
+ if (bus_dmamem_alloc(sc->rx_bd_chain_tag,
+ (void **)&sc->rx_bd_chain[i],
+ BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT,
+ &sc->rx_bd_chain_map[i])) {
+ BCE_PRINTF("%s(%d): Could not allocate RX descriptor "
+ "chain DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ error = bus_dmamap_load(sc->rx_bd_chain_tag,
+ sc->rx_bd_chain_map[i], sc->rx_bd_chain[i],
+ BCE_RX_CHAIN_PAGE_SZ, bce_dma_map_addr,
+ &sc->rx_bd_chain_paddr[i], BUS_DMA_NOWAIT);
+
+ if (error) {
+ BCE_PRINTF("%s(%d): Could not map RX descriptor "
+ "chain DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): rx_bd_chain_paddr[%d] = "
+ "0x%jX\n", __FUNCTION__, i,
+ (uintmax_t) sc->rx_bd_chain_paddr[i]);
+ }
+
+ /*
+ * Create a DMA tag for RX mbufs.
+ */
+#ifdef BCE_JUMBO_HDRSPLIT
+ max_size = max_seg_size = ((sc->rx_bd_mbuf_alloc_size < MCLBYTES) ?
+ MCLBYTES : sc->rx_bd_mbuf_alloc_size);
+#else
+ max_size = max_seg_size = MJUM9BYTES;
+#endif
+ max_segments = 1;
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): Creating rx_mbuf_tag "
+ "(max size = 0x%jX max segments = %d, max segment "
+ "size = 0x%jX)\n", __FUNCTION__, (uintmax_t) max_size,
+ max_segments, (uintmax_t) max_seg_size);
+
+ if (bus_dma_tag_create(sc->parent_tag, BCE_RX_BUF_ALIGN,
+ BCE_DMA_BOUNDARY, sc->max_bus_addr, BUS_SPACE_MAXADDR, NULL, NULL,
+ max_size, max_segments, max_seg_size, 0, NULL, NULL,
+ &sc->rx_mbuf_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate RX mbuf DMA tag!\n",
+ __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ /* Create DMA maps for the RX mbuf clusters. */
+ for (i = 0; i < TOTAL_RX_BD; i++) {
+ if (bus_dmamap_create(sc->rx_mbuf_tag, BUS_DMA_NOWAIT,
+ &sc->rx_mbuf_map[i])) {
+ BCE_PRINTF("%s(%d): Unable to create RX mbuf "
+ "DMA map!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+ }
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ /*
+ * Create a DMA tag for the page buffer descriptor chain,
+ * allocate and clear the memory, and fetch the physical
+ * address of the blocks.
+ */
+ if (bus_dma_tag_create(sc->parent_tag, BCM_PAGE_SIZE,
+ BCE_DMA_BOUNDARY, BUS_SPACE_MAXADDR, sc->max_bus_addr,
+ NULL, NULL, BCE_PG_CHAIN_PAGE_SZ, 1, BCE_PG_CHAIN_PAGE_SZ,
+ 0, NULL, NULL, &sc->pg_bd_chain_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate page descriptor "
+ "chain DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ for (i = 0; i < PG_PAGES; i++) {
+
+ if (bus_dmamem_alloc(sc->pg_bd_chain_tag,
+ (void **)&sc->pg_bd_chain[i],
+ BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT,
+ &sc->pg_bd_chain_map[i])) {
+ BCE_PRINTF("%s(%d): Could not allocate page "
+ "descriptor chain DMA memory!\n",
+ __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ error = bus_dmamap_load(sc->pg_bd_chain_tag,
+ sc->pg_bd_chain_map[i], sc->pg_bd_chain[i],
+ BCE_PG_CHAIN_PAGE_SZ, bce_dma_map_addr,
+ &sc->pg_bd_chain_paddr[i], BUS_DMA_NOWAIT);
+
+ if (error) {
+ BCE_PRINTF("%s(%d): Could not map page descriptor "
+ "chain DMA memory!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): pg_bd_chain_paddr[%d] = "
+ "0x%jX\n", __FUNCTION__, i,
+ (uintmax_t) sc->pg_bd_chain_paddr[i]);
+ }
+
+ /*
+ * Create a DMA tag for page mbufs.
+ */
+ max_size = max_seg_size = ((sc->pg_bd_mbuf_alloc_size < MCLBYTES) ?
+ MCLBYTES : sc->pg_bd_mbuf_alloc_size);
+
+ if (bus_dma_tag_create(sc->parent_tag, 1, BCE_DMA_BOUNDARY,
+ sc->max_bus_addr, BUS_SPACE_MAXADDR, NULL, NULL,
+ max_size, 1, max_seg_size, 0, NULL, NULL, &sc->pg_mbuf_tag)) {
+ BCE_PRINTF("%s(%d): Could not allocate page mbuf "
+ "DMA tag!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+
+ /* Create DMA maps for the page mbuf clusters. */
+ for (i = 0; i < TOTAL_PG_BD; i++) {
+ if (bus_dmamap_create(sc->pg_mbuf_tag, BUS_DMA_NOWAIT,
+ &sc->pg_mbuf_map[i])) {
+ BCE_PRINTF("%s(%d): Unable to create page mbuf "
+ "DMA map!\n", __FILE__, __LINE__);
+ rc = ENOMEM;
+ goto bce_dma_alloc_exit;
+ }
+ }
+#endif
+
+bce_dma_alloc_exit:
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_CTX);
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Release all resources used by the driver. */
+/* */
+/* Releases all resources acquired by the driver including interrupts, */
+/* interrupt handler, interfaces, mutexes, and DMA memory. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_release_resources(struct bce_softc *sc)
+{
+ device_t dev;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ dev = sc->bce_dev;
+
+ bce_dma_free(sc);
+
+ if (sc->bce_intrhand != NULL) {
+ DBPRINT(sc, BCE_INFO_RESET, "Removing interrupt handler.\n");
+ bus_teardown_intr(dev, sc->bce_res_irq, sc->bce_intrhand);
+ }
+
+ if (sc->bce_res_irq != NULL) {
+ DBPRINT(sc, BCE_INFO_RESET, "Releasing IRQ.\n");
+ bus_release_resource(dev, SYS_RES_IRQ, sc->bce_irq_rid,
+ sc->bce_res_irq);
+ }
+
+ if (sc->bce_flags & (BCE_USING_MSI_FLAG | BCE_USING_MSIX_FLAG)) {
+ DBPRINT(sc, BCE_INFO_RESET, "Releasing MSI/MSI-X vector.\n");
+ pci_release_msi(dev);
+ }
+
+ if (sc->bce_res_mem != NULL) {
+ DBPRINT(sc, BCE_INFO_RESET, "Releasing PCI memory.\n");
+ bus_release_resource(dev, SYS_RES_MEMORY, PCIR_BAR(0),
+ sc->bce_res_mem);
+ }
+
+ if (sc->bce_ifp != NULL) {
+ DBPRINT(sc, BCE_INFO_RESET, "Releasing IF.\n");
+ if_free(sc->bce_ifp);
+ }
+
+ if (mtx_initialized(&sc->bce_mtx))
+ BCE_LOCK_DESTROY(sc);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Firmware synchronization. */
+/* */
+/* Before performing certain events such as a chip reset, synchronize with */
+/* the firmware first. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_fw_sync(struct bce_softc *sc, u32 msg_data)
+{
+ int i, rc = 0;
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ /* Don't waste any time if we've timed out before. */
+ if (sc->bce_fw_timed_out == TRUE) {
+ rc = EBUSY;
+ goto bce_fw_sync_exit;
+ }
+
+ /* Increment the message sequence number. */
+ sc->bce_fw_wr_seq++;
+ msg_data |= sc->bce_fw_wr_seq;
+
+ DBPRINT(sc, BCE_VERBOSE_FIRMWARE, "bce_fw_sync(): msg_data = "
+ "0x%08X\n", msg_data);
+
+ /* Send the message to the bootcode driver mailbox. */
+ bce_shmem_wr(sc, BCE_DRV_MB, msg_data);
+
+ /* Wait for the bootcode to acknowledge the message. */
+ for (i = 0; i < FW_ACK_TIME_OUT_MS; i++) {
+ /* Check for a response in the bootcode firmware mailbox. */
+ val = bce_shmem_rd(sc, BCE_FW_MB);
+ if ((val & BCE_FW_MSG_ACK) == (msg_data & BCE_DRV_MSG_SEQ))
+ break;
+ DELAY(1000);
+ }
+
+ /* If we've timed out, tell bootcode that we've stopped waiting. */
+ if (((val & BCE_FW_MSG_ACK) != (msg_data & BCE_DRV_MSG_SEQ)) &&
+ ((msg_data & BCE_DRV_MSG_DATA) != BCE_DRV_MSG_DATA_WAIT0)) {
+
+ BCE_PRINTF("%s(%d): Firmware synchronization timeout! "
+ "msg_data = 0x%08X\n", __FILE__, __LINE__, msg_data);
+
+ msg_data &= ~BCE_DRV_MSG_CODE;
+ msg_data |= BCE_DRV_MSG_CODE_FW_TIMEOUT;
+
+ bce_shmem_wr(sc, BCE_DRV_MB, msg_data);
+
+ sc->bce_fw_timed_out = TRUE;
+ rc = EBUSY;
+ }
+
+bce_fw_sync_exit:
+ DBEXIT(BCE_VERBOSE_RESET);
+ return (rc);
+}
+
+
+/****************************************************************************/
+/* Load Receive Virtual 2 Physical (RV2P) processor firmware. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_load_rv2p_fw(struct bce_softc *sc, u32 *rv2p_code,
+ u32 rv2p_code_len, u32 rv2p_proc)
+{
+ int i;
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ /* Set the page size used by RV2P. */
+ if (rv2p_proc == RV2P_PROC2) {
+ BCE_RV2P_PROC2_CHG_MAX_BD_PAGE(USABLE_RX_BD_PER_PAGE);
+ }
+
+ for (i = 0; i < rv2p_code_len; i += 8) {
+ REG_WR(sc, BCE_RV2P_INSTR_HIGH, *rv2p_code);
+ rv2p_code++;
+ REG_WR(sc, BCE_RV2P_INSTR_LOW, *rv2p_code);
+ rv2p_code++;
+
+ if (rv2p_proc == RV2P_PROC1) {
+ val = (i / 8) | BCE_RV2P_PROC1_ADDR_CMD_RDWR;
+ REG_WR(sc, BCE_RV2P_PROC1_ADDR_CMD, val);
+ }
+ else {
+ val = (i / 8) | BCE_RV2P_PROC2_ADDR_CMD_RDWR;
+ REG_WR(sc, BCE_RV2P_PROC2_ADDR_CMD, val);
+ }
+ }
+
+ /* Reset the processor, un-stall is done later. */
+ if (rv2p_proc == RV2P_PROC1) {
+ REG_WR(sc, BCE_RV2P_COMMAND, BCE_RV2P_COMMAND_PROC1_RESET);
+ }
+ else {
+ REG_WR(sc, BCE_RV2P_COMMAND, BCE_RV2P_COMMAND_PROC2_RESET);
+ }
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Load RISC processor firmware. */
+/* */
+/* Loads firmware from the file if_bcefw.h into the scratchpad memory */
+/* associated with a particular processor. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_load_cpu_fw(struct bce_softc *sc, struct cpu_reg *cpu_reg,
+ struct fw_info *fw)
+{
+ u32 offset;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ bce_halt_cpu(sc, cpu_reg);
+
+ /* Load the Text area. */
+ offset = cpu_reg->spad_base + (fw->text_addr - cpu_reg->mips_view_base);
+ if (fw->text) {
+ int j;
+
+ for (j = 0; j < (fw->text_len / 4); j++, offset += 4) {
+ REG_WR_IND(sc, offset, fw->text[j]);
+ }
+ }
+
+ /* Load the Data area. */
+ offset = cpu_reg->spad_base + (fw->data_addr - cpu_reg->mips_view_base);
+ if (fw->data) {
+ int j;
+
+ for (j = 0; j < (fw->data_len / 4); j++, offset += 4) {
+ REG_WR_IND(sc, offset, fw->data[j]);
+ }
+ }
+
+ /* Load the SBSS area. */
+ offset = cpu_reg->spad_base + (fw->sbss_addr - cpu_reg->mips_view_base);
+ if (fw->sbss) {
+ int j;
+
+ for (j = 0; j < (fw->sbss_len / 4); j++, offset += 4) {
+ REG_WR_IND(sc, offset, fw->sbss[j]);
+ }
+ }
+
+ /* Load the BSS area. */
+ offset = cpu_reg->spad_base + (fw->bss_addr - cpu_reg->mips_view_base);
+ if (fw->bss) {
+ int j;
+
+ for (j = 0; j < (fw->bss_len/4); j++, offset += 4) {
+ REG_WR_IND(sc, offset, fw->bss[j]);
+ }
+ }
+
+ /* Load the Read-Only area. */
+ offset = cpu_reg->spad_base +
+ (fw->rodata_addr - cpu_reg->mips_view_base);
+ if (fw->rodata) {
+ int j;
+
+ for (j = 0; j < (fw->rodata_len / 4); j++, offset += 4) {
+ REG_WR_IND(sc, offset, fw->rodata[j]);
+ }
+ }
+
+ /* Clear the pre-fetch instruction and set the FW start address. */
+ REG_WR_IND(sc, cpu_reg->inst, 0);
+ REG_WR_IND(sc, cpu_reg->pc, fw->start_addr);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Starts the RISC processor. */
+/* */
+/* Assumes the CPU starting address has already been set. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_start_cpu(struct bce_softc *sc, struct cpu_reg *cpu_reg)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ /* Start the CPU. */
+ val = REG_RD_IND(sc, cpu_reg->mode);
+ val &= ~cpu_reg->mode_value_halt;
+ REG_WR_IND(sc, cpu_reg->state, cpu_reg->state_value_clear);
+ REG_WR_IND(sc, cpu_reg->mode, val);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Halts the RISC processor. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_halt_cpu(struct bce_softc *sc, struct cpu_reg *cpu_reg)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ /* Halt the CPU. */
+ val = REG_RD_IND(sc, cpu_reg->mode);
+ val |= cpu_reg->mode_value_halt;
+ REG_WR_IND(sc, cpu_reg->mode, val);
+ REG_WR_IND(sc, cpu_reg->state, cpu_reg->state_value_clear);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the RX CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_start_rxp_cpu(struct bce_softc *sc)
+{
+ struct cpu_reg cpu_reg;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ cpu_reg.mode = BCE_RXP_CPU_MODE;
+ cpu_reg.mode_value_halt = BCE_RXP_CPU_MODE_SOFT_HALT;
+ cpu_reg.mode_value_sstep = BCE_RXP_CPU_MODE_STEP_ENA;
+ cpu_reg.state = BCE_RXP_CPU_STATE;
+ cpu_reg.state_value_clear = 0xffffff;
+ cpu_reg.gpr0 = BCE_RXP_CPU_REG_FILE;
+ cpu_reg.evmask = BCE_RXP_CPU_EVENT_MASK;
+ cpu_reg.pc = BCE_RXP_CPU_PROGRAM_COUNTER;
+ cpu_reg.inst = BCE_RXP_CPU_INSTRUCTION;
+ cpu_reg.bp = BCE_RXP_CPU_HW_BREAKPOINT;
+ cpu_reg.spad_base = BCE_RXP_SCRATCH;
+ cpu_reg.mips_view_base = 0x8000000;
+
+ DBPRINT(sc, BCE_INFO_RESET, "Starting RX firmware.\n");
+ bce_start_cpu(sc, &cpu_reg);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the RX CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_rxp_cpu(struct bce_softc *sc)
+{
+ struct cpu_reg cpu_reg;
+ struct fw_info fw;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ cpu_reg.mode = BCE_RXP_CPU_MODE;
+ cpu_reg.mode_value_halt = BCE_RXP_CPU_MODE_SOFT_HALT;
+ cpu_reg.mode_value_sstep = BCE_RXP_CPU_MODE_STEP_ENA;
+ cpu_reg.state = BCE_RXP_CPU_STATE;
+ cpu_reg.state_value_clear = 0xffffff;
+ cpu_reg.gpr0 = BCE_RXP_CPU_REG_FILE;
+ cpu_reg.evmask = BCE_RXP_CPU_EVENT_MASK;
+ cpu_reg.pc = BCE_RXP_CPU_PROGRAM_COUNTER;
+ cpu_reg.inst = BCE_RXP_CPU_INSTRUCTION;
+ cpu_reg.bp = BCE_RXP_CPU_HW_BREAKPOINT;
+ cpu_reg.spad_base = BCE_RXP_SCRATCH;
+ cpu_reg.mips_view_base = 0x8000000;
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ fw.ver_major = bce_RXP_b09FwReleaseMajor;
+ fw.ver_minor = bce_RXP_b09FwReleaseMinor;
+ fw.ver_fix = bce_RXP_b09FwReleaseFix;
+ fw.start_addr = bce_RXP_b09FwStartAddr;
+
+ fw.text_addr = bce_RXP_b09FwTextAddr;
+ fw.text_len = bce_RXP_b09FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_RXP_b09FwText;
+
+ fw.data_addr = bce_RXP_b09FwDataAddr;
+ fw.data_len = bce_RXP_b09FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_RXP_b09FwData;
+
+ fw.sbss_addr = bce_RXP_b09FwSbssAddr;
+ fw.sbss_len = bce_RXP_b09FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_RXP_b09FwSbss;
+
+ fw.bss_addr = bce_RXP_b09FwBssAddr;
+ fw.bss_len = bce_RXP_b09FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_RXP_b09FwBss;
+
+ fw.rodata_addr = bce_RXP_b09FwRodataAddr;
+ fw.rodata_len = bce_RXP_b09FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_RXP_b09FwRodata;
+ } else {
+ fw.ver_major = bce_RXP_b06FwReleaseMajor;
+ fw.ver_minor = bce_RXP_b06FwReleaseMinor;
+ fw.ver_fix = bce_RXP_b06FwReleaseFix;
+ fw.start_addr = bce_RXP_b06FwStartAddr;
+
+ fw.text_addr = bce_RXP_b06FwTextAddr;
+ fw.text_len = bce_RXP_b06FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_RXP_b06FwText;
+
+ fw.data_addr = bce_RXP_b06FwDataAddr;
+ fw.data_len = bce_RXP_b06FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_RXP_b06FwData;
+
+ fw.sbss_addr = bce_RXP_b06FwSbssAddr;
+ fw.sbss_len = bce_RXP_b06FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_RXP_b06FwSbss;
+
+ fw.bss_addr = bce_RXP_b06FwBssAddr;
+ fw.bss_len = bce_RXP_b06FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_RXP_b06FwBss;
+
+ fw.rodata_addr = bce_RXP_b06FwRodataAddr;
+ fw.rodata_len = bce_RXP_b06FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_RXP_b06FwRodata;
+ }
+
+ DBPRINT(sc, BCE_INFO_RESET, "Loading RX firmware.\n");
+ bce_load_cpu_fw(sc, &cpu_reg, &fw);
+
+ /* Delay RXP start until initialization is complete. */
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the TX CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_txp_cpu(struct bce_softc *sc)
+{
+ struct cpu_reg cpu_reg;
+ struct fw_info fw;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ cpu_reg.mode = BCE_TXP_CPU_MODE;
+ cpu_reg.mode_value_halt = BCE_TXP_CPU_MODE_SOFT_HALT;
+ cpu_reg.mode_value_sstep = BCE_TXP_CPU_MODE_STEP_ENA;
+ cpu_reg.state = BCE_TXP_CPU_STATE;
+ cpu_reg.state_value_clear = 0xffffff;
+ cpu_reg.gpr0 = BCE_TXP_CPU_REG_FILE;
+ cpu_reg.evmask = BCE_TXP_CPU_EVENT_MASK;
+ cpu_reg.pc = BCE_TXP_CPU_PROGRAM_COUNTER;
+ cpu_reg.inst = BCE_TXP_CPU_INSTRUCTION;
+ cpu_reg.bp = BCE_TXP_CPU_HW_BREAKPOINT;
+ cpu_reg.spad_base = BCE_TXP_SCRATCH;
+ cpu_reg.mips_view_base = 0x8000000;
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ fw.ver_major = bce_TXP_b09FwReleaseMajor;
+ fw.ver_minor = bce_TXP_b09FwReleaseMinor;
+ fw.ver_fix = bce_TXP_b09FwReleaseFix;
+ fw.start_addr = bce_TXP_b09FwStartAddr;
+
+ fw.text_addr = bce_TXP_b09FwTextAddr;
+ fw.text_len = bce_TXP_b09FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_TXP_b09FwText;
+
+ fw.data_addr = bce_TXP_b09FwDataAddr;
+ fw.data_len = bce_TXP_b09FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_TXP_b09FwData;
+
+ fw.sbss_addr = bce_TXP_b09FwSbssAddr;
+ fw.sbss_len = bce_TXP_b09FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_TXP_b09FwSbss;
+
+ fw.bss_addr = bce_TXP_b09FwBssAddr;
+ fw.bss_len = bce_TXP_b09FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_TXP_b09FwBss;
+
+ fw.rodata_addr = bce_TXP_b09FwRodataAddr;
+ fw.rodata_len = bce_TXP_b09FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_TXP_b09FwRodata;
+ } else {
+ fw.ver_major = bce_TXP_b06FwReleaseMajor;
+ fw.ver_minor = bce_TXP_b06FwReleaseMinor;
+ fw.ver_fix = bce_TXP_b06FwReleaseFix;
+ fw.start_addr = bce_TXP_b06FwStartAddr;
+
+ fw.text_addr = bce_TXP_b06FwTextAddr;
+ fw.text_len = bce_TXP_b06FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_TXP_b06FwText;
+
+ fw.data_addr = bce_TXP_b06FwDataAddr;
+ fw.data_len = bce_TXP_b06FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_TXP_b06FwData;
+
+ fw.sbss_addr = bce_TXP_b06FwSbssAddr;
+ fw.sbss_len = bce_TXP_b06FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_TXP_b06FwSbss;
+
+ fw.bss_addr = bce_TXP_b06FwBssAddr;
+ fw.bss_len = bce_TXP_b06FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_TXP_b06FwBss;
+
+ fw.rodata_addr = bce_TXP_b06FwRodataAddr;
+ fw.rodata_len = bce_TXP_b06FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_TXP_b06FwRodata;
+ }
+
+ DBPRINT(sc, BCE_INFO_RESET, "Loading TX firmware.\n");
+ bce_load_cpu_fw(sc, &cpu_reg, &fw);
+ bce_start_cpu(sc, &cpu_reg);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the TPAT CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_tpat_cpu(struct bce_softc *sc)
+{
+ struct cpu_reg cpu_reg;
+ struct fw_info fw;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ cpu_reg.mode = BCE_TPAT_CPU_MODE;
+ cpu_reg.mode_value_halt = BCE_TPAT_CPU_MODE_SOFT_HALT;
+ cpu_reg.mode_value_sstep = BCE_TPAT_CPU_MODE_STEP_ENA;
+ cpu_reg.state = BCE_TPAT_CPU_STATE;
+ cpu_reg.state_value_clear = 0xffffff;
+ cpu_reg.gpr0 = BCE_TPAT_CPU_REG_FILE;
+ cpu_reg.evmask = BCE_TPAT_CPU_EVENT_MASK;
+ cpu_reg.pc = BCE_TPAT_CPU_PROGRAM_COUNTER;
+ cpu_reg.inst = BCE_TPAT_CPU_INSTRUCTION;
+ cpu_reg.bp = BCE_TPAT_CPU_HW_BREAKPOINT;
+ cpu_reg.spad_base = BCE_TPAT_SCRATCH;
+ cpu_reg.mips_view_base = 0x8000000;
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ fw.ver_major = bce_TPAT_b09FwReleaseMajor;
+ fw.ver_minor = bce_TPAT_b09FwReleaseMinor;
+ fw.ver_fix = bce_TPAT_b09FwReleaseFix;
+ fw.start_addr = bce_TPAT_b09FwStartAddr;
+
+ fw.text_addr = bce_TPAT_b09FwTextAddr;
+ fw.text_len = bce_TPAT_b09FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_TPAT_b09FwText;
+
+ fw.data_addr = bce_TPAT_b09FwDataAddr;
+ fw.data_len = bce_TPAT_b09FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_TPAT_b09FwData;
+
+ fw.sbss_addr = bce_TPAT_b09FwSbssAddr;
+ fw.sbss_len = bce_TPAT_b09FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_TPAT_b09FwSbss;
+
+ fw.bss_addr = bce_TPAT_b09FwBssAddr;
+ fw.bss_len = bce_TPAT_b09FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_TPAT_b09FwBss;
+
+ fw.rodata_addr = bce_TPAT_b09FwRodataAddr;
+ fw.rodata_len = bce_TPAT_b09FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_TPAT_b09FwRodata;
+ } else {
+ fw.ver_major = bce_TPAT_b06FwReleaseMajor;
+ fw.ver_minor = bce_TPAT_b06FwReleaseMinor;
+ fw.ver_fix = bce_TPAT_b06FwReleaseFix;
+ fw.start_addr = bce_TPAT_b06FwStartAddr;
+
+ fw.text_addr = bce_TPAT_b06FwTextAddr;
+ fw.text_len = bce_TPAT_b06FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_TPAT_b06FwText;
+
+ fw.data_addr = bce_TPAT_b06FwDataAddr;
+ fw.data_len = bce_TPAT_b06FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_TPAT_b06FwData;
+
+ fw.sbss_addr = bce_TPAT_b06FwSbssAddr;
+ fw.sbss_len = bce_TPAT_b06FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_TPAT_b06FwSbss;
+
+ fw.bss_addr = bce_TPAT_b06FwBssAddr;
+ fw.bss_len = bce_TPAT_b06FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_TPAT_b06FwBss;
+
+ fw.rodata_addr = bce_TPAT_b06FwRodataAddr;
+ fw.rodata_len = bce_TPAT_b06FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_TPAT_b06FwRodata;
+ }
+
+ DBPRINT(sc, BCE_INFO_RESET, "Loading TPAT firmware.\n");
+ bce_load_cpu_fw(sc, &cpu_reg, &fw);
+ bce_start_cpu(sc, &cpu_reg);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the CP CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_cp_cpu(struct bce_softc *sc)
+{
+ struct cpu_reg cpu_reg;
+ struct fw_info fw;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ cpu_reg.mode = BCE_CP_CPU_MODE;
+ cpu_reg.mode_value_halt = BCE_CP_CPU_MODE_SOFT_HALT;
+ cpu_reg.mode_value_sstep = BCE_CP_CPU_MODE_STEP_ENA;
+ cpu_reg.state = BCE_CP_CPU_STATE;
+ cpu_reg.state_value_clear = 0xffffff;
+ cpu_reg.gpr0 = BCE_CP_CPU_REG_FILE;
+ cpu_reg.evmask = BCE_CP_CPU_EVENT_MASK;
+ cpu_reg.pc = BCE_CP_CPU_PROGRAM_COUNTER;
+ cpu_reg.inst = BCE_CP_CPU_INSTRUCTION;
+ cpu_reg.bp = BCE_CP_CPU_HW_BREAKPOINT;
+ cpu_reg.spad_base = BCE_CP_SCRATCH;
+ cpu_reg.mips_view_base = 0x8000000;
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ fw.ver_major = bce_CP_b09FwReleaseMajor;
+ fw.ver_minor = bce_CP_b09FwReleaseMinor;
+ fw.ver_fix = bce_CP_b09FwReleaseFix;
+ fw.start_addr = bce_CP_b09FwStartAddr;
+
+ fw.text_addr = bce_CP_b09FwTextAddr;
+ fw.text_len = bce_CP_b09FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_CP_b09FwText;
+
+ fw.data_addr = bce_CP_b09FwDataAddr;
+ fw.data_len = bce_CP_b09FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_CP_b09FwData;
+
+ fw.sbss_addr = bce_CP_b09FwSbssAddr;
+ fw.sbss_len = bce_CP_b09FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_CP_b09FwSbss;
+
+ fw.bss_addr = bce_CP_b09FwBssAddr;
+ fw.bss_len = bce_CP_b09FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_CP_b09FwBss;
+
+ fw.rodata_addr = bce_CP_b09FwRodataAddr;
+ fw.rodata_len = bce_CP_b09FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_CP_b09FwRodata;
+ } else {
+ fw.ver_major = bce_CP_b06FwReleaseMajor;
+ fw.ver_minor = bce_CP_b06FwReleaseMinor;
+ fw.ver_fix = bce_CP_b06FwReleaseFix;
+ fw.start_addr = bce_CP_b06FwStartAddr;
+
+ fw.text_addr = bce_CP_b06FwTextAddr;
+ fw.text_len = bce_CP_b06FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_CP_b06FwText;
+
+ fw.data_addr = bce_CP_b06FwDataAddr;
+ fw.data_len = bce_CP_b06FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_CP_b06FwData;
+
+ fw.sbss_addr = bce_CP_b06FwSbssAddr;
+ fw.sbss_len = bce_CP_b06FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_CP_b06FwSbss;
+
+ fw.bss_addr = bce_CP_b06FwBssAddr;
+ fw.bss_len = bce_CP_b06FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_CP_b06FwBss;
+
+ fw.rodata_addr = bce_CP_b06FwRodataAddr;
+ fw.rodata_len = bce_CP_b06FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_CP_b06FwRodata;
+ }
+
+ DBPRINT(sc, BCE_INFO_RESET, "Loading CP firmware.\n");
+ bce_load_cpu_fw(sc, &cpu_reg, &fw);
+ bce_start_cpu(sc, &cpu_reg);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the COM CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_com_cpu(struct bce_softc *sc)
+{
+ struct cpu_reg cpu_reg;
+ struct fw_info fw;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ cpu_reg.mode = BCE_COM_CPU_MODE;
+ cpu_reg.mode_value_halt = BCE_COM_CPU_MODE_SOFT_HALT;
+ cpu_reg.mode_value_sstep = BCE_COM_CPU_MODE_STEP_ENA;
+ cpu_reg.state = BCE_COM_CPU_STATE;
+ cpu_reg.state_value_clear = 0xffffff;
+ cpu_reg.gpr0 = BCE_COM_CPU_REG_FILE;
+ cpu_reg.evmask = BCE_COM_CPU_EVENT_MASK;
+ cpu_reg.pc = BCE_COM_CPU_PROGRAM_COUNTER;
+ cpu_reg.inst = BCE_COM_CPU_INSTRUCTION;
+ cpu_reg.bp = BCE_COM_CPU_HW_BREAKPOINT;
+ cpu_reg.spad_base = BCE_COM_SCRATCH;
+ cpu_reg.mips_view_base = 0x8000000;
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ fw.ver_major = bce_COM_b09FwReleaseMajor;
+ fw.ver_minor = bce_COM_b09FwReleaseMinor;
+ fw.ver_fix = bce_COM_b09FwReleaseFix;
+ fw.start_addr = bce_COM_b09FwStartAddr;
+
+ fw.text_addr = bce_COM_b09FwTextAddr;
+ fw.text_len = bce_COM_b09FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_COM_b09FwText;
+
+ fw.data_addr = bce_COM_b09FwDataAddr;
+ fw.data_len = bce_COM_b09FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_COM_b09FwData;
+
+ fw.sbss_addr = bce_COM_b09FwSbssAddr;
+ fw.sbss_len = bce_COM_b09FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_COM_b09FwSbss;
+
+ fw.bss_addr = bce_COM_b09FwBssAddr;
+ fw.bss_len = bce_COM_b09FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_COM_b09FwBss;
+
+ fw.rodata_addr = bce_COM_b09FwRodataAddr;
+ fw.rodata_len = bce_COM_b09FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_COM_b09FwRodata;
+ } else {
+ fw.ver_major = bce_COM_b06FwReleaseMajor;
+ fw.ver_minor = bce_COM_b06FwReleaseMinor;
+ fw.ver_fix = bce_COM_b06FwReleaseFix;
+ fw.start_addr = bce_COM_b06FwStartAddr;
+
+ fw.text_addr = bce_COM_b06FwTextAddr;
+ fw.text_len = bce_COM_b06FwTextLen;
+ fw.text_index = 0;
+ fw.text = bce_COM_b06FwText;
+
+ fw.data_addr = bce_COM_b06FwDataAddr;
+ fw.data_len = bce_COM_b06FwDataLen;
+ fw.data_index = 0;
+ fw.data = bce_COM_b06FwData;
+
+ fw.sbss_addr = bce_COM_b06FwSbssAddr;
+ fw.sbss_len = bce_COM_b06FwSbssLen;
+ fw.sbss_index = 0;
+ fw.sbss = bce_COM_b06FwSbss;
+
+ fw.bss_addr = bce_COM_b06FwBssAddr;
+ fw.bss_len = bce_COM_b06FwBssLen;
+ fw.bss_index = 0;
+ fw.bss = bce_COM_b06FwBss;
+
+ fw.rodata_addr = bce_COM_b06FwRodataAddr;
+ fw.rodata_len = bce_COM_b06FwRodataLen;
+ fw.rodata_index = 0;
+ fw.rodata = bce_COM_b06FwRodata;
+ }
+
+ DBPRINT(sc, BCE_INFO_RESET, "Loading COM firmware.\n");
+ bce_load_cpu_fw(sc, &cpu_reg, &fw);
+ bce_start_cpu(sc, &cpu_reg);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the RV2P, RX, TX, TPAT, COM, and CP CPUs. */
+/* */
+/* Loads the firmware for each CPU and starts the CPU. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_cpus(struct bce_softc *sc)
+{
+ DBENTER(BCE_VERBOSE_RESET);
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+
+ if ((BCE_CHIP_REV(sc) == BCE_CHIP_REV_Ax)) {
+ bce_load_rv2p_fw(sc, bce_xi90_rv2p_proc1,
+ sizeof(bce_xi90_rv2p_proc1), RV2P_PROC1);
+ bce_load_rv2p_fw(sc, bce_xi90_rv2p_proc2,
+ sizeof(bce_xi90_rv2p_proc2), RV2P_PROC2);
+ } else {
+ bce_load_rv2p_fw(sc, bce_xi_rv2p_proc1,
+ sizeof(bce_xi_rv2p_proc1), RV2P_PROC1);
+ bce_load_rv2p_fw(sc, bce_xi_rv2p_proc2,
+ sizeof(bce_xi_rv2p_proc2), RV2P_PROC2);
+ }
+
+ } else {
+ bce_load_rv2p_fw(sc, bce_rv2p_proc1,
+ sizeof(bce_rv2p_proc1), RV2P_PROC1);
+ bce_load_rv2p_fw(sc, bce_rv2p_proc2,
+ sizeof(bce_rv2p_proc2), RV2P_PROC2);
+ }
+
+ bce_init_rxp_cpu(sc);
+ bce_init_txp_cpu(sc);
+ bce_init_tpat_cpu(sc);
+ bce_init_com_cpu(sc);
+ bce_init_cp_cpu(sc);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize context memory. */
+/* */
+/* Clears the memory associated with each Context ID (CID). */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static int
+bce_init_ctx(struct bce_softc *sc)
+{
+ u32 offset, val, vcid_addr;
+ int i, j, rc, retry_cnt;
+
+ rc = 0;
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_CTX);
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ retry_cnt = CTX_INIT_RETRY_COUNT;
+
+ DBPRINT(sc, BCE_INFO_CTX, "Initializing 5709 context.\n");
+
+ /*
+ * BCM5709 context memory may be cached
+ * in host memory so prepare the host memory
+ * for access.
+ */
+ val = BCE_CTX_COMMAND_ENABLED |
+ BCE_CTX_COMMAND_MEM_INIT | (1 << 12);
+ val |= (BCM_PAGE_BITS - 8) << 16;
+ REG_WR(sc, BCE_CTX_COMMAND, val);
+
+ /* Wait for mem init command to complete. */
+ for (i = 0; i < retry_cnt; i++) {
+ val = REG_RD(sc, BCE_CTX_COMMAND);
+ if (!(val & BCE_CTX_COMMAND_MEM_INIT))
+ break;
+ DELAY(2);
+ }
+ if ((val & BCE_CTX_COMMAND_MEM_INIT) != 0) {
+ BCE_PRINTF("%s(): Context memory initialization failed!\n",
+ __FUNCTION__);
+ rc = EBUSY;
+ goto init_ctx_fail;
+ }
+
+ for (i = 0; i < sc->ctx_pages; i++) {
+ /* Set the physical address of the context memory. */
+ REG_WR(sc, BCE_CTX_HOST_PAGE_TBL_DATA0,
+ BCE_ADDR_LO(sc->ctx_paddr[i] & 0xfffffff0) |
+ BCE_CTX_HOST_PAGE_TBL_DATA0_VALID);
+ REG_WR(sc, BCE_CTX_HOST_PAGE_TBL_DATA1,
+ BCE_ADDR_HI(sc->ctx_paddr[i]));
+ REG_WR(sc, BCE_CTX_HOST_PAGE_TBL_CTRL, i |
+ BCE_CTX_HOST_PAGE_TBL_CTRL_WRITE_REQ);
+
+ /* Verify the context memory write was successful. */
+ for (j = 0; j < retry_cnt; j++) {
+ val = REG_RD(sc, BCE_CTX_HOST_PAGE_TBL_CTRL);
+ if ((val &
+ BCE_CTX_HOST_PAGE_TBL_CTRL_WRITE_REQ) == 0)
+ break;
+ DELAY(5);
+ }
+ if ((val & BCE_CTX_HOST_PAGE_TBL_CTRL_WRITE_REQ) != 0) {
+ BCE_PRINTF("%s(): Failed to initialize "
+ "context page %d!\n", __FUNCTION__, i);
+ rc = EBUSY;
+ goto init_ctx_fail;
+ }
+ }
+ } else {
+
+ DBPRINT(sc, BCE_INFO, "Initializing 5706/5708 context.\n");
+
+ /*
+ * For the 5706/5708, context memory is local to
+ * the controller, so initialize the controller
+ * context memory.
+ */
+
+ vcid_addr = GET_CID_ADDR(96);
+ while (vcid_addr) {
+
+ vcid_addr -= PHY_CTX_SIZE;
+
+ REG_WR(sc, BCE_CTX_VIRT_ADDR, 0);
+ REG_WR(sc, BCE_CTX_PAGE_TBL, vcid_addr);
+
+ for(offset = 0; offset < PHY_CTX_SIZE; offset += 4) {
+ CTX_WR(sc, 0x00, offset, 0);
+ }
+
+ REG_WR(sc, BCE_CTX_VIRT_ADDR, vcid_addr);
+ REG_WR(sc, BCE_CTX_PAGE_TBL, vcid_addr);
+ }
+
+ }
+init_ctx_fail:
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_CTX);
+ return (rc);
+}
+
+
+/****************************************************************************/
+/* Fetch the permanent MAC address of the controller. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_get_mac_addr(struct bce_softc *sc)
+{
+ u32 mac_lo = 0, mac_hi = 0;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ /*
+ * The NetXtreme II bootcode populates various NIC
+ * power-on and runtime configuration items in a
+ * shared memory area. The factory configured MAC
+ * address is available from both NVRAM and the
+ * shared memory area so we'll read the value from
+ * shared memory for speed.
+ */
+
+ mac_hi = bce_shmem_rd(sc, BCE_PORT_HW_CFG_MAC_UPPER);
+ mac_lo = bce_shmem_rd(sc, BCE_PORT_HW_CFG_MAC_LOWER);
+
+ if ((mac_lo == 0) && (mac_hi == 0)) {
+ BCE_PRINTF("%s(%d): Invalid Ethernet address!\n",
+ __FILE__, __LINE__);
+ } else {
+ sc->eaddr[0] = (u_char)(mac_hi >> 8);
+ sc->eaddr[1] = (u_char)(mac_hi >> 0);
+ sc->eaddr[2] = (u_char)(mac_lo >> 24);
+ sc->eaddr[3] = (u_char)(mac_lo >> 16);
+ sc->eaddr[4] = (u_char)(mac_lo >> 8);
+ sc->eaddr[5] = (u_char)(mac_lo >> 0);
+ }
+
+ DBPRINT(sc, BCE_INFO_MISC, "Permanent Ethernet "
+ "address = %6D\n", sc->eaddr, ":");
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Program the MAC address. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_set_mac_addr(struct bce_softc *sc)
+{
+ u32 val;
+ u8 *mac_addr = sc->eaddr;
+
+ /* ToDo: Add support for setting multiple MAC addresses. */
+
+ DBENTER(BCE_VERBOSE_RESET);
+ DBPRINT(sc, BCE_INFO_MISC, "Setting Ethernet address = "
+ "%6D\n", sc->eaddr, ":");
+
+ val = (mac_addr[0] << 8) | mac_addr[1];
+
+ REG_WR(sc, BCE_EMAC_MAC_MATCH0, val);
+
+ val = (mac_addr[2] << 24) | (mac_addr[3] << 16) |
+ (mac_addr[4] << 8) | mac_addr[5];
+
+ REG_WR(sc, BCE_EMAC_MAC_MATCH1, val);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Stop the controller. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_stop(struct bce_softc *sc)
+{
+ struct ifnet *ifp;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ BCE_LOCK_ASSERT(sc);
+
+ ifp = sc->bce_ifp;
+
+ callout_stop(&sc->bce_tick_callout);
+
+ /* Disable the transmit/receive blocks. */
+ REG_WR(sc, BCE_MISC_ENABLE_CLR_BITS, BCE_MISC_ENABLE_CLR_DEFAULT);
+ REG_RD(sc, BCE_MISC_ENABLE_CLR_BITS);
+ DELAY(20);
+
+ bce_disable_intr(sc);
+
+ /* Free RX buffers. */
+#ifdef BCE_JUMBO_HDRSPLIT
+ bce_free_pg_chain(sc);
+#endif
+ bce_free_rx_chain(sc);
+
+ /* Free TX buffers. */
+ bce_free_tx_chain(sc);
+
+ sc->watchdog_timer = 0;
+
+ sc->bce_link_up = FALSE;
+
+ ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+static int
+bce_reset(struct bce_softc *sc, u32 reset_code)
+{
+ u32 val;
+ int i, rc = 0;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ DBPRINT(sc, BCE_VERBOSE_RESET, "%s(): reset_code = 0x%08X\n",
+ __FUNCTION__, reset_code);
+
+ /* Wait for pending PCI transactions to complete. */
+ REG_WR(sc, BCE_MISC_ENABLE_CLR_BITS,
+ BCE_MISC_ENABLE_CLR_BITS_TX_DMA_ENABLE |
+ BCE_MISC_ENABLE_CLR_BITS_DMA_ENGINE_ENABLE |
+ BCE_MISC_ENABLE_CLR_BITS_RX_DMA_ENABLE |
+ BCE_MISC_ENABLE_CLR_BITS_HOST_COALESCE_ENABLE);
+ val = REG_RD(sc, BCE_MISC_ENABLE_CLR_BITS);
+ DELAY(5);
+
+ /* Disable DMA */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ val = REG_RD(sc, BCE_MISC_NEW_CORE_CTL);
+ val &= ~BCE_MISC_NEW_CORE_CTL_DMA_ENABLE;
+ REG_WR(sc, BCE_MISC_NEW_CORE_CTL, val);
+ }
+
+ /* Assume bootcode is running. */
+ sc->bce_fw_timed_out = FALSE;
+ sc->bce_drv_cardiac_arrest = FALSE;
+
+ /* Give the firmware a chance to prepare for the reset. */
+ rc = bce_fw_sync(sc, BCE_DRV_MSG_DATA_WAIT0 | reset_code);
+ if (rc)
+ goto bce_reset_exit;
+
+ /* Set a firmware reminder that this is a soft reset. */
+ bce_shmem_wr(sc, BCE_DRV_RESET_SIGNATURE, BCE_DRV_RESET_SIGNATURE_MAGIC);
+
+ /* Dummy read to force the chip to complete all current transactions. */
+ val = REG_RD(sc, BCE_MISC_ID);
+
+ /* Chip reset. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ REG_WR(sc, BCE_MISC_COMMAND, BCE_MISC_COMMAND_SW_RESET);
+ REG_RD(sc, BCE_MISC_COMMAND);
+ DELAY(5);
+
+ val = BCE_PCICFG_MISC_CONFIG_REG_WINDOW_ENA |
+ BCE_PCICFG_MISC_CONFIG_TARGET_MB_WORD_SWAP;
+
+ pci_write_config(sc->bce_dev, BCE_PCICFG_MISC_CONFIG, val, 4);
+ } else {
+ val = BCE_PCICFG_MISC_CONFIG_CORE_RST_REQ |
+ BCE_PCICFG_MISC_CONFIG_REG_WINDOW_ENA |
+ BCE_PCICFG_MISC_CONFIG_TARGET_MB_WORD_SWAP;
+ REG_WR(sc, BCE_PCICFG_MISC_CONFIG, val);
+
+ /* Allow up to 30us for reset to complete. */
+ for (i = 0; i < 10; i++) {
+ val = REG_RD(sc, BCE_PCICFG_MISC_CONFIG);
+ if ((val & (BCE_PCICFG_MISC_CONFIG_CORE_RST_REQ |
+ BCE_PCICFG_MISC_CONFIG_CORE_RST_BSY)) == 0) {
+ break;
+ }
+ DELAY(10);
+ }
+
+ /* Check that reset completed successfully. */
+ if (val & (BCE_PCICFG_MISC_CONFIG_CORE_RST_REQ |
+ BCE_PCICFG_MISC_CONFIG_CORE_RST_BSY)) {
+ BCE_PRINTF("%s(%d): Reset failed!\n",
+ __FILE__, __LINE__);
+ rc = EBUSY;
+ goto bce_reset_exit;
+ }
+ }
+
+ /* Make sure byte swapping is properly configured. */
+ val = REG_RD(sc, BCE_PCI_SWAP_DIAG0);
+ if (val != 0x01020304) {
+ BCE_PRINTF("%s(%d): Byte swap is incorrect!\n",
+ __FILE__, __LINE__);
+ rc = ENODEV;
+ goto bce_reset_exit;
+ }
+
+ /* Just completed a reset, assume that firmware is running again. */
+ sc->bce_fw_timed_out = FALSE;
+ sc->bce_drv_cardiac_arrest = FALSE;
+
+ /* Wait for the firmware to finish its initialization. */
+ rc = bce_fw_sync(sc, BCE_DRV_MSG_DATA_WAIT1 | reset_code);
+ if (rc)
+ BCE_PRINTF("%s(%d): Firmware did not complete "
+ "initialization!\n", __FILE__, __LINE__);
+
+bce_reset_exit:
+ DBEXIT(BCE_VERBOSE_RESET);
+ return (rc);
+}
+
+
+static int
+bce_chipinit(struct bce_softc *sc)
+{
+ u32 val;
+ int rc = 0;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ bce_disable_intr(sc);
+
+ /*
+ * Initialize DMA byte/word swapping, configure the number of DMA
+ * channels and PCI clock compensation delay.
+ */
+ val = BCE_DMA_CONFIG_DATA_BYTE_SWAP |
+ BCE_DMA_CONFIG_DATA_WORD_SWAP |
+#if BYTE_ORDER == BIG_ENDIAN
+ BCE_DMA_CONFIG_CNTL_BYTE_SWAP |
+#endif
+ BCE_DMA_CONFIG_CNTL_WORD_SWAP |
+ DMA_READ_CHANS << 12 |
+ DMA_WRITE_CHANS << 16;
+
+ val |= (0x2 << 20) | BCE_DMA_CONFIG_CNTL_PCI_COMP_DLY;
+
+ if ((sc->bce_flags & BCE_PCIX_FLAG) && (sc->bus_speed_mhz == 133))
+ val |= BCE_DMA_CONFIG_PCI_FAST_CLK_CMP;
+
+ /*
+ * This setting resolves a problem observed on certain Intel PCI
+ * chipsets that cannot handle multiple outstanding DMA operations.
+ * See errata E9_5706A1_65.
+ */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5706) &&
+ (BCE_CHIP_ID(sc) != BCE_CHIP_ID_5706_A0) &&
+ !(sc->bce_flags & BCE_PCIX_FLAG))
+ val |= BCE_DMA_CONFIG_CNTL_PING_PONG_DMA;
+
+ REG_WR(sc, BCE_DMA_CONFIG, val);
+
+ /* Enable the RX_V2P and Context state machines before access. */
+ REG_WR(sc, BCE_MISC_ENABLE_SET_BITS,
+ BCE_MISC_ENABLE_SET_BITS_HOST_COALESCE_ENABLE |
+ BCE_MISC_ENABLE_STATUS_BITS_RX_V2P_ENABLE |
+ BCE_MISC_ENABLE_STATUS_BITS_CONTEXT_ENABLE);
+
+ /* Initialize context mapping and zero out the quick contexts. */
+ if ((rc = bce_init_ctx(sc)) != 0)
+ goto bce_chipinit_exit;
+
+ /* Initialize the on-boards CPUs */
+ bce_init_cpus(sc);
+
+ /* Enable management frames (NC-SI) to flow to the MCP. */
+ if (sc->bce_flags & BCE_MFW_ENABLE_FLAG) {
+ val = REG_RD(sc, BCE_RPM_MGMT_PKT_CTRL) | BCE_RPM_MGMT_PKT_CTRL_MGMT_EN;
+ REG_WR(sc, BCE_RPM_MGMT_PKT_CTRL, val);
+ }
+
+ /* Prepare NVRAM for access. */
+ if ((rc = bce_init_nvram(sc)) != 0)
+ goto bce_chipinit_exit;
+
+ /* Set the kernel bypass block size */
+ val = REG_RD(sc, BCE_MQ_CONFIG);
+ val &= ~BCE_MQ_CONFIG_KNL_BYP_BLK_SIZE;
+ val |= BCE_MQ_CONFIG_KNL_BYP_BLK_SIZE_256;
+
+ /* Enable bins used on the 5709. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ val |= BCE_MQ_CONFIG_BIN_MQ_MODE;
+ if (BCE_CHIP_ID(sc) == BCE_CHIP_ID_5709_A1)
+ val |= BCE_MQ_CONFIG_HALT_DIS;
+ }
+
+ REG_WR(sc, BCE_MQ_CONFIG, val);
+
+ val = 0x10000 + (MAX_CID_CNT * MB_KERNEL_CTX_SIZE);
+ REG_WR(sc, BCE_MQ_KNL_BYP_WIND_START, val);
+ REG_WR(sc, BCE_MQ_KNL_WIND_END, val);
+
+ /* Set the page size and clear the RV2P processor stall bits. */
+ val = (BCM_PAGE_BITS - 8) << 24;
+ REG_WR(sc, BCE_RV2P_CONFIG, val);
+
+ /* Configure page size. */
+ val = REG_RD(sc, BCE_TBDR_CONFIG);
+ val &= ~BCE_TBDR_CONFIG_PAGE_SIZE;
+ val |= (BCM_PAGE_BITS - 8) << 24 | 0x40;
+ REG_WR(sc, BCE_TBDR_CONFIG, val);
+
+ /* Set the perfect match control register to default. */
+ REG_WR_IND(sc, BCE_RXP_PM_CTRL, 0);
+
+bce_chipinit_exit:
+ DBEXIT(BCE_VERBOSE_RESET);
+
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Initialize the controller in preparation to send/receive traffic. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_blockinit(struct bce_softc *sc)
+{
+ u32 reg, val;
+ int rc = 0;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ /* Load the hardware default MAC address. */
+ bce_set_mac_addr(sc);
+
+ /* Set the Ethernet backoff seed value */
+ val = sc->eaddr[0] + (sc->eaddr[1] << 8) +
+ (sc->eaddr[2] << 16) + (sc->eaddr[3] ) +
+ (sc->eaddr[4] << 8) + (sc->eaddr[5] << 16);
+ REG_WR(sc, BCE_EMAC_BACKOFF_SEED, val);
+
+ sc->last_status_idx = 0;
+ sc->rx_mode = BCE_EMAC_RX_MODE_SORT_MODE;
+
+ /* Set up link change interrupt generation. */
+ REG_WR(sc, BCE_EMAC_ATTENTION_ENA, BCE_EMAC_ATTENTION_ENA_LINK);
+
+ /* Program the physical address of the status block. */
+ REG_WR(sc, BCE_HC_STATUS_ADDR_L,
+ BCE_ADDR_LO(sc->status_block_paddr));
+ REG_WR(sc, BCE_HC_STATUS_ADDR_H,
+ BCE_ADDR_HI(sc->status_block_paddr));
+
+ /* Program the physical address of the statistics block. */
+ REG_WR(sc, BCE_HC_STATISTICS_ADDR_L,
+ BCE_ADDR_LO(sc->stats_block_paddr));
+ REG_WR(sc, BCE_HC_STATISTICS_ADDR_H,
+ BCE_ADDR_HI(sc->stats_block_paddr));
+
+ /* Program various host coalescing parameters. */
+ REG_WR(sc, BCE_HC_TX_QUICK_CONS_TRIP,
+ (sc->bce_tx_quick_cons_trip_int << 16) | sc->bce_tx_quick_cons_trip);
+ REG_WR(sc, BCE_HC_RX_QUICK_CONS_TRIP,
+ (sc->bce_rx_quick_cons_trip_int << 16) | sc->bce_rx_quick_cons_trip);
+ REG_WR(sc, BCE_HC_COMP_PROD_TRIP,
+ (sc->bce_comp_prod_trip_int << 16) | sc->bce_comp_prod_trip);
+ REG_WR(sc, BCE_HC_TX_TICKS,
+ (sc->bce_tx_ticks_int << 16) | sc->bce_tx_ticks);
+ REG_WR(sc, BCE_HC_RX_TICKS,
+ (sc->bce_rx_ticks_int << 16) | sc->bce_rx_ticks);
+ REG_WR(sc, BCE_HC_COM_TICKS,
+ (sc->bce_com_ticks_int << 16) | sc->bce_com_ticks);
+ REG_WR(sc, BCE_HC_CMD_TICKS,
+ (sc->bce_cmd_ticks_int << 16) | sc->bce_cmd_ticks);
+ REG_WR(sc, BCE_HC_STATS_TICKS,
+ (sc->bce_stats_ticks & 0xffff00));
+ REG_WR(sc, BCE_HC_STAT_COLLECT_TICKS, 0xbb8); /* 3ms */
+
+ /* Configure the Host Coalescing block. */
+ val = BCE_HC_CONFIG_RX_TMR_MODE | BCE_HC_CONFIG_TX_TMR_MODE |
+ BCE_HC_CONFIG_COLLECT_STATS;
+
+#if 0
+ /* ToDo: Add MSI-X support. */
+ if (sc->bce_flags & BCE_USING_MSIX_FLAG) {
+ u32 base = ((BCE_TX_VEC - 1) * BCE_HC_SB_CONFIG_SIZE) +
+ BCE_HC_SB_CONFIG_1;
+
+ REG_WR(sc, BCE_HC_MSIX_BIT_VECTOR, BCE_HC_MSIX_BIT_VECTOR_VAL);
+
+ REG_WR(sc, base, BCE_HC_SB_CONFIG_1_TX_TMR_MODE |
+ BCE_HC_SB_CONFIG_1_ONE_SHOT);
+
+ REG_WR(sc, base + BCE_HC_TX_QUICK_CONS_TRIP_OFF,
+ (sc->tx_quick_cons_trip_int << 16) |
+ sc->tx_quick_cons_trip);
+
+ REG_WR(sc, base + BCE_HC_TX_TICKS_OFF,
+ (sc->tx_ticks_int << 16) | sc->tx_ticks);
+
+ val |= BCE_HC_CONFIG_SB_ADDR_INC_128B;
+ }
+
+ /*
+ * Tell the HC block to automatically set the
+ * INT_MASK bit after an MSI/MSI-X interrupt
+ * is generated so the driver doesn't have to.
+ */
+ if (sc->bce_flags & BCE_ONE_SHOT_MSI_FLAG)
+ val |= BCE_HC_CONFIG_ONE_SHOT;
+
+ /* Set the MSI-X status blocks to 128 byte boundaries. */
+ if (sc->bce_flags & BCE_USING_MSIX_FLAG)
+ val |= BCE_HC_CONFIG_SB_ADDR_INC_128B;
+#endif
+
+ REG_WR(sc, BCE_HC_CONFIG, val);
+
+ /* Clear the internal statistics counters. */
+ REG_WR(sc, BCE_HC_COMMAND, BCE_HC_COMMAND_CLR_STAT_NOW);
+
+ /* Verify that bootcode is running. */
+ reg = bce_shmem_rd(sc, BCE_DEV_INFO_SIGNATURE);
+
+ DBRUNIF(DB_RANDOMTRUE(bootcode_running_failure_sim_control),
+ BCE_PRINTF("%s(%d): Simulating bootcode failure.\n",
+ __FILE__, __LINE__);
+ reg = 0);
+
+ if ((reg & BCE_DEV_INFO_SIGNATURE_MAGIC_MASK) !=
+ BCE_DEV_INFO_SIGNATURE_MAGIC) {
+ BCE_PRINTF("%s(%d): Bootcode not running! Found: 0x%08X, "
+ "Expected: 08%08X\n", __FILE__, __LINE__,
+ (reg & BCE_DEV_INFO_SIGNATURE_MAGIC_MASK),
+ BCE_DEV_INFO_SIGNATURE_MAGIC);
+ rc = ENODEV;
+ goto bce_blockinit_exit;
+ }
+
+ /* Enable DMA */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ val = REG_RD(sc, BCE_MISC_NEW_CORE_CTL);
+ val |= BCE_MISC_NEW_CORE_CTL_DMA_ENABLE;
+ REG_WR(sc, BCE_MISC_NEW_CORE_CTL, val);
+ }
+
+ /* Allow bootcode to apply additional fixes before enabling MAC. */
+ rc = bce_fw_sync(sc, BCE_DRV_MSG_DATA_WAIT2 |
+ BCE_DRV_MSG_CODE_RESET);
+
+ /* Enable link state change interrupt generation. */
+ REG_WR(sc, BCE_HC_ATTN_BITS_ENABLE, STATUS_ATTN_BITS_LINK_STATE);
+
+ /* Enable the RXP. */
+ bce_start_rxp_cpu(sc);
+
+ /* Disable management frames (NC-SI) from flowing to the MCP. */
+ if (sc->bce_flags & BCE_MFW_ENABLE_FLAG) {
+ val = REG_RD(sc, BCE_RPM_MGMT_PKT_CTRL) &
+ ~BCE_RPM_MGMT_PKT_CTRL_MGMT_EN;
+ REG_WR(sc, BCE_RPM_MGMT_PKT_CTRL, val);
+ }
+
+ /* Enable all remaining blocks in the MAC. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716))
+ REG_WR(sc, BCE_MISC_ENABLE_SET_BITS,
+ BCE_MISC_ENABLE_DEFAULT_XI);
+ else
+ REG_WR(sc, BCE_MISC_ENABLE_SET_BITS,
+ BCE_MISC_ENABLE_DEFAULT);
+
+ REG_RD(sc, BCE_MISC_ENABLE_SET_BITS);
+ DELAY(20);
+
+ /* Save the current host coalescing block settings. */
+ sc->hc_command = REG_RD(sc, BCE_HC_COMMAND);
+
+bce_blockinit_exit:
+ DBEXIT(BCE_VERBOSE_RESET);
+
+ return (rc);
+}
+
+
+/****************************************************************************/
+/* Encapsulate an mbuf into the rx_bd chain. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_get_rx_buf(struct bce_softc *sc, struct mbuf *m, u16 *prod,
+ u16 *chain_prod, u32 *prod_bseq)
+{
+ bus_dmamap_t map;
+ bus_dma_segment_t segs[BCE_MAX_SEGMENTS];
+ struct mbuf *m_new = NULL;
+ struct rx_bd *rxbd;
+ int nsegs, error, rc = 0;
+#ifdef BCE_DEBUG
+ u16 debug_chain_prod = *chain_prod;
+#endif
+
+ DBENTER(BCE_EXTREME_RESET | BCE_EXTREME_RECV | BCE_EXTREME_LOAD);
+
+ /* Make sure the inputs are valid. */
+ DBRUNIF((*chain_prod > MAX_RX_BD),
+ BCE_PRINTF("%s(%d): RX producer out of range: "
+ "0x%04X > 0x%04X\n", __FILE__, __LINE__,
+ *chain_prod, (u16) MAX_RX_BD));
+
+ DBPRINT(sc, BCE_EXTREME_RECV, "%s(enter): prod = 0x%04X, "
+ "chain_prod = 0x%04X, prod_bseq = 0x%08X\n", __FUNCTION__,
+ *prod, *chain_prod, *prod_bseq);
+
+ /* Update some debug statistic counters */
+ DBRUNIF((sc->free_rx_bd < sc->rx_low_watermark),
+ sc->rx_low_watermark = sc->free_rx_bd);
+ DBRUNIF((sc->free_rx_bd == sc->max_rx_bd),
+ sc->rx_empty_count++);
+
+ /* Check whether this is a new mbuf allocation. */
+ if (m == NULL) {
+
+ /* Simulate an mbuf allocation failure. */
+ DBRUNIF(DB_RANDOMTRUE(mbuf_alloc_failed_sim_control),
+ sc->mbuf_alloc_failed_count++;
+ sc->mbuf_alloc_failed_sim_count++;
+ rc = ENOBUFS;
+ goto bce_get_rx_buf_exit);
+
+ /* This is a new mbuf allocation. */
+#ifdef BCE_JUMBO_HDRSPLIT
+ MGETHDR(m_new, M_DONTWAIT, MT_DATA);
+#else
+ if (sc->rx_bd_mbuf_alloc_size <= MCLBYTES)
+ m_new = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
+ else
+ m_new = m_getjcl(M_DONTWAIT, MT_DATA, M_PKTHDR,
+ sc->rx_bd_mbuf_alloc_size);
+#endif
+
+ if (m_new == NULL) {
+ sc->mbuf_alloc_failed_count++;
+ rc = ENOBUFS;
+ goto bce_get_rx_buf_exit;
+ }
+
+ DBRUN(sc->debug_rx_mbuf_alloc++);
+ } else {
+ /* Reuse an existing mbuf. */
+ m_new = m;
+ }
+
+ /* Make sure we have a valid packet header. */
+ M_ASSERTPKTHDR(m_new);
+
+ /* Initialize the mbuf size and pad if necessary for alignment. */
+ m_new->m_pkthdr.len = m_new->m_len = sc->rx_bd_mbuf_alloc_size;
+ m_adj(m_new, sc->rx_bd_mbuf_align_pad);
+
+ /* ToDo: Consider calling m_fragment() to test error handling. */
+
+ /* Map the mbuf cluster into device memory. */
+ map = sc->rx_mbuf_map[*chain_prod];
+ error = bus_dmamap_load_mbuf_sg(sc->rx_mbuf_tag, map, m_new,
+ segs, &nsegs, BUS_DMA_NOWAIT);
+
+ /* Handle any mapping errors. */
+ if (error) {
+ BCE_PRINTF("%s(%d): Error mapping mbuf into RX "
+ "chain (%d)!\n", __FILE__, __LINE__, error);
+
+ sc->dma_map_addr_rx_failed_count++;
+ m_freem(m_new);
+
+ DBRUN(sc->debug_rx_mbuf_alloc--);
+
+ rc = ENOBUFS;
+ goto bce_get_rx_buf_exit;
+ }
+
+ /* All mbufs must map to a single segment. */
+ KASSERT(nsegs == 1, ("%s(): Too many segments returned (%d)!",
+ __FUNCTION__, nsegs));
+
+ /* Setup the rx_bd for the segment. */
+ rxbd = &sc->rx_bd_chain[RX_PAGE(*chain_prod)][RX_IDX(*chain_prod)];
+
+ rxbd->rx_bd_haddr_lo = htole32(BCE_ADDR_LO(segs[0].ds_addr));
+ rxbd->rx_bd_haddr_hi = htole32(BCE_ADDR_HI(segs[0].ds_addr));
+ rxbd->rx_bd_len = htole32(segs[0].ds_len);
+ rxbd->rx_bd_flags = htole32(RX_BD_FLAGS_START | RX_BD_FLAGS_END);
+ *prod_bseq += segs[0].ds_len;
+
+ /* Save the mbuf and update our counter. */
+ sc->rx_mbuf_ptr[*chain_prod] = m_new;
+ sc->free_rx_bd -= nsegs;
+
+ DBRUNMSG(BCE_INSANE_RECV,
+ bce_dump_rx_mbuf_chain(sc, debug_chain_prod, nsegs));
+
+ DBPRINT(sc, BCE_EXTREME_RECV, "%s(exit): prod = 0x%04X, "
+ "chain_prod = 0x%04X, prod_bseq = 0x%08X\n",
+ __FUNCTION__, *prod, *chain_prod, *prod_bseq);
+
+bce_get_rx_buf_exit:
+ DBEXIT(BCE_EXTREME_RESET | BCE_EXTREME_RECV | BCE_EXTREME_LOAD);
+
+ return(rc);
+}
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+/****************************************************************************/
+/* Encapsulate an mbuf cluster into the page chain. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_get_pg_buf(struct bce_softc *sc, struct mbuf *m, u16 *prod,
+ u16 *prod_idx)
+{
+ bus_dmamap_t map;
+ bus_addr_t busaddr;
+ struct mbuf *m_new = NULL;
+ struct rx_bd *pgbd;
+ int error, rc = 0;
+#ifdef BCE_DEBUG
+ u16 debug_prod_idx = *prod_idx;
+#endif
+
+ DBENTER(BCE_EXTREME_RESET | BCE_EXTREME_RECV | BCE_EXTREME_LOAD);
+
+ /* Make sure the inputs are valid. */
+ DBRUNIF((*prod_idx > MAX_PG_BD),
+ BCE_PRINTF("%s(%d): page producer out of range: "
+ "0x%04X > 0x%04X\n", __FILE__, __LINE__,
+ *prod_idx, (u16) MAX_PG_BD));
+
+ DBPRINT(sc, BCE_EXTREME_RECV, "%s(enter): prod = 0x%04X, "
+ "chain_prod = 0x%04X\n", __FUNCTION__, *prod, *prod_idx);
+
+ /* Update counters if we've hit a new low or run out of pages. */
+ DBRUNIF((sc->free_pg_bd < sc->pg_low_watermark),
+ sc->pg_low_watermark = sc->free_pg_bd);
+ DBRUNIF((sc->free_pg_bd == sc->max_pg_bd), sc->pg_empty_count++);
+
+ /* Check whether this is a new mbuf allocation. */
+ if (m == NULL) {
+
+ /* Simulate an mbuf allocation failure. */
+ DBRUNIF(DB_RANDOMTRUE(mbuf_alloc_failed_sim_control),
+ sc->mbuf_alloc_failed_count++;
+ sc->mbuf_alloc_failed_sim_count++;
+ rc = ENOBUFS;
+ goto bce_get_pg_buf_exit);
+
+ /* This is a new mbuf allocation. */
+ m_new = m_getcl(M_DONTWAIT, MT_DATA, 0);
+ if (m_new == NULL) {
+ sc->mbuf_alloc_failed_count++;
+ rc = ENOBUFS;
+ goto bce_get_pg_buf_exit;
+ }
+
+ DBRUN(sc->debug_pg_mbuf_alloc++);
+ } else {
+ /* Reuse an existing mbuf. */
+ m_new = m;
+ m_new->m_data = m_new->m_ext.ext_buf;
+ }
+
+ m_new->m_len = sc->pg_bd_mbuf_alloc_size;
+
+ /* ToDo: Consider calling m_fragment() to test error handling. */
+
+ /* Map the mbuf cluster into device memory. */
+ map = sc->pg_mbuf_map[*prod_idx];
+ error = bus_dmamap_load(sc->pg_mbuf_tag, map, mtod(m_new, void *),
+ sc->pg_bd_mbuf_alloc_size, bce_dma_map_addr,
+ &busaddr, BUS_DMA_NOWAIT);
+
+ /* Handle any mapping errors. */
+ if (error) {
+ BCE_PRINTF("%s(%d): Error mapping mbuf into page chain!\n",
+ __FILE__, __LINE__);
+
+ m_freem(m_new);
+ DBRUN(sc->debug_pg_mbuf_alloc--);
+
+ rc = ENOBUFS;
+ goto bce_get_pg_buf_exit;
+ }
+
+ /* ToDo: Do we need bus_dmamap_sync(,,BUS_DMASYNC_PREREAD) here? */
+
+ /*
+ * The page chain uses the same rx_bd data structure
+ * as the receive chain but doesn't require a byte sequence (bseq).
+ */
+ pgbd = &sc->pg_bd_chain[PG_PAGE(*prod_idx)][PG_IDX(*prod_idx)];
+
+ pgbd->rx_bd_haddr_lo = htole32(BCE_ADDR_LO(busaddr));
+ pgbd->rx_bd_haddr_hi = htole32(BCE_ADDR_HI(busaddr));
+ pgbd->rx_bd_len = htole32(sc->pg_bd_mbuf_alloc_size);
+ pgbd->rx_bd_flags = htole32(RX_BD_FLAGS_START | RX_BD_FLAGS_END);
+
+ /* Save the mbuf and update our counter. */
+ sc->pg_mbuf_ptr[*prod_idx] = m_new;
+ sc->free_pg_bd--;
+
+ DBRUNMSG(BCE_INSANE_RECV,
+ bce_dump_pg_mbuf_chain(sc, debug_prod_idx, 1));
+
+ DBPRINT(sc, BCE_EXTREME_RECV, "%s(exit): prod = 0x%04X, "
+ "prod_idx = 0x%04X\n", __FUNCTION__, *prod, *prod_idx);
+
+bce_get_pg_buf_exit:
+ DBEXIT(BCE_EXTREME_RESET | BCE_EXTREME_RECV | BCE_EXTREME_LOAD);
+
+ return(rc);
+}
+#endif /* BCE_JUMBO_HDRSPLIT */
+
+
+/****************************************************************************/
+/* Initialize the TX context memory. */
+/* */
+/* Returns: */
+/* Nothing */
+/****************************************************************************/
+static void
+bce_init_tx_context(struct bce_softc *sc)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_SEND | BCE_VERBOSE_CTX);
+
+ /* Initialize the context ID for an L2 TX chain. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ /* Set the CID type to support an L2 connection. */
+ val = BCE_L2CTX_TX_TYPE_TYPE_L2_XI |
+ BCE_L2CTX_TX_TYPE_SIZE_L2_XI;
+ CTX_WR(sc, GET_CID_ADDR(TX_CID), BCE_L2CTX_TX_TYPE_XI, val);
+ val = BCE_L2CTX_TX_CMD_TYPE_TYPE_L2_XI | (8 << 16);
+ CTX_WR(sc, GET_CID_ADDR(TX_CID),
+ BCE_L2CTX_TX_CMD_TYPE_XI, val);
+
+ /* Point the hardware to the first page in the chain. */
+ val = BCE_ADDR_HI(sc->tx_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(TX_CID),
+ BCE_L2CTX_TX_TBDR_BHADDR_HI_XI, val);
+ val = BCE_ADDR_LO(sc->tx_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(TX_CID),
+ BCE_L2CTX_TX_TBDR_BHADDR_LO_XI, val);
+ } else {
+ /* Set the CID type to support an L2 connection. */
+ val = BCE_L2CTX_TX_TYPE_TYPE_L2 | BCE_L2CTX_TX_TYPE_SIZE_L2;
+ CTX_WR(sc, GET_CID_ADDR(TX_CID), BCE_L2CTX_TX_TYPE, val);
+ val = BCE_L2CTX_TX_CMD_TYPE_TYPE_L2 | (8 << 16);
+ CTX_WR(sc, GET_CID_ADDR(TX_CID), BCE_L2CTX_TX_CMD_TYPE, val);
+
+ /* Point the hardware to the first page in the chain. */
+ val = BCE_ADDR_HI(sc->tx_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(TX_CID),
+ BCE_L2CTX_TX_TBDR_BHADDR_HI, val);
+ val = BCE_ADDR_LO(sc->tx_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(TX_CID),
+ BCE_L2CTX_TX_TBDR_BHADDR_LO, val);
+ }
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_SEND | BCE_VERBOSE_CTX);
+}
+
+
+/****************************************************************************/
+/* Allocate memory and initialize the TX data structures. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_init_tx_chain(struct bce_softc *sc)
+{
+ struct tx_bd *txbd;
+ int i, rc = 0;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_SEND | BCE_VERBOSE_LOAD);
+
+ /* Set the initial TX producer/consumer indices. */
+ sc->tx_prod = 0;
+ sc->tx_cons = 0;
+ sc->tx_prod_bseq = 0;
+ sc->used_tx_bd = 0;
+ sc->max_tx_bd = USABLE_TX_BD;
+ DBRUN(sc->tx_hi_watermark = 0);
+ DBRUN(sc->tx_full_count = 0);
+
+ /*
+ * The NetXtreme II supports a linked-list structre called
+ * a Buffer Descriptor Chain (or BD chain). A BD chain
+ * consists of a series of 1 or more chain pages, each of which
+ * consists of a fixed number of BD entries.
+ * The last BD entry on each page is a pointer to the next page
+ * in the chain, and the last pointer in the BD chain
+ * points back to the beginning of the chain.
+ */
+
+ /* Set the TX next pointer chain entries. */
+ for (i = 0; i < TX_PAGES; i++) {
+ int j;
+
+ txbd = &sc->tx_bd_chain[i][USABLE_TX_BD_PER_PAGE];
+
+ /* Check if we've reached the last page. */
+ if (i == (TX_PAGES - 1))
+ j = 0;
+ else
+ j = i + 1;
+
+ txbd->tx_bd_haddr_hi = htole32(BCE_ADDR_HI(sc->tx_bd_chain_paddr[j]));
+ txbd->tx_bd_haddr_lo = htole32(BCE_ADDR_LO(sc->tx_bd_chain_paddr[j]));
+ }
+
+ bce_init_tx_context(sc);
+
+ DBRUNMSG(BCE_INSANE_SEND, bce_dump_tx_chain(sc, 0, TOTAL_TX_BD));
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_SEND | BCE_VERBOSE_LOAD);
+
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Free memory and clear the TX data structures. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_free_tx_chain(struct bce_softc *sc)
+{
+ int i;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_SEND | BCE_VERBOSE_UNLOAD);
+
+ /* Unmap, unload, and free any mbufs still in the TX mbuf chain. */
+ for (i = 0; i < TOTAL_TX_BD; i++) {
+ if (sc->tx_mbuf_ptr[i] != NULL) {
+ if (sc->tx_mbuf_map[i] != NULL)
+ bus_dmamap_sync(sc->tx_mbuf_tag,
+ sc->tx_mbuf_map[i],
+ BUS_DMASYNC_POSTWRITE);
+ m_freem(sc->tx_mbuf_ptr[i]);
+ sc->tx_mbuf_ptr[i] = NULL;
+ DBRUN(sc->debug_tx_mbuf_alloc--);
+ }
+ }
+
+ /* Clear each TX chain page. */
+ for (i = 0; i < TX_PAGES; i++)
+ bzero((char *)sc->tx_bd_chain[i], BCE_TX_CHAIN_PAGE_SZ);
+
+ sc->used_tx_bd = 0;
+
+ /* Check if we lost any mbufs in the process. */
+ DBRUNIF((sc->debug_tx_mbuf_alloc),
+ BCE_PRINTF("%s(%d): Memory leak! Lost %d mbufs "
+ "from tx chain!\n", __FILE__, __LINE__,
+ sc->debug_tx_mbuf_alloc));
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_SEND | BCE_VERBOSE_UNLOAD);
+}
+
+
+/****************************************************************************/
+/* Initialize the RX context memory. */
+/* */
+/* Returns: */
+/* Nothing */
+/****************************************************************************/
+static void
+bce_init_rx_context(struct bce_softc *sc)
+{
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_CTX);
+
+ /* Init the type, size, and BD cache levels for the RX context. */
+ val = BCE_L2CTX_RX_CTX_TYPE_CTX_BD_CHN_TYPE_VALUE |
+ BCE_L2CTX_RX_CTX_TYPE_SIZE_L2 |
+ (0x02 << BCE_L2CTX_RX_BD_PRE_READ_SHIFT);
+
+ /*
+ * Set the level for generating pause frames
+ * when the number of available rx_bd's gets
+ * too low (the low watermark) and the level
+ * when pause frames can be stopped (the high
+ * watermark).
+ */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ u32 lo_water, hi_water;
+
+ if (sc->bce_flags && BCE_USING_TX_FLOW_CONTROL) {
+ lo_water = BCE_L2CTX_RX_LO_WATER_MARK_DEFAULT;
+ } else {
+ lo_water = 0;
+ }
+
+ if (lo_water >= USABLE_RX_BD) {
+ lo_water = 0;
+ }
+
+ hi_water = USABLE_RX_BD / 4;
+
+ if (hi_water <= lo_water) {
+ lo_water = 0;
+ }
+
+ lo_water /= BCE_L2CTX_RX_LO_WATER_MARK_SCALE;
+ hi_water /= BCE_L2CTX_RX_HI_WATER_MARK_SCALE;
+
+ if (hi_water > 0xf)
+ hi_water = 0xf;
+ else if (hi_water == 0)
+ lo_water = 0;
+
+ val |= (lo_water << BCE_L2CTX_RX_LO_WATER_MARK_SHIFT) |
+ (hi_water << BCE_L2CTX_RX_HI_WATER_MARK_SHIFT);
+ }
+
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_CTX_TYPE, val);
+
+ /* Setup the MQ BIN mapping for l2_ctx_host_bseq. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ val = REG_RD(sc, BCE_MQ_MAP_L2_5);
+ REG_WR(sc, BCE_MQ_MAP_L2_5, val | BCE_MQ_MAP_L2_5_ARM);
+ }
+
+ /* Point the hardware to the first page in the chain. */
+ val = BCE_ADDR_HI(sc->rx_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_NX_BDHADDR_HI, val);
+ val = BCE_ADDR_LO(sc->rx_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_NX_BDHADDR_LO, val);
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_CTX);
+}
+
+
+/****************************************************************************/
+/* Allocate memory and initialize the RX data structures. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_init_rx_chain(struct bce_softc *sc)
+{
+ struct rx_bd *rxbd;
+ int i, rc = 0;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+
+ /* Initialize the RX producer and consumer indices. */
+ sc->rx_prod = 0;
+ sc->rx_cons = 0;
+ sc->rx_prod_bseq = 0;
+ sc->free_rx_bd = USABLE_RX_BD;
+ sc->max_rx_bd = USABLE_RX_BD;
+
+ /* Initialize the RX next pointer chain entries. */
+ for (i = 0; i < RX_PAGES; i++) {
+ int j;
+
+ rxbd = &sc->rx_bd_chain[i][USABLE_RX_BD_PER_PAGE];
+
+ /* Check if we've reached the last page. */
+ if (i == (RX_PAGES - 1))
+ j = 0;
+ else
+ j = i + 1;
+
+ /* Setup the chain page pointers. */
+ rxbd->rx_bd_haddr_hi =
+ htole32(BCE_ADDR_HI(sc->rx_bd_chain_paddr[j]));
+ rxbd->rx_bd_haddr_lo =
+ htole32(BCE_ADDR_LO(sc->rx_bd_chain_paddr[j]));
+ }
+
+ /* Fill up the RX chain. */
+ bce_fill_rx_chain(sc);
+
+ DBRUN(sc->rx_low_watermark = USABLE_RX_BD);
+ DBRUN(sc->rx_empty_count = 0);
+ for (i = 0; i < RX_PAGES; i++) {
+ bus_dmamap_sync(sc->rx_bd_chain_tag, sc->rx_bd_chain_map[i],
+ BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
+ }
+
+ bce_init_rx_context(sc);
+
+ DBRUNMSG(BCE_EXTREME_RECV, bce_dump_rx_bd_chain(sc, 0, TOTAL_RX_BD));
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+
+ /* ToDo: Are there possible failure modes here? */
+
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Add mbufs to the RX chain until its full or an mbuf allocation error */
+/* occurs. */
+/* */
+/* Returns: */
+/* Nothing */
+/****************************************************************************/
+static void
+bce_fill_rx_chain(struct bce_softc *sc)
+{
+ u16 prod, prod_idx;
+ u32 prod_bseq;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_EXTREME_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+
+ /* Get the RX chain producer indices. */
+ prod = sc->rx_prod;
+ prod_bseq = sc->rx_prod_bseq;
+
+ /* Keep filling the RX chain until it's full. */
+ while (sc->free_rx_bd > 0) {
+ prod_idx = RX_CHAIN_IDX(prod);
+ if (bce_get_rx_buf(sc, NULL, &prod, &prod_idx, &prod_bseq)) {
+ /* Bail out if we can't add an mbuf to the chain. */
+ break;
+ }
+ prod = NEXT_RX_BD(prod);
+ }
+
+ /* Save the RX chain producer indices. */
+ sc->rx_prod = prod;
+ sc->rx_prod_bseq = prod_bseq;
+
+ /* We should never end up pointing to a next page pointer. */
+ DBRUNIF(((prod & USABLE_RX_BD_PER_PAGE) == USABLE_RX_BD_PER_PAGE),
+ BCE_PRINTF("%s(): Invalid rx_prod value: 0x%04X\n",
+ __FUNCTION__, sc->rx_prod));
+
+ /* Write the mailbox and tell the chip about the waiting rx_bd's. */
+ REG_WR16(sc, MB_GET_CID_ADDR(RX_CID) +
+ BCE_L2MQ_RX_HOST_BDIDX, sc->rx_prod);
+ REG_WR(sc, MB_GET_CID_ADDR(RX_CID) +
+ BCE_L2MQ_RX_HOST_BSEQ, sc->rx_prod_bseq);
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_EXTREME_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+}
+
+
+/****************************************************************************/
+/* Free memory and clear the RX data structures. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_free_rx_chain(struct bce_softc *sc)
+{
+ int i;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_UNLOAD);
+
+ /* Free any mbufs still in the RX mbuf chain. */
+ for (i = 0; i < TOTAL_RX_BD; i++) {
+ if (sc->rx_mbuf_ptr[i] != NULL) {
+ if (sc->rx_mbuf_map[i] != NULL)
+ bus_dmamap_sync(sc->rx_mbuf_tag,
+ sc->rx_mbuf_map[i],
+ BUS_DMASYNC_POSTREAD);
+ m_freem(sc->rx_mbuf_ptr[i]);
+ sc->rx_mbuf_ptr[i] = NULL;
+ DBRUN(sc->debug_rx_mbuf_alloc--);
+ }
+ }
+
+ /* Clear each RX chain page. */
+ for (i = 0; i < RX_PAGES; i++)
+ if (sc->rx_bd_chain[i] != NULL) {
+ bzero((char *)sc->rx_bd_chain[i],
+ BCE_RX_CHAIN_PAGE_SZ);
+ }
+
+ sc->free_rx_bd = sc->max_rx_bd;
+
+ /* Check if we lost any mbufs in the process. */
+ DBRUNIF((sc->debug_rx_mbuf_alloc),
+ BCE_PRINTF("%s(): Memory leak! Lost %d mbufs from rx chain!\n",
+ __FUNCTION__, sc->debug_rx_mbuf_alloc));
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_UNLOAD);
+}
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+/****************************************************************************/
+/* Allocate memory and initialize the page data structures. */
+/* Assumes that bce_init_rx_chain() has not already been called. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_init_pg_chain(struct bce_softc *sc)
+{
+ struct rx_bd *pgbd;
+ int i, rc = 0;
+ u32 val;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+
+ /* Initialize the page producer and consumer indices. */
+ sc->pg_prod = 0;
+ sc->pg_cons = 0;
+ sc->free_pg_bd = USABLE_PG_BD;
+ sc->max_pg_bd = USABLE_PG_BD;
+ DBRUN(sc->pg_low_watermark = sc->max_pg_bd);
+ DBRUN(sc->pg_empty_count = 0);
+
+ /* Initialize the page next pointer chain entries. */
+ for (i = 0; i < PG_PAGES; i++) {
+ int j;
+
+ pgbd = &sc->pg_bd_chain[i][USABLE_PG_BD_PER_PAGE];
+
+ /* Check if we've reached the last page. */
+ if (i == (PG_PAGES - 1))
+ j = 0;
+ else
+ j = i + 1;
+
+ /* Setup the chain page pointers. */
+ pgbd->rx_bd_haddr_hi = htole32(BCE_ADDR_HI(sc->pg_bd_chain_paddr[j]));
+ pgbd->rx_bd_haddr_lo = htole32(BCE_ADDR_LO(sc->pg_bd_chain_paddr[j]));
+ }
+
+ /* Setup the MQ BIN mapping for host_pg_bidx. */
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716))
+ REG_WR(sc, BCE_MQ_MAP_L2_3, BCE_MQ_MAP_L2_3_DEFAULT);
+
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_PG_BUF_SIZE, 0);
+
+ /* Configure the rx_bd and page chain mbuf cluster size. */
+ val = (sc->rx_bd_mbuf_data_len << 16) | sc->pg_bd_mbuf_alloc_size;
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_PG_BUF_SIZE, val);
+
+ /* Configure the context reserved for jumbo support. */
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_RBDC_KEY,
+ BCE_L2CTX_RX_RBDC_JUMBO_KEY);
+
+ /* Point the hardware to the first page in the page chain. */
+ val = BCE_ADDR_HI(sc->pg_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_NX_PG_BDHADDR_HI, val);
+ val = BCE_ADDR_LO(sc->pg_bd_chain_paddr[0]);
+ CTX_WR(sc, GET_CID_ADDR(RX_CID), BCE_L2CTX_RX_NX_PG_BDHADDR_LO, val);
+
+ /* Fill up the page chain. */
+ bce_fill_pg_chain(sc);
+
+ for (i = 0; i < PG_PAGES; i++) {
+ bus_dmamap_sync(sc->pg_bd_chain_tag, sc->pg_bd_chain_map[i],
+ BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
+ }
+
+ DBRUNMSG(BCE_EXTREME_RECV, bce_dump_pg_chain(sc, 0, TOTAL_PG_BD));
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Add mbufs to the page chain until its full or an mbuf allocation error */
+/* occurs. */
+/* */
+/* Returns: */
+/* Nothing */
+/****************************************************************************/
+static void
+bce_fill_pg_chain(struct bce_softc *sc)
+{
+ u16 prod, prod_idx;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_EXTREME_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+
+ /* Get the page chain prodcuer index. */
+ prod = sc->pg_prod;
+
+ /* Keep filling the page chain until it's full. */
+ while (sc->free_pg_bd > 0) {
+ prod_idx = PG_CHAIN_IDX(prod);
+ if (bce_get_pg_buf(sc, NULL, &prod, &prod_idx)) {
+ /* Bail out if we can't add an mbuf to the chain. */
+ break;
+ }
+ prod = NEXT_PG_BD(prod);
+ }
+
+ /* Save the page chain producer index. */
+ sc->pg_prod = prod;
+
+ DBRUNIF(((prod & USABLE_RX_BD_PER_PAGE) == USABLE_RX_BD_PER_PAGE),
+ BCE_PRINTF("%s(): Invalid pg_prod value: 0x%04X\n",
+ __FUNCTION__, sc->pg_prod));
+
+ /*
+ * Write the mailbox and tell the chip about
+ * the new rx_bd's in the page chain.
+ */
+ REG_WR16(sc, MB_GET_CID_ADDR(RX_CID) +
+ BCE_L2MQ_RX_HOST_PG_BDIDX, sc->pg_prod);
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_EXTREME_RECV | BCE_VERBOSE_LOAD |
+ BCE_VERBOSE_CTX);
+}
+
+
+/****************************************************************************/
+/* Free memory and clear the RX data structures. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_free_pg_chain(struct bce_softc *sc)
+{
+ int i;
+
+ DBENTER(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_UNLOAD);
+
+ /* Free any mbufs still in the mbuf page chain. */
+ for (i = 0; i < TOTAL_PG_BD; i++) {
+ if (sc->pg_mbuf_ptr[i] != NULL) {
+ if (sc->pg_mbuf_map[i] != NULL)
+ bus_dmamap_sync(sc->pg_mbuf_tag,
+ sc->pg_mbuf_map[i],
+ BUS_DMASYNC_POSTREAD);
+ m_freem(sc->pg_mbuf_ptr[i]);
+ sc->pg_mbuf_ptr[i] = NULL;
+ DBRUN(sc->debug_pg_mbuf_alloc--);
+ }
+ }
+
+ /* Clear each page chain pages. */
+ for (i = 0; i < PG_PAGES; i++)
+ bzero((char *)sc->pg_bd_chain[i], BCE_PG_CHAIN_PAGE_SZ);
+
+ sc->free_pg_bd = sc->max_pg_bd;
+
+ /* Check if we lost any mbufs in the process. */
+ DBRUNIF((sc->debug_pg_mbuf_alloc),
+ BCE_PRINTF("%s(): Memory leak! Lost %d mbufs from page chain!\n",
+ __FUNCTION__, sc->debug_pg_mbuf_alloc));
+
+ DBEXIT(BCE_VERBOSE_RESET | BCE_VERBOSE_RECV | BCE_VERBOSE_UNLOAD);
+}
+#endif /* BCE_JUMBO_HDRSPLIT */
+
+
+/****************************************************************************/
+/* Set media options. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_ifmedia_upd(struct ifnet *ifp)
+{
+ struct bce_softc *sc = ifp->if_softc;
+ int error;
+
+ DBENTER(BCE_VERBOSE);
+
+ BCE_LOCK(sc);
+ error = bce_ifmedia_upd_locked(ifp);
+ BCE_UNLOCK(sc);
+
+ DBEXIT(BCE_VERBOSE);
+ return (error);
+}
+
+
+/****************************************************************************/
+/* Set media options. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static int
+bce_ifmedia_upd_locked(struct ifnet *ifp)
+{
+ struct bce_softc *sc = ifp->if_softc;
+ struct mii_data *mii;
+ int error;
+
+ DBENTER(BCE_VERBOSE_PHY);
+
+ error = 0;
+ BCE_LOCK_ASSERT(sc);
+
+ mii = device_get_softc(sc->bce_miibus);
+
+ /* Make sure the MII bus has been enumerated. */
+ if (mii) {
+ sc->bce_link_up = FALSE;
+ if (mii->mii_instance) {
+ struct mii_softc *miisc;
+
+ LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
+ mii_phy_reset(miisc);
+ }
+ error = mii_mediachg(mii);
+ }
+
+ DBEXIT(BCE_VERBOSE_PHY);
+ return (error);
+}
+
+
+/****************************************************************************/
+/* Reports current media status. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
+{
+ struct bce_softc *sc = ifp->if_softc;
+ struct mii_data *mii;
+
+ DBENTER(BCE_VERBOSE_PHY);
+
+ BCE_LOCK(sc);
+
+ if ((ifp->if_flags & IFF_UP) == 0) {
+ BCE_UNLOCK(sc);
+ return;
+ }
+ mii = device_get_softc(sc->bce_miibus);
+
+ mii_pollstat(mii);
+ ifmr->ifm_active = mii->mii_media_active;
+ ifmr->ifm_status = mii->mii_media_status;
+
+ BCE_UNLOCK(sc);
+
+ DBEXIT(BCE_VERBOSE_PHY);
+}
+
+
+/****************************************************************************/
+/* Handles PHY generated interrupt events. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_phy_intr(struct bce_softc *sc)
+{
+ u32 new_link_state, old_link_state;
+
+ DBENTER(BCE_VERBOSE_PHY | BCE_VERBOSE_INTR);
+
+ DBRUN(sc->phy_interrupts++);
+
+ new_link_state = sc->status_block->status_attn_bits &
+ STATUS_ATTN_BITS_LINK_STATE;
+ old_link_state = sc->status_block->status_attn_bits_ack &
+ STATUS_ATTN_BITS_LINK_STATE;
+
+ /* Handle any changes if the link state has changed. */
+ if (new_link_state != old_link_state) {
+
+ /* Update the status_attn_bits_ack field. */
+ if (new_link_state) {
+ REG_WR(sc, BCE_PCICFG_STATUS_BIT_SET_CMD,
+ STATUS_ATTN_BITS_LINK_STATE);
+ DBPRINT(sc, BCE_INFO_PHY, "%s(): Link is now UP.\n",
+ __FUNCTION__);
+ }
+ else {
+ REG_WR(sc, BCE_PCICFG_STATUS_BIT_CLEAR_CMD,
+ STATUS_ATTN_BITS_LINK_STATE);
+ DBPRINT(sc, BCE_INFO_PHY, "%s(): Link is now DOWN.\n",
+ __FUNCTION__);
+ }
+
+ /*
+ * Assume link is down and allow
+ * tick routine to update the state
+ * based on the actual media state.
+ */
+ sc->bce_link_up = FALSE;
+ callout_stop(&sc->bce_tick_callout);
+ bce_tick(sc);
+ }
+
+ /* Acknowledge the link change interrupt. */
+ REG_WR(sc, BCE_EMAC_STATUS, BCE_EMAC_STATUS_LINK_CHANGE);
+
+ DBEXIT(BCE_VERBOSE_PHY | BCE_VERBOSE_INTR);
+}
+
+
+/****************************************************************************/
+/* Reads the receive consumer value from the status block (skipping over */
+/* chain page pointer if necessary). */
+/* */
+/* Returns: */
+/* hw_cons */
+/****************************************************************************/
+static inline u16
+bce_get_hw_rx_cons(struct bce_softc *sc)
+{
+ u16 hw_cons;
+
+ rmb();
+ hw_cons = sc->status_block->status_rx_quick_consumer_index0;
+ if ((hw_cons & USABLE_RX_BD_PER_PAGE) == USABLE_RX_BD_PER_PAGE)
+ hw_cons++;
+
+ return hw_cons;
+}
+
+/****************************************************************************/
+/* Handles received frame interrupt events. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_rx_intr(struct bce_softc *sc)
+{
+ struct ifnet *ifp = sc->bce_ifp;
+ struct l2_fhdr *l2fhdr;
+ struct ether_vlan_header *vh;
+ unsigned int pkt_len;
+ u16 sw_rx_cons, sw_rx_cons_idx, hw_rx_cons;
+ u32 status;
+#ifdef BCE_JUMBO_HDRSPLIT
+ unsigned int rem_len;
+ u16 sw_pg_cons, sw_pg_cons_idx;
+#endif
+
+ DBENTER(BCE_VERBOSE_RECV | BCE_VERBOSE_INTR);
+ DBRUN(sc->interrupts_rx++);
+ DBPRINT(sc, BCE_EXTREME_RECV, "%s(enter): rx_prod = 0x%04X, "
+ "rx_cons = 0x%04X, rx_prod_bseq = 0x%08X\n",
+ __FUNCTION__, sc->rx_prod, sc->rx_cons, sc->rx_prod_bseq);
+
+ /* Prepare the RX chain pages to be accessed by the host CPU. */
+ for (int i = 0; i < RX_PAGES; i++)
+ bus_dmamap_sync(sc->rx_bd_chain_tag,
+ sc->rx_bd_chain_map[i], BUS_DMASYNC_POSTREAD);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ /* Prepare the page chain pages to be accessed by the host CPU. */
+ for (int i = 0; i < PG_PAGES; i++)
+ bus_dmamap_sync(sc->pg_bd_chain_tag,
+ sc->pg_bd_chain_map[i], BUS_DMASYNC_POSTREAD);
+#endif
+
+ /* Get the hardware's view of the RX consumer index. */
+ hw_rx_cons = sc->hw_rx_cons = bce_get_hw_rx_cons(sc);
+
+ /* Get working copies of the driver's view of the consumer indices. */
+ sw_rx_cons = sc->rx_cons;
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ sw_pg_cons = sc->pg_cons;
+#endif
+
+ /* Update some debug statistics counters */
+ DBRUNIF((sc->free_rx_bd < sc->rx_low_watermark),
+ sc->rx_low_watermark = sc->free_rx_bd);
+ DBRUNIF((sc->free_rx_bd == sc->max_rx_bd),
+ sc->rx_empty_count++);
+
+ /* Scan through the receive chain as long as there is work to do */
+ /* ToDo: Consider setting a limit on the number of packets processed. */
+ rmb();
+ while (sw_rx_cons != hw_rx_cons) {
+ struct mbuf *m0;
+
+ /* Convert the producer/consumer indices to an actual rx_bd index. */
+ sw_rx_cons_idx = RX_CHAIN_IDX(sw_rx_cons);
+
+ /* Unmap the mbuf from DMA space. */
+ bus_dmamap_sync(sc->rx_mbuf_tag,
+ sc->rx_mbuf_map[sw_rx_cons_idx],
+ BUS_DMASYNC_POSTREAD);
+ bus_dmamap_unload(sc->rx_mbuf_tag,
+ sc->rx_mbuf_map[sw_rx_cons_idx]);
+
+ /* Remove the mbuf from the RX chain. */
+ m0 = sc->rx_mbuf_ptr[sw_rx_cons_idx];
+ sc->rx_mbuf_ptr[sw_rx_cons_idx] = NULL;
+ DBRUN(sc->debug_rx_mbuf_alloc--);
+ sc->free_rx_bd++;
+
+ if(m0 == NULL) {
+ DBPRINT(sc, BCE_EXTREME_RECV,
+ "%s(): Oops! Empty mbuf pointer "
+ "found in sc->rx_mbuf_ptr[0x%04X]!\n",
+ __FUNCTION__, sw_rx_cons_idx);
+ goto bce_rx_int_next_rx;
+ }
+
+ /*
+ * Frames received on the NetXteme II are prepended
+ * with an l2_fhdr structure which provides status
+ * information about the received frame (including
+ * VLAN tags and checksum info). The frames are
+ * also automatically adjusted to align the IP
+ * header (i.e. two null bytes are inserted before
+ * the Ethernet header). As a result the data
+ * DMA'd by the controller into the mbuf looks
+ * like this:
+ *
+ * +---------+-----+---------------------+-----+
+ * | l2_fhdr | pad | packet data | FCS |
+ * +---------+-----+---------------------+-----+
+ *
+ * The l2_fhdr needs to be checked and skipped and
+ * the FCS needs to be stripped before sending the
+ * packet up the stack.
+ */
+ l2fhdr = mtod(m0, struct l2_fhdr *);
+
+ /* Get the packet data + FCS length and the status. */
+ pkt_len = l2fhdr->l2_fhdr_pkt_len;
+ status = l2fhdr->l2_fhdr_status;
+
+ /*
+ * Skip over the l2_fhdr and pad, resulting in the
+ * following data in the mbuf:
+ * +---------------------+-----+
+ * | packet data | FCS |
+ * +---------------------+-----+
+ */
+ m_adj(m0, sizeof(struct l2_fhdr) + ETHER_ALIGN);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ /*
+ * Check whether the received frame fits in a single
+ * mbuf or not (i.e. packet data + FCS <=
+ * sc->rx_bd_mbuf_data_len bytes).
+ */
+ if (pkt_len > m0->m_len) {
+ /*
+ * The received frame is larger than a single mbuf.
+ * If the frame was a TCP frame then only the TCP
+ * header is placed in the mbuf, the remaining
+ * payload (including FCS) is placed in the page
+ * chain, the SPLIT flag is set, and the header
+ * length is placed in the IP checksum field.
+ * If the frame is not a TCP frame then the mbuf
+ * is filled and the remaining bytes are placed
+ * in the page chain.
+ */
+
+ DBPRINT(sc, BCE_INFO_RECV, "%s(): Found a large "
+ "packet.\n", __FUNCTION__);
+
+ /*
+ * When the page chain is enabled and the TCP
+ * header has been split from the TCP payload,
+ * the ip_xsum structure will reflect the length
+ * of the TCP header, not the IP checksum. Set
+ * the packet length of the mbuf accordingly.
+ */
+ if (status & L2_FHDR_STATUS_SPLIT)
+ m0->m_len = l2fhdr->l2_fhdr_ip_xsum;
+
+ rem_len = pkt_len - m0->m_len;
+
+ /* Pull mbufs off the page chain for the remaining data. */
+ while (rem_len > 0) {
+ struct mbuf *m_pg;
+
+ sw_pg_cons_idx = PG_CHAIN_IDX(sw_pg_cons);
+
+ /* Remove the mbuf from the page chain. */
+ m_pg = sc->pg_mbuf_ptr[sw_pg_cons_idx];
+ sc->pg_mbuf_ptr[sw_pg_cons_idx] = NULL;
+ DBRUN(sc->debug_pg_mbuf_alloc--);
+ sc->free_pg_bd++;
+
+ /* Unmap the page chain mbuf from DMA space. */
+ bus_dmamap_sync(sc->pg_mbuf_tag,
+ sc->pg_mbuf_map[sw_pg_cons_idx],
+ BUS_DMASYNC_POSTREAD);
+ bus_dmamap_unload(sc->pg_mbuf_tag,
+ sc->pg_mbuf_map[sw_pg_cons_idx]);
+
+ /* Adjust the mbuf length. */
+ if (rem_len < m_pg->m_len) {
+ /* The mbuf chain is complete. */
+ m_pg->m_len = rem_len;
+ rem_len = 0;
+ } else {
+ /* More packet data is waiting. */
+ rem_len -= m_pg->m_len;
+ }
+
+ /* Concatenate the mbuf cluster to the mbuf. */
+ m_cat(m0, m_pg);
+
+ sw_pg_cons = NEXT_PG_BD(sw_pg_cons);
+ }
+
+ /* Set the total packet length. */
+ m0->m_pkthdr.len = pkt_len;
+
+ } else {
+ /*
+ * The received packet is small and fits in a
+ * single mbuf (i.e. the l2_fhdr + pad + packet +
+ * FCS <= MHLEN). In other words, the packet is
+ * 154 bytes or less in size.
+ */
+
+ DBPRINT(sc, BCE_INFO_RECV, "%s(): Found a small "
+ "packet.\n", __FUNCTION__);
+
+ /* Set the total packet length. */
+ m0->m_pkthdr.len = m0->m_len = pkt_len;
+ }
+#else
+ /* Set the total packet length. */
+ m0->m_pkthdr.len = m0->m_len = pkt_len;
+#endif
+
+ /* Remove the trailing Ethernet FCS. */
+ m_adj(m0, -ETHER_CRC_LEN);
+
+ /* Check that the resulting mbuf chain is valid. */
+ DBRUN(m_sanity(m0, FALSE));
+ DBRUNIF(((m0->m_len < ETHER_HDR_LEN) |
+ (m0->m_pkthdr.len > BCE_MAX_JUMBO_ETHER_MTU_VLAN)),
+ BCE_PRINTF("Invalid Ethernet frame size!\n");
+ m_print(m0, 128));
+
+ DBRUNIF(DB_RANDOMTRUE(l2fhdr_error_sim_control),
+ sc->l2fhdr_error_sim_count++;
+ status = status | L2_FHDR_ERRORS_PHY_DECODE);
+
+ /* Check the received frame for errors. */
+ if (status & (L2_FHDR_ERRORS_BAD_CRC |
+ L2_FHDR_ERRORS_PHY_DECODE | L2_FHDR_ERRORS_ALIGNMENT |
+ L2_FHDR_ERRORS_TOO_SHORT | L2_FHDR_ERRORS_GIANT_FRAME)) {
+
+ /* Log the error and release the mbuf. */
+ ifp->if_ierrors++;
+ sc->l2fhdr_error_count++;
+
+ m_freem(m0);
+ m0 = NULL;
+ goto bce_rx_int_next_rx;
+ }
+
+ /* Send the packet to the appropriate interface. */
+ m0->m_pkthdr.rcvif = ifp;
+
+ /* Assume no hardware checksum. */
+ m0->m_pkthdr.csum_flags = 0;
+
+ /* Validate the checksum if offload enabled. */
+ if (ifp->if_capenable & IFCAP_RXCSUM) {
+
+ /* Check for an IP datagram. */
+ if (!(status & L2_FHDR_STATUS_SPLIT) &&
+ (status & L2_FHDR_STATUS_IP_DATAGRAM)) {
+ m0->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
+ DBRUN(sc->csum_offload_ip++);
+ /* Check if the IP checksum is valid. */
+ if ((l2fhdr->l2_fhdr_ip_xsum ^ 0xffff) == 0)
+ m0->m_pkthdr.csum_flags |=
+ CSUM_IP_VALID;
+ }
+
+ /* Check for a valid TCP/UDP frame. */
+ if (status & (L2_FHDR_STATUS_TCP_SEGMENT |
+ L2_FHDR_STATUS_UDP_DATAGRAM)) {
+
+ /* Check for a good TCP/UDP checksum. */
+ if ((status & (L2_FHDR_ERRORS_TCP_XSUM |
+ L2_FHDR_ERRORS_UDP_XSUM)) == 0) {
+ DBRUN(sc->csum_offload_tcp_udp++);
+ m0->m_pkthdr.csum_data =
+ l2fhdr->l2_fhdr_tcp_udp_xsum;
+ m0->m_pkthdr.csum_flags |=
+ (CSUM_DATA_VALID
+ | CSUM_PSEUDO_HDR);
+ }
+ }
+ }
+
+ /* Attach the VLAN tag. */
+ if (status & L2_FHDR_STATUS_L2_VLAN_TAG) {
+ if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) {
+#if __FreeBSD_version < 700000
+ VLAN_INPUT_TAG(ifp, m0,
+ l2fhdr->l2_fhdr_vlan_tag, continue);
+#else
+ m0->m_pkthdr.ether_vtag =
+ l2fhdr->l2_fhdr_vlan_tag;
+ m0->m_flags |= M_VLANTAG;
+#endif
+ } else {
+ /*
+ * bce(4) controllers can't disable VLAN
+ * tag stripping if management firmware
+ * (ASF/IPMI/UMP) is running. So we always
+ * strip VLAN tag and manually reconstruct
+ * the VLAN frame by appending stripped
+ * VLAN tag in driver if VLAN tag stripping
+ * was disabled.
+ *
+ * TODO: LLC SNAP handling.
+ */
+ bcopy(mtod(m0, uint8_t *),
+ mtod(m0, uint8_t *) - ETHER_VLAN_ENCAP_LEN,
+ ETHER_ADDR_LEN * 2);
+ m0->m_data -= ETHER_VLAN_ENCAP_LEN;
+ vh = mtod(m0, struct ether_vlan_header *);
+ vh->evl_encap_proto = htons(ETHERTYPE_VLAN);
+ vh->evl_tag = htons(l2fhdr->l2_fhdr_vlan_tag);
+ m0->m_pkthdr.len += ETHER_VLAN_ENCAP_LEN;
+ m0->m_len += ETHER_VLAN_ENCAP_LEN;
+ }
+ }
+
+ /* Increment received packet statistics. */
+ ifp->if_ipackets++;
+
+bce_rx_int_next_rx:
+ sw_rx_cons = NEXT_RX_BD(sw_rx_cons);
+
+ /* If we have a packet, pass it up the stack */
+ if (m0) {
+ /* Make sure we don't lose our place when we release the lock. */
+ sc->rx_cons = sw_rx_cons;
+#ifdef BCE_JUMBO_HDRSPLIT
+ sc->pg_cons = sw_pg_cons;
+#endif
+
+ BCE_UNLOCK(sc);
+ (*ifp->if_input)(ifp, m0);
+ BCE_LOCK(sc);
+
+ /* Recover our place. */
+ sw_rx_cons = sc->rx_cons;
+#ifdef BCE_JUMBO_HDRSPLIT
+ sw_pg_cons = sc->pg_cons;
+#endif
+ }
+
+ /* Refresh hw_cons to see if there's new work */
+ if (sw_rx_cons == hw_rx_cons)
+ hw_rx_cons = sc->hw_rx_cons = bce_get_hw_rx_cons(sc);
+ }
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ /* No new packets. Refill the page chain. */
+ sc->pg_cons = sw_pg_cons;
+ bce_fill_pg_chain(sc);
+#endif
+
+ /* No new packets. Refill the RX chain. */
+ sc->rx_cons = sw_rx_cons;
+ bce_fill_rx_chain(sc);
+
+ /* Prepare the page chain pages to be accessed by the NIC. */
+ for (int i = 0; i < RX_PAGES; i++)
+ bus_dmamap_sync(sc->rx_bd_chain_tag,
+ sc->rx_bd_chain_map[i], BUS_DMASYNC_PREWRITE);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ for (int i = 0; i < PG_PAGES; i++)
+ bus_dmamap_sync(sc->pg_bd_chain_tag,
+ sc->pg_bd_chain_map[i], BUS_DMASYNC_PREWRITE);
+#endif
+
+ DBPRINT(sc, BCE_EXTREME_RECV, "%s(exit): rx_prod = 0x%04X, "
+ "rx_cons = 0x%04X, rx_prod_bseq = 0x%08X\n",
+ __FUNCTION__, sc->rx_prod, sc->rx_cons, sc->rx_prod_bseq);
+ DBEXIT(BCE_VERBOSE_RECV | BCE_VERBOSE_INTR);
+}
+
+
+/****************************************************************************/
+/* Reads the transmit consumer value from the status block (skipping over */
+/* chain page pointer if necessary). */
+/* */
+/* Returns: */
+/* hw_cons */
+/****************************************************************************/
+static inline u16
+bce_get_hw_tx_cons(struct bce_softc *sc)
+{
+ u16 hw_cons;
+
+ mb();
+ hw_cons = sc->status_block->status_tx_quick_consumer_index0;
+ if ((hw_cons & USABLE_TX_BD_PER_PAGE) == USABLE_TX_BD_PER_PAGE)
+ hw_cons++;
+
+ return hw_cons;
+}
+
+
+/****************************************************************************/
+/* Handles transmit completion interrupt events. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_tx_intr(struct bce_softc *sc)
+{
+ struct ifnet *ifp = sc->bce_ifp;
+ u16 hw_tx_cons, sw_tx_cons, sw_tx_chain_cons;
+
+ DBENTER(BCE_VERBOSE_SEND | BCE_VERBOSE_INTR);
+ DBRUN(sc->interrupts_tx++);
+ DBPRINT(sc, BCE_EXTREME_SEND, "%s(enter): tx_prod = 0x%04X, "
+ "tx_cons = 0x%04X, tx_prod_bseq = 0x%08X\n",
+ __FUNCTION__, sc->tx_prod, sc->tx_cons, sc->tx_prod_bseq);
+
+ BCE_LOCK_ASSERT(sc);
+
+ /* Get the hardware's view of the TX consumer index. */
+ hw_tx_cons = sc->hw_tx_cons = bce_get_hw_tx_cons(sc);
+ sw_tx_cons = sc->tx_cons;
+
+ /* Prevent speculative reads of the status block. */
+ bus_space_barrier(sc->bce_btag, sc->bce_bhandle, 0, 0,
+ BUS_SPACE_BARRIER_READ);
+
+ /* Cycle through any completed TX chain page entries. */
+ while (sw_tx_cons != hw_tx_cons) {
+#ifdef BCE_DEBUG
+ struct tx_bd *txbd = NULL;
+#endif
+ sw_tx_chain_cons = TX_CHAIN_IDX(sw_tx_cons);
+
+ DBPRINT(sc, BCE_INFO_SEND,
+ "%s(): hw_tx_cons = 0x%04X, sw_tx_cons = 0x%04X, "
+ "sw_tx_chain_cons = 0x%04X\n",
+ __FUNCTION__, hw_tx_cons, sw_tx_cons, sw_tx_chain_cons);
+
+ DBRUNIF((sw_tx_chain_cons > MAX_TX_BD),
+ BCE_PRINTF("%s(%d): TX chain consumer out of range! "
+ " 0x%04X > 0x%04X\n", __FILE__, __LINE__, sw_tx_chain_cons,
+ (int) MAX_TX_BD);
+ bce_breakpoint(sc));
+
+ DBRUN(txbd = &sc->tx_bd_chain[TX_PAGE(sw_tx_chain_cons)]
+ [TX_IDX(sw_tx_chain_cons)]);
+
+ DBRUNIF((txbd == NULL),
+ BCE_PRINTF("%s(%d): Unexpected NULL tx_bd[0x%04X]!\n",
+ __FILE__, __LINE__, sw_tx_chain_cons);
+ bce_breakpoint(sc));
+
+ DBRUNMSG(BCE_INFO_SEND, BCE_PRINTF("%s(): ", __FUNCTION__);
+ bce_dump_txbd(sc, sw_tx_chain_cons, txbd));
+
+ /*
+ * Free the associated mbuf. Remember
+ * that only the last tx_bd of a packet
+ * has an mbuf pointer and DMA map.
+ */
+ if (sc->tx_mbuf_ptr[sw_tx_chain_cons] != NULL) {
+
+ /* Validate that this is the last tx_bd. */
+ DBRUNIF((!(txbd->tx_bd_flags & TX_BD_FLAGS_END)),
+ BCE_PRINTF("%s(%d): tx_bd END flag not set but "
+ "txmbuf == NULL!\n", __FILE__, __LINE__);
+ bce_breakpoint(sc));
+
+ DBRUNMSG(BCE_INFO_SEND,
+ BCE_PRINTF("%s(): Unloading map/freeing mbuf "
+ "from tx_bd[0x%04X]\n", __FUNCTION__,
+ sw_tx_chain_cons));
+
+ /* Unmap the mbuf. */
+ bus_dmamap_unload(sc->tx_mbuf_tag,
+ sc->tx_mbuf_map[sw_tx_chain_cons]);
+
+ /* Free the mbuf. */
+ m_freem(sc->tx_mbuf_ptr[sw_tx_chain_cons]);
+ sc->tx_mbuf_ptr[sw_tx_chain_cons] = NULL;
+ DBRUN(sc->debug_tx_mbuf_alloc--);
+
+ ifp->if_opackets++;
+ }
+
+ sc->used_tx_bd--;
+ sw_tx_cons = NEXT_TX_BD(sw_tx_cons);
+
+ /* Refresh hw_cons to see if there's new work. */
+ hw_tx_cons = sc->hw_tx_cons = bce_get_hw_tx_cons(sc);
+
+ /* Prevent speculative reads of the status block. */
+ bus_space_barrier(sc->bce_btag, sc->bce_bhandle, 0, 0,
+ BUS_SPACE_BARRIER_READ);
+ }
+
+ /* Clear the TX timeout timer. */
+ sc->watchdog_timer = 0;
+
+ /* Clear the tx hardware queue full flag. */
+ if (sc->used_tx_bd < sc->max_tx_bd) {
+ DBRUNIF((ifp->if_drv_flags & IFF_DRV_OACTIVE),
+ DBPRINT(sc, BCE_INFO_SEND,
+ "%s(): Open TX chain! %d/%d (used/total)\n",
+ __FUNCTION__, sc->used_tx_bd, sc->max_tx_bd));
+ ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
+ }
+
+ sc->tx_cons = sw_tx_cons;
+
+ DBPRINT(sc, BCE_EXTREME_SEND, "%s(exit): tx_prod = 0x%04X, "
+ "tx_cons = 0x%04X, tx_prod_bseq = 0x%08X\n",
+ __FUNCTION__, sc->tx_prod, sc->tx_cons, sc->tx_prod_bseq);
+ DBEXIT(BCE_VERBOSE_SEND | BCE_VERBOSE_INTR);
+}
+
+
+/****************************************************************************/
+/* Disables interrupt generation. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_disable_intr(struct bce_softc *sc)
+{
+ DBENTER(BCE_VERBOSE_INTR);
+
+ REG_WR(sc, BCE_PCICFG_INT_ACK_CMD, BCE_PCICFG_INT_ACK_CMD_MASK_INT);
+ REG_RD(sc, BCE_PCICFG_INT_ACK_CMD);
+
+ DBEXIT(BCE_VERBOSE_INTR);
+}
+
+
+/****************************************************************************/
+/* Enables interrupt generation. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_enable_intr(struct bce_softc *sc, int coal_now)
+{
+ DBENTER(BCE_VERBOSE_INTR);
+
+ REG_WR(sc, BCE_PCICFG_INT_ACK_CMD,
+ BCE_PCICFG_INT_ACK_CMD_INDEX_VALID |
+ BCE_PCICFG_INT_ACK_CMD_MASK_INT | sc->last_status_idx);
+
+ REG_WR(sc, BCE_PCICFG_INT_ACK_CMD,
+ BCE_PCICFG_INT_ACK_CMD_INDEX_VALID | sc->last_status_idx);
+
+ /* Force an immediate interrupt (whether there is new data or not). */
+ if (coal_now)
+ REG_WR(sc, BCE_HC_COMMAND, sc->hc_command | BCE_HC_COMMAND_COAL_NOW);
+
+ DBEXIT(BCE_VERBOSE_INTR);
+}
+
+
+/****************************************************************************/
+/* Handles controller initialization. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init_locked(struct bce_softc *sc)
+{
+ struct ifnet *ifp;
+ u32 ether_mtu = 0;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ BCE_LOCK_ASSERT(sc);
+
+ ifp = sc->bce_ifp;
+
+ /* Check if the driver is still running and bail out if it is. */
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING)
+ goto bce_init_locked_exit;
+
+ bce_stop(sc);
+
+ if (bce_reset(sc, BCE_DRV_MSG_CODE_RESET)) {
+ BCE_PRINTF("%s(%d): Controller reset failed!\n",
+ __FILE__, __LINE__);
+ goto bce_init_locked_exit;
+ }
+
+ if (bce_chipinit(sc)) {
+ BCE_PRINTF("%s(%d): Controller initialization failed!\n",
+ __FILE__, __LINE__);
+ goto bce_init_locked_exit;
+ }
+
+ if (bce_blockinit(sc)) {
+ BCE_PRINTF("%s(%d): Block initialization failed!\n",
+ __FILE__, __LINE__);
+ goto bce_init_locked_exit;
+ }
+
+ /* Load our MAC address. */
+ bcopy(IF_LLADDR(sc->bce_ifp), sc->eaddr, ETHER_ADDR_LEN);
+ bce_set_mac_addr(sc);
+
+ /*
+ * Calculate and program the hardware Ethernet MTU
+ * size. Be generous on the receive if we have room.
+ */
+#ifdef BCE_JUMBO_HDRSPLIT
+ if (ifp->if_mtu <= (sc->rx_bd_mbuf_data_len +
+ sc->pg_bd_mbuf_alloc_size))
+ ether_mtu = sc->rx_bd_mbuf_data_len +
+ sc->pg_bd_mbuf_alloc_size;
+#else
+ if (ifp->if_mtu <= sc->rx_bd_mbuf_data_len)
+ ether_mtu = sc->rx_bd_mbuf_data_len;
+#endif
+ else
+ ether_mtu = ifp->if_mtu;
+
+ ether_mtu += ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN + ETHER_CRC_LEN;
+
+ DBPRINT(sc, BCE_INFO_MISC, "%s(): setting h/w mtu = %d\n",
+ __FUNCTION__, ether_mtu);
+
+ /* Program the mtu, enabling jumbo frame support if necessary. */
+ if (ether_mtu > (ETHER_MAX_LEN + ETHER_VLAN_ENCAP_LEN))
+ REG_WR(sc, BCE_EMAC_RX_MTU_SIZE,
+ min(ether_mtu, BCE_MAX_JUMBO_ETHER_MTU) |
+ BCE_EMAC_RX_MTU_SIZE_JUMBO_ENA);
+ else
+ REG_WR(sc, BCE_EMAC_RX_MTU_SIZE, ether_mtu);
+
+ DBPRINT(sc, BCE_INFO_LOAD,
+ "%s(): rx_bd_mbuf_alloc_size = %d, rx_bce_mbuf_data_len = %d, "
+ "rx_bd_mbuf_align_pad = %d\n", __FUNCTION__,
+ sc->rx_bd_mbuf_alloc_size, sc->rx_bd_mbuf_data_len,
+ sc->rx_bd_mbuf_align_pad);
+
+ /* Program appropriate promiscuous/multicast filtering. */
+ bce_set_rx_mode(sc);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ DBPRINT(sc, BCE_INFO_LOAD, "%s(): pg_bd_mbuf_alloc_size = %d\n",
+ __FUNCTION__, sc->pg_bd_mbuf_alloc_size);
+
+ /* Init page buffer descriptor chain. */
+ bce_init_pg_chain(sc);
+#endif
+
+ /* Init RX buffer descriptor chain. */
+ bce_init_rx_chain(sc);
+
+ /* Init TX buffer descriptor chain. */
+ bce_init_tx_chain(sc);
+
+ /* Enable host interrupts. */
+ bce_enable_intr(sc, 1);
+
+ bce_ifmedia_upd_locked(ifp);
+
+ /* Let the OS know the driver is up and running. */
+ ifp->if_drv_flags |= IFF_DRV_RUNNING;
+ ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
+
+ callout_reset(&sc->bce_tick_callout, hz, bce_tick, sc);
+
+bce_init_locked_exit:
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Initialize the controller just enough so that any management firmware */
+/* running on the device will continue to operate correctly. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_mgmt_init_locked(struct bce_softc *sc)
+{
+ struct ifnet *ifp;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ BCE_LOCK_ASSERT(sc);
+
+ /* Bail out if management firmware is not running. */
+ if (!(sc->bce_flags & BCE_MFW_ENABLE_FLAG)) {
+ DBPRINT(sc, BCE_VERBOSE_SPECIAL,
+ "No management firmware running...\n");
+ goto bce_mgmt_init_locked_exit;
+ }
+
+ ifp = sc->bce_ifp;
+
+ /* Enable all critical blocks in the MAC. */
+ REG_WR(sc, BCE_MISC_ENABLE_SET_BITS, BCE_MISC_ENABLE_DEFAULT);
+ REG_RD(sc, BCE_MISC_ENABLE_SET_BITS);
+ DELAY(20);
+
+ bce_ifmedia_upd_locked(ifp);
+
+bce_mgmt_init_locked_exit:
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Handles controller initialization when called from an unlocked routine. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_init(void *xsc)
+{
+ struct bce_softc *sc = xsc;
+
+ DBENTER(BCE_VERBOSE_RESET);
+
+ BCE_LOCK(sc);
+ bce_init_locked(sc);
+ BCE_UNLOCK(sc);
+
+ DBEXIT(BCE_VERBOSE_RESET);
+}
+
+
+/****************************************************************************/
+/* Modifies an mbuf for TSO on the hardware. */
+/* */
+/* Returns: */
+/* Pointer to a modified mbuf. */
+/****************************************************************************/
+static struct mbuf *
+bce_tso_setup(struct bce_softc *sc, struct mbuf **m_head, u16 *flags)
+{
+ struct mbuf *m;
+ struct ether_header *eh;
+ struct ip *ip;
+ struct tcphdr *th;
+ u16 etype;
+ int hdr_len, ip_hlen = 0, tcp_hlen = 0, ip_len = 0;
+
+ DBRUN(sc->tso_frames_requested++);
+
+ /* Controller may modify mbuf chains. */
+ if (M_WRITABLE(*m_head) == 0) {
+ m = m_dup(*m_head, M_DONTWAIT);
+ m_freem(*m_head);
+ if (m == NULL) {
+ sc->mbuf_alloc_failed_count++;
+ *m_head = NULL;
+ return (NULL);
+ }
+ *m_head = m;
+ }
+
+ /*
+ * For TSO the controller needs two pieces of info,
+ * the MSS and the IP+TCP options length.
+ */
+ m = m_pullup(*m_head, sizeof(struct ether_header) + sizeof(struct ip));
+ if (m == NULL) {
+ *m_head = NULL;
+ return (NULL);
+ }
+ eh = mtod(m, struct ether_header *);
+ etype = ntohs(eh->ether_type);
+
+ /* Check for supported TSO Ethernet types (only IPv4 for now) */
+ switch (etype) {
+ case ETHERTYPE_IP:
+ ip = (struct ip *)(m->m_data + sizeof(struct ether_header));
+ /* TSO only supported for TCP protocol. */
+ if (ip->ip_p != IPPROTO_TCP) {
+ BCE_PRINTF("%s(%d): TSO enabled for non-TCP frame!.\n",
+ __FILE__, __LINE__);
+ m_freem(*m_head);
+ *m_head = NULL;
+ return (NULL);
+ }
+
+ /* Get IP header length in bytes (min 20) */
+ ip_hlen = ip->ip_hl << 2;
+ m = m_pullup(*m_head, sizeof(struct ether_header) + ip_hlen +
+ sizeof(struct tcphdr));
+ if (m == NULL) {
+ *m_head = NULL;
+ return (NULL);
+ }
+
+ /* Get the TCP header length in bytes (min 20) */
+ ip = (struct ip *)(m->m_data + sizeof(struct ether_header));
+ th = (struct tcphdr *)((caddr_t)ip + ip_hlen);
+ tcp_hlen = (th->th_off << 2);
+
+ /* Make sure all IP/TCP options live in the same buffer. */
+ m = m_pullup(*m_head, sizeof(struct ether_header)+ ip_hlen +
+ tcp_hlen);
+ if (m == NULL) {
+ *m_head = NULL;
+ return (NULL);
+ }
+
+ /* IP header length and checksum will be calc'd by hardware */
+ ip = (struct ip *)(m->m_data + sizeof(struct ether_header));
+ ip_len = ip->ip_len;
+ ip->ip_len = 0;
+ ip->ip_sum = 0;
+ break;
+ case ETHERTYPE_IPV6:
+ BCE_PRINTF("%s(%d): TSO over IPv6 not supported!.\n",
+ __FILE__, __LINE__);
+ m_freem(*m_head);
+ *m_head = NULL;
+ return (NULL);
+ /* NOT REACHED */
+ default:
+ BCE_PRINTF("%s(%d): TSO enabled for unsupported protocol!.\n",
+ __FILE__, __LINE__);
+ m_freem(*m_head);
+ *m_head = NULL;
+ return (NULL);
+ }
+
+ hdr_len = sizeof(struct ether_header) + ip_hlen + tcp_hlen;
+
+ DBPRINT(sc, BCE_EXTREME_SEND, "%s(): hdr_len = %d, e_hlen = %d, "
+ "ip_hlen = %d, tcp_hlen = %d, ip_len = %d\n",
+ __FUNCTION__, hdr_len, (int) sizeof(struct ether_header), ip_hlen,
+ tcp_hlen, ip_len);
+
+ /* Set the LSO flag in the TX BD */
+ *flags |= TX_BD_FLAGS_SW_LSO;
+
+ /* Set the length of IP + TCP options (in 32 bit words) */
+ *flags |= (((ip_hlen + tcp_hlen - sizeof(struct ip) -
+ sizeof(struct tcphdr)) >> 2) << 8);
+
+ DBRUN(sc->tso_frames_completed++);
+ return (*m_head);
+}
+
+
+/****************************************************************************/
+/* Encapsultes an mbuf cluster into the tx_bd chain structure and makes the */
+/* memory visible to the controller. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/* Modified: */
+/* m_head: May be set to NULL if MBUF is excessively fragmented. */
+/****************************************************************************/
+static int
+bce_tx_encap(struct bce_softc *sc, struct mbuf **m_head)
+{
+ bus_dma_segment_t segs[BCE_MAX_SEGMENTS];
+ bus_dmamap_t map;
+ struct tx_bd *txbd = NULL;
+ struct mbuf *m0;
+ u16 prod, chain_prod, mss = 0, vlan_tag = 0, flags = 0;
+ u32 prod_bseq;
+
+#ifdef BCE_DEBUG
+ u16 debug_prod;
+#endif
+
+ int i, error, nsegs, rc = 0;
+
+ DBENTER(BCE_VERBOSE_SEND);
+
+ /* Make sure we have room in the TX chain. */
+ if (sc->used_tx_bd >= sc->max_tx_bd)
+ goto bce_tx_encap_exit;
+
+ /* Transfer any checksum offload flags to the bd. */
+ m0 = *m_head;
+ if (m0->m_pkthdr.csum_flags) {
+ if (m0->m_pkthdr.csum_flags & CSUM_TSO) {
+ m0 = bce_tso_setup(sc, m_head, &flags);
+ if (m0 == NULL) {
+ DBRUN(sc->tso_frames_failed++);
+ goto bce_tx_encap_exit;
+ }
+ mss = htole16(m0->m_pkthdr.tso_segsz);
+ } else {
+ if (m0->m_pkthdr.csum_flags & CSUM_IP)
+ flags |= TX_BD_FLAGS_IP_CKSUM;
+ if (m0->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP))
+ flags |= TX_BD_FLAGS_TCP_UDP_CKSUM;
+ }
+ }
+
+ /* Transfer any VLAN tags to the bd. */
+ if (m0->m_flags & M_VLANTAG) {
+ flags |= TX_BD_FLAGS_VLAN_TAG;
+ vlan_tag = m0->m_pkthdr.ether_vtag;
+ }
+
+ /* Map the mbuf into DMAable memory. */
+ prod = sc->tx_prod;
+ chain_prod = TX_CHAIN_IDX(prod);
+ map = sc->tx_mbuf_map[chain_prod];
+
+ /* Map the mbuf into our DMA address space. */
+ error = bus_dmamap_load_mbuf_sg(sc->tx_mbuf_tag, map, m0,
+ segs, &nsegs, BUS_DMA_NOWAIT);
+
+ /* Check if the DMA mapping was successful */
+ if (error == EFBIG) {
+ sc->mbuf_frag_count++;
+
+ /* Try to defrag the mbuf. */
+ m0 = m_collapse(*m_head, M_DONTWAIT, BCE_MAX_SEGMENTS);
+ if (m0 == NULL) {
+ /* Defrag was unsuccessful */
+ m_freem(*m_head);
+ *m_head = NULL;
+ sc->mbuf_alloc_failed_count++;
+ rc = ENOBUFS;
+ goto bce_tx_encap_exit;
+ }
+
+ /* Defrag was successful, try mapping again */
+ *m_head = m0;
+ error = bus_dmamap_load_mbuf_sg(sc->tx_mbuf_tag,
+ map, m0, segs, &nsegs, BUS_DMA_NOWAIT);
+
+ /* Still getting an error after a defrag. */
+ if (error == ENOMEM) {
+ /* Insufficient DMA buffers available. */
+ sc->dma_map_addr_tx_failed_count++;
+ rc = error;
+ goto bce_tx_encap_exit;
+ } else if (error != 0) {
+ /* Release it and return an error. */
+ BCE_PRINTF("%s(%d): Unknown error mapping mbuf into "
+ "TX chain!\n", __FILE__, __LINE__);
+ m_freem(m0);
+ *m_head = NULL;
+ sc->dma_map_addr_tx_failed_count++;
+ rc = ENOBUFS;
+ goto bce_tx_encap_exit;
+ }
+ } else if (error == ENOMEM) {
+ /* Insufficient DMA buffers available. */
+ sc->dma_map_addr_tx_failed_count++;
+ rc = error;
+ goto bce_tx_encap_exit;
+ } else if (error != 0) {
+ m_freem(m0);
+ *m_head = NULL;
+ sc->dma_map_addr_tx_failed_count++;
+ rc = error;
+ goto bce_tx_encap_exit;
+ }
+
+ /* Make sure there's room in the chain */
+ if (nsegs > (sc->max_tx_bd - sc->used_tx_bd)) {
+ bus_dmamap_unload(sc->tx_mbuf_tag, map);
+ rc = ENOBUFS;
+ goto bce_tx_encap_exit;
+ }
+
+ /* prod points to an empty tx_bd at this point. */
+ prod_bseq = sc->tx_prod_bseq;
+
+#ifdef BCE_DEBUG
+ debug_prod = chain_prod;
+#endif
+
+ DBPRINT(sc, BCE_INFO_SEND,
+ "%s(start): prod = 0x%04X, chain_prod = 0x%04X, "
+ "prod_bseq = 0x%08X\n",
+ __FUNCTION__, prod, chain_prod, prod_bseq);
+
+ /*
+ * Cycle through each mbuf segment that makes up
+ * the outgoing frame, gathering the mapping info
+ * for that segment and creating a tx_bd for
+ * the mbuf.
+ */
+ for (i = 0; i < nsegs ; i++) {
+
+ chain_prod = TX_CHAIN_IDX(prod);
+ txbd= &sc->tx_bd_chain[TX_PAGE(chain_prod)]
+ [TX_IDX(chain_prod)];
+
+ txbd->tx_bd_haddr_lo =
+ htole32(BCE_ADDR_LO(segs[i].ds_addr));
+ txbd->tx_bd_haddr_hi =
+ htole32(BCE_ADDR_HI(segs[i].ds_addr));
+ txbd->tx_bd_mss_nbytes = htole32(mss << 16) |
+ htole16(segs[i].ds_len);
+ txbd->tx_bd_vlan_tag = htole16(vlan_tag);
+ txbd->tx_bd_flags = htole16(flags);
+ prod_bseq += segs[i].ds_len;
+ if (i == 0)
+ txbd->tx_bd_flags |= htole16(TX_BD_FLAGS_START);
+ prod = NEXT_TX_BD(prod);
+ }
+
+ /* Set the END flag on the last TX buffer descriptor. */
+ txbd->tx_bd_flags |= htole16(TX_BD_FLAGS_END);
+
+ DBRUNMSG(BCE_EXTREME_SEND,
+ bce_dump_tx_chain(sc, debug_prod, nsegs));
+
+ /*
+ * Ensure that the mbuf pointer for this transmission
+ * is placed at the array index of the last
+ * descriptor in this chain. This is done
+ * because a single map is used for all
+ * segments of the mbuf and we don't want to
+ * unload the map before all of the segments
+ * have been freed.
+ */
+ sc->tx_mbuf_ptr[chain_prod] = m0;
+ sc->used_tx_bd += nsegs;
+
+ /* Update some debug statistic counters */
+ DBRUNIF((sc->used_tx_bd > sc->tx_hi_watermark),
+ sc->tx_hi_watermark = sc->used_tx_bd);
+ DBRUNIF((sc->used_tx_bd == sc->max_tx_bd), sc->tx_full_count++);
+ DBRUNIF(sc->debug_tx_mbuf_alloc++);
+
+ DBRUNMSG(BCE_EXTREME_SEND, bce_dump_tx_mbuf_chain(sc, chain_prod, 1));
+
+ /* prod points to the next free tx_bd at this point. */
+ sc->tx_prod = prod;
+ sc->tx_prod_bseq = prod_bseq;
+
+ /* Tell the chip about the waiting TX frames. */
+ REG_WR16(sc, MB_GET_CID_ADDR(TX_CID) +
+ BCE_L2MQ_TX_HOST_BIDX, sc->tx_prod);
+ REG_WR(sc, MB_GET_CID_ADDR(TX_CID) +
+ BCE_L2MQ_TX_HOST_BSEQ, sc->tx_prod_bseq);
+
+bce_tx_encap_exit:
+ DBEXIT(BCE_VERBOSE_SEND);
+ return(rc);
+}
+
+
+/****************************************************************************/
+/* Main transmit routine when called from another routine with a lock. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_start_locked(struct ifnet *ifp)
+{
+ struct bce_softc *sc = ifp->if_softc;
+ struct mbuf *m_head = NULL;
+ int count = 0;
+ u16 tx_prod, tx_chain_prod;
+
+ DBENTER(BCE_VERBOSE_SEND | BCE_VERBOSE_CTX);
+
+ BCE_LOCK_ASSERT(sc);
+
+ /* prod points to the next free tx_bd. */
+ tx_prod = sc->tx_prod;
+ tx_chain_prod = TX_CHAIN_IDX(tx_prod);
+
+ DBPRINT(sc, BCE_INFO_SEND,
+ "%s(enter): tx_prod = 0x%04X, tx_chain_prod = 0x%04X, "
+ "tx_prod_bseq = 0x%08X\n",
+ __FUNCTION__, tx_prod, tx_chain_prod, sc->tx_prod_bseq);
+
+ /* If there's no link or the transmit queue is empty then just exit. */
+ if (sc->bce_link_up == FALSE) {
+ DBPRINT(sc, BCE_INFO_SEND, "%s(): No link.\n",
+ __FUNCTION__);
+ goto bce_start_locked_exit;
+ }
+
+ if (IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
+ DBPRINT(sc, BCE_INFO_SEND, "%s(): Transmit queue empty.\n",
+ __FUNCTION__);
+ goto bce_start_locked_exit;
+ }
+
+ /*
+ * Keep adding entries while there is space in the ring.
+ */
+ while (sc->used_tx_bd < sc->max_tx_bd) {
+
+ /* Check for any frames to send. */
+ IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
+
+ /* Stop when the transmit queue is empty. */
+ if (m_head == NULL)
+ break;
+
+ /*
+ * Pack the data into the transmit ring. If we
+ * don't have room, place the mbuf back at the
+ * head of the queue and set the OACTIVE flag
+ * to wait for the NIC to drain the chain.
+ */
+ if (bce_tx_encap(sc, &m_head)) {
+ if (m_head != NULL)
+ IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
+ ifp->if_drv_flags |= IFF_DRV_OACTIVE;
+ DBPRINT(sc, BCE_INFO_SEND,
+ "TX chain is closed for business! Total "
+ "tx_bd used = %d\n", sc->used_tx_bd);
+ break;
+ }
+
+ count++;
+
+ /* Send a copy of the frame to any BPF listeners. */
+ ETHER_BPF_MTAP(ifp, m_head);
+ }
+
+ /* Exit if no packets were dequeued. */
+ if (count == 0) {
+ DBPRINT(sc, BCE_VERBOSE_SEND, "%s(): No packets were "
+ "dequeued\n", __FUNCTION__);
+ goto bce_start_locked_exit;
+ }
+
+ DBPRINT(sc, BCE_VERBOSE_SEND, "%s(): Inserted %d frames into "
+ "send queue.\n", __FUNCTION__, count);
+
+ /* Set the tx timeout. */
+ sc->watchdog_timer = BCE_TX_TIMEOUT;
+
+ DBRUNMSG(BCE_VERBOSE_SEND, bce_dump_ctx(sc, TX_CID));
+ DBRUNMSG(BCE_VERBOSE_SEND, bce_dump_mq_regs(sc));
+
+bce_start_locked_exit:
+ DBEXIT(BCE_VERBOSE_SEND | BCE_VERBOSE_CTX);
+ return;
+}
+
+
+/****************************************************************************/
+/* Main transmit routine when called from another routine without a lock. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_start(struct ifnet *ifp)
+{
+ struct bce_softc *sc = ifp->if_softc;
+
+ DBENTER(BCE_VERBOSE_SEND);
+
+ BCE_LOCK(sc);
+ bce_start_locked(ifp);
+ BCE_UNLOCK(sc);
+
+ DBEXIT(BCE_VERBOSE_SEND);
+}
+
+
+/****************************************************************************/
+/* Handles any IOCTL calls from the operating system. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
+{
+ struct bce_softc *sc = ifp->if_softc;
+ struct ifreq *ifr = (struct ifreq *) data;
+ struct mii_data *mii;
+ int mask, error = 0, reinit;
+
+ DBENTER(BCE_VERBOSE_MISC);
+
+ switch(command) {
+
+ /* Set the interface MTU. */
+ case SIOCSIFMTU:
+ /* Check that the MTU setting is supported. */
+ if ((ifr->ifr_mtu < BCE_MIN_MTU) ||
+ (ifr->ifr_mtu > BCE_MAX_JUMBO_MTU)) {
+ error = EINVAL;
+ break;
+ }
+
+ DBPRINT(sc, BCE_INFO_MISC,
+ "SIOCSIFMTU: Changing MTU from %d to %d\n",
+ (int) ifp->if_mtu, (int) ifr->ifr_mtu);
+
+ BCE_LOCK(sc);
+ ifp->if_mtu = ifr->ifr_mtu;
+ reinit = 0;
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
+ /*
+ * Because allocation size is used in RX
+ * buffer allocation, stop controller if
+ * it is already running.
+ */
+ bce_stop(sc);
+ reinit = 1;
+ }
+#ifdef BCE_JUMBO_HDRSPLIT
+ /* No buffer allocation size changes are necessary. */
+#else
+ /* Recalculate our buffer allocation sizes. */
+ if ((ifp->if_mtu + ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN +
+ ETHER_CRC_LEN) > MCLBYTES) {
+ sc->rx_bd_mbuf_alloc_size = MJUM9BYTES;
+ sc->rx_bd_mbuf_align_pad =
+ roundup2(MJUM9BYTES, 16) - MJUM9BYTES;
+ sc->rx_bd_mbuf_data_len =
+ sc->rx_bd_mbuf_alloc_size -
+ sc->rx_bd_mbuf_align_pad;
+ } else {
+ sc->rx_bd_mbuf_alloc_size = MCLBYTES;
+ sc->rx_bd_mbuf_align_pad =
+ roundup2(MCLBYTES, 16) - MCLBYTES;
+ sc->rx_bd_mbuf_data_len =
+ sc->rx_bd_mbuf_alloc_size -
+ sc->rx_bd_mbuf_align_pad;
+ }
+#endif
+
+ if (reinit != 0)
+ bce_init_locked(sc);
+ BCE_UNLOCK(sc);
+ break;
+
+ /* Set interface flags. */
+ case SIOCSIFFLAGS:
+ DBPRINT(sc, BCE_VERBOSE_SPECIAL, "Received SIOCSIFFLAGS\n");
+
+ BCE_LOCK(sc);
+
+ /* Check if the interface is up. */
+ if (ifp->if_flags & IFF_UP) {
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
+ /* Change promiscuous/multicast flags as necessary. */
+ bce_set_rx_mode(sc);
+ } else {
+ /* Start the HW */
+ bce_init_locked(sc);
+ }
+ } else {
+ /* The interface is down, check if driver is running. */
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
+ bce_stop(sc);
+
+ /* If MFW is running, restart the controller a bit. */
+ if (sc->bce_flags & BCE_MFW_ENABLE_FLAG) {
+ bce_reset(sc, BCE_DRV_MSG_CODE_RESET);
+ bce_chipinit(sc);
+ bce_mgmt_init_locked(sc);
+ }
+ }
+ }
+
+ BCE_UNLOCK(sc);
+ break;
+
+ /* Add/Delete multicast address */
+ case SIOCADDMULTI:
+ case SIOCDELMULTI:
+ DBPRINT(sc, BCE_VERBOSE_MISC,
+ "Received SIOCADDMULTI/SIOCDELMULTI\n");
+
+ BCE_LOCK(sc);
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING)
+ bce_set_rx_mode(sc);
+ BCE_UNLOCK(sc);
+
+ break;
+
+ /* Set/Get Interface media */
+ case SIOCSIFMEDIA:
+ case SIOCGIFMEDIA:
+ DBPRINT(sc, BCE_VERBOSE_MISC,
+ "Received SIOCSIFMEDIA/SIOCGIFMEDIA\n");
+
+ mii = device_get_softc(sc->bce_miibus);
+ error = ifmedia_ioctl(ifp, ifr,
+ &mii->mii_media, command);
+ break;
+
+ /* Set interface capability */
+ case SIOCSIFCAP:
+ mask = ifr->ifr_reqcap ^ ifp->if_capenable;
+ DBPRINT(sc, BCE_INFO_MISC,
+ "Received SIOCSIFCAP = 0x%08X\n", (u32) mask);
+
+ /* Toggle the TX checksum capabilities enable flag. */
+ if (mask & IFCAP_TXCSUM &&
+ ifp->if_capabilities & IFCAP_TXCSUM) {
+ ifp->if_capenable ^= IFCAP_TXCSUM;
+ if (IFCAP_TXCSUM & ifp->if_capenable)
+ ifp->if_hwassist |= BCE_IF_HWASSIST;
+ else
+ ifp->if_hwassist &= ~BCE_IF_HWASSIST;
+ }
+
+ /* Toggle the RX checksum capabilities enable flag. */
+ if (mask & IFCAP_RXCSUM &&
+ ifp->if_capabilities & IFCAP_RXCSUM)
+ ifp->if_capenable ^= IFCAP_RXCSUM;
+
+ /* Toggle the TSO capabilities enable flag. */
+ if (bce_tso_enable && (mask & IFCAP_TSO4) &&
+ ifp->if_capabilities & IFCAP_TSO4) {
+ ifp->if_capenable ^= IFCAP_TSO4;
+ if (IFCAP_TSO4 & ifp->if_capenable)
+ ifp->if_hwassist |= CSUM_TSO;
+ else
+ ifp->if_hwassist &= ~CSUM_TSO;
+ }
+
+ if (mask & IFCAP_VLAN_HWCSUM &&
+ ifp->if_capabilities & IFCAP_VLAN_HWCSUM)
+ ifp->if_capenable ^= IFCAP_VLAN_HWCSUM;
+
+ if ((mask & IFCAP_VLAN_HWTSO) != 0 &&
+ (ifp->if_capabilities & IFCAP_VLAN_HWTSO) != 0)
+ ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
+ /*
+ * Don't actually disable VLAN tag stripping as
+ * management firmware (ASF/IPMI/UMP) requires the
+ * feature. If VLAN tag stripping is disabled driver
+ * will manually reconstruct the VLAN frame by
+ * appending stripped VLAN tag.
+ */
+ if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
+ (ifp->if_capabilities & IFCAP_VLAN_HWTAGGING)) {
+ ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
+ if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
+ == 0)
+ ifp->if_capenable &= ~IFCAP_VLAN_HWTSO;
+ }
+ VLAN_CAPABILITIES(ifp);
+ break;
+ default:
+ /* We don't know how to handle the IOCTL, pass it on. */
+ error = ether_ioctl(ifp, command, data);
+ break;
+ }
+
+ DBEXIT(BCE_VERBOSE_MISC);
+ return(error);
+}
+
+
+/****************************************************************************/
+/* Transmit timeout handler. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_watchdog(struct bce_softc *sc)
+{
+ DBENTER(BCE_EXTREME_SEND);
+
+ BCE_LOCK_ASSERT(sc);
+
+ /* If the watchdog timer hasn't expired then just exit. */
+ if (sc->watchdog_timer == 0 || --sc->watchdog_timer)
+ goto bce_watchdog_exit;
+
+ /* If pause frames are active then don't reset the hardware. */
+ /* ToDo: Should we reset the timer here? */
+ if (REG_RD(sc, BCE_EMAC_TX_STATUS) & BCE_EMAC_TX_STATUS_XOFFED)
+ goto bce_watchdog_exit;
+
+ BCE_PRINTF("%s(%d): Watchdog timeout occurred, resetting!\n",
+ __FILE__, __LINE__);
+
+ DBRUNMSG(BCE_INFO,
+ bce_dump_driver_state(sc);
+ bce_dump_status_block(sc);
+ bce_dump_stats_block(sc);
+ bce_dump_ftqs(sc);
+ bce_dump_txp_state(sc, 0);
+ bce_dump_rxp_state(sc, 0);
+ bce_dump_tpat_state(sc, 0);
+ bce_dump_cp_state(sc, 0);
+ bce_dump_com_state(sc, 0));
+
+ DBRUN(bce_breakpoint(sc));
+
+ sc->bce_ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
+
+ bce_init_locked(sc);
+ sc->bce_ifp->if_oerrors++;
+
+bce_watchdog_exit:
+ DBEXIT(BCE_EXTREME_SEND);
+}
+
+
+/*
+ * Interrupt handler.
+ */
+/****************************************************************************/
+/* Main interrupt entry point. Verifies that the controller generated the */
+/* interrupt and then calls a separate routine for handle the various */
+/* interrupt causes (PHY, TX, RX). */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static void
+bce_intr(void *xsc)
+{
+ struct bce_softc *sc;
+ struct ifnet *ifp;
+ u32 status_attn_bits;
+ u16 hw_rx_cons, hw_tx_cons;
+
+ sc = xsc;
+ ifp = sc->bce_ifp;
+
+ DBENTER(BCE_VERBOSE_SEND | BCE_VERBOSE_RECV | BCE_VERBOSE_INTR);
+ DBRUNMSG(BCE_VERBOSE_INTR, bce_dump_status_block(sc));
+ DBRUNMSG(BCE_VERBOSE_INTR, bce_dump_stats_block(sc));
+
+ BCE_LOCK(sc);
+
+ DBRUN(sc->interrupts_generated++);
+
+ /* Synchnorize before we read from interface's status block */
+ bus_dmamap_sync(sc->status_tag, sc->status_map,
+ BUS_DMASYNC_POSTREAD);
+
+ /*
+ * If the hardware status block index
+ * matches the last value read by the
+ * driver and we haven't asserted our
+ * interrupt then there's nothing to do.
+ */
+ if ((sc->status_block->status_idx == sc->last_status_idx) &&
+ (REG_RD(sc, BCE_PCICFG_MISC_STATUS) &
+ BCE_PCICFG_MISC_STATUS_INTA_VALUE)) {
+ DBPRINT(sc, BCE_VERBOSE_INTR, "%s(): Spurious interrupt.\n",
+ __FUNCTION__);
+ goto bce_intr_exit;
+ }
+
+ /* Ack the interrupt and stop others from occuring. */
+ REG_WR(sc, BCE_PCICFG_INT_ACK_CMD,
+ BCE_PCICFG_INT_ACK_CMD_USE_INT_HC_PARAM |
+ BCE_PCICFG_INT_ACK_CMD_MASK_INT);
+
+ /* Check if the hardware has finished any work. */
+ hw_rx_cons = bce_get_hw_rx_cons(sc);
+ hw_tx_cons = bce_get_hw_tx_cons(sc);
+
+ /* Keep processing data as long as there is work to do. */
+ for (;;) {
+
+ status_attn_bits = sc->status_block->status_attn_bits;
+
+ DBRUNIF(DB_RANDOMTRUE(unexpected_attention_sim_control),
+ BCE_PRINTF("Simulating unexpected status attention "
+ "bit set.");
+ sc->unexpected_attention_sim_count++;
+ status_attn_bits = status_attn_bits |
+ STATUS_ATTN_BITS_PARITY_ERROR);
+
+ /* Was it a link change interrupt? */
+ if ((status_attn_bits & STATUS_ATTN_BITS_LINK_STATE) !=
+ (sc->status_block->status_attn_bits_ack &
+ STATUS_ATTN_BITS_LINK_STATE)) {
+ bce_phy_intr(sc);
+
+ /* Clear transient updates during link state change. */
+ REG_WR(sc, BCE_HC_COMMAND, sc->hc_command |
+ BCE_HC_COMMAND_COAL_NOW_WO_INT);
+ REG_RD(sc, BCE_HC_COMMAND);
+ }
+
+ /* If any other attention is asserted, the chip is toast. */
+ if (((status_attn_bits & ~STATUS_ATTN_BITS_LINK_STATE) !=
+ (sc->status_block->status_attn_bits_ack &
+ ~STATUS_ATTN_BITS_LINK_STATE))) {
+
+ sc->unexpected_attention_count++;
+
+ BCE_PRINTF("%s(%d): Fatal attention detected: "
+ "0x%08X\n", __FILE__, __LINE__,
+ sc->status_block->status_attn_bits);
+
+ DBRUNMSG(BCE_FATAL,
+ if (unexpected_attention_sim_control == 0)
+ bce_breakpoint(sc));
+
+ bce_init_locked(sc);
+ goto bce_intr_exit;
+ }
+
+ /* Check for any completed RX frames. */
+ if (hw_rx_cons != sc->hw_rx_cons)
+ bce_rx_intr(sc);
+
+ /* Check for any completed TX frames. */
+ if (hw_tx_cons != sc->hw_tx_cons)
+ bce_tx_intr(sc);
+
+ /* Save status block index value for the next interrupt. */
+ sc->last_status_idx = sc->status_block->status_idx;
+
+ /*
+ * Prevent speculative reads from getting
+ * ahead of the status block.
+ */
+ bus_space_barrier(sc->bce_btag, sc->bce_bhandle, 0, 0,
+ BUS_SPACE_BARRIER_READ);
+
+ /*
+ * If there's no work left then exit the
+ * interrupt service routine.
+ */
+ hw_rx_cons = bce_get_hw_rx_cons(sc);
+ hw_tx_cons = bce_get_hw_tx_cons(sc);
+
+ if ((hw_rx_cons == sc->hw_rx_cons) &&
+ (hw_tx_cons == sc->hw_tx_cons))
+ break;
+
+ }
+
+ bus_dmamap_sync(sc->status_tag, sc->status_map,
+ BUS_DMASYNC_PREREAD);
+
+ /* Re-enable interrupts. */
+ bce_enable_intr(sc, 0);
+
+ /* Handle any frames that arrived while handling the interrupt. */
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
+ !IFQ_DRV_IS_EMPTY(&ifp->if_snd))
+ bce_start_locked(ifp);
+
+bce_intr_exit:
+ BCE_UNLOCK(sc);
+
+ DBEXIT(BCE_VERBOSE_SEND | BCE_VERBOSE_RECV | BCE_VERBOSE_INTR);
+}
+
+
+/****************************************************************************/
+/* Programs the various packet receive modes (broadcast and multicast). */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_set_rx_mode(struct bce_softc *sc)
+{
+ struct ifnet *ifp;
+ struct ifmultiaddr *ifma;
+ u32 hashes[NUM_MC_HASH_REGISTERS] = { 0, 0, 0, 0, 0, 0, 0, 0 };
+ u32 rx_mode, sort_mode;
+ int h, i;
+
+ DBENTER(BCE_VERBOSE_MISC);
+
+ BCE_LOCK_ASSERT(sc);
+
+ ifp = sc->bce_ifp;
+
+ /* Initialize receive mode default settings. */
+ rx_mode = sc->rx_mode & ~(BCE_EMAC_RX_MODE_PROMISCUOUS |
+ BCE_EMAC_RX_MODE_KEEP_VLAN_TAG);
+ sort_mode = 1 | BCE_RPM_SORT_USER0_BC_EN;
+
+ /*
+ * ASF/IPMI/UMP firmware requires that VLAN tag stripping
+ * be enbled.
+ */
+ if (!(BCE_IF_CAPABILITIES & IFCAP_VLAN_HWTAGGING) &&
+ (!(sc->bce_flags & BCE_MFW_ENABLE_FLAG)))
+ rx_mode |= BCE_EMAC_RX_MODE_KEEP_VLAN_TAG;
+
+ /*
+ * Check for promiscuous, all multicast, or selected
+ * multicast address filtering.
+ */
+ if (ifp->if_flags & IFF_PROMISC) {
+ DBPRINT(sc, BCE_INFO_MISC, "Enabling promiscuous mode.\n");
+
+ /* Enable promiscuous mode. */
+ rx_mode |= BCE_EMAC_RX_MODE_PROMISCUOUS;
+ sort_mode |= BCE_RPM_SORT_USER0_PROM_EN;
+ } else if (ifp->if_flags & IFF_ALLMULTI) {
+ DBPRINT(sc, BCE_INFO_MISC, "Enabling all multicast mode.\n");
+
+ /* Enable all multicast addresses. */
+ for (i = 0; i < NUM_MC_HASH_REGISTERS; i++) {
+ REG_WR(sc, BCE_EMAC_MULTICAST_HASH0 + (i * 4), 0xffffffff);
+ }
+ sort_mode |= BCE_RPM_SORT_USER0_MC_EN;
+ } else {
+ /* Accept one or more multicast(s). */
+ DBPRINT(sc, BCE_INFO_MISC, "Enabling selective multicast mode.\n");
+
+ if_maddr_rlock(ifp);
+ TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
+ if (ifma->ifma_addr->sa_family != AF_LINK)
+ continue;
+ h = ether_crc32_le(LLADDR((struct sockaddr_dl *)
+ ifma->ifma_addr), ETHER_ADDR_LEN) & 0xFF;
+ hashes[(h & 0xE0) >> 5] |= 1 << (h & 0x1F);
+ }
+ if_maddr_runlock(ifp);
+
+ for (i = 0; i < NUM_MC_HASH_REGISTERS; i++)
+ REG_WR(sc, BCE_EMAC_MULTICAST_HASH0 + (i * 4), hashes[i]);
+
+ sort_mode |= BCE_RPM_SORT_USER0_MC_HSH_EN;
+ }
+
+ /* Only make changes if the recive mode has actually changed. */
+ if (rx_mode != sc->rx_mode) {
+ DBPRINT(sc, BCE_VERBOSE_MISC, "Enabling new receive mode: "
+ "0x%08X\n", rx_mode);
+
+ sc->rx_mode = rx_mode;
+ REG_WR(sc, BCE_EMAC_RX_MODE, rx_mode);
+ }
+
+ /* Disable and clear the exisitng sort before enabling a new sort. */
+ REG_WR(sc, BCE_RPM_SORT_USER0, 0x0);
+ REG_WR(sc, BCE_RPM_SORT_USER0, sort_mode);
+ REG_WR(sc, BCE_RPM_SORT_USER0, sort_mode | BCE_RPM_SORT_USER0_ENA);
+
+ DBEXIT(BCE_VERBOSE_MISC);
+}
+
+
+/****************************************************************************/
+/* Called periodically to updates statistics from the controllers */
+/* statistics block. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_stats_update(struct bce_softc *sc)
+{
+ struct ifnet *ifp;
+ struct statistics_block *stats;
+
+ DBENTER(BCE_EXTREME_MISC);
+
+ ifp = sc->bce_ifp;
+
+ stats = (struct statistics_block *) sc->stats_block;
+
+ /*
+ * Certain controllers don't report
+ * carrier sense errors correctly.
+ * See errata E11_5708CA0_1165.
+ */
+ if (!(BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5706) &&
+ !(BCE_CHIP_ID(sc) == BCE_CHIP_ID_5708_A0))
+ ifp->if_oerrors +=
+ (u_long) stats->stat_Dot3StatsCarrierSenseErrors;
+
+ /*
+ * Update the sysctl statistics from the
+ * hardware statistics.
+ */
+ sc->stat_IfHCInOctets =
+ ((u64) stats->stat_IfHCInOctets_hi << 32) +
+ (u64) stats->stat_IfHCInOctets_lo;
+
+ sc->stat_IfHCInBadOctets =
+ ((u64) stats->stat_IfHCInBadOctets_hi << 32) +
+ (u64) stats->stat_IfHCInBadOctets_lo;
+
+ sc->stat_IfHCOutOctets =
+ ((u64) stats->stat_IfHCOutOctets_hi << 32) +
+ (u64) stats->stat_IfHCOutOctets_lo;
+
+ sc->stat_IfHCOutBadOctets =
+ ((u64) stats->stat_IfHCOutBadOctets_hi << 32) +
+ (u64) stats->stat_IfHCOutBadOctets_lo;
+
+ sc->stat_IfHCInUcastPkts =
+ ((u64) stats->stat_IfHCInUcastPkts_hi << 32) +
+ (u64) stats->stat_IfHCInUcastPkts_lo;
+
+ sc->stat_IfHCInMulticastPkts =
+ ((u64) stats->stat_IfHCInMulticastPkts_hi << 32) +
+ (u64) stats->stat_IfHCInMulticastPkts_lo;
+
+ sc->stat_IfHCInBroadcastPkts =
+ ((u64) stats->stat_IfHCInBroadcastPkts_hi << 32) +
+ (u64) stats->stat_IfHCInBroadcastPkts_lo;
+
+ sc->stat_IfHCOutUcastPkts =
+ ((u64) stats->stat_IfHCOutUcastPkts_hi << 32) +
+ (u64) stats->stat_IfHCOutUcastPkts_lo;
+
+ sc->stat_IfHCOutMulticastPkts =
+ ((u64) stats->stat_IfHCOutMulticastPkts_hi << 32) +
+ (u64) stats->stat_IfHCOutMulticastPkts_lo;
+
+ sc->stat_IfHCOutBroadcastPkts =
+ ((u64) stats->stat_IfHCOutBroadcastPkts_hi << 32) +
+ (u64) stats->stat_IfHCOutBroadcastPkts_lo;
+
+ /* ToDo: Preserve counters beyond 32 bits? */
+ /* ToDo: Read the statistics from auto-clear regs? */
+
+ sc->stat_emac_tx_stat_dot3statsinternalmactransmiterrors =
+ stats->stat_emac_tx_stat_dot3statsinternalmactransmiterrors;
+
+ sc->stat_Dot3StatsCarrierSenseErrors =
+ stats->stat_Dot3StatsCarrierSenseErrors;
+
+ sc->stat_Dot3StatsFCSErrors =
+ stats->stat_Dot3StatsFCSErrors;
+
+ sc->stat_Dot3StatsAlignmentErrors =
+ stats->stat_Dot3StatsAlignmentErrors;
+
+ sc->stat_Dot3StatsSingleCollisionFrames =
+ stats->stat_Dot3StatsSingleCollisionFrames;
+
+ sc->stat_Dot3StatsMultipleCollisionFrames =
+ stats->stat_Dot3StatsMultipleCollisionFrames;
+
+ sc->stat_Dot3StatsDeferredTransmissions =
+ stats->stat_Dot3StatsDeferredTransmissions;
+
+ sc->stat_Dot3StatsExcessiveCollisions =
+ stats->stat_Dot3StatsExcessiveCollisions;
+
+ sc->stat_Dot3StatsLateCollisions =
+ stats->stat_Dot3StatsLateCollisions;
+
+ sc->stat_EtherStatsCollisions =
+ stats->stat_EtherStatsCollisions;
+
+ sc->stat_EtherStatsFragments =
+ stats->stat_EtherStatsFragments;
+
+ sc->stat_EtherStatsJabbers =
+ stats->stat_EtherStatsJabbers;
+
+ sc->stat_EtherStatsUndersizePkts =
+ stats->stat_EtherStatsUndersizePkts;
+
+ sc->stat_EtherStatsOversizePkts =
+ stats->stat_EtherStatsOversizePkts;
+
+ sc->stat_EtherStatsPktsRx64Octets =
+ stats->stat_EtherStatsPktsRx64Octets;
+
+ sc->stat_EtherStatsPktsRx65Octetsto127Octets =
+ stats->stat_EtherStatsPktsRx65Octetsto127Octets;
+
+ sc->stat_EtherStatsPktsRx128Octetsto255Octets =
+ stats->stat_EtherStatsPktsRx128Octetsto255Octets;
+
+ sc->stat_EtherStatsPktsRx256Octetsto511Octets =
+ stats->stat_EtherStatsPktsRx256Octetsto511Octets;
+
+ sc->stat_EtherStatsPktsRx512Octetsto1023Octets =
+ stats->stat_EtherStatsPktsRx512Octetsto1023Octets;
+
+ sc->stat_EtherStatsPktsRx1024Octetsto1522Octets =
+ stats->stat_EtherStatsPktsRx1024Octetsto1522Octets;
+
+ sc->stat_EtherStatsPktsRx1523Octetsto9022Octets =
+ stats->stat_EtherStatsPktsRx1523Octetsto9022Octets;
+
+ sc->stat_EtherStatsPktsTx64Octets =
+ stats->stat_EtherStatsPktsTx64Octets;
+
+ sc->stat_EtherStatsPktsTx65Octetsto127Octets =
+ stats->stat_EtherStatsPktsTx65Octetsto127Octets;
+
+ sc->stat_EtherStatsPktsTx128Octetsto255Octets =
+ stats->stat_EtherStatsPktsTx128Octetsto255Octets;
+
+ sc->stat_EtherStatsPktsTx256Octetsto511Octets =
+ stats->stat_EtherStatsPktsTx256Octetsto511Octets;
+
+ sc->stat_EtherStatsPktsTx512Octetsto1023Octets =
+ stats->stat_EtherStatsPktsTx512Octetsto1023Octets;
+
+ sc->stat_EtherStatsPktsTx1024Octetsto1522Octets =
+ stats->stat_EtherStatsPktsTx1024Octetsto1522Octets;
+
+ sc->stat_EtherStatsPktsTx1523Octetsto9022Octets =
+ stats->stat_EtherStatsPktsTx1523Octetsto9022Octets;
+
+ sc->stat_XonPauseFramesReceived =
+ stats->stat_XonPauseFramesReceived;
+
+ sc->stat_XoffPauseFramesReceived =
+ stats->stat_XoffPauseFramesReceived;
+
+ sc->stat_OutXonSent =
+ stats->stat_OutXonSent;
+
+ sc->stat_OutXoffSent =
+ stats->stat_OutXoffSent;
+
+ sc->stat_FlowControlDone =
+ stats->stat_FlowControlDone;
+
+ sc->stat_MacControlFramesReceived =
+ stats->stat_MacControlFramesReceived;
+
+ sc->stat_XoffStateEntered =
+ stats->stat_XoffStateEntered;
+
+ sc->stat_IfInFramesL2FilterDiscards =
+ stats->stat_IfInFramesL2FilterDiscards;
+
+ sc->stat_IfInRuleCheckerDiscards =
+ stats->stat_IfInRuleCheckerDiscards;
+
+ sc->stat_IfInFTQDiscards =
+ stats->stat_IfInFTQDiscards;
+
+ sc->stat_IfInMBUFDiscards =
+ stats->stat_IfInMBUFDiscards;
+
+ sc->stat_IfInRuleCheckerP4Hit =
+ stats->stat_IfInRuleCheckerP4Hit;
+
+ sc->stat_CatchupInRuleCheckerDiscards =
+ stats->stat_CatchupInRuleCheckerDiscards;
+
+ sc->stat_CatchupInFTQDiscards =
+ stats->stat_CatchupInFTQDiscards;
+
+ sc->stat_CatchupInMBUFDiscards =
+ stats->stat_CatchupInMBUFDiscards;
+
+ sc->stat_CatchupInRuleCheckerP4Hit =
+ stats->stat_CatchupInRuleCheckerP4Hit;
+
+ sc->com_no_buffers = REG_RD_IND(sc, 0x120084);
+
+ /*
+ * Update the interface statistics from the
+ * hardware statistics.
+ */
+ ifp->if_collisions =
+ (u_long) sc->stat_EtherStatsCollisions;
+
+ /* ToDo: This method loses soft errors. */
+ ifp->if_ierrors =
+ (u_long) sc->stat_EtherStatsUndersizePkts +
+ (u_long) sc->stat_EtherStatsOversizePkts +
+ (u_long) sc->stat_IfInMBUFDiscards +
+ (u_long) sc->stat_Dot3StatsAlignmentErrors +
+ (u_long) sc->stat_Dot3StatsFCSErrors +
+ (u_long) sc->stat_IfInRuleCheckerDiscards +
+ (u_long) sc->stat_IfInFTQDiscards +
+ (u_long) sc->com_no_buffers;
+
+ /* ToDo: This method loses soft errors. */
+ ifp->if_oerrors =
+ (u_long) sc->stat_emac_tx_stat_dot3statsinternalmactransmiterrors +
+ (u_long) sc->stat_Dot3StatsExcessiveCollisions +
+ (u_long) sc->stat_Dot3StatsLateCollisions;
+
+ /* ToDo: Add additional statistics? */
+
+ DBEXIT(BCE_EXTREME_MISC);
+}
+
+
+/****************************************************************************/
+/* Periodic function to notify the bootcode that the driver is still */
+/* present. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_pulse(void *xsc)
+{
+ struct bce_softc *sc = xsc;
+ u32 msg;
+
+ DBENTER(BCE_EXTREME_MISC);
+
+ BCE_LOCK_ASSERT(sc);
+
+ /* Tell the firmware that the driver is still running. */
+ msg = (u32) ++sc->bce_fw_drv_pulse_wr_seq;
+ bce_shmem_wr(sc, BCE_DRV_PULSE_MB, msg);
+
+ /* Update the bootcode condition. */
+ sc->bc_state = bce_shmem_rd(sc, BCE_BC_STATE_CONDITION);
+
+ /* Report whether the bootcode still knows the driver is running. */
+ if (bootverbose) {
+ if (sc->bce_drv_cardiac_arrest == FALSE) {
+ if (!(sc->bc_state & BCE_CONDITION_DRV_PRESENT)) {
+ sc->bce_drv_cardiac_arrest = TRUE;
+ BCE_PRINTF("%s(): Warning: bootcode "
+ "thinks driver is absent! "
+ "(bc_state = 0x%08X)\n",
+ __FUNCTION__, sc->bc_state);
+ }
+ } else {
+ /*
+ * Not supported by all bootcode versions.
+ * (v5.0.11+ and v5.2.1+) Older bootcode
+ * will require the driver to reset the
+ * controller to clear this condition.
+ */
+ if (sc->bc_state & BCE_CONDITION_DRV_PRESENT) {
+ sc->bce_drv_cardiac_arrest = FALSE;
+ BCE_PRINTF("%s(): Bootcode found the "
+ "driver pulse! (bc_state = 0x%08X)\n",
+ __FUNCTION__, sc->bc_state);
+ }
+ }
+ }
+
+
+ /* Schedule the next pulse. */
+ callout_reset(&sc->bce_pulse_callout, hz, bce_pulse, sc);
+
+ DBEXIT(BCE_EXTREME_MISC);
+}
+
+
+/****************************************************************************/
+/* Periodic function to perform maintenance tasks. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static void
+bce_tick(void *xsc)
+{
+ struct bce_softc *sc = xsc;
+ struct mii_data *mii;
+ struct ifnet *ifp;
+
+ ifp = sc->bce_ifp;
+
+ DBENTER(BCE_EXTREME_MISC);
+
+ BCE_LOCK_ASSERT(sc);
+
+ /* Schedule the next tick. */
+ callout_reset(&sc->bce_tick_callout, hz, bce_tick, sc);
+
+ /* Update the statistics from the hardware statistics block. */
+ bce_stats_update(sc);
+
+ /* Top off the receive and page chains. */
+#ifdef BCE_JUMBO_HDRSPLIT
+ bce_fill_pg_chain(sc);
+#endif
+ bce_fill_rx_chain(sc);
+
+ /* Check that chip hasn't hung. */
+ bce_watchdog(sc);
+
+ /* If link is up already up then we're done. */
+ if (sc->bce_link_up == TRUE)
+ goto bce_tick_exit;
+
+ /* Link is down. Check what the PHY's doing. */
+ mii = device_get_softc(sc->bce_miibus);
+ mii_tick(mii);
+
+ /* Check if the link has come up. */
+ if ((mii->mii_media_status & IFM_ACTIVE) &&
+ (IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE)) {
+ DBPRINT(sc, BCE_VERBOSE_MISC,
+ "%s(): Link up!\n", __FUNCTION__);
+ sc->bce_link_up = TRUE;
+ if ((IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_T ||
+ IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_SX ||
+ IFM_SUBTYPE(mii->mii_media_active) == IFM_2500_SX) &&
+ bootverbose)
+ BCE_PRINTF("Gigabit link up!\n");
+
+ /* Now that link is up, handle any outstanding TX traffic. */
+ if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
+ DBPRINT(sc, BCE_VERBOSE_MISC, "%s(): Found "
+ "pending TX traffic.\n", __FUNCTION__);
+ bce_start_locked(ifp);
+ }
+ }
+
+bce_tick_exit:
+ DBEXIT(BCE_EXTREME_MISC);
+ return;
+}
+
+
+#ifdef BCE_DEBUG
+/****************************************************************************/
+/* Allows the driver state to be dumped through the sysctl interface. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_driver_state(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_driver_state(sc);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Allows the hardware state to be dumped through the sysctl interface. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_hw_state(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_hw_state(sc);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Allows the status block to be dumped through the sysctl interface. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_status_block(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_status_block(sc);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Allows the stats block to be dumped through the sysctl interface. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_stats_block(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_stats_block(sc);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Allows the stat counters to be cleared without unloading/reloading the */
+/* driver. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_stats_clear(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+
+ /* Clear the internal H/W statistics counters. */
+ REG_WR(sc, BCE_HC_COMMAND, BCE_HC_COMMAND_CLR_STAT_NOW);
+
+ /* Reset the driver maintained statistics. */
+ sc->interrupts_rx =
+ sc->interrupts_tx = 0;
+ sc->tso_frames_requested =
+ sc->tso_frames_completed =
+ sc->tso_frames_failed = 0;
+ sc->rx_empty_count =
+ sc->tx_full_count = 0;
+ sc->rx_low_watermark = USABLE_RX_BD;
+ sc->tx_hi_watermark = 0;
+ sc->l2fhdr_error_count =
+ sc->l2fhdr_error_sim_count = 0;
+ sc->mbuf_alloc_failed_count =
+ sc->mbuf_alloc_failed_sim_count = 0;
+ sc->dma_map_addr_rx_failed_count =
+ sc->dma_map_addr_tx_failed_count = 0;
+ sc->mbuf_frag_count = 0;
+ sc->csum_offload_tcp_udp =
+ sc->csum_offload_ip = 0;
+ sc->vlan_tagged_frames_rcvd =
+ sc->vlan_tagged_frames_stripped = 0;
+
+ /* Clear firmware maintained statistics. */
+ REG_WR_IND(sc, 0x120084, 0);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Allows the bootcode state to be dumped through the sysctl interface. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_bc_state(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_bc_state(sc);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow dumping the RX BD chain. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_dump_rx_bd_chain(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_rx_bd_chain(sc, 0, TOTAL_RX_BD);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow dumping the RX MBUF chain. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_dump_rx_mbuf_chain(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_rx_mbuf_chain(sc, 0, USABLE_RX_BD);
+ }
+
+ return error;
+}
+
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow dumping the TX chain. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_dump_tx_chain(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_tx_chain(sc, 0, TOTAL_TX_BD);
+ }
+
+ return error;
+}
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+/****************************************************************************/
+/* Provides a sysctl interface to allow dumping the page chain. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_dump_pg_chain(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_pg_chain(sc, 0, TOTAL_PG_BD);
+ }
+
+ return error;
+}
+#endif
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow reading arbitrary NVRAM offsets in */
+/* the device. DO NOT ENABLE ON PRODUCTION SYSTEMS! */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_nvram_read(SYSCTL_HANDLER_ARGS)
+{
+ struct bce_softc *sc = (struct bce_softc *)arg1;
+ int error;
+ u32 result;
+ u32 val[1];
+ u8 *data = (u8 *) val;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+ if (error || (req->newptr == NULL))
+ return (error);
+
+ bce_nvram_read(sc, result, data, 4);
+ BCE_PRINTF("offset 0x%08X = 0x%08X\n", result, bce_be32toh(val[0]));
+
+ return (error);
+}
+
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow reading arbitrary registers in the */
+/* device. DO NOT ENABLE ON PRODUCTION SYSTEMS! */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_reg_read(SYSCTL_HANDLER_ARGS)
+{
+ struct bce_softc *sc = (struct bce_softc *)arg1;
+ int error;
+ u32 val, result;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+ if (error || (req->newptr == NULL))
+ return (error);
+
+ /* Make sure the register is accessible. */
+ if (result < 0x8000) {
+ val = REG_RD(sc, result);
+ BCE_PRINTF("reg 0x%08X = 0x%08X\n", result, val);
+ } else if (result < 0x0280000) {
+ val = REG_RD_IND(sc, result);
+ BCE_PRINTF("reg 0x%08X = 0x%08X\n", result, val);
+ }
+
+ return (error);
+}
+
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow reading arbitrary PHY registers in */
+/* the device. DO NOT ENABLE ON PRODUCTION SYSTEMS! */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_phy_read(SYSCTL_HANDLER_ARGS)
+{
+ struct bce_softc *sc;
+ device_t dev;
+ int error, result;
+ u16 val;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+ if (error || (req->newptr == NULL))
+ return (error);
+
+ /* Make sure the register is accessible. */
+ if (result < 0x20) {
+ sc = (struct bce_softc *)arg1;
+ dev = sc->bce_dev;
+ val = bce_miibus_read_reg(dev, sc->bce_phy_addr, result);
+ BCE_PRINTF("phy 0x%02X = 0x%04X\n", result, val);
+ }
+ return (error);
+}
+
+
+/****************************************************************************/
+/* Provides a sysctl interface to allow reading a CID. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_dump_ctx(SYSCTL_HANDLER_ARGS)
+{
+ struct bce_softc *sc;
+ int error, result;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+ if (error || (req->newptr == NULL))
+ return (error);
+
+ /* Make sure the register is accessible. */
+ if (result <= TX_CID) {
+ sc = (struct bce_softc *)arg1;
+ bce_dump_ctx(sc, result);
+ }
+
+ return (error);
+}
+
+
+ /****************************************************************************/
+/* Provides a sysctl interface to forcing the driver to dump state and */
+/* enter the debugger. DO NOT ENABLE ON PRODUCTION SYSTEMS! */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static int
+bce_sysctl_breakpoint(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int result;
+ struct bce_softc *sc;
+
+ result = -1;
+ error = sysctl_handle_int(oidp, &result, 0, req);
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (result == 1) {
+ sc = (struct bce_softc *)arg1;
+ bce_breakpoint(sc);
+ }
+
+ return error;
+}
+#endif
+
+
+/****************************************************************************/
+/* Adds any sysctl parameters for tuning or debugging purposes. */
+/* */
+/* Returns: */
+/* 0 for success, positive value for failure. */
+/****************************************************************************/
+static void
+bce_add_sysctls(struct bce_softc *sc)
+{
+ struct sysctl_ctx_list *ctx;
+ struct sysctl_oid_list *children;
+
+ DBENTER(BCE_VERBOSE_MISC);
+
+ ctx = device_get_sysctl_ctx(sc->bce_dev);
+ children = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->bce_dev));
+
+#ifdef BCE_DEBUG
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "l2fhdr_error_sim_control",
+ CTLFLAG_RW, &l2fhdr_error_sim_control,
+ 0, "Debug control to force l2fhdr errors");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "l2fhdr_error_sim_count",
+ CTLFLAG_RD, &sc->l2fhdr_error_sim_count,
+ 0, "Number of simulated l2_fhdr errors");
+#endif
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "l2fhdr_error_count",
+ CTLFLAG_RD, &sc->l2fhdr_error_count,
+ 0, "Number of l2_fhdr errors");
+
+#ifdef BCE_DEBUG
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "mbuf_alloc_failed_sim_control",
+ CTLFLAG_RW, &mbuf_alloc_failed_sim_control,
+ 0, "Debug control to force mbuf allocation failures");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "mbuf_alloc_failed_sim_count",
+ CTLFLAG_RD, &sc->mbuf_alloc_failed_sim_count,
+ 0, "Number of simulated mbuf cluster allocation failures");
+#endif
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "mbuf_alloc_failed_count",
+ CTLFLAG_RD, &sc->mbuf_alloc_failed_count,
+ 0, "Number of mbuf allocation failures");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "mbuf_frag_count",
+ CTLFLAG_RD, &sc->mbuf_frag_count,
+ 0, "Number of fragmented mbufs");
+
+#ifdef BCE_DEBUG
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "dma_map_addr_failed_sim_control",
+ CTLFLAG_RW, &dma_map_addr_failed_sim_control,
+ 0, "Debug control to force DMA mapping failures");
+
+ /* ToDo: Figure out how to update this value in bce_dma_map_addr(). */
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "dma_map_addr_failed_sim_count",
+ CTLFLAG_RD, &sc->dma_map_addr_failed_sim_count,
+ 0, "Number of simulated DMA mapping failures");
+
+#endif
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "dma_map_addr_rx_failed_count",
+ CTLFLAG_RD, &sc->dma_map_addr_rx_failed_count,
+ 0, "Number of RX DMA mapping failures");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "dma_map_addr_tx_failed_count",
+ CTLFLAG_RD, &sc->dma_map_addr_tx_failed_count,
+ 0, "Number of TX DMA mapping failures");
+
+#ifdef BCE_DEBUG
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "unexpected_attention_sim_control",
+ CTLFLAG_RW, &unexpected_attention_sim_control,
+ 0, "Debug control to simulate unexpected attentions");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "unexpected_attention_sim_count",
+ CTLFLAG_RW, &sc->unexpected_attention_sim_count,
+ 0, "Number of simulated unexpected attentions");
+#endif
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "unexpected_attention_count",
+ CTLFLAG_RW, &sc->unexpected_attention_count,
+ 0, "Number of unexpected attentions");
+
+#ifdef BCE_DEBUG
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "debug_bootcode_running_failure",
+ CTLFLAG_RW, &bootcode_running_failure_sim_control,
+ 0, "Debug control to force bootcode running failures");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "rx_low_watermark",
+ CTLFLAG_RD, &sc->rx_low_watermark,
+ 0, "Lowest level of free rx_bd's");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "rx_empty_count",
+ CTLFLAG_RD, &sc->rx_empty_count,
+ 0, "Number of times the RX chain was empty");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "tx_hi_watermark",
+ CTLFLAG_RD, &sc->tx_hi_watermark,
+ 0, "Highest level of used tx_bd's");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "tx_full_count",
+ CTLFLAG_RD, &sc->tx_full_count,
+ 0, "Number of times the TX chain was full");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "tso_frames_requested",
+ CTLFLAG_RD, &sc->tso_frames_requested,
+ 0, "Number of TSO frames requested");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "tso_frames_completed",
+ CTLFLAG_RD, &sc->tso_frames_completed,
+ 0, "Number of TSO frames completed");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "tso_frames_failed",
+ CTLFLAG_RD, &sc->tso_frames_failed,
+ 0, "Number of TSO frames failed");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "csum_offload_ip",
+ CTLFLAG_RD, &sc->csum_offload_ip,
+ 0, "Number of IP checksum offload frames");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "csum_offload_tcp_udp",
+ CTLFLAG_RD, &sc->csum_offload_tcp_udp,
+ 0, "Number of TCP/UDP checksum offload frames");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "vlan_tagged_frames_rcvd",
+ CTLFLAG_RD, &sc->vlan_tagged_frames_rcvd,
+ 0, "Number of VLAN tagged frames received");
+
+ SYSCTL_ADD_INT(ctx, children, OID_AUTO,
+ "vlan_tagged_frames_stripped",
+ CTLFLAG_RD, &sc->vlan_tagged_frames_stripped,
+ 0, "Number of VLAN tagged frames stripped");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "interrupts_rx",
+ CTLFLAG_RD, &sc->interrupts_rx,
+ 0, "Number of RX interrupts");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "interrupts_tx",
+ CTLFLAG_RD, &sc->interrupts_tx,
+ 0, "Number of TX interrupts");
+#endif
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHcInOctets",
+ CTLFLAG_RD, &sc->stat_IfHCInOctets,
+ "Bytes received");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCInBadOctets",
+ CTLFLAG_RD, &sc->stat_IfHCInBadOctets,
+ "Bad bytes received");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCOutOctets",
+ CTLFLAG_RD, &sc->stat_IfHCOutOctets,
+ "Bytes sent");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCOutBadOctets",
+ CTLFLAG_RD, &sc->stat_IfHCOutBadOctets,
+ "Bad bytes sent");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCInUcastPkts",
+ CTLFLAG_RD, &sc->stat_IfHCInUcastPkts,
+ "Unicast packets received");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCInMulticastPkts",
+ CTLFLAG_RD, &sc->stat_IfHCInMulticastPkts,
+ "Multicast packets received");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCInBroadcastPkts",
+ CTLFLAG_RD, &sc->stat_IfHCInBroadcastPkts,
+ "Broadcast packets received");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCOutUcastPkts",
+ CTLFLAG_RD, &sc->stat_IfHCOutUcastPkts,
+ "Unicast packets sent");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCOutMulticastPkts",
+ CTLFLAG_RD, &sc->stat_IfHCOutMulticastPkts,
+ "Multicast packets sent");
+
+ SYSCTL_ADD_ULONG(ctx, children, OID_AUTO,
+ "stat_IfHCOutBroadcastPkts",
+ CTLFLAG_RD, &sc->stat_IfHCOutBroadcastPkts,
+ "Broadcast packets sent");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_emac_tx_stat_dot3statsinternalmactransmiterrors",
+ CTLFLAG_RD, &sc->stat_emac_tx_stat_dot3statsinternalmactransmiterrors,
+ 0, "Internal MAC transmit errors");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsCarrierSenseErrors",
+ CTLFLAG_RD, &sc->stat_Dot3StatsCarrierSenseErrors,
+ 0, "Carrier sense errors");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsFCSErrors",
+ CTLFLAG_RD, &sc->stat_Dot3StatsFCSErrors,
+ 0, "Frame check sequence errors");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsAlignmentErrors",
+ CTLFLAG_RD, &sc->stat_Dot3StatsAlignmentErrors,
+ 0, "Alignment errors");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsSingleCollisionFrames",
+ CTLFLAG_RD, &sc->stat_Dot3StatsSingleCollisionFrames,
+ 0, "Single Collision Frames");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsMultipleCollisionFrames",
+ CTLFLAG_RD, &sc->stat_Dot3StatsMultipleCollisionFrames,
+ 0, "Multiple Collision Frames");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsDeferredTransmissions",
+ CTLFLAG_RD, &sc->stat_Dot3StatsDeferredTransmissions,
+ 0, "Deferred Transmissions");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsExcessiveCollisions",
+ CTLFLAG_RD, &sc->stat_Dot3StatsExcessiveCollisions,
+ 0, "Excessive Collisions");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_Dot3StatsLateCollisions",
+ CTLFLAG_RD, &sc->stat_Dot3StatsLateCollisions,
+ 0, "Late Collisions");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsCollisions",
+ CTLFLAG_RD, &sc->stat_EtherStatsCollisions,
+ 0, "Collisions");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsFragments",
+ CTLFLAG_RD, &sc->stat_EtherStatsFragments,
+ 0, "Fragments");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsJabbers",
+ CTLFLAG_RD, &sc->stat_EtherStatsJabbers,
+ 0, "Jabbers");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsUndersizePkts",
+ CTLFLAG_RD, &sc->stat_EtherStatsUndersizePkts,
+ 0, "Undersize packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsOversizePkts",
+ CTLFLAG_RD, &sc->stat_EtherStatsOversizePkts,
+ 0, "stat_EtherStatsOversizePkts");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx64Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx64Octets,
+ 0, "Bytes received in 64 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx65Octetsto127Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx65Octetsto127Octets,
+ 0, "Bytes received in 65 to 127 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx128Octetsto255Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx128Octetsto255Octets,
+ 0, "Bytes received in 128 to 255 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx256Octetsto511Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx256Octetsto511Octets,
+ 0, "Bytes received in 256 to 511 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx512Octetsto1023Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx512Octetsto1023Octets,
+ 0, "Bytes received in 512 to 1023 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx1024Octetsto1522Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx1024Octetsto1522Octets,
+ 0, "Bytes received in 1024 t0 1522 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsRx1523Octetsto9022Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsRx1523Octetsto9022Octets,
+ 0, "Bytes received in 1523 to 9022 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx64Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx64Octets,
+ 0, "Bytes sent in 64 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx65Octetsto127Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx65Octetsto127Octets,
+ 0, "Bytes sent in 65 to 127 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx128Octetsto255Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx128Octetsto255Octets,
+ 0, "Bytes sent in 128 to 255 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx256Octetsto511Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx256Octetsto511Octets,
+ 0, "Bytes sent in 256 to 511 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx512Octetsto1023Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx512Octetsto1023Octets,
+ 0, "Bytes sent in 512 to 1023 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx1024Octetsto1522Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx1024Octetsto1522Octets,
+ 0, "Bytes sent in 1024 to 1522 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_EtherStatsPktsTx1523Octetsto9022Octets",
+ CTLFLAG_RD, &sc->stat_EtherStatsPktsTx1523Octetsto9022Octets,
+ 0, "Bytes sent in 1523 to 9022 byte packets");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_XonPauseFramesReceived",
+ CTLFLAG_RD, &sc->stat_XonPauseFramesReceived,
+ 0, "XON pause frames receved");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_XoffPauseFramesReceived",
+ CTLFLAG_RD, &sc->stat_XoffPauseFramesReceived,
+ 0, "XOFF pause frames received");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_OutXonSent",
+ CTLFLAG_RD, &sc->stat_OutXonSent,
+ 0, "XON pause frames sent");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_OutXoffSent",
+ CTLFLAG_RD, &sc->stat_OutXoffSent,
+ 0, "XOFF pause frames sent");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_FlowControlDone",
+ CTLFLAG_RD, &sc->stat_FlowControlDone,
+ 0, "Flow control done");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_MacControlFramesReceived",
+ CTLFLAG_RD, &sc->stat_MacControlFramesReceived,
+ 0, "MAC control frames received");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_XoffStateEntered",
+ CTLFLAG_RD, &sc->stat_XoffStateEntered,
+ 0, "XOFF state entered");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_IfInFramesL2FilterDiscards",
+ CTLFLAG_RD, &sc->stat_IfInFramesL2FilterDiscards,
+ 0, "Received L2 packets discarded");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_IfInRuleCheckerDiscards",
+ CTLFLAG_RD, &sc->stat_IfInRuleCheckerDiscards,
+ 0, "Received packets discarded by rule");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_IfInFTQDiscards",
+ CTLFLAG_RD, &sc->stat_IfInFTQDiscards,
+ 0, "Received packet FTQ discards");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_IfInMBUFDiscards",
+ CTLFLAG_RD, &sc->stat_IfInMBUFDiscards,
+ 0, "Received packets discarded due to lack "
+ "of controller buffer memory");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_IfInRuleCheckerP4Hit",
+ CTLFLAG_RD, &sc->stat_IfInRuleCheckerP4Hit,
+ 0, "Received packets rule checker hits");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_CatchupInRuleCheckerDiscards",
+ CTLFLAG_RD, &sc->stat_CatchupInRuleCheckerDiscards,
+ 0, "Received packets discarded in Catchup path");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_CatchupInFTQDiscards",
+ CTLFLAG_RD, &sc->stat_CatchupInFTQDiscards,
+ 0, "Received packets discarded in FTQ in Catchup path");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_CatchupInMBUFDiscards",
+ CTLFLAG_RD, &sc->stat_CatchupInMBUFDiscards,
+ 0, "Received packets discarded in controller "
+ "buffer memory in Catchup path");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "stat_CatchupInRuleCheckerP4Hit",
+ CTLFLAG_RD, &sc->stat_CatchupInRuleCheckerP4Hit,
+ 0, "Received packets rule checker hits in Catchup path");
+
+ SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
+ "com_no_buffers",
+ CTLFLAG_RD, &sc->com_no_buffers,
+ 0, "Valid packets received but no RX buffers available");
+
+#ifdef BCE_DEBUG
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "driver_state", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_driver_state, "I", "Drive state information");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "hw_state", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_hw_state, "I", "Hardware state information");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "status_block", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_status_block, "I", "Dump status block");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "stats_block", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_stats_block, "I", "Dump statistics block");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "stats_clear", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_stats_clear, "I", "Clear statistics block");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "bc_state", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_bc_state, "I", "Bootcode state information");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "dump_rx_bd_chain", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_dump_rx_bd_chain, "I", "Dump RX BD chain");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "dump_rx_mbuf_chain", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_dump_rx_mbuf_chain, "I", "Dump RX MBUF chain");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "dump_tx_chain", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_dump_tx_chain, "I", "Dump tx_bd chain");
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "dump_pg_chain", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_dump_pg_chain, "I", "Dump page chain");
+#endif
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "dump_ctx", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_dump_ctx, "I", "Dump context memory");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "breakpoint", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_breakpoint, "I", "Driver breakpoint");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "reg_read", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_reg_read, "I", "Register read");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "nvram_read", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_nvram_read, "I", "NVRAM read");
+
+ SYSCTL_ADD_PROC(ctx, children, OID_AUTO,
+ "phy_read", CTLTYPE_INT | CTLFLAG_RW,
+ (void *)sc, 0,
+ bce_sysctl_phy_read, "I", "PHY register read");
+
+#endif
+
+ DBEXIT(BCE_VERBOSE_MISC);
+}
+
+
+/****************************************************************************/
+/* BCE Debug Routines */
+/****************************************************************************/
+#ifdef BCE_DEBUG
+
+/****************************************************************************/
+/* Freezes the controller to allow for a cohesive state dump. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_freeze_controller(struct bce_softc *sc)
+{
+ u32 val;
+ val = REG_RD(sc, BCE_MISC_COMMAND);
+ val |= BCE_MISC_COMMAND_DISABLE_ALL;
+ REG_WR(sc, BCE_MISC_COMMAND, val);
+}
+
+
+/****************************************************************************/
+/* Unfreezes the controller after a freeze operation. This may not always */
+/* work and the controller will require a reset! */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_unfreeze_controller(struct bce_softc *sc)
+{
+ u32 val;
+ val = REG_RD(sc, BCE_MISC_COMMAND);
+ val |= BCE_MISC_COMMAND_ENABLE_ALL;
+ REG_WR(sc, BCE_MISC_COMMAND, val);
+}
+
+
+/****************************************************************************/
+/* Prints out Ethernet frame information from an mbuf. */
+/* */
+/* Partially decode an Ethernet frame to look at some important headers. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_enet(struct bce_softc *sc, struct mbuf *m)
+{
+ struct ether_vlan_header *eh;
+ u16 etype;
+ int ehlen;
+ struct ip *ip;
+ struct tcphdr *th;
+ struct udphdr *uh;
+ struct arphdr *ah;
+
+ BCE_PRINTF(
+ "-----------------------------"
+ " Frame Decode "
+ "-----------------------------\n");
+
+ eh = mtod(m, struct ether_vlan_header *);
+
+ /* Handle VLAN encapsulation if present. */
+ if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
+ etype = ntohs(eh->evl_proto);
+ ehlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
+ } else {
+ etype = ntohs(eh->evl_encap_proto);
+ ehlen = ETHER_HDR_LEN;
+ }
+
+ /* ToDo: Add VLAN output. */
+ BCE_PRINTF("enet: dest = %6D, src = %6D, type = 0x%04X, hlen = %d\n",
+ eh->evl_dhost, ":", eh->evl_shost, ":", etype, ehlen);
+
+ switch (etype) {
+ case ETHERTYPE_IP:
+ ip = (struct ip *)(m->m_data + ehlen);
+ BCE_PRINTF("--ip: dest = 0x%08X , src = 0x%08X, "
+ "len = %d bytes, protocol = 0x%02X, xsum = 0x%04X\n",
+ ntohl(ip->ip_dst.s_addr), ntohl(ip->ip_src.s_addr),
+ ntohs(ip->ip_len), ip->ip_p, ntohs(ip->ip_sum));
+
+ switch (ip->ip_p) {
+ case IPPROTO_TCP:
+ th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
+ BCE_PRINTF("-tcp: dest = %d, src = %d, hlen = "
+ "%d bytes, flags = 0x%b, csum = 0x%04X\n",
+ ntohs(th->th_dport), ntohs(th->th_sport),
+ (th->th_off << 2), th->th_flags,
+ "\20\10CWR\07ECE\06URG\05ACK\04PSH\03RST"
+ "\02SYN\01FIN", ntohs(th->th_sum));
+ break;
+ case IPPROTO_UDP:
+ uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
+ BCE_PRINTF("-udp: dest = %d, src = %d, len = %d "
+ "bytes, csum = 0x%04X\n", ntohs(uh->uh_dport),
+ ntohs(uh->uh_sport), ntohs(uh->uh_ulen),
+ ntohs(uh->uh_sum));
+ break;
+ case IPPROTO_ICMP:
+ BCE_PRINTF("icmp:\n");
+ break;
+ default:
+ BCE_PRINTF("----: Other IP protocol.\n");
+ }
+ break;
+ case ETHERTYPE_IPV6:
+ BCE_PRINTF("ipv6: No decode supported.\n");
+ break;
+ case ETHERTYPE_ARP:
+ BCE_PRINTF("-arp: ");
+ ah = (struct arphdr *) (m->m_data + ehlen);
+ switch (ntohs(ah->ar_op)) {
+ case ARPOP_REVREQUEST:
+ printf("reverse ARP request\n");
+ break;
+ case ARPOP_REVREPLY:
+ printf("reverse ARP reply\n");
+ break;
+ case ARPOP_REQUEST:
+ printf("ARP request\n");
+ break;
+ case ARPOP_REPLY:
+ printf("ARP reply\n");
+ break;
+ default:
+ printf("other ARP operation\n");
+ }
+ break;
+ default:
+ BCE_PRINTF("----: Other protocol.\n");
+ }
+
+ BCE_PRINTF(
+ "-----------------------------"
+ "--------------"
+ "-----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out information about an mbuf. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_mbuf(struct bce_softc *sc, struct mbuf *m)
+{
+ struct mbuf *mp = m;
+
+ if (m == NULL) {
+ BCE_PRINTF("mbuf: null pointer\n");
+ return;
+ }
+
+ while (mp) {
+ BCE_PRINTF("mbuf: %p, m_len = %d, m_flags = 0x%b, "
+ "m_data = %p\n", mp, mp->m_len, mp->m_flags,
+ "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY", mp->m_data);
+
+ if (mp->m_flags & M_PKTHDR) {
+ BCE_PRINTF("- m_pkthdr: len = %d, flags = 0x%b, "
+ "csum_flags = %b\n", mp->m_pkthdr.len,
+ mp->m_flags, "\20\12M_BCAST\13M_MCAST\14M_FRAG"
+ "\15M_FIRSTFRAG\16M_LASTFRAG\21M_VLANTAG"
+ "\22M_PROMISC\23M_NOFREE",
+ mp->m_pkthdr.csum_flags,
+ "\20\1CSUM_IP\2CSUM_TCP\3CSUM_UDP\4CSUM_IP_FRAGS"
+ "\5CSUM_FRAGMENT\6CSUM_TSO\11CSUM_IP_CHECKED"
+ "\12CSUM_IP_VALID\13CSUM_DATA_VALID"
+ "\14CSUM_PSEUDO_HDR");
+ }
+
+ if (mp->m_flags & M_EXT) {
+ BCE_PRINTF("- m_ext: %p, ext_size = %d, type = ",
+ mp->m_ext.ext_buf, mp->m_ext.ext_size);
+ switch (mp->m_ext.ext_type) {
+ case EXT_CLUSTER:
+ printf("EXT_CLUSTER\n"); break;
+ case EXT_SFBUF:
+ printf("EXT_SFBUF\n"); break;
+ case EXT_JUMBO9:
+ printf("EXT_JUMBO9\n"); break;
+ case EXT_JUMBO16:
+ printf("EXT_JUMBO16\n"); break;
+ case EXT_PACKET:
+ printf("EXT_PACKET\n"); break;
+ case EXT_MBUF:
+ printf("EXT_MBUF\n"); break;
+ case EXT_NET_DRV:
+ printf("EXT_NET_DRV\n"); break;
+ case EXT_MOD_TYPE:
+ printf("EXT_MDD_TYPE\n"); break;
+ case EXT_DISPOSABLE:
+ printf("EXT_DISPOSABLE\n"); break;
+ case EXT_EXTREF:
+ printf("EXT_EXTREF\n"); break;
+ default:
+ printf("UNKNOWN\n");
+ }
+ }
+
+ mp = mp->m_next;
+ }
+}
+
+
+/****************************************************************************/
+/* Prints out the mbufs in the TX mbuf chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_tx_mbuf_chain(struct bce_softc *sc, u16 chain_prod, int count)
+{
+ struct mbuf *m;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " tx mbuf data "
+ "----------------------------\n");
+
+ for (int i = 0; i < count; i++) {
+ m = sc->tx_mbuf_ptr[chain_prod];
+ BCE_PRINTF("txmbuf[0x%04X]\n", chain_prod);
+ bce_dump_mbuf(sc, m);
+ chain_prod = TX_CHAIN_IDX(NEXT_TX_BD(chain_prod));
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the mbufs in the RX mbuf chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_rx_mbuf_chain(struct bce_softc *sc, u16 chain_prod, int count)
+{
+ struct mbuf *m;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " rx mbuf data "
+ "----------------------------\n");
+
+ for (int i = 0; i < count; i++) {
+ m = sc->rx_mbuf_ptr[chain_prod];
+ BCE_PRINTF("rxmbuf[0x%04X]\n", chain_prod);
+ bce_dump_mbuf(sc, m);
+ chain_prod = RX_CHAIN_IDX(NEXT_RX_BD(chain_prod));
+ }
+
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+/****************************************************************************/
+/* Prints out the mbufs in the mbuf page chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_pg_mbuf_chain(struct bce_softc *sc, u16 chain_prod, int count)
+{
+ struct mbuf *m;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " pg mbuf data "
+ "----------------------------\n");
+
+ for (int i = 0; i < count; i++) {
+ m = sc->pg_mbuf_ptr[chain_prod];
+ BCE_PRINTF("pgmbuf[0x%04X]\n", chain_prod);
+ bce_dump_mbuf(sc, m);
+ chain_prod = PG_CHAIN_IDX(NEXT_PG_BD(chain_prod));
+ }
+
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+#endif
+
+
+/****************************************************************************/
+/* Prints out a tx_bd structure. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_txbd(struct bce_softc *sc, int idx, struct tx_bd *txbd)
+{
+ int i = 0;
+
+ if (idx > MAX_TX_BD)
+ /* Index out of range. */
+ BCE_PRINTF("tx_bd[0x%04X]: Invalid tx_bd index!\n", idx);
+ else if ((idx & USABLE_TX_BD_PER_PAGE) == USABLE_TX_BD_PER_PAGE)
+ /* TX Chain page pointer. */
+ BCE_PRINTF("tx_bd[0x%04X]: haddr = 0x%08X:%08X, chain page "
+ "pointer\n", idx, txbd->tx_bd_haddr_hi,
+ txbd->tx_bd_haddr_lo);
+ else {
+ /* Normal tx_bd entry. */
+ BCE_PRINTF("tx_bd[0x%04X]: haddr = 0x%08X:%08X, "
+ "mss_nbytes = 0x%08X, vlan tag = 0x%04X, flags = "
+ "0x%04X (", idx, txbd->tx_bd_haddr_hi,
+ txbd->tx_bd_haddr_lo, txbd->tx_bd_mss_nbytes,
+ txbd->tx_bd_vlan_tag, txbd->tx_bd_flags);
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_CONN_FAULT) {
+ if (i>0)
+ printf("|");
+ printf("CONN_FAULT");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_TCP_UDP_CKSUM) {
+ if (i>0)
+ printf("|");
+ printf("TCP_UDP_CKSUM");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_IP_CKSUM) {
+ if (i>0)
+ printf("|");
+ printf("IP_CKSUM");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_VLAN_TAG) {
+ if (i>0)
+ printf("|");
+ printf("VLAN");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_COAL_NOW) {
+ if (i>0)
+ printf("|");
+ printf("COAL_NOW");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_DONT_GEN_CRC) {
+ if (i>0)
+ printf("|");
+ printf("DONT_GEN_CRC");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_START) {
+ if (i>0)
+ printf("|");
+ printf("START");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_END) {
+ if (i>0)
+ printf("|");
+ printf("END");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_SW_LSO) {
+ if (i>0)
+ printf("|");
+ printf("LSO");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_SW_OPTION_WORD) {
+ if (i>0)
+ printf("|");
+ printf("SW_OPTION=%d", ((txbd->tx_bd_flags &
+ TX_BD_FLAGS_SW_OPTION_WORD) >> 8)); i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_SW_FLAGS) {
+ if (i>0)
+ printf("|");
+ printf("SW_FLAGS");
+ i++;
+ }
+
+ if (txbd->tx_bd_flags & TX_BD_FLAGS_SW_SNAP) {
+ if (i>0)
+ printf("|");
+ printf("SNAP)");
+ } else {
+ printf(")\n");
+ }
+ }
+}
+
+
+/****************************************************************************/
+/* Prints out a rx_bd structure. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_rxbd(struct bce_softc *sc, int idx, struct rx_bd *rxbd)
+{
+ if (idx > MAX_RX_BD)
+ /* Index out of range. */
+ BCE_PRINTF("rx_bd[0x%04X]: Invalid rx_bd index!\n", idx);
+ else if ((idx & USABLE_RX_BD_PER_PAGE) == USABLE_RX_BD_PER_PAGE)
+ /* RX Chain page pointer. */
+ BCE_PRINTF("rx_bd[0x%04X]: haddr = 0x%08X:%08X, chain page "
+ "pointer\n", idx, rxbd->rx_bd_haddr_hi,
+ rxbd->rx_bd_haddr_lo);
+ else
+ /* Normal rx_bd entry. */
+ BCE_PRINTF("rx_bd[0x%04X]: haddr = 0x%08X:%08X, nbytes = "
+ "0x%08X, flags = 0x%08X\n", idx, rxbd->rx_bd_haddr_hi,
+ rxbd->rx_bd_haddr_lo, rxbd->rx_bd_len,
+ rxbd->rx_bd_flags);
+}
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+/****************************************************************************/
+/* Prints out a rx_bd structure in the page chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_pgbd(struct bce_softc *sc, int idx, struct rx_bd *pgbd)
+{
+ if (idx > MAX_PG_BD)
+ /* Index out of range. */
+ BCE_PRINTF("pg_bd[0x%04X]: Invalid pg_bd index!\n", idx);
+ else if ((idx & USABLE_PG_BD_PER_PAGE) == USABLE_PG_BD_PER_PAGE)
+ /* Page Chain page pointer. */
+ BCE_PRINTF("px_bd[0x%04X]: haddr = 0x%08X:%08X, chain page pointer\n",
+ idx, pgbd->rx_bd_haddr_hi, pgbd->rx_bd_haddr_lo);
+ else
+ /* Normal rx_bd entry. */
+ BCE_PRINTF("pg_bd[0x%04X]: haddr = 0x%08X:%08X, nbytes = 0x%08X, "
+ "flags = 0x%08X\n", idx,
+ pgbd->rx_bd_haddr_hi, pgbd->rx_bd_haddr_lo,
+ pgbd->rx_bd_len, pgbd->rx_bd_flags);
+}
+#endif
+
+
+/****************************************************************************/
+/* Prints out a l2_fhdr structure. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_l2fhdr(struct bce_softc *sc, int idx, struct l2_fhdr *l2fhdr)
+{
+ BCE_PRINTF("l2_fhdr[0x%04X]: status = 0x%b, "
+ "pkt_len = %d, vlan = 0x%04x, ip_xsum/hdr_len = 0x%04X, "
+ "tcp_udp_xsum = 0x%04X\n", idx,
+ l2fhdr->l2_fhdr_status, BCE_L2FHDR_PRINTFB,
+ l2fhdr->l2_fhdr_pkt_len, l2fhdr->l2_fhdr_vlan_tag,
+ l2fhdr->l2_fhdr_ip_xsum, l2fhdr->l2_fhdr_tcp_udp_xsum);
+}
+
+
+/****************************************************************************/
+/* Prints out context memory info. (Only useful for CID 0 to 16.) */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_ctx(struct bce_softc *sc, u16 cid)
+{
+ if (cid > TX_CID) {
+ BCE_PRINTF(" Unknown CID\n");
+ return;
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ " CTX Data "
+ "----------------------------\n");
+
+ BCE_PRINTF(" 0x%04X - (CID) Context ID\n", cid);
+
+ if (cid == RX_CID) {
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_HOST_BDIDX) host rx "
+ "producer index\n",
+ CTX_RD(sc, GET_CID_ADDR(cid), BCE_L2CTX_RX_HOST_BDIDX));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_HOST_BSEQ) host "
+ "byte sequence\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_HOST_BSEQ));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_BSEQ) h/w byte sequence\n",
+ CTX_RD(sc, GET_CID_ADDR(cid), BCE_L2CTX_RX_NX_BSEQ));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_BDHADDR_HI) h/w buffer "
+ "descriptor address\n",
+ CTX_RD(sc, GET_CID_ADDR(cid), BCE_L2CTX_RX_NX_BDHADDR_HI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_BDHADDR_LO) h/w buffer "
+ "descriptor address\n",
+ CTX_RD(sc, GET_CID_ADDR(cid), BCE_L2CTX_RX_NX_BDHADDR_LO));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_BDIDX) h/w rx consumer "
+ "index\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_NX_BDIDX));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_HOST_PG_BDIDX) host page "
+ "producer index\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_HOST_PG_BDIDX));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_PG_BUF_SIZE) host rx_bd/page "
+ "buffer size\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_PG_BUF_SIZE));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_PG_BDHADDR_HI) h/w page "
+ "chain address\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_NX_PG_BDHADDR_HI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_PG_BDHADDR_LO) h/w page "
+ "chain address\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_NX_PG_BDHADDR_LO));
+ BCE_PRINTF(" 0x%08X - (L2CTX_RX_NX_PG_BDIDX) h/w page "
+ "consumer index\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_RX_NX_PG_BDIDX));
+ } else if (cid == TX_CID) {
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_TYPE_XI) ctx type\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_TYPE_XI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_CMD_TX_TYPE_XI) ctx "
+ "cmd\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_CMD_TYPE_XI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_TBDR_BDHADDR_HI_XI) "
+ "h/w buffer descriptor address\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_TBDR_BHADDR_HI_XI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_TBDR_BHADDR_LO_XI) "
+ "h/w buffer descriptor address\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_TBDR_BHADDR_LO_XI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_HOST_BIDX_XI) "
+ "host producer index\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_HOST_BIDX_XI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_HOST_BSEQ_XI) "
+ "host byte sequence\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_HOST_BSEQ_XI));
+ } else {
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_TYPE) ctx type\n",
+ CTX_RD(sc, GET_CID_ADDR(cid), BCE_L2CTX_TX_TYPE));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_CMD_TYPE) ctx cmd\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_CMD_TYPE));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_TBDR_BDHADDR_HI) "
+ "h/w buffer descriptor address\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_TBDR_BHADDR_HI));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_TBDR_BHADDR_LO) "
+ "h/w buffer descriptor address\n",
+ CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_TBDR_BHADDR_LO));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_HOST_BIDX) host "
+ "producer index\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_HOST_BIDX));
+ BCE_PRINTF(" 0x%08X - (L2CTX_TX_HOST_BSEQ) host byte "
+ "sequence\n", CTX_RD(sc, GET_CID_ADDR(cid),
+ BCE_L2CTX_TX_HOST_BSEQ));
+ }
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ " Raw CTX "
+ "----------------------------\n");
+
+ for (int i = 0x0; i < 0x300; i += 0x10) {
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X 0x%08X 0x%08X\n", i,
+ CTX_RD(sc, GET_CID_ADDR(cid), i),
+ CTX_RD(sc, GET_CID_ADDR(cid), i + 0x4),
+ CTX_RD(sc, GET_CID_ADDR(cid), i + 0x8),
+ CTX_RD(sc, GET_CID_ADDR(cid), i + 0xc));
+ }
+
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the FTQ data. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_ftqs(struct bce_softc *sc)
+{
+ u32 cmd, ctl, cur_depth, max_depth, valid_cnt, val;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " FTQ Data "
+ "----------------------------\n");
+
+ BCE_PRINTF(" FTQ Command Control Depth_Now "
+ "Max_Depth Valid_Cnt \n");
+ BCE_PRINTF(" ------- ---------- ---------- ---------- "
+ "---------- ----------\n");
+
+ /* Setup the generic statistic counters for the FTQ valid count. */
+ val = (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RV2PPQ_VALID_CNT << 24) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RXPCQ_VALID_CNT << 16) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RXPQ_VALID_CNT << 8) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RLUPQ_VALID_CNT);
+ REG_WR(sc, BCE_HC_STAT_GEN_SEL_0, val);
+
+ val = (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_TSCHQ_VALID_CNT << 24) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RDMAQ_VALID_CNT << 16) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RV2PTQ_VALID_CNT << 8) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RV2PMQ_VALID_CNT);
+ REG_WR(sc, BCE_HC_STAT_GEN_SEL_1, val);
+
+ val = (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_TPATQ_VALID_CNT << 24) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_TDMAQ_VALID_CNT << 16) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_TXPQ_VALID_CNT << 8) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_TBDRQ_VALID_CNT);
+ REG_WR(sc, BCE_HC_STAT_GEN_SEL_2, val);
+
+ val = (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_COMQ_VALID_CNT << 24) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_COMTQ_VALID_CNT << 16) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_COMXQ_VALID_CNT << 8) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_TASQ_VALID_CNT);
+ REG_WR(sc, BCE_HC_STAT_GEN_SEL_3, val);
+
+ /* Input queue to the Receive Lookup state machine */
+ cmd = REG_RD(sc, BCE_RLUP_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_RLUP_FTQ_CTL);
+ cur_depth = (ctl & BCE_RLUP_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RLUP_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT0);
+ BCE_PRINTF(" RLUP 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Receive Processor */
+ cmd = REG_RD_IND(sc, BCE_RXP_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_RXP_FTQ_CTL);
+ cur_depth = (ctl & BCE_RXP_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RXP_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT1);
+ BCE_PRINTF(" RXP 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Recevie Processor */
+ cmd = REG_RD_IND(sc, BCE_RXP_CFTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_RXP_CFTQ_CTL);
+ cur_depth = (ctl & BCE_RXP_CFTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RXP_CFTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT2);
+ BCE_PRINTF(" RXPC 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Receive Virtual to Physical state machine */
+ cmd = REG_RD(sc, BCE_RV2P_PFTQ_CMD);
+ ctl = REG_RD(sc, BCE_RV2P_PFTQ_CTL);
+ cur_depth = (ctl & BCE_RV2P_PFTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RV2P_PFTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT3);
+ BCE_PRINTF(" RV2PP 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Recevie Virtual to Physical state machine */
+ cmd = REG_RD(sc, BCE_RV2P_MFTQ_CMD);
+ ctl = REG_RD(sc, BCE_RV2P_MFTQ_CTL);
+ cur_depth = (ctl & BCE_RV2P_MFTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RV2P_MFTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT4);
+ BCE_PRINTF(" RV2PM 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Receive Virtual to Physical state machine */
+ cmd = REG_RD(sc, BCE_RV2P_TFTQ_CMD);
+ ctl = REG_RD(sc, BCE_RV2P_TFTQ_CTL);
+ cur_depth = (ctl & BCE_RV2P_TFTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RV2P_TFTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT5);
+ BCE_PRINTF(" RV2PT 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Receive DMA state machine */
+ cmd = REG_RD(sc, BCE_RDMA_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_RDMA_FTQ_CTL);
+ cur_depth = (ctl & BCE_RDMA_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_RDMA_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT6);
+ BCE_PRINTF(" RDMA 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Transmit Scheduler state machine */
+ cmd = REG_RD(sc, BCE_TSCH_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_TSCH_FTQ_CTL);
+ cur_depth = (ctl & BCE_TSCH_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_TSCH_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT7);
+ BCE_PRINTF(" TSCH 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Transmit Buffer Descriptor state machine */
+ cmd = REG_RD(sc, BCE_TBDR_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_TBDR_FTQ_CTL);
+ cur_depth = (ctl & BCE_TBDR_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_TBDR_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT8);
+ BCE_PRINTF(" TBDR 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Transmit Processor */
+ cmd = REG_RD_IND(sc, BCE_TXP_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_TXP_FTQ_CTL);
+ cur_depth = (ctl & BCE_TXP_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_TXP_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT9);
+ BCE_PRINTF(" TXP 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Transmit DMA state machine */
+ cmd = REG_RD(sc, BCE_TDMA_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_TDMA_FTQ_CTL);
+ cur_depth = (ctl & BCE_TDMA_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_TDMA_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT10);
+ BCE_PRINTF(" TDMA 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Transmit Patch-Up Processor */
+ cmd = REG_RD_IND(sc, BCE_TPAT_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_TPAT_FTQ_CTL);
+ cur_depth = (ctl & BCE_TPAT_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_TPAT_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT11);
+ BCE_PRINTF(" TPAT 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Transmit Assembler state machine */
+ cmd = REG_RD_IND(sc, BCE_TAS_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_TAS_FTQ_CTL);
+ cur_depth = (ctl & BCE_TAS_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_TAS_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT12);
+ BCE_PRINTF(" TAS 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Completion Processor */
+ cmd = REG_RD_IND(sc, BCE_COM_COMXQ_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_COM_COMXQ_FTQ_CTL);
+ cur_depth = (ctl & BCE_COM_COMXQ_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_COM_COMXQ_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT13);
+ BCE_PRINTF(" COMX 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Completion Processor */
+ cmd = REG_RD_IND(sc, BCE_COM_COMTQ_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_COM_COMTQ_FTQ_CTL);
+ cur_depth = (ctl & BCE_COM_COMTQ_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_COM_COMTQ_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT14);
+ BCE_PRINTF(" COMT 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Completion Processor */
+ cmd = REG_RD_IND(sc, BCE_COM_COMQ_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_COM_COMQ_FTQ_CTL);
+ cur_depth = (ctl & BCE_COM_COMQ_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_COM_COMQ_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT15);
+ BCE_PRINTF(" COMX 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Setup the generic statistic counters for the FTQ valid count. */
+ val = (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_CSQ_VALID_CNT << 16) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_CPQ_VALID_CNT << 8) |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_MGMQ_VALID_CNT);
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716))
+ val = val |
+ (BCE_HC_STAT_GEN_SEL_0_GEN_SEL_0_RV2PCSQ_VALID_CNT_XI <<
+ 24);
+ REG_WR(sc, BCE_HC_STAT_GEN_SEL_0, val);
+
+ /* Input queue to the Management Control Processor */
+ cmd = REG_RD_IND(sc, BCE_MCP_MCPQ_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_MCP_MCPQ_FTQ_CTL);
+ cur_depth = (ctl & BCE_MCP_MCPQ_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_MCP_MCPQ_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT0);
+ BCE_PRINTF(" MCP 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Command Processor */
+ cmd = REG_RD_IND(sc, BCE_CP_CPQ_FTQ_CMD);
+ ctl = REG_RD_IND(sc, BCE_CP_CPQ_FTQ_CTL);
+ cur_depth = (ctl & BCE_CP_CPQ_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_CP_CPQ_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT1);
+ BCE_PRINTF(" CP 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ /* Input queue to the Completion Scheduler state machine */
+ cmd = REG_RD(sc, BCE_CSCH_CH_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_CSCH_CH_FTQ_CTL);
+ cur_depth = (ctl & BCE_CSCH_CH_FTQ_CTL_CUR_DEPTH) >> 22;
+ max_depth = (ctl & BCE_CSCH_CH_FTQ_CTL_MAX_DEPTH) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT2);
+ BCE_PRINTF(" CS 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+
+ if ((BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5709) ||
+ (BCE_CHIP_NUM(sc) == BCE_CHIP_NUM_5716)) {
+ /* Input queue to the RV2P Command Scheduler */
+ cmd = REG_RD(sc, BCE_RV2PCSR_FTQ_CMD);
+ ctl = REG_RD(sc, BCE_RV2PCSR_FTQ_CTL);
+ cur_depth = (ctl & 0xFFC00000) >> 22;
+ max_depth = (ctl & 0x003FF000) >> 12;
+ valid_cnt = REG_RD(sc, BCE_HC_STAT_GEN_STAT3);
+ BCE_PRINTF(" RV2PCSR 0x%08X 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ cmd, ctl, cur_depth, max_depth, valid_cnt);
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the TX chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_tx_chain(struct bce_softc *sc, u16 tx_prod, int count)
+{
+ struct tx_bd *txbd;
+
+ /* First some info about the tx_bd chain structure. */
+ BCE_PRINTF(
+ "----------------------------"
+ " tx_bd chain "
+ "----------------------------\n");
+
+ BCE_PRINTF("page size = 0x%08X, tx chain pages = 0x%08X\n",
+ (u32) BCM_PAGE_SIZE, (u32) TX_PAGES);
+ BCE_PRINTF("tx_bd per page = 0x%08X, usable tx_bd per page = 0x%08X\n",
+ (u32) TOTAL_TX_BD_PER_PAGE, (u32) USABLE_TX_BD_PER_PAGE);
+ BCE_PRINTF("total tx_bd = 0x%08X\n", (u32) TOTAL_TX_BD);
+
+ BCE_PRINTF(
+ "----------------------------"
+ " tx_bd data "
+ "----------------------------\n");
+
+ /* Now print out a decoded list of TX buffer descriptors. */
+ for (int i = 0; i < count; i++) {
+ txbd = &sc->tx_bd_chain[TX_PAGE(tx_prod)][TX_IDX(tx_prod)];
+ bce_dump_txbd(sc, tx_prod, txbd);
+ tx_prod++;
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the RX chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_rx_bd_chain(struct bce_softc *sc, u16 rx_prod, int count)
+{
+ struct rx_bd *rxbd;
+
+ /* First some info about the rx_bd chain structure. */
+ BCE_PRINTF(
+ "----------------------------"
+ " rx_bd chain "
+ "----------------------------\n");
+
+ BCE_PRINTF("page size = 0x%08X, rx chain pages = 0x%08X\n",
+ (u32) BCM_PAGE_SIZE, (u32) RX_PAGES);
+
+ BCE_PRINTF("rx_bd per page = 0x%08X, usable rx_bd per page = 0x%08X\n",
+ (u32) TOTAL_RX_BD_PER_PAGE, (u32) USABLE_RX_BD_PER_PAGE);
+
+ BCE_PRINTF("total rx_bd = 0x%08X\n", (u32) TOTAL_RX_BD);
+
+ BCE_PRINTF(
+ "----------------------------"
+ " rx_bd data "
+ "----------------------------\n");
+
+ /* Now print out the rx_bd's themselves. */
+ for (int i = 0; i < count; i++) {
+ rxbd = &sc->rx_bd_chain[RX_PAGE(rx_prod)][RX_IDX(rx_prod)];
+ bce_dump_rxbd(sc, rx_prod, rxbd);
+ rx_prod = RX_CHAIN_IDX(rx_prod + 1);
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+#ifdef BCE_JUMBO_HDRSPLIT
+/****************************************************************************/
+/* Prints out the page chain. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_pg_chain(struct bce_softc *sc, u16 pg_prod, int count)
+{
+ struct rx_bd *pgbd;
+
+ /* First some info about the page chain structure. */
+ BCE_PRINTF(
+ "----------------------------"
+ " page chain "
+ "----------------------------\n");
+
+ BCE_PRINTF("page size = 0x%08X, pg chain pages = 0x%08X\n",
+ (u32) BCM_PAGE_SIZE, (u32) PG_PAGES);
+
+ BCE_PRINTF("rx_bd per page = 0x%08X, usable rx_bd per page = 0x%08X\n",
+ (u32) TOTAL_PG_BD_PER_PAGE, (u32) USABLE_PG_BD_PER_PAGE);
+
+ BCE_PRINTF("total rx_bd = 0x%08X, max_pg_bd = 0x%08X\n",
+ (u32) TOTAL_PG_BD, (u32) MAX_PG_BD);
+
+ BCE_PRINTF(
+ "----------------------------"
+ " page data "
+ "----------------------------\n");
+
+ /* Now print out the rx_bd's themselves. */
+ for (int i = 0; i < count; i++) {
+ pgbd = &sc->pg_bd_chain[PG_PAGE(pg_prod)][PG_IDX(pg_prod)];
+ bce_dump_pgbd(sc, pg_prod, pgbd);
+ pg_prod = PG_CHAIN_IDX(pg_prod + 1);
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+#endif
+
+
+#define BCE_PRINT_RX_CONS(arg) \
+if (sblk->status_rx_quick_consumer_index##arg) \
+ BCE_PRINTF("0x%04X(0x%04X) - rx_quick_consumer_index%d\n", \
+ sblk->status_rx_quick_consumer_index##arg, (u16) \
+ RX_CHAIN_IDX(sblk->status_rx_quick_consumer_index##arg), \
+ arg);
+
+
+#define BCE_PRINT_TX_CONS(arg) \
+if (sblk->status_tx_quick_consumer_index##arg) \
+ BCE_PRINTF("0x%04X(0x%04X) - tx_quick_consumer_index%d\n", \
+ sblk->status_tx_quick_consumer_index##arg, (u16) \
+ TX_CHAIN_IDX(sblk->status_tx_quick_consumer_index##arg), \
+ arg);
+
+/****************************************************************************/
+/* Prints out the status block from host memory. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_status_block(struct bce_softc *sc)
+{
+ struct status_block *sblk;
+
+ sblk = sc->status_block;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " Status Block "
+ "----------------------------\n");
+
+ /* Theses indices are used for normal L2 drivers. */
+ BCE_PRINTF(" 0x%08X - attn_bits\n",
+ sblk->status_attn_bits);
+
+ BCE_PRINTF(" 0x%08X - attn_bits_ack\n",
+ sblk->status_attn_bits_ack);
+
+ BCE_PRINT_RX_CONS(0);
+ BCE_PRINT_TX_CONS(0)
+
+ BCE_PRINTF(" 0x%04X - status_idx\n", sblk->status_idx);
+
+ /* Theses indices are not used for normal L2 drivers. */
+ BCE_PRINT_RX_CONS(1); BCE_PRINT_RX_CONS(2); BCE_PRINT_RX_CONS(3);
+ BCE_PRINT_RX_CONS(4); BCE_PRINT_RX_CONS(5); BCE_PRINT_RX_CONS(6);
+ BCE_PRINT_RX_CONS(7); BCE_PRINT_RX_CONS(8); BCE_PRINT_RX_CONS(9);
+ BCE_PRINT_RX_CONS(10); BCE_PRINT_RX_CONS(11); BCE_PRINT_RX_CONS(12);
+ BCE_PRINT_RX_CONS(13); BCE_PRINT_RX_CONS(14); BCE_PRINT_RX_CONS(15);
+
+ BCE_PRINT_TX_CONS(1); BCE_PRINT_TX_CONS(2); BCE_PRINT_TX_CONS(3);
+
+ if (sblk->status_completion_producer_index ||
+ sblk->status_cmd_consumer_index)
+ BCE_PRINTF("com_prod = 0x%08X, cmd_cons = 0x%08X\n",
+ sblk->status_completion_producer_index,
+ sblk->status_cmd_consumer_index);
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+#define BCE_PRINT_64BIT_STAT(arg) \
+if (sblk->arg##_lo || sblk->arg##_hi) \
+ BCE_PRINTF("0x%08X:%08X : %s\n", sblk->arg##_hi, \
+ sblk->arg##_lo, #arg);
+
+#define BCE_PRINT_32BIT_STAT(arg) \
+if (sblk->arg) \
+ BCE_PRINTF(" 0x%08X : %s\n", \
+ sblk->arg, #arg);
+
+/****************************************************************************/
+/* Prints out the statistics block from host memory. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_stats_block(struct bce_softc *sc)
+{
+ struct statistics_block *sblk;
+
+ sblk = sc->stats_block;
+
+ BCE_PRINTF(
+ "---------------"
+ " Stats Block (All Stats Not Shown Are 0) "
+ "---------------\n");
+
+ BCE_PRINT_64BIT_STAT(stat_IfHCInOctets);
+ BCE_PRINT_64BIT_STAT(stat_IfHCInBadOctets);
+ BCE_PRINT_64BIT_STAT(stat_IfHCOutOctets);
+ BCE_PRINT_64BIT_STAT(stat_IfHCOutBadOctets);
+ BCE_PRINT_64BIT_STAT(stat_IfHCInUcastPkts);
+ BCE_PRINT_64BIT_STAT(stat_IfHCInBroadcastPkts);
+ BCE_PRINT_64BIT_STAT(stat_IfHCInMulticastPkts);
+ BCE_PRINT_64BIT_STAT(stat_IfHCOutUcastPkts);
+ BCE_PRINT_64BIT_STAT(stat_IfHCOutBroadcastPkts);
+ BCE_PRINT_64BIT_STAT(stat_IfHCOutMulticastPkts);
+ BCE_PRINT_32BIT_STAT(
+ stat_emac_tx_stat_dot3statsinternalmactransmiterrors);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsCarrierSenseErrors);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsFCSErrors);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsAlignmentErrors);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsSingleCollisionFrames);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsMultipleCollisionFrames);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsDeferredTransmissions);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsExcessiveCollisions);
+ BCE_PRINT_32BIT_STAT(stat_Dot3StatsLateCollisions);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsCollisions);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsFragments);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsJabbers);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsUndersizePkts);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsOversizePkts);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx64Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx65Octetsto127Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx128Octetsto255Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx256Octetsto511Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx512Octetsto1023Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx1024Octetsto1522Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsRx1523Octetsto9022Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx64Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx65Octetsto127Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx128Octetsto255Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx256Octetsto511Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx512Octetsto1023Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx1024Octetsto1522Octets);
+ BCE_PRINT_32BIT_STAT(stat_EtherStatsPktsTx1523Octetsto9022Octets);
+ BCE_PRINT_32BIT_STAT(stat_XonPauseFramesReceived);
+ BCE_PRINT_32BIT_STAT(stat_XoffPauseFramesReceived);
+ BCE_PRINT_32BIT_STAT(stat_OutXonSent);
+ BCE_PRINT_32BIT_STAT(stat_OutXoffSent);
+ BCE_PRINT_32BIT_STAT(stat_FlowControlDone);
+ BCE_PRINT_32BIT_STAT(stat_MacControlFramesReceived);
+ BCE_PRINT_32BIT_STAT(stat_XoffStateEntered);
+ BCE_PRINT_32BIT_STAT(stat_IfInFramesL2FilterDiscards);
+ BCE_PRINT_32BIT_STAT(stat_IfInRuleCheckerDiscards);
+ BCE_PRINT_32BIT_STAT(stat_IfInFTQDiscards);
+ BCE_PRINT_32BIT_STAT(stat_IfInMBUFDiscards);
+ BCE_PRINT_32BIT_STAT(stat_IfInRuleCheckerP4Hit);
+ BCE_PRINT_32BIT_STAT(stat_CatchupInRuleCheckerDiscards);
+ BCE_PRINT_32BIT_STAT(stat_CatchupInFTQDiscards);
+ BCE_PRINT_32BIT_STAT(stat_CatchupInMBUFDiscards);
+ BCE_PRINT_32BIT_STAT(stat_CatchupInRuleCheckerP4Hit);
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out a summary of the driver state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_driver_state(struct bce_softc *sc)
+{
+ u32 val_hi, val_lo;
+
+ BCE_PRINTF(
+ "-----------------------------"
+ " Driver State "
+ "-----------------------------\n");
+
+ val_hi = BCE_ADDR_HI(sc);
+ val_lo = BCE_ADDR_LO(sc);
+ BCE_PRINTF("0x%08X:%08X - (sc) driver softc structure virtual "
+ "address\n", val_hi, val_lo);
+
+ val_hi = BCE_ADDR_HI(sc->bce_vhandle);
+ val_lo = BCE_ADDR_LO(sc->bce_vhandle);
+ BCE_PRINTF("0x%08X:%08X - (sc->bce_vhandle) PCI BAR virtual "
+ "address\n", val_hi, val_lo);
+
+ val_hi = BCE_ADDR_HI(sc->status_block);
+ val_lo = BCE_ADDR_LO(sc->status_block);
+ BCE_PRINTF("0x%08X:%08X - (sc->status_block) status block "
+ "virtual address\n", val_hi, val_lo);
+
+ val_hi = BCE_ADDR_HI(sc->stats_block);
+ val_lo = BCE_ADDR_LO(sc->stats_block);
+ BCE_PRINTF("0x%08X:%08X - (sc->stats_block) statistics block "
+ "virtual address\n", val_hi, val_lo);
+
+ val_hi = BCE_ADDR_HI(sc->tx_bd_chain);
+ val_lo = BCE_ADDR_LO(sc->tx_bd_chain);
+ BCE_PRINTF("0x%08X:%08X - (sc->tx_bd_chain) tx_bd chain "
+ "virtual adddress\n", val_hi, val_lo);
+
+ val_hi = BCE_ADDR_HI(sc->rx_bd_chain);
+ val_lo = BCE_ADDR_LO(sc->rx_bd_chain);
+ BCE_PRINTF("0x%08X:%08X - (sc->rx_bd_chain) rx_bd chain "
+ "virtual address\n", val_hi, val_lo);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ val_hi = BCE_ADDR_HI(sc->pg_bd_chain);
+ val_lo = BCE_ADDR_LO(sc->pg_bd_chain);
+ BCE_PRINTF("0x%08X:%08X - (sc->pg_bd_chain) page chain "
+ "virtual address\n", val_hi, val_lo);
+#endif
+
+ val_hi = BCE_ADDR_HI(sc->tx_mbuf_ptr);
+ val_lo = BCE_ADDR_LO(sc->tx_mbuf_ptr);
+ BCE_PRINTF("0x%08X:%08X - (sc->tx_mbuf_ptr) tx mbuf chain "
+ "virtual address\n", val_hi, val_lo);
+
+ val_hi = BCE_ADDR_HI(sc->rx_mbuf_ptr);
+ val_lo = BCE_ADDR_LO(sc->rx_mbuf_ptr);
+ BCE_PRINTF("0x%08X:%08X - (sc->rx_mbuf_ptr) rx mbuf chain "
+ "virtual address\n", val_hi, val_lo);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ val_hi = BCE_ADDR_HI(sc->pg_mbuf_ptr);
+ val_lo = BCE_ADDR_LO(sc->pg_mbuf_ptr);
+ BCE_PRINTF("0x%08X:%08X - (sc->pg_mbuf_ptr) page mbuf chain "
+ "virtual address\n", val_hi, val_lo);
+#endif
+
+ BCE_PRINTF(" 0x%08X - (sc->interrupts_generated) "
+ "h/w intrs\n", sc->interrupts_generated);
+
+ BCE_PRINTF(" 0x%08X - (sc->interrupts_rx) "
+ "rx interrupts handled\n", sc->interrupts_rx);
+
+ BCE_PRINTF(" 0x%08X - (sc->interrupts_tx) "
+ "tx interrupts handled\n", sc->interrupts_tx);
+
+ BCE_PRINTF(" 0x%08X - (sc->phy_interrupts) "
+ "phy interrupts handled\n", sc->phy_interrupts);
+
+ BCE_PRINTF(" 0x%08X - (sc->last_status_idx) "
+ "status block index\n", sc->last_status_idx);
+
+ BCE_PRINTF(" 0x%04X(0x%04X) - (sc->tx_prod) tx producer "
+ "index\n", sc->tx_prod, (u16) TX_CHAIN_IDX(sc->tx_prod));
+
+ BCE_PRINTF(" 0x%04X(0x%04X) - (sc->tx_cons) tx consumer "
+ "index\n", sc->tx_cons, (u16) TX_CHAIN_IDX(sc->tx_cons));
+
+ BCE_PRINTF(" 0x%08X - (sc->tx_prod_bseq) tx producer "
+ "byte seq index\n", sc->tx_prod_bseq);
+
+ BCE_PRINTF(" 0x%08X - (sc->debug_tx_mbuf_alloc) tx "
+ "mbufs allocated\n", sc->debug_tx_mbuf_alloc);
+
+ BCE_PRINTF(" 0x%08X - (sc->used_tx_bd) used "
+ "tx_bd's\n", sc->used_tx_bd);
+
+ BCE_PRINTF("0x%08X/%08X - (sc->tx_hi_watermark) tx hi "
+ "watermark\n", sc->tx_hi_watermark, sc->max_tx_bd);
+
+ BCE_PRINTF(" 0x%04X(0x%04X) - (sc->rx_prod) rx producer "
+ "index\n", sc->rx_prod, (u16) RX_CHAIN_IDX(sc->rx_prod));
+
+ BCE_PRINTF(" 0x%04X(0x%04X) - (sc->rx_cons) rx consumer "
+ "index\n", sc->rx_cons, (u16) RX_CHAIN_IDX(sc->rx_cons));
+
+ BCE_PRINTF(" 0x%08X - (sc->rx_prod_bseq) rx producer "
+ "byte seq index\n", sc->rx_prod_bseq);
+
+ BCE_PRINTF(" 0x%08X - (sc->debug_rx_mbuf_alloc) rx "
+ "mbufs allocated\n", sc->debug_rx_mbuf_alloc);
+
+ BCE_PRINTF(" 0x%08X - (sc->free_rx_bd) free "
+ "rx_bd's\n", sc->free_rx_bd);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ BCE_PRINTF(" 0x%04X(0x%04X) - (sc->pg_prod) page producer "
+ "index\n", sc->pg_prod, (u16) PG_CHAIN_IDX(sc->pg_prod));
+
+ BCE_PRINTF(" 0x%04X(0x%04X) - (sc->pg_cons) page consumer "
+ "index\n", sc->pg_cons, (u16) PG_CHAIN_IDX(sc->pg_cons));
+
+ BCE_PRINTF(" 0x%08X - (sc->debug_pg_mbuf_alloc) page "
+ "mbufs allocated\n", sc->debug_pg_mbuf_alloc);
+
+ BCE_PRINTF(" 0x%08X - (sc->free_pg_bd) free page "
+ "rx_bd's\n", sc->free_pg_bd);
+
+ BCE_PRINTF("0x%08X/%08X - (sc->pg_low_watermark) page low "
+ "watermark\n", sc->pg_low_watermark, sc->max_pg_bd);
+#endif
+
+ BCE_PRINTF(" 0x%08X - (sc->mbuf_alloc_failed_count) "
+ "mbuf alloc failures\n", sc->mbuf_alloc_failed_count);
+
+ BCE_PRINTF(" 0x%08X - (sc->bce_flags) "
+ "bce mac flags\n", sc->bce_flags);
+
+ BCE_PRINTF(" 0x%08X - (sc->bce_phy_flags) "
+ "bce phy flags\n", sc->bce_phy_flags);
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the hardware state through a summary of important register, */
+/* followed by a complete register dump. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_hw_state(struct bce_softc *sc)
+{
+ u32 val;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " Hardware State "
+ "----------------------------\n");
+
+ BCE_PRINTF("%s - bootcode version\n", sc->bce_bc_ver);
+
+ val = REG_RD(sc, BCE_MISC_ENABLE_STATUS_BITS);
+ BCE_PRINTF("0x%08X - (0x%06X) misc_enable_status_bits\n",
+ val, BCE_MISC_ENABLE_STATUS_BITS);
+
+ val = REG_RD(sc, BCE_DMA_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) dma_status\n",
+ val, BCE_DMA_STATUS);
+
+ val = REG_RD(sc, BCE_CTX_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) ctx_status\n",
+ val, BCE_CTX_STATUS);
+
+ val = REG_RD(sc, BCE_EMAC_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) emac_status\n",
+ val, BCE_EMAC_STATUS);
+
+ val = REG_RD(sc, BCE_RPM_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) rpm_status\n",
+ val, BCE_RPM_STATUS);
+
+ /* ToDo: Create a #define for this constant. */
+ val = REG_RD(sc, 0x2004);
+ BCE_PRINTF("0x%08X - (0x%06X) rlup_status\n",
+ val, 0x2004);
+
+ val = REG_RD(sc, BCE_RV2P_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) rv2p_status\n",
+ val, BCE_RV2P_STATUS);
+
+ /* ToDo: Create a #define for this constant. */
+ val = REG_RD(sc, 0x2c04);
+ BCE_PRINTF("0x%08X - (0x%06X) rdma_status\n",
+ val, 0x2c04);
+
+ val = REG_RD(sc, BCE_TBDR_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) tbdr_status\n",
+ val, BCE_TBDR_STATUS);
+
+ val = REG_RD(sc, BCE_TDMA_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) tdma_status\n",
+ val, BCE_TDMA_STATUS);
+
+ val = REG_RD(sc, BCE_HC_STATUS);
+ BCE_PRINTF("0x%08X - (0x%06X) hc_status\n",
+ val, BCE_HC_STATUS);
+
+ val = REG_RD_IND(sc, BCE_TXP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) txp_cpu_state\n",
+ val, BCE_TXP_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_TPAT_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) tpat_cpu_state\n",
+ val, BCE_TPAT_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_RXP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) rxp_cpu_state\n",
+ val, BCE_RXP_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_COM_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) com_cpu_state\n",
+ val, BCE_COM_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_MCP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) mcp_cpu_state\n",
+ val, BCE_MCP_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_CP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) cp_cpu_state\n",
+ val, BCE_CP_CPU_STATE);
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+
+ BCE_PRINTF(
+ "----------------------------"
+ " Register Dump "
+ "----------------------------\n");
+
+ for (int i = 0x400; i < 0x8000; i += 0x10) {
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ i, REG_RD(sc, i), REG_RD(sc, i + 0x4),
+ REG_RD(sc, i + 0x8), REG_RD(sc, i + 0xC));
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the mailbox queue registers. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_mq_regs(struct bce_softc *sc)
+{
+ BCE_PRINTF(
+ "----------------------------"
+ " MQ Regs "
+ "----------------------------\n");
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+
+ for (int i = 0x3c00; i < 0x4000; i += 0x10) {
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X 0x%08X 0x%08X\n",
+ i, REG_RD(sc, i), REG_RD(sc, i + 0x4),
+ REG_RD(sc, i + 0x8), REG_RD(sc, i + 0xC));
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the bootcode state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_bc_state(struct bce_softc *sc)
+{
+ u32 val;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " Bootcode State "
+ "----------------------------\n");
+
+ BCE_PRINTF("%s - bootcode version\n", sc->bce_bc_ver);
+
+ val = bce_shmem_rd(sc, BCE_BC_RESET_TYPE);
+ BCE_PRINTF("0x%08X - (0x%06X) reset_type\n",
+ val, BCE_BC_RESET_TYPE);
+
+ val = bce_shmem_rd(sc, BCE_BC_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) state\n",
+ val, BCE_BC_STATE);
+
+ val = bce_shmem_rd(sc, BCE_BC_STATE_CONDITION);
+ BCE_PRINTF("0x%08X - (0x%06X) condition\n",
+ val, BCE_BC_STATE_CONDITION);
+
+ val = bce_shmem_rd(sc, BCE_BC_STATE_DEBUG_CMD);
+ BCE_PRINTF("0x%08X - (0x%06X) debug_cmd\n",
+ val, BCE_BC_STATE_DEBUG_CMD);
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the TXP processor state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_txp_state(struct bce_softc *sc, int regs)
+{
+ u32 val;
+ u32 fw_version[3];
+
+ BCE_PRINTF(
+ "----------------------------"
+ " TXP State "
+ "----------------------------\n");
+
+ for (int i = 0; i < 3; i++)
+ fw_version[i] = htonl(REG_RD_IND(sc,
+ (BCE_TXP_SCRATCH + 0x10 + i * 4)));
+ BCE_PRINTF("Firmware version - %s\n", (char *) fw_version);
+
+ val = REG_RD_IND(sc, BCE_TXP_CPU_MODE);
+ BCE_PRINTF("0x%08X - (0x%06X) txp_cpu_mode\n",
+ val, BCE_TXP_CPU_MODE);
+
+ val = REG_RD_IND(sc, BCE_TXP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) txp_cpu_state\n",
+ val, BCE_TXP_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_TXP_CPU_EVENT_MASK);
+ BCE_PRINTF("0x%08X - (0x%06X) txp_cpu_event_mask\n",
+ val, BCE_TXP_CPU_EVENT_MASK);
+
+ if (regs) {
+ BCE_PRINTF(
+ "----------------------------"
+ " Register Dump "
+ "----------------------------\n");
+
+ for (int i = BCE_TXP_CPU_MODE; i < 0x68000; i += 0x10) {
+ /* Skip the big blank spaces */
+ if (i < 0x454000 && i > 0x5ffff)
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X "
+ "0x%08X 0x%08X\n", i,
+ REG_RD_IND(sc, i),
+ REG_RD_IND(sc, i + 0x4),
+ REG_RD_IND(sc, i + 0x8),
+ REG_RD_IND(sc, i + 0xC));
+ }
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the RXP processor state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_rxp_state(struct bce_softc *sc, int regs)
+{
+ u32 val;
+ u32 fw_version[3];
+
+ BCE_PRINTF(
+ "----------------------------"
+ " RXP State "
+ "----------------------------\n");
+
+ for (int i = 0; i < 3; i++)
+ fw_version[i] = htonl(REG_RD_IND(sc,
+ (BCE_RXP_SCRATCH + 0x10 + i * 4)));
+
+ BCE_PRINTF("Firmware version - %s\n", (char *) fw_version);
+
+ val = REG_RD_IND(sc, BCE_RXP_CPU_MODE);
+ BCE_PRINTF("0x%08X - (0x%06X) rxp_cpu_mode\n",
+ val, BCE_RXP_CPU_MODE);
+
+ val = REG_RD_IND(sc, BCE_RXP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) rxp_cpu_state\n",
+ val, BCE_RXP_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_RXP_CPU_EVENT_MASK);
+ BCE_PRINTF("0x%08X - (0x%06X) rxp_cpu_event_mask\n",
+ val, BCE_RXP_CPU_EVENT_MASK);
+
+ if (regs) {
+ BCE_PRINTF(
+ "----------------------------"
+ " Register Dump "
+ "----------------------------\n");
+
+ for (int i = BCE_RXP_CPU_MODE; i < 0xe8fff; i += 0x10) {
+ /* Skip the big blank sapces */
+ if (i < 0xc5400 && i > 0xdffff)
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X "
+ "0x%08X 0x%08X\n", i,
+ REG_RD_IND(sc, i),
+ REG_RD_IND(sc, i + 0x4),
+ REG_RD_IND(sc, i + 0x8),
+ REG_RD_IND(sc, i + 0xC));
+ }
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the TPAT processor state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_tpat_state(struct bce_softc *sc, int regs)
+{
+ u32 val;
+ u32 fw_version[3];
+
+ BCE_PRINTF(
+ "----------------------------"
+ " TPAT State "
+ "----------------------------\n");
+
+ for (int i = 0; i < 3; i++)
+ fw_version[i] = htonl(REG_RD_IND(sc,
+ (BCE_TPAT_SCRATCH + 0x410 + i * 4)));
+
+ BCE_PRINTF("Firmware version - %s\n", (char *) fw_version);
+
+ val = REG_RD_IND(sc, BCE_TPAT_CPU_MODE);
+ BCE_PRINTF("0x%08X - (0x%06X) tpat_cpu_mode\n",
+ val, BCE_TPAT_CPU_MODE);
+
+ val = REG_RD_IND(sc, BCE_TPAT_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) tpat_cpu_state\n",
+ val, BCE_TPAT_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_TPAT_CPU_EVENT_MASK);
+ BCE_PRINTF("0x%08X - (0x%06X) tpat_cpu_event_mask\n",
+ val, BCE_TPAT_CPU_EVENT_MASK);
+
+ if (regs) {
+ BCE_PRINTF(
+ "----------------------------"
+ " Register Dump "
+ "----------------------------\n");
+
+ for (int i = BCE_TPAT_CPU_MODE; i < 0xa3fff; i += 0x10) {
+ /* Skip the big blank spaces */
+ if (i < 0x854000 && i > 0x9ffff)
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X "
+ "0x%08X 0x%08X\n", i,
+ REG_RD_IND(sc, i),
+ REG_RD_IND(sc, i + 0x4),
+ REG_RD_IND(sc, i + 0x8),
+ REG_RD_IND(sc, i + 0xC));
+ }
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the Command Procesor (CP) state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_cp_state(struct bce_softc *sc, int regs)
+{
+ u32 val;
+ u32 fw_version[3];
+
+ BCE_PRINTF(
+ "----------------------------"
+ " CP State "
+ "----------------------------\n");
+
+ for (int i = 0; i < 3; i++)
+ fw_version[i] = htonl(REG_RD_IND(sc,
+ (BCE_CP_SCRATCH + 0x10 + i * 4)));
+
+ BCE_PRINTF("Firmware version - %s\n", (char *) fw_version);
+
+ val = REG_RD_IND(sc, BCE_CP_CPU_MODE);
+ BCE_PRINTF("0x%08X - (0x%06X) cp_cpu_mode\n",
+ val, BCE_CP_CPU_MODE);
+
+ val = REG_RD_IND(sc, BCE_CP_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) cp_cpu_state\n",
+ val, BCE_CP_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_CP_CPU_EVENT_MASK);
+ BCE_PRINTF("0x%08X - (0x%06X) cp_cpu_event_mask\n", val,
+ BCE_CP_CPU_EVENT_MASK);
+
+ if (regs) {
+ BCE_PRINTF(
+ "----------------------------"
+ " Register Dump "
+ "----------------------------\n");
+
+ for (int i = BCE_CP_CPU_MODE; i < 0x1aa000; i += 0x10) {
+ /* Skip the big blank spaces */
+ if (i < 0x185400 && i > 0x19ffff)
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X "
+ "0x%08X 0x%08X\n", i,
+ REG_RD_IND(sc, i),
+ REG_RD_IND(sc, i + 0x4),
+ REG_RD_IND(sc, i + 0x8),
+ REG_RD_IND(sc, i + 0xC));
+ }
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the Completion Procesor (COM) state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_com_state(struct bce_softc *sc, int regs)
+{
+ u32 val;
+ u32 fw_version[4];
+
+ BCE_PRINTF(
+ "----------------------------"
+ " COM State "
+ "----------------------------\n");
+
+ for (int i = 0; i < 3; i++)
+ fw_version[i] = htonl(REG_RD_IND(sc,
+ (BCE_COM_SCRATCH + 0x10 + i * 4)));
+
+ BCE_PRINTF("Firmware version - %s\n", (char *) fw_version);
+
+ val = REG_RD_IND(sc, BCE_COM_CPU_MODE);
+ BCE_PRINTF("0x%08X - (0x%06X) com_cpu_mode\n",
+ val, BCE_COM_CPU_MODE);
+
+ val = REG_RD_IND(sc, BCE_COM_CPU_STATE);
+ BCE_PRINTF("0x%08X - (0x%06X) com_cpu_state\n",
+ val, BCE_COM_CPU_STATE);
+
+ val = REG_RD_IND(sc, BCE_COM_CPU_EVENT_MASK);
+ BCE_PRINTF("0x%08X - (0x%06X) com_cpu_event_mask\n", val,
+ BCE_COM_CPU_EVENT_MASK);
+
+ if (regs) {
+ BCE_PRINTF(
+ "----------------------------"
+ " Register Dump "
+ "----------------------------\n");
+
+ for (int i = BCE_COM_CPU_MODE; i < 0x1053e8; i += 0x10) {
+ BCE_PRINTF("0x%04X: 0x%08X 0x%08X "
+ "0x%08X 0x%08X\n", i,
+ REG_RD_IND(sc, i),
+ REG_RD_IND(sc, i + 0x4),
+ REG_RD_IND(sc, i + 0x8),
+ REG_RD_IND(sc, i + 0xC));
+ }
+ }
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the Receive Virtual 2 Physical (RV2P) state. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_dump_rv2p_state(struct bce_softc *sc)
+{
+ u32 val, pc1, pc2, fw_ver_high, fw_ver_low;
+
+ BCE_PRINTF(
+ "----------------------------"
+ " RV2P State "
+ "----------------------------\n");
+
+ /* Stall the RV2P processors. */
+ val = REG_RD_IND(sc, BCE_RV2P_CONFIG);
+ val |= BCE_RV2P_CONFIG_STALL_PROC1 | BCE_RV2P_CONFIG_STALL_PROC2;
+ REG_WR_IND(sc, BCE_RV2P_CONFIG, val);
+
+ /* Read the firmware version. */
+ val = 0x00000001;
+ REG_WR_IND(sc, BCE_RV2P_PROC1_ADDR_CMD, val);
+ fw_ver_low = REG_RD_IND(sc, BCE_RV2P_INSTR_LOW);
+ fw_ver_high = REG_RD_IND(sc, BCE_RV2P_INSTR_HIGH) &
+ BCE_RV2P_INSTR_HIGH_HIGH;
+ BCE_PRINTF("RV2P1 Firmware version - 0x%08X:0x%08X\n",
+ fw_ver_high, fw_ver_low);
+
+ val = 0x00000001;
+ REG_WR_IND(sc, BCE_RV2P_PROC2_ADDR_CMD, val);
+ fw_ver_low = REG_RD_IND(sc, BCE_RV2P_INSTR_LOW);
+ fw_ver_high = REG_RD_IND(sc, BCE_RV2P_INSTR_HIGH) &
+ BCE_RV2P_INSTR_HIGH_HIGH;
+ BCE_PRINTF("RV2P2 Firmware version - 0x%08X:0x%08X\n",
+ fw_ver_high, fw_ver_low);
+
+ /* Resume the RV2P processors. */
+ val = REG_RD_IND(sc, BCE_RV2P_CONFIG);
+ val &= ~(BCE_RV2P_CONFIG_STALL_PROC1 | BCE_RV2P_CONFIG_STALL_PROC2);
+ REG_WR_IND(sc, BCE_RV2P_CONFIG, val);
+
+ /* Fetch the program counter value. */
+ val = 0x68007800;
+ REG_WR_IND(sc, BCE_RV2P_DEBUG_VECT_PEEK, val);
+ val = REG_RD_IND(sc, BCE_RV2P_DEBUG_VECT_PEEK);
+ pc1 = (val & BCE_RV2P_DEBUG_VECT_PEEK_1_VALUE);
+ pc2 = (val & BCE_RV2P_DEBUG_VECT_PEEK_2_VALUE) >> 16;
+ BCE_PRINTF("0x%08X - RV2P1 program counter (1st read)\n", pc1);
+ BCE_PRINTF("0x%08X - RV2P2 program counter (1st read)\n", pc2);
+
+ /* Fetch the program counter value again to see if it is advancing. */
+ val = 0x68007800;
+ REG_WR_IND(sc, BCE_RV2P_DEBUG_VECT_PEEK, val);
+ val = REG_RD_IND(sc, BCE_RV2P_DEBUG_VECT_PEEK);
+ pc1 = (val & BCE_RV2P_DEBUG_VECT_PEEK_1_VALUE);
+ pc2 = (val & BCE_RV2P_DEBUG_VECT_PEEK_2_VALUE) >> 16;
+ BCE_PRINTF("0x%08X - RV2P1 program counter (2nd read)\n", pc1);
+ BCE_PRINTF("0x%08X - RV2P2 program counter (2nd read)\n", pc2);
+
+ BCE_PRINTF(
+ "----------------------------"
+ "----------------"
+ "----------------------------\n");
+}
+
+
+/****************************************************************************/
+/* Prints out the driver state and then enters the debugger. */
+/* */
+/* Returns: */
+/* Nothing. */
+/****************************************************************************/
+static __attribute__ ((noinline)) void
+bce_breakpoint(struct bce_softc *sc)
+{
+
+ /*
+ * Unreachable code to silence compiler warnings
+ * about unused functions.
+ */
+ if (0) {
+ bce_freeze_controller(sc);
+ bce_unfreeze_controller(sc);
+ bce_dump_enet(sc, NULL);
+ bce_dump_txbd(sc, 0, NULL);
+ bce_dump_rxbd(sc, 0, NULL);
+ bce_dump_tx_mbuf_chain(sc, 0, USABLE_TX_BD);
+ bce_dump_rx_mbuf_chain(sc, 0, USABLE_RX_BD);
+ bce_dump_l2fhdr(sc, 0, NULL);
+ bce_dump_ctx(sc, RX_CID);
+ bce_dump_ftqs(sc);
+ bce_dump_tx_chain(sc, 0, USABLE_TX_BD);
+ bce_dump_rx_bd_chain(sc, 0, USABLE_RX_BD);
+ bce_dump_status_block(sc);
+ bce_dump_stats_block(sc);
+ bce_dump_driver_state(sc);
+ bce_dump_hw_state(sc);
+ bce_dump_bc_state(sc);
+ bce_dump_txp_state(sc, 0);
+ bce_dump_rxp_state(sc, 0);
+ bce_dump_tpat_state(sc, 0);
+ bce_dump_cp_state(sc, 0);
+ bce_dump_com_state(sc, 0);
+ bce_dump_rv2p_state(sc);
+
+#ifdef BCE_JUMBO_HDRSPLIT
+ bce_dump_pgbd(sc, 0, NULL);
+ bce_dump_pg_mbuf_chain(sc, 0, USABLE_PG_BD);
+ bce_dump_pg_chain(sc, 0, USABLE_PG_BD);
+#endif
+ }
+
+ bce_dump_status_block(sc);
+ bce_dump_driver_state(sc);
+
+ /* Call the debugger. */
+ breakpoint();
+
+ return;
+}
+#endif
+