diff options
Diffstat (limited to 'drivers/net/sfc')
-rw-r--r-- | drivers/net/sfc/Makefile | 4 | ||||
-rw-r--r-- | drivers/net/sfc/bitfield.h | 7 | ||||
-rw-r--r-- | drivers/net/sfc/boards.c | 9 | ||||
-rw-r--r-- | drivers/net/sfc/boards.h | 2 | ||||
-rw-r--r-- | drivers/net/sfc/efx.c | 88 | ||||
-rw-r--r-- | drivers/net/sfc/enum.h | 49 | ||||
-rw-r--r-- | drivers/net/sfc/ethtool.c | 259 | ||||
-rw-r--r-- | drivers/net/sfc/falcon.c | 97 | ||||
-rw-r--r-- | drivers/net/sfc/falcon.h | 5 | ||||
-rw-r--r-- | drivers/net/sfc/falcon_hwdefs.h | 20 | ||||
-rw-r--r-- | drivers/net/sfc/falcon_io.h | 29 | ||||
-rw-r--r-- | drivers/net/sfc/falcon_xmac.c | 92 | ||||
-rw-r--r-- | drivers/net/sfc/mdio_10g.c | 78 | ||||
-rw-r--r-- | drivers/net/sfc/mdio_10g.h | 24 | ||||
-rw-r--r-- | drivers/net/sfc/net_driver.h | 72 | ||||
-rw-r--r-- | drivers/net/sfc/rx.c | 59 | ||||
-rw-r--r-- | drivers/net/sfc/selftest.c | 719 | ||||
-rw-r--r-- | drivers/net/sfc/selftest.h | 50 | ||||
-rw-r--r-- | drivers/net/sfc/sfe4001.c | 28 | ||||
-rw-r--r-- | drivers/net/sfc/tenxpress.c | 93 | ||||
-rw-r--r-- | drivers/net/sfc/tx.c | 669 | ||||
-rw-r--r-- | drivers/net/sfc/workarounds.h | 2 | ||||
-rw-r--r-- | drivers/net/sfc/xfp_phy.c | 38 |
23 files changed, 2300 insertions, 193 deletions
diff --git a/drivers/net/sfc/Makefile b/drivers/net/sfc/Makefile index 0f023447eafd..1d2daeec7ac1 100644 --- a/drivers/net/sfc/Makefile +++ b/drivers/net/sfc/Makefile @@ -1,5 +1,5 @@ sfc-y += efx.o falcon.o tx.o rx.o falcon_xmac.o \ - i2c-direct.o ethtool.o xfp_phy.o mdio_10g.o \ - tenxpress.o boards.o sfe4001.o + i2c-direct.o selftest.o ethtool.o xfp_phy.o \ + mdio_10g.o tenxpress.o boards.o sfe4001.o obj-$(CONFIG_SFC) += sfc.o diff --git a/drivers/net/sfc/bitfield.h b/drivers/net/sfc/bitfield.h index 2806201644cc..2c79d27404e0 100644 --- a/drivers/net/sfc/bitfield.h +++ b/drivers/net/sfc/bitfield.h @@ -483,7 +483,7 @@ typedef union efx_oword { #endif #define EFX_SET_OWORD_FIELD_VER(efx, oword, field, value) do { \ - if (FALCON_REV(efx) >= FALCON_REV_B0) { \ + if (falcon_rev(efx) >= FALCON_REV_B0) { \ EFX_SET_OWORD_FIELD((oword), field##_B0, (value)); \ } else { \ EFX_SET_OWORD_FIELD((oword), field##_A1, (value)); \ @@ -491,7 +491,7 @@ typedef union efx_oword { } while (0) #define EFX_QWORD_FIELD_VER(efx, qword, field) \ - (FALCON_REV(efx) >= FALCON_REV_B0 ? \ + (falcon_rev(efx) >= FALCON_REV_B0 ? \ EFX_QWORD_FIELD((qword), field##_B0) : \ EFX_QWORD_FIELD((qword), field##_A1)) @@ -501,8 +501,5 @@ typedef union efx_oword { #define DMA_ADDR_T_WIDTH (8 * sizeof(dma_addr_t)) #define EFX_DMA_TYPE_WIDTH(width) \ (((width) < DMA_ADDR_T_WIDTH) ? (width) : DMA_ADDR_T_WIDTH) -#define EFX_DMA_MAX_MASK ((DMA_ADDR_T_WIDTH == 64) ? \ - ~((u64) 0) : ~((u32) 0)) -#define EFX_DMA_MASK(mask) ((mask) & EFX_DMA_MAX_MASK) #endif /* EFX_BITFIELD_H */ diff --git a/drivers/net/sfc/boards.c b/drivers/net/sfc/boards.c index eecaa6d58584..7fc0328dc055 100644 --- a/drivers/net/sfc/boards.c +++ b/drivers/net/sfc/boards.c @@ -27,10 +27,8 @@ static void blink_led_timer(unsigned long context) struct efx_blinker *bl = &efx->board_info.blinker; efx->board_info.set_fault_led(efx, bl->state); bl->state = !bl->state; - if (bl->resubmit) { - bl->timer.expires = jiffies + BLINK_INTERVAL; - add_timer(&bl->timer); - } + if (bl->resubmit) + mod_timer(&bl->timer, jiffies + BLINK_INTERVAL); } static void board_blink(struct efx_nic *efx, int blink) @@ -44,8 +42,7 @@ static void board_blink(struct efx_nic *efx, int blink) blinker->state = 0; setup_timer(&blinker->timer, blink_led_timer, (unsigned long)efx); - blinker->timer.expires = jiffies + BLINK_INTERVAL; - add_timer(&blinker->timer); + mod_timer(&blinker->timer, jiffies + BLINK_INTERVAL); } else { blinker->resubmit = 0; if (blinker->timer.function) diff --git a/drivers/net/sfc/boards.h b/drivers/net/sfc/boards.h index f56341d428e1..695764dc2e64 100644 --- a/drivers/net/sfc/boards.h +++ b/drivers/net/sfc/boards.h @@ -22,5 +22,7 @@ enum efx_board_type { extern int efx_set_board_info(struct efx_nic *efx, u16 revision_info); extern int sfe4001_poweron(struct efx_nic *efx); extern void sfe4001_poweroff(struct efx_nic *efx); +/* Are we putting the PHY into flash config mode */ +extern unsigned int sfe4001_phy_flash_cfg; #endif diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c index 59edcf793c19..449760642e31 100644 --- a/drivers/net/sfc/efx.c +++ b/drivers/net/sfc/efx.c @@ -199,11 +199,12 @@ static inline int efx_process_channel(struct efx_channel *channel, int rx_quota) */ static inline void efx_channel_processed(struct efx_channel *channel) { - /* Write to EVQ_RPTR_REG. If a new event arrived in a race - * with finishing processing, a new interrupt will be raised. - */ + /* The interrupt handler for this channel may set work_pending + * as soon as we acknowledge the events we've seen. Make sure + * it's cleared before then. */ channel->work_pending = 0; - smp_wmb(); /* Ensure channel updated before any new interrupt. */ + smp_wmb(); + falcon_eventq_read_ack(channel); } @@ -265,7 +266,7 @@ void efx_process_channel_now(struct efx_channel *channel) napi_disable(&channel->napi_str); /* Poll the channel */ - (void) efx_process_channel(channel, efx->type->evq_size); + efx_process_channel(channel, efx->type->evq_size); /* Ack the eventq. This may cause an interrupt to be generated * when they are reenabled */ @@ -317,26 +318,6 @@ static void efx_remove_eventq(struct efx_channel *channel) * *************************************************************************/ -/* Setup per-NIC RX buffer parameters. - * Calculate the rx buffer allocation parameters required to support - * the current MTU, including padding for header alignment and overruns. - */ -static void efx_calc_rx_buffer_params(struct efx_nic *efx) -{ - unsigned int order, len; - - len = (max(EFX_PAGE_IP_ALIGN, NET_IP_ALIGN) + - EFX_MAX_FRAME_LEN(efx->net_dev->mtu) + - efx->type->rx_buffer_padding); - - /* Calculate page-order */ - for (order = 0; ((1u << order) * PAGE_SIZE) < len; ++order) - ; - - efx->rx_buffer_len = len; - efx->rx_buffer_order = order; -} - static int efx_probe_channel(struct efx_channel *channel) { struct efx_tx_queue *tx_queue; @@ -387,7 +368,14 @@ static int efx_init_channels(struct efx_nic *efx) struct efx_channel *channel; int rc = 0; - efx_calc_rx_buffer_params(efx); + /* Calculate the rx buffer allocation parameters required to + * support the current MTU, including padding for header + * alignment and overruns. + */ + efx->rx_buffer_len = (max(EFX_PAGE_IP_ALIGN, NET_IP_ALIGN) + + EFX_MAX_FRAME_LEN(efx->net_dev->mtu) + + efx->type->rx_buffer_padding); + efx->rx_buffer_order = get_order(efx->rx_buffer_len); /* Initialise the channels */ efx_for_each_channel(channel, efx) { @@ -440,9 +428,12 @@ static void efx_start_channel(struct efx_channel *channel) netif_napi_add(channel->napi_dev, &channel->napi_str, efx_poll, napi_weight); + /* The interrupt handler for this channel may set work_pending + * as soon as we enable it. Make sure it's cleared before + * then. Similarly, make sure it sees the enabled flag set. */ channel->work_pending = 0; channel->enabled = 1; - smp_wmb(); /* ensure channel updated before first interrupt */ + smp_wmb(); napi_enable(&channel->napi_str); @@ -704,7 +695,7 @@ static void efx_stop_port(struct efx_nic *efx) mutex_unlock(&efx->mac_lock); /* Serialise against efx_set_multicast_list() */ - if (NET_DEV_REGISTERED(efx)) { + if (efx_dev_registered(efx)) { netif_tx_lock_bh(efx->net_dev); netif_tx_unlock_bh(efx->net_dev); } @@ -791,22 +782,23 @@ static int efx_init_io(struct efx_nic *efx) efx->membase = ioremap_nocache(efx->membase_phys, efx->type->mem_map_size); if (!efx->membase) { - EFX_ERR(efx, "could not map memory BAR %d at %lx+%x\n", - efx->type->mem_bar, efx->membase_phys, + EFX_ERR(efx, "could not map memory BAR %d at %llx+%x\n", + efx->type->mem_bar, + (unsigned long long)efx->membase_phys, efx->type->mem_map_size); rc = -ENOMEM; goto fail4; } - EFX_LOG(efx, "memory BAR %u at %lx+%x (virtual %p)\n", - efx->type->mem_bar, efx->membase_phys, efx->type->mem_map_size, - efx->membase); + EFX_LOG(efx, "memory BAR %u at %llx+%x (virtual %p)\n", + efx->type->mem_bar, (unsigned long long)efx->membase_phys, + efx->type->mem_map_size, efx->membase); return 0; fail4: release_mem_region(efx->membase_phys, efx->type->mem_map_size); fail3: - efx->membase_phys = 0UL; + efx->membase_phys = 0; fail2: pci_disable_device(efx->pci_dev); fail1: @@ -824,7 +816,7 @@ static void efx_fini_io(struct efx_nic *efx) if (efx->membase_phys) { pci_release_region(efx->pci_dev, efx->type->mem_bar); - efx->membase_phys = 0UL; + efx->membase_phys = 0; } pci_disable_device(efx->pci_dev); @@ -1043,7 +1035,7 @@ static void efx_start_all(struct efx_nic *efx) return; if ((efx->state != STATE_RUNNING) && (efx->state != STATE_INIT)) return; - if (NET_DEV_REGISTERED(efx) && !netif_running(efx->net_dev)) + if (efx_dev_registered(efx) && !netif_running(efx->net_dev)) return; /* Mark the port as enabled so port reconfigurations can start, then @@ -1073,9 +1065,8 @@ static void efx_flush_all(struct efx_nic *efx) cancel_delayed_work_sync(&efx->monitor_work); /* Ensure that all RX slow refills are complete. */ - efx_for_each_rx_queue(rx_queue, efx) { + efx_for_each_rx_queue(rx_queue, efx) cancel_delayed_work_sync(&rx_queue->work); - } /* Stop scheduled port reconfigurations */ cancel_work_sync(&efx->reconfigure_work); @@ -1101,9 +1092,10 @@ static void efx_stop_all(struct efx_nic *efx) falcon_disable_interrupts(efx); if (efx->legacy_irq) synchronize_irq(efx->legacy_irq); - efx_for_each_channel_with_interrupt(channel, efx) + efx_for_each_channel_with_interrupt(channel, efx) { if (channel->irq) synchronize_irq(channel->irq); + } /* Stop all NAPI processing and synchronous rx refills */ efx_for_each_channel(channel, efx) @@ -1125,7 +1117,7 @@ static void efx_stop_all(struct efx_nic *efx) /* Stop the kernel transmit interface late, so the watchdog * timer isn't ticking over the flush */ efx_stop_queue(efx); - if (NET_DEV_REGISTERED(efx)) { + if (efx_dev_registered(efx)) { netif_tx_lock_bh(efx->net_dev); netif_tx_unlock_bh(efx->net_dev); } @@ -1344,13 +1336,17 @@ static int efx_net_stop(struct net_device *net_dev) return 0; } -/* Context: process, dev_base_lock held, non-blocking. */ +/* Context: process, dev_base_lock or RTNL held, non-blocking. */ static struct net_device_stats *efx_net_stats(struct net_device *net_dev) { struct efx_nic *efx = net_dev->priv; struct efx_mac_stats *mac_stats = &efx->mac_stats; struct net_device_stats *stats = &net_dev->stats; + /* Update stats if possible, but do not wait if another thread + * is updating them (or resetting the NIC); slightly stale + * stats are acceptable. + */ if (!spin_trylock(&efx->stats_lock)) return stats; if (efx->state == STATE_RUNNING) { @@ -1494,7 +1490,7 @@ static void efx_set_multicast_list(struct net_device *net_dev) static int efx_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) { - struct net_device *net_dev = (struct net_device *)ptr; + struct net_device *net_dev = ptr; if (net_dev->open == efx_net_open && event == NETDEV_CHANGENAME) { struct efx_nic *efx = net_dev->priv; @@ -1563,7 +1559,7 @@ static void efx_unregister_netdev(struct efx_nic *efx) efx_for_each_tx_queue(tx_queue, efx) efx_release_tx_buffers(tx_queue); - if (NET_DEV_REGISTERED(efx)) { + if (efx_dev_registered(efx)) { strlcpy(efx->name, pci_name(efx->pci_dev), sizeof(efx->name)); unregister_netdev(efx->net_dev); } @@ -1688,7 +1684,7 @@ static int efx_reset(struct efx_nic *efx) if (method == RESET_TYPE_DISABLE) { /* Reinitialise the device anyway so the driver unload sequence * can talk to the external SRAM */ - (void) falcon_init_nic(efx); + falcon_init_nic(efx); rc = -EIO; goto fail4; } @@ -1873,6 +1869,7 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type, tx_queue->queue = i; tx_queue->buffer = NULL; tx_queue->channel = &efx->channel[0]; /* for safety */ + tx_queue->tso_headers_free = NULL; } for (i = 0; i < EFX_MAX_RX_QUEUES; i++) { rx_queue = &efx->rx_queue[i]; @@ -2071,7 +2068,8 @@ static int __devinit efx_pci_probe(struct pci_dev *pci_dev, net_dev = alloc_etherdev(sizeof(*efx)); if (!net_dev) return -ENOMEM; - net_dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG | NETIF_F_HIGHDMA; + net_dev->features |= (NETIF_F_IP_CSUM | NETIF_F_SG | + NETIF_F_HIGHDMA | NETIF_F_TSO); if (lro) net_dev->features |= NETIF_F_LRO; efx = net_dev->priv; diff --git a/drivers/net/sfc/enum.h b/drivers/net/sfc/enum.h index 43663a4619da..c53290d08e2b 100644 --- a/drivers/net/sfc/enum.h +++ b/drivers/net/sfc/enum.h @@ -10,6 +10,55 @@ #ifndef EFX_ENUM_H #define EFX_ENUM_H +/** + * enum efx_loopback_mode - loopback modes + * @LOOPBACK_NONE: no loopback + * @LOOPBACK_XGMII: loopback within MAC at XGMII level + * @LOOPBACK_XGXS: loopback within MAC at XGXS level + * @LOOPBACK_XAUI: loopback within MAC at XAUI level + * @LOOPBACK_PHYXS: loopback within PHY at PHYXS level + * @LOOPBACK_PCS: loopback within PHY at PCS level + * @LOOPBACK_PMAPMD: loopback within PHY at PMAPMD level + * @LOOPBACK_NETWORK: reflecting loopback (even further than furthest!) + */ +/* Please keep in order and up-to-date w.r.t the following two #defines */ +enum efx_loopback_mode { + LOOPBACK_NONE = 0, + LOOPBACK_MAC = 1, + LOOPBACK_XGMII = 2, + LOOPBACK_XGXS = 3, + LOOPBACK_XAUI = 4, + LOOPBACK_PHY = 5, + LOOPBACK_PHYXS = 6, + LOOPBACK_PCS = 7, + LOOPBACK_PMAPMD = 8, + LOOPBACK_NETWORK = 9, + LOOPBACK_MAX +}; + +#define LOOPBACK_TEST_MAX LOOPBACK_PMAPMD + +extern const char *efx_loopback_mode_names[]; +#define LOOPBACK_MODE_NAME(mode) \ + STRING_TABLE_LOOKUP(mode, efx_loopback_mode) +#define LOOPBACK_MODE(efx) \ + LOOPBACK_MODE_NAME(efx->loopback_mode) + +/* These loopbacks occur within the controller */ +#define LOOPBACKS_10G_INTERNAL ((1 << LOOPBACK_XGMII)| \ + (1 << LOOPBACK_XGXS) | \ + (1 << LOOPBACK_XAUI)) + +#define LOOPBACK_MASK(_efx) \ + (1 << (_efx)->loopback_mode) + +#define LOOPBACK_INTERNAL(_efx) \ + ((LOOPBACKS_10G_INTERNAL & LOOPBACK_MASK(_efx)) ? 1 : 0) + +#define LOOPBACK_OUT_OF(_from, _to, _mask) \ + (((LOOPBACK_MASK(_from) & (_mask)) && \ + ((LOOPBACK_MASK(_to) & (_mask)) == 0)) ? 1 : 0) + /*****************************************************************************/ /** diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c index ad541badbd98..e2c75d101610 100644 --- a/drivers/net/sfc/ethtool.c +++ b/drivers/net/sfc/ethtool.c @@ -12,12 +12,26 @@ #include <linux/ethtool.h> #include <linux/rtnetlink.h> #include "net_driver.h" +#include "selftest.h" #include "efx.h" #include "ethtool.h" #include "falcon.h" #include "gmii.h" #include "mac.h" +const char *efx_loopback_mode_names[] = { + [LOOPBACK_NONE] = "NONE", + [LOOPBACK_MAC] = "MAC", + [LOOPBACK_XGMII] = "XGMII", + [LOOPBACK_XGXS] = "XGXS", + [LOOPBACK_XAUI] = "XAUI", + [LOOPBACK_PHY] = "PHY", + [LOOPBACK_PHYXS] = "PHY(XS)", + [LOOPBACK_PCS] = "PHY(PCS)", + [LOOPBACK_PMAPMD] = "PHY(PMAPMD)", + [LOOPBACK_NETWORK] = "NETWORK", +}; + static int efx_ethtool_set_tx_csum(struct net_device *net_dev, u32 enable); struct ethtool_string { @@ -217,23 +231,179 @@ static void efx_ethtool_get_drvinfo(struct net_device *net_dev, strlcpy(info->bus_info, pci_name(efx->pci_dev), sizeof(info->bus_info)); } +/** + * efx_fill_test - fill in an individual self-test entry + * @test_index: Index of the test + * @strings: Ethtool strings, or %NULL + * @data: Ethtool test results, or %NULL + * @test: Pointer to test result (used only if data != %NULL) + * @unit_format: Unit name format (e.g. "channel\%d") + * @unit_id: Unit id (e.g. 0 for "channel0") + * @test_format: Test name format (e.g. "loopback.\%s.tx.sent") + * @test_id: Test id (e.g. "PHY" for "loopback.PHY.tx_sent") + * + * Fill in an individual self-test entry. + */ +static void efx_fill_test(unsigned int test_index, + struct ethtool_string *strings, u64 *data, + int *test, const char *unit_format, int unit_id, + const char *test_format, const char *test_id) +{ + struct ethtool_string unit_str, test_str; + + /* Fill data value, if applicable */ + if (data) + data[test_index] = *test; + + /* Fill string, if applicable */ + if (strings) { + snprintf(unit_str.name, sizeof(unit_str.name), + unit_format, unit_id); + snprintf(test_str.name, sizeof(test_str.name), + test_format, test_id); + snprintf(strings[test_index].name, + sizeof(strings[test_index].name), + "%-9s%-17s", unit_str.name, test_str.name); + } +} + +#define EFX_PORT_NAME "port%d", 0 +#define EFX_CHANNEL_NAME(_channel) "channel%d", _channel->channel +#define EFX_TX_QUEUE_NAME(_tx_queue) "txq%d", _tx_queue->queue +#define EFX_RX_QUEUE_NAME(_rx_queue) "rxq%d", _rx_queue->queue +#define EFX_LOOPBACK_NAME(_mode, _counter) \ + "loopback.%s." _counter, LOOPBACK_MODE_NAME(mode) + +/** + * efx_fill_loopback_test - fill in a block of loopback self-test entries + * @efx: Efx NIC + * @lb_tests: Efx loopback self-test results structure + * @mode: Loopback test mode + * @test_index: Starting index of the test + * @strings: Ethtool strings, or %NULL + * @data: Ethtool test results, or %NULL + */ +static int efx_fill_loopback_test(struct efx_nic *efx, + struct efx_loopback_self_tests *lb_tests, + enum efx_loopback_mode mode, + unsigned int test_index, + struct ethtool_string *strings, u64 *data) +{ + struct efx_tx_queue *tx_queue; + + efx_for_each_tx_queue(tx_queue, efx) { + efx_fill_test(test_index++, strings, data, + &lb_tests->tx_sent[tx_queue->queue], + EFX_TX_QUEUE_NAME(tx_queue), + EFX_LOOPBACK_NAME(mode, "tx_sent")); + efx_fill_test(test_index++, strings, data, + &lb_tests->tx_done[tx_queue->queue], + EFX_TX_QUEUE_NAME(tx_queue), + EFX_LOOPBACK_NAME(mode, "tx_done")); + } + efx_fill_test(test_index++, strings, data, + &lb_tests->rx_good, + EFX_PORT_NAME, + EFX_LOOPBACK_NAME(mode, "rx_good")); + efx_fill_test(test_index++, strings, data, + &lb_tests->rx_bad, + EFX_PORT_NAME, + EFX_LOOPBACK_NAME(mode, "rx_bad")); + + return test_index; +} + +/** + * efx_ethtool_fill_self_tests - get self-test details + * @efx: Efx NIC + * @tests: Efx self-test results structure, or %NULL + * @strings: Ethtool strings, or %NULL + * @data: Ethtool test results, or %NULL + */ +static int efx_ethtool_fill_self_tests(struct efx_nic *efx, + struct efx_self_tests *tests, + struct ethtool_string *strings, + u64 *data) +{ + struct efx_channel *channel; + unsigned int n = 0; + enum efx_loopback_mode mode; + + /* Interrupt */ + efx_fill_test(n++, strings, data, &tests->interrupt, + "core", 0, "interrupt", NULL); + + /* Event queues */ + efx_for_each_channel(channel, efx) { + efx_fill_test(n++, strings, data, + &tests->eventq_dma[channel->channel], + EFX_CHANNEL_NAME(channel), + "eventq.dma", NULL); + efx_fill_test(n++, strings, data, + &tests->eventq_int[channel->channel], + EFX_CHANNEL_NAME(channel), + "eventq.int", NULL); + efx_fill_test(n++, strings, data, + &tests->eventq_poll[channel->channel], + EFX_CHANNEL_NAME(channel), + "eventq.poll", NULL); + } + + /* PHY presence */ + efx_fill_test(n++, strings, data, &tests->phy_ok, + EFX_PORT_NAME, "phy_ok", NULL); + + /* Loopback tests */ + efx_fill_test(n++, strings, data, &tests->loopback_speed, + EFX_PORT_NAME, "loopback.speed", NULL); + efx_fill_test(n++, strings, data, &tests->loopback_full_duplex, + EFX_PORT_NAME, "loopback.full_duplex", NULL); + for (mode = LOOPBACK_NONE; mode < LOOPBACK_TEST_MAX; mode++) { + if (!(efx->loopback_modes & (1 << mode))) + continue; + n = efx_fill_loopback_test(efx, + &tests->loopback[mode], mode, n, + strings, data); + } + + return n; +} + static int efx_ethtool_get_stats_count(struct net_device *net_dev) { return EFX_ETHTOOL_NUM_STATS; } +static int efx_ethtool_self_test_count(struct net_device *net_dev) +{ + struct efx_nic *efx = net_dev->priv; + + return efx_ethtool_fill_self_tests(efx, NULL, NULL, NULL); +} + static void efx_ethtool_get_strings(struct net_device *net_dev, u32 string_set, u8 *strings) { + struct efx_nic *efx = net_dev->priv; struct ethtool_string *ethtool_strings = (struct ethtool_string *)strings; int i; - if (string_set == ETH_SS_STATS) + switch (string_set) { + case ETH_SS_STATS: for (i = 0; i < EFX_ETHTOOL_NUM_STATS; i++) strncpy(ethtool_strings[i].name, efx_ethtool_stats[i].name, sizeof(ethtool_strings[i].name)); + break; + case ETH_SS_TEST: + efx_ethtool_fill_self_tests(efx, NULL, + ethtool_strings, NULL); + break; + default: + /* No other string sets */ + break; + } } static void efx_ethtool_get_stats(struct net_device *net_dev, @@ -272,6 +442,22 @@ static void efx_ethtool_get_stats(struct net_device *net_dev, } } +static int efx_ethtool_set_tso(struct net_device *net_dev, u32 enable) +{ + int rc; + + /* Our TSO requires TX checksumming, so force TX checksumming + * on when TSO is enabled. + */ + if (enable) { + rc = efx_ethtool_set_tx_csum(net_dev, 1); + if (rc) + return rc; + } + + return ethtool_op_set_tso(net_dev, enable); +} + static int efx_ethtool_set_tx_csum(struct net_device *net_dev, u32 enable) { struct efx_nic *efx = net_dev->priv; @@ -283,6 +469,15 @@ static int efx_ethtool_set_tx_csum(struct net_device *net_dev, u32 enable) efx_flush_queues(efx); + /* Our TSO requires TX checksumming, so disable TSO when + * checksumming is disabled + */ + if (!enable) { + rc = efx_ethtool_set_tso(net_dev, 0); + if (rc) + return rc; + } + return 0; } @@ -305,6 +500,64 @@ static u32 efx_ethtool_get_rx_csum(struct net_device *net_dev) return efx->rx_checksum_enabled; } +static void efx_ethtool_self_test(struct net_device *net_dev, + struct ethtool_test *test, u64 *data) +{ + struct efx_nic *efx = net_dev->priv; + struct efx_self_tests efx_tests; + int offline, already_up; + int rc; + + ASSERT_RTNL(); + if (efx->state != STATE_RUNNING) { + rc = -EIO; + goto fail1; + } + + /* We need rx buffers and interrupts. */ + already_up = (efx->net_dev->flags & IFF_UP); + if (!already_up) { + rc = dev_open(efx->net_dev); + if (rc) { + EFX_ERR(efx, "failed opening device.\n"); + goto fail2; + } + } + + memset(&efx_tests, 0, sizeof(efx_tests)); + offline = (test->flags & ETH_TEST_FL_OFFLINE); + + /* Perform online self tests first */ + rc = efx_online_test(efx, &efx_tests); + if (rc) + goto out; + + /* Perform offline tests only if online tests passed */ + if (offline) { + /* Stop the kernel from sending packets during the test. */ + efx_stop_queue(efx); + rc = efx_flush_queues(efx); + if (!rc) + rc = efx_offline_test(efx, &efx_tests, + efx->loopback_modes); + efx_wake_queue(efx); + } + + out: + if (!already_up) + dev_close(efx->net_dev); + + EFX_LOG(efx, "%s all %sline self-tests\n", + rc == 0 ? "passed" : "failed", offline ? "off" : "on"); + + fail2: + fail1: + /* Fill ethtool results structures */ + efx_ethtool_fill_self_tests(efx, &efx_tests, NULL, data); + if (rc) + test->flags |= ETH_TEST_FL_FAILED; +} + /* Restart autonegotiation */ static int efx_ethtool_nway_reset(struct net_device *net_dev) { @@ -451,8 +704,12 @@ struct ethtool_ops efx_ethtool_ops = { .set_tx_csum = efx_ethtool_set_tx_csum, .get_sg = ethtool_op_get_sg, .set_sg = ethtool_op_set_sg, + .get_tso = ethtool_op_get_tso, + .set_tso = efx_ethtool_set_tso, .get_flags = ethtool_op_get_flags, .set_flags = ethtool_op_set_flags, + .self_test_count = efx_ethtool_self_test_count, + .self_test = efx_ethtool_self_test, .get_strings = efx_ethtool_get_strings, .phys_id = efx_ethtool_phys_id, .get_stats_count = efx_ethtool_get_stats_count, diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c index 46db549ce580..790db89db345 100644 --- a/drivers/net/sfc/falcon.c +++ b/drivers/net/sfc/falcon.c @@ -116,17 +116,8 @@ MODULE_PARM_DESC(rx_xon_thresh_bytes, "RX fifo XON threshold"); ************************************************************************** */ -/* DMA address mask (up to 46-bit, avoiding compiler warnings) - * - * Note that it is possible to have a platform with 64-bit longs and - * 32-bit DMA addresses, or vice versa. EFX_DMA_MASK takes care of the - * platform DMA mask. - */ -#if BITS_PER_LONG == 64 -#define FALCON_DMA_MASK EFX_DMA_MASK(0x00003fffffffffffUL) -#else -#define FALCON_DMA_MASK EFX_DMA_MASK(0x00003fffffffffffULL) -#endif +/* DMA address mask */ +#define FALCON_DMA_MASK DMA_BIT_MASK(46) /* TX DMA length mask (13-bit) */ #define FALCON_TX_DMA_MASK (4096 - 1) @@ -145,7 +136,7 @@ MODULE_PARM_DESC(rx_xon_thresh_bytes, "RX fifo XON threshold"); #define PCI_EXP_LNKSTA_LNK_WID_LBN 4 #define FALCON_IS_DUAL_FUNC(efx) \ - (FALCON_REV(efx) < FALCON_REV_B0) + (falcon_rev(efx) < FALCON_REV_B0) /************************************************************************** * @@ -465,7 +456,7 @@ int falcon_init_tx(struct efx_tx_queue *tx_queue) TX_DESCQ_TYPE, 0, TX_NON_IP_DROP_DIS_B0, 1); - if (FALCON_REV(efx) >= FALCON_REV_B0) { + if (falcon_rev(efx) >= FALCON_REV_B0) { int csum = !(efx->net_dev->features & NETIF_F_IP_CSUM); EFX_SET_OWORD_FIELD(tx_desc_ptr, TX_IP_CHKSM_DIS_B0, csum); EFX_SET_OWORD_FIELD(tx_desc_ptr, TX_TCP_CHKSM_DIS_B0, csum); @@ -474,7 +465,7 @@ int falcon_init_tx(struct efx_tx_queue *tx_queue) falcon_write_table(efx, &tx_desc_ptr, efx->type->txd_ptr_tbl_base, tx_queue->queue); - if (FALCON_REV(efx) < FALCON_REV_B0) { + if (falcon_rev(efx) < FALCON_REV_B0) { efx_oword_t reg; BUG_ON(tx_queue->queue >= 128); /* HW limit */ @@ -635,7 +626,7 @@ int falcon_init_rx(struct efx_rx_queue *rx_queue) efx_oword_t rx_desc_ptr; struct efx_nic *efx = rx_queue->efx; int rc; - int is_b0 = FALCON_REV(efx) >= FALCON_REV_B0; + int is_b0 = falcon_rev(efx) >= FALCON_REV_B0; int iscsi_digest_en = is_b0; EFX_LOG(efx, "RX queue %d ring in special buffers %d-%d\n", @@ -742,8 +733,10 @@ void falcon_fini_rx(struct efx_rx_queue *rx_queue) continue; break; } - if (rc) + if (rc) { EFX_ERR(efx, "failed to flush rx queue %d\n", rx_queue->queue); + efx_schedule_reset(efx, RESET_TYPE_INVISIBLE); + } /* Remove RX descriptor ring from card */ EFX_ZERO_OWORD(rx_desc_ptr); @@ -822,10 +815,10 @@ static inline void falcon_handle_tx_event(struct efx_channel *channel, tx_ev_q_label = EFX_QWORD_FIELD(*event, TX_EV_Q_LABEL); tx_queue = &efx->tx_queue[tx_ev_q_label]; - if (NET_DEV_REGISTERED(efx)) + if (efx_dev_registered(efx)) netif_tx_lock(efx->net_dev); falcon_notify_tx_desc(tx_queue); - if (NET_DEV_REGISTERED(efx)) + if (efx_dev_registered(efx)) netif_tx_unlock(efx->net_dev); } else if (EFX_QWORD_FIELD(*event, TX_EV_PKT_ERR) && EFX_WORKAROUND_10727(efx)) { @@ -884,7 +877,7 @@ static void falcon_handle_rx_not_ok(struct efx_rx_queue *rx_queue, RX_EV_TCP_UDP_CHKSUM_ERR); rx_ev_eth_crc_err = EFX_QWORD_FIELD(*event, RX_EV_ETH_CRC_ERR); rx_ev_frm_trunc = EFX_QWORD_FIELD(*event, RX_EV_FRM_TRUNC); - rx_ev_drib_nib = ((FALCON_REV(efx) >= FALCON_REV_B0) ? + rx_ev_drib_nib = ((falcon_rev(efx) >= FALCON_REV_B0) ? 0 : EFX_QWORD_FIELD(*event, RX_EV_DRIB_NIB)); rx_ev_pause_frm = EFX_QWORD_FIELD(*event, RX_EV_PAUSE_FRM_ERR); @@ -1065,7 +1058,7 @@ static void falcon_handle_global_event(struct efx_channel *channel, EFX_QWORD_FIELD(*event, XG_PHY_INTR)) is_phy_event = 1; - if ((FALCON_REV(efx) >= FALCON_REV_B0) && + if ((falcon_rev(efx) >= FALCON_REV_B0) && EFX_OWORD_FIELD(*event, XG_MNT_INTR_B0)) is_phy_event = 1; @@ -1129,6 +1122,7 @@ static void falcon_handle_driver_event(struct efx_channel *channel, case RX_RECOVERY_EV_DECODE: EFX_ERR(efx, "channel %d seen DRIVER RX_RESET event. " "Resetting.\n", channel->channel); + atomic_inc(&efx->rx_reset); efx_schedule_reset(efx, EFX_WORKAROUND_6555(efx) ? RESET_TYPE_RX_RECOVERY : @@ -1404,7 +1398,7 @@ static inline void falcon_irq_ack_a1(struct efx_nic *efx) static irqreturn_t falcon_fatal_interrupt(struct efx_nic *efx) { struct falcon_nic_data *nic_data = efx->nic_data; - efx_oword_t *int_ker = (efx_oword_t *) efx->irq_status.addr; + efx_oword_t *int_ker = efx->irq_status.addr; efx_oword_t fatal_intr; int error, mem_perr; static int n_int_errors; @@ -1450,8 +1444,8 @@ out: */ static irqreturn_t falcon_legacy_interrupt_b0(int irq, void *dev_id) { - struct efx_nic *efx = (struct efx_nic *)dev_id; - efx_oword_t *int_ker = (efx_oword_t *) efx->irq_status.addr; + struct efx_nic *efx = dev_id; + efx_oword_t *int_ker = efx->irq_status.addr; struct efx_channel *channel; efx_dword_t reg; u32 queues; @@ -1488,8 +1482,8 @@ static irqreturn_t falcon_legacy_interrupt_b0(int irq, void *dev_id) static irqreturn_t falcon_legacy_interrupt_a1(int irq, void *dev_id) { - struct efx_nic *efx = (struct efx_nic *)dev_id; - efx_oword_t *int_ker = (efx_oword_t *) efx->irq_status.addr; + struct efx_nic *efx = dev_id; + efx_oword_t *int_ker = efx->irq_status.addr; struct efx_channel *channel; int syserr; int queues; @@ -1541,9 +1535,9 @@ static irqreturn_t falcon_legacy_interrupt_a1(int irq, void *dev_id) */ static irqreturn_t falcon_msi_interrupt(int irq, void *dev_id) { - struct efx_channel *channel = (struct efx_channel *)dev_id; + struct efx_channel *channel = dev_id; struct efx_nic *efx = channel->efx; - efx_oword_t *int_ker = (efx_oword_t *) efx->irq_status.addr; + efx_oword_t *int_ker = efx->irq_status.addr; int syserr; efx->last_irq_cpu = raw_smp_processor_id(); @@ -1571,7 +1565,7 @@ static void falcon_setup_rss_indir_table(struct efx_nic *efx) unsigned long offset; efx_dword_t dword; - if (FALCON_REV(efx) < FALCON_REV_B0) + if (falcon_rev(efx) < FALCON_REV_B0) return; for (offset = RX_RSS_INDIR_TBL_B0; @@ -1594,7 +1588,7 @@ int falcon_init_interrupt(struct efx_nic *efx) if (!EFX_INT_MODE_USE_MSI(efx)) { irq_handler_t handler; - if (FALCON_REV(efx) >= FALCON_REV_B0) + if (falcon_rev(efx) >= FALCON_REV_B0) handler = falcon_legacy_interrupt_b0; else handler = falcon_legacy_interrupt_a1; @@ -1635,12 +1629,13 @@ void falcon_fini_interrupt(struct efx_nic *efx) efx_oword_t reg; /* Disable MSI/MSI-X interrupts */ - efx_for_each_channel_with_interrupt(channel, efx) + efx_for_each_channel_with_interrupt(channel, efx) { if (channel->irq) free_irq(channel->irq, channel); + } /* ACK legacy interrupt */ - if (FALCON_REV(efx) >= FALCON_REV_B0) + if (falcon_rev(efx) >= FALCON_REV_B0) falcon_read(efx, ®, INT_ISR0_B0); else falcon_irq_ack_a1(efx); @@ -1731,7 +1726,8 @@ void falcon_drain_tx_fifo(struct efx_nic *efx) efx_oword_t temp; int count; - if (FALCON_REV(efx) < FALCON_REV_B0) + if ((falcon_rev(efx) < FALCON_REV_B0) || + (efx->loopback_mode != LOOPBACK_NONE)) return; falcon_read(efx, &temp, MAC0_CTRL_REG_KER); @@ -1783,7 +1779,7 @@ void falcon_deconfigure_mac_wrapper(struct efx_nic *efx) { efx_oword_t temp; - if (FALCON_REV(efx) < FALCON_REV_B0) + if (falcon_rev(efx) < FALCON_REV_B0) return; /* Isolate the MAC -> RX */ @@ -1821,7 +1817,7 @@ void falcon_reconfigure_mac_wrapper(struct efx_nic *efx) MAC_SPEED, link_speed); /* On B0, MAC backpressure can be disabled and packets get * discarded. */ - if (FALCON_REV(efx) >= FALCON_REV_B0) { + if (falcon_rev(efx) >= FALCON_REV_B0) { EFX_SET_OWORD_FIELD(reg, TXFIFO_DRAIN_EN_B0, !efx->link_up); } @@ -1839,7 +1835,7 @@ void falcon_reconfigure_mac_wrapper(struct efx_nic *efx) EFX_SET_OWORD_FIELD_VER(efx, reg, RX_XOFF_MAC_EN, tx_fc); /* Unisolate the MAC -> RX */ - if (FALCON_REV(efx) >= FALCON_REV_B0) + if (falcon_rev(efx) >= FALCON_REV_B0) EFX_SET_OWORD_FIELD(reg, RX_INGR_EN_B0, 1); falcon_write(efx, ®, RX_CFG_REG_KER); } @@ -1854,7 +1850,7 @@ int falcon_dma_stats(struct efx_nic *efx, unsigned int done_offset) return 0; /* Statistics fetch will fail if the MAC is in TX drain */ - if (FALCON_REV(efx) >= FALCON_REV_B0) { + if (falcon_rev(efx) >= FALCON_REV_B0) { efx_oword_t temp; falcon_read(efx, &temp, MAC0_CTRL_REG_KER); if (EFX_OWORD_FIELD(temp, TXFIFO_DRAIN_EN_B0)) @@ -1938,7 +1934,7 @@ static int falcon_gmii_wait(struct efx_nic *efx) static void falcon_mdio_write(struct net_device *net_dev, int phy_id, int addr, int value) { - struct efx_nic *efx = (struct efx_nic *)net_dev->priv; + struct efx_nic *efx = net_dev->priv; unsigned int phy_id2 = phy_id & FALCON_PHY_ID_ID_MASK; efx_oword_t reg; @@ -2006,7 +2002,7 @@ static void falcon_mdio_write(struct net_device *net_dev, int phy_id, * could be read, -1 will be returned. */ static int falcon_mdio_read(struct net_device *net_dev, int phy_id, int addr) { - struct efx_nic *efx = (struct efx_nic *)net_dev->priv; + struct efx_nic *efx = net_dev->priv; unsigned int phy_addr = phy_id & FALCON_PHY_ID_ID_MASK; efx_oword_t reg; int value = -1; @@ -2091,6 +2087,8 @@ static int falcon_probe_phy(struct efx_nic *efx) efx->phy_type); return -1; } + + efx->loopback_modes = LOOPBACKS_10G_INTERNAL | efx->phy_op->loopbacks; return 0; } @@ -2109,7 +2107,7 @@ int falcon_probe_port(struct efx_nic *efx) falcon_init_mdio(&efx->mii); /* Hardware flow ctrl. FalconA RX FIFO too small for pause generation */ - if (FALCON_REV(efx) >= FALCON_REV_B0) + if (falcon_rev(efx) >= FALCON_REV_B0) efx->flow_control = EFX_FC_RX | EFX_FC_TX; else efx->flow_control = EFX_FC_RX; @@ -2369,7 +2367,7 @@ static int falcon_probe_nic_variant(struct efx_nic *efx) return -ENODEV; } - switch (FALCON_REV(efx)) { + switch (falcon_rev(efx)) { case FALCON_REV_A0: case 0xff: EFX_ERR(efx, "Falcon rev A0 not supported\n"); @@ -2395,7 +2393,7 @@ static int falcon_probe_nic_variant(struct efx_nic *efx) break; default: - EFX_ERR(efx, "Unknown Falcon rev %d\n", FALCON_REV(efx)); + EFX_ERR(efx, "Unknown Falcon rev %d\n", falcon_rev(efx)); return -ENODEV; } @@ -2415,7 +2413,7 @@ int falcon_probe_nic(struct efx_nic *efx) /* Allocate storage for hardware specific data */ nic_data = kzalloc(sizeof(*nic_data), GFP_KERNEL); - efx->nic_data = (void *) nic_data; + efx->nic_data = nic_data; /* Determine number of ports etc. */ rc = falcon_probe_nic_variant(efx); @@ -2468,14 +2466,12 @@ int falcon_probe_nic(struct efx_nic *efx) fail5: falcon_free_buffer(efx, &efx->irq_status); fail4: - /* fall-thru */ fail3: if (nic_data->pci_dev2) { pci_dev_put(nic_data->pci_dev2); nic_data->pci_dev2 = NULL; } fail2: - /* fall-thru */ fail1: kfree(efx->nic_data); return rc; @@ -2487,13 +2483,10 @@ int falcon_probe_nic(struct efx_nic *efx) */ int falcon_init_nic(struct efx_nic *efx) { - struct falcon_nic_data *data; efx_oword_t temp; unsigned thresh; int rc; - data = (struct falcon_nic_data *)efx->nic_data; - /* Set up the address region register. This is only needed * for the B0 FPGA, but since we are just pushing in the * reset defaults this may as well be unconditional. */ @@ -2560,7 +2553,7 @@ int falcon_init_nic(struct efx_nic *efx) /* Set number of RSS queues for receive path. */ falcon_read(efx, &temp, RX_FILTER_CTL_REG); - if (FALCON_REV(efx) >= FALCON_REV_B0) + if (falcon_rev(efx) >= FALCON_REV_B0) EFX_SET_OWORD_FIELD(temp, NUM_KER, 0); else EFX_SET_OWORD_FIELD(temp, NUM_KER, efx->rss_queues - 1); @@ -2598,7 +2591,7 @@ int falcon_init_nic(struct efx_nic *efx) /* Prefetch threshold 2 => fetch when descriptor cache half empty */ EFX_SET_OWORD_FIELD(temp, TX_PREF_THRESHOLD, 2); /* Squash TX of packets of 16 bytes or less */ - if (FALCON_REV(efx) >= FALCON_REV_B0 && EFX_WORKAROUND_9141(efx)) + if (falcon_rev(efx) >= FALCON_REV_B0 && EFX_WORKAROUND_9141(efx)) EFX_SET_OWORD_FIELD(temp, TX_FLUSH_MIN_LEN_EN_B0, 1); falcon_write(efx, &temp, TX_CFG2_REG_KER); @@ -2615,7 +2608,7 @@ int falcon_init_nic(struct efx_nic *efx) if (EFX_WORKAROUND_7575(efx)) EFX_SET_OWORD_FIELD_VER(efx, temp, RX_USR_BUF_SIZE, (3 * 4096) / 32); - if (FALCON_REV(efx) >= FALCON_REV_B0) + if (falcon_rev(efx) >= FALCON_REV_B0) EFX_SET_OWORD_FIELD(temp, RX_INGR_EN_B0, 1); /* RX FIFO flow control thresholds */ @@ -2631,7 +2624,7 @@ int falcon_init_nic(struct efx_nic *efx) falcon_write(efx, &temp, RX_CFG_REG_KER); /* Set destination of both TX and RX Flush events */ - if (FALCON_REV(efx) >= FALCON_REV_B0) { + if (falcon_rev(efx) >= FALCON_REV_B0) { EFX_POPULATE_OWORD_1(temp, FLS_EVQ_ID, 0); falcon_write(efx, &temp, DP_CTRL_REG); } @@ -2645,7 +2638,7 @@ void falcon_remove_nic(struct efx_nic *efx) falcon_free_buffer(efx, &efx->irq_status); - (void) falcon_reset_hw(efx, RESET_TYPE_ALL); + falcon_reset_hw(efx, RESET_TYPE_ALL); /* Release the second function after the reset */ if (nic_data->pci_dev2) { diff --git a/drivers/net/sfc/falcon.h b/drivers/net/sfc/falcon.h index 6117403b0c03..492f9bc28840 100644 --- a/drivers/net/sfc/falcon.h +++ b/drivers/net/sfc/falcon.h @@ -23,7 +23,10 @@ enum falcon_revision { FALCON_REV_B0 = 2, }; -#define FALCON_REV(efx) ((efx)->pci_dev->revision) +static inline int falcon_rev(struct efx_nic *efx) +{ + return efx->pci_dev->revision; +} extern struct efx_nic_type falcon_a_nic_type; extern struct efx_nic_type falcon_b_nic_type; diff --git a/drivers/net/sfc/falcon_hwdefs.h b/drivers/net/sfc/falcon_hwdefs.h index 0485a63eaff6..6d003114eeab 100644 --- a/drivers/net/sfc/falcon_hwdefs.h +++ b/drivers/net/sfc/falcon_hwdefs.h @@ -636,6 +636,14 @@ #define XX_HIDRVA_WIDTH 1 #define XX_LODRVA_LBN 8 #define XX_LODRVA_WIDTH 1 +#define XX_LPBKD_LBN 3 +#define XX_LPBKD_WIDTH 1 +#define XX_LPBKC_LBN 2 +#define XX_LPBKC_WIDTH 1 +#define XX_LPBKB_LBN 1 +#define XX_LPBKB_WIDTH 1 +#define XX_LPBKA_LBN 0 +#define XX_LPBKA_WIDTH 1 #define XX_TXDRV_CTL_REG_MAC 0x12 #define XX_DEQD_LBN 28 @@ -656,8 +664,14 @@ #define XX_DTXA_WIDTH 4 /* XAUI XGXS core status register */ -#define XX_FORCE_SIG_DECODE_FORCED 0xff #define XX_CORE_STAT_REG_MAC 0x16 +#define XX_FORCE_SIG_LBN 24 +#define XX_FORCE_SIG_WIDTH 8 +#define XX_FORCE_SIG_DECODE_FORCED 0xff +#define XX_XGXS_LB_EN_LBN 23 +#define XX_XGXS_LB_EN_WIDTH 1 +#define XX_XGMII_LB_EN_LBN 22 +#define XX_XGMII_LB_EN_WIDTH 1 #define XX_ALIGN_DONE_LBN 20 #define XX_ALIGN_DONE_WIDTH 1 #define XX_SYNC_STAT_LBN 16 @@ -1111,7 +1125,7 @@ struct falcon_nvconfig_board_v2 { u8 port1_phy_type; __le16 asic_sub_revision; __le16 board_revision; -} __attribute__ ((packed)); +} __packed; #define NVCONFIG_BASE 0x300 #define NVCONFIG_BOARD_MAGIC_NUM 0xFA1C @@ -1130,6 +1144,6 @@ struct falcon_nvconfig { __le16 board_struct_ver; __le16 board_checksum; struct falcon_nvconfig_board_v2 board_v2; -} __attribute__ ((packed)); +} __packed; #endif /* EFX_FALCON_HWDEFS_H */ diff --git a/drivers/net/sfc/falcon_io.h b/drivers/net/sfc/falcon_io.h index ea08184ddfa9..6670cdfc41ab 100644 --- a/drivers/net/sfc/falcon_io.h +++ b/drivers/net/sfc/falcon_io.h @@ -56,14 +56,27 @@ #define FALCON_USE_QWORD_IO 1 #endif -#define _falcon_writeq(efx, value, reg) \ - __raw_writeq((__force u64) (value), (efx)->membase + (reg)) -#define _falcon_writel(efx, value, reg) \ - __raw_writel((__force u32) (value), (efx)->membase + (reg)) -#define _falcon_readq(efx, reg) \ - ((__force __le64) __raw_readq((efx)->membase + (reg))) -#define _falcon_readl(efx, reg) \ - ((__force __le32) __raw_readl((efx)->membase + (reg))) +#ifdef FALCON_USE_QWORD_IO +static inline void _falcon_writeq(struct efx_nic *efx, __le64 value, + unsigned int reg) +{ + __raw_writeq((__force u64)value, efx->membase + reg); +} +static inline __le64 _falcon_readq(struct efx_nic *efx, unsigned int reg) +{ + return (__force __le64)__raw_readq(efx->membase + reg); +} +#endif + +static inline void _falcon_writel(struct efx_nic *efx, __le32 value, + unsigned int reg) +{ + __raw_writel((__force u32)value, efx->membase + reg); +} +static inline __le32 _falcon_readl(struct efx_nic *efx, unsigned int reg) +{ + return (__force __le32)__raw_readl(efx->membase + reg); +} /* Writes to a normal 16-byte Falcon register, locking as appropriate. */ static inline void falcon_write(struct efx_nic *efx, efx_oword_t *value, diff --git a/drivers/net/sfc/falcon_xmac.c b/drivers/net/sfc/falcon_xmac.c index aa7521b24a5d..55c0d9760be8 100644 --- a/drivers/net/sfc/falcon_xmac.c +++ b/drivers/net/sfc/falcon_xmac.c @@ -32,7 +32,7 @@ (FALCON_XMAC_REGBANK + ((mac_reg) * FALCON_XMAC_REG_SIZE)) void falcon_xmac_writel(struct efx_nic *efx, - efx_dword_t *value, unsigned int mac_reg) + efx_dword_t *value, unsigned int mac_reg) { efx_oword_t temp; @@ -69,6 +69,10 @@ static int falcon_reset_xmac(struct efx_nic *efx) udelay(10); } + /* This often fails when DSP is disabled, ignore it */ + if (sfe4001_phy_flash_cfg != 0) + return 0; + EFX_ERR(efx, "timed out waiting for XMAC core reset\n"); return -ETIMEDOUT; } @@ -217,13 +221,13 @@ static int falcon_xgmii_status(struct efx_nic *efx) { efx_dword_t reg; - if (FALCON_REV(efx) < FALCON_REV_B0) + if (falcon_rev(efx) < FALCON_REV_B0) return 1; /* The ISR latches, so clear it and re-read */ falcon_xmac_readl(efx, ®, XM_MGT_INT_REG_MAC_B0); falcon_xmac_readl(efx, ®, XM_MGT_INT_REG_MAC_B0); - + if (EFX_DWORD_FIELD(reg, XM_LCLFLT) || EFX_DWORD_FIELD(reg, XM_RMTFLT)) { EFX_INFO(efx, "MGT_INT: "EFX_DWORD_FMT"\n", EFX_DWORD_VAL(reg)); @@ -237,7 +241,7 @@ static void falcon_mask_status_intr(struct efx_nic *efx, int enable) { efx_dword_t reg; - if (FALCON_REV(efx) < FALCON_REV_B0) + if ((falcon_rev(efx) < FALCON_REV_B0) || LOOPBACK_INTERNAL(efx)) return; /* Flush the ISR */ @@ -284,6 +288,9 @@ int falcon_xaui_link_ok(struct efx_nic *efx) efx_dword_t reg; int align_done, sync_status, link_ok = 0; + if (LOOPBACK_INTERNAL(efx)) + return 1; + /* Read link status */ falcon_xmac_readl(efx, ®, XX_CORE_STAT_REG_MAC); @@ -374,6 +381,61 @@ static void falcon_reconfigure_xmac_core(struct efx_nic *efx) falcon_xmac_writel(efx, ®, XM_ADR_HI_REG_MAC); } +static void falcon_reconfigure_xgxs_core(struct efx_nic *efx) +{ + efx_dword_t reg; + int xgxs_loopback = (efx->loopback_mode == LOOPBACK_XGXS) ? 1 : 0; + int xaui_loopback = (efx->loopback_mode == LOOPBACK_XAUI) ? 1 : 0; + int xgmii_loopback = + (efx->loopback_mode == LOOPBACK_XGMII) ? 1 : 0; + + /* XGXS block is flaky and will need to be reset if moving + * into our out of XGMII, XGXS or XAUI loopbacks. */ + if (EFX_WORKAROUND_5147(efx)) { + int old_xgmii_loopback, old_xgxs_loopback, old_xaui_loopback; + int reset_xgxs; + + falcon_xmac_readl(efx, ®, XX_CORE_STAT_REG_MAC); + old_xgxs_loopback = EFX_DWORD_FIELD(reg, XX_XGXS_LB_EN); + old_xgmii_loopback = EFX_DWORD_FIELD(reg, XX_XGMII_LB_EN); + + falcon_xmac_readl(efx, ®, XX_SD_CTL_REG_MAC); + old_xaui_loopback = EFX_DWORD_FIELD(reg, XX_LPBKA); + + /* The PHY driver may have turned XAUI off */ + reset_xgxs = ((xgxs_loopback != old_xgxs_loopback) || + (xaui_loopback != old_xaui_loopback) || + (xgmii_loopback != old_xgmii_loopback)); + if (reset_xgxs) { + falcon_xmac_readl(efx, ®, XX_PWR_RST_REG_MAC); + EFX_SET_DWORD_FIELD(reg, XX_RSTXGXSTX_EN, 1); + EFX_SET_DWORD_FIELD(reg, XX_RSTXGXSRX_EN, 1); + falcon_xmac_writel(efx, ®, XX_PWR_RST_REG_MAC); + udelay(1); + EFX_SET_DWORD_FIELD(reg, XX_RSTXGXSTX_EN, 0); + EFX_SET_DWORD_FIELD(reg, XX_RSTXGXSRX_EN, 0); + falcon_xmac_writel(efx, ®, XX_PWR_RST_REG_MAC); + udelay(1); + } + } + + falcon_xmac_readl(efx, ®, XX_CORE_STAT_REG_MAC); + EFX_SET_DWORD_FIELD(reg, XX_FORCE_SIG, + (xgxs_loopback || xaui_loopback) ? + XX_FORCE_SIG_DECODE_FORCED : 0); + EFX_SET_DWORD_FIELD(reg, XX_XGXS_LB_EN, xgxs_loopback); + EFX_SET_DWORD_FIELD(reg, XX_XGMII_LB_EN, xgmii_loopback); + falcon_xmac_writel(efx, ®, XX_CORE_STAT_REG_MAC); + + falcon_xmac_readl(efx, ®, XX_SD_CTL_REG_MAC); + EFX_SET_DWORD_FIELD(reg, XX_LPBKD, xaui_loopback); + EFX_SET_DWORD_FIELD(reg, XX_LPBKC, xaui_loopback); + EFX_SET_DWORD_FIELD(reg, XX_LPBKB, xaui_loopback); + EFX_SET_DWORD_FIELD(reg, XX_LPBKA, xaui_loopback); + falcon_xmac_writel(efx, ®, XX_SD_CTL_REG_MAC); +} + + /* Try and bring the Falcon side of the Falcon-Phy XAUI link fails * to come back up. Bash it until it comes back up */ static int falcon_check_xaui_link_up(struct efx_nic *efx) @@ -382,7 +444,8 @@ static int falcon_check_xaui_link_up(struct efx_nic *efx) tries = EFX_WORKAROUND_5147(efx) ? 5 : 1; max_tries = tries; - if (efx->phy_type == PHY_TYPE_NONE) + if ((efx->loopback_mode == LOOPBACK_NETWORK) || + (efx->phy_type == PHY_TYPE_NONE)) return 0; while (tries) { @@ -391,12 +454,12 @@ static int falcon_check_xaui_link_up(struct efx_nic *efx) EFX_LOG(efx, "%s Clobbering XAUI (%d tries left).\n", __func__, tries); - (void) falcon_reset_xaui(efx); + falcon_reset_xaui(efx); udelay(200); tries--; } - EFX_ERR(efx, "Failed to bring XAUI link back up in %d tries!\n", + EFX_LOG(efx, "Failed to bring XAUI link back up in %d tries!\n", max_tries); return 0; } @@ -408,8 +471,13 @@ void falcon_reconfigure_xmac(struct efx_nic *efx) falcon_mask_status_intr(efx, 0); falcon_deconfigure_mac_wrapper(efx); + + efx->tx_disabled = LOOPBACK_INTERNAL(efx); efx->phy_op->reconfigure(efx); + + falcon_reconfigure_xgxs_core(efx); falcon_reconfigure_xmac_core(efx); + falcon_reconfigure_mac_wrapper(efx); /* Ensure XAUI link is up */ @@ -491,18 +559,20 @@ void falcon_update_stats_xmac(struct efx_nic *efx) (mac_stats->rx_bytes - mac_stats->rx_good_bytes); } -#define EFX_XAUI_RETRAIN_MAX 8 - int falcon_check_xmac(struct efx_nic *efx) { unsigned xaui_link_ok; int rc; + if ((efx->loopback_mode == LOOPBACK_NETWORK) || + (efx->phy_type == PHY_TYPE_NONE)) + return 0; + falcon_mask_status_intr(efx, 0); xaui_link_ok = falcon_xaui_link_ok(efx); if (EFX_WORKAROUND_5147(efx) && !xaui_link_ok) - (void) falcon_reset_xaui(efx); + falcon_reset_xaui(efx); /* Call the PHY check_hw routine */ rc = efx->phy_op->check_hw(efx); @@ -569,7 +639,7 @@ int falcon_xmac_set_pause(struct efx_nic *efx, enum efx_fc_type flow_control) reset = ((flow_control & EFX_FC_TX) && !(efx->flow_control & EFX_FC_TX)); if (EFX_WORKAROUND_11482(efx) && reset) { - if (FALCON_REV(efx) >= FALCON_REV_B0) { + if (falcon_rev(efx) >= FALCON_REV_B0) { /* Recover by resetting the EM block */ if (efx->link_up) falcon_drain_tx_fifo(efx); diff --git a/drivers/net/sfc/mdio_10g.c b/drivers/net/sfc/mdio_10g.c index dc06bb0aa575..c4f540e93b79 100644 --- a/drivers/net/sfc/mdio_10g.c +++ b/drivers/net/sfc/mdio_10g.c @@ -44,6 +44,9 @@ static int mdio_clause45_check_mmd(struct efx_nic *efx, int mmd, int status; int phy_id = efx->mii.phy_id; + if (LOOPBACK_INTERNAL(efx)) + return 0; + /* Read MMD STATUS2 to check it is responding. */ status = mdio_clause45_read(efx, phy_id, mmd, MDIO_MMDREG_STAT2); if (((status >> MDIO_MMDREG_STAT2_PRESENT_LBN) & @@ -164,6 +167,22 @@ int mdio_clause45_links_ok(struct efx_nic *efx, unsigned int mmd_mask) int mmd = 0; int good; + /* If the port is in loopback, then we should only consider a subset + * of mmd's */ + if (LOOPBACK_INTERNAL(efx)) + return 1; + else if (efx->loopback_mode == LOOPBACK_NETWORK) + return 0; + else if (efx->loopback_mode == LOOPBACK_PHYXS) + mmd_mask &= ~(MDIO_MMDREG_DEVS0_PHYXS | + MDIO_MMDREG_DEVS0_PCS | + MDIO_MMDREG_DEVS0_PMAPMD); + else if (efx->loopback_mode == LOOPBACK_PCS) + mmd_mask &= ~(MDIO_MMDREG_DEVS0_PCS | + MDIO_MMDREG_DEVS0_PMAPMD); + else if (efx->loopback_mode == LOOPBACK_PMAPMD) + mmd_mask &= ~MDIO_MMDREG_DEVS0_PMAPMD; + while (mmd_mask) { if (mmd_mask & 1) { /* Double reads because link state is latched, and a @@ -182,6 +201,65 @@ int mdio_clause45_links_ok(struct efx_nic *efx, unsigned int mmd_mask) return ok; } +void mdio_clause45_transmit_disable(struct efx_nic *efx) +{ + int phy_id = efx->mii.phy_id; + int ctrl1, ctrl2; + + ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD, + MDIO_MMDREG_TXDIS); + if (efx->tx_disabled) + ctrl2 |= (1 << MDIO_MMDREG_TXDIS_GLOBAL_LBN); + else + ctrl1 &= ~(1 << MDIO_MMDREG_TXDIS_GLOBAL_LBN); + if (ctrl1 != ctrl2) + mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD, + MDIO_MMDREG_TXDIS, ctrl2); +} + +void mdio_clause45_phy_reconfigure(struct efx_nic *efx) +{ + int phy_id = efx->mii.phy_id; + int ctrl1, ctrl2; + + /* Handle (with debouncing) PMA/PMD loopback */ + ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD, + MDIO_MMDREG_CTRL1); + + if (efx->loopback_mode == LOOPBACK_PMAPMD) + ctrl2 |= (1 << MDIO_PMAPMD_CTRL1_LBACK_LBN); + else + ctrl2 &= ~(1 << MDIO_PMAPMD_CTRL1_LBACK_LBN); + + if (ctrl1 != ctrl2) + mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD, + MDIO_MMDREG_CTRL1, ctrl2); + + /* Handle (with debouncing) PCS loopback */ + ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PCS, + MDIO_MMDREG_CTRL1); + if (efx->loopback_mode == LOOPBACK_PCS) + ctrl2 |= (1 << MDIO_MMDREG_CTRL1_LBACK_LBN); + else + ctrl2 &= ~(1 << MDIO_MMDREG_CTRL1_LBACK_LBN); + + if (ctrl1 != ctrl2) + mdio_clause45_write(efx, phy_id, MDIO_MMD_PCS, + MDIO_MMDREG_CTRL1, ctrl2); + + /* Handle (with debouncing) PHYXS network loopback */ + ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PHYXS, + MDIO_MMDREG_CTRL1); + if (efx->loopback_mode == LOOPBACK_NETWORK) + ctrl2 |= (1 << MDIO_MMDREG_CTRL1_LBACK_LBN); + else + ctrl2 &= ~(1 << MDIO_MMDREG_CTRL1_LBACK_LBN); + + if (ctrl1 != ctrl2) + mdio_clause45_write(efx, phy_id, MDIO_MMD_PHYXS, + MDIO_MMDREG_CTRL1, ctrl2); +} + /** * mdio_clause45_get_settings - Read (some of) the PHY settings over MDIO. * @efx: Efx NIC diff --git a/drivers/net/sfc/mdio_10g.h b/drivers/net/sfc/mdio_10g.h index 2214b6d820a7..cb99f3f4491c 100644 --- a/drivers/net/sfc/mdio_10g.h +++ b/drivers/net/sfc/mdio_10g.h @@ -44,11 +44,16 @@ #define MDIO_MMDREG_DEVS1 (6) #define MDIO_MMDREG_CTRL2 (7) #define MDIO_MMDREG_STAT2 (8) +#define MDIO_MMDREG_TXDIS (9) /* Bits in MMDREG_CTRL1 */ /* Reset */ #define MDIO_MMDREG_CTRL1_RESET_LBN (15) #define MDIO_MMDREG_CTRL1_RESET_WIDTH (1) +/* Loopback */ +/* Loopback bit for WIS, PCS, PHYSX and DTEXS */ +#define MDIO_MMDREG_CTRL1_LBACK_LBN (14) +#define MDIO_MMDREG_CTRL1_LBACK_WIDTH (1) /* Bits in MMDREG_STAT1 */ #define MDIO_MMDREG_STAT1_FAULT_LBN (7) @@ -56,6 +61,9 @@ /* Link state */ #define MDIO_MMDREG_STAT1_LINK_LBN (2) #define MDIO_MMDREG_STAT1_LINK_WIDTH (1) +/* Low power ability */ +#define MDIO_MMDREG_STAT1_LPABLE_LBN (1) +#define MDIO_MMDREG_STAT1_LPABLE_WIDTH (1) /* Bits in ID reg */ #define MDIO_ID_REV(_id32) (_id32 & 0xf) @@ -76,6 +84,14 @@ #define MDIO_MMDREG_STAT2_PRESENT_LBN (14) #define MDIO_MMDREG_STAT2_PRESENT_WIDTH (2) +/* Bits in MMDREG_TXDIS */ +#define MDIO_MMDREG_TXDIS_GLOBAL_LBN (0) +#define MDIO_MMDREG_TXDIS_GLOBAL_WIDTH (1) + +/* MMD-specific bits, ordered by MMD, then register */ +#define MDIO_PMAPMD_CTRL1_LBACK_LBN (0) +#define MDIO_PMAPMD_CTRL1_LBACK_WIDTH (1) + /* PMA type (4 bits) */ #define MDIO_PMAPMD_CTRL2_10G_CX4 (0x0) #define MDIO_PMAPMD_CTRL2_10G_EW (0x1) @@ -95,7 +111,7 @@ #define MDIO_PMAPMD_CTRL2_10_BT (0xf) #define MDIO_PMAPMD_CTRL2_TYPE_MASK (0xf) -/* /\* PHY XGXS lane state *\/ */ +/* PHY XGXS lane state */ #define MDIO_PHYXS_LANE_STATE (0x18) #define MDIO_PHYXS_LANE_ALIGNED_LBN (12) @@ -217,6 +233,12 @@ int mdio_clause45_check_mmds(struct efx_nic *efx, extern int mdio_clause45_links_ok(struct efx_nic *efx, unsigned int mmd_mask); +/* Generic transmit disable support though PMAPMD */ +extern void mdio_clause45_transmit_disable(struct efx_nic *efx); + +/* Generic part of reconfigure: set/clear loopback bits */ +extern void mdio_clause45_phy_reconfigure(struct efx_nic *efx); + /* Read (some of) the PHY settings over MDIO */ extern void mdio_clause45_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd); diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h index c505482c2520..5e20e7551dae 100644 --- a/drivers/net/sfc/net_driver.h +++ b/drivers/net/sfc/net_driver.h @@ -42,7 +42,7 @@ #ifndef EFX_DRIVER_NAME #define EFX_DRIVER_NAME "sfc" #endif -#define EFX_DRIVER_VERSION "2.2.0136" +#define EFX_DRIVER_VERSION "2.2" #ifdef EFX_ENABLE_DEBUG #define EFX_BUG_ON_PARANOID(x) BUG_ON(x) @@ -52,28 +52,19 @@ #define EFX_WARN_ON_PARANOID(x) do {} while (0) #endif -#define NET_DEV_REGISTERED(efx) \ - ((efx)->net_dev->reg_state == NETREG_REGISTERED) - -/* Include net device name in log messages if it has been registered. - * Use efx->name not efx->net_dev->name so that races with (un)registration - * are harmless. - */ -#define NET_DEV_NAME(efx) (NET_DEV_REGISTERED(efx) ? (efx)->name : "") - /* Un-rate-limited logging */ #define EFX_ERR(efx, fmt, args...) \ -dev_err(&((efx)->pci_dev->dev), "ERR: %s " fmt, NET_DEV_NAME(efx), ##args) +dev_err(&((efx)->pci_dev->dev), "ERR: %s " fmt, efx_dev_name(efx), ##args) #define EFX_INFO(efx, fmt, args...) \ -dev_info(&((efx)->pci_dev->dev), "INFO: %s " fmt, NET_DEV_NAME(efx), ##args) +dev_info(&((efx)->pci_dev->dev), "INFO: %s " fmt, efx_dev_name(efx), ##args) #ifdef EFX_ENABLE_DEBUG #define EFX_LOG(efx, fmt, args...) \ -dev_info(&((efx)->pci_dev->dev), "DBG: %s " fmt, NET_DEV_NAME(efx), ##args) +dev_info(&((efx)->pci_dev->dev), "DBG: %s " fmt, efx_dev_name(efx), ##args) #else #define EFX_LOG(efx, fmt, args...) \ -dev_dbg(&((efx)->pci_dev->dev), "DBG: %s " fmt, NET_DEV_NAME(efx), ##args) +dev_dbg(&((efx)->pci_dev->dev), "DBG: %s " fmt, efx_dev_name(efx), ##args) #endif #define EFX_TRACE(efx, fmt, args...) do {} while (0) @@ -90,11 +81,6 @@ do {if (net_ratelimit()) EFX_INFO(efx, fmt, ##args); } while (0) #define EFX_LOG_RL(efx, fmt, args...) \ do {if (net_ratelimit()) EFX_LOG(efx, fmt, ##args); } while (0) -/* Kernel headers may redefine inline anyway */ -#ifndef inline -#define inline inline __attribute__ ((always_inline)) -#endif - /************************************************************************** * * Efx data structures @@ -134,6 +120,8 @@ struct efx_special_buffer { * Set only on the final fragment of a packet; %NULL for all other * fragments. When this fragment completes, then we can free this * skb. + * @tsoh: The associated TSO header structure, or %NULL if this + * buffer is not a TSO header. * @dma_addr: DMA address of the fragment. * @len: Length of this fragment. * This field is zero when the queue slot is empty. @@ -144,6 +132,7 @@ struct efx_special_buffer { */ struct efx_tx_buffer { const struct sk_buff *skb; + struct efx_tso_header *tsoh; dma_addr_t dma_addr; unsigned short len; unsigned char continuation; @@ -187,6 +176,13 @@ struct efx_tx_buffer { * variable indicates that the queue is full. This is to * avoid cache-line ping-pong between the xmit path and the * completion path. + * @tso_headers_free: A list of TSO headers allocated for this TX queue + * that are not in use, and so available for new TSO sends. The list + * is protected by the TX queue lock. + * @tso_bursts: Number of times TSO xmit invoked by kernel + * @tso_long_headers: Number of packets with headers too long for standard + * blocks + * @tso_packets: Number of packets via the TSO xmit path */ struct efx_tx_queue { /* Members which don't change on the fast path */ @@ -206,6 +202,10 @@ struct efx_tx_queue { unsigned int insert_count ____cacheline_aligned_in_smp; unsigned int write_count; unsigned int old_read_count; + struct efx_tso_header *tso_headers_free; + unsigned int tso_bursts; + unsigned int tso_long_headers; + unsigned int tso_packets; }; /** @@ -434,6 +434,9 @@ struct efx_board { struct efx_blinker blinker; }; +#define STRING_TABLE_LOOKUP(val, member) \ + member ## _names[val] + enum efx_int_mode { /* Be careful if altering to correct macro below */ EFX_INT_MODE_MSIX = 0, @@ -506,6 +509,7 @@ enum efx_fc_type { * @check_hw: Check hardware * @reset_xaui: Reset XAUI side of PHY for (software sequenced reset) * @mmds: MMD presence mask + * @loopbacks: Supported loopback modes mask */ struct efx_phy_operations { int (*init) (struct efx_nic *efx); @@ -515,6 +519,7 @@ struct efx_phy_operations { int (*check_hw) (struct efx_nic *efx); void (*reset_xaui) (struct efx_nic *efx); int mmds; + unsigned loopbacks; }; /* @@ -653,7 +658,6 @@ union efx_multicast_hash { * @phy_op: PHY interface * @phy_data: PHY private data (including PHY-specific stats) * @mii: PHY interface - * @phy_powered: PHY power state * @tx_disabled: PHY transmitter turned off * @link_up: Link status * @link_options: Link options (MII/GMII format) @@ -662,6 +666,9 @@ union efx_multicast_hash { * @multicast_hash: Multicast hash table * @flow_control: Flow control flags - separate RX/TX so can't use link_options * @reconfigure_work: work item for dealing with PHY events + * @loopback_mode: Loopback status + * @loopback_modes: Supported loopback mode bitmask + * @loopback_selftest: Offline self-test private state * * The @priv field of the corresponding &struct net_device points to * this. @@ -674,7 +681,7 @@ struct efx_nic { struct workqueue_struct *workqueue; struct work_struct reset_work; struct delayed_work monitor_work; - unsigned long membase_phys; + resource_size_t membase_phys; void __iomem *membase; spinlock_t biu_lock; enum efx_int_mode interrupt_mode; @@ -698,7 +705,7 @@ struct efx_nic { unsigned n_rx_nodesc_drop_cnt; - void *nic_data; + struct falcon_nic_data *nic_data; struct mutex mac_lock; int port_enabled; @@ -721,6 +728,7 @@ struct efx_nic { struct efx_phy_operations *phy_op; void *phy_data; struct mii_if_info mii; + unsigned tx_disabled; int link_up; unsigned int link_options; @@ -732,8 +740,26 @@ struct efx_nic { struct work_struct reconfigure_work; atomic_t rx_reset; + enum efx_loopback_mode loopback_mode; + unsigned int loopback_modes; + + void *loopback_selftest; }; +static inline int efx_dev_registered(struct efx_nic *efx) +{ + return efx->net_dev->reg_state == NETREG_REGISTERED; +} + +/* Net device name, for inclusion in log messages if it has been registered. + * Use efx->name not efx->net_dev->name so that races with (un)registration + * are harmless. + */ +static inline const char *efx_dev_name(struct efx_nic *efx) +{ + return efx_dev_registered(efx) ? efx->name : ""; +} + /** * struct efx_nic_type - Efx device type definition * @mem_bar: Memory BAR number @@ -769,7 +795,7 @@ struct efx_nic_type { unsigned int txd_ring_mask; unsigned int rxd_ring_mask; unsigned int evq_size; - dma_addr_t max_dma_mask; + u64 max_dma_mask; unsigned int tx_dma_mask; unsigned bug5391_mask; diff --git a/drivers/net/sfc/rx.c b/drivers/net/sfc/rx.c index 551299b462ae..601b001437c0 100644 --- a/drivers/net/sfc/rx.c +++ b/drivers/net/sfc/rx.c @@ -19,6 +19,7 @@ #include "rx.h" #include "efx.h" #include "falcon.h" +#include "selftest.h" #include "workarounds.h" /* Number of RX descriptors pushed at once. */ @@ -85,14 +86,17 @@ static unsigned int rx_refill_limit = 95; */ #define EFX_RXD_HEAD_ROOM 2 -/* Macros for zero-order pages (potentially) containing multiple RX buffers */ -#define RX_DATA_OFFSET(_data) \ - (((unsigned long) (_data)) & (PAGE_SIZE-1)) -#define RX_BUF_OFFSET(_rx_buf) \ - RX_DATA_OFFSET((_rx_buf)->data) - -#define RX_PAGE_SIZE(_efx) \ - (PAGE_SIZE * (1u << (_efx)->rx_buffer_order)) +static inline unsigned int efx_rx_buf_offset(struct efx_rx_buffer *buf) +{ + /* Offset is always within one page, so we don't need to consider + * the page order. + */ + return (__force unsigned long) buf->data & (PAGE_SIZE - 1); +} +static inline unsigned int efx_rx_buf_size(struct efx_nic *efx) +{ + return PAGE_SIZE << efx->rx_buffer_order; +} /************************************************************************** @@ -105,7 +109,7 @@ static unsigned int rx_refill_limit = 95; static int efx_lro_get_skb_hdr(struct sk_buff *skb, void **ip_hdr, void **tcpudp_hdr, u64 *hdr_flags, void *priv) { - struct efx_channel *channel = (struct efx_channel *)priv; + struct efx_channel *channel = priv; struct iphdr *iph; struct tcphdr *th; @@ -130,12 +134,12 @@ static int efx_get_frag_hdr(struct skb_frag_struct *frag, void **mac_hdr, void **ip_hdr, void **tcpudp_hdr, u64 *hdr_flags, void *priv) { - struct efx_channel *channel = (struct efx_channel *)priv; + struct efx_channel *channel = priv; struct ethhdr *eh; struct iphdr *iph; /* We support EtherII and VLAN encapsulated IPv4 */ - eh = (struct ethhdr *)(page_address(frag->page) + frag->page_offset); + eh = page_address(frag->page) + frag->page_offset; *mac_hdr = eh; if (eh->h_proto == htons(ETH_P_IP)) { @@ -268,7 +272,7 @@ static inline int efx_init_rx_buffer_page(struct efx_rx_queue *rx_queue, return -ENOMEM; dma_addr = pci_map_page(efx->pci_dev, rx_buf->page, - 0, RX_PAGE_SIZE(efx), + 0, efx_rx_buf_size(efx), PCI_DMA_FROMDEVICE); if (unlikely(pci_dma_mapping_error(dma_addr))) { @@ -279,14 +283,14 @@ static inline int efx_init_rx_buffer_page(struct efx_rx_queue *rx_queue, rx_queue->buf_page = rx_buf->page; rx_queue->buf_dma_addr = dma_addr; - rx_queue->buf_data = ((char *) page_address(rx_buf->page) + + rx_queue->buf_data = (page_address(rx_buf->page) + EFX_PAGE_IP_ALIGN); } - offset = RX_DATA_OFFSET(rx_queue->buf_data); rx_buf->len = bytes; - rx_buf->dma_addr = rx_queue->buf_dma_addr + offset; rx_buf->data = rx_queue->buf_data; + offset = efx_rx_buf_offset(rx_buf); + rx_buf->dma_addr = rx_queue->buf_dma_addr + offset; /* Try to pack multiple buffers per page */ if (efx->rx_buffer_order == 0) { @@ -294,7 +298,7 @@ static inline int efx_init_rx_buffer_page(struct efx_rx_queue *rx_queue, rx_queue->buf_data += ((bytes + 0x1ff) & ~0x1ff); offset += ((bytes + 0x1ff) & ~0x1ff); - space = RX_PAGE_SIZE(efx) - offset; + space = efx_rx_buf_size(efx) - offset; if (space >= bytes) { /* Refs dropped on kernel releasing each skb */ get_page(rx_queue->buf_page); @@ -343,7 +347,8 @@ static inline void efx_unmap_rx_buffer(struct efx_nic *efx, EFX_BUG_ON_PARANOID(rx_buf->skb); if (rx_buf->unmap_addr) { pci_unmap_page(efx->pci_dev, rx_buf->unmap_addr, - RX_PAGE_SIZE(efx), PCI_DMA_FROMDEVICE); + efx_rx_buf_size(efx), + PCI_DMA_FROMDEVICE); rx_buf->unmap_addr = 0; } } else if (likely(rx_buf->skb)) { @@ -399,9 +404,10 @@ static int __efx_fast_push_rx_descriptors(struct efx_rx_queue *rx_queue, return 0; /* Record minimum fill level */ - if (unlikely(fill_level < rx_queue->min_fill)) + if (unlikely(fill_level < rx_queue->min_fill)) { if (fill_level) rx_queue->min_fill = fill_level; + } /* Acquire RX add lock. If this lock is contended, then a fast * fill must already be in progress (e.g. in the refill @@ -551,7 +557,7 @@ static inline void efx_rx_packet_lro(struct efx_channel *channel, struct skb_frag_struct frags; frags.page = rx_buf->page; - frags.page_offset = RX_BUF_OFFSET(rx_buf); + frags.page_offset = efx_rx_buf_offset(rx_buf); frags.size = rx_buf->len; lro_receive_frags(lro_mgr, &frags, rx_buf->len, @@ -596,7 +602,7 @@ static inline struct sk_buff *efx_rx_mk_skb(struct efx_rx_buffer *rx_buf, if (unlikely(rx_buf->len > hdr_len)) { struct skb_frag_struct *frag = skb_shinfo(skb)->frags; frag->page = rx_buf->page; - frag->page_offset = RX_BUF_OFFSET(rx_buf) + hdr_len; + frag->page_offset = efx_rx_buf_offset(rx_buf) + hdr_len; frag->size = skb->len - hdr_len; skb_shinfo(skb)->nr_frags = 1; skb->data_len = frag->size; @@ -683,6 +689,15 @@ void __efx_rx_packet(struct efx_channel *channel, struct sk_buff *skb; int lro = efx->net_dev->features & NETIF_F_LRO; + /* If we're in loopback test, then pass the packet directly to the + * loopback layer, and free the rx_buf here + */ + if (unlikely(efx->loopback_selftest)) { + efx_loopback_rx_packet(efx, rx_buf->data, rx_buf->len); + efx_free_rx_buffer(efx, rx_buf); + goto done; + } + if (rx_buf->skb) { prefetch(skb_shinfo(rx_buf->skb)); @@ -736,7 +751,6 @@ void __efx_rx_packet(struct efx_channel *channel, /* Update allocation strategy method */ channel->rx_alloc_level += RX_ALLOC_FACTOR_SKB; - /* fall-thru */ done: efx->net_dev->last_rx = jiffies; } @@ -842,7 +856,8 @@ void efx_fini_rx_queue(struct efx_rx_queue *rx_queue) /* For a page that is part-way through splitting into RX buffers */ if (rx_queue->buf_page != NULL) { pci_unmap_page(rx_queue->efx->pci_dev, rx_queue->buf_dma_addr, - RX_PAGE_SIZE(rx_queue->efx), PCI_DMA_FROMDEVICE); + efx_rx_buf_size(rx_queue->efx), + PCI_DMA_FROMDEVICE); __free_pages(rx_queue->buf_page, rx_queue->efx->rx_buffer_order); rx_queue->buf_page = NULL; diff --git a/drivers/net/sfc/selftest.c b/drivers/net/sfc/selftest.c new file mode 100644 index 000000000000..3b2de9fe7f27 --- /dev/null +++ b/drivers/net/sfc/selftest.c @@ -0,0 +1,719 @@ +/**************************************************************************** + * Driver for Solarflare Solarstorm network controllers and boards + * Copyright 2005-2006 Fen Systems Ltd. + * Copyright 2006-2008 Solarflare Communications Inc. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published + * by the Free Software Foundation, incorporated herein by reference. + */ + +#include <linux/netdevice.h> +#include <linux/module.h> +#include <linux/delay.h> +#include <linux/kernel_stat.h> +#include <linux/pci.h> +#include <linux/ethtool.h> +#include <linux/ip.h> +#include <linux/in.h> +#include <linux/udp.h> +#include <linux/rtnetlink.h> +#include <asm/io.h> +#include "net_driver.h" +#include "ethtool.h" +#include "efx.h" +#include "falcon.h" +#include "selftest.h" +#include "boards.h" +#include "workarounds.h" +#include "mac.h" + +/* + * Loopback test packet structure + * + * The self-test should stress every RSS vector, and unfortunately + * Falcon only performs RSS on TCP/UDP packets. + */ +struct efx_loopback_payload { + struct ethhdr header; + struct iphdr ip; + struct udphdr udp; + __be16 iteration; + const char msg[64]; +} __attribute__ ((packed)); + +/* Loopback test source MAC address */ +static const unsigned char payload_source[ETH_ALEN] = { + 0x00, 0x0f, 0x53, 0x1b, 0x1b, 0x1b, +}; + +static const char *payload_msg = + "Hello world! This is an Efx loopback test in progress!"; + +/** + * efx_selftest_state - persistent state during a selftest + * @flush: Drop all packets in efx_loopback_rx_packet + * @packet_count: Number of packets being used in this test + * @skbs: An array of skbs transmitted + * @rx_good: RX good packet count + * @rx_bad: RX bad packet count + * @payload: Payload used in tests + */ +struct efx_selftest_state { + int flush; + int packet_count; + struct sk_buff **skbs; + atomic_t rx_good; + atomic_t rx_bad; + struct efx_loopback_payload payload; +}; + +/************************************************************************** + * + * Configurable values + * + **************************************************************************/ + +/* Level of loopback testing + * + * The maximum packet burst length is 16**(n-1), i.e. + * + * - Level 0 : no packets + * - Level 1 : 1 packet + * - Level 2 : 17 packets (1 * 1 packet, 1 * 16 packets) + * - Level 3 : 273 packets (1 * 1 packet, 1 * 16 packet, 1 * 256 packets) + * + */ +static unsigned int loopback_test_level = 3; + +/************************************************************************** + * + * Interrupt and event queue testing + * + **************************************************************************/ + +/* Test generation and receipt of interrupts */ +static int efx_test_interrupts(struct efx_nic *efx, + struct efx_self_tests *tests) +{ + struct efx_channel *channel; + + EFX_LOG(efx, "testing interrupts\n"); + tests->interrupt = -1; + + /* Reset interrupt flag */ + efx->last_irq_cpu = -1; + smp_wmb(); + + /* ACK each interrupting event queue. Receiving an interrupt due to + * traffic before a test event is raised is considered a pass */ + efx_for_each_channel_with_interrupt(channel, efx) { + if (channel->work_pending) + efx_process_channel_now(channel); + if (efx->last_irq_cpu >= 0) + goto success; + } + + falcon_generate_interrupt(efx); + + /* Wait for arrival of test interrupt. */ + EFX_LOG(efx, "waiting for test interrupt\n"); + schedule_timeout_uninterruptible(HZ / 10); + if (efx->last_irq_cpu >= 0) + goto success; + + EFX_ERR(efx, "timed out waiting for interrupt\n"); + return -ETIMEDOUT; + + success: + EFX_LOG(efx, "test interrupt (mode %d) seen on CPU%d\n", + efx->interrupt_mode, efx->last_irq_cpu); + tests->interrupt = 1; + return 0; +} + +/* Test generation and receipt of non-interrupting events */ +static int efx_test_eventq(struct efx_channel *channel, + struct efx_self_tests *tests) +{ + unsigned int magic; + + /* Channel specific code, limited to 20 bits */ + magic = (0x00010150 + channel->channel); + EFX_LOG(channel->efx, "channel %d testing event queue with code %x\n", + channel->channel, magic); + + tests->eventq_dma[channel->channel] = -1; + tests->eventq_int[channel->channel] = 1; /* fake pass */ + tests->eventq_poll[channel->channel] = 1; /* fake pass */ + + /* Reset flag and zero magic word */ + channel->efx->last_irq_cpu = -1; + channel->eventq_magic = 0; + smp_wmb(); + + falcon_generate_test_event(channel, magic); + udelay(1); + + efx_process_channel_now(channel); + if (channel->eventq_magic != magic) { + EFX_ERR(channel->efx, "channel %d failed to see test event\n", + channel->channel); + return -ETIMEDOUT; + } else { + tests->eventq_dma[channel->channel] = 1; + } + + return 0; +} + +/* Test generation and receipt of interrupting events */ +static int efx_test_eventq_irq(struct efx_channel *channel, + struct efx_self_tests *tests) +{ + unsigned int magic, count; + + /* Channel specific code, limited to 20 bits */ + magic = (0x00010150 + channel->channel); + EFX_LOG(channel->efx, "channel %d testing event queue with code %x\n", + channel->channel, magic); + + tests->eventq_dma[channel->channel] = -1; + tests->eventq_int[channel->channel] = -1; + tests->eventq_poll[channel->channel] = -1; + + /* Reset flag and zero magic word */ + channel->efx->last_irq_cpu = -1; + channel->eventq_magic = 0; + smp_wmb(); + + falcon_generate_test_event(channel, magic); + + /* Wait for arrival of interrupt */ + count = 0; + do { + schedule_timeout_uninterruptible(HZ / 100); + + if (channel->work_pending) + efx_process_channel_now(channel); + + if (channel->eventq_magic == magic) + goto eventq_ok; + } while (++count < 2); + + EFX_ERR(channel->efx, "channel %d timed out waiting for event queue\n", + channel->channel); + + /* See if interrupt arrived */ + if (channel->efx->last_irq_cpu >= 0) { + EFX_ERR(channel->efx, "channel %d saw interrupt on CPU%d " + "during event queue test\n", channel->channel, + raw_smp_processor_id()); + tests->eventq_int[channel->channel] = 1; + } + + /* Check to see if event was received even if interrupt wasn't */ + efx_process_channel_now(channel); + if (channel->eventq_magic == magic) { + EFX_ERR(channel->efx, "channel %d event was generated, but " + "failed to trigger an interrupt\n", channel->channel); + tests->eventq_dma[channel->channel] = 1; + } + + return -ETIMEDOUT; + eventq_ok: + EFX_LOG(channel->efx, "channel %d event queue passed\n", + channel->channel); + tests->eventq_dma[channel->channel] = 1; + tests->eventq_int[channel->channel] = 1; + tests->eventq_poll[channel->channel] = 1; + return 0; +} + +/************************************************************************** + * + * PHY testing + * + **************************************************************************/ + +/* Check PHY presence by reading the PHY ID registers */ +static int efx_test_phy(struct efx_nic *efx, + struct efx_self_tests *tests) +{ + u16 physid1, physid2; + struct mii_if_info *mii = &efx->mii; + struct net_device *net_dev = efx->net_dev; + + if (efx->phy_type == PHY_TYPE_NONE) + return 0; + + EFX_LOG(efx, "testing PHY presence\n"); + tests->phy_ok = -1; + + physid1 = mii->mdio_read(net_dev, mii->phy_id, MII_PHYSID1); + physid2 = mii->mdio_read(net_dev, mii->phy_id, MII_PHYSID2); + + if ((physid1 != 0x0000) && (physid1 != 0xffff) && + (physid2 != 0x0000) && (physid2 != 0xffff)) { + EFX_LOG(efx, "found MII PHY %d ID 0x%x:%x\n", + mii->phy_id, physid1, physid2); + tests->phy_ok = 1; + return 0; + } + + EFX_ERR(efx, "no MII PHY present with ID %d\n", mii->phy_id); + return -ENODEV; +} + +/************************************************************************** + * + * Loopback testing + * NB Only one loopback test can be executing concurrently. + * + **************************************************************************/ + +/* Loopback test RX callback + * This is called for each received packet during loopback testing. + */ +void efx_loopback_rx_packet(struct efx_nic *efx, + const char *buf_ptr, int pkt_len) +{ + struct efx_selftest_state *state = efx->loopback_selftest; + struct efx_loopback_payload *received; + struct efx_loopback_payload *payload; + + BUG_ON(!buf_ptr); + + /* If we are just flushing, then drop the packet */ + if ((state == NULL) || state->flush) + return; + + payload = &state->payload; + + received = (struct efx_loopback_payload *) buf_ptr; + received->ip.saddr = payload->ip.saddr; + received->ip.check = payload->ip.check; + + /* Check that header exists */ + if (pkt_len < sizeof(received->header)) { + EFX_ERR(efx, "saw runt RX packet (length %d) in %s loopback " + "test\n", pkt_len, LOOPBACK_MODE(efx)); + goto err; + } + + /* Check that the ethernet header exists */ + if (memcmp(&received->header, &payload->header, ETH_HLEN) != 0) { + EFX_ERR(efx, "saw non-loopback RX packet in %s loopback test\n", + LOOPBACK_MODE(efx)); + goto err; + } + + /* Check packet length */ + if (pkt_len != sizeof(*payload)) { + EFX_ERR(efx, "saw incorrect RX packet length %d (wanted %d) in " + "%s loopback test\n", pkt_len, (int)sizeof(*payload), + LOOPBACK_MODE(efx)); + goto err; + } + + /* Check that IP header matches */ + if (memcmp(&received->ip, &payload->ip, sizeof(payload->ip)) != 0) { + EFX_ERR(efx, "saw corrupted IP header in %s loopback test\n", + LOOPBACK_MODE(efx)); + goto err; + } + + /* Check that msg and padding matches */ + if (memcmp(&received->msg, &payload->msg, sizeof(received->msg)) != 0) { + EFX_ERR(efx, "saw corrupted RX packet in %s loopback test\n", + LOOPBACK_MODE(efx)); + goto err; + } + + /* Check that iteration matches */ + if (received->iteration != payload->iteration) { + EFX_ERR(efx, "saw RX packet from iteration %d (wanted %d) in " + "%s loopback test\n", ntohs(received->iteration), + ntohs(payload->iteration), LOOPBACK_MODE(efx)); + goto err; + } + + /* Increase correct RX count */ + EFX_TRACE(efx, "got loopback RX in %s loopback test\n", + LOOPBACK_MODE(efx)); + + atomic_inc(&state->rx_good); + return; + + err: +#ifdef EFX_ENABLE_DEBUG + if (atomic_read(&state->rx_bad) == 0) { + EFX_ERR(efx, "received packet:\n"); + print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 0x10, 1, + buf_ptr, pkt_len, 0); + EFX_ERR(efx, "expected packet:\n"); + print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 0x10, 1, + &state->payload, sizeof(state->payload), 0); + } +#endif + atomic_inc(&state->rx_bad); +} + +/* Initialise an efx_selftest_state for a new iteration */ +static void efx_iterate_state(struct efx_nic *efx) +{ + struct efx_selftest_state *state = efx->loopback_selftest; + struct net_device *net_dev = efx->net_dev; + struct efx_loopback_payload *payload = &state->payload; + + /* Initialise the layerII header */ + memcpy(&payload->header.h_dest, net_dev->dev_addr, ETH_ALEN); + memcpy(&payload->header.h_source, &payload_source, ETH_ALEN); + payload->header.h_proto = htons(ETH_P_IP); + + /* saddr set later and used as incrementing count */ + payload->ip.daddr = htonl(INADDR_LOOPBACK); + payload->ip.ihl = 5; + payload->ip.check = htons(0xdead); + payload->ip.tot_len = htons(sizeof(*payload) - sizeof(struct ethhdr)); + payload->ip.version = IPVERSION; + payload->ip.protocol = IPPROTO_UDP; + + /* Initialise udp header */ + payload->udp.source = 0; + payload->udp.len = htons(sizeof(*payload) - sizeof(struct ethhdr) - + sizeof(struct iphdr)); + payload->udp.check = 0; /* checksum ignored */ + + /* Fill out payload */ + payload->iteration = htons(ntohs(payload->iteration) + 1); + memcpy(&payload->msg, payload_msg, sizeof(payload_msg)); + + /* Fill out remaining state members */ + atomic_set(&state->rx_good, 0); + atomic_set(&state->rx_bad, 0); + smp_wmb(); +} + +static int efx_tx_loopback(struct efx_tx_queue *tx_queue) +{ + struct efx_nic *efx = tx_queue->efx; + struct efx_selftest_state *state = efx->loopback_selftest; + struct efx_loopback_payload *payload; + struct sk_buff *skb; + int i, rc; + + /* Transmit N copies of buffer */ + for (i = 0; i < state->packet_count; i++) { + /* Allocate an skb, holding an extra reference for + * transmit completion counting */ + skb = alloc_skb(sizeof(state->payload), GFP_KERNEL); + if (!skb) + return -ENOMEM; + state->skbs[i] = skb; + skb_get(skb); + + /* Copy the payload in, incrementing the source address to + * exercise the rss vectors */ + payload = ((struct efx_loopback_payload *) + skb_put(skb, sizeof(state->payload))); + memcpy(payload, &state->payload, sizeof(state->payload)); + payload->ip.saddr = htonl(INADDR_LOOPBACK | (i << 2)); + + /* Ensure everything we've written is visible to the + * interrupt handler. */ + smp_wmb(); + + if (efx_dev_registered(efx)) + netif_tx_lock_bh(efx->net_dev); + rc = efx_xmit(efx, tx_queue, skb); + if (efx_dev_registered(efx)) + netif_tx_unlock_bh(efx->net_dev); + + if (rc != NETDEV_TX_OK) { + EFX_ERR(efx, "TX queue %d could not transmit packet %d " + "of %d in %s loopback test\n", tx_queue->queue, + i + 1, state->packet_count, LOOPBACK_MODE(efx)); + + /* Defer cleaning up the other skbs for the caller */ + kfree_skb(skb); + return -EPIPE; + } + } + + return 0; +} + +static int efx_rx_loopback(struct efx_tx_queue *tx_queue, + struct efx_loopback_self_tests *lb_tests) +{ + struct efx_nic *efx = tx_queue->efx; + struct efx_selftest_state *state = efx->loopback_selftest; + struct sk_buff *skb; + int tx_done = 0, rx_good, rx_bad; + int i, rc = 0; + + if (efx_dev_registered(efx)) + netif_tx_lock_bh(efx->net_dev); + + /* Count the number of tx completions, and decrement the refcnt. Any + * skbs not already completed will be free'd when the queue is flushed */ + for (i=0; i < state->packet_count; i++) { + skb = state->skbs[i]; + if (skb && !skb_shared(skb)) + ++tx_done; + dev_kfree_skb_any(skb); + } + + if (efx_dev_registered(efx)) + netif_tx_unlock_bh(efx->net_dev); + + /* Check TX completion and received packet counts */ + rx_good = atomic_read(&state->rx_good); + rx_bad = atomic_read(&state->rx_bad); + if (tx_done != state->packet_count) { + /* Don't free the skbs; they will be picked up on TX + * overflow or channel teardown. + */ + EFX_ERR(efx, "TX queue %d saw only %d out of an expected %d " + "TX completion events in %s loopback test\n", + tx_queue->queue, tx_done, state->packet_count, + LOOPBACK_MODE(efx)); + rc = -ETIMEDOUT; + /* Allow to fall through so we see the RX errors as well */ + } + + /* We may always be up to a flush away from our desired packet total */ + if (rx_good != state->packet_count) { + EFX_LOG(efx, "TX queue %d saw only %d out of an expected %d " + "received packets in %s loopback test\n", + tx_queue->queue, rx_good, state->packet_count, + LOOPBACK_MODE(efx)); + rc = -ETIMEDOUT; + /* Fall through */ + } + + /* Update loopback test structure */ + lb_tests->tx_sent[tx_queue->queue] += state->packet_count; + lb_tests->tx_done[tx_queue->queue] += tx_done; + lb_tests->rx_good += rx_good; + lb_tests->rx_bad += rx_bad; + + return rc; +} + +static int +efx_test_loopback(struct efx_tx_queue *tx_queue, + struct efx_loopback_self_tests *lb_tests) +{ + struct efx_nic *efx = tx_queue->efx; + struct efx_selftest_state *state = efx->loopback_selftest; + struct efx_channel *channel; + int i, rc = 0; + + for (i = 0; i < loopback_test_level; i++) { + /* Determine how many packets to send */ + state->packet_count = (efx->type->txd_ring_mask + 1) / 3; + state->packet_count = min(1 << (i << 2), state->packet_count); + state->skbs = kzalloc(sizeof(state->skbs[0]) * + state->packet_count, GFP_KERNEL); + if (!state->skbs) + return -ENOMEM; + state->flush = 0; + + EFX_LOG(efx, "TX queue %d testing %s loopback with %d " + "packets\n", tx_queue->queue, LOOPBACK_MODE(efx), + state->packet_count); + + efx_iterate_state(efx); + rc = efx_tx_loopback(tx_queue); + + /* NAPI polling is not enabled, so process channels synchronously */ + schedule_timeout_uninterruptible(HZ / 50); + efx_for_each_channel_with_interrupt(channel, efx) { + if (channel->work_pending) + efx_process_channel_now(channel); + } + + rc |= efx_rx_loopback(tx_queue, lb_tests); + kfree(state->skbs); + + if (rc) { + /* Wait a while to ensure there are no packets + * floating around after a failure. */ + schedule_timeout_uninterruptible(HZ / 10); + return rc; + } + } + + EFX_LOG(efx, "TX queue %d passed %s loopback test with a burst length " + "of %d packets\n", tx_queue->queue, LOOPBACK_MODE(efx), + state->packet_count); + + return rc; +} + +static int efx_test_loopbacks(struct efx_nic *efx, + struct efx_self_tests *tests, + unsigned int loopback_modes) +{ + struct efx_selftest_state *state = efx->loopback_selftest; + struct ethtool_cmd ecmd, ecmd_loopback; + struct efx_tx_queue *tx_queue; + enum efx_loopback_mode old_mode, mode; + int count, rc = 0, link_up; + + rc = efx_ethtool_get_settings(efx->net_dev, &ecmd); + if (rc) { + EFX_ERR(efx, "could not get GMII settings\n"); + return rc; + } + old_mode = efx->loopback_mode; + + /* Disable autonegotiation for the purposes of loopback */ + memcpy(&ecmd_loopback, &ecmd, sizeof(ecmd_loopback)); + if (ecmd_loopback.autoneg == AUTONEG_ENABLE) { + ecmd_loopback.autoneg = AUTONEG_DISABLE; + ecmd_loopback.duplex = DUPLEX_FULL; + ecmd_loopback.speed = SPEED_10000; + } + + rc = efx_ethtool_set_settings(efx->net_dev, &ecmd_loopback); + if (rc) { + EFX_ERR(efx, "could not disable autonegotiation\n"); + goto out; + } + tests->loopback_speed = ecmd_loopback.speed; + tests->loopback_full_duplex = ecmd_loopback.duplex; + + /* Test all supported loopback modes */ + for (mode = LOOPBACK_NONE; mode < LOOPBACK_TEST_MAX; mode++) { + if (!(loopback_modes & (1 << mode))) + continue; + + /* Move the port into the specified loopback mode. */ + state->flush = 1; + efx->loopback_mode = mode; + efx_reconfigure_port(efx); + + /* Wait for the PHY to signal the link is up */ + count = 0; + do { + struct efx_channel *channel = &efx->channel[0]; + + falcon_check_xmac(efx); + schedule_timeout_uninterruptible(HZ / 10); + if (channel->work_pending) + efx_process_channel_now(channel); + /* Wait for PHY events to be processed */ + flush_workqueue(efx->workqueue); + rmb(); + + /* efx->link_up can be 1 even if the XAUI link is down, + * (bug5762). Usually, it's not worth bothering with the + * difference, but for selftests, we need that extra + * guarantee that the link is really, really, up. + */ + link_up = efx->link_up; + if (!falcon_xaui_link_ok(efx)) + link_up = 0; + + } while ((++count < 20) && !link_up); + + /* The link should now be up. If it isn't, there is no point + * in attempting a loopback test */ + if (!link_up) { + EFX_ERR(efx, "loopback %s never came up\n", + LOOPBACK_MODE(efx)); + rc = -EIO; + goto out; + } + + EFX_LOG(efx, "link came up in %s loopback in %d iterations\n", + LOOPBACK_MODE(efx), count); + + /* Test every TX queue */ + efx_for_each_tx_queue(tx_queue, efx) { + rc |= efx_test_loopback(tx_queue, + &tests->loopback[mode]); + if (rc) + goto out; + } + } + + out: + /* Take out of loopback and restore PHY settings */ + state->flush = 1; + efx->loopback_mode = old_mode; + efx_ethtool_set_settings(efx->net_dev, &ecmd); + + return rc; +} + +/************************************************************************** + * + * Entry points + * + *************************************************************************/ + +/* Online (i.e. non-disruptive) testing + * This checks interrupt generation, event delivery and PHY presence. */ +int efx_online_test(struct efx_nic *efx, struct efx_self_tests *tests) +{ + struct efx_channel *channel; + int rc = 0; + + EFX_LOG(efx, "performing online self-tests\n"); + + rc |= efx_test_interrupts(efx, tests); + efx_for_each_channel(channel, efx) { + if (channel->has_interrupt) + rc |= efx_test_eventq_irq(channel, tests); + else + rc |= efx_test_eventq(channel, tests); + } + rc |= efx_test_phy(efx, tests); + + if (rc) + EFX_ERR(efx, "failed online self-tests\n"); + + return rc; +} + +/* Offline (i.e. disruptive) testing + * This checks MAC and PHY loopback on the specified port. */ +int efx_offline_test(struct efx_nic *efx, + struct efx_self_tests *tests, unsigned int loopback_modes) +{ + struct efx_selftest_state *state; + int rc = 0; + + EFX_LOG(efx, "performing offline self-tests\n"); + + /* Create a selftest_state structure to hold state for the test */ + state = kzalloc(sizeof(*state), GFP_KERNEL); + if (state == NULL) { + rc = -ENOMEM; + goto out; + } + + /* Set the port loopback_selftest member. From this point on + * all received packets will be dropped. Mark the state as + * "flushing" so all inflight packets are dropped */ + BUG_ON(efx->loopback_selftest); + state->flush = 1; + efx->loopback_selftest = state; + + rc = efx_test_loopbacks(efx, tests, loopback_modes); + + efx->loopback_selftest = NULL; + wmb(); + kfree(state); + + out: + if (rc) + EFX_ERR(efx, "failed offline self-tests\n"); + + return rc; +} + diff --git a/drivers/net/sfc/selftest.h b/drivers/net/sfc/selftest.h new file mode 100644 index 000000000000..f6999c2b622d --- /dev/null +++ b/drivers/net/sfc/selftest.h @@ -0,0 +1,50 @@ +/**************************************************************************** + * Driver for Solarflare Solarstorm network controllers and boards + * Copyright 2005-2006 Fen Systems Ltd. + * Copyright 2006-2008 Solarflare Communications Inc. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published + * by the Free Software Foundation, incorporated herein by reference. + */ + +#ifndef EFX_SELFTEST_H +#define EFX_SELFTEST_H + +#include "net_driver.h" + +/* + * Self tests + */ + +struct efx_loopback_self_tests { + int tx_sent[EFX_MAX_TX_QUEUES]; + int tx_done[EFX_MAX_TX_QUEUES]; + int rx_good; + int rx_bad; +}; + +/* Efx self test results + * For fields which are not counters, 1 indicates success and -1 + * indicates failure. + */ +struct efx_self_tests { + int interrupt; + int eventq_dma[EFX_MAX_CHANNELS]; + int eventq_int[EFX_MAX_CHANNELS]; + int eventq_poll[EFX_MAX_CHANNELS]; + int phy_ok; + int loopback_speed; + int loopback_full_duplex; + struct efx_loopback_self_tests loopback[LOOPBACK_TEST_MAX]; +}; + +extern void efx_loopback_rx_packet(struct efx_nic *efx, + const char *buf_ptr, int pkt_len); +extern int efx_online_test(struct efx_nic *efx, + struct efx_self_tests *tests); +extern int efx_offline_test(struct efx_nic *efx, + struct efx_self_tests *tests, + unsigned int loopback_modes); + +#endif /* EFX_SELFTEST_H */ diff --git a/drivers/net/sfc/sfe4001.c b/drivers/net/sfc/sfe4001.c index 11fa9fb8f48b..66a0d1442aba 100644 --- a/drivers/net/sfc/sfe4001.c +++ b/drivers/net/sfc/sfe4001.c @@ -116,20 +116,29 @@ void sfe4001_poweroff(struct efx_nic *efx) /* Turn off all power rails */ out = 0xff; - (void) efx_i2c_write(i2c, PCA9539, P0_OUT, &out, 1); + efx_i2c_write(i2c, PCA9539, P0_OUT, &out, 1); /* Disable port 1 outputs on IO expander */ cfg = 0xff; - (void) efx_i2c_write(i2c, PCA9539, P1_CONFIG, &cfg, 1); + efx_i2c_write(i2c, PCA9539, P1_CONFIG, &cfg, 1); /* Disable port 0 outputs on IO expander */ cfg = 0xff; - (void) efx_i2c_write(i2c, PCA9539, P0_CONFIG, &cfg, 1); + efx_i2c_write(i2c, PCA9539, P0_CONFIG, &cfg, 1); /* Clear any over-temperature alert */ - (void) efx_i2c_read(i2c, MAX6647, RSL, &in, 1); + efx_i2c_read(i2c, MAX6647, RSL, &in, 1); } +/* The P0_EN_3V3X line on SFE4001 boards (from A2 onward) is connected + * to the FLASH_CFG_1 input on the DSP. We must keep it high at power- + * up to allow writing the flash (done through MDIO from userland). + */ +unsigned int sfe4001_phy_flash_cfg; +module_param_named(phy_flash_cfg, sfe4001_phy_flash_cfg, uint, 0444); +MODULE_PARM_DESC(phy_flash_cfg, + "Force PHY to enter flash configuration mode"); + /* This board uses an I2C expander to provider power to the PHY, which needs to * be turned on before the PHY can be used. * Context: Process context, rtnl lock held @@ -203,6 +212,8 @@ int sfe4001_poweron(struct efx_nic *efx) out = 0xff & ~((1 << P0_EN_1V2_LBN) | (1 << P0_EN_2V5_LBN) | (1 << P0_EN_3V3X_LBN) | (1 << P0_EN_5V_LBN) | (1 << P0_X_TRST_LBN)); + if (sfe4001_phy_flash_cfg) + out |= 1 << P0_EN_3V3X_LBN; rc = efx_i2c_write(i2c, PCA9539, P0_OUT, &out, 1); if (rc) @@ -226,6 +237,9 @@ int sfe4001_poweron(struct efx_nic *efx) if (in & (1 << P1_AFE_PWD_LBN)) goto done; + /* DSP doesn't look powered in flash config mode */ + if (sfe4001_phy_flash_cfg) + goto done; } while (++count < 20); EFX_INFO(efx, "timed out waiting for power\n"); @@ -239,14 +253,14 @@ done: fail3: /* Turn off all power rails */ out = 0xff; - (void) efx_i2c_write(i2c, PCA9539, P0_OUT, &out, 1); + efx_i2c_write(i2c, PCA9539, P0_OUT, &out, 1); /* Disable port 1 outputs on IO expander */ out = 0xff; - (void) efx_i2c_write(i2c, PCA9539, P1_CONFIG, &out, 1); + efx_i2c_write(i2c, PCA9539, P1_CONFIG, &out, 1); fail2: /* Disable port 0 outputs on IO expander */ out = 0xff; - (void) efx_i2c_write(i2c, PCA9539, P0_CONFIG, &out, 1); + efx_i2c_write(i2c, PCA9539, P0_CONFIG, &out, 1); fail1: return rc; } diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c index a2e9f79e47b1..c0146061c326 100644 --- a/drivers/net/sfc/tenxpress.c +++ b/drivers/net/sfc/tenxpress.c @@ -24,6 +24,11 @@ MDIO_MMDREG_DEVS0_PCS | \ MDIO_MMDREG_DEVS0_PHYXS) +#define TENXPRESS_LOOPBACKS ((1 << LOOPBACK_PHYXS) | \ + (1 << LOOPBACK_PCS) | \ + (1 << LOOPBACK_PMAPMD) | \ + (1 << LOOPBACK_NETWORK)) + /* We complain if we fail to see the link partner as 10G capable this many * times in a row (must be > 1 as sampling the autoneg. registers is racy) */ @@ -72,6 +77,10 @@ #define PMA_PMD_BIST_RXD_LBN (1) #define PMA_PMD_BIST_AFE_LBN (0) +/* Special Software reset register */ +#define PMA_PMD_EXT_CTRL_REG 49152 +#define PMA_PMD_EXT_SSR_LBN 15 + #define BIST_MAX_DELAY (1000) #define BIST_POLL_DELAY (10) @@ -86,6 +95,11 @@ #define PCS_TEST_SELECT_REG 0xd807 /* PRM 10.5.8 */ #define CLK312_EN_LBN 3 +/* PHYXS registers */ +#define PHYXS_TEST1 (49162) +#define LOOPBACK_NEAR_LBN (8) +#define LOOPBACK_NEAR_WIDTH (1) + /* Boot status register */ #define PCS_BOOT_STATUS_REG (0xd000) #define PCS_BOOT_FATAL_ERR_LBN (0) @@ -106,7 +120,9 @@ MODULE_PARM_DESC(crc_error_reset_threshold, struct tenxpress_phy_data { enum tenxpress_state state; + enum efx_loopback_mode loopback_mode; atomic_t bad_crc_count; + int tx_disabled; int bad_lp_tries; }; @@ -195,14 +211,18 @@ static int tenxpress_phy_init(struct efx_nic *efx) int rc = 0; phy_data = kzalloc(sizeof(*phy_data), GFP_KERNEL); + if (!phy_data) + return -ENOMEM; efx->phy_data = phy_data; tenxpress_set_state(efx, TENXPRESS_STATUS_NORMAL); - rc = mdio_clause45_wait_reset_mmds(efx, - TENXPRESS_REQUIRED_DEVS); - if (rc < 0) - goto fail; + if (!sfe4001_phy_flash_cfg) { + rc = mdio_clause45_wait_reset_mmds(efx, + TENXPRESS_REQUIRED_DEVS); + if (rc < 0) + goto fail; + } rc = mdio_clause45_check_mmds(efx, TENXPRESS_REQUIRED_DEVS, 0); if (rc < 0) @@ -225,6 +245,35 @@ static int tenxpress_phy_init(struct efx_nic *efx) return rc; } +static int tenxpress_special_reset(struct efx_nic *efx) +{ + int rc, reg; + + EFX_TRACE(efx, "%s\n", __func__); + + /* Initiate reset */ + reg = mdio_clause45_read(efx, efx->mii.phy_id, + MDIO_MMD_PMAPMD, PMA_PMD_EXT_CTRL_REG); + reg |= (1 << PMA_PMD_EXT_SSR_LBN); + mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD, + PMA_PMD_EXT_CTRL_REG, reg); + + msleep(200); + + /* Wait for the blocks to come out of reset */ + rc = mdio_clause45_wait_reset_mmds(efx, + TENXPRESS_REQUIRED_DEVS); + if (rc < 0) + return rc; + + /* Try and reconfigure the device */ + rc = tenxpress_init(efx); + if (rc < 0) + return rc; + + return 0; +} + static void tenxpress_set_bad_lp(struct efx_nic *efx, int bad_lp) { struct tenxpress_phy_data *pd = efx->phy_data; @@ -299,11 +348,46 @@ static int tenxpress_link_ok(struct efx_nic *efx, int check_lp) return ok; } +static void tenxpress_phyxs_loopback(struct efx_nic *efx) +{ + int phy_id = efx->mii.phy_id; + int ctrl1, ctrl2; + + ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PHYXS, + PHYXS_TEST1); + if (efx->loopback_mode == LOOPBACK_PHYXS) + ctrl2 |= (1 << LOOPBACK_NEAR_LBN); + else + ctrl2 &= ~(1 << LOOPBACK_NEAR_LBN); + if (ctrl1 != ctrl2) + mdio_clause45_write(efx, phy_id, MDIO_MMD_PHYXS, + PHYXS_TEST1, ctrl2); +} + static void tenxpress_phy_reconfigure(struct efx_nic *efx) { + struct tenxpress_phy_data *phy_data = efx->phy_data; + int loop_change = LOOPBACK_OUT_OF(phy_data, efx, + TENXPRESS_LOOPBACKS); + if (!tenxpress_state_is(efx, TENXPRESS_STATUS_NORMAL)) return; + /* When coming out of transmit disable, coming out of low power + * mode, or moving out of any PHY internal loopback mode, + * perform a special software reset */ + if ((phy_data->tx_disabled && !efx->tx_disabled) || + loop_change) { + tenxpress_special_reset(efx); + falcon_reset_xaui(efx); + } + + mdio_clause45_transmit_disable(efx); + mdio_clause45_phy_reconfigure(efx); + tenxpress_phyxs_loopback(efx); + + phy_data->tx_disabled = efx->tx_disabled; + phy_data->loopback_mode = efx->loopback_mode; efx->link_up = tenxpress_link_ok(efx, 0); efx->link_options = GM_LPA_10000FULL; } @@ -431,4 +515,5 @@ struct efx_phy_operations falcon_tenxpress_phy_ops = { .clear_interrupt = tenxpress_phy_clear_interrupt, .reset_xaui = tenxpress_reset_xaui, .mmds = TENXPRESS_REQUIRED_DEVS, + .loopbacks = TENXPRESS_LOOPBACKS, }; diff --git a/drivers/net/sfc/tx.c b/drivers/net/sfc/tx.c index fbb866b2185e..5cdd082ab8f6 100644 --- a/drivers/net/sfc/tx.c +++ b/drivers/net/sfc/tx.c @@ -82,6 +82,46 @@ static inline void efx_dequeue_buffer(struct efx_tx_queue *tx_queue, } } +/** + * struct efx_tso_header - a DMA mapped buffer for packet headers + * @next: Linked list of free ones. + * The list is protected by the TX queue lock. + * @dma_unmap_len: Length to unmap for an oversize buffer, or 0. + * @dma_addr: The DMA address of the header below. + * + * This controls the memory used for a TSO header. Use TSOH_DATA() + * to find the packet header data. Use TSOH_SIZE() to calculate the + * total size required for a given packet header length. TSO headers + * in the free list are exactly %TSOH_STD_SIZE bytes in size. + */ +struct efx_tso_header { + union { + struct efx_tso_header *next; + size_t unmap_len; + }; + dma_addr_t dma_addr; +}; + +static int efx_enqueue_skb_tso(struct efx_tx_queue *tx_queue, + const struct sk_buff *skb); +static void efx_fini_tso(struct efx_tx_queue *tx_queue); +static void efx_tsoh_heap_free(struct efx_tx_queue *tx_queue, + struct efx_tso_header *tsoh); + +static inline void efx_tsoh_free(struct efx_tx_queue *tx_queue, + struct efx_tx_buffer *buffer) +{ + if (buffer->tsoh) { + if (likely(!buffer->tsoh->unmap_len)) { + buffer->tsoh->next = tx_queue->tso_headers_free; + tx_queue->tso_headers_free = buffer->tsoh; + } else { + efx_tsoh_heap_free(tx_queue, buffer->tsoh); + } + buffer->tsoh = NULL; + } +} + /* * Add a socket buffer to a TX queue @@ -114,6 +154,9 @@ static inline int efx_enqueue_skb(struct efx_tx_queue *tx_queue, EFX_BUG_ON_PARANOID(tx_queue->write_count != tx_queue->insert_count); + if (skb_shinfo((struct sk_buff *)skb)->gso_size) + return efx_enqueue_skb_tso(tx_queue, skb); + /* Get size of the initial fragment */ len = skb_headlen(skb); @@ -166,6 +209,8 @@ static inline int efx_enqueue_skb(struct efx_tx_queue *tx_queue, insert_ptr = (tx_queue->insert_count & efx->type->txd_ring_mask); buffer = &tx_queue->buffer[insert_ptr]; + efx_tsoh_free(tx_queue, buffer); + EFX_BUG_ON_PARANOID(buffer->tsoh); EFX_BUG_ON_PARANOID(buffer->skb); EFX_BUG_ON_PARANOID(buffer->len); EFX_BUG_ON_PARANOID(buffer->continuation != 1); @@ -342,7 +387,7 @@ void efx_xmit_done(struct efx_tx_queue *tx_queue, unsigned int index) if (unlikely(tx_queue->stopped)) { fill_level = tx_queue->insert_count - tx_queue->read_count; if (fill_level < EFX_NETDEV_TX_THRESHOLD(tx_queue)) { - EFX_BUG_ON_PARANOID(!NET_DEV_REGISTERED(efx)); + EFX_BUG_ON_PARANOID(!efx_dev_registered(efx)); /* Do this under netif_tx_lock(), to avoid racing * with efx_xmit(). */ @@ -432,6 +477,9 @@ void efx_fini_tx_queue(struct efx_tx_queue *tx_queue) efx_release_tx_buffers(tx_queue); + /* Free up TSO header cache */ + efx_fini_tso(tx_queue); + /* Release queue's stop on port, if any */ if (tx_queue->stopped) { tx_queue->stopped = 0; @@ -450,3 +498,622 @@ void efx_remove_tx_queue(struct efx_tx_queue *tx_queue) } +/* Efx TCP segmentation acceleration. + * + * Why? Because by doing it here in the driver we can go significantly + * faster than the GSO. + * + * Requires TX checksum offload support. + */ + +/* Number of bytes inserted at the start of a TSO header buffer, + * similar to NET_IP_ALIGN. + */ +#if defined(__i386__) || defined(__x86_64__) +#define TSOH_OFFSET 0 +#else +#define TSOH_OFFSET NET_IP_ALIGN +#endif + +#define TSOH_BUFFER(tsoh) ((u8 *)(tsoh + 1) + TSOH_OFFSET) + +/* Total size of struct efx_tso_header, buffer and padding */ +#define TSOH_SIZE(hdr_len) \ + (sizeof(struct efx_tso_header) + TSOH_OFFSET + hdr_len) + +/* Size of blocks on free list. Larger blocks must be allocated from + * the heap. + */ +#define TSOH_STD_SIZE 128 + +#define PTR_DIFF(p1, p2) ((u8 *)(p1) - (u8 *)(p2)) +#define ETH_HDR_LEN(skb) (skb_network_header(skb) - (skb)->data) +#define SKB_TCP_OFF(skb) PTR_DIFF(tcp_hdr(skb), (skb)->data) +#define SKB_IPV4_OFF(skb) PTR_DIFF(ip_hdr(skb), (skb)->data) + +/** + * struct tso_state - TSO state for an SKB + * @remaining_len: Bytes of data we've yet to segment + * @seqnum: Current sequence number + * @packet_space: Remaining space in current packet + * @ifc: Input fragment cursor. + * Where we are in the current fragment of the incoming SKB. These + * values get updated in place when we split a fragment over + * multiple packets. + * @p: Parameters. + * These values are set once at the start of the TSO send and do + * not get changed as the routine progresses. + * + * The state used during segmentation. It is put into this data structure + * just to make it easy to pass into inline functions. + */ +struct tso_state { + unsigned remaining_len; + unsigned seqnum; + unsigned packet_space; + + struct { + /* DMA address of current position */ + dma_addr_t dma_addr; + /* Remaining length */ + unsigned int len; + /* DMA address and length of the whole fragment */ + unsigned int unmap_len; + dma_addr_t unmap_addr; + struct page *page; + unsigned page_off; + } ifc; + + struct { + /* The number of bytes of header */ + unsigned int header_length; + + /* The number of bytes to put in each outgoing segment. */ + int full_packet_size; + + /* Current IPv4 ID, host endian. */ + unsigned ipv4_id; + } p; +}; + + +/* + * Verify that our various assumptions about sk_buffs and the conditions + * under which TSO will be attempted hold true. + */ +static inline void efx_tso_check_safe(const struct sk_buff *skb) +{ + EFX_BUG_ON_PARANOID(skb->protocol != htons(ETH_P_IP)); + EFX_BUG_ON_PARANOID(((struct ethhdr *)skb->data)->h_proto != + skb->protocol); + EFX_BUG_ON_PARANOID(ip_hdr(skb)->protocol != IPPROTO_TCP); + EFX_BUG_ON_PARANOID((PTR_DIFF(tcp_hdr(skb), skb->data) + + (tcp_hdr(skb)->doff << 2u)) > + skb_headlen(skb)); +} + + +/* + * Allocate a page worth of efx_tso_header structures, and string them + * into the tx_queue->tso_headers_free linked list. Return 0 or -ENOMEM. + */ +static int efx_tsoh_block_alloc(struct efx_tx_queue *tx_queue) +{ + + struct pci_dev *pci_dev = tx_queue->efx->pci_dev; + struct efx_tso_header *tsoh; + dma_addr_t dma_addr; + u8 *base_kva, *kva; + + base_kva = pci_alloc_consistent(pci_dev, PAGE_SIZE, &dma_addr); + if (base_kva == NULL) { + EFX_ERR(tx_queue->efx, "Unable to allocate page for TSO" + " headers\n"); + return -ENOMEM; + } + + /* pci_alloc_consistent() allocates pages. */ + EFX_BUG_ON_PARANOID(dma_addr & (PAGE_SIZE - 1u)); + + for (kva = base_kva; kva < base_kva + PAGE_SIZE; kva += TSOH_STD_SIZE) { + tsoh = (struct efx_tso_header *)kva; + tsoh->dma_addr = dma_addr + (TSOH_BUFFER(tsoh) - base_kva); + tsoh->next = tx_queue->tso_headers_free; + tx_queue->tso_headers_free = tsoh; + } + + return 0; +} + + +/* Free up a TSO header, and all others in the same page. */ +static void efx_tsoh_block_free(struct efx_tx_queue *tx_queue, + struct efx_tso_header *tsoh, + struct pci_dev *pci_dev) +{ + struct efx_tso_header **p; + unsigned long base_kva; + dma_addr_t base_dma; + + base_kva = (unsigned long)tsoh & PAGE_MASK; + base_dma = tsoh->dma_addr & PAGE_MASK; + + p = &tx_queue->tso_headers_free; + while (*p != NULL) { + if (((unsigned long)*p & PAGE_MASK) == base_kva) + *p = (*p)->next; + else + p = &(*p)->next; + } + + pci_free_consistent(pci_dev, PAGE_SIZE, (void *)base_kva, base_dma); +} + +static struct efx_tso_header * +efx_tsoh_heap_alloc(struct efx_tx_queue *tx_queue, size_t header_len) +{ + struct efx_tso_header *tsoh; + + tsoh = kmalloc(TSOH_SIZE(header_len), GFP_ATOMIC | GFP_DMA); + if (unlikely(!tsoh)) + return NULL; + + tsoh->dma_addr = pci_map_single(tx_queue->efx->pci_dev, + TSOH_BUFFER(tsoh), header_len, + PCI_DMA_TODEVICE); + if (unlikely(pci_dma_mapping_error(tsoh->dma_addr))) { + kfree(tsoh); + return NULL; + } + + tsoh->unmap_len = header_len; + return tsoh; +} + +static void +efx_tsoh_heap_free(struct efx_tx_queue *tx_queue, struct efx_tso_header *tsoh) +{ + pci_unmap_single(tx_queue->efx->pci_dev, + tsoh->dma_addr, tsoh->unmap_len, + PCI_DMA_TODEVICE); + kfree(tsoh); +} + +/** + * efx_tx_queue_insert - push descriptors onto the TX queue + * @tx_queue: Efx TX queue + * @dma_addr: DMA address of fragment + * @len: Length of fragment + * @skb: Only non-null for end of last segment + * @end_of_packet: True if last fragment in a packet + * @unmap_addr: DMA address of fragment for unmapping + * @unmap_len: Only set this in last segment of a fragment + * + * Push descriptors onto the TX queue. Return 0 on success or 1 if + * @tx_queue full. + */ +static int efx_tx_queue_insert(struct efx_tx_queue *tx_queue, + dma_addr_t dma_addr, unsigned len, + const struct sk_buff *skb, int end_of_packet, + dma_addr_t unmap_addr, unsigned unmap_len) +{ + struct efx_tx_buffer *buffer; + struct efx_nic *efx = tx_queue->efx; + unsigned dma_len, fill_level, insert_ptr, misalign; + int q_space; + + EFX_BUG_ON_PARANOID(len <= 0); + + fill_level = tx_queue->insert_count - tx_queue->old_read_count; + /* -1 as there is no way to represent all descriptors used */ + q_space = efx->type->txd_ring_mask - 1 - fill_level; + + while (1) { + if (unlikely(q_space-- <= 0)) { + /* It might be that completions have happened + * since the xmit path last checked. Update + * the xmit path's copy of read_count. + */ + ++tx_queue->stopped; + /* This memory barrier protects the change of + * stopped from the access of read_count. */ + smp_mb(); + tx_queue->old_read_count = + *(volatile unsigned *)&tx_queue->read_count; + fill_level = (tx_queue->insert_count + - tx_queue->old_read_count); + q_space = efx->type->txd_ring_mask - 1 - fill_level; + if (unlikely(q_space-- <= 0)) + return 1; + smp_mb(); + --tx_queue->stopped; + } + + insert_ptr = tx_queue->insert_count & efx->type->txd_ring_mask; + buffer = &tx_queue->buffer[insert_ptr]; + ++tx_queue->insert_count; + + EFX_BUG_ON_PARANOID(tx_queue->insert_count - + tx_queue->read_count > + efx->type->txd_ring_mask); + + efx_tsoh_free(tx_queue, buffer); + EFX_BUG_ON_PARANOID(buffer->len); + EFX_BUG_ON_PARANOID(buffer->unmap_len); + EFX_BUG_ON_PARANOID(buffer->skb); + EFX_BUG_ON_PARANOID(buffer->continuation != 1); + EFX_BUG_ON_PARANOID(buffer->tsoh); + + buffer->dma_addr = dma_addr; + + /* Ensure we do not cross a boundary unsupported by H/W */ + dma_len = (~dma_addr & efx->type->tx_dma_mask) + 1; + + misalign = (unsigned)dma_addr & efx->type->bug5391_mask; + if (misalign && dma_len + misalign > 512) + dma_len = 512 - misalign; + + /* If there is enough space to send then do so */ + if (dma_len >= len) + break; + + buffer->len = dma_len; /* Don't set the other members */ + dma_addr += dma_len; + len -= dma_len; + } + + EFX_BUG_ON_PARANOID(!len); + buffer->len = len; + buffer->skb = skb; + buffer->continuation = !end_of_packet; + buffer->unmap_addr = unmap_addr; + buffer->unmap_len = unmap_len; + return 0; +} + + +/* + * Put a TSO header into the TX queue. + * + * This is special-cased because we know that it is small enough to fit in + * a single fragment, and we know it doesn't cross a page boundary. It + * also allows us to not worry about end-of-packet etc. + */ +static inline void efx_tso_put_header(struct efx_tx_queue *tx_queue, + struct efx_tso_header *tsoh, unsigned len) +{ + struct efx_tx_buffer *buffer; + + buffer = &tx_queue->buffer[tx_queue->insert_count & + tx_queue->efx->type->txd_ring_mask]; + efx_tsoh_free(tx_queue, buffer); + EFX_BUG_ON_PARANOID(buffer->len); + EFX_BUG_ON_PARANOID(buffer->unmap_len); + EFX_BUG_ON_PARANOID(buffer->skb); + EFX_BUG_ON_PARANOID(buffer->continuation != 1); + EFX_BUG_ON_PARANOID(buffer->tsoh); + buffer->len = len; + buffer->dma_addr = tsoh->dma_addr; + buffer->tsoh = tsoh; + + ++tx_queue->insert_count; +} + + +/* Remove descriptors put into a tx_queue. */ +static void efx_enqueue_unwind(struct efx_tx_queue *tx_queue) +{ + struct efx_tx_buffer *buffer; + + /* Work backwards until we hit the original insert pointer value */ + while (tx_queue->insert_count != tx_queue->write_count) { + --tx_queue->insert_count; + buffer = &tx_queue->buffer[tx_queue->insert_count & + tx_queue->efx->type->txd_ring_mask]; + efx_tsoh_free(tx_queue, buffer); + EFX_BUG_ON_PARANOID(buffer->skb); + buffer->len = 0; + buffer->continuation = 1; + if (buffer->unmap_len) { + pci_unmap_page(tx_queue->efx->pci_dev, + buffer->unmap_addr, + buffer->unmap_len, PCI_DMA_TODEVICE); + buffer->unmap_len = 0; + } + } +} + + +/* Parse the SKB header and initialise state. */ +static inline void tso_start(struct tso_state *st, const struct sk_buff *skb) +{ + /* All ethernet/IP/TCP headers combined size is TCP header size + * plus offset of TCP header relative to start of packet. + */ + st->p.header_length = ((tcp_hdr(skb)->doff << 2u) + + PTR_DIFF(tcp_hdr(skb), skb->data)); + st->p.full_packet_size = (st->p.header_length + + skb_shinfo(skb)->gso_size); + + st->p.ipv4_id = ntohs(ip_hdr(skb)->id); + st->seqnum = ntohl(tcp_hdr(skb)->seq); + + EFX_BUG_ON_PARANOID(tcp_hdr(skb)->urg); + EFX_BUG_ON_PARANOID(tcp_hdr(skb)->syn); + EFX_BUG_ON_PARANOID(tcp_hdr(skb)->rst); + + st->packet_space = st->p.full_packet_size; + st->remaining_len = skb->len - st->p.header_length; +} + + +/** + * tso_get_fragment - record fragment details and map for DMA + * @st: TSO state + * @efx: Efx NIC + * @data: Pointer to fragment data + * @len: Length of fragment + * + * Record fragment details and map for DMA. Return 0 on success, or + * -%ENOMEM if DMA mapping fails. + */ +static inline int tso_get_fragment(struct tso_state *st, struct efx_nic *efx, + int len, struct page *page, int page_off) +{ + + st->ifc.unmap_addr = pci_map_page(efx->pci_dev, page, page_off, + len, PCI_DMA_TODEVICE); + if (likely(!pci_dma_mapping_error(st->ifc.unmap_addr))) { + st->ifc.unmap_len = len; + st->ifc.len = len; + st->ifc.dma_addr = st->ifc.unmap_addr; + st->ifc.page = page; + st->ifc.page_off = page_off; + return 0; + } + return -ENOMEM; +} + + +/** + * tso_fill_packet_with_fragment - form descriptors for the current fragment + * @tx_queue: Efx TX queue + * @skb: Socket buffer + * @st: TSO state + * + * Form descriptors for the current fragment, until we reach the end + * of fragment or end-of-packet. Return 0 on success, 1 if not enough + * space in @tx_queue. + */ +static inline int tso_fill_packet_with_fragment(struct efx_tx_queue *tx_queue, + const struct sk_buff *skb, + struct tso_state *st) +{ + + int n, end_of_packet, rc; + + if (st->ifc.len == 0) + return 0; + if (st->packet_space == 0) + return 0; + + EFX_BUG_ON_PARANOID(st->ifc.len <= 0); + EFX_BUG_ON_PARANOID(st->packet_space <= 0); + + n = min(st->ifc.len, st->packet_space); + + st->packet_space -= n; + st->remaining_len -= n; + st->ifc.len -= n; + st->ifc.page_off += n; + end_of_packet = st->remaining_len == 0 || st->packet_space == 0; + + rc = efx_tx_queue_insert(tx_queue, st->ifc.dma_addr, n, + st->remaining_len ? NULL : skb, + end_of_packet, st->ifc.unmap_addr, + st->ifc.len ? 0 : st->ifc.unmap_len); + + st->ifc.dma_addr += n; + + return rc; +} + + +/** + * tso_start_new_packet - generate a new header and prepare for the new packet + * @tx_queue: Efx TX queue + * @skb: Socket buffer + * @st: TSO state + * + * Generate a new header and prepare for the new packet. Return 0 on + * success, or -1 if failed to alloc header. + */ +static inline int tso_start_new_packet(struct efx_tx_queue *tx_queue, + const struct sk_buff *skb, + struct tso_state *st) +{ + struct efx_tso_header *tsoh; + struct iphdr *tsoh_iph; + struct tcphdr *tsoh_th; + unsigned ip_length; + u8 *header; + + /* Allocate a DMA-mapped header buffer. */ + if (likely(TSOH_SIZE(st->p.header_length) <= TSOH_STD_SIZE)) { + if (tx_queue->tso_headers_free == NULL) { + if (efx_tsoh_block_alloc(tx_queue)) + return -1; + } + EFX_BUG_ON_PARANOID(!tx_queue->tso_headers_free); + tsoh = tx_queue->tso_headers_free; + tx_queue->tso_headers_free = tsoh->next; + tsoh->unmap_len = 0; + } else { + tx_queue->tso_long_headers++; + tsoh = efx_tsoh_heap_alloc(tx_queue, st->p.header_length); + if (unlikely(!tsoh)) + return -1; + } + + header = TSOH_BUFFER(tsoh); + tsoh_th = (struct tcphdr *)(header + SKB_TCP_OFF(skb)); + tsoh_iph = (struct iphdr *)(header + SKB_IPV4_OFF(skb)); + + /* Copy and update the headers. */ + memcpy(header, skb->data, st->p.header_length); + + tsoh_th->seq = htonl(st->seqnum); + st->seqnum += skb_shinfo(skb)->gso_size; + if (st->remaining_len > skb_shinfo(skb)->gso_size) { + /* This packet will not finish the TSO burst. */ + ip_length = st->p.full_packet_size - ETH_HDR_LEN(skb); + tsoh_th->fin = 0; + tsoh_th->psh = 0; + } else { + /* This packet will be the last in the TSO burst. */ + ip_length = (st->p.header_length - ETH_HDR_LEN(skb) + + st->remaining_len); + tsoh_th->fin = tcp_hdr(skb)->fin; + tsoh_th->psh = tcp_hdr(skb)->psh; + } + tsoh_iph->tot_len = htons(ip_length); + + /* Linux leaves suitable gaps in the IP ID space for us to fill. */ + tsoh_iph->id = htons(st->p.ipv4_id); + st->p.ipv4_id++; + + st->packet_space = skb_shinfo(skb)->gso_size; + ++tx_queue->tso_packets; + + /* Form a descriptor for this header. */ + efx_tso_put_header(tx_queue, tsoh, st->p.header_length); + + return 0; +} + + +/** + * efx_enqueue_skb_tso - segment and transmit a TSO socket buffer + * @tx_queue: Efx TX queue + * @skb: Socket buffer + * + * Context: You must hold netif_tx_lock() to call this function. + * + * Add socket buffer @skb to @tx_queue, doing TSO or return != 0 if + * @skb was not enqueued. In all cases @skb is consumed. Return + * %NETDEV_TX_OK or %NETDEV_TX_BUSY. + */ +static int efx_enqueue_skb_tso(struct efx_tx_queue *tx_queue, + const struct sk_buff *skb) +{ + int frag_i, rc, rc2 = NETDEV_TX_OK; + struct tso_state state; + skb_frag_t *f; + + /* Verify TSO is safe - these checks should never fail. */ + efx_tso_check_safe(skb); + + EFX_BUG_ON_PARANOID(tx_queue->write_count != tx_queue->insert_count); + + tso_start(&state, skb); + + /* Assume that skb header area contains exactly the headers, and + * all payload is in the frag list. + */ + if (skb_headlen(skb) == state.p.header_length) { + /* Grab the first payload fragment. */ + EFX_BUG_ON_PARANOID(skb_shinfo(skb)->nr_frags < 1); + frag_i = 0; + f = &skb_shinfo(skb)->frags[frag_i]; + rc = tso_get_fragment(&state, tx_queue->efx, + f->size, f->page, f->page_offset); + if (rc) + goto mem_err; + } else { + /* It may look like this code fragment assumes that the + * skb->data portion does not cross a page boundary, but + * that is not the case. It is guaranteed to be direct + * mapped memory, and therefore is physically contiguous, + * and so DMA will work fine. kmap_atomic() on this region + * will just return the direct mapping, so that will work + * too. + */ + int page_off = (unsigned long)skb->data & (PAGE_SIZE - 1); + int hl = state.p.header_length; + rc = tso_get_fragment(&state, tx_queue->efx, + skb_headlen(skb) - hl, + virt_to_page(skb->data), page_off + hl); + if (rc) + goto mem_err; + frag_i = -1; + } + + if (tso_start_new_packet(tx_queue, skb, &state) < 0) + goto mem_err; + + while (1) { + rc = tso_fill_packet_with_fragment(tx_queue, skb, &state); + if (unlikely(rc)) + goto stop; + + /* Move onto the next fragment? */ + if (state.ifc.len == 0) { + if (++frag_i >= skb_shinfo(skb)->nr_frags) + /* End of payload reached. */ + break; + f = &skb_shinfo(skb)->frags[frag_i]; + rc = tso_get_fragment(&state, tx_queue->efx, + f->size, f->page, f->page_offset); + if (rc) + goto mem_err; + } + + /* Start at new packet? */ + if (state.packet_space == 0 && + tso_start_new_packet(tx_queue, skb, &state) < 0) + goto mem_err; + } + + /* Pass off to hardware */ + falcon_push_buffers(tx_queue); + + tx_queue->tso_bursts++; + return NETDEV_TX_OK; + + mem_err: + EFX_ERR(tx_queue->efx, "Out of memory for TSO headers, or PCI mapping" + " error\n"); + dev_kfree_skb_any((struct sk_buff *)skb); + goto unwind; + + stop: + rc2 = NETDEV_TX_BUSY; + + /* Stop the queue if it wasn't stopped before. */ + if (tx_queue->stopped == 1) + efx_stop_queue(tx_queue->efx); + + unwind: + efx_enqueue_unwind(tx_queue); + return rc2; +} + + +/* + * Free up all TSO datastructures associated with tx_queue. This + * routine should be called only once the tx_queue is both empty and + * will no longer be used. + */ +static void efx_fini_tso(struct efx_tx_queue *tx_queue) +{ + unsigned i; + + if (tx_queue->buffer) { + for (i = 0; i <= tx_queue->efx->type->txd_ring_mask; ++i) + efx_tsoh_free(tx_queue, &tx_queue->buffer[i]); + } + + while (tx_queue->tso_headers_free != NULL) + efx_tsoh_block_free(tx_queue, tx_queue->tso_headers_free, + tx_queue->efx->pci_dev); +} diff --git a/drivers/net/sfc/workarounds.h b/drivers/net/sfc/workarounds.h index dca62f190198..35ab19c27f8d 100644 --- a/drivers/net/sfc/workarounds.h +++ b/drivers/net/sfc/workarounds.h @@ -16,7 +16,7 @@ */ #define EFX_WORKAROUND_ALWAYS(efx) 1 -#define EFX_WORKAROUND_FALCON_A(efx) (FALCON_REV(efx) <= FALCON_REV_A1) +#define EFX_WORKAROUND_FALCON_A(efx) (falcon_rev(efx) <= FALCON_REV_A1) /* XAUI resets if link not detected */ #define EFX_WORKAROUND_5147 EFX_WORKAROUND_ALWAYS diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c index 66dd5bf1eaa9..f3684ad28887 100644 --- a/drivers/net/sfc/xfp_phy.c +++ b/drivers/net/sfc/xfp_phy.c @@ -24,6 +24,10 @@ MDIO_MMDREG_DEVS0_PMAPMD | \ MDIO_MMDREG_DEVS0_PHYXS) +#define XFP_LOOPBACKS ((1 << LOOPBACK_PCS) | \ + (1 << LOOPBACK_PMAPMD) | \ + (1 << LOOPBACK_NETWORK)) + /****************************************************************************/ /* Quake-specific MDIO registers */ #define MDIO_QUAKE_LED0_REG (0xD006) @@ -35,6 +39,10 @@ void xfp_set_led(struct efx_nic *p, int led, int mode) mode); } +struct xfp_phy_data { + int tx_disabled; +}; + #define XFP_MAX_RESET_TIME 500 #define XFP_RESET_WAIT 10 @@ -72,18 +80,33 @@ static int xfp_reset_phy(struct efx_nic *efx) static int xfp_phy_init(struct efx_nic *efx) { + struct xfp_phy_data *phy_data; u32 devid = mdio_clause45_read_id(efx, MDIO_MMD_PHYXS); int rc; + phy_data = kzalloc(sizeof(struct xfp_phy_data), GFP_KERNEL); + if (!phy_data) + return -ENOMEM; + efx->phy_data = phy_data; + EFX_INFO(efx, "XFP: PHY ID reg %x (OUI %x model %x revision" " %x)\n", devid, MDIO_ID_OUI(devid), MDIO_ID_MODEL(devid), MDIO_ID_REV(devid)); + phy_data->tx_disabled = efx->tx_disabled; + rc = xfp_reset_phy(efx); EFX_INFO(efx, "XFP: PHY init %s.\n", rc ? "failed" : "successful"); + if (rc < 0) + goto fail; + return 0; + + fail: + kfree(efx->phy_data); + efx->phy_data = NULL; return rc; } @@ -110,6 +133,16 @@ static int xfp_phy_check_hw(struct efx_nic *efx) static void xfp_phy_reconfigure(struct efx_nic *efx) { + struct xfp_phy_data *phy_data = efx->phy_data; + + /* Reset the PHY when moving from tx off to tx on */ + if (phy_data->tx_disabled && !efx->tx_disabled) + xfp_reset_phy(efx); + + mdio_clause45_transmit_disable(efx); + mdio_clause45_phy_reconfigure(efx); + + phy_data->tx_disabled = efx->tx_disabled; efx->link_up = xfp_link_ok(efx); efx->link_options = GM_LPA_10000FULL; } @@ -119,6 +152,10 @@ static void xfp_phy_fini(struct efx_nic *efx) { /* Clobber the LED if it was blinking */ efx->board_info.blink(efx, 0); + + /* Free the context block */ + kfree(efx->phy_data); + efx->phy_data = NULL; } struct efx_phy_operations falcon_xfp_phy_ops = { @@ -129,4 +166,5 @@ struct efx_phy_operations falcon_xfp_phy_ops = { .clear_interrupt = xfp_phy_clear_interrupt, .reset_xaui = efx_port_dummy_op_void, .mmds = XFP_REQUIRED_DEVS, + .loopbacks = XFP_LOOPBACKS, }; |