sfc: Move and rename Falcon/Siena common NIC operations

Add efx_nic_type operations for the many efx_nic functions that need to be implemented different on EF10. For now, change most of the existing efx_nic_*() functions into inline wrappers. As a later step, we may be able to improve branch prediction for operations used on the fast path by copying the pointers into each queue/channel structure. Move the Falcon/Siena implementations to new file farch.c and rename the functions and static data to use a prefix of 'efx_farch_'. Move efx_may_push_tx_desc() to nic.h, as the EF10 TX code will also use it. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
author: Ben Hutchings <bhutchings@solarflare.com> 2013-08-21 19:51:04 +0100
committer: Ben Hutchings <bhutchings@solarflare.com> 2013-08-21 20:19:05 +0100
commit: 86094f7f38ff711f3db8497fcb4d2e109100f497 (patch)
tree: c9c8d9418b32d0a2abf6220a22328ceeb9130244
parent: e42c3d85af629697699c89aecba481527a1da898 (diff)
9 files changed, 2088 insertions, 1858 deletions
diff --git a/drivers/net/ethernet/sfc/Makefile b/drivers/net/ethernet/sfc/Makefile
index 5b31d8a4ae5..ef7410f014d 100644
--- a/drivers/net/ethernet/sfc/Makefile
+++ b/drivers/net/ethernet/sfc/Makefile
@@ -1,4 +1,5 @@
-sfc-y			+= efx.o nic.o falcon.o siena.o tx.o rx.o filter.o \
+sfc-y			+= efx.o nic.o farch.o falcon.o siena.o tx.o rx.o \
+			   filter.o \
 			   selftest.o ethtool.o qt202x_phy.o mdio_10g.o \
 			   tenxpress.o txc43128_phy.o falcon_boards.o \
 			   mcdi.o mcdi_port.o mcdi_mon.o ptp.o
diff --git a/drivers/net/ethernet/sfc/efx.c b/drivers/net/ethernet/sfc/efx.c
index 9c6555c12ac..872b9f5b38a 100644
--- a/drivers/net/ethernet/sfc/efx.c
+++ b/drivers/net/ethernet/sfc/efx.c
@@ -1386,7 +1386,7 @@ static void efx_enable_interrupts(struct efx_nic *efx)
 		efx->eeh_disabled_legacy_irq = false;
 	}
 
-	efx_nic_enable_interrupts(efx);
+	efx->type->irq_enable_master(efx);
 
 	efx_for_each_channel(channel, efx) {
 		if (channel->type->keep_eventq)
@@ -1407,7 +1407,7 @@ static void efx_disable_interrupts(struct efx_nic *efx)
 			efx_fini_eventq(channel);
 	}
 
-	efx_nic_disable_interrupts(efx);
+	efx->type->irq_disable_non_ev(efx);
 }
 
 static void efx_remove_interrupts(struct efx_nic *efx)
diff --git a/drivers/net/ethernet/sfc/falcon.c b/drivers/net/ethernet/sfc/falcon.c
index c8efcb0efde..fe83c26c4b8 100644
--- a/drivers/net/ethernet/sfc/falcon.c
+++ b/drivers/net/ethernet/sfc/falcon.c
@@ -346,7 +346,7 @@ static inline void falcon_irq_ack_a1(struct efx_nic *efx)
 }
 
 
-irqreturn_t falcon_legacy_interrupt_a1(int irq, void *dev_id)
+static irqreturn_t falcon_legacy_interrupt_a1(int irq, void *dev_id)
 {
 	struct efx_nic *efx = dev_id;
 	efx_oword_t *int_ker = efx->irq_status.addr;
@@ -373,7 +373,7 @@ irqreturn_t falcon_legacy_interrupt_a1(int irq, void *dev_id)
 	/* Check to see if we have a serious error condition */
 	syserr = EFX_OWORD_FIELD(*int_ker, FSF_AZ_NET_IVEC_FATAL_INT);
 	if (unlikely(syserr))
-		return efx_nic_fatal_interrupt(efx);
+		return efx_farch_fatal_interrupt(efx);
 
 	/* Determine interrupting queues, clear interrupt status
 	 * register and acknowledge the device interrupt.
@@ -1558,7 +1558,7 @@ static int falcon_test_nvram(struct efx_nic *efx)
 	return falcon_read_nvram(efx, NULL);
 }
 
-static const struct efx_nic_register_test falcon_b0_register_tests[] = {
+static const struct efx_farch_register_test falcon_b0_register_tests[] = {
 	{ FR_AZ_ADR_REGION,
 	  EFX_OWORD32(0x0003FFFF, 0x0003FFFF, 0x0003FFFF, 0x0003FFFF) },
 	{ FR_AZ_RX_CFG,
@@ -1618,8 +1618,8 @@ falcon_b0_test_chip(struct efx_nic *efx, struct efx_self_tests *tests)
 	efx_reset_down(efx, reset_method);
 
 	tests->registers =
-		efx_nic_test_registers(efx, falcon_b0_register_tests,
-				       ARRAY_SIZE(falcon_b0_register_tests))
+		efx_farch_test_registers(efx, falcon_b0_register_tests,
+					 ARRAY_SIZE(falcon_b0_register_tests))
 		? -1 : 1;
 
 	rc = falcon_reset_hw(efx, reset_method);
@@ -1984,7 +1984,7 @@ static int falcon_probe_nic(struct efx_nic *efx)
 
 	rc = -ENODEV;
 
-	if (efx_nic_fpga_ver(efx) != 0) {
+	if (efx_farch_fpga_ver(efx) != 0) {
 		netif_err(efx, probe, efx->net_dev,
 			  "Falcon FPGA not supported\n");
 		goto fail1;
@@ -2218,7 +2218,7 @@ static int falcon_init_nic(struct efx_nic *efx)
 		efx_writeo(efx, &temp, FR_BZ_DP_CTRL);
 	}
 
-	efx_nic_init_common(efx);
+	efx_farch_init_common(efx);
 
 	return 0;
 }
@@ -2367,6 +2367,28 @@ const struct efx_nic_type falcon_a1_nic_type = {
 	.set_wol = falcon_set_wol,
 	.resume_wol = efx_port_dummy_op_void,
 	.test_nvram = falcon_test_nvram,
+	.irq_enable_master = efx_farch_irq_enable_master,
+	.irq_test_generate = efx_farch_irq_test_generate,
+	.irq_disable_non_ev = efx_farch_irq_disable_master,
+	.irq_handle_msi = efx_farch_msi_interrupt,
+	.irq_handle_legacy = falcon_legacy_interrupt_a1,
+	.tx_probe = efx_farch_tx_probe,
+	.tx_init = efx_farch_tx_init,
+	.tx_remove = efx_farch_tx_remove,
+	.tx_write = efx_farch_tx_write,
+	.rx_push_indir_table = efx_farch_rx_push_indir_table,
+	.rx_probe = efx_farch_rx_probe,
+	.rx_init = efx_farch_rx_init,
+	.rx_remove = efx_farch_rx_remove,
+	.rx_write = efx_farch_rx_write,
+	.rx_defer_refill = efx_farch_rx_defer_refill,
+	.ev_probe = efx_farch_ev_probe,
+	.ev_init = efx_farch_ev_init,
+	.ev_fini = efx_farch_ev_fini,
+	.ev_remove = efx_farch_ev_remove,
+	.ev_process = efx_farch_ev_process,
+	.ev_read_ack = efx_farch_ev_read_ack,
+	.ev_test_generate = efx_farch_ev_test_generate,
 
 	.revision = EFX_REV_FALCON_A1,
 	.mem_map_size = 0x20000,
@@ -2414,6 +2436,28 @@ const struct efx_nic_type falcon_b0_nic_type = {
 	.resume_wol = efx_port_dummy_op_void,
 	.test_chip = falcon_b0_test_chip,
 	.test_nvram = falcon_test_nvram,
+	.irq_enable_master = efx_farch_irq_enable_master,
+	.irq_test_generate = efx_farch_irq_test_generate,
+	.irq_disable_non_ev = efx_farch_irq_disable_master,
+	.irq_handle_msi = efx_farch_msi_interrupt,
+	.irq_handle_legacy = efx_farch_legacy_interrupt,
+	.tx_probe = efx_farch_tx_probe,
+	.tx_init = efx_farch_tx_init,
+	.tx_remove = efx_farch_tx_remove,
+	.tx_write = efx_farch_tx_write,
+	.rx_push_indir_table = efx_farch_rx_push_indir_table,
+	.rx_probe = efx_farch_rx_probe,
+	.rx_init = efx_farch_rx_init,
+	.rx_remove = efx_farch_rx_remove,
+	.rx_write = efx_farch_rx_write,
+	.rx_defer_refill = efx_farch_rx_defer_refill,
+	.ev_probe = efx_farch_ev_probe,
+	.ev_init = efx_farch_ev_init,
+	.ev_fini = efx_farch_ev_fini,
+	.ev_remove = efx_farch_ev_remove,
+	.ev_process = efx_farch_ev_process,
+	.ev_read_ack = efx_farch_ev_read_ack,
+	.ev_test_generate = efx_farch_ev_test_generate,
 
 	.revision = EFX_REV_FALCON_B0,
 	/* Map everything up to and including the RSS indirection
diff --git a/drivers/net/ethernet/sfc/farch.c b/drivers/net/ethernet/sfc/farch.c
new file mode 100644
index 00000000000..c3d07c55656
--- /dev/null
+++ b/drivers/net/ethernet/sfc/farch.c
@@ -0,0 +1,1781 @@
+/****************************************************************************
+ * Driver for Solarflare Solarstorm network controllers and boards
+ * Copyright 2005-2006 Fen Systems Ltd.
+ * Copyright 2006-2011 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/bitops.h>
+#include <linux/delay.h>
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/module.h>
+#include <linux/seq_file.h>
+#include "net_driver.h"
+#include "bitfield.h"
+#include "efx.h"
+#include "nic.h"
+#include "farch_regs.h"
+#include "io.h"
+#include "workarounds.h"
+
+/* Falcon-architecture (SFC4000 and SFC9000-family) support */
+
+/**************************************************************************
+ *
+ * Configurable values
+ *
+ **************************************************************************
+ */
+
+/* This is set to 16 for a good reason.  In summary, if larger than
+ * 16, the descriptor cache holds more than a default socket
+ * buffer's worth of packets (for UDP we can only have at most one
+ * socket buffer's worth outstanding).  This combined with the fact
+ * that we only get 1 TX event per descriptor cache means the NIC
+ * goes idle.
+ */
+#define TX_DC_ENTRIES 16
+#define TX_DC_ENTRIES_ORDER 1
+
+#define RX_DC_ENTRIES 64
+#define RX_DC_ENTRIES_ORDER 3
+
+/* If EFX_MAX_INT_ERRORS internal errors occur within
+ * EFX_INT_ERROR_EXPIRE seconds, we consider the NIC broken and
+ * disable it.
+ */
+#define EFX_INT_ERROR_EXPIRE 3600
+#define EFX_MAX_INT_ERRORS 5
+
+/* Depth of RX flush request fifo */
+#define EFX_RX_FLUSH_COUNT 4
+
+/* Driver generated events */
+#define _EFX_CHANNEL_MAGIC_TEST		0x000101
+#define _EFX_CHANNEL_MAGIC_FILL		0x000102
+#define _EFX_CHANNEL_MAGIC_RX_DRAIN	0x000103
+#define _EFX_CHANNEL_MAGIC_TX_DRAIN	0x000104
+
+#define _EFX_CHANNEL_MAGIC(_code, _data)	((_code) << 8 | (_data))
+#define _EFX_CHANNEL_MAGIC_CODE(_magic)		((_magic) >> 8)
+
+#define EFX_CHANNEL_MAGIC_TEST(_channel)				\
+	_EFX_CHANNEL_MAGIC(_EFX_CHANNEL_MAGIC_TEST, (_channel)->channel)
+#define EFX_CHANNEL_MAGIC_FILL(_rx_queue)				\
+	_EFX_CHANNEL_MAGIC(_EFX_CHANNEL_MAGIC_FILL,			\
+			   efx_rx_queue_index(_rx_queue))
+#define EFX_CHANNEL_MAGIC_RX_DRAIN(_rx_queue)				\
+	_EFX_CHANNEL_MAGIC(_EFX_CHANNEL_MAGIC_RX_DRAIN,			\
+			   efx_rx_queue_index(_rx_queue))
+#define EFX_CHANNEL_MAGIC_TX_DRAIN(_tx_queue)				\
+	_EFX_CHANNEL_MAGIC(_EFX_CHANNEL_MAGIC_TX_DRAIN,			\
+			   (_tx_queue)->queue)
+
+static void efx_farch_magic_event(struct efx_channel *channel, u32 magic);
+
+/**************************************************************************
+ *
+ * Hardware access
+ *
+ **************************************************************************/
+
+static inline void efx_write_buf_tbl(struct efx_nic *efx, efx_qword_t *value,
+				     unsigned int index)
+{
+	efx_sram_writeq(efx, efx->membase + efx->type->buf_tbl_base,
+			value, index);
+}
+
+static bool efx_masked_compare_oword(const efx_oword_t *a, const efx_oword_t *b,
+				     const efx_oword_t *mask)
+{
+	return ((a->u64[0] ^ b->u64[0]) & mask->u64[0]) ||
+		((a->u64[1] ^ b->u64[1]) & mask->u64[1]);
+}
+
+int efx_farch_test_registers(struct efx_nic *efx,
+			     const struct efx_farch_register_test *regs,
+			     size_t n_regs)
+{
+	unsigned address = 0, i, j;
+	efx_oword_t mask, imask, original, reg, buf;
+
+	for (i = 0; i < n_regs; ++i) {
+		address = regs[i].address;
+		mask = imask = regs[i].mask;
+		EFX_INVERT_OWORD(imask);
+
+		efx_reado(efx, &original, address);
+
+		/* bit sweep on and off */
+		for (j = 0; j < 128; j++) {
+			if (!EFX_EXTRACT_OWORD32(mask, j, j))
+				continue;
+
+			/* Test this testable bit can be set in isolation */
+			EFX_AND_OWORD(reg, original, mask);
+			EFX_SET_OWORD32(reg, j, j, 1);
+
+			efx_writeo(efx, &reg, address);
+			efx_reado(efx, &buf, address);
+
+			if (efx_masked_compare_oword(&reg, &buf, &mask))
+				goto fail;
+
+			/* Test this testable bit can be cleared in isolation */
+			EFX_OR_OWORD(reg, original, mask);
+			EFX_SET_OWORD32(reg, j, j, 0);
+
+			efx_writeo(efx, &reg, address);
+			efx_reado(efx, &buf, address);
+
+			if (efx_masked_compare_oword(&reg, &buf, &mask))
+				goto fail;
+		}
+
+		efx_writeo(efx, &original, address);
+	}
+
+	return 0;
+
+fail:
+	netif_err(efx, hw, efx->net_dev,
+		  "wrote "EFX_OWORD_FMT" read "EFX_OWORD_FMT
+		  " at address 0x%x mask "EFX_OWORD_FMT"\n", EFX_OWORD_VAL(reg),
+		  EFX_OWORD_VAL(buf), address, EFX_OWORD_VAL(mask));
+	return -EIO;
+}
+
+/**************************************************************************
+ *
+ * Special buffer handling
+ * Special buffers are used for event queues and the TX and RX
+ * descriptor rings.
+ *
+ *************************************************************************/
+
+/*
+ * Initialise a special buffer
+ *
+ * This will define a buffer (previously allocated via
+ * efx_alloc_special_buffer()) in the buffer table, allowing
+ * it to be used for event queues, descriptor rings etc.
+ */
+static void
+efx_init_special_buffer(struct efx_nic *efx, struct efx_special_buffer *buffer)
+{
+	efx_qword_t buf_desc;
+	unsigned int index;
+	dma_addr_t dma_addr;
+	int i;
+
+	EFX_BUG_ON_PARANOID(!buffer->buf.addr);
+
+	/* Write buffer descriptors to NIC */
+	for (i = 0; i < buffer->entries; i++) {
+		index = buffer->index + i;
+		dma_addr = buffer->buf.dma_addr + (i * EFX_BUF_SIZE);
+		netif_dbg(efx, probe, efx->net_dev,
+			  "mapping special buffer %d at %llx\n",
+			  index, (unsigned long long)dma_addr);
+		EFX_POPULATE_QWORD_3(buf_desc,
+				     FRF_AZ_BUF_ADR_REGION, 0,
+				     FRF_AZ_BUF_ADR_FBUF, dma_addr >> 12,
+				     FRF_AZ_BUF_OWNER_ID_FBUF, 0);
+		efx_write_buf_tbl(efx, &buf_desc, index);
+	}
+}
+
+/* Unmaps a buffer and clears the buffer table entries */
+static void
+efx_fini_special_buffer(struct efx_nic *efx, struct efx_special_buffer *buffer)
+{
+	efx_oword_t buf_tbl_upd;
+	unsigned int start = buffer->index;
+	unsigned int end = (buffer->index + buffer->entries - 1);
+
+	if (!buffer->entries)
+		return;
+
+	netif_dbg(efx, hw, efx->net_dev, "unmapping special buffers %d-%d\n",
+		  buffer->index, buffer->index + buffer->entries - 1);
+
+	EFX_POPULATE_OWORD_4(buf_tbl_upd,
+			     FRF_AZ_BUF_UPD_CMD, 0,
+			     FRF_AZ_BUF_CLR_CMD, 1,
+			     FRF_AZ_BUF_CLR_END_ID, end,
+			     FRF_AZ_BUF_CLR_START_ID, start);
+	efx_writeo(efx, &buf_tbl_upd, FR_AZ_BUF_TBL_UPD);
+}
+
+/*
+ * Allocate a new special buffer
+ *
+ * This allocates memory for a new buffer, clears it and allocates a
+ * new buffer ID range.  It does not write into the buffer table.
+ *
+ * This call will allocate 4KB buffers, since 8KB buffers can't be
+ * used for event queues and descriptor rings.
+ */
+static int efx_alloc_special_buffer(struct efx_nic *efx,
+				    struct efx_special_buffer *buffer,
+				    unsigned int len)
+{
+	len = ALIGN(len, EFX_BUF_SIZE);
+
+	if (efx_nic_alloc_buffer(efx, &buffer->buf, len, GFP_KERNEL))
+		return -ENOMEM;
+	buffer->entries = len / EFX_BUF_SIZE;
+	BUG_ON(buffer->buf.dma_addr & (EFX_BUF_SIZE - 1));
+
+	/* Select new buffer ID */
+	buffer->index = efx->next_buffer_table;
+	efx->next_buffer_table += buffer->entries;
+#ifdef CONFIG_SFC_SRIOV
+	BUG_ON(efx_sriov_enabled(efx) &&
+	       efx->vf_buftbl_base < efx->next_buffer_table);
+#endif
+
+	netif_dbg(efx, probe, efx->net_dev,
+		  "allocating special buffers %d-%d at %llx+%x "
+		  "(virt %p phys %llx)\n", buffer->index,
+		  buffer->index + buffer->entries - 1,
+		  (u64)buffer->buf.dma_addr, len,
+		  buffer->buf.addr, (u64)virt_to_phys(buffer->buf.addr));
+
+	return 0;
+}
+
+static void
+efx_free_special_buffer(struct efx_nic *efx, struct efx_special_buffer *buffer)
+{
+	if (!buffer->buf.addr)
+		return;
+
+	netif_dbg(efx, hw, efx->net_dev,
+		  "deallocating special buffers %d-%d at %llx+%x "
+		  "(virt %p phys %llx)\n", buffer->index,
+		  buffer->index + buffer->entries - 1,
+		  (u64)buffer->buf.dma_addr, buffer->buf.len,
+		  buffer->buf.addr, (u64)virt_to_phys(buffer->buf.addr));
+
+	efx_nic_free_buffer(efx, &buffer->buf);
+	buffer->entries = 0;
+}
+
+/**************************************************************************
+ *
+ * TX path
+ *
+ **************************************************************************/
+
+/* This writes to the TX_DESC_WPTR; write pointer for TX descriptor ring */
+static inline void efx_farch_notify_tx_desc(struct efx_tx_queue *tx_queue)
+{
+	unsigned write_ptr;
+	efx_dword_t reg;
+
+	write_ptr = tx_queue->write_count & tx_queue->ptr_mask;
+	EFX_POPULATE_DWORD_1(reg, FRF_AZ_TX_DESC_WPTR_DWORD, write_ptr);
+	efx_writed_page(tx_queue->efx, &reg,
+			FR_AZ_TX_DESC_UPD_DWORD_P0, tx_queue->queue);
+}
+
+/* Write pointer and first descriptor for TX descriptor ring */
+static inline void efx_farch_push_tx_desc(struct efx_tx_queue *tx_queue,
+					  const efx_qword_t *txd)
+{
+	unsigned write_ptr;
+	efx_oword_t reg;
+
+	BUILD_BUG_ON(FRF_AZ_TX_DESC_LBN != 0);
+	BUILD_BUG_ON(FR_AA_TX_DESC_UPD_KER != FR_BZ_TX_DESC_UPD_P0);
+
+	write_ptr = tx_queue->write_count & tx_queue->ptr_mask;
+	EFX_POPULATE_OWORD_2(reg, FRF_AZ_TX_DESC_PUSH_CMD, true,
+			     FRF_AZ_TX_DESC_WPTR, write_ptr);
+	reg.qword[0] = *txd;
+	efx_writeo_page(tx_queue->efx, &reg,
+			FR_BZ_TX_DESC_UPD_P0, tx_queue->queue);
+}
+
+
+/* For each entry inserted into the software descriptor ring, create a
+ * descriptor in the hardware TX descriptor ring (in host memory), and
+ * write a doorbell.
+ */
+void efx_farch_tx_write(struct efx_tx_queue *tx_queue)
+{
+
+	struct efx_tx_buffer *buffer;
+	efx_qword_t *txd;
+	unsigned write_ptr;
+	unsigned old_write_count = tx_queue->write_count;
+
+	BUG_ON(tx_queue->write_count == tx_queue->insert_count);
+
+	do {
+		write_ptr = tx_queue->write_count & tx_queue->ptr_mask;
+		buffer = &tx_queue->buffer[write_ptr];
+		txd = efx_tx_desc(tx_queue, write_ptr);
+		++tx_queue->write_count;
+
+		/* Create TX descriptor ring entry */
+		BUILD_BUG_ON(EFX_TX_BUF_CONT != 1);
+		EFX_POPULATE_QWORD_4(*txd,
+				     FSF_AZ_TX_KER_CONT,
+				     buffer->flags & EFX_TX_BUF_CONT,
+				     FSF_AZ_TX_KER_BYTE_COUNT, buffer->len,
+				     FSF_AZ_TX_KER_BUF_REGION, 0,
+				     FSF_AZ_TX_KER_BUF_ADDR, buffer->dma_addr);
+	} while (tx_queue->write_count != tx_queue->insert_count);
+
+	wmb(); /* Ensure descriptors are written before they are fetched */
+
+	if (efx_nic_may_push_tx_desc(tx_queue, old_write_count)) {
+		txd = efx_tx_desc(tx_queue,
+				  old_write_count & tx_queue->ptr_mask);
+		efx_farch_push_tx_desc(tx_queue, txd);
+		++tx_queue->pushes;
+	} else {
+		efx_farch_notify_tx_desc(tx_queue);
+	}
+}
+
+/* Allocate hardware resources for a TX queue */
+int efx_farch_tx_probe(struct efx_tx_queue *tx_queue)
+{
+	struct efx_nic *efx = tx_queue->efx;
+	unsigned entries;
+
+	entries = tx_queue->ptr_mask + 1;
+	return efx_alloc_special_buffer(efx, &tx_queue->txd,
+					entries * sizeof(efx_qword_t));
+}
+
+void efx_farch_tx_init(struct efx_tx_queue *tx_queue)
+{
+	struct efx_nic *efx = tx_queue->efx;
+	efx_oword_t reg;
+
+	/* Pin TX descriptor ring */
+	efx_init_special_buffer(efx, &tx_queue->txd);
+
+	/* Push TX descriptor ring to card */
+	EFX_POPULATE_OWORD_10(reg,
+			      FRF_AZ_TX_DESCQ_EN, 1,
+			      FRF_AZ_TX_ISCSI_DDIG_EN, 0,
+			      FRF_AZ_TX_ISCSI_HDIG_EN, 0,
+			      FRF_AZ_TX_DESCQ_BUF_BASE_ID, tx_queue->txd.index,
+			      FRF_AZ_TX_DESCQ_EVQ_ID,
+			      tx_queue->channel->channel,
+			      FRF_AZ_TX_DESCQ_OWNER_ID, 0,
+			      FRF_AZ_TX_DESCQ_LABEL, tx_queue->queue,
+			      FRF_AZ_TX_DESCQ_SIZE,
+			      __ffs(tx_queue->txd.entries),
+			      FRF_AZ_TX_DESCQ_TYPE, 0,
+			      FRF_BZ_TX_NON_IP_DROP_DIS, 1);
+
+	if (efx_nic_rev(efx) >= EFX_REV_FALCON_B0) {
+		int csum = tx_queue->queue & EFX_TXQ_TYPE_OFFLOAD;
+		EFX_SET_OWORD_FIELD(reg, FRF_BZ_TX_IP_CHKSM_DIS, !csum);
+		EFX_SET_OWORD_FIELD(reg, FRF_BZ_TX_TCP_CHKSM_DIS,
+				    !csum);
+	}
+
+	efx_writeo_table(efx, &reg, efx->type->txd_ptr_tbl_base,
+			 tx_queue->queue);
+
+	if (efx_nic_rev(efx) < EFX_REV_FALCON_B0) {
+		/* Only 128 bits in this register */
+		BUILD_BUG_ON(EFX_MAX_TX_QUEUES > 128);
+
+		efx_reado(efx, &reg, FR_AA_TX_CHKSM_CFG);
+		if (tx_queue->queue & EFX_TXQ_TYPE_OFFLOAD)
+			__clear_bit_le(tx_queue->queue, &reg);
+		else
+			__set_bit_le(tx_queue->queue, &reg);
+		efx_writeo(efx, &reg, FR_AA_TX_CHKSM_CFG);
+	}
+
+	if (efx_nic_rev(efx) >= EFX_REV_FALCON_B0) {
+		EFX_POPULATE_OWORD_1(reg,
+				     FRF_BZ_TX_PACE,
+				     (tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI) ?
+				     FFE_BZ_TX_PACE_OFF :
+				     FFE_BZ_TX_PACE_RESERVED);
+		efx_writeo_table(efx, &reg, FR_BZ_TX_PACE_TBL,
+				 tx_queue->queue);
+	}
+}
+
+static void efx_farch_flush_tx_queue(struct efx_tx_queue *tx_queue)
+{
+	struct efx_nic *efx = tx_queue->efx;
+	efx_oword_t tx_flush_descq;
+
+	WARN_ON(atomic_read(&tx_queue->flush_outstanding));
+	atomic_set(&tx_queue->flush_outstanding, 1);
+
+	EFX_POPULATE_OWORD_2(tx_flush_descq,
+			     FRF_AZ_TX_FLUSH_DESCQ_CMD, 1,
+			     FRF_AZ_TX_FLUSH_DESCQ, tx_queue->queue);
+	efx_writeo(efx, &tx_flush_descq, FR_AZ_TX_FLUSH_DESCQ);
+}
+
+void efx_farch_tx_fini(struct efx_tx_queue *tx_queue)
+{
+	struct efx_nic *efx = tx_queue->efx;
+	efx_oword_t tx_desc_ptr;
+
+	/* Remove TX descriptor ring from card */
+	EFX_ZERO_OWORD(tx_desc_ptr);
+	efx_writeo_table(efx, &tx_desc_ptr, efx->type->txd_ptr_tbl_base,
+			 tx_queue->queue);
+
+	/* Unpin TX descriptor ring */
+	efx_fini_special_buffer(efx, &tx_queue->txd);
+}
+
+/* Free buffers backing TX queue */
+void efx_farch_tx_remove(struct efx_tx_queue *tx_queue)
+{
+	efx_free_special_buffer(tx_queue->efx, &tx_queue->txd);
+}
+
+/**************************************************************************
+ *
+ * RX path
+ *
+ **************************************************************************/
+
+/* This creates an entry in the RX descriptor queue */
+static inline void
+efx_farch_build_rx_desc(struct efx_rx_queue *rx_queue, unsigned index)
+{
+	struct efx_rx_buffer *rx_buf;
+	efx_qword_t *rxd;
+
+	rxd = efx_rx_desc(rx_queue, index);
+	rx_buf = efx_rx_buffer(rx_queue, index);
+	EFX_POPULATE_QWORD_3(*rxd,
+			     FSF_AZ_RX_KER_BUF_SIZE,
+			     rx_buf->len -
+			     rx_queue->efx->type->rx_buffer_padding,
+			     FSF_AZ_RX_KER_BUF_REGION, 0,
+			     FSF_AZ_RX_KER_BUF_ADDR, rx_buf->dma_addr);
+}
+
+/* This writes to the RX_DESC_WPTR register for the specified receive
+ * descriptor ring.
+ */
+void efx_farch_rx_write(struct efx_rx_queue *rx_queue)
+{
+	struct efx_nic *efx = rx_queue->efx;
+	efx_dword_t reg;
+	unsigned write_ptr;
+
+	while (rx_queue->notified_count != rx_queue->added_count) {
+		efx_farch_build_rx_desc(
+			rx_queue,
+			rx_queue->notified_count & rx_queue->ptr_mask);
+		++rx_queue->notified_count;
+	}
+
+	wmb();
+	write_ptr = rx_queue->added_count & rx_queue->ptr_mask;
+	EFX_POPULATE_DWORD_1(reg, FRF_AZ_RX_DESC_WPTR_DWORD, write_ptr);
+	efx_writed_page(efx, &reg, FR_AZ_RX_DESC_UPD_DWORD_P0,
+			efx_rx_queue_index(rx_queue));
+}
+
+int efx_farch_rx_probe(struct efx_rx_queue *rx_queue)
+{
+	struct efx_nic *efx = rx_queue->efx;
+	unsigned entries;
+
+	entries = rx_queue->ptr_mask + 1;
+	return efx_alloc_special_buffer(efx, &rx_queue->rxd,
+					entries * sizeof(efx_qword_t));
+}
+
+void efx_farch_rx_init(struct efx_rx_queue *rx_queue)
+{
+	efx_oword_t rx_desc_ptr;
+	struct efx_nic *efx = rx_queue->efx;
+	bool is_b0 = efx_nic_rev(efx) >= EFX_REV_FALCON_B0;
+	bool iscsi_digest_en = is_b0;
+	bool jumbo_en;
+
+	/* For kernel-mode queues in Falcon A1, the JUMBO flag enables
+	 * DMA to continue after a PCIe page boundary (and scattering
+	 * is not possible).  In Falcon B0 and Siena, it enables
+	 * scatter.
+	 */
+	jumbo_en = !is_b0 || efx->rx_scatter;
+
+	netif_dbg(efx, hw, efx->net_dev,
+		  "RX queue %d ring in special buffers %d-%d\n",
+		  efx_rx_queue_index(rx_queue), rx_queue->rxd.index,
+		  rx_queue->rxd.index + rx_queue->rxd.entries - 1);
+
+	rx_queue->scatter_n = 0;
+
+	/* Pin RX descriptor ring */
+	efx_init_special_buffer(efx, &rx_queue->rxd);
+
+	/* Push RX descriptor ring to card */
+	EFX_POPULATE_OWORD_10(rx_desc_ptr,
+			      FRF_AZ_RX_ISCSI_DDIG_EN, iscsi_digest_en,
+			      FRF_AZ_RX_ISCSI_HDIG_EN, iscsi_digest_en,
+			      FRF_AZ_RX_DESCQ_BUF_BASE_ID, rx_queue->rxd.index,
+			      FRF_AZ_RX_DESCQ_EVQ_ID,
+			      efx_rx_queue_channel(rx_queue)->channel,
+			      FRF_AZ_RX_DESCQ_OWNER_ID, 0,
+			      FRF_AZ_RX_DESCQ_LABEL,
+			      efx_rx_queue_index(rx_queue),
+			      FRF_AZ_RX_DESCQ_SIZE,
+			      __ffs(rx_queue->rxd.entries),
+			      FRF_AZ_RX_DESCQ_TYPE, 0 /* kernel queue */ ,
+			      FRF_AZ_RX_DESCQ_JUMBO, jumbo_en,
+			      FRF_AZ_RX_DESCQ_EN, 1);
+	efx_writeo_table(efx, &rx_desc_ptr, efx->type->rxd_ptr_tbl_base,
+			 efx_rx_queue_index(rx_queue));
+}
+
+static void efx_farch_flush_rx_queue(struct efx_rx_queue *rx_queue)
+{
+	struct efx_nic *efx = rx_queue->efx;
+	efx_oword_t rx_flush_descq;
+
+	EFX_POPULATE_OWORD_2(rx_flush_descq,
+			     FRF_AZ_RX_FLUSH_DESCQ_CMD, 1,
+			     FRF_AZ_RX_FLUSH_DESCQ,
+			     efx_rx_queue_index(rx_queue));
+	efx_writeo(efx, &rx_flush_descq, FR_AZ_RX_FLUSH_DESCQ);
+}
+
+void efx_farch_rx_fini(struct efx_rx_queue *rx_queue)
+{
+	efx_oword_t rx_desc_ptr;
+	struct efx_nic *efx = rx_queue->efx;
+
+	/* Remove RX descriptor ring from card */
+	EFX_ZERO_OWORD(rx_desc_ptr);
+	efx_writeo_table(efx, &rx_desc_ptr, efx->type->rxd_ptr_tbl_base,
+			 efx_rx_queue_index(rx_queue));
+
+	/* Unpin RX descriptor ring */
+	efx_fini_special_buffer(efx, &rx_queue->rxd);
+}
+
+/* Free buffers backing RX queue */
+void efx_farch_rx_remove(struct efx_rx_queue *rx_queue)
+{
+	efx_free_special_buffer(rx_queue->efx, &rx_queue->rxd);
+}
+
+/**************************************************************************
+ *
+ * Flush handling
+ *
+ **************************************************************************/
+
+/* efx_farch_flush_queues() must be woken up when all flushes are completed,
+ * or more RX flushes can be kicked off.
+ */
+static bool efx_farch_flush_wake(struct efx_nic *efx)
+{
+	/* Ensure that all updates are visible to efx_farch_flush_queues() */
+	smp_mb();
+
+	return (atomic_read(&efx->drain_pending) == 0 ||
+		(atomic_read(&efx->rxq_flush_outstanding) < EFX_RX_FLUSH_COUNT
+		 && atomic_read(&efx->rxq_flush_pending) > 0));
+}
+
+static bool efx_check_tx_flush_complete(struct efx_nic *efx)
+{
+	bool i = true;
+	efx_oword_t txd_ptr_tbl;
+	struct efx_channel *channel;
+	struct efx_tx_queue *tx_queue;
+
+	efx_for_each_channel(channel, efx) {
+		efx_for_each_channel_tx_queue(tx_queue, channel) {
+			efx_reado_table(efx, &txd_ptr_tbl,
+					FR_BZ_TX_DESC_PTR_TBL, tx_queue->queue);
+			if (EFX_OWORD_FIELD(txd_ptr_tbl,
+					    FRF_AZ_TX_DESCQ_FLUSH) ||
+			    EFX_OWORD_FIELD(txd_ptr_tbl,
+					    FRF_AZ_TX_DESCQ_EN)) {
+				netif_dbg(efx, hw, efx->net_dev,
+					  "flush did not complete on TXQ %d\n",
+					  tx_queue->queue);
+				i = false;
+			} else if (atomic_cmpxchg(&tx_queue->flush_outstanding,
+						  1, 0)) {
+				/* The flush is complete, but we didn't
+				 * receive a flush completion event
+				 */
+				netif_dbg(efx, hw, efx->net_dev,
+					  "flush complete on TXQ %d, so drain "
+					  "the queue\n", tx_queue->queue);
+				/* Don't need to increment drain_pending as it
+				 * has already been incremented for the queues
+				 * which did not drain
+				 */
+				efx_farch_magic_event(channel,
+						      EFX_CHANNEL_MAGIC_TX_DRAIN(
+							      tx_queue));
+			}
+		}
+	}
+
+	return i;
+}
+
+/* Flush all the transmit queues, and continue flushing receive queues until
+ * they're all flushed. Wait for the DRAIN events to be recieved so that there
+ * are no more RX and TX events left on any channel. */
+static int efx_farch_do_flush(struct efx_nic *efx)
+{
+	unsigned timeout = msecs_to_jiffies(5000); /* 5s for all flushes and drains */
+	struct efx_channel *channel;
+	struct efx_rx_queue *rx_queue;
+	struct efx_tx_queue *tx_queue;
+	int rc = 0;
+
+	efx_for_each_channel(channel, efx) {
+		efx_for_each_channel_tx_queue(tx_queue, channel) {
+			atomic_inc(&efx->drain_pending);
+			efx_farch_flush_tx_queue(tx_queue);
+		}
+		efx_for_each_channel_rx_queue(rx_queue, channel) {
+			atomic_inc(&efx->drain_pending);
+			rx_queue->flush_pending = true;
+			atomic_inc(&efx->rxq_flush_pending);
+		}
+	}
+
+	while (timeout && atomic_read(&efx->drain_pending) > 0) {
+		/* If SRIOV is enabled, then offload receive queue flushing to
+		 * the firmware (though we will still have to poll for
+		 * completion). If that fails, fall back to the old scheme.
+		 */
+		if (efx_sriov_enabled(efx)) {
+			rc = efx_mcdi_flush_rxqs(efx);
+			if (!rc)
+				goto wait;
+		}
+
+		/* The hardware supports four concurrent rx flushes, each of
+		 * which may need to be retried if there is an outstanding
+		 * descriptor fetch
+		 */
+		efx_for_each_channel(channel, efx) {
+			efx_for_each_channel_rx_queue(rx_queue, channel) {
+				if (atomic_read(&efx->rxq_flush_outstanding) >=
+				    EFX_RX_FLUSH_COUNT)
+					break;
+
+				if (rx_queue->flush_pending) {
+					rx_queue->flush_pending = false;
+					atomic_dec(&efx->rxq_flush_pending);
+					atomic_inc(&efx->rxq_flush_outstanding);
+					efx_farch_flush_rx_queue(rx_queue);
+				}
+			}
+		}
+
+	wait:
+		timeout = wait_event_timeout(efx->flush_wq,
+					     efx_farch_flush_wake(efx),
+					     timeout);
+	}
+
+	if (atomic_read(&efx->drain_pending) &&
+	    !efx_check_tx_flush_complete(efx)) {
+		netif_err(efx, hw, efx->net_dev, "failed to flush %d queues "
+			  "(rx %d+%d)\n", atomic_read(&efx->drain_pending),
+			  atomic_read(&efx->rxq_flush_outstanding),
+			  atomic_read(&efx->rxq_flush_pending));
+		rc = -ETIMEDOUT;
+
+		atomic_set(&efx->drain_pending, 0);
+		atomic_set(&efx->rxq_flush_pending, 0);
+		atomic_set(&efx->rxq_flush_outstanding, 0);
+	}
+
+	return rc;
+}
+
+int efx_farch_fini_dmaq(struct efx_nic *efx)
+{
+	struct efx_channel *channel;
+	struct efx_tx_queue *tx_queue;
+	struct efx_rx_queue *rx_queue;
+	int rc = 0;
+
+	/* Do not attempt to write to the NIC during EEH recovery */
+	if (efx->state != STATE_RECOVERY) {
+		/* Only perform flush if DMA is enabled */
+		if (efx->pci_dev->is_busmaster) {
+			efx->type->prepare_flush(efx);
+			rc = efx_farch_do_flush(efx);
+			efx->type->finish_flush(efx);
+		}
+
+		efx_for_each_channel(channel, efx) {
+			efx_for_each_channel_rx_queue(rx_queue, channel)
+				efx_farch_rx_fini(rx_queue);
+			efx_for_each_channel_tx_queue(tx_queue, channel)
+				efx_farch_tx_fini(tx_queue);
+		}
+	}
+
+	return rc;
+}
+
+/**************************************************************************
+ *
+ * Event queue processing
+ * Event queues are processed by per-channel tasklets.
+ *
+ **************************************************************************/
+
+/* Update a channel's event queue's read pointer (RPTR) register
+ *
+ * This writes the EVQ_RPTR_REG register for the specified channel's
+ * event queue.
+ */
+void efx_farch_ev_read_ack(struct efx_channel *channel)
+{
+	efx_dword_t reg;
+	struct efx_nic *efx = channel->efx;
+
+	EFX_POPULATE_DWORD_1(reg, FRF_AZ_EVQ_RPTR,
+			     channel->eventq_read_ptr & channel->eventq_mask);
+
+	/* For Falcon A1, EVQ_RPTR_KER is documented as having a step size
+	 * of 4 bytes, but it is really 16 bytes just like later revisions.
+	 */
+	efx_writed(efx, &reg,
+		   efx->type->evq_rptr_tbl_base +
+		   FR_BZ_EVQ_RPTR_STEP * channel->channel);
+}
+
+/* Use HW to insert a SW defined event */
+void efx_farch_generate_event(struct efx_nic *efx, unsigned int evq,
+			      efx_qword_t *event)
+{
+	efx_oword_t drv_ev_reg;
+
+	BUILD_BUG_ON(FRF_AZ_DRV_EV_DATA_LBN != 0 ||
+		     FRF_AZ_DRV_EV_DATA_WIDTH != 64);
+	drv_ev_reg.u32[0] = event->u32[0];
+	drv_ev_reg.u32[1] = event->u32[1];
+	drv_ev_reg.u32[2] = 0;
+	drv_ev_reg.u32[3] = 0;
+	EFX_SET_OWORD_FIELD(drv_ev_reg, FRF_AZ_DRV_EV_QID, evq);
+	efx_writeo(efx, &drv_ev_reg, FR_AZ_DRV_EV);
+}
+
+static void efx_farch_magic_event(struct efx_channel *channel, u32 magic)
+{
+	efx_qword_t event;
+
+	EFX_POPULATE_QWORD_2(event, FSF_AZ_EV_CODE,
+			     FSE_AZ_EV_CODE_DRV_GEN_EV,
+			     FSF_AZ_DRV_GEN_EV_MAGIC, magic);
+	efx_farch_generate_event(channel->efx, channel->channel, &event);
+}
+
+/* Handle a transmit completion event
+ *
+ * The NIC batches TX completion events; the message we receive is of
+ * the form "complete all TX events up to this index".
+ */
+static int
+efx_farch_handle_tx_event(struct efx_channel *channel, efx_qword_t *event)
+{
+	unsigned int tx_ev_desc_ptr;
+	unsigned int tx_ev_q_label;
+	struct efx_tx_queue *tx_queue;
+	struct efx_nic *efx = channel->efx;
+	int tx_packets = 0;
+
+	if (unlikely(ACCESS_ONCE(efx->reset_pending)))
+		return 0;
+
+	if (likely(EFX_QWORD_FIELD(*event, FSF_AZ_TX_EV_COMP))) {
+		/* Transmit completion */
+		tx_ev_desc_ptr = EFX_QWORD_FIELD(*event, FSF_AZ_TX_EV_DESC_PTR);
+		tx_ev_q_label = EFX_QWORD_FIELD(*event, FSF_AZ_TX_EV_Q_LABEL);
+		tx_queue = efx_channel_get_tx_queue(
+			channel, tx_ev_q_label % EFX_TXQ_TYPES);
+		tx_packets = ((tx_ev_desc_ptr - tx_queue->read_count) &
+			      tx_queue->ptr_mask);
+		efx_xmit_done(tx_queue, tx_ev_desc_ptr);
+	} else if (EFX_QWORD_FIELD(*event, FSF_AZ_TX_EV_WQ_FF_FULL)) {
+		/* Rewrite the FIFO write pointer */
+		tx_ev_q_label = EFX_QWORD_FIELD(*event, FSF_AZ_TX_EV_Q_LABEL);
+		tx_queue = efx_channel_get_tx_queue(
+			channel, tx_ev_q_label % EFX_TXQ_TYPES);
+
+		netif_tx_lock(efx->net_dev);
+		efx_farch_notify_tx_desc(tx_queue);
+		netif_tx_unlock(efx->net_dev);
+	} else if (EFX_QWORD_FIELD(*event, FSF_AZ_TX_EV_PKT_ERR) &&
+		   EFX_WORKAROUND_10727(efx)) {
+		efx_schedule_reset(efx, RESET_TYPE_TX_DESC_FETCH);
+	} else {
+		netif_err(efx, tx_err, efx->net_dev,
+			  "channel %d unexpected TX event "
+			  EFX_QWORD_FMT"\n", channel->channel,
+			  EFX_QWORD_VAL(*event));
+	}
+
+	return tx_packets;
+}
+
+/* Detect errors included in the rx_evt_pkt_ok bit. */
+static u16 efx_farch_handle_rx_not_ok(struct efx_rx_queue *rx_queue,
+				      const efx_qword_t *event)
+{
+	struct efx_channel *channel = efx_rx_queue_channel(rx_queue);
+	struct efx_nic *efx = rx_queue->efx;
author	Ben Hutchings <bhutchings@solarflare.com>	2013-08-21 19:51:04 +0100
committer	Ben Hutchings <bhutchings@solarflare.com>	2013-08-21 20:19:05 +0100
commit	86094f7f38ff711f3db8497fcb4d2e109100f497 (patch)
tree	c9c8d9418b32d0a2abf6220a22328ceeb9130244
parent	e42c3d85af629697699c89aecba481527a1da898 (diff)