falcon.c

/****************************************************************************
 * 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/bitops.h>
#include <linux/delay.h>
#include <linux/pci.h>
#include <linux/module.h>
#include <linux/seq_file.h>
#include <linux/i2c.h>
#include <linux/mii.h>
#include "net_driver.h"
#include "bitfield.h"
#include "efx.h"
#include "mac.h"
#include "spi.h"
#include "falcon.h"
#include "regs.h"
#include "io.h"
#include "mdio_10g.h"
#include "phy.h"
#include "workarounds.h"

/* Hardware control for SFC4000 (aka Falcon). */

/**************************************************************************
 *
 * 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 TX_DC_BASE 0x130000

#define RX_DC_ENTRIES 64
#define RX_DC_ENTRIES_ORDER 3
#define RX_DC_BASE 0x100000

static const unsigned int
/* "Large" EEPROM device: Atmel AT25640 or similar
 * 8 KB, 16-bit address, 32 B write block */
large_eeprom_type = ((13 << SPI_DEV_TYPE_SIZE_LBN)
		     | (2 << SPI_DEV_TYPE_ADDR_LEN_LBN)
		     | (5 << SPI_DEV_TYPE_BLOCK_SIZE_LBN)),
/* Default flash device: Atmel AT25F1024
 * 128 KB, 24-bit address, 32 KB erase block, 256 B write block */
default_flash_type = ((17 << SPI_DEV_TYPE_SIZE_LBN)
		      | (3 << SPI_DEV_TYPE_ADDR_LEN_LBN)
		      | (0x52 << SPI_DEV_TYPE_ERASE_CMD_LBN)
		      | (15 << SPI_DEV_TYPE_ERASE_SIZE_LBN)
		      | (8 << SPI_DEV_TYPE_BLOCK_SIZE_LBN));

/* RX FIFO XOFF watermark
 *
 * When the amount of the RX FIFO increases used increases past this
 * watermark send XOFF. Only used if RX flow control is enabled (ethtool -A)
 * This also has an effect on RX/TX arbitration
 */
static int rx_xoff_thresh_bytes = -1;
module_param(rx_xoff_thresh_bytes, int, 0644);
MODULE_PARM_DESC(rx_xoff_thresh_bytes, "RX fifo XOFF threshold");

/* RX FIFO XON watermark
 *
 * When the amount of the RX FIFO used decreases below this
 * watermark send XON. Only used if TX flow control is enabled (ethtool -A)
 * This also has an effect on RX/TX arbitration
 */
static int rx_xon_thresh_bytes = -1;
module_param(rx_xon_thresh_bytes, int, 0644);
MODULE_PARM_DESC(rx_xon_thresh_bytes, "RX fifo XON threshold");

/* If FALCON_MAX_INT_ERRORS internal errors occur within
 * FALCON_INT_ERROR_EXPIRE seconds, we consider the NIC broken and
 * disable it.
 */
#define FALCON_INT_ERROR_EXPIRE 3600
#define FALCON_MAX_INT_ERRORS 5

/* We poll for events every FLUSH_INTERVAL ms, and check FLUSH_POLL_COUNT times
 */
#define FALCON_FLUSH_INTERVAL 10
#define FALCON_FLUSH_POLL_COUNT 100

/**************************************************************************
 *
 * Falcon constants
 *
 **************************************************************************
 */

/* Size and alignment of special buffers (4KB) */
#define FALCON_BUF_SIZE 4096

/* Depth of RX flush request fifo */
#define FALCON_RX_FLUSH_COUNT 4

#define FALCON_IS_DUAL_FUNC(efx)		\
	(falcon_rev(efx) < FALCON_REV_B0)

/**************************************************************************
 *
 * Falcon hardware access
 *
 **************************************************************************/

static inline void falcon_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);
}

/* Read the current event from the event queue */
static inline efx_qword_t *falcon_event(struct efx_channel *channel,
					unsigned int index)
{
	return (((efx_qword_t *) (channel->eventq.addr)) + index);
}

/* See if an event is present
 *
 * We check both the high and low dword of the event for all ones.  We
 * wrote all ones when we cleared the event, and no valid event can
 * have all ones in either its high or low dwords.  This approach is
 * robust against reordering.
 *
 * Note that using a single 64-bit comparison is incorrect; even
 * though the CPU read will be atomic, the DMA write may not be.
 */
static inline int falcon_event_present(efx_qword_t *event)
{
	return (!(EFX_DWORD_IS_ALL_ONES(event->dword[0]) |
		  EFX_DWORD_IS_ALL_ONES(event->dword[1])));
}

/**************************************************************************
 *
 * I2C bus - this is a bit-bashing interface using GPIO pins
 * Note that it uses the output enables to tristate the outputs
 * SDA is the data pin and SCL is the clock
 *
 **************************************************************************
 */
static void falcon_setsda(void *data, int state)
{
	struct efx_nic *efx = (struct efx_nic *)data;
	efx_oword_t reg;

	efx_reado(efx, &reg, FR_AB_GPIO_CTL);
	EFX_SET_OWORD_FIELD(reg, FRF_AB_GPIO3_OEN, !state);
	efx_writeo(efx, &reg, FR_AB_GPIO_CTL);
}

static void falcon_setscl(void *data, int state)
{
	struct efx_nic *efx = (struct efx_nic *)data;
	efx_oword_t reg;

	efx_reado(efx, &reg, FR_AB_GPIO_CTL);
	EFX_SET_OWORD_FIELD(reg, FRF_AB_GPIO0_OEN, !state);
	efx_writeo(efx, &reg, FR_AB_GPIO_CTL);
}

static int falcon_getsda(void *data)
{
	struct efx_nic *efx = (struct efx_nic *)data;
	efx_oword_t reg;

	efx_reado(efx, &reg, FR_AB_GPIO_CTL);
	return EFX_OWORD_FIELD(reg, FRF_AB_GPIO3_IN);
}

static int falcon_getscl(void *data)
{
	struct efx_nic *efx = (struct efx_nic *)data;
	efx_oword_t reg;

	efx_reado(efx, &reg, FR_AB_GPIO_CTL);
	return EFX_OWORD_FIELD(reg, FRF_AB_GPIO0_IN);
}

static struct i2c_algo_bit_data falcon_i2c_bit_operations = {
	.setsda		= falcon_setsda,
	.setscl		= falcon_setscl,
	.getsda		= falcon_getsda,
	.getscl		= falcon_getscl,
	.udelay		= 5,
	/* Wait up to 50 ms for slave to let us pull SCL high */
	.timeout	= DIV_ROUND_UP(HZ, 20),
};

/**************************************************************************
 *
 * Falcon special buffer handling
 * Special buffers are used for event queues and the TX and RX
 * descriptor rings.
 *
 *************************************************************************/

/*
 * Initialise a Falcon special buffer
 *
 * This will define a buffer (previously allocated via
 * falcon_alloc_special_buffer()) in Falcon's buffer table, allowing
 * it to be used for event queues, descriptor rings etc.
 */
static void
falcon_init_special_buffer(struct efx_nic *efx,
			   struct efx_special_buffer *buffer)
{
	efx_qword_t buf_desc;
	int index;
	dma_addr_t dma_addr;
	int i;

	EFX_BUG_ON_PARANOID(!buffer->addr);

	/* Write buffer descriptors to NIC */
	for (i = 0; i < buffer->entries; i++) {
		index = buffer->index + i;
		dma_addr = buffer->dma_addr + (i * 4096);
		EFX_LOG(efx, "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);
		falcon_write_buf_tbl(efx, &buf_desc, index);
	}
}

/* Unmaps a buffer from Falcon and clears the buffer table entries */
static void
falcon_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;

	EFX_LOG(efx, "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 Falcon special buffer
 *
 * This allocates memory for a new buffer, clears it and allocates a
 * new buffer ID range.  It does not write into Falcon's buffer table.
 *
 * This call will allocate 4KB buffers, since Falcon can't use 8KB
 * buffers for event queues and descriptor rings.
 */
static int falcon_alloc_special_buffer(struct efx_nic *efx,
				       struct efx_special_buffer *buffer,
				       unsigned int len)
{
	len = ALIGN(len, FALCON_BUF_SIZE);

	buffer->addr = pci_alloc_consistent(efx->pci_dev, len,
					    &buffer->dma_addr);
	if (!buffer->addr)
		return -ENOMEM;
	buffer->len = len;
	buffer->entries = len / FALCON_BUF_SIZE;
	BUG_ON(buffer->dma_addr & (FALCON_BUF_SIZE - 1));

	/* All zeros is a potentially valid event so memset to 0xff */
	memset(buffer->addr, 0xff, len);

	/* Select new buffer ID */
	buffer->index = efx->next_buffer_table;
	efx->next_buffer_table += buffer->entries;

	EFX_LOG(efx, "allocating special buffers %d-%d at %llx+%x "
		"(virt %p phys %llx)\n", buffer->index,
		buffer->index + buffer->entries - 1,
		(u64)buffer->dma_addr, len,
		buffer->addr, (u64)virt_to_phys(buffer->addr));

	return 0;
}

static void falcon_free_special_buffer(struct efx_nic *efx,
				       struct efx_special_buffer *buffer)
{
	if (!buffer->addr)
		return;

	EFX_LOG(efx, "deallocating special buffers %d-%d at %llx+%x "
		"(virt %p phys %llx)\n", buffer->index,
		buffer->index + buffer->entries - 1,
		(u64)buffer->dma_addr, buffer->len,
		buffer->addr, (u64)virt_to_phys(buffer->addr));

	pci_free_consistent(efx->pci_dev, buffer->len, buffer->addr,
			    buffer->dma_addr);
	buffer->addr = NULL;
	buffer->entries = 0;
}

/**************************************************************************
 *
 * Falcon generic buffer handling
 * These buffers are used for interrupt status and MAC stats
 *
 **************************************************************************/

static int falcon_alloc_buffer(struct efx_nic *efx,
			       struct efx_buffer *buffer, unsigned int len)
{
	buffer->addr = pci_alloc_consistent(efx->pci_dev, len,
					    &buffer->dma_addr);
	if (!buffer->addr)
		return -ENOMEM;
	buffer->len = len;
	memset(buffer->addr, 0, len);
	return 0;
}

static void falcon_free_buffer(struct efx_nic *efx, struct efx_buffer *buffer)
{
	if (buffer->addr) {
		pci_free_consistent(efx->pci_dev, buffer->len,
				    buffer->addr, buffer->dma_addr);
		buffer->addr = NULL;
	}
}

/**************************************************************************
 *
 * Falcon TX path
 *
 **************************************************************************/

/* Returns a pointer to the specified transmit descriptor in the TX
 * descriptor queue belonging to the specified channel.
 */
static inline efx_qword_t *falcon_tx_desc(struct efx_tx_queue *tx_queue,
					       unsigned int index)
{
	return (((efx_qword_t *) (tx_queue->txd.addr)) + index);
}

/* This writes to the TX_DESC_WPTR; write pointer for TX descriptor ring */
static inline void falcon_notify_tx_desc(struct efx_tx_queue *tx_queue)
{
	unsigned write_ptr;
	efx_dword_t reg;

	write_ptr = tx_queue->write_count & EFX_TXQ_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);
}


/* 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 falcon_push_buffers(struct efx_tx_queue *tx_queue)
{

	struct efx_tx_buffer *buffer;
	efx_qword_t *txd;
	unsigned write_ptr;

	BUG_ON(tx_queue->write_count == tx_queue->insert_count);

	do {
		write_ptr = tx_queue->write_count & EFX_TXQ_MASK;
		buffer = &tx_queue->buffer[write_ptr];
		txd = falcon_tx_desc(tx_queue, write_ptr);
		++tx_queue->write_count;

		/* Create TX descriptor ring entry */
		EFX_POPULATE_QWORD_4(*txd,
				     FSF_AZ_TX_KER_CONT, buffer->continuation,
				     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 */
	falcon_notify_tx_desc(tx_queue);
}

/* Allocate hardware resources for a TX queue */
int falcon_probe_tx(struct efx_tx_queue *tx_queue)
{
	struct efx_nic *efx = tx_queue->efx;
	BUILD_BUG_ON(EFX_TXQ_SIZE < 512 || EFX_TXQ_SIZE > 4096 ||
		     EFX_TXQ_SIZE & EFX_TXQ_MASK);
	return falcon_alloc_special_buffer(efx, &tx_queue->txd,
					   EFX_TXQ_SIZE * sizeof(efx_qword_t));
}

void falcon_init_tx(struct efx_tx_queue *tx_queue)
{
	efx_oword_t tx_desc_ptr;
	struct efx_nic *efx = tx_queue->efx;

	tx_queue->flushed = FLUSH_NONE;

	/* Pin TX descriptor ring */
	falcon_init_special_buffer(efx, &tx_queue->txd);

	/* Push TX descriptor ring to card */
	EFX_POPULATE_OWORD_10(tx_desc_ptr,
			      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 (falcon_rev(efx) >= FALCON_REV_B0) {
		int csum = tx_queue->queue == EFX_TX_QUEUE_OFFLOAD_CSUM;
		EFX_SET_OWORD_FIELD(tx_desc_ptr, FRF_BZ_TX_IP_CHKSM_DIS, !csum);
		EFX_SET_OWORD_FIELD(tx_desc_ptr, FRF_BZ_TX_TCP_CHKSM_DIS,
				    !csum);
	}

	efx_writeo_table(efx, &tx_desc_ptr, efx->type->txd_ptr_tbl_base,
			 tx_queue->queue);

	if (falcon_rev(efx) < FALCON_REV_B0) {
		efx_oword_t reg;

		/* Only 128 bits in this register */
		BUILD_BUG_ON(EFX_TX_QUEUE_COUNT >= 128);

		efx_reado(efx, &reg, FR_AA_TX_CHKSM_CFG);
		if (tx_queue->queue == EFX_TX_QUEUE_OFFLOAD_CSUM)
			clear_bit_le(tx_queue->queue, (void *)&reg);
		else
			set_bit_le(tx_queue->queue, (void *)&reg);
		efx_writeo(efx, &reg, FR_AA_TX_CHKSM_CFG);
	}
}

static void falcon_flush_tx_queue(struct efx_tx_queue *tx_queue)
{
	struct efx_nic *efx = tx_queue->efx;
	efx_oword_t tx_flush_descq;

	tx_queue->flushed = FLUSH_PENDING;

	/* Post a flush command */
	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 falcon_fini_tx(struct efx_tx_queue *tx_queue)
{
	struct efx_nic *efx = tx_queue->efx;
	efx_oword_t tx_desc_ptr;

	/* The queue should have been flushed */
	WARN_ON(tx_queue->flushed != FLUSH_DONE);

	/* 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 */
	falcon_fini_special_buffer(efx, &tx_queue->txd);
}

/* Free buffers backing TX queue */
void falcon_remove_tx(struct efx_tx_queue *tx_queue)
{
	falcon_free_special_buffer(tx_queue->efx, &tx_queue->txd);
}

/**************************************************************************
 *
 * Falcon RX path
 *
 **************************************************************************/

/* Returns a pointer to the specified descriptor in the RX descriptor queue */
static inline efx_qword_t *falcon_rx_desc(struct efx_rx_queue *rx_queue,
					       unsigned int index)
{
	return (((efx_qword_t *) (rx_queue->rxd.addr)) + index);
}

/* This creates an entry in the RX descriptor queue */
static inline void falcon_build_rx_desc(struct efx_rx_queue *rx_queue,
					unsigned index)
{
	struct efx_rx_buffer *rx_buf;
	efx_qword_t *rxd;

	rxd = falcon_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 falcon_notify_rx_desc(struct efx_rx_queue *rx_queue)
{
	efx_dword_t reg;
	unsigned write_ptr;

	while (rx_queue->notified_count != rx_queue->added_count) {
		falcon_build_rx_desc(rx_queue,
				     rx_queue->notified_count &
				     EFX_RXQ_MASK);
		++rx_queue->notified_count;
	}

	wmb();
	write_ptr = rx_queue->added_count & EFX_RXQ_MASK;
	EFX_POPULATE_DWORD_1(reg, FRF_AZ_RX_DESC_WPTR_DWORD, write_ptr);
	efx_writed_page(rx_queue->efx, &reg,
			FR_AZ_RX_DESC_UPD_DWORD_P0, rx_queue->queue);
}

int falcon_probe_rx(struct efx_rx_queue *rx_queue)
{
	struct efx_nic *efx = rx_queue->efx;
	BUILD_BUG_ON(EFX_RXQ_SIZE < 512 || EFX_RXQ_SIZE > 4096 ||
		     EFX_RXQ_SIZE & EFX_RXQ_MASK);
	return falcon_alloc_special_buffer(efx, &rx_queue->rxd,
					   EFX_RXQ_SIZE * sizeof(efx_qword_t));
}

void falcon_init_rx(struct efx_rx_queue *rx_queue)
{
	efx_oword_t rx_desc_ptr;
	struct efx_nic *efx = rx_queue->efx;
	bool is_b0 = falcon_rev(efx) >= FALCON_REV_B0;
	bool iscsi_digest_en = is_b0;

	EFX_LOG(efx, "RX queue %d ring in special buffers %d-%d\n",
		rx_queue->queue, rx_queue->rxd.index,
		rx_queue->rxd.index + rx_queue->rxd.entries - 1);

	rx_queue->flushed = FLUSH_NONE;

	/* Pin RX descriptor ring */
	falcon_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,
			      rx_queue->channel->channel,
			      FRF_AZ_RX_DESCQ_OWNER_ID, 0,
			      FRF_AZ_RX_DESCQ_LABEL, rx_queue->queue,
			      FRF_AZ_RX_DESCQ_SIZE,
			      __ffs(rx_queue->rxd.entries),
			      FRF_AZ_RX_DESCQ_TYPE, 0 /* kernel queue */ ,
			      /* For >=B0 this is scatter so disable */
			      FRF_AZ_RX_DESCQ_JUMBO, !is_b0,
			      FRF_AZ_RX_DESCQ_EN, 1);
	efx_writeo_table(efx, &rx_desc_ptr, efx->type->rxd_ptr_tbl_base,
			 rx_queue->queue);
}

static void falcon_flush_rx_queue(struct efx_rx_queue *rx_queue)
{
	struct efx_nic *efx = rx_queue->efx;
	efx_oword_t rx_flush_descq;

	rx_queue->flushed = FLUSH_PENDING;

	/* Post a flush command */
	EFX_POPULATE_OWORD_2(rx_flush_descq,
			     FRF_AZ_RX_FLUSH_DESCQ_CMD, 1,
			     FRF_AZ_RX_FLUSH_DESCQ, rx_queue->queue);
	efx_writeo(efx, &rx_flush_descq, FR_AZ_RX_FLUSH_DESCQ);
}

void falcon_fini_rx(struct efx_rx_queue *rx_queue)
{
	efx_oword_t rx_desc_ptr;
	struct efx_nic *efx = rx_queue->efx;

	/* The queue should already have been flushed */
	WARN_ON(rx_queue->flushed != FLUSH_DONE);

	/* 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,
			 rx_queue->queue);

	/* Unpin RX descriptor ring */
	falcon_fini_special_buffer(efx, &rx_queue->rxd);
}

/* Free buffers backing RX queue */
void falcon_remove_rx(struct efx_rx_queue *rx_queue)
{
	falcon_free_special_buffer(rx_queue->efx, &rx_queue->rxd);
}

/**************************************************************************
 *
 * Falcon 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.
 *
 * Note that EVQ_RPTR_REG contains the index of the "last read" event,
 * whereas channel->eventq_read_ptr contains the index of the "next to
 * read" event.
 */
void falcon_eventq_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);
	efx_writed_table(efx, &reg, efx->type->evq_rptr_tbl_base,
			    channel->channel);
}

/* Use HW to insert a SW defined event */
void falcon_generate_event(struct efx_channel *channel, 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, channel->channel);
	efx_writeo(channel->efx, &drv_ev_reg, FR_AZ_DRV_EV);
}

/* Handle a transmit completion event
 *
 * Falcon batches TX completion events; the message we receive is of
 * the form "complete all TX events up to this index".
 */
static void falcon_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;

	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->tx_queue[tx_ev_q_label];
		channel->irq_mod_score +=
			(tx_ev_desc_ptr - tx_queue->read_count) &
			EFX_TXQ_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->tx_queue[tx_ev_q_label];

		if (efx_dev_registered(efx))
			netif_tx_lock(efx->net_dev);
		falcon_notify_tx_desc(tx_queue);
		if (efx_dev_registered(efx))
			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 {
		EFX_ERR(efx, "channel %d unexpected TX event "
			EFX_QWORD_FMT"\n", channel->channel,
			EFX_QWORD_VAL(*event));
	}
}

/* Detect errors included in the rx_evt_pkt_ok bit. */
static void falcon_handle_rx_not_ok(struct efx_rx_queue *rx_queue,
				    const efx_qword_t *event,
				    bool *rx_ev_pkt_ok,
				    bool *discard)
{
	struct efx_nic *efx = rx_queue->efx;
	bool rx_ev_buf_owner_id_err, rx_ev_ip_hdr_chksum_err;
	bool rx_ev_tcp_udp_chksum_err, rx_ev_eth_crc_err;
	bool rx_ev_frm_trunc, rx_ev_drib_nib, rx_ev_tobe_disc;
	bool rx_ev_other_err, rx_ev_pause_frm;
	bool rx_ev_ip_frag_err, rx_ev_hdr_type, rx_ev_mcast_pkt;
	unsigned rx_ev_pkt_type;

	rx_ev_hdr_type = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_HDR_TYPE);
	rx_ev_mcast_pkt = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_MCAST_PKT);
	rx_ev_tobe_disc = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_TOBE_DISC);
	rx_ev_pkt_type = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_PKT_TYPE);
	rx_ev_buf_owner_id_err = EFX_QWORD_FIELD(*event,
						 FSF_AZ_RX_EV_BUF_OWNER_ID_ERR);
	rx_ev_ip_frag_err = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_IP_FRAG_ERR);
	rx_ev_ip_hdr_chksum_err = EFX_QWORD_FIELD(*event,
						  FSF_AZ_RX_EV_IP_HDR_CHKSUM_ERR);
	rx_ev_tcp_udp_chksum_err = EFX_QWORD_FIELD(*event,
						   FSF_AZ_RX_EV_TCP_UDP_CHKSUM_ERR);
	rx_ev_eth_crc_err = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_ETH_CRC_ERR);
	rx_ev_frm_trunc = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_FRM_TRUNC);
	rx_ev_drib_nib = ((falcon_rev(efx) >= FALCON_REV_B0) ?
			  0 : EFX_QWORD_FIELD(*event, FSF_AA_RX_EV_DRIB_NIB));
	rx_ev_pause_frm = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_PAUSE_FRM_ERR);

	/* Every error apart from tobe_disc and pause_frm */
	rx_ev_other_err = (rx_ev_drib_nib | rx_ev_tcp_udp_chksum_err |
			   rx_ev_buf_owner_id_err | rx_ev_eth_crc_err |
			   rx_ev_frm_trunc | rx_ev_ip_hdr_chksum_err);

	/* Count errors that are not in MAC stats.  Ignore expected
	 * checksum errors during self-test. */
	if (rx_ev_frm_trunc)
		++rx_queue->channel->n_rx_frm_trunc;
	else if (rx_ev_tobe_disc)
		++rx_queue->channel->n_rx_tobe_disc;
	else if (!efx->loopback_selftest) {
		if (rx_ev_ip_hdr_chksum_err)
			++rx_queue->channel->n_rx_ip_hdr_chksum_err;
		else if (rx_ev_tcp_udp_chksum_err)
			++rx_queue->channel->n_rx_tcp_udp_chksum_err;
	}
	if (rx_ev_ip_frag_err)
		++rx_queue->channel->n_rx_ip_frag_err;

	/* The frame must be discarded if any of these are true. */
	*discard = (rx_ev_eth_crc_err | rx_ev_frm_trunc | rx_ev_drib_nib |
		    rx_ev_tobe_disc | rx_ev_pause_frm);

	/* TOBE_DISC is expected on unicast mismatches; don't print out an
	 * error message.  FRM_TRUNC indicates RXDP dropped the packet due
	 * to a FIFO overflow.
	 */
#ifdef EFX_ENABLE_DEBUG
	if (rx_ev_other_err) {
		EFX_INFO_RL(efx, " RX queue %d unexpected RX event "
			    EFX_QWORD_FMT "%s%s%s%s%s%s%s%s\n",
			    rx_queue->queue, EFX_QWORD_VAL(*event),
			    rx_ev_buf_owner_id_err ? " [OWNER_ID_ERR]" : "",
			    rx_ev_ip_hdr_chksum_err ?
			    " [IP_HDR_CHKSUM_ERR]" : "",
			    rx_ev_tcp_udp_chksum_err ?
			    " [TCP_UDP_CHKSUM_ERR]" : "",
			    rx_ev_eth_crc_err ? " [ETH_CRC_ERR]" : "",
			    rx_ev_frm_trunc ? " [FRM_TRUNC]" : "",
			    rx_ev_drib_nib ? " [DRIB_NIB]" : "",
			    rx_ev_tobe_disc ? " [TOBE_DISC]" : "",
			    rx_ev_pause_frm ? " [PAUSE]" : "");
	}
#endif
}

/* Handle receive events that are not in-order. */
static void falcon_handle_rx_bad_index(struct efx_rx_queue *rx_queue,
				       unsigned index)
{
	struct efx_nic *efx = rx_queue->efx;
	unsigned expected, dropped;

	expected = rx_queue->removed_count & EFX_RXQ_MASK;
	dropped = (index - expected) & EFX_RXQ_MASK;
	EFX_INFO(efx, "dropped %d events (index=%d expected=%d)\n",
		dropped, index, expected);

	efx_schedule_reset(efx, EFX_WORKAROUND_5676(efx) ?
			   RESET_TYPE_RX_RECOVERY : RESET_TYPE_DISABLE);
}

/* Handle a packet received event
 *
 * Falcon silicon gives a "discard" flag if it's a unicast packet with the
 * wrong destination address
 * Also "is multicast" and "matches multicast filter" flags can be used to
 * discard non-matching multicast packets.
 */
static void falcon_handle_rx_event(struct efx_channel *channel,
				   const efx_qword_t *event)
{
	unsigned int rx_ev_desc_ptr, rx_ev_byte_cnt;
	unsigned int rx_ev_hdr_type, rx_ev_mcast_pkt;
	unsigned expected_ptr;
	bool rx_ev_pkt_ok, discard = false, checksummed;
	struct efx_rx_queue *rx_queue;
	struct efx_nic *efx = channel->efx;

	/* Basic packet information */
	rx_ev_byte_cnt = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_BYTE_CNT);
	rx_ev_pkt_ok = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_PKT_OK);
	rx_ev_hdr_type = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_HDR_TYPE);
	WARN_ON(EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_JUMBO_CONT));
	WARN_ON(EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_SOP) != 1);
	WARN_ON(EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_Q_LABEL) !=
		channel->channel);

	rx_queue = &efx->rx_queue[channel->channel];

	rx_ev_desc_ptr = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_DESC_PTR);
	expected_ptr = rx_queue->removed_count & EFX_RXQ_MASK;
	if (unlikely(rx_ev_desc_ptr != expected_ptr))
		falcon_handle_rx_bad_index(rx_queue, rx_ev_desc_ptr);

	if (likely(rx_ev_pkt_ok)) {
		/* If packet is marked as OK and packet type is TCP/IPv4 or
		 * UDP/IPv4, then we can rely on the hardware checksum.
		 */
		checksummed =
			efx->rx_checksum_enabled &&
			(rx_ev_hdr_type == FSE_AB_RX_EV_HDR_TYPE_IPV4_TCP ||
			 rx_ev_hdr_type == FSE_AB_RX_EV_HDR_TYPE_IPV4_UDP);
	} else {
		falcon_handle_rx_not_ok(rx_queue, event, &rx_ev_pkt_ok,
					&discard);
		checksummed = false;
	}

	/* Detect multicast packets that didn't match the filter */
	rx_ev_mcast_pkt = EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_MCAST_PKT);
	if (rx_ev_mcast_pkt) {
		unsigned int rx_ev_mcast_hash_match =
			EFX_QWORD_FIELD(*event, FSF_AZ_RX_EV_MCAST_HASH_MATCH);

		if (unlikely(!rx_ev_mcast_hash_match))
			discard = true;
	}

	channel->irq_mod_score += 2;

	/* Handle received packet */
	efx_rx_packet(rx_queue, rx_ev_desc_ptr, rx_ev_byte_cnt,
		      checksummed, discard);
}

/* Global events are basically PHY events */
static void falcon_handle_global_event(struct efx_channel *channel,
				       efx_qword_t *event)
{
	struct efx_nic *efx = channel->efx;
	bool handled = false;

	if (EFX_QWORD_FIELD(*event, FSF_AB_GLB_EV_G_PHY0_INTR) ||
	    EFX_QWORD_FIELD(*event, FSF_AB_GLB_EV_XG_PHY0_INTR) ||
	    EFX_QWORD_FIELD(*event, FSF_AB_GLB_EV_XFP_PHY0_INTR)) {
		/* Ignored */
		handled = true;
	}

	if ((falcon_rev(efx) >= FALCON_REV_B0) &&
	    EFX_QWORD_FIELD(*event, FSF_BB_GLB_EV_XG_MGT_INTR)) {
		efx->xmac_poll_required = true;
		handled = true;
	}

	if (falcon_rev(efx) <= FALCON_REV_A1 ?
	    EFX_QWORD_FIELD(*event, FSF_AA_GLB_EV_RX_RECOVERY) :
	    EFX_QWORD_FIELD(*event, FSF_BB_GLB_EV_RX_RECOVERY)) {
		EFX_ERR(efx, "channel %d seen global RX_RESET "
			"event. Resetting.\n", channel->channel);

		atomic_inc(&efx->rx_reset);
		efx_schedule_reset(efx, EFX_WORKAROUND_6555(efx) ?
				   RESET_TYPE_RX_RECOVERY : RESET_TYPE_DISABLE);
		handled = true;
	}

	if (!handled)
		EFX_ERR(efx, "channel %d unknown global event "
			EFX_QWORD_FMT "\n", channel->channel,
			EFX_QWORD_VAL(*event));
}

static void falcon_handle_driver_event(struct efx_channel *channel,
				       efx_qword_t *event)
{
	struct efx_nic *efx = channel->efx;
	unsigned int ev_sub_code;
	unsigned int ev_sub_data;

	ev_sub_code = EFX_QWORD_FIELD(*event, FSF_AZ_DRIVER_EV_SUBCODE);
	ev_sub_data = EFX_QWORD_FIELD(*event, FSF_AZ_DRIVER_EV_SUBDATA);

	switch (ev_sub_code) {
	case FSE_AZ_TX_DESCQ_FLS_DONE_EV:
		EFX_TRACE(efx, "channel %d TXQ %d flushed\n",
			  channel->channel, ev_sub_data);
		break;
	case FSE_AZ_RX_DESCQ_FLS_DONE_EV:
		EFX_TRACE(efx, "channel %d RXQ %d flushed\n",
			  channel->channel, ev_sub_data);
		break;
	case FSE_AZ_EVQ_INIT_DONE_EV:
		EFX_LOG(efx, "channel %d EVQ %d initialised\n",
			channel->channel, ev_sub_data);
		break;
	case FSE_AZ_SRM_UPD_DONE_EV:
		EFX_TRACE(efx, "channel %d SRAM update done\n",
			  channel->channel);
		break;
	case FSE_AZ_WAKE_UP_EV:
		EFX_TRACE(efx, "channel %d RXQ %d wakeup event\n",
			  channel->channel, ev_sub_data);
		break;
	case FSE_AZ_TIMER_EV:
		EFX_TRACE(efx, "channel %d RX queue %d timer expired\n",
			  channel->channel, ev_sub_data);
		break;
	case FSE_AA_RX_RECOVER_EV:
		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 :
				   RESET_TYPE_DISABLE);
		break;
	case FSE_BZ_RX_DSC_ERROR_EV:
		EFX_ERR(efx, "RX DMA Q %d reports descriptor fetch error."
			" RX Q %d is disabled.\n", ev_sub_data, ev_sub_data);
		efx_schedule_reset(efx, RESET_TYPE_RX_DESC_FETCH);
		break;
	case FSE_BZ_TX_DSC_ERROR_EV:
		EFX_ERR(efx, "TX DMA Q %d reports descriptor fetch error."
			" TX Q %d is disabled.\n", ev_sub_data, ev_sub_data);
		efx_schedule_reset(efx, RESET_TYPE_TX_DESC_FETCH);
		break;
	default:
		EFX_TRACE(efx, "channel %d unknown driver event code %d "
			  "data %04x\n", channel->channel, ev_sub_code,
			  ev_sub_data);
		break;
	}
}

int falcon_process_eventq(struct efx_channel *channel, int rx_quota)
{
	unsigned int read_ptr;
	efx_qword_t event, *p_event;
	int ev_code;
	int rx_packets = 0;

	read_ptr = channel->eventq_read_ptr;

	do {
		p_event = falcon_event(channel, read_ptr);
		event = *p_event;

		if (!falcon_event_present(&event))
			/* End of events */