linux/drivers/net/ethernet/sfc/net_driver.h

1357 lines
50 KiB
C
Raw Normal View History

/****************************************************************************
* Driver for Solarflare network controllers and boards
* Copyright 2005-2006 Fen Systems Ltd.
* Copyright 2005-2013 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.
*/
/* Common definitions for all Efx net driver code */
#ifndef EFX_NET_DRIVER_H
#define EFX_NET_DRIVER_H
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/ethtool.h>
#include <linux/if_vlan.h>
#include <linux/timer.h>
#include <linux/mdio.h>
#include <linux/list.h>
#include <linux/pci.h>
#include <linux/device.h>
#include <linux/highmem.h>
#include <linux/workqueue.h>
#include <linux/mutex.h>
#include <linux/vmalloc.h>
#include <linux/i2c.h>
#include <linux/mtd/mtd.h>
#include "enum.h"
#include "bitfield.h"
#include "filter.h"
/**************************************************************************
*
* Build definitions
*
**************************************************************************/
#define EFX_DRIVER_VERSION "4.0"
#ifdef DEBUG
#define EFX_BUG_ON_PARANOID(x) BUG_ON(x)
#define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
#else
#define EFX_BUG_ON_PARANOID(x) do {} while (0)
#define EFX_WARN_ON_PARANOID(x) do {} while (0)
#endif
/**************************************************************************
*
* Efx data structures
*
**************************************************************************/
#define EFX_MAX_CHANNELS 32U
#define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
#define EFX_EXTRA_CHANNEL_IOV 0
#define EFX_EXTRA_CHANNEL_PTP 1
#define EFX_MAX_EXTRA_CHANNELS 2U
/* Checksum generation is a per-queue option in hardware, so each
* queue visible to the networking core is backed by two hardware TX
* queues. */
#define EFX_MAX_TX_TC 2
#define EFX_MAX_CORE_TX_QUEUES (EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
#define EFX_TXQ_TYPE_OFFLOAD 1 /* flag */
#define EFX_TXQ_TYPE_HIGHPRI 2 /* flag */
#define EFX_TXQ_TYPES 4
#define EFX_MAX_TX_QUEUES (EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
/* Maximum possible MTU the driver supports */
#define EFX_MAX_MTU (9 * 1024)
sfc: Reduce RX scatter buffer size, and reduce alignment if appropriate efx_start_datapath() asserts that we can fit 2 RX scatter buffers plus a software structure, each appropriately aligned, into a single page. Where L1_CACHE_BYTES == 256 and PAGE_SIZE == 4096, which is the case on s390, this assertion fails. The current scatter buffer size is also not a multiple of 64 or 128, which are more common cache line sizes. If we can make both the start and end of a scatter buffer cache-aligned, this will reduce the need for read-modify-write operations on inter- processor links. Fix the alignment by reducing EFX_RX_USR_BUF_SIZE to 2048 - 256 == 1792. (We could use 2048 - L1_CACHE_BYTES, but EFX_RX_USR_BUF_SIZE also affects user-level networking where a larger amount of housekeeping data may be needed. Although this version of the driver does not support user-level networking, I prefer to keep scattering behaviour consistent with the out-of-tree version.) This still doesn't fix the s390 build because like most architectures it has NET_IP_ALIGN == 2. When NET_IP_ALIGN != 0 we cannot achieve cache line alignment at either the start or end of a scatter buffer, so there is actually no point in padding the buffers to a multiple of the cache line size. All we need is 4-byte alignment of the network header, so do that. Adjust the assertions accordingly. Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Acked-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-05-13 20:01:22 +08:00
/* Size of an RX scatter buffer. Small enough to pack 2 into a 4K page,
* and should be a multiple of the cache line size.
*/
#define EFX_RX_USR_BUF_SIZE (2048 - 256)
/* If possible, we should ensure cache line alignment at start and end
* of every buffer. Otherwise, we just need to ensure 4-byte
* alignment of the network header.
*/
#if NET_IP_ALIGN == 0
#define EFX_RX_BUF_ALIGNMENT L1_CACHE_BYTES
#else
#define EFX_RX_BUF_ALIGNMENT 4
#endif
/* Forward declare Precision Time Protocol (PTP) support structure. */
struct efx_ptp_data;
struct hwtstamp_config;
struct efx_self_tests;
/**
* struct efx_buffer - A general-purpose DMA buffer
* @addr: host base address of the buffer
* @dma_addr: DMA base address of the buffer
* @len: Buffer length, in bytes
*
* The NIC uses these buffers for its interrupt status registers and
* MAC stats dumps.
*/
struct efx_buffer {
void *addr;
dma_addr_t dma_addr;
unsigned int len;
};
/**
* struct efx_special_buffer - DMA buffer entered into buffer table
* @buf: Standard &struct efx_buffer
* @index: Buffer index within controller;s buffer table
* @entries: Number of buffer table entries
*
* The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
* Event and descriptor rings are addressed via one or more buffer
* table entries (and so can be physically non-contiguous, although we
* currently do not take advantage of that). On Falcon and Siena we
* have to take care of allocating and initialising the entries
* ourselves. On later hardware this is managed by the firmware and
* @index and @entries are left as 0.
*/
struct efx_special_buffer {
struct efx_buffer buf;
unsigned int index;
unsigned int entries;
};
/**
* struct efx_tx_buffer - buffer state for a TX descriptor
* @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
* freed when descriptor completes
* @heap_buf: When @flags & %EFX_TX_BUF_HEAP, the associated heap buffer to be
* freed when descriptor completes.
* @option: When @flags & %EFX_TX_BUF_OPTION, a NIC-specific option descriptor.
* @dma_addr: DMA address of the fragment.
* @flags: Flags for allocation and DMA mapping type
* @len: Length of this fragment.
* This field is zero when the queue slot is empty.
* @unmap_len: Length of this fragment to unmap
* @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
* Only valid if @unmap_len != 0.
*/
struct efx_tx_buffer {
union {
const struct sk_buff *skb;
void *heap_buf;
};
union {
efx_qword_t option;
dma_addr_t dma_addr;
};
unsigned short flags;
unsigned short len;
unsigned short unmap_len;
unsigned short dma_offset;
};
#define EFX_TX_BUF_CONT 1 /* not last descriptor of packet */
#define EFX_TX_BUF_SKB 2 /* buffer is last part of skb */
#define EFX_TX_BUF_HEAP 4 /* buffer was allocated with kmalloc() */
#define EFX_TX_BUF_MAP_SINGLE 8 /* buffer was mapped with dma_map_single() */
#define EFX_TX_BUF_OPTION 0x10 /* empty buffer for option descriptor */
/**
* struct efx_tx_queue - An Efx TX queue
*
* This is a ring buffer of TX fragments.
* Since the TX completion path always executes on the same
* CPU and the xmit path can operate on different CPUs,
* performance is increased by ensuring that the completion
* path and the xmit path operate on different cache lines.
* This is particularly important if the xmit path is always
* executing on one CPU which is different from the completion
* path. There is also a cache line for members which are
* read but not written on the fast path.
*
* @efx: The associated Efx NIC
* @queue: DMA queue number
* @channel: The associated channel
* @core_txq: The networking core TX queue structure
* @buffer: The software buffer ring
* @tsoh_page: Array of pages of TSO header buffers
* @txd: The hardware descriptor ring
* @ptr_mask: The size of the ring minus 1.
* @piobuf: PIO buffer region for this TX queue (shared with its partner).
* Size of the region is efx_piobuf_size.
* @piobuf_offset: Buffer offset to be specified in PIO descriptors
* @initialised: Has hardware queue been initialised?
* @read_count: Current read pointer.
* This is the number of buffers that have been removed from both rings.
sfc: Use TX push whenever adding descriptors to an empty queue Whenever we add DMA descriptors to a TX ring and update the ring pointer, the TX DMA engine must first read the new DMA descriptors and then start reading packet data. However, all released Solarflare 10G controllers have a 'TX push' feature that allows us to reduce latency by writing the first new DMA descriptor along with the pointer update. This is only useful when the queue is empty. The hardware should ignore the pushed descriptor if the queue is not empty, but this check is buggy, so we must do it in software. In order to tell whether a TX queue is empty, we need to compare the previous transmission count (write_count) and completion count (read_count). However, if we do that every time we update the ring pointer then read_count may ping-pong between the caches of two CPUs running the transmission and completion paths for the queue. Therefore, we split the check for an empty queue between the completion path and the transmission path: - Add an empty_read_count field representing a point at which the completion path saw the TX queue as empty. - Add an old_write_count field for use on the completion path. - On the completion path, whenever read_count reaches or passes old_write_count the TX queue may be empty. We then read write_count, set empty_read_count if read_count == write_count, and update old_write_count. - On the transmission path, we read empty_read_count. If it's set, we compare it with the value of write_count before the current set of descriptors was added. If they match, the queue really is empty and we can use TX push. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2010-11-16 07:53:11 +08:00
* @old_write_count: The value of @write_count when last checked.
* This is here for performance reasons. The xmit path will
* only get the up-to-date value of @write_count if this
* variable indicates that the queue is empty. This is to
* avoid cache-line ping-pong between the xmit path and the
* completion path.
* @merge_events: Number of TX merged completion events
* @insert_count: Current insert pointer
* This is the number of buffers that have been added to the
* software ring.
* @write_count: Current write pointer
* This is the number of buffers that have been added to the
* hardware ring.
* @old_read_count: The value of read_count when last checked.
* This is here for performance reasons. The xmit path will
* only get the up-to-date value of read_count if this
* variable indicates that the queue is full. This is to
* avoid cache-line ping-pong between the xmit path and the
* completion path.
* @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
sfc: Use TX push whenever adding descriptors to an empty queue Whenever we add DMA descriptors to a TX ring and update the ring pointer, the TX DMA engine must first read the new DMA descriptors and then start reading packet data. However, all released Solarflare 10G controllers have a 'TX push' feature that allows us to reduce latency by writing the first new DMA descriptor along with the pointer update. This is only useful when the queue is empty. The hardware should ignore the pushed descriptor if the queue is not empty, but this check is buggy, so we must do it in software. In order to tell whether a TX queue is empty, we need to compare the previous transmission count (write_count) and completion count (read_count). However, if we do that every time we update the ring pointer then read_count may ping-pong between the caches of two CPUs running the transmission and completion paths for the queue. Therefore, we split the check for an empty queue between the completion path and the transmission path: - Add an empty_read_count field representing a point at which the completion path saw the TX queue as empty. - Add an old_write_count field for use on the completion path. - On the completion path, whenever read_count reaches or passes old_write_count the TX queue may be empty. We then read write_count, set empty_read_count if read_count == write_count, and update old_write_count. - On the transmission path, we read empty_read_count. If it's set, we compare it with the value of write_count before the current set of descriptors was added. If they match, the queue really is empty and we can use TX push. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2010-11-16 07:53:11 +08:00
* @pushes: Number of times the TX push feature has been used
* @pio_packets: Number of times the TX PIO feature has been used
sfc: Use TX push whenever adding descriptors to an empty queue Whenever we add DMA descriptors to a TX ring and update the ring pointer, the TX DMA engine must first read the new DMA descriptors and then start reading packet data. However, all released Solarflare 10G controllers have a 'TX push' feature that allows us to reduce latency by writing the first new DMA descriptor along with the pointer update. This is only useful when the queue is empty. The hardware should ignore the pushed descriptor if the queue is not empty, but this check is buggy, so we must do it in software. In order to tell whether a TX queue is empty, we need to compare the previous transmission count (write_count) and completion count (read_count). However, if we do that every time we update the ring pointer then read_count may ping-pong between the caches of two CPUs running the transmission and completion paths for the queue. Therefore, we split the check for an empty queue between the completion path and the transmission path: - Add an empty_read_count field representing a point at which the completion path saw the TX queue as empty. - Add an old_write_count field for use on the completion path. - On the completion path, whenever read_count reaches or passes old_write_count the TX queue may be empty. We then read write_count, set empty_read_count if read_count == write_count, and update old_write_count. - On the transmission path, we read empty_read_count. If it's set, we compare it with the value of write_count before the current set of descriptors was added. If they match, the queue really is empty and we can use TX push. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2010-11-16 07:53:11 +08:00
* @empty_read_count: If the completion path has seen the queue as empty
* and the transmission path has not yet checked this, the value of
* @read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
*/
struct efx_tx_queue {
/* Members which don't change on the fast path */
struct efx_nic *efx ____cacheline_aligned_in_smp;
unsigned queue;
struct efx_channel *channel;
struct netdev_queue *core_txq;
struct efx_tx_buffer *buffer;
struct efx_buffer *tsoh_page;
struct efx_special_buffer txd;
unsigned int ptr_mask;
void __iomem *piobuf;
unsigned int piobuf_offset;
bool initialised;
/* Members used mainly on the completion path */
unsigned int read_count ____cacheline_aligned_in_smp;
sfc: Use TX push whenever adding descriptors to an empty queue Whenever we add DMA descriptors to a TX ring and update the ring pointer, the TX DMA engine must first read the new DMA descriptors and then start reading packet data. However, all released Solarflare 10G controllers have a 'TX push' feature that allows us to reduce latency by writing the first new DMA descriptor along with the pointer update. This is only useful when the queue is empty. The hardware should ignore the pushed descriptor if the queue is not empty, but this check is buggy, so we must do it in software. In order to tell whether a TX queue is empty, we need to compare the previous transmission count (write_count) and completion count (read_count). However, if we do that every time we update the ring pointer then read_count may ping-pong between the caches of two CPUs running the transmission and completion paths for the queue. Therefore, we split the check for an empty queue between the completion path and the transmission path: - Add an empty_read_count field representing a point at which the completion path saw the TX queue as empty. - Add an old_write_count field for use on the completion path. - On the completion path, whenever read_count reaches or passes old_write_count the TX queue may be empty. We then read write_count, set empty_read_count if read_count == write_count, and update old_write_count. - On the transmission path, we read empty_read_count. If it's set, we compare it with the value of write_count before the current set of descriptors was added. If they match, the queue really is empty and we can use TX push. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2010-11-16 07:53:11 +08:00
unsigned int old_write_count;
unsigned int merge_events;
/* Members used only on the xmit path */
unsigned int insert_count ____cacheline_aligned_in_smp;
unsigned int write_count;
unsigned int old_read_count;
unsigned int tso_bursts;
unsigned int tso_long_headers;
unsigned int tso_packets;
sfc: Use TX push whenever adding descriptors to an empty queue Whenever we add DMA descriptors to a TX ring and update the ring pointer, the TX DMA engine must first read the new DMA descriptors and then start reading packet data. However, all released Solarflare 10G controllers have a 'TX push' feature that allows us to reduce latency by writing the first new DMA descriptor along with the pointer update. This is only useful when the queue is empty. The hardware should ignore the pushed descriptor if the queue is not empty, but this check is buggy, so we must do it in software. In order to tell whether a TX queue is empty, we need to compare the previous transmission count (write_count) and completion count (read_count). However, if we do that every time we update the ring pointer then read_count may ping-pong between the caches of two CPUs running the transmission and completion paths for the queue. Therefore, we split the check for an empty queue between the completion path and the transmission path: - Add an empty_read_count field representing a point at which the completion path saw the TX queue as empty. - Add an old_write_count field for use on the completion path. - On the completion path, whenever read_count reaches or passes old_write_count the TX queue may be empty. We then read write_count, set empty_read_count if read_count == write_count, and update old_write_count. - On the transmission path, we read empty_read_count. If it's set, we compare it with the value of write_count before the current set of descriptors was added. If they match, the queue really is empty and we can use TX push. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2010-11-16 07:53:11 +08:00
unsigned int pushes;
unsigned int pio_packets;
sfc: Use TX push whenever adding descriptors to an empty queue Whenever we add DMA descriptors to a TX ring and update the ring pointer, the TX DMA engine must first read the new DMA descriptors and then start reading packet data. However, all released Solarflare 10G controllers have a 'TX push' feature that allows us to reduce latency by writing the first new DMA descriptor along with the pointer update. This is only useful when the queue is empty. The hardware should ignore the pushed descriptor if the queue is not empty, but this check is buggy, so we must do it in software. In order to tell whether a TX queue is empty, we need to compare the previous transmission count (write_count) and completion count (read_count). However, if we do that every time we update the ring pointer then read_count may ping-pong between the caches of two CPUs running the transmission and completion paths for the queue. Therefore, we split the check for an empty queue between the completion path and the transmission path: - Add an empty_read_count field representing a point at which the completion path saw the TX queue as empty. - Add an old_write_count field for use on the completion path. - On the completion path, whenever read_count reaches or passes old_write_count the TX queue may be empty. We then read write_count, set empty_read_count if read_count == write_count, and update old_write_count. - On the transmission path, we read empty_read_count. If it's set, we compare it with the value of write_count before the current set of descriptors was added. If they match, the queue really is empty and we can use TX push. Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2010-11-16 07:53:11 +08:00
/* Members shared between paths and sometimes updated */
unsigned int empty_read_count ____cacheline_aligned_in_smp;
#define EFX_EMPTY_COUNT_VALID 0x80000000
atomic_t flush_outstanding;
};
/**
* struct efx_rx_buffer - An Efx RX data buffer
* @dma_addr: DMA base address of the buffer
* @page: The associated page buffer.
* Will be %NULL if the buffer slot is currently free.
* @page_offset: If pending: offset in @page of DMA base address.
* If completed: offset in @page of Ethernet header.
* @len: If pending: length for DMA descriptor.
* If completed: received length, excluding hash prefix.
* @flags: Flags for buffer and packet state. These are only set on the
* first buffer of a scattered packet.
*/
struct efx_rx_buffer {
dma_addr_t dma_addr;
struct page *page;
u16 page_offset;
u16 len;
u16 flags;
};
#define EFX_RX_BUF_LAST_IN_PAGE 0x0001
#define EFX_RX_PKT_CSUMMED 0x0002
#define EFX_RX_PKT_DISCARD 0x0004
#define EFX_RX_PKT_TCP 0x0040
#define EFX_RX_PKT_PREFIX_LEN 0x0080 /* length is in prefix only */
/**
* struct efx_rx_page_state - Page-based rx buffer state
*
* Inserted at the start of every page allocated for receive buffers.
* Used to facilitate sharing dma mappings between recycled rx buffers
* and those passed up to the kernel.
*
* @dma_addr: The dma address of this page.
*/
struct efx_rx_page_state {
dma_addr_t dma_addr;
unsigned int __pad[0] ____cacheline_aligned;
};
/**
* struct efx_rx_queue - An Efx RX queue
* @efx: The associated Efx NIC
* @core_index: Index of network core RX queue. Will be >= 0 iff this
* is associated with a real RX queue.
* @buffer: The software buffer ring
* @rxd: The hardware descriptor ring
* @ptr_mask: The size of the ring minus 1.
* @refill_enabled: Enable refill whenever fill level is low
* @flush_pending: Set when a RX flush is pending. Has the same lifetime as
* @rxq_flush_pending.
* @added_count: Number of buffers added to the receive queue.
* @notified_count: Number of buffers given to NIC (<= @added_count).
* @removed_count: Number of buffers removed from the receive queue.
* @scatter_n: Used by NIC specific receive code.
* @scatter_len: Used by NIC specific receive code.
sfc: reuse pages to avoid DMA mapping/unmapping costs On POWER systems, DMA mapping/unmapping operations are very expensive. These changes reduce these costs by trying to reuse DMA mapped pages. After all the buffers associated with a page have been processed and passed up, the page is placed into a ring (if there is room). For each page that is required for a refill operation, a page in the ring is examined to determine if its page count has fallen to 1, ie. the kernel has released its reference to these packets. If this is the case, the page can be immediately added back into the RX descriptor ring, without having to re-map it for DMA. If the kernel is still holding a reference to this page, it is removed from the ring and unmapped for DMA. Then a new page, which can immediately be used by RX buffers in the descriptor ring, is allocated and DMA mapped. The time a page needs to spend in the recycle ring before the kernel has released its page references is based on the number of buffers that use this page. As large pages can hold more RX buffers, the RX recycle ring can be shorter. This reduces memory usage on POWER systems, while maintaining the performance gain achieved by recycling pages, following the driver change to pack more than two RX buffers into large pages. When an IOMMU is not present, the recycle ring can be small to reduce memory usage, since DMA mapping operations are inexpensive. With a small recycle ring, attempting to refill the descriptor queue with more buffers than the equivalent size of the recycle ring could ultimately lead to memory leaks if page entries in the recycle ring were overwritten. To prevent this, the check to see if the recycle ring is full is changed to check if the next entry to be written is NULL. [bwh: Combine and rebase several commits so this is complete before the following buffer-packing changes. Remove module parameter.] Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2013-02-13 18:54:41 +08:00
* @page_ring: The ring to store DMA mapped pages for reuse.
* @page_add: Counter to calculate the write pointer for the recycle ring.
* @page_remove: Counter to calculate the read pointer for the recycle ring.
* @page_recycle_count: The number of pages that have been recycled.
* @page_recycle_failed: The number of pages that couldn't be recycled because
* the kernel still held a reference to them.
* @page_recycle_full: The number of pages that were released because the
* recycle ring was full.
* @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
* @max_fill: RX descriptor maximum fill level (<= ring size)
* @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
* (<= @max_fill)
* @min_fill: RX descriptor minimum non-zero fill level.
* This records the minimum fill level observed when a ring
* refill was triggered.
sfc: reuse pages to avoid DMA mapping/unmapping costs On POWER systems, DMA mapping/unmapping operations are very expensive. These changes reduce these costs by trying to reuse DMA mapped pages. After all the buffers associated with a page have been processed and passed up, the page is placed into a ring (if there is room). For each page that is required for a refill operation, a page in the ring is examined to determine if its page count has fallen to 1, ie. the kernel has released its reference to these packets. If this is the case, the page can be immediately added back into the RX descriptor ring, without having to re-map it for DMA. If the kernel is still holding a reference to this page, it is removed from the ring and unmapped for DMA. Then a new page, which can immediately be used by RX buffers in the descriptor ring, is allocated and DMA mapped. The time a page needs to spend in the recycle ring before the kernel has released its page references is based on the number of buffers that use this page. As large pages can hold more RX buffers, the RX recycle ring can be shorter. This reduces memory usage on POWER systems, while maintaining the performance gain achieved by recycling pages, following the driver change to pack more than two RX buffers into large pages. When an IOMMU is not present, the recycle ring can be small to reduce memory usage, since DMA mapping operations are inexpensive. With a small recycle ring, attempting to refill the descriptor queue with more buffers than the equivalent size of the recycle ring could ultimately lead to memory leaks if page entries in the recycle ring were overwritten. To prevent this, the check to see if the recycle ring is full is changed to check if the next entry to be written is NULL. [bwh: Combine and rebase several commits so this is complete before the following buffer-packing changes. Remove module parameter.] Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2013-02-13 18:54:41 +08:00
* @recycle_count: RX buffer recycle counter.
* @slow_fill: Timer used to defer efx_nic_generate_fill_event().
*/
struct efx_rx_queue {
struct efx_nic *efx;
int core_index;
struct efx_rx_buffer *buffer;
struct efx_special_buffer rxd;
unsigned int ptr_mask;
bool refill_enabled;
bool flush_pending;
unsigned int added_count;
unsigned int notified_count;
unsigned int removed_count;
unsigned int scatter_n;
unsigned int scatter_len;
sfc: reuse pages to avoid DMA mapping/unmapping costs On POWER systems, DMA mapping/unmapping operations are very expensive. These changes reduce these costs by trying to reuse DMA mapped pages. After all the buffers associated with a page have been processed and passed up, the page is placed into a ring (if there is room). For each page that is required for a refill operation, a page in the ring is examined to determine if its page count has fallen to 1, ie. the kernel has released its reference to these packets. If this is the case, the page can be immediately added back into the RX descriptor ring, without having to re-map it for DMA. If the kernel is still holding a reference to this page, it is removed from the ring and unmapped for DMA. Then a new page, which can immediately be used by RX buffers in the descriptor ring, is allocated and DMA mapped. The time a page needs to spend in the recycle ring before the kernel has released its page references is based on the number of buffers that use this page. As large pages can hold more RX buffers, the RX recycle ring can be shorter. This reduces memory usage on POWER systems, while maintaining the performance gain achieved by recycling pages, following the driver change to pack more than two RX buffers into large pages. When an IOMMU is not present, the recycle ring can be small to reduce memory usage, since DMA mapping operations are inexpensive. With a small recycle ring, attempting to refill the descriptor queue with more buffers than the equivalent size of the recycle ring could ultimately lead to memory leaks if page entries in the recycle ring were overwritten. To prevent this, the check to see if the recycle ring is full is changed to check if the next entry to be written is NULL. [bwh: Combine and rebase several commits so this is complete before the following buffer-packing changes. Remove module parameter.] Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2013-02-13 18:54:41 +08:00
struct page **page_ring;
unsigned int page_add;
unsigned int page_remove;
unsigned int page_recycle_count;
unsigned int page_recycle_failed;
unsigned int page_recycle_full;
unsigned int page_ptr_mask;
unsigned int max_fill;
unsigned int fast_fill_trigger;
unsigned int min_fill;
unsigned int min_overfill;
sfc: reuse pages to avoid DMA mapping/unmapping costs On POWER systems, DMA mapping/unmapping operations are very expensive. These changes reduce these costs by trying to reuse DMA mapped pages. After all the buffers associated with a page have been processed and passed up, the page is placed into a ring (if there is room). For each page that is required for a refill operation, a page in the ring is examined to determine if its page count has fallen to 1, ie. the kernel has released its reference to these packets. If this is the case, the page can be immediately added back into the RX descriptor ring, without having to re-map it for DMA. If the kernel is still holding a reference to this page, it is removed from the ring and unmapped for DMA. Then a new page, which can immediately be used by RX buffers in the descriptor ring, is allocated and DMA mapped. The time a page needs to spend in the recycle ring before the kernel has released its page references is based on the number of buffers that use this page. As large pages can hold more RX buffers, the RX recycle ring can be shorter. This reduces memory usage on POWER systems, while maintaining the performance gain achieved by recycling pages, following the driver change to pack more than two RX buffers into large pages. When an IOMMU is not present, the recycle ring can be small to reduce memory usage, since DMA mapping operations are inexpensive. With a small recycle ring, attempting to refill the descriptor queue with more buffers than the equivalent size of the recycle ring could ultimately lead to memory leaks if page entries in the recycle ring were overwritten. To prevent this, the check to see if the recycle ring is full is changed to check if the next entry to be written is NULL. [bwh: Combine and rebase several commits so this is complete before the following buffer-packing changes. Remove module parameter.] Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2013-02-13 18:54:41 +08:00
unsigned int recycle_count;
struct timer_list slow_fill;
unsigned int slow_fill_count;
};
enum efx_sync_events_state {
SYNC_EVENTS_DISABLED = 0,
SYNC_EVENTS_QUIESCENT,
SYNC_EVENTS_REQUESTED,
SYNC_EVENTS_VALID,
};
/**
* struct efx_channel - An Efx channel
*
* A channel comprises an event queue, at least one TX queue, at least
* one RX queue, and an associated tasklet for processing the event
* queue.
*
* @efx: Associated Efx NIC
* @channel: Channel instance number
* @type: Channel type definition
* @eventq_init: Event queue initialised flag
* @enabled: Channel enabled indicator
* @irq: IRQ number (MSI and MSI-X only)
* @irq_moderation: IRQ moderation value (in hardware ticks)
* @napi_dev: Net device used with NAPI
* @napi_str: NAPI control structure
* @eventq: Event queue buffer
* @eventq_mask: Event queue pointer mask
* @eventq_read_ptr: Event queue read pointer
* @event_test_cpu: Last CPU to handle interrupt or test event for this channel
* @irq_count: Number of IRQs since last adaptive moderation decision
* @irq_mod_score: IRQ moderation score
* @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
* @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
* @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
* @n_rx_mcast_mismatch: Count of unmatched multicast frames
* @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
* @n_rx_overlength: Count of RX_OVERLENGTH errors
* @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
* @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
* lack of descriptors
* @n_rx_merge_events: Number of RX merged completion events
* @n_rx_merge_packets: Number of RX packets completed by merged events
* @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
* __efx_rx_packet(), or zero if there is none
* @rx_pkt_index: Ring index of first buffer for next packet to be delivered
* by __efx_rx_packet(), if @rx_pkt_n_frags != 0
* @rx_queue: RX queue for this channel
* @tx_queue: TX queues for this channel
* @sync_events_state: Current state of sync events on this channel
* @sync_timestamp_major: Major part of the last ptp sync event
* @sync_timestamp_minor: Minor part of the last ptp sync event
*/
struct efx_channel {
struct efx_nic *efx;
int channel;
const struct efx_channel_type *type;
bool eventq_init;
bool enabled;
int irq;
unsigned int irq_moderation;
struct net_device *napi_dev;
struct napi_struct napi_str;
struct efx_special_buffer eventq;
unsigned int eventq_mask;
unsigned int eventq_read_ptr;
int event_test_cpu;
unsigned int irq_count;
unsigned int irq_mod_score;
#ifdef CONFIG_RFS_ACCEL
unsigned int rfs_filters_added;
#endif
unsigned n_rx_tobe_disc;
unsigned n_rx_ip_hdr_chksum_err;
unsigned n_rx_tcp_udp_chksum_err;
unsigned n_rx_mcast_mismatch;
unsigned n_rx_frm_trunc;
unsigned n_rx_overlength;
unsigned n_skbuff_leaks;
unsigned int n_rx_nodesc_trunc;
unsigned int n_rx_merge_events;
unsigned int n_rx_merge_packets;
unsigned int rx_pkt_n_frags;
unsigned int rx_pkt_index;
struct efx_rx_queue rx_queue;
struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
enum efx_sync_events_state sync_events_state;
u32 sync_timestamp_major;
u32 sync_timestamp_minor;
};
/**
* struct efx_msi_context - Context for each MSI
* @efx: The associated NIC
* @index: Index of the channel/IRQ
* @name: Name of the channel/IRQ
*
* Unlike &struct efx_channel, this is never reallocated and is always
* safe for the IRQ handler to access.
*/
struct efx_msi_context {
struct efx_nic *efx;
unsigned int index;
char name[IFNAMSIZ + 6];
};
/**
* struct efx_channel_type - distinguishes traffic and extra channels
* @handle_no_channel: Handle failure to allocate an extra channel
* @pre_probe: Set up extra state prior to initialisation
* @post_remove: Tear down extra state after finalisation, if allocated.
* May be called on channels that have not been probed.
* @get_name: Generate the channel's name (used for its IRQ handler)
* @copy: Copy the channel state prior to reallocation. May be %NULL if
* reallocation is not supported.
* @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
* @keep_eventq: Flag for whether event queue should be kept initialised
* while the device is stopped
*/
struct efx_channel_type {
void (*handle_no_channel)(struct efx_nic *);
int (*pre_probe)(struct efx_channel *);
void (*post_remove)(struct efx_channel *);
void (*get_name)(struct efx_channel *, char *buf, size_t len);
struct efx_channel *(*copy)(const struct efx_channel *);
bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
bool keep_eventq;
};
enum efx_led_mode {
EFX_LED_OFF = 0,
EFX_LED_ON = 1,
EFX_LED_DEFAULT = 2
};
#define STRING_TABLE_LOOKUP(val, member) \
((val) < member ## _max) ? member ## _names[val] : "(invalid)"
extern const char *const efx_loopback_mode_names[];
extern const unsigned int efx_loopback_mode_max;
#define LOOPBACK_MODE(efx) \
STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
extern const char *const efx_reset_type_names[];
extern const unsigned int efx_reset_type_max;
#define RESET_TYPE(type) \
STRING_TABLE_LOOKUP(type, efx_reset_type)
enum efx_int_mode {
/* Be careful if altering to correct macro below */
EFX_INT_MODE_MSIX = 0,
EFX_INT_MODE_MSI = 1,
EFX_INT_MODE_LEGACY = 2,
EFX_INT_MODE_MAX /* Insert any new items before this */
};
#define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
enum nic_state {
STATE_UNINIT = 0, /* device being probed/removed or is frozen */
STATE_READY = 1, /* hardware ready and netdev registered */
STATE_DISABLED = 2, /* device disabled due to hardware errors */
STATE_RECOVERY = 3, /* device recovering from PCI error */
};
/* Forward declaration */
struct efx_nic;
/* Pseudo bit-mask flow control field */
#define EFX_FC_RX FLOW_CTRL_RX
#define EFX_FC_TX FLOW_CTRL_TX
#define EFX_FC_AUTO 4
/**
* struct efx_link_state - Current state of the link
* @up: Link is up
* @fd: Link is full-duplex
* @fc: Actual flow control flags
* @speed: Link speed (Mbps)
*/
struct efx_link_state {
bool up;
bool fd;
u8 fc;
unsigned int speed;
};
static inline bool efx_link_state_equal(const struct efx_link_state *left,
const struct efx_link_state *right)
{
return left->up == right->up && left->fd == right->fd &&
left->fc == right->fc && left->speed == right->speed;
}
/**
* struct efx_phy_operations - Efx PHY operations table
* @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
* efx->loopback_modes.
* @init: Initialise PHY
* @fini: Shut down PHY
* @reconfigure: Reconfigure PHY (e.g. for new link parameters)
* @poll: Update @link_state and report whether it changed.
* Serialised by the mac_lock.
* @get_settings: Get ethtool settings. Serialised by the mac_lock.
* @set_settings: Set ethtool settings. Serialised by the mac_lock.
* @set_npage_adv: Set abilities advertised in (Extended) Next Page
* (only needed where AN bit is set in mmds)
* @test_alive: Test that PHY is 'alive' (online)
* @test_name: Get the name of a PHY-specific test/result
* @run_tests: Run tests and record results as appropriate (offline).
* Flags are the ethtool tests flags.
*/
struct efx_phy_operations {
int (*probe) (struct efx_nic *efx);
int (*init) (struct efx_nic *efx);
void (*fini) (struct efx_nic *efx);
void (*remove) (struct efx_nic *efx);
int (*reconfigure) (struct efx_nic *efx);
bool (*poll) (struct efx_nic *efx);
void (*get_settings) (struct efx_nic *efx,
struct ethtool_cmd *ecmd);
int (*set_settings) (struct efx_nic *efx,
struct ethtool_cmd *ecmd);
void (*set_npage_adv) (struct efx_nic *efx, u32);
int (*test_alive) (struct efx_nic *efx);
const char *(*test_name) (struct efx_nic *efx, unsigned int index);
int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
int (*get_module_eeprom) (struct efx_nic *efx,
struct ethtool_eeprom *ee,
u8 *data);
int (*get_module_info) (struct efx_nic *efx,
struct ethtool_modinfo *modinfo);
};
/**
* enum efx_phy_mode - PHY operating mode flags
* @PHY_MODE_NORMAL: on and should pass traffic
* @PHY_MODE_TX_DISABLED: on with TX disabled
* @PHY_MODE_LOW_POWER: set to low power through MDIO
* @PHY_MODE_OFF: switched off through external control
* @PHY_MODE_SPECIAL: on but will not pass traffic
*/
enum efx_phy_mode {
PHY_MODE_NORMAL = 0,
PHY_MODE_TX_DISABLED = 1,
PHY_MODE_LOW_POWER = 2,
PHY_MODE_OFF = 4,
PHY_MODE_SPECIAL = 8,
};
static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
{
return !!(mode & ~PHY_MODE_TX_DISABLED);
}
/**
* struct efx_hw_stat_desc - Description of a hardware statistic
* @name: Name of the statistic as visible through ethtool, or %NULL if
* it should not be exposed
* @dma_width: Width in bits (0 for non-DMA statistics)
* @offset: Offset within stats (ignored for non-DMA statistics)
*/
struct efx_hw_stat_desc {
const char *name;
u16 dma_width;
u16 offset;
};
/* Number of bits used in a multicast filter hash address */
#define EFX_MCAST_HASH_BITS 8
/* Number of (single-bit) entries in a multicast filter hash */
#define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
/* An Efx multicast filter hash */
union efx_multicast_hash {
u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
};
struct efx_vf;
struct vfdi_status;
/**
* struct efx_nic - an Efx NIC
* @name: Device name (net device name or bus id before net device registered)
* @pci_dev: The PCI device
* @node: List node for maintaning primary/secondary function lists
* @primary: &struct efx_nic instance for the primary function of this
* controller. May be the same structure, and may be %NULL if no
* primary function is bound. Serialised by rtnl_lock.
* @secondary_list: List of &struct efx_nic instances for the secondary PCI
* functions of the controller, if this is for the primary function.
* Serialised by rtnl_lock.
* @type: Controller type attributes
* @legacy_irq: IRQ number
* @workqueue: Workqueue for port reconfigures and the HW monitor.
* Work items do not hold and must not acquire RTNL.
* @workqueue_name: Name of workqueue
* @reset_work: Scheduled reset workitem
* @membase_phys: Memory BAR value as physical address
* @membase: Memory BAR value
* @interrupt_mode: Interrupt mode
* @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
* @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
* @irq_rx_moderation: IRQ moderation time for RX event queues
* @msg_enable: Log message enable flags
* @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
* @reset_pending: Bitmask for pending resets
* @tx_queue: TX DMA queues
* @rx_queue: RX DMA queues
* @channel: Channels
* @msi_context: Context for each MSI
* @extra_channel_types: Types of extra (non-traffic) channels that
* should be allocated for this NIC
* @rxq_entries: Size of receive queues requested by user.
* @txq_entries: Size of transmit queues requested by user.
* @txq_stop_thresh: TX queue fill level at or above which we stop it.
* @txq_wake_thresh: TX queue fill level at or below which we wake it.
* @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
* @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
* @sram_lim_qw: Qword address limit of SRAM
* @next_buffer_table: First available buffer table id
* @n_channels: Number of channels in use
* @n_rx_channels: Number of channels used for RX (= number of RX queues)
* @n_tx_channels: Number of channels used for TX
* @rx_ip_align: RX DMA address offset to have IP header aligned in
* in accordance with NET_IP_ALIGN
* @rx_dma_len: Current maximum RX DMA length
* @rx_buffer_order: Order (log2) of number of pages for each RX buffer
* @rx_buffer_truesize: Amortised allocation size of an RX buffer,
* for use in sk_buff::truesize
* @rx_prefix_size: Size of RX prefix before packet data
* @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
* (valid only if @rx_prefix_size != 0; always negative)
* @rx_packet_len_offset: Offset of RX packet length from start of packet data
* (valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
* @rx_packet_ts_offset: Offset of timestamp from start of packet data
* (valid only if channel->sync_timestamps_enabled; always negative)
* @rx_hash_key: Toeplitz hash key for RSS
* @rx_indir_table: Indirection table for RSS
* @rx_scatter: Scatter mode enabled for receives
* @int_error_count: Number of internal errors seen recently
* @int_error_expire: Time at which error count will be expired
* @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
* acknowledge but do nothing else.
* @irq_status: Interrupt status buffer
* @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
* @irq_level: IRQ level/index for IRQs not triggered by an event queue
* @selftest_work: Work item for asynchronous self-test
* @mtd_list: List of MTDs attached to the NIC
* @nic_data: Hardware dependent state
* @mcdi: Management-Controller-to-Driver Interface state
* @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
* efx_monitor() and efx_reconfigure_port()
* @port_enabled: Port enabled indicator.
* Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
* efx_mac_work() with kernel interfaces. Safe to read under any
* one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
* be held to modify it.
* @port_initialized: Port initialized?
* @net_dev: Operating system network device. Consider holding the rtnl lock
* @stats_buffer: DMA buffer for statistics
* @phy_type: PHY type
* @phy_op: PHY interface
* @phy_data: PHY private data (including PHY-specific stats)
* @mdio: PHY MDIO interface
* @mdio_bus: PHY MDIO bus ID (only used by Siena)
* @phy_mode: PHY operating mode. Serialised by @mac_lock.
* @link_advertising: Autonegotiation advertising flags
* @link_state: Current state of the link
* @n_link_state_changes: Number of times the link has changed state
* @unicast_filter: Flag for Falcon-arch simple unicast filter.
* Protected by @mac_lock.
* @multicast_hash: Multicast hash table for Falcon-arch.
* Protected by @mac_lock.
* @wanted_fc: Wanted flow control flags
* @fc_disable: When non-zero flow control is disabled. Typically used to
* ensure that network back pressure doesn't delay dma queue flushes.
* Serialised by the rtnl lock.
* @mac_work: Work item for changing MAC promiscuity and multicast hash
* @loopback_mode: Loopback status
* @loopback_modes: Supported loopback mode bitmask
* @loopback_selftest: Offline self-test private state
* @filter_lock: Filter table lock
* @filter_state: Architecture-dependent filter table state
* @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
* indexed by filter ID
* @rps_expire_index: Next index to check for expiry in @rps_flow_id
* @active_queues: Count of RX and TX queues that haven't been flushed and drained.
* @rxq_flush_pending: Count of number of receive queues that need to be flushed.
* Decremented when the efx_flush_rx_queue() is called.
* @rxq_flush_outstanding: Count of number of RX flushes started but not yet
* completed (either success or failure). Not used when MCDI is used to
* flush receive queues.
* @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
* @vf: Array of &struct efx_vf objects.
* @vf_count: Number of VFs intended to be enabled.
* @vf_init_count: Number of VFs that have been fully initialised.
* @vi_scale: log2 number of vnics per VF.
* @vf_buftbl_base: The zeroth buffer table index used to back VF queues.
* @vfdi_status: Common VFDI status page to be dmad to VF address space.
* @local_addr_list: List of local addresses. Protected by %local_lock.
* @local_page_list: List of DMA addressable pages used to broadcast
* %local_addr_list. Protected by %local_lock.
* @local_lock: Mutex protecting %local_addr_list and %local_page_list.
* @peer_work: Work item to broadcast peer addresses to VMs.
* @ptp_data: PTP state data
* @vpd_sn: Serial number read from VPD
* @monitor_work: Hardware monitor workitem
* @biu_lock: BIU (bus interface unit) lock
* @last_irq_cpu: Last CPU to handle a possible test interrupt. This
* field is used by efx_test_interrupts() to verify that an
* interrupt has occurred.
* @stats_lock: Statistics update lock. Must be held when calling
* efx_nic_type::{update,start,stop}_stats.
*
* This is stored in the private area of the &struct net_device.
*/
struct efx_nic {
/* The following fields should be written very rarely */
char name[IFNAMSIZ];
struct list_head node;
struct efx_nic *primary;
struct list_head secondary_list;
struct pci_dev *pci_dev;
unsigned int port_num;
const struct efx_nic_type *type;
int legacy_irq;
bool eeh_disabled_legacy_irq;
struct workqueue_struct *workqueue;
char workqueue_name[16];
struct work_struct reset_work;
resource_size_t membase_phys;
void __iomem *membase;
enum efx_int_mode interrupt_mode;
unsigned int timer_quantum_ns;
bool irq_rx_adaptive;
unsigned int irq_rx_moderation;
u32 msg_enable;
enum nic_state state;
unsigned long reset_pending;
struct efx_channel *channel[EFX_MAX_CHANNELS];
struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
const struct efx_channel_type *
extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
unsigned rxq_entries;
unsigned txq_entries;
unsigned int txq_stop_thresh;
unsigned int txq_wake_thresh;
unsigned tx_dc_base;
unsigned rx_dc_base;
unsigned sram_lim_qw;
unsigned next_buffer_table;
unsigned int max_channels;
unsigned n_channels;
unsigned n_rx_channels;
unsigned rss_spread;
unsigned tx_channel_offset;
unsigned n_tx_channels;
unsigned int rx_ip_align;
unsigned int rx_dma_len;
unsigned int rx_buffer_order;
unsigned int rx_buffer_truesize;
unsigned int rx_page_buf_step;
sfc: reuse pages to avoid DMA mapping/unmapping costs On POWER systems, DMA mapping/unmapping operations are very expensive. These changes reduce these costs by trying to reuse DMA mapped pages. After all the buffers associated with a page have been processed and passed up, the page is placed into a ring (if there is room). For each page that is required for a refill operation, a page in the ring is examined to determine if its page count has fallen to 1, ie. the kernel has released its reference to these packets. If this is the case, the page can be immediately added back into the RX descriptor ring, without having to re-map it for DMA. If the kernel is still holding a reference to this page, it is removed from the ring and unmapped for DMA. Then a new page, which can immediately be used by RX buffers in the descriptor ring, is allocated and DMA mapped. The time a page needs to spend in the recycle ring before the kernel has released its page references is based on the number of buffers that use this page. As large pages can hold more RX buffers, the RX recycle ring can be shorter. This reduces memory usage on POWER systems, while maintaining the performance gain achieved by recycling pages, following the driver change to pack more than two RX buffers into large pages. When an IOMMU is not present, the recycle ring can be small to reduce memory usage, since DMA mapping operations are inexpensive. With a small recycle ring, attempting to refill the descriptor queue with more buffers than the equivalent size of the recycle ring could ultimately lead to memory leaks if page entries in the recycle ring were overwritten. To prevent this, the check to see if the recycle ring is full is changed to check if the next entry to be written is NULL. [bwh: Combine and rebase several commits so this is complete before the following buffer-packing changes. Remove module parameter.] Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
2013-02-13 18:54:41 +08:00
unsigned int rx_bufs_per_page;
unsigned int rx_pages_per_batch;
unsigned int rx_prefix_size;
int rx_packet_hash_offset;
int rx_packet_len_offset;
int rx_packet_ts_offset;
u8 rx_hash_key[40];
u32 rx_indir_table[128];
bool rx_scatter;
unsigned int_error_count;
unsigned long int_error_expire;
bool irq_soft_enabled;
struct efx_buffer irq_status;
unsigned irq_zero_count;
unsigned irq_level;
struct delayed_work selftest_work;
#ifdef CONFIG_SFC_MTD
struct list_head mtd_list;
#endif
void *nic_data;
struct efx_mcdi_data *mcdi;
struct mutex mac_lock;
struct work_struct mac_work;
bool port_enabled;
bool mc_bist_for_other_fn;
bool port_initialized;
struct net_device *net_dev;
struct efx_buffer stats_buffer;
u64 rx_nodesc_drops_total;
u64 rx_nodesc_drops_while_down;
bool rx_nodesc_drops_prev_state;
unsigned int phy_type;
const struct efx_phy_operations *phy_op;
void *phy_data;
struct mdio_if_info mdio;
unsigned int mdio_bus;
enum efx_phy_mode phy_mode;
u32 link_advertising;
struct efx_link_state link_state;
unsigned int n_link_state_changes;
bool unicast_filter;
union efx_multicast_hash multicast_hash;
u8 wanted_fc;
unsigned fc_disable;
atomic_t rx_reset;
enum efx_loopback_mode loopback_mode;
u64 loopback_modes;
void *loopback_selftest;
spinlock_t filter_lock;
void *filter_state;
#ifdef CONFIG_RFS_ACCEL
u32 *rps_flow_id;
unsigned int rps_expire_index;
#endif
atomic_t active_queues;
atomic_t rxq_flush_pending;
atomic_t rxq_flush_outstanding;
wait_queue_head_t flush_wq;
#ifdef CONFIG_SFC_SRIOV
struct efx_channel *vfdi_channel;
struct efx_vf *vf;
unsigned vf_count;
unsigned vf_init_count;
unsigned vi_scale;
unsigned vf_buftbl_base;
struct efx_buffer vfdi_status;
struct list_head local_addr_list;
struct list_head local_page_list;
struct mutex local_lock;
struct work_struct peer_work;
#endif
struct efx_ptp_data *ptp_data;
char *vpd_sn;
/* The following fields may be written more often */
struct delayed_work monitor_work ____cacheline_aligned_in_smp;
spinlock_t biu_lock;
int last_irq_cpu;
spinlock_t stats_lock;
};
static inline int efx_dev_registered(struct efx_nic *efx)
{
return efx->net_dev->reg_state == NETREG_REGISTERED;
}
static inline unsigned int efx_port_num(struct efx_nic *efx)
{
return efx->port_num;
}
struct efx_mtd_partition {
struct list_head node;
struct mtd_info mtd;
const char *dev_type_name;
const char *type_name;
char name[IFNAMSIZ + 20];
};
/**
* struct efx_nic_type - Efx device type definition
* @mem_map_size: Get memory BAR mapped size
* @probe: Probe the controller
* @remove: Free resources allocated by probe()
* @init: Initialise the controller
* @dimension_resources: Dimension controller resources (buffer table,
* and VIs once the available interrupt resources are clear)
* @fini: Shut down the controller
* @monitor: Periodic function for polling link state and hardware monitor
* @map_reset_reason: Map ethtool reset reason to a reset method
* @map_reset_flags: Map ethtool reset flags to a reset method, if possible
* @reset: Reset the controller hardware and possibly the PHY. This will
* be called while the controller is uninitialised.
* @probe_port: Probe the MAC and PHY
* @remove_port: Free resources allocated by probe_port()
* @handle_global_event: Handle a "global" event (may be %NULL)
* @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
* @prepare_flush: Prepare the hardware for flushing the DMA queues
* (for Falcon architecture)
* @finish_flush: Clean up after flushing the DMA queues (for Falcon
* architecture)
* @describe_stats: Describe statistics for ethtool
* @update_stats: Update statistics not provided by event handling.
* Either argument may be %NULL.
* @start_stats: Start the regular fetching of statistics
* @pull_stats: Pull stats from the NIC and wait until they arrive.
* @stop_stats: Stop the regular fetching of statistics
* @set_id_led: Set state of identifying LED or revert to automatic function
* @push_irq_moderation: Apply interrupt moderation value
* @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
* @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
* @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
* to the hardware. Serialised by the mac_lock.
* @check_mac_fault: Check MAC fault state. True if fault present.
* @get_wol: Get WoL configuration from driver state
* @set_wol: Push WoL configuration to the NIC
* @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
* @test_chip: Test registers. May use efx_farch_test_registers(), and is
* expected to reset the NIC.
* @test_nvram: Test validity of NVRAM contents
* @mcdi_request: Send an MCDI request with the given header and SDU.
* The SDU length may be any value from 0 up to the protocol-
* defined maximum, but its buffer will be padded to a multiple
* of 4 bytes.
* @mcdi_poll_response: Test whether an MCDI response is available.
* @mcdi_read_response: Read the MCDI response PDU. The offset will
* be a multiple of 4. The length may not be, but the buffer
* will be padded so it is safe to round up.
* @mcdi_poll_reboot: Test whether the MCDI has rebooted. If so,
* return an appropriate error code for aborting any current
* request; otherwise return 0.
* @irq_enable_master: Enable IRQs on the NIC. Each event queue must
* be separately enabled after this.
* @irq_test_generate: Generate a test IRQ
* @irq_disable_non_ev: Disable non-event IRQs on the NIC. Each event
* queue must be separately disabled before this.
* @irq_handle_msi: Handle MSI for a channel. The @dev_id argument is
* a pointer to the &struct efx_msi_context for the channel.
* @irq_handle_legacy: Handle legacy interrupt. The @dev_id argument
* is a pointer to the &struct efx_nic.
* @tx_probe: Allocate resources for TX queue
* @tx_init: Initialise TX queue on the NIC
* @tx_remove: Free resources for TX queue
* @tx_write: Write TX descriptors and doorbell
* @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
* @rx_probe: Allocate resources for RX queue
* @rx_init: Initialise RX queue on the NIC
* @rx_remove: Free resources for RX queue
* @rx_write: Write RX descriptors and doorbell
* @rx_defer_refill: Generate a refill reminder event
* @ev_probe: Allocate resources for event queue
* @ev_init: Initialise event queue on the NIC
* @ev_fini: Deinitialise event queue on the NIC
* @ev_remove: Free resources for event queue
* @ev_process: Process events for a queue, up to the given NAPI quota
* @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
* @ev_test_generate: Generate a test event
* @filter_table_probe: Probe filter capabilities and set up filter software state
* @filter_table_restore: Restore filters removed from hardware
* @filter_table_remove: Remove filters from hardware and tear down software state
* @filter_update_rx_scatter: Update filters after change to rx scatter setting
* @filter_insert: add or replace a filter
* @filter_remove_safe: remove a filter by ID, carefully
* @filter_get_safe: retrieve a filter by ID, carefully
* @filter_clear_rx: Remove all RX filters whose priority is less than or
* equal to the given priority and is not %EFX_FILTER_PRI_AUTO
* @filter_count_rx_used: Get the number of filters in use at a given priority
* @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
* @filter_get_rx_ids: Get list of RX filters at a given priority
* @filter_rfs_insert: Add or replace a filter for RFS. This must be
* atomic. The hardware change may be asynchronous but should
* not be delayed for long. It may fail if this can't be done
* atomically.
* @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
* This must check whether the specified table entry is used by RFS
* and that rps_may_expire_flow() returns true for it.
* @mtd_probe: Probe and add MTD partitions associated with this net device,
* using efx_mtd_add()
* @mtd_rename: Set an MTD partition name using the net device name
* @mtd_read: Read from an MTD partition
* @mtd_erase: Erase part of an MTD partition
* @mtd_write: Write to an MTD partition
* @mtd_sync: Wait for write-back to complete on MTD partition. This
* also notifies the driver that a writer has finished using this
* partition.
* @ptp_write_host_time: Send host time to MC as part of sync protocol
* @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
* timestamping, possibly only temporarily for the purposes of a reset.
* @ptp_set_ts_config: Set hardware timestamp configuration. The flags
* and tx_type will already have been validated but this operation
* must validate and update rx_filter.
* @revision: Hardware architecture revision
* @txd_ptr_tbl_base: TX descriptor ring base address
* @rxd_ptr_tbl_base: RX descriptor ring base address
* @buf_tbl_base: Buffer table base address
* @evq_ptr_tbl_base: Event queue pointer table base address
* @evq_rptr_tbl_base: Event queue read-pointer table base address
* @max_dma_mask: Maximum possible DMA mask
* @rx_prefix_size: Size of RX prefix before packet data
* @rx_hash_offset: Offset of RX flow hash within prefix
* @rx_ts_offset: Offset of timestamp within prefix
* @rx_buffer_padding: Size of padding at end of RX packet
* @can_rx_scatter: NIC is able to scatter packets to multiple buffers
* @always_rx_scatter: NIC will always scatter packets to multiple buffers
* @max_interrupt_mode: Highest capability interrupt mode supported
* from &enum efx_init_mode.
* @timer_period_max: Maximum period of interrupt timer (in ticks)
* @offload_features: net_device feature flags for protocol offload
* features implemented in hardware
* @mcdi_max_ver: Maximum MCDI version supported
* @hwtstamp_filters: Mask of hardware timestamp filter types supported
*/
struct efx_nic_type {
unsigned int (*mem_map_size)(struct efx_nic *efx);
int (*probe)(struct efx_nic *efx);
void (*remove)(struct efx_nic *efx);
int (*init)(struct efx_nic *efx);
int (*dimension_resources)(struct efx_nic *efx);
void (*fini)(struct efx_nic *efx);
void (*monitor)(struct efx_nic *efx);
enum reset_type (*map_reset_reason)(enum reset_type reason);
int (*map_reset_flags)(u32 *flags);
int (*reset)(struct efx_nic *efx, enum reset_type method);
int (*probe_port)(struct efx_nic *efx);
void (*remove_port)(struct efx_nic *efx);
bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
int (*fini_dmaq)(struct efx_nic *efx);
void (*prepare_flush)(struct efx_nic *efx);
void (*finish_flush)(struct efx_nic *efx);
size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
struct rtnl_link_stats64 *core_stats);
void (*start_stats)(struct efx_nic *efx);
void (*pull_stats)(struct efx_nic *efx);
void (*stop_stats)(struct efx_nic *efx);
void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
void (*push_irq_moderation)(struct efx_channel *channel);
int (*reconfigure_port)(struct efx_nic *efx);
void (*prepare_enable_fc_tx)(struct efx_nic *efx);
int (*reconfigure_mac)(struct efx_nic *efx);
bool (*check_mac_fault)(struct efx_nic *efx);
void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
int (*set_wol)(struct efx_nic *efx, u32 type);
void (*resume_wol)(struct efx_nic *efx);
int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
int (*test_nvram)(struct efx_nic *efx);
void (*mcdi_request)(struct efx_nic *efx,
const efx_dword_t *hdr, size_t hdr_len,
const efx_dword_t *sdu, size_t sdu_len);
bool (*mcdi_poll_response)(struct efx_nic *efx);
void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
size_t pdu_offset, size_t pdu_len);
int (*mcdi_poll_reboot)(struct efx_nic *efx);
void (*irq_enable_master)(struct efx_nic *efx);
void (*irq_test_generate)(struct efx_nic *efx);
void (*irq_disable_non_ev)(struct efx_nic *efx);
irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
int (*tx_probe)(struct efx_tx_queue *tx_queue);
void (*tx_init)(struct efx_tx_queue *tx_queue);
void (*tx_remove)(struct efx_tx_queue *tx_queue);
void (*tx_write)(struct efx_tx_queue *tx_queue);
void (*rx_push_rss_config)(struct efx_nic *efx);
int (*rx_probe)(struct efx_rx_queue *rx_queue);
void (*rx_init)(struct efx_rx_queue *rx_queue);
void (*rx_remove)(struct efx_rx_queue *rx_queue);
void (*rx_write)(struct efx_rx_queue *rx_queue);
void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
int (*ev_probe)(struct efx_channel *channel);
int (*ev_init)(struct efx_channel *channel);
void (*ev_fini)(struct efx_channel *channel);
void (*ev_remove)(struct efx_channel *channel);
int (*ev_process)(struct efx_channel *channel, int quota);
void (*ev_read_ack)(struct efx_channel *channel);
void (*ev_test_generate)(struct efx_channel *channel);
int (*filter_table_probe)(struct efx_nic *efx);
void (*filter_table_restore)(struct efx_nic *efx);
void (*filter_table_remove)(struct efx_nic *efx);
void (*filter_update_rx_scatter)(struct efx_nic *efx);
s32 (*filter_insert)(struct efx_nic *efx,
struct efx_filter_spec *spec, bool replace);
int (*filter_remove_safe)(struct efx_nic *efx,
enum efx_filter_priority priority,
u32 filter_id);
int (*filter_get_safe)(struct efx_nic *efx,
enum efx_filter_priority priority,
u32 filter_id, struct efx_filter_spec *);
int (*filter_clear_rx)(struct efx_nic *efx,
enum efx_filter_priority priority);
u32 (*filter_count_rx_used)(struct efx_nic *efx,
enum efx_filter_priority priority);
u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
s32 (*filter_get_rx_ids)(struct efx_nic *efx,
enum efx_filter_priority priority,
u32 *buf, u32 size);
#ifdef CONFIG_RFS_ACCEL
s32 (*filter_rfs_insert)(struct efx_nic *efx,
struct efx_filter_spec *spec);
bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
unsigned int index);
#endif
#ifdef CONFIG_SFC_MTD
int (*mtd_probe)(struct efx_nic *efx);
void (*mtd_rename)(struct efx_mtd_partition *part);
int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
size_t *retlen, u8 *buffer);
int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
size_t *retlen, const u8 *buffer);
int (*mtd_sync)(struct mtd_info *mtd);
#endif
void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
int (*ptp_set_ts_config)(struct efx_nic *efx,
struct hwtstamp_config *init);
int revision;
unsigned int txd_ptr_tbl_base;
unsigned int rxd_ptr_tbl_base;
unsigned int buf_tbl_base;
unsigned int evq_ptr_tbl_base;
unsigned int evq_rptr_tbl_base;
u64 max_dma_mask;
unsigned int rx_prefix_size;
unsigned int rx_hash_offset;
unsigned int rx_ts_offset;
unsigned int rx_buffer_padding;
bool can_rx_scatter;
bool always_rx_scatter;
unsigned int max_interrupt_mode;
unsigned int timer_period_max;
netdev_features_t offload_features;
int mcdi_max_ver;
unsigned int max_rx_ip_filters;
u32 hwtstamp_filters;
};
/**************************************************************************
*
* Prototypes and inline functions
*
*************************************************************************/
static inline struct efx_channel *
efx_get_channel(struct efx_nic *efx, unsigned index)
{
EFX_BUG_ON_PARANOID(index >= efx->n_channels);
return efx->channel[index];
}
/* Iterate over all used channels */
#define efx_for_each_channel(_channel, _efx) \
for (_channel = (_efx)->channel[0]; \
_channel; \
_channel = (_channel->channel + 1 < (_efx)->n_channels) ? \
(_efx)->channel[_channel->channel + 1] : NULL)
/* Iterate over all used channels in reverse */
#define efx_for_each_channel_rev(_channel, _efx) \
for (_channel = (_efx)->channel[(_efx)->n_channels - 1]; \
_channel; \
_channel = _channel->channel ? \
(_efx)->channel[_channel->channel - 1] : NULL)
static inline struct efx_tx_queue *
efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
{
EFX_BUG_ON_PARANOID(index >= efx->n_tx_channels ||
type >= EFX_TXQ_TYPES);
return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
}
static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
{
return channel->channel - channel->efx->tx_channel_offset <
channel->efx->n_tx_channels;
}
static inline struct efx_tx_queue *
efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
{
EFX_BUG_ON_PARANOID(!efx_channel_has_tx_queues(channel) ||
type >= EFX_TXQ_TYPES);
return &channel->tx_queue[type];
}
static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
{
return !(tx_queue->efx->net_dev->num_tc < 2 &&
tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
}
/* Iterate over all TX queues belonging to a channel */
#define efx_for_each_channel_tx_queue(_tx_queue, _channel) \
if (!efx_channel_has_tx_queues(_channel)) \
; \
else \
for (_tx_queue = (_channel)->tx_queue; \
_tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
efx_tx_queue_used(_tx_queue); \
_tx_queue++)
/* Iterate over all possible TX queues belonging to a channel */
#define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel) \
if (!efx_channel_has_tx_queues(_channel)) \
; \
else \
for (_tx_queue = (_channel)->tx_queue; \
_tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES; \
_tx_queue++)
static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
{
return channel->rx_queue.core_index >= 0;
}
static inline struct efx_rx_queue *
efx_channel_get_rx_queue(struct efx_channel *channel)
{
EFX_BUG_ON_PARANOID(!efx_channel_has_rx_queue(channel));
return &channel->rx_queue;
}
/* Iterate over all RX queues belonging to a channel */
#define efx_for_each_channel_rx_queue(_rx_queue, _channel) \
if (!efx_channel_has_rx_queue(_channel)) \
; \
else \
for (_rx_queue = &(_channel)->rx_queue; \
_rx_queue; \
_rx_queue = NULL)
static inline struct efx_channel *
efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
{
return container_of(rx_queue, struct efx_channel, rx_queue);
}
static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
{
return efx_rx_queue_channel(rx_queue)->channel;
}
/* Returns a pointer to the specified receive buffer in the RX
* descriptor queue.
*/
static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
unsigned int index)
{
return &rx_queue->buffer[index];
}
/**
* EFX_MAX_FRAME_LEN - calculate maximum frame length
*
* This calculates the maximum frame length that will be used for a
* given MTU. The frame length will be equal to the MTU plus a
* constant amount of header space and padding. This is the quantity
* that the net driver will program into the MAC as the maximum frame
* length.
*
* The 10G MAC requires 8-byte alignment on the frame
* length, so we round up to the nearest 8.
*
* Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
* XGMII cycle). If the frame length reaches the maximum value in the
* same cycle, the XMAC can miss the IPG altogether. We work around
* this by adding a further 16 bytes.
*/
#define EFX_MAX_FRAME_LEN(mtu) \
((((mtu) + ETH_HLEN + VLAN_HLEN + 4/* FCS */ + 7) & ~7) + 16)
static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
{
return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
}
static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
{
skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
}
#endif /* EFX_NET_DRIVER_H */