freebsd-skq/sys/dev/nxge/xgehal/xgehal-fifo.c
Robert Watson 3be4cb0b4a Merge Neterion if_nxge driver version 2.0.9.11230 with the following
changes:

  01 -  Enhanced LRO:
  LRO feature is extended to support multi-buffer mode. Previously,
  Ethernet frames received in contiguous buffers were offloaded.
  Now, frames received in multiple non-contiguous buffers can be
  offloaded, as well. The driver now supports LRO for jumbo frames.

  02 - Locks Optimization:
  The driver code was re-organized to limit the use of locks.
  Moreover, lock contention was reduced by replacing wait locks
  with try locks.

  03 - Code Optimization:
  The driver code was re-factored  to eliminate some memcpy
  operations.  Fast path loops were optimized.

  04 - Tag Creations:
  Physical Buffer Tags are now optimized based upon frame size.
  For better performance, Physical Memory Maps are now re-used.

  05 - Configuration:
  Features such as TSO, LRO, and Interrupt Mode can be configured
  either at load or at run time. Rx buffer mode (mode 1 or mode 2)
  can be configured at load time through kenv.

  06 - Driver Statistics:
  Run time statistics are enhanced to provide better visibility
  into the driver performance.

  07 - Bug Fixes:
  The driver contains fixes for the problems discovered and
  reported since last submission.

  08 - MSI support:
  Added Message Signaled Interrupt feature which currently uses 1
  message.

  09  Removed feature:
  Rx 3 buffer mode feature has been removed. Driver now supports 1,
  2 and 5 buffer modes of which 2 and 5 buffer modes can be used
  for header separation.

  10  Compiler warning:
  Fixed compiler warning when compiled for 32 bit system.

  11 Copyright notice:
  Source files are updated with the proper copyright notice.

MFC after:	3 days
Submitted by:	Alicia Pena <Alicia dot Pena at neterion dot com>,
		Muhammad Shafiq <Muhammad dot Shafiq at neterion dot com>
2007-10-29 14:19:32 +00:00

561 lines
18 KiB
C

/*-
* Copyright (c) 2002-2007 Neterion, Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <dev/nxge/include/xgehal-fifo.h>
#include <dev/nxge/include/xgehal-device.h>
static xge_hal_status_e
__hal_fifo_mempool_item_alloc(xge_hal_mempool_h mempoolh,
void *memblock,
int memblock_index,
xge_hal_mempool_dma_t *dma_object,
void *item,
int index,
int is_last,
void *userdata)
{
int memblock_item_idx;
xge_hal_fifo_txdl_priv_t *txdl_priv;
xge_hal_fifo_txd_t *txdp = (xge_hal_fifo_txd_t *)item;
xge_hal_fifo_t *fifo = (xge_hal_fifo_t *)userdata;
xge_assert(item);
txdl_priv = (xge_hal_fifo_txdl_priv_t *) \
__hal_mempool_item_priv((xge_hal_mempool_t *) mempoolh,
memblock_index,
item,
&memblock_item_idx);
xge_assert(txdl_priv);
/* pre-format HAL's TxDL's private */
txdl_priv->dma_offset = (char*)item - (char*)memblock;
txdl_priv->dma_addr = dma_object->addr + txdl_priv->dma_offset;
txdl_priv->dma_handle = dma_object->handle;
txdl_priv->memblock = memblock;
txdl_priv->first_txdp = (xge_hal_fifo_txd_t *)item;
txdl_priv->next_txdl_priv = NULL;
txdl_priv->dang_txdl = NULL;
txdl_priv->dang_frags = 0;
txdl_priv->alloc_frags = 0;
#ifdef XGE_DEBUG_ASSERT
txdl_priv->dma_object = dma_object;
#endif
txdp->host_control = (u64)(ulong_t)txdl_priv;
#ifdef XGE_HAL_ALIGN_XMIT
txdl_priv->align_vaddr = NULL;
txdl_priv->align_dma_addr = (dma_addr_t)0;
#ifndef XGE_HAL_ALIGN_XMIT_ALLOC_RT
{
xge_hal_status_e status;
if (fifo->config->alignment_size) {
status =__hal_fifo_dtr_align_alloc_map(fifo, txdp);
if (status != XGE_HAL_OK) {
xge_debug_mm(XGE_ERR,
"align buffer[%d] %d bytes, status %d",
index,
fifo->align_size,
status);
return status;
}
}
}
#endif
#endif
if (fifo->channel.dtr_init) {
fifo->channel.dtr_init(fifo, (xge_hal_dtr_h)txdp, index,
fifo->channel.userdata, XGE_HAL_CHANNEL_OC_NORMAL);
}
return XGE_HAL_OK;
}
static xge_hal_status_e
__hal_fifo_mempool_item_free(xge_hal_mempool_h mempoolh,
void *memblock,
int memblock_index,
xge_hal_mempool_dma_t *dma_object,
void *item,
int index,
int is_last,
void *userdata)
{
int memblock_item_idx;
xge_hal_fifo_txdl_priv_t *txdl_priv;
#ifdef XGE_HAL_ALIGN_XMIT
xge_hal_fifo_t *fifo = (xge_hal_fifo_t *)userdata;
#endif
xge_assert(item);
txdl_priv = (xge_hal_fifo_txdl_priv_t *) \
__hal_mempool_item_priv((xge_hal_mempool_t *) mempoolh,
memblock_index,
item,
&memblock_item_idx);
xge_assert(txdl_priv);
#ifdef XGE_HAL_ALIGN_XMIT
if (fifo->config->alignment_size) {
if (txdl_priv->align_dma_addr != 0) {
xge_os_dma_unmap(fifo->channel.pdev,
txdl_priv->align_dma_handle,
txdl_priv->align_dma_addr,
fifo->align_size,
XGE_OS_DMA_DIR_TODEVICE);
txdl_priv->align_dma_addr = 0;
}
if (txdl_priv->align_vaddr != NULL) {
xge_os_dma_free(fifo->channel.pdev,
txdl_priv->align_vaddr,
fifo->align_size,
&txdl_priv->align_dma_acch,
&txdl_priv->align_dma_handle);
txdl_priv->align_vaddr = NULL;
}
}
#endif
return XGE_HAL_OK;
}
xge_hal_status_e
__hal_fifo_open(xge_hal_channel_h channelh, xge_hal_channel_attr_t *attr)
{
xge_hal_device_t *hldev;
xge_hal_status_e status;
xge_hal_fifo_t *fifo = (xge_hal_fifo_t *)channelh;
xge_hal_fifo_queue_t *queue;
int i, txdl_size, max_arr_index, mid_point;
xge_hal_dtr_h dtrh;
hldev = (xge_hal_device_t *)fifo->channel.devh;
fifo->config = &hldev->config.fifo;
queue = &fifo->config->queue[attr->post_qid];
#if defined(XGE_HAL_TX_MULTI_RESERVE)
xge_os_spin_lock_init(&fifo->channel.reserve_lock, hldev->pdev);
#elif defined(XGE_HAL_TX_MULTI_RESERVE_IRQ)
xge_os_spin_lock_init_irq(&fifo->channel.reserve_lock, hldev->irqh);
#endif
#if defined(XGE_HAL_TX_MULTI_POST)
if (xge_hal_device_check_id(hldev) == XGE_HAL_CARD_XENA) {
fifo->post_lock_ptr = &hldev->xena_post_lock;
} else {
xge_os_spin_lock_init(&fifo->channel.post_lock, hldev->pdev);
fifo->post_lock_ptr = &fifo->channel.post_lock;
}
#elif defined(XGE_HAL_TX_MULTI_POST_IRQ)
if (xge_hal_device_check_id(hldev) == XGE_HAL_CARD_XENA) {
fifo->post_lock_ptr = &hldev->xena_post_lock;
} else {
xge_os_spin_lock_init_irq(&fifo->channel.post_lock,
hldev->irqh);
fifo->post_lock_ptr = &fifo->channel.post_lock;
}
#endif
fifo->align_size =
fifo->config->alignment_size * fifo->config->max_aligned_frags;
/* Initializing the BAR1 address as the start of
* the FIFO queue pointer and as a location of FIFO control
* word. */
fifo->hw_pair =
(xge_hal_fifo_hw_pair_t *) (void *)(hldev->bar1 +
(attr->post_qid * XGE_HAL_FIFO_HW_PAIR_OFFSET));
/* apply "interrupts per txdl" attribute */
fifo->interrupt_type = XGE_HAL_TXD_INT_TYPE_UTILZ;
if (queue->intr) {
fifo->interrupt_type = XGE_HAL_TXD_INT_TYPE_PER_LIST;
}
fifo->no_snoop_bits =
(int)(XGE_HAL_TX_FIFO_NO_SNOOP(queue->no_snoop_bits));
/*
* FIFO memory management strategy:
*
* TxDL splitted into three independent parts:
* - set of TxD's
* - TxD HAL private part
* - upper layer private part
*
* Adaptative memory allocation used. i.e. Memory allocated on
* demand with the size which will fit into one memory block.
* One memory block may contain more than one TxDL. In simple case
* memory block size can be equal to CPU page size. On more
* sophisticated OS's memory block can be contigious across
* several pages.
*
* During "reserve" operations more memory can be allocated on demand
* for example due to FIFO full condition.
*
* Pool of memory memblocks never shrinks except __hal_fifo_close
* routine which will essentially stop channel and free the resources.
*/
/* TxDL common private size == TxDL private + ULD private */
fifo->priv_size = sizeof(xge_hal_fifo_txdl_priv_t) +
attr->per_dtr_space;
fifo->priv_size = ((fifo->priv_size + __xge_os_cacheline_size -1) /
__xge_os_cacheline_size) *
__xge_os_cacheline_size;
/* recompute txdl size to be cacheline aligned */
fifo->txdl_size = fifo->config->max_frags * sizeof(xge_hal_fifo_txd_t);
txdl_size = ((fifo->txdl_size + __xge_os_cacheline_size - 1) /
__xge_os_cacheline_size) * __xge_os_cacheline_size;
if (fifo->txdl_size != txdl_size)
xge_debug_fifo(XGE_ERR, "cacheline > 128 ( ?? ): %d, %d, %d, %d",
fifo->config->max_frags, fifo->txdl_size, txdl_size,
__xge_os_cacheline_size);
fifo->txdl_size = txdl_size;
/* since dtr_init() callback will be called from item_alloc(),
* the same way channels userdata might be used prior to
* channel_initialize() */
fifo->channel.dtr_init = attr->dtr_init;
fifo->channel.userdata = attr->userdata;
fifo->txdl_per_memblock = fifo->config->memblock_size /
fifo->txdl_size;
fifo->mempool = __hal_mempool_create(hldev->pdev,
fifo->config->memblock_size,
fifo->txdl_size,
fifo->priv_size,
queue->initial,
queue->max,
__hal_fifo_mempool_item_alloc,
__hal_fifo_mempool_item_free,
fifo);
if (fifo->mempool == NULL) {
return XGE_HAL_ERR_OUT_OF_MEMORY;
}
status = __hal_channel_initialize(channelh, attr,
(void **) __hal_mempool_items_arr(fifo->mempool),
queue->initial, queue->max,
fifo->config->reserve_threshold);
if (status != XGE_HAL_OK) {
__hal_fifo_close(channelh);
return status;
}
xge_debug_fifo(XGE_TRACE,
"DTR reserve_length:%d reserve_top:%d\n"
"max_frags:%d reserve_threshold:%d\n"
"memblock_size:%d alignment_size:%d max_aligned_frags:%d",
fifo->channel.reserve_length, fifo->channel.reserve_top,
fifo->config->max_frags, fifo->config->reserve_threshold,
fifo->config->memblock_size, fifo->config->alignment_size,
fifo->config->max_aligned_frags);
#ifdef XGE_DEBUG_ASSERT
for ( i = 0; i < fifo->channel.reserve_length; i++) {
xge_debug_fifo(XGE_TRACE, "DTR before reversing index:%d"
" handle:%p", i, fifo->channel.reserve_arr[i]);
}
#endif
xge_assert(fifo->channel.reserve_length);
/* reverse the FIFO dtr array */
max_arr_index = fifo->channel.reserve_length - 1;
max_arr_index -=fifo->channel.reserve_top;
xge_assert(max_arr_index);
mid_point = (fifo->channel.reserve_length - fifo->channel.reserve_top)/2;
for (i = 0; i < mid_point; i++) {
dtrh = fifo->channel.reserve_arr[i];
fifo->channel.reserve_arr[i] =
fifo->channel.reserve_arr[max_arr_index - i];
fifo->channel.reserve_arr[max_arr_index - i] = dtrh;
}
#ifdef XGE_DEBUG_ASSERT
for ( i = 0; i < fifo->channel.reserve_length; i++) {
xge_debug_fifo(XGE_TRACE, "DTR after reversing index:%d"
" handle:%p", i, fifo->channel.reserve_arr[i]);
}
#endif
return XGE_HAL_OK;
}
void
__hal_fifo_close(xge_hal_channel_h channelh)
{
xge_hal_fifo_t *fifo = (xge_hal_fifo_t *)channelh;
xge_hal_device_t *hldev = (xge_hal_device_t *)fifo->channel.devh;
if (fifo->mempool) {
__hal_mempool_destroy(fifo->mempool);
}
__hal_channel_terminate(channelh);
#if defined(XGE_HAL_TX_MULTI_RESERVE)
xge_os_spin_lock_destroy(&fifo->channel.reserve_lock, hldev->pdev);
#elif defined(XGE_HAL_TX_MULTI_RESERVE_IRQ)
xge_os_spin_lock_destroy_irq(&fifo->channel.reserve_lock, hldev->pdev);
#endif
if (xge_hal_device_check_id(hldev) == XGE_HAL_CARD_HERC) {
#if defined(XGE_HAL_TX_MULTI_POST)
xge_os_spin_lock_destroy(&fifo->channel.post_lock, hldev->pdev);
#elif defined(XGE_HAL_TX_MULTI_POST_IRQ)
xge_os_spin_lock_destroy_irq(&fifo->channel.post_lock,
hldev->pdev);
#endif
}
}
void
__hal_fifo_hw_initialize(xge_hal_device_h devh)
{
xge_hal_device_t *hldev = (xge_hal_device_t *)devh;
xge_hal_pci_bar0_t *bar0 = (xge_hal_pci_bar0_t *)(void *)hldev->bar0;
u64* tx_fifo_partitions[4];
u64* tx_fifo_wrr[5];
u64 tx_fifo_wrr_value[5];
u64 val64, part0;
int i;
/* Tx DMA Initialization */
tx_fifo_partitions[0] = &bar0->tx_fifo_partition_0;
tx_fifo_partitions[1] = &bar0->tx_fifo_partition_1;
tx_fifo_partitions[2] = &bar0->tx_fifo_partition_2;
tx_fifo_partitions[3] = &bar0->tx_fifo_partition_3;
tx_fifo_wrr[0] = &bar0->tx_w_round_robin_0;
tx_fifo_wrr[1] = &bar0->tx_w_round_robin_1;
tx_fifo_wrr[2] = &bar0->tx_w_round_robin_2;
tx_fifo_wrr[3] = &bar0->tx_w_round_robin_3;
tx_fifo_wrr[4] = &bar0->tx_w_round_robin_4;
tx_fifo_wrr_value[0] = XGE_HAL_FIFO_WRR_0;
tx_fifo_wrr_value[1] = XGE_HAL_FIFO_WRR_1;
tx_fifo_wrr_value[2] = XGE_HAL_FIFO_WRR_2;
tx_fifo_wrr_value[3] = XGE_HAL_FIFO_WRR_3;
tx_fifo_wrr_value[4] = XGE_HAL_FIFO_WRR_4;
/* Note: WRR calendar must be configured before the transmit
* FIFOs are enabled! page 6-77 user guide */
if (!hldev->config.rts_qos_en) {
/* all zeroes for Round-Robin */
for (i = 0; i < XGE_HAL_FIFO_MAX_WRR; i++) {
xge_os_pio_mem_write64(hldev->pdev, hldev->regh0, 0,
tx_fifo_wrr[i]);
}
/* reset all of them but '0' */
for (i=1; i < XGE_HAL_FIFO_MAX_PARTITION; i++) {
xge_os_pio_mem_write64(hldev->pdev, hldev->regh0, 0ULL,
tx_fifo_partitions[i]);
}
} else { /* Change the default settings */
for (i = 0; i < XGE_HAL_FIFO_MAX_WRR; i++) {
xge_os_pio_mem_write64(hldev->pdev, hldev->regh0,
tx_fifo_wrr_value[i], tx_fifo_wrr[i]);
}
}
/* configure only configured FIFOs */
val64 = 0; part0 = 0;
for (i = 0; i < XGE_HAL_MAX_FIFO_NUM; i++) {
int reg_half = i % 2;
int reg_num = i / 2;
if (hldev->config.fifo.queue[i].configured) {
int priority = hldev->config.fifo.queue[i].priority;
val64 |=
vBIT((hldev->config.fifo.queue[i].max-1),
(((reg_half) * 32) + 19),
13) | vBIT(priority, (((reg_half)*32) + 5), 3);
}
/* NOTE: do write operation for each second u64 half
* or force for first one if configured number
* is even */
if (reg_half) {
if (reg_num == 0) {
/* skip partition '0', must write it once at
* the end */
part0 = val64;
} else {
xge_os_pio_mem_write64(hldev->pdev, hldev->regh0,
val64, tx_fifo_partitions[reg_num]);
xge_debug_fifo(XGE_TRACE,
"fifo partition_%d at: "
"0x"XGE_OS_LLXFMT" is: 0x"XGE_OS_LLXFMT,
reg_num, (unsigned long long)(ulong_t)
tx_fifo_partitions[reg_num],
(unsigned long long)val64);
}
val64 = 0;
}
}
part0 |= BIT(0); /* to enable the FIFO partition. */
__hal_pio_mem_write32_lower(hldev->pdev, hldev->regh0, (u32)part0,
tx_fifo_partitions[0]);
xge_os_wmb();
__hal_pio_mem_write32_upper(hldev->pdev, hldev->regh0, (u32)(part0>>32),
tx_fifo_partitions[0]);
xge_debug_fifo(XGE_TRACE, "fifo partition_0 at: "
"0x"XGE_OS_LLXFMT" is: 0x"XGE_OS_LLXFMT,
(unsigned long long)(ulong_t)
tx_fifo_partitions[0],
(unsigned long long) part0);
/*
* Initialization of Tx_PA_CONFIG register to ignore packet
* integrity checking.
*/
val64 = xge_os_pio_mem_read64(hldev->pdev, hldev->regh0,
&bar0->tx_pa_cfg);
val64 |= XGE_HAL_TX_PA_CFG_IGNORE_FRM_ERR |
XGE_HAL_TX_PA_CFG_IGNORE_SNAP_OUI |
XGE_HAL_TX_PA_CFG_IGNORE_LLC_CTRL |
XGE_HAL_TX_PA_CFG_IGNORE_L2_ERR;
xge_os_pio_mem_write64(hldev->pdev, hldev->regh0, val64,
&bar0->tx_pa_cfg);
/*
* Assign MSI-X vectors
*/
for (i = 0; i < XGE_HAL_MAX_FIFO_NUM; i++) {
xge_list_t *item;
xge_hal_channel_t *channel = NULL;
if (!hldev->config.fifo.queue[i].configured ||
!hldev->config.fifo.queue[i].intr_vector ||
!hldev->config.intr_mode != XGE_HAL_INTR_MODE_MSIX)
continue;
/* find channel */
xge_list_for_each(item, &hldev->free_channels) {
xge_hal_channel_t *tmp;
tmp = xge_container_of(item, xge_hal_channel_t,
item);
if (tmp->type == XGE_HAL_CHANNEL_TYPE_FIFO &&
tmp->post_qid == i) {
channel = tmp;
break;
}
}
if (channel) {
xge_hal_channel_msix_set(channel,
hldev->config.fifo.queue[i].intr_vector);
}
}
xge_debug_fifo(XGE_TRACE, "%s", "fifo channels initialized");
}
#ifdef XGE_HAL_ALIGN_XMIT
void
__hal_fifo_dtr_align_free_unmap(xge_hal_channel_h channelh, xge_hal_dtr_h dtrh)
{
xge_hal_fifo_txdl_priv_t *txdl_priv;
xge_hal_fifo_txd_t *txdp = (xge_hal_fifo_txd_t *)dtrh;
xge_hal_fifo_t *fifo = (xge_hal_fifo_t *)channelh;
txdl_priv = __hal_fifo_txdl_priv(txdp);
if (txdl_priv->align_dma_addr != 0) {
xge_os_dma_unmap(fifo->channel.pdev,
txdl_priv->align_dma_handle,
txdl_priv->align_dma_addr,
fifo->align_size,
XGE_OS_DMA_DIR_TODEVICE);
txdl_priv->align_dma_addr = 0;
}
if (txdl_priv->align_vaddr != NULL) {
xge_os_dma_free(fifo->channel.pdev,
txdl_priv->align_vaddr,
fifo->align_size,
&txdl_priv->align_dma_acch,
&txdl_priv->align_dma_handle);
txdl_priv->align_vaddr = NULL;
}
}
xge_hal_status_e
__hal_fifo_dtr_align_alloc_map(xge_hal_channel_h channelh, xge_hal_dtr_h dtrh)
{
xge_hal_fifo_txdl_priv_t *txdl_priv;
xge_hal_fifo_txd_t *txdp = (xge_hal_fifo_txd_t *)dtrh;
xge_hal_fifo_t *fifo = (xge_hal_fifo_t *)channelh;
xge_assert(txdp);
txdl_priv = __hal_fifo_txdl_priv(txdp);
/* allocate alignment DMA-buffer */
txdl_priv->align_vaddr = (char *)xge_os_dma_malloc(fifo->channel.pdev,
fifo->align_size,
XGE_OS_DMA_CACHELINE_ALIGNED |
XGE_OS_DMA_STREAMING,
&txdl_priv->align_dma_handle,
&txdl_priv->align_dma_acch);
if (txdl_priv->align_vaddr == NULL) {
return XGE_HAL_ERR_OUT_OF_MEMORY;
}
/* map it */
txdl_priv->align_dma_addr = xge_os_dma_map(fifo->channel.pdev,
txdl_priv->align_dma_handle, txdl_priv->align_vaddr,
fifo->align_size,
XGE_OS_DMA_DIR_TODEVICE, XGE_OS_DMA_STREAMING);
if (txdl_priv->align_dma_addr == XGE_OS_INVALID_DMA_ADDR) {
__hal_fifo_dtr_align_free_unmap(channelh, dtrh);
return XGE_HAL_ERR_OUT_OF_MAPPING;
}
return XGE_HAL_OK;
}
#endif