The new design: * Supports a generic batching capability in the accel_fw layer that keeps track of hw accelerated vs sw commands based on the capabilities of the engine and processes sw commands in the generic layer while sending a list of commands (not a batch) to the engines for processing. * Batch completions are managed via the generic layer, when using the accel_fw the engines only process commands. With DSA however, if a list of commands is sent down it will use the DSA public API to create and send batches but will rely on the generic layer to complete the batch task itself. When using DSA directly, batching works as usual (DSA handles batch completion). * The engine function tables were greatly simplified by replacing all of the individual entries (copy, fill, crc32c, etc) with one `submit_tasks` function that is used to both send lists of tasks for batches or just one task for single shot API. * Internally batching is now used to re-submit tasks that were queued for flow control reasons. Signed-off-by: paul luse <paul.e.luse@intel.com> Change-Id: I99c28751df32017c43490a90f4904bdabe79a270 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3555 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Community-CI: Mellanox Build Bot
386 lines
9.2 KiB
C
386 lines
9.2 KiB
C
/*-
|
|
* BSD LICENSE
|
|
*
|
|
* Copyright (c) Intel Corporation.
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
*
|
|
* * Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* * 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.
|
|
* * Neither the name of Intel Corporation nor the names of its
|
|
* contributors may be used to endorse or promote products derived
|
|
* from this software without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT
|
|
* OWNER 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.
|
|
*/
|
|
|
|
#include "accel_engine_ioat.h"
|
|
|
|
#include "spdk/stdinc.h"
|
|
|
|
#include "spdk_internal/accel_engine.h"
|
|
#include "spdk/log.h"
|
|
|
|
#include "spdk/env.h"
|
|
#include "spdk/event.h"
|
|
#include "spdk/thread.h"
|
|
#include "spdk/ioat.h"
|
|
|
|
static bool g_ioat_enable = false;
|
|
static bool g_ioat_initialized = false;
|
|
|
|
struct ioat_probe_ctx {
|
|
int num_whitelist_devices;
|
|
struct spdk_pci_addr whitelist[IOAT_MAX_CHANNELS];
|
|
};
|
|
|
|
static struct ioat_probe_ctx g_probe_ctx;
|
|
|
|
struct ioat_device {
|
|
struct spdk_ioat_chan *ioat;
|
|
bool is_allocated;
|
|
/** linked list pointer for device list */
|
|
TAILQ_ENTRY(ioat_device) tailq;
|
|
};
|
|
|
|
struct pci_device {
|
|
struct spdk_pci_device *pci_dev;
|
|
TAILQ_ENTRY(pci_device) tailq;
|
|
};
|
|
|
|
static TAILQ_HEAD(, ioat_device) g_devices = TAILQ_HEAD_INITIALIZER(g_devices);
|
|
static pthread_mutex_t g_ioat_mutex = PTHREAD_MUTEX_INITIALIZER;
|
|
|
|
static TAILQ_HEAD(, pci_device) g_pci_devices = TAILQ_HEAD_INITIALIZER(g_pci_devices);
|
|
|
|
struct ioat_io_channel {
|
|
struct spdk_ioat_chan *ioat_ch;
|
|
struct ioat_device *ioat_dev;
|
|
struct spdk_poller *poller;
|
|
};
|
|
|
|
static int
|
|
ioat_find_dev_by_whitelist_bdf(const struct spdk_pci_addr *pci_addr,
|
|
const struct spdk_pci_addr *whitelist,
|
|
int num_whitelist_devices)
|
|
{
|
|
int i;
|
|
|
|
for (i = 0; i < num_whitelist_devices; i++) {
|
|
if (spdk_pci_addr_compare(pci_addr, &whitelist[i]) == 0) {
|
|
return 1;
|
|
}
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
static struct ioat_device *
|
|
ioat_allocate_device(void)
|
|
{
|
|
struct ioat_device *dev;
|
|
|
|
pthread_mutex_lock(&g_ioat_mutex);
|
|
TAILQ_FOREACH(dev, &g_devices, tailq) {
|
|
if (!dev->is_allocated) {
|
|
dev->is_allocated = true;
|
|
pthread_mutex_unlock(&g_ioat_mutex);
|
|
return dev;
|
|
}
|
|
}
|
|
pthread_mutex_unlock(&g_ioat_mutex);
|
|
|
|
return NULL;
|
|
}
|
|
|
|
static void
|
|
ioat_free_device(struct ioat_device *dev)
|
|
{
|
|
pthread_mutex_lock(&g_ioat_mutex);
|
|
dev->is_allocated = false;
|
|
pthread_mutex_unlock(&g_ioat_mutex);
|
|
}
|
|
|
|
static int accel_engine_ioat_init(void);
|
|
static void accel_engine_ioat_exit(void *ctx);
|
|
|
|
static size_t
|
|
accel_engine_ioat_get_ctx_size(void)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
SPDK_ACCEL_MODULE_REGISTER(accel_engine_ioat_init, accel_engine_ioat_exit,
|
|
NULL, accel_engine_ioat_get_ctx_size)
|
|
|
|
static void
|
|
ioat_done(void *cb_arg)
|
|
{
|
|
struct spdk_accel_task *accel_task = cb_arg;
|
|
|
|
spdk_accel_task_complete(accel_task, 0);
|
|
}
|
|
|
|
static int
|
|
ioat_poll(void *arg)
|
|
{
|
|
struct spdk_ioat_chan *chan = arg;
|
|
|
|
return spdk_ioat_process_events(chan) != 0 ? SPDK_POLLER_BUSY :
|
|
SPDK_POLLER_IDLE;
|
|
}
|
|
|
|
static struct spdk_io_channel *ioat_get_io_channel(void);
|
|
|
|
static uint64_t
|
|
ioat_get_capabilities(void)
|
|
{
|
|
return ACCEL_COPY | ACCEL_FILL;
|
|
}
|
|
|
|
static uint32_t
|
|
ioat_batch_get_max(struct spdk_io_channel *ch)
|
|
{
|
|
struct ioat_io_channel *ioat_ch = spdk_io_channel_get_ctx(ch);
|
|
|
|
return spdk_ioat_get_max_descriptors(ioat_ch->ioat_dev->ioat);
|
|
}
|
|
|
|
static int
|
|
ioat_submit_tasks(struct spdk_io_channel *ch, struct spdk_accel_task *accel_task)
|
|
{
|
|
struct ioat_io_channel *ioat_ch = spdk_io_channel_get_ctx(ch);
|
|
struct spdk_accel_task *tmp;
|
|
int rc = 0;
|
|
|
|
do {
|
|
switch (accel_task->op_code) {
|
|
case ACCEL_OPCODE_MEMFILL:
|
|
rc = spdk_ioat_build_fill(ioat_ch->ioat_ch, accel_task, ioat_done,
|
|
accel_task->dst, accel_task->fill_pattern, accel_task->nbytes);
|
|
break;
|
|
case ACCEL_OPCODE_MEMMOVE:
|
|
rc = spdk_ioat_build_copy(ioat_ch->ioat_ch, accel_task, ioat_done,
|
|
accel_task->dst, accel_task->src, accel_task->nbytes);
|
|
break;
|
|
default:
|
|
assert(false);
|
|
break;
|
|
}
|
|
|
|
tmp = TAILQ_NEXT(accel_task, link);
|
|
|
|
/* Report any build errors via the callback now. */
|
|
if (rc) {
|
|
spdk_accel_task_complete(accel_task, rc);
|
|
}
|
|
|
|
accel_task = tmp;
|
|
} while (accel_task);
|
|
|
|
spdk_ioat_flush(ioat_ch->ioat_ch);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static struct spdk_accel_engine ioat_accel_engine = {
|
|
.get_capabilities = ioat_get_capabilities,
|
|
.get_io_channel = ioat_get_io_channel,
|
|
.batch_get_max = ioat_batch_get_max,
|
|
.submit_tasks = ioat_submit_tasks,
|
|
};
|
|
|
|
static int
|
|
ioat_create_cb(void *io_device, void *ctx_buf)
|
|
{
|
|
struct ioat_io_channel *ch = ctx_buf;
|
|
struct ioat_device *ioat_dev;
|
|
|
|
ioat_dev = ioat_allocate_device();
|
|
if (ioat_dev == NULL) {
|
|
return -1;
|
|
}
|
|
|
|
ch->ioat_dev = ioat_dev;
|
|
ch->ioat_ch = ioat_dev->ioat;
|
|
ch->poller = SPDK_POLLER_REGISTER(ioat_poll, ch->ioat_ch, 0);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void
|
|
ioat_destroy_cb(void *io_device, void *ctx_buf)
|
|
{
|
|
struct ioat_io_channel *ch = ctx_buf;
|
|
|
|
ioat_free_device(ch->ioat_dev);
|
|
spdk_poller_unregister(&ch->poller);
|
|
}
|
|
|
|
static struct spdk_io_channel *
|
|
ioat_get_io_channel(void)
|
|
{
|
|
return spdk_get_io_channel(&ioat_accel_engine);
|
|
}
|
|
|
|
static bool
|
|
probe_cb(void *cb_ctx, struct spdk_pci_device *pci_dev)
|
|
{
|
|
struct ioat_probe_ctx *ctx = cb_ctx;
|
|
struct spdk_pci_addr pci_addr = spdk_pci_device_get_addr(pci_dev);
|
|
struct pci_device *pdev;
|
|
|
|
SPDK_INFOLOG(accel_ioat,
|
|
" Found matching device at %04x:%02x:%02x.%x vendor:0x%04x device:0x%04x\n",
|
|
pci_addr.domain,
|
|
pci_addr.bus,
|
|
pci_addr.dev,
|
|
pci_addr.func,
|
|
spdk_pci_device_get_vendor_id(pci_dev),
|
|
spdk_pci_device_get_device_id(pci_dev));
|
|
|
|
pdev = calloc(1, sizeof(*pdev));
|
|
if (pdev == NULL) {
|
|
return false;
|
|
}
|
|
pdev->pci_dev = pci_dev;
|
|
TAILQ_INSERT_TAIL(&g_pci_devices, pdev, tailq);
|
|
|
|
if (ctx->num_whitelist_devices > 0 &&
|
|
!ioat_find_dev_by_whitelist_bdf(&pci_addr, ctx->whitelist, ctx->num_whitelist_devices)) {
|
|
return false;
|
|
}
|
|
|
|
/* Claim the device in case conflict with other process */
|
|
if (spdk_pci_device_claim(pci_dev) < 0) {
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
static void
|
|
attach_cb(void *cb_ctx, struct spdk_pci_device *pci_dev, struct spdk_ioat_chan *ioat)
|
|
{
|
|
struct ioat_device *dev;
|
|
|
|
dev = calloc(1, sizeof(*dev));
|
|
if (dev == NULL) {
|
|
SPDK_ERRLOG("Failed to allocate device struct\n");
|
|
return;
|
|
}
|
|
|
|
dev->ioat = ioat;
|
|
TAILQ_INSERT_TAIL(&g_devices, dev, tailq);
|
|
}
|
|
|
|
void
|
|
accel_engine_ioat_enable_probe(void)
|
|
{
|
|
g_ioat_enable = true;
|
|
}
|
|
|
|
static int
|
|
accel_engine_ioat_add_whitelist_device(const char *pci_bdf)
|
|
{
|
|
if (pci_bdf == NULL) {
|
|
return -1;
|
|
}
|
|
|
|
if (g_probe_ctx.num_whitelist_devices >= IOAT_MAX_CHANNELS) {
|
|
SPDK_ERRLOG("Ioat whitelist is full (max size is %d)\n",
|
|
IOAT_MAX_CHANNELS);
|
|
return -1;
|
|
}
|
|
|
|
if (spdk_pci_addr_parse(&g_probe_ctx.whitelist[g_probe_ctx.num_whitelist_devices],
|
|
pci_bdf) < 0) {
|
|
SPDK_ERRLOG("Invalid address %s\n", pci_bdf);
|
|
return -1;
|
|
}
|
|
|
|
g_probe_ctx.num_whitelist_devices++;
|
|
|
|
return 0;
|
|
}
|
|
|
|
int
|
|
accel_engine_ioat_add_whitelist_devices(const char *pci_bdfs[], size_t num_pci_bdfs)
|
|
{
|
|
size_t i;
|
|
|
|
for (i = 0; i < num_pci_bdfs; i++) {
|
|
if (accel_engine_ioat_add_whitelist_device(pci_bdfs[i]) < 0) {
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
accel_engine_ioat_init(void)
|
|
{
|
|
if (!g_ioat_enable) {
|
|
return 0;
|
|
}
|
|
|
|
if (spdk_ioat_probe(&g_probe_ctx, probe_cb, attach_cb) != 0) {
|
|
SPDK_ERRLOG("spdk_ioat_probe() failed\n");
|
|
return -1;
|
|
}
|
|
|
|
g_ioat_initialized = true;
|
|
SPDK_NOTICELOG("Accel engine updated to use IOAT engine.\n");
|
|
spdk_accel_hw_engine_register(&ioat_accel_engine);
|
|
spdk_io_device_register(&ioat_accel_engine, ioat_create_cb, ioat_destroy_cb,
|
|
sizeof(struct ioat_io_channel), "ioat_accel_engine");
|
|
return 0;
|
|
}
|
|
|
|
static void
|
|
accel_engine_ioat_exit(void *ctx)
|
|
{
|
|
struct ioat_device *dev;
|
|
struct pci_device *pci_dev;
|
|
|
|
if (g_ioat_initialized) {
|
|
spdk_io_device_unregister(&ioat_accel_engine, NULL);
|
|
}
|
|
|
|
while (!TAILQ_EMPTY(&g_devices)) {
|
|
dev = TAILQ_FIRST(&g_devices);
|
|
TAILQ_REMOVE(&g_devices, dev, tailq);
|
|
spdk_ioat_detach(dev->ioat);
|
|
ioat_free_device(dev);
|
|
free(dev);
|
|
}
|
|
|
|
while (!TAILQ_EMPTY(&g_pci_devices)) {
|
|
pci_dev = TAILQ_FIRST(&g_pci_devices);
|
|
TAILQ_REMOVE(&g_pci_devices, pci_dev, tailq);
|
|
spdk_pci_device_detach(pci_dev->pci_dev);
|
|
free(pci_dev);
|
|
}
|
|
|
|
spdk_accel_engine_module_finish();
|
|
}
|
|
|
|
SPDK_LOG_REGISTER_COMPONENT(accel_ioat)
|