numam-spdk/lib/vhost/vhost_blk.c

1093 lines
30 KiB
C
Raw Normal View History

/*-
* BSD LICENSE
*
* Copyright(c) Intel Corporation. All rights reserved.
* 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 <linux/virtio_blk.h>
#include "spdk/env.h"
#include "spdk/bdev.h"
#include "spdk/conf.h"
#include "spdk/thread.h"
#include "spdk/likely.h"
#include "spdk/string.h"
#include "spdk/util.h"
#include "spdk/vhost.h"
#include "vhost_internal.h"
struct spdk_vhost_blk_task {
struct spdk_bdev_io *bdev_io;
struct spdk_vhost_blk_session *bvsession;
struct spdk_vhost_virtqueue *vq;
volatile uint8_t *status;
uint16_t req_idx;
/* for io wait */
struct spdk_bdev_io_wait_entry bdev_io_wait;
/* If set, the task is currently used for I/O processing. */
bool used;
/** Number of bytes that were written. */
uint32_t used_len;
uint16_t iovcnt;
struct iovec iovs[SPDK_VHOST_IOVS_MAX];
};
struct spdk_vhost_blk_dev {
struct spdk_vhost_dev vdev;
struct spdk_bdev *bdev;
struct spdk_bdev_desc *bdev_desc;
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
bool readonly;
};
struct spdk_vhost_blk_session {
/* The parent session must be the very first field in this struct */
struct spdk_vhost_session vsession;
struct spdk_vhost_blk_dev *bvdev;
struct spdk_poller *requestq_poller;
struct spdk_io_channel *io_channel;
struct spdk_poller *stop_poller;
};
/* forward declaration */
static const struct spdk_vhost_dev_backend vhost_blk_device_backend;
static int
process_blk_request(struct spdk_vhost_blk_task *task,
struct spdk_vhost_blk_session *bvsession,
struct spdk_vhost_virtqueue *vq);
static void
blk_task_finish(struct spdk_vhost_blk_task *task)
{
assert(task->bvsession->vsession.task_cnt > 0);
task->bvsession->vsession.task_cnt--;
task->used = false;
}
static void
invalid_blk_request(struct spdk_vhost_blk_task *task, uint8_t status)
{
if (task->status) {
*task->status = status;
}
spdk_vhost_vq_used_ring_enqueue(&task->bvsession->vsession, task->vq, task->req_idx,
task->used_len);
blk_task_finish(task);
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK_DATA, "Invalid request (status=%" PRIu8")\n", status);
}
/*
* Process task's descriptor chain and setup data related fields.
* Return
* total size of suplied buffers
*
* FIXME: Make this function return to rd_cnt and wr_cnt
*/
static int
blk_iovs_setup(struct spdk_vhost_blk_session *bvsession, struct spdk_vhost_virtqueue *vq,
uint16_t req_idx, struct iovec *iovs, uint16_t *iovs_cnt, uint32_t *length)
{
struct spdk_vhost_session *vsession = &bvsession->vsession;
struct spdk_vhost_dev *vdev = vsession->vdev;
struct vring_desc *desc, *desc_table;
uint16_t out_cnt = 0, cnt = 0;
uint32_t desc_table_size, len = 0;
uint32_t desc_handled_cnt;
int rc;
rc = spdk_vhost_vq_get_desc(vsession, vq, req_idx, &desc, &desc_table, &desc_table_size);
if (rc != 0) {
SPDK_ERRLOG("%s: Invalid descriptor at index %"PRIu16".\n", vdev->name, req_idx);
return -1;
}
desc_handled_cnt = 0;
while (1) {
/*
* Maximum cnt reached?
* Should not happen if request is well formatted, otherwise this is a BUG.
*/
if (spdk_unlikely(cnt == *iovs_cnt)) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "Max IOVs in request reached (req_idx = %"PRIu16").\n",
req_idx);
return -1;
}
if (spdk_unlikely(spdk_vhost_vring_desc_to_iov(vsession, iovs, &cnt, desc))) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "Invalid descriptor %" PRIu16" (req_idx = %"PRIu16").\n",
req_idx, cnt);
return -1;
}
len += desc->len;
out_cnt += spdk_vhost_vring_desc_is_wr(desc);
rc = spdk_vhost_vring_desc_get_next(&desc, desc_table, desc_table_size);
if (rc != 0) {
SPDK_ERRLOG("%s: Descriptor chain at index %"PRIu16" terminated unexpectedly.\n",
vdev->name, req_idx);
return -1;
} else if (desc == NULL) {
break;
}
desc_handled_cnt++;
if (spdk_unlikely(desc_handled_cnt > desc_table_size)) {
/* Break a cycle and report an error, if any. */
SPDK_ERRLOG("%s: found a cycle in the descriptor chain: desc_table_size = %d, desc_handled_cnt = %d.\n",
vdev->name, desc_table_size, desc_handled_cnt);
return -1;
}
}
/*
* There must be least two descriptors.
* First contain request so it must be readable.
* Last descriptor contain buffer for response so it must be writable.
*/
if (spdk_unlikely(out_cnt == 0 || cnt < 2)) {
return -1;
}
*length = len;
*iovs_cnt = cnt;
return 0;
}
static void
blk_request_finish(bool success, struct spdk_vhost_blk_task *task)
{
*task->status = success ? VIRTIO_BLK_S_OK : VIRTIO_BLK_S_IOERR;
spdk_vhost_vq_used_ring_enqueue(&task->bvsession->vsession, task->vq, task->req_idx,
task->used_len);
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "Finished task (%p) req_idx=%d\n status: %s\n", task,
task->req_idx, success ? "OK" : "FAIL");
blk_task_finish(task);
}
static void
blk_request_complete_cb(struct spdk_bdev_io *bdev_io, bool success, void *cb_arg)
{
struct spdk_vhost_blk_task *task = cb_arg;
spdk_bdev_free_io(bdev_io);
blk_request_finish(success, task);
}
static void
blk_request_resubmit(void *arg)
{
struct spdk_vhost_blk_task *task = (struct spdk_vhost_blk_task *)arg;
int rc = 0;
rc = process_blk_request(task, task->bvsession, task->vq);
if (rc == 0) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "====== Task %p resubmitted ======\n", task);
} else {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "====== Task %p failed ======\n", task);
}
}
static inline void
blk_request_queue_io(struct spdk_vhost_blk_task *task)
{
int rc;
struct spdk_vhost_blk_session *bvsession = task->bvsession;
struct spdk_bdev *bdev = bvsession->bvdev->bdev;
task->bdev_io_wait.bdev = bdev;
task->bdev_io_wait.cb_fn = blk_request_resubmit;
task->bdev_io_wait.cb_arg = task;
rc = spdk_bdev_queue_io_wait(bdev, bvsession->io_channel, &task->bdev_io_wait);
if (rc != 0) {
SPDK_ERRLOG("Queue io failed in vhost_blk, rc=%d\n", rc);
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
}
}
static int
process_blk_request(struct spdk_vhost_blk_task *task,
struct spdk_vhost_blk_session *bvsession,
struct spdk_vhost_virtqueue *vq)
{
struct spdk_vhost_blk_dev *bvdev = bvsession->bvdev;
const struct virtio_blk_outhdr *req;
struct virtio_blk_discard_write_zeroes *desc;
struct iovec *iov;
uint32_t type;
uint32_t payload_len;
uint64_t flush_bytes;
int rc;
if (blk_iovs_setup(bvsession, vq, task->req_idx, task->iovs, &task->iovcnt, &payload_len)) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "Invalid request (req_idx = %"PRIu16").\n", task->req_idx);
/* Only READ and WRITE are supported for now. */
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
iov = &task->iovs[0];
if (spdk_unlikely(iov->iov_len != sizeof(*req))) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK,
"First descriptor size is %zu but expected %zu (req_idx = %"PRIu16").\n",
iov->iov_len, sizeof(*req), task->req_idx);
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
req = iov->iov_base;
iov = &task->iovs[task->iovcnt - 1];
if (spdk_unlikely(iov->iov_len != 1)) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK,
"Last descriptor size is %zu but expected %d (req_idx = %"PRIu16").\n",
iov->iov_len, 1, task->req_idx);
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
task->status = iov->iov_base;
payload_len -= sizeof(*req) + sizeof(*task->status);
task->iovcnt -= 2;
type = req->type;
#ifdef VIRTIO_BLK_T_BARRIER
/* Don't care about barier for now (as QEMU's virtio-blk do). */
type &= ~VIRTIO_BLK_T_BARRIER;
#endif
switch (type) {
case VIRTIO_BLK_T_IN:
case VIRTIO_BLK_T_OUT:
if (spdk_unlikely(payload_len == 0 || (payload_len & (512 - 1)) != 0)) {
SPDK_ERRLOG("%s - passed IO buffer is not multiple of 512b (req_idx = %"PRIu16").\n",
type ? "WRITE" : "READ", task->req_idx);
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
if (type == VIRTIO_BLK_T_IN) {
task->used_len = payload_len + sizeof(*task->status);
rc = spdk_bdev_readv(bvdev->bdev_desc, bvsession->io_channel,
&task->iovs[1], task->iovcnt, req->sector * 512,
payload_len, blk_request_complete_cb, task);
} else if (!bvdev->readonly) {
task->used_len = sizeof(*task->status);
rc = spdk_bdev_writev(bvdev->bdev_desc, bvsession->io_channel,
&task->iovs[1], task->iovcnt, req->sector * 512,
payload_len, blk_request_complete_cb, task);
} else {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "Device is in read-only mode!\n");
rc = -1;
}
if (rc) {
if (rc == -ENOMEM) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "No memory, start to queue io.\n");
blk_request_queue_io(task);
} else {
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
}
break;
case VIRTIO_BLK_T_DISCARD:
desc = task->iovs[1].iov_base;
if (payload_len != sizeof(*desc)) {
SPDK_NOTICELOG("Invalid discard payload size: %u\n", payload_len);
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
rc = spdk_bdev_unmap(bvdev->bdev_desc, bvsession->io_channel,
desc->sector * 512, desc->num_sectors * 512,
blk_request_complete_cb, task);
if (rc) {
if (rc == -ENOMEM) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "No memory, start to queue io.\n");
blk_request_queue_io(task);
} else {
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
}
break;
case VIRTIO_BLK_T_WRITE_ZEROES:
desc = task->iovs[1].iov_base;
if (payload_len != sizeof(*desc)) {
SPDK_NOTICELOG("Invalid write zeroes payload size: %u\n", payload_len);
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
/* Zeroed and Unmap the range, SPDK doen't support it. */
if (desc->flags & VIRTIO_BLK_WRITE_ZEROES_FLAG_UNMAP) {
SPDK_NOTICELOG("Can't support Write Zeroes with Unmap flag\n");
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
rc = spdk_bdev_write_zeroes(bvdev->bdev_desc, bvsession->io_channel,
desc->sector * 512, desc->num_sectors * 512,
blk_request_complete_cb, task);
if (rc) {
if (rc == -ENOMEM) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "No memory, start to queue io.\n");
blk_request_queue_io(task);
} else {
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
}
break;
case VIRTIO_BLK_T_FLUSH:
flush_bytes = spdk_bdev_get_num_blocks(bvdev->bdev) * spdk_bdev_get_block_size(bvdev->bdev);
if (req->sector != 0) {
SPDK_NOTICELOG("sector must be zero for flush command\n");
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
rc = spdk_bdev_flush(bvdev->bdev_desc, bvsession->io_channel,
0, flush_bytes,
blk_request_complete_cb, task);
if (rc) {
if (rc == -ENOMEM) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "No memory, start to queue io.\n");
blk_request_queue_io(task);
} else {
invalid_blk_request(task, VIRTIO_BLK_S_IOERR);
return -1;
}
}
break;
case VIRTIO_BLK_T_GET_ID:
if (!task->iovcnt || !payload_len) {
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
task->used_len = spdk_min((size_t)VIRTIO_BLK_ID_BYTES, task->iovs[1].iov_len);
spdk_strcpy_pad(task->iovs[1].iov_base, spdk_bdev_get_product_name(bvdev->bdev),
task->used_len, ' ');
blk_request_finish(true, task);
break;
default:
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "Not supported request type '%"PRIu32"'.\n", type);
invalid_blk_request(task, VIRTIO_BLK_S_UNSUPP);
return -1;
}
return 0;
}
static void
process_vq(struct spdk_vhost_blk_session *bvsession, struct spdk_vhost_virtqueue *vq)
{
struct spdk_vhost_blk_dev *bvdev = bvsession->bvdev;
struct spdk_vhost_blk_task *task;
struct spdk_vhost_session *vsession = &bvsession->vsession;
int rc;
uint16_t reqs[32];
uint16_t reqs_cnt, i;
reqs_cnt = spdk_vhost_vq_avail_ring_get(vq, reqs, SPDK_COUNTOF(reqs));
if (!reqs_cnt) {
return;
}
for (i = 0; i < reqs_cnt; i++) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "====== Starting processing request idx %"PRIu16"======\n",
reqs[i]);
if (spdk_unlikely(reqs[i] >= vq->vring.size)) {
SPDK_ERRLOG("%s: request idx '%"PRIu16"' exceeds virtqueue size (%"PRIu16").\n",
bvdev->vdev.name, reqs[i], vq->vring.size);
2018-12-13 10:51:34 +00:00
spdk_vhost_vq_used_ring_enqueue(vsession, vq, reqs[i], 0);
continue;
}
task = &((struct spdk_vhost_blk_task *)vq->tasks)[reqs[i]];
if (spdk_unlikely(task->used)) {
SPDK_ERRLOG("%s: request with idx '%"PRIu16"' is already pending.\n",
bvdev->vdev.name, reqs[i]);
2018-12-13 10:51:34 +00:00
spdk_vhost_vq_used_ring_enqueue(vsession, vq, reqs[i], 0);
continue;
}
2018-12-13 10:51:34 +00:00
vsession->task_cnt++;
task->used = true;
task->iovcnt = SPDK_COUNTOF(task->iovs);
task->status = NULL;
task->used_len = 0;
rc = process_blk_request(task, bvsession, vq);
if (rc == 0) {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "====== Task %p req_idx %d submitted ======\n", task,
reqs[i]);
} else {
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK, "====== Task %p req_idx %d failed ======\n", task, reqs[i]);
}
}
}
static int
vdev_worker(void *arg)
{
struct spdk_vhost_blk_session *bvsession = arg;
struct spdk_vhost_session *vsession = &bvsession->vsession;
uint16_t q_idx;
2018-12-13 10:51:34 +00:00
for (q_idx = 0; q_idx < vsession->max_queues; q_idx++) {
process_vq(bvsession, &vsession->virtqueue[q_idx]);
}
2018-12-13 10:51:34 +00:00
spdk_vhost_session_used_signal(vsession);
return -1;
}
static void
no_bdev_process_vq(struct spdk_vhost_blk_session *bvsession, struct spdk_vhost_virtqueue *vq)
{
struct spdk_vhost_session *vsession = &bvsession->vsession;
struct iovec iovs[SPDK_VHOST_IOVS_MAX];
uint32_t length;
uint16_t iovcnt, req_idx;
if (spdk_vhost_vq_avail_ring_get(vq, &req_idx, 1) != 1) {
return;
}
iovcnt = SPDK_COUNTOF(iovs);
if (blk_iovs_setup(bvsession, vq, req_idx, iovs, &iovcnt, &length) == 0) {
*(volatile uint8_t *)iovs[iovcnt - 1].iov_base = VIRTIO_BLK_S_IOERR;
SPDK_DEBUGLOG(SPDK_LOG_VHOST_BLK_DATA, "Aborting request %" PRIu16"\n", req_idx);
}
2018-12-13 10:51:34 +00:00
spdk_vhost_vq_used_ring_enqueue(vsession, vq, req_idx, 0);
}
static int
no_bdev_vdev_worker(void *arg)
{
struct spdk_vhost_blk_session *bvsession = arg;
struct spdk_vhost_session *vsession = &bvsession->vsession;
uint16_t q_idx;
2018-12-13 10:51:34 +00:00
for (q_idx = 0; q_idx < vsession->max_queues; q_idx++) {
no_bdev_process_vq(bvsession, &vsession->virtqueue[q_idx]);
}
2018-12-13 10:51:34 +00:00
spdk_vhost_session_used_signal(vsession);
if (vsession->task_cnt == 0 && bvsession->io_channel) {
spdk_put_io_channel(bvsession->io_channel);
bvsession->io_channel = NULL;
}
return -1;
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
static struct spdk_vhost_blk_session *
to_blk_session(struct spdk_vhost_session *vsession)
{
if (vsession == NULL) {
return NULL;
}
if (vsession->vdev->backend != &vhost_blk_device_backend) {
SPDK_ERRLOG("%s: not a vhost-blk device\n", vsession->vdev->name);
return NULL;
}
return (struct spdk_vhost_blk_session *)vsession;
}
static struct spdk_vhost_blk_dev *
to_blk_dev(struct spdk_vhost_dev *vdev)
{
if (vdev == NULL) {
return NULL;
}
if (vdev->backend != &vhost_blk_device_backend) {
SPDK_ERRLOG("%s: not a vhost-blk device\n", vdev->name);
return NULL;
}
return SPDK_CONTAINEROF(vdev, struct spdk_vhost_blk_dev, vdev);
}
struct spdk_bdev *
spdk_vhost_blk_get_dev(struct spdk_vhost_dev *vdev)
{
struct spdk_vhost_blk_dev *bvdev = to_blk_dev(vdev);
assert(bvdev != NULL);
return bvdev->bdev;
}
static int
_spdk_vhost_session_bdev_remove_cb(struct spdk_vhost_dev *vdev, struct spdk_vhost_session *vsession,
void *ctx)
{
struct spdk_vhost_blk_session *bvsession;
if (vdev == NULL) {
/* Nothing to do */
return 0;
}
if (vsession == NULL) {
/* All sessions have been notified, time to close the bdev */
struct spdk_vhost_blk_dev *bvdev = to_blk_dev(vdev);
assert(bvdev != NULL);
spdk_bdev_close(bvdev->bdev_desc);
bvdev->bdev_desc = NULL;
bvdev->bdev = NULL;
return 0;
}
bvsession = (struct spdk_vhost_blk_session *)vsession;
if (bvsession->requestq_poller) {
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
spdk_poller_unregister(&bvsession->requestq_poller);
bvsession->requestq_poller = spdk_poller_register(no_bdev_vdev_worker, bvsession, 0);
}
return 0;
}
static void
bdev_remove_cb(void *remove_ctx)
{
struct spdk_vhost_blk_dev *bvdev = remove_ctx;
SPDK_WARNLOG("Controller %s: Hot-removing bdev - all further requests will fail.\n",
bvdev->vdev.name);
spdk_vhost_lock();
spdk_vhost_dev_foreach_session(&bvdev->vdev, _spdk_vhost_session_bdev_remove_cb, NULL);
spdk_vhost_unlock();
}
static void
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
free_task_pool(struct spdk_vhost_blk_session *bvsession)
{
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
struct spdk_vhost_session *vsession = &bvsession->vsession;
struct spdk_vhost_virtqueue *vq;
uint16_t i;
2018-12-13 10:51:34 +00:00
for (i = 0; i < vsession->max_queues; i++) {
vq = &vsession->virtqueue[i];
if (vq->tasks == NULL) {
continue;
}
spdk_free(vq->tasks);
vq->tasks = NULL;
}
}
static int
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
alloc_task_pool(struct spdk_vhost_blk_session *bvsession)
{
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
struct spdk_vhost_session *vsession = &bvsession->vsession;
struct spdk_vhost_blk_dev *bvdev = bvsession->bvdev;
struct spdk_vhost_virtqueue *vq;
struct spdk_vhost_blk_task *task;
uint32_t task_cnt;
uint16_t i;
uint32_t j;
2018-12-13 10:51:34 +00:00
for (i = 0; i < vsession->max_queues; i++) {
vq = &vsession->virtqueue[i];
if (vq->vring.desc == NULL) {
continue;
}
task_cnt = vq->vring.size;
if (task_cnt > SPDK_VHOST_MAX_VQ_SIZE) {
/* sanity check */
SPDK_ERRLOG("Controller %s: virtuque %"PRIu16" is too big. (size = %"PRIu32", max = %"PRIu32")\n",
bvdev->vdev.name, i, task_cnt, SPDK_VHOST_MAX_VQ_SIZE);
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
free_task_pool(bvsession);
return -1;
}
vq->tasks = spdk_zmalloc(sizeof(struct spdk_vhost_blk_task) * task_cnt,
SPDK_CACHE_LINE_SIZE, NULL,
SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
if (vq->tasks == NULL) {
SPDK_ERRLOG("Controller %s: failed to allocate %"PRIu32" tasks for virtqueue %"PRIu16"\n",
bvdev->vdev.name, task_cnt, i);
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
free_task_pool(bvsession);
return -1;
}
for (j = 0; j < task_cnt; j++) {
task = &((struct spdk_vhost_blk_task *)vq->tasks)[j];
task->bvsession = bvsession;
task->req_idx = j;
task->vq = vq;
}
}
return 0;
}
static int
spdk_vhost_blk_start_cb(struct spdk_vhost_dev *vdev,
struct spdk_vhost_session *vsession, void *unused)
{
struct spdk_vhost_blk_dev *bvdev;
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
struct spdk_vhost_blk_session *bvsession;
int i, rc = 0;
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
bvsession = to_blk_session(vsession);
if (bvsession == NULL) {
SPDK_ERRLOG("Trying to start non-blk controller as a blk one.\n");
rc = -1;
goto out;
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
bvdev = to_blk_dev(vdev);
assert(bvdev != NULL);
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
bvsession->bvdev = bvdev;
/* validate all I/O queues are in a contiguous index range */
2018-12-13 10:51:34 +00:00
for (i = 0; i < vsession->max_queues; i++) {
if (vsession->virtqueue[i].vring.desc == NULL) {
SPDK_ERRLOG("%s: queue %"PRIu32" is empty\n", vdev->name, i);
rc = -1;
goto out;
}
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
rc = alloc_task_pool(bvsession);
if (rc != 0) {
SPDK_ERRLOG("%s: failed to alloc task pool.\n", bvdev->vdev.name);
goto out;
}
if (bvdev->bdev) {
bvsession->io_channel = spdk_bdev_get_io_channel(bvdev->bdev_desc);
if (!bvsession->io_channel) {
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
free_task_pool(bvsession);
SPDK_ERRLOG("Controller %s: IO channel allocation failed\n", vdev->name);
rc = -1;
goto out;
}
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
bvsession->requestq_poller = spdk_poller_register(bvdev->bdev ? vdev_worker : no_bdev_vdev_worker,
bvsession, 0);
SPDK_INFOLOG(SPDK_LOG_VHOST, "Started poller for vhost controller %s on lcore %d\n",
vhost: change vsession->lcore only within that lcore There is currently a small window after we stop session's pollers and before we mark the session as stopped (by setting vsession->lcore to -1). If spdk_vhost_dev_foreach_session() is called within this window, its callback could assume the session is still running and for example in vhost scsi target hotremove case, could destroy an io_channel for the second time - as it'd first done when the session was stopped. That's a bug. A similar case exists for session start. We fix the above by setting vsession->lcore directly after starting or stopping the session, hence eliminating the possible window for data races. This has a few implications: * spdk_vhost_session_send_event() called before session start can't operate on vsession->lcore, so it needs to be provided with the lcore as an additional parameter now. * the vsession->lcore can't be accessed until spdk_vhost_session_start_done() is called, so its existing usages were replaced with spdk_env_get_current_core() * active_session_num is decremented right after spdk_vhost_session_stop_done() is called and before spdk_vhost_session_send_event() returns, so some active_session_num == 1 checks meaning "the last session gets stopped now" needed to be changed to check against == 0, as if "the last session has been just stopped" Change-Id: I5781bb0ce247425130c9672e0df27d06b6234317 Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448229 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
2019-03-17 11:13:01 +00:00
vdev->name, spdk_env_get_current_core());
out:
spdk_vhost_session_start_done(vsession, rc);
return rc;
}
static int
spdk_vhost_blk_start(struct spdk_vhost_session *vsession)
{
vhost: change vsession->lcore only within that lcore There is currently a small window after we stop session's pollers and before we mark the session as stopped (by setting vsession->lcore to -1). If spdk_vhost_dev_foreach_session() is called within this window, its callback could assume the session is still running and for example in vhost scsi target hotremove case, could destroy an io_channel for the second time - as it'd first done when the session was stopped. That's a bug. A similar case exists for session start. We fix the above by setting vsession->lcore directly after starting or stopping the session, hence eliminating the possible window for data races. This has a few implications: * spdk_vhost_session_send_event() called before session start can't operate on vsession->lcore, so it needs to be provided with the lcore as an additional parameter now. * the vsession->lcore can't be accessed until spdk_vhost_session_start_done() is called, so its existing usages were replaced with spdk_env_get_current_core() * active_session_num is decremented right after spdk_vhost_session_stop_done() is called and before spdk_vhost_session_send_event() returns, so some active_session_num == 1 checks meaning "the last session gets stopped now" needed to be changed to check against == 0, as if "the last session has been just stopped" Change-Id: I5781bb0ce247425130c9672e0df27d06b6234317 Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448229 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
2019-03-17 11:13:01 +00:00
int32_t lcore;
int rc;
vhost: change vsession->lcore only within that lcore There is currently a small window after we stop session's pollers and before we mark the session as stopped (by setting vsession->lcore to -1). If spdk_vhost_dev_foreach_session() is called within this window, its callback could assume the session is still running and for example in vhost scsi target hotremove case, could destroy an io_channel for the second time - as it'd first done when the session was stopped. That's a bug. A similar case exists for session start. We fix the above by setting vsession->lcore directly after starting or stopping the session, hence eliminating the possible window for data races. This has a few implications: * spdk_vhost_session_send_event() called before session start can't operate on vsession->lcore, so it needs to be provided with the lcore as an additional parameter now. * the vsession->lcore can't be accessed until spdk_vhost_session_start_done() is called, so its existing usages were replaced with spdk_env_get_current_core() * active_session_num is decremented right after spdk_vhost_session_stop_done() is called and before spdk_vhost_session_send_event() returns, so some active_session_num == 1 checks meaning "the last session gets stopped now" needed to be changed to check against == 0, as if "the last session has been just stopped" Change-Id: I5781bb0ce247425130c9672e0df27d06b6234317 Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448229 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
2019-03-17 11:13:01 +00:00
lcore = spdk_vhost_allocate_reactor(vsession->vdev->cpumask);
rc = spdk_vhost_session_send_event(lcore, vsession, spdk_vhost_blk_start_cb,
3, "start session");
if (rc != 0) {
vhost: change vsession->lcore only within that lcore There is currently a small window after we stop session's pollers and before we mark the session as stopped (by setting vsession->lcore to -1). If spdk_vhost_dev_foreach_session() is called within this window, its callback could assume the session is still running and for example in vhost scsi target hotremove case, could destroy an io_channel for the second time - as it'd first done when the session was stopped. That's a bug. A similar case exists for session start. We fix the above by setting vsession->lcore directly after starting or stopping the session, hence eliminating the possible window for data races. This has a few implications: * spdk_vhost_session_send_event() called before session start can't operate on vsession->lcore, so it needs to be provided with the lcore as an additional parameter now. * the vsession->lcore can't be accessed until spdk_vhost_session_start_done() is called, so its existing usages were replaced with spdk_env_get_current_core() * active_session_num is decremented right after spdk_vhost_session_stop_done() is called and before spdk_vhost_session_send_event() returns, so some active_session_num == 1 checks meaning "the last session gets stopped now" needed to be changed to check against == 0, as if "the last session has been just stopped" Change-Id: I5781bb0ce247425130c9672e0df27d06b6234317 Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448229 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
2019-03-17 11:13:01 +00:00
spdk_vhost_free_reactor(lcore);
}
return rc;
}
static int
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
destroy_session_poller_cb(void *arg)
{
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
struct spdk_vhost_blk_session *bvsession = arg;
struct spdk_vhost_session *vsession = &bvsession->vsession;
int i;
2018-12-13 10:51:34 +00:00
if (vsession->task_cnt > 0) {
return -1;
}
if (spdk_vhost_trylock() != 0) {
return -1;
}
2018-12-13 10:51:34 +00:00
for (i = 0; i < vsession->max_queues; i++) {
vsession->virtqueue[i].next_event_time = 0;
spdk_vhost_vq_used_signal(vsession, &vsession->virtqueue[i]);
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
SPDK_INFOLOG(SPDK_LOG_VHOST, "Stopping poller for vhost controller %s\n", vsession->vdev->name);
if (bvsession->io_channel) {
spdk_put_io_channel(bvsession->io_channel);
bvsession->io_channel = NULL;
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
free_task_pool(bvsession);
spdk_poller_unregister(&bvsession->stop_poller);
spdk_vhost_session_stop_done(vsession, 0);
spdk_vhost_unlock();
return -1;
}
static int
spdk_vhost_blk_stop_cb(struct spdk_vhost_dev *vdev,
struct spdk_vhost_session *vsession, void *unused)
{
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
struct spdk_vhost_blk_session *bvsession;
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
bvsession = to_blk_session(vsession);
if (bvsession == NULL) {
SPDK_ERRLOG("Trying to stop non-blk controller as a blk one.\n");
goto err;
}
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
spdk_poller_unregister(&bvsession->requestq_poller);
bvsession->stop_poller = spdk_poller_register(destroy_session_poller_cb,
bvsession, 1000);
return 0;
err:
spdk_vhost_session_stop_done(vsession, -1);
return -1;
}
static int
spdk_vhost_blk_stop(struct spdk_vhost_session *vsession)
{
vhost: change vsession->lcore only within that lcore There is currently a small window after we stop session's pollers and before we mark the session as stopped (by setting vsession->lcore to -1). If spdk_vhost_dev_foreach_session() is called within this window, its callback could assume the session is still running and for example in vhost scsi target hotremove case, could destroy an io_channel for the second time - as it'd first done when the session was stopped. That's a bug. A similar case exists for session start. We fix the above by setting vsession->lcore directly after starting or stopping the session, hence eliminating the possible window for data races. This has a few implications: * spdk_vhost_session_send_event() called before session start can't operate on vsession->lcore, so it needs to be provided with the lcore as an additional parameter now. * the vsession->lcore can't be accessed until spdk_vhost_session_start_done() is called, so its existing usages were replaced with spdk_env_get_current_core() * active_session_num is decremented right after spdk_vhost_session_stop_done() is called and before spdk_vhost_session_send_event() returns, so some active_session_num == 1 checks meaning "the last session gets stopped now" needed to be changed to check against == 0, as if "the last session has been just stopped" Change-Id: I5781bb0ce247425130c9672e0df27d06b6234317 Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448229 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
2019-03-17 11:13:01 +00:00
return spdk_vhost_session_send_event(vsession->lcore, vsession,
spdk_vhost_blk_stop_cb, 3, "stop session");
}
static void
spdk_vhost_blk_dump_info_json(struct spdk_vhost_dev *vdev, struct spdk_json_write_ctx *w)
{
struct spdk_bdev *bdev = spdk_vhost_blk_get_dev(vdev);
struct spdk_vhost_blk_dev *bvdev;
bvdev = to_blk_dev(vdev);
if (bvdev == NULL) {
return;
}
assert(bvdev != NULL);
spdk_json_write_named_object_begin(w, "block");
spdk_json_write_named_bool(w, "readonly", bvdev->readonly);
spdk_json_write_name(w, "bdev");
if (bdev) {
spdk_json_write_string(w, spdk_bdev_get_name(bdev));
} else {
spdk_json_write_null(w);
}
spdk_json_write_object_end(w);
}
static void
spdk_vhost_blk_write_config_json(struct spdk_vhost_dev *vdev, struct spdk_json_write_ctx *w)
{
struct spdk_vhost_blk_dev *bvdev;
bvdev = to_blk_dev(vdev);
if (bvdev == NULL) {
return;
}
if (!bvdev->bdev) {
return;
}
spdk_json_write_object_begin(w);
spdk_json_write_named_string(w, "method", "construct_vhost_blk_controller");
spdk_json_write_named_object_begin(w, "params");
spdk_json_write_named_string(w, "ctrlr", vdev->name);
spdk_json_write_named_string(w, "dev_name", spdk_bdev_get_name(bvdev->bdev));
spdk_json_write_named_string(w, "cpumask", spdk_cpuset_fmt(vdev->cpumask));
spdk_json_write_named_bool(w, "readonly", bvdev->readonly);
spdk_json_write_object_end(w);
spdk_json_write_object_end(w);
}
static int spdk_vhost_blk_destroy(struct spdk_vhost_dev *dev);
static int
spdk_vhost_blk_get_config(struct spdk_vhost_dev *vdev, uint8_t *config,
uint32_t len)
{
struct virtio_blk_config blkcfg;
struct spdk_vhost_blk_dev *bvdev;
struct spdk_bdev *bdev;
uint32_t blk_size;
uint64_t blkcnt;
bvdev = to_blk_dev(vdev);
if (bvdev == NULL) {
SPDK_ERRLOG("Trying to get virito_blk configuration failed\n");
return -1;
}
bdev = bvdev->bdev;
if (bdev == NULL) {
/* We can't just return -1 here as this GET_CONFIG message might
* be caused by a QEMU VM reboot. Returning -1 will indicate an
* error to QEMU, who might then decide to terminate itself.
* We don't want that. A simple reboot shouldn't break the system.
*
* Presenting a block device with block size 0 and block count 0
* doesn't cause any problems on QEMU side and the virtio-pci
* device is even still available inside the VM, but there will
* be no block device created for it - the kernel drivers will
* silently reject it.
*/
blk_size = 0;
blkcnt = 0;
} else {
blk_size = spdk_bdev_get_block_size(bdev);
blkcnt = spdk_bdev_get_num_blocks(bdev);
}
memset(&blkcfg, 0, sizeof(blkcfg));
blkcfg.blk_size = blk_size;
/* minimum I/O size in blocks */
blkcfg.min_io_size = 1;
/* expressed in 512 Bytes sectors */
blkcfg.capacity = (blkcnt * blk_size) / 512;
blkcfg.size_max = 131072;
/* -2 for REQ and RESP and -1 for region boundary splitting */
blkcfg.seg_max = SPDK_VHOST_IOVS_MAX - 2 - 1;
/* QEMU can overwrite this value when started */
blkcfg.num_queues = SPDK_VHOST_MAX_VQUEUES;
if (bdev && spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_UNMAP)) {
/* 16MiB, expressed in 512 Bytes */
blkcfg.max_discard_sectors = 32768;
blkcfg.max_discard_seg = 1;
blkcfg.discard_sector_alignment = blk_size / 512;
}
if (bdev && spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_WRITE_ZEROES)) {
blkcfg.max_write_zeroes_sectors = 32768;
blkcfg.max_write_zeroes_seg = 1;
}
memcpy(config, &blkcfg, spdk_min(len, sizeof(blkcfg)));
return 0;
}
static const struct spdk_vhost_dev_backend vhost_blk_device_backend = {
.virtio_features = SPDK_VHOST_FEATURES |
(1ULL << VIRTIO_BLK_F_SIZE_MAX) | (1ULL << VIRTIO_BLK_F_SEG_MAX) |
(1ULL << VIRTIO_BLK_F_GEOMETRY) | (1ULL << VIRTIO_BLK_F_RO) |
(1ULL << VIRTIO_BLK_F_BLK_SIZE) | (1ULL << VIRTIO_BLK_F_TOPOLOGY) |
(1ULL << VIRTIO_BLK_F_BARRIER) | (1ULL << VIRTIO_BLK_F_SCSI) |
(1ULL << VIRTIO_BLK_F_FLUSH) | (1ULL << VIRTIO_BLK_F_CONFIG_WCE) |
(1ULL << VIRTIO_BLK_F_MQ) | (1ULL << VIRTIO_BLK_F_DISCARD) |
(1ULL << VIRTIO_BLK_F_WRITE_ZEROES),
.disabled_features = SPDK_VHOST_DISABLED_FEATURES | (1ULL << VIRTIO_BLK_F_GEOMETRY) |
(1ULL << VIRTIO_BLK_F_RO) | (1ULL << VIRTIO_BLK_F_FLUSH) | (1ULL << VIRTIO_BLK_F_CONFIG_WCE) |
(1ULL << VIRTIO_BLK_F_BARRIER) | (1ULL << VIRTIO_BLK_F_SCSI) | (1ULL << VIRTIO_BLK_F_DISCARD) |
(1ULL << VIRTIO_BLK_F_WRITE_ZEROES),
vhost/blk: separate sessions from devices With all the core vhost changes in place, we can refactor the upper layers now. We start with the vhost block since it's the easiest one. Vhost Block session specific fields were moved from the device struct into a new session struct. What's tricky, is that the blk-specific struct directly contains the generic session struct. This gives us handy access to the generic session data from the blk-specific session, and also allows us to directly upcast generic session objects. Most of the functions inside vhost block still accept vhost device as a parameter and then get the session object internally. Those functions will be refactored to accept session object directly in a separate patch since the amount of changes required is too big to be done here. Because of the above, some parts of this patch might seem overcomplicated. Especially the to_blk_session() funtion, which checks the device backend inside. The ultimate goal is to receive session object through start/stop callbacks - and for that the backend check does make sense. Change-Id: If222c31aec16a8cbe2d0cfb98c828e1ac75b91fc Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com> Reviewed-on: https://review.gerrithub.io/c/438678 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
2018-12-14 12:51:19 +00:00
.session_ctx_size = sizeof(struct spdk_vhost_blk_session) - sizeof(struct spdk_vhost_session),
.start_session = spdk_vhost_blk_start,
.stop_session = spdk_vhost_blk_stop,
.vhost_get_config = spdk_vhost_blk_get_config,
.dump_info_json = spdk_vhost_blk_dump_info_json,
.write_config_json = spdk_vhost_blk_write_config_json,
.remove_device = spdk_vhost_blk_destroy,
};
int
spdk_vhost_blk_controller_construct(void)
{
struct spdk_conf_section *sp;
unsigned ctrlr_num;
char *bdev_name;
char *cpumask;
char *name;
bool readonly;
for (sp = spdk_conf_first_section(NULL); sp != NULL; sp = spdk_conf_next_section(sp)) {
if (!spdk_conf_section_match_prefix(sp, "VhostBlk")) {
continue;
}
if (sscanf(spdk_conf_section_get_name(sp), "VhostBlk%u", &ctrlr_num) != 1) {
SPDK_ERRLOG("Section '%s' has non-numeric suffix.\n",
spdk_conf_section_get_name(sp));
return -1;
}
name = spdk_conf_section_get_val(sp, "Name");
if (name == NULL) {
SPDK_ERRLOG("VhostBlk%u: missing Name\n", ctrlr_num);
return -1;
}
cpumask = spdk_conf_section_get_val(sp, "Cpumask");
readonly = spdk_conf_section_get_boolval(sp, "ReadOnly", false);
bdev_name = spdk_conf_section_get_val(sp, "Dev");
if (bdev_name == NULL) {
continue;
}
if (spdk_vhost_blk_construct(name, cpumask, bdev_name, readonly) < 0) {
return -1;
}
}
return 0;
}
int
spdk_vhost_blk_construct(const char *name, const char *cpumask, const char *dev_name, bool readonly)
{
struct spdk_vhost_blk_dev *bvdev = NULL;
struct spdk_bdev *bdev;
uint64_t features = 0;
int ret = 0;
spdk_vhost_lock();
bdev = spdk_bdev_get_by_name(dev_name);
if (bdev == NULL) {
SPDK_ERRLOG("Controller %s: bdev '%s' not found\n",
name, dev_name);
ret = -ENODEV;
goto out;
}
bvdev = calloc(1, sizeof(*bvdev));
if (bvdev == NULL) {
ret = -ENOMEM;
goto out;
}
ret = spdk_bdev_open(bdev, true, bdev_remove_cb, bvdev, &bvdev->bdev_desc);
if (ret != 0) {
SPDK_ERRLOG("Controller %s: could not open bdev '%s', error=%d\n",
name, dev_name, ret);
goto out;
}
bvdev->bdev = bdev;
bvdev->readonly = readonly;
ret = spdk_vhost_dev_register(&bvdev->vdev, name, cpumask, &vhost_blk_device_backend);
if (ret != 0) {
spdk_bdev_close(bvdev->bdev_desc);
goto out;
}
if (spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_UNMAP)) {
features |= (1ULL << VIRTIO_BLK_F_DISCARD);
}
if (spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_WRITE_ZEROES)) {
features |= (1ULL << VIRTIO_BLK_F_WRITE_ZEROES);
}
if (readonly) {
features |= (1ULL << VIRTIO_BLK_F_RO);
}
if (spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_FLUSH)) {
features |= (1ULL << VIRTIO_BLK_F_FLUSH);
}
if (features && rte_vhost_driver_enable_features(bvdev->vdev.path, features)) {
SPDK_ERRLOG("Controller %s: failed to enable features 0x%"PRIx64"\n", name, features);
if (spdk_vhost_dev_unregister(&bvdev->vdev) != 0) {
SPDK_ERRLOG("Controller %s: failed to remove controller\n", name);
}
spdk_bdev_close(bvdev->bdev_desc);
ret = -1;
goto out;
}
SPDK_INFOLOG(SPDK_LOG_VHOST, "Controller %s: using bdev '%s'\n", name, dev_name);
out:
if (ret != 0 && bvdev) {
free(bvdev);
}
spdk_vhost_unlock();
return ret;
}
static int
spdk_vhost_blk_destroy(struct spdk_vhost_dev *vdev)
{
struct spdk_vhost_blk_dev *bvdev = to_blk_dev(vdev);
int rc;
if (!bvdev) {
return -EINVAL;
}
rc = spdk_vhost_dev_unregister(&bvdev->vdev);
if (rc != 0) {
return rc;
}
if (bvdev->bdev_desc) {
spdk_bdev_close(bvdev->bdev_desc);
bvdev->bdev_desc = NULL;
}
bvdev->bdev = NULL;
free(bvdev);
return 0;
}
SPDK_LOG_REGISTER_COMPONENT("vhost_blk", SPDK_LOG_VHOST_BLK)
SPDK_LOG_REGISTER_COMPONENT("vhost_blk_data", SPDK_LOG_VHOST_BLK_DATA)