d/numam-spdk
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

bdev/rbd: Poller can now handle channel deletion inline

Browse Source 
An I/O completion callback may delete the channel, so
restructure the poller to avoid touching channel memory
while triggering completions.

Change-Id: I612f10ff172481084386c9a3056fdd5e2f19e854
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/370526
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
This commit is contained in:
Ben Walker 2017-07-20 07:47:34 -07:00
parent 12b53038f6
commit b1617bf80f
1 changed files with 27 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

69
lib/bdev/rbd/bdev_rbd.c
View File

@ -48,13 +48,11 @@
#include "spdk_internal/bdev.h"
#define SPDK_RBD_QUEUE_DEPTH 128
static TAILQ_HEAD(, bdev_rbd) g_rbds = TAILQ_HEAD_INITIALIZER(g_rbds);
static int bdev_rbd_count = 0;
struct bdev_rbd_io {
rbd_completion_t completion;
};
struct bdev_rbd {
struct spdk_bdev disk;
char *rbd_name;
@ -68,8 +66,6 @@ struct bdev_rbd_io_channel {
rados_t cluster;
struct pollfd pfd;
rbd_image_t image;
rbd_completion_t *comps;
uint32_t queue_depth;
struct bdev_rbd *disk;
struct spdk_bdev_poller *poller;
};
@ -171,30 +167,30 @@ bdev_rbd_finish_aiocb(rbd_completion_t cb, void *arg)
}
static int
bdev_rbd_start_aio(rbd_image_t image, struct bdev_rbd_io *cmd,
bdev_rbd_start_aio(rbd_image_t image, struct spdk_bdev_io *bdev_io,
void *buf, uint64_t offset, size_t len)
{
int ret;
struct spdk_bdev_io *bdev_io = spdk_bdev_io_from_ctx(cmd);
rbd_completion_t comp;
ret = rbd_aio_create_completion((void *)cmd, bdev_rbd_finish_aiocb,
&cmd->completion);
ret = rbd_aio_create_completion(bdev_io, bdev_rbd_finish_aiocb,
&comp);
if (ret < 0) {
return -1;
}
if (bdev_io->type == SPDK_BDEV_IO_TYPE_READ) {
ret = rbd_aio_read(image, offset, len,
buf, cmd->completion);
buf, comp);
} else if (bdev_io->type == SPDK_BDEV_IO_TYPE_WRITE) {
ret = rbd_aio_write(image, offset, len,
buf, cmd->completion);
buf, comp);
} else if (bdev_io->type == SPDK_BDEV_IO_TYPE_FLUSH) {
ret = rbd_aio_flush(image, cmd->completion);
ret = rbd_aio_flush(image, comp);
}
if (ret < 0) {
rbd_aio_release(cmd->completion);
rbd_aio_release(comp);
return -1;
}
@ -207,7 +203,7 @@ static void bdev_rbd_library_fini(void);
static int
bdev_rbd_get_ctx_size(void)
{
return sizeof(struct bdev_rbd_io);
return 0;
}
SPDK_BDEV_MODULE_REGISTER(rbd, bdev_rbd_library_init, bdev_rbd_library_fini, NULL,
@ -215,7 +211,7 @@ SPDK_BDEV_MODULE_REGISTER(rbd, bdev_rbd_library_init, bdev_rbd_library_fini, NUL
static int64_t
bdev_rbd_readv(struct bdev_rbd *disk, struct spdk_io_channel *ch,
struct bdev_rbd_io *cmd, struct iovec *iov,
struct spdk_bdev_io *bdev_io, struct iovec *iov,
int iovcnt, size_t len, uint64_t offset)
{
struct bdev_rbd_io_channel *rbdio_ch = spdk_io_channel_get_ctx(ch);
@ -223,12 +219,12 @@ bdev_rbd_readv(struct bdev_rbd *disk, struct spdk_io_channel *ch,
if (iovcnt != 1 || iov->iov_len != len)
return -1;
return bdev_rbd_start_aio(rbdio_ch->image, cmd, iov->iov_base, offset, len);
return bdev_rbd_start_aio(rbdio_ch->image, bdev_io, iov->iov_base, offset, len);
}
static int64_t
bdev_rbd_writev(struct bdev_rbd *disk, struct spdk_io_channel *ch,
struct bdev_rbd_io *cmd, struct iovec *iov,
struct spdk_bdev_io *bdev_io, struct iovec *iov,
int iovcnt, size_t len, uint64_t offset)
{
struct bdev_rbd_io_channel *rbdio_ch = spdk_io_channel_get_ctx(ch);
@ -236,16 +232,16 @@ bdev_rbd_writev(struct bdev_rbd *disk, struct spdk_io_channel *ch,
if ((iovcnt != 1) || (iov->iov_len != len))
return -1;
return bdev_rbd_start_aio(rbdio_ch->image, cmd, (void *)iov->iov_base, offset, len);
return bdev_rbd_start_aio(rbdio_ch->image, bdev_io, (void *)iov->iov_base, offset, len);
}
static int64_t
bdev_rbd_flush(struct bdev_rbd *disk, struct spdk_io_channel *ch,
struct bdev_rbd_io *cmd, uint64_t offset, uint64_t nbytes)
struct spdk_bdev_io *bdev_io, uint64_t offset, uint64_t nbytes)
{
struct bdev_rbd_io_channel *rbdio_ch = spdk_io_channel_get_ctx(ch);
return bdev_rbd_start_aio(rbdio_ch->image, cmd, NULL, offset, nbytes);
return bdev_rbd_start_aio(rbdio_ch->image, bdev_io, NULL, offset, nbytes);
}
static int
@ -260,7 +256,7 @@ static void bdev_rbd_get_buf_cb(struct spdk_io_channel *ch, struct spdk_bdev_io
ret = bdev_rbd_readv(bdev_io->bdev->ctxt,
ch,
(struct bdev_rbd_io *)bdev_io->driver_ctx,
bdev_io,
bdev_io->u.read.iovs,
bdev_io->u.read.iovcnt,
bdev_io->u.read.len,
@ -281,7 +277,7 @@ static int _bdev_rbd_submit_request(struct spdk_io_channel *ch, struct spdk_bdev
case SPDK_BDEV_IO_TYPE_WRITE:
return bdev_rbd_writev((struct bdev_rbd *)bdev_io->bdev->ctxt,
ch,
(struct bdev_rbd_io *)bdev_io->driver_ctx,
bdev_io,
bdev_io->u.write.iovs,
bdev_io->u.write.iovcnt,
bdev_io->u.write.len,
@ -289,7 +285,7 @@ static int _bdev_rbd_submit_request(struct spdk_io_channel *ch, struct spdk_bdev
case SPDK_BDEV_IO_TYPE_FLUSH:
return bdev_rbd_flush((struct bdev_rbd *)bdev_io->bdev->ctxt,
ch,
(struct bdev_rbd_io *)bdev_io->driver_ctx,
bdev_io,
bdev_io->u.flush.offset,
bdev_io->u.flush.length);
default:
@ -323,9 +319,10 @@ static void
bdev_rbd_io_poll(void *arg)
{
struct bdev_rbd_io_channel *ch = arg;
struct bdev_rbd_io *req;
int i, io_status, rc;
rbd_completion_t comps[SPDK_RBD_QUEUE_DEPTH];
struct spdk_bdev_io *bdev_io;
int i, io_status, status, rc;
enum spdk_bdev_io_status status;
rc = poll(&ch->pfd, 1, 0);
@ -334,11 +331,10 @@ bdev_rbd_io_poll(void *arg)
return;
}
rc = rbd_poll_io_events(ch->image, ch->comps, ch->queue_depth);
rc = rbd_poll_io_events(ch->image, comps, SPDK_RBD_QUEUE_DEPTH);
for (i = 0; i < rc; i++) {
req = (struct bdev_rbd_io *)rbd_aio_get_arg(ch->comps[i]);
bdev_io = spdk_bdev_io_from_ctx(req);
io_status = rbd_aio_get_return_value(ch->comps[i]);
bdev_io = rbd_aio_get_arg(comps[i]);
io_status = rbd_aio_get_return_value(comps[i]);
if (bdev_io->type == SPDK_BDEV_IO_TYPE_READ) {
if ((int)bdev_io->u.read.len == io_status) {
status = SPDK_BDEV_IO_STATUS_SUCCESS;
@ -353,8 +349,8 @@ bdev_rbd_io_poll(void *arg)
status = SPDK_BDEV_IO_STATUS_FAILED;
}
}
rbd_aio_release(comps[i]);
spdk_bdev_io_complete(bdev_io, status);
rbd_aio_release(req->completion);
}
}
@ -377,10 +373,6 @@ bdev_rbd_free_channel(struct bdev_rbd_io_channel *ch)
rados_shutdown(ch->cluster);
}
if (ch->comps) {
free(ch->comps);
}
if (ch->pfd.fd >= 0) {
close(ch->pfd.fd);
}
@ -435,13 +427,6 @@ bdev_rbd_create_cb(void *io_device, void *ctx_buf)
goto err;
}
ch->queue_depth = 128;
ch->comps = calloc(sizeof(rbd_completion_t), ch->queue_depth);
if (!ch->comps) {
SPDK_ERRLOG("Failed to allocate rbd completion array\n");
goto err;
}
spdk_bdev_poller_start(&ch->poller, bdev_rbd_io_poll, ch,
spdk_env_get_current_core(), 0);