From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Subject: [Qemu-devel] [PULL 32/33] nbd: make it thread-safe, fix qcow2 over nbd
Date: Thu, 1 Jun 2017 14:41:50 +0200 [thread overview]
Message-ID: <1496320911-51305-33-git-send-email-pbonzini@redhat.com> (raw)
In-Reply-To: <1496320911-51305-1-git-send-email-pbonzini@redhat.com>
NBD is not thread safe, because it accesses s->in_flight without
a CoMutex. Fixing this will be required for multiqueue.
CoQueue doesn't have spurious wakeups but, when another coroutine can
run between qemu_co_queue_next's wakeup and qemu_co_queue_wait's
re-locking of the mutex, the wait condition can become false and
a loop is necessary.
In fact, it turns out that the loop is necessary even without this
multi-threaded scenario. A particular sequence of coroutine wakeups
is happening ~80% of the time when starting a guest with qcow2 image
served over NBD (i.e. qemu-nbd --format=raw, and QEMU's -drive option
has -format=qcow2). This patch fixes that issue too.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
block/nbd-client.c | 30 +++++++++---------------------
1 file changed, 9 insertions(+), 21 deletions(-)
diff --git a/block/nbd-client.c b/block/nbd-client.c
index 09d955b..87d19c7 100644
--- a/block/nbd-client.c
+++ b/block/nbd-client.c
@@ -119,6 +119,10 @@ static int nbd_co_send_request(BlockDriverState *bs,
int rc, ret, i;
qemu_co_mutex_lock(&s->send_mutex);
+ while (s->in_flight == MAX_NBD_REQUESTS) {
+ qemu_co_queue_wait(&s->free_sema, &s->send_mutex);
+ }
+ s->in_flight++;
for (i = 0; i < MAX_NBD_REQUESTS; i++) {
if (s->recv_coroutine[i] == NULL) {
@@ -181,20 +185,6 @@ static void nbd_co_receive_reply(NBDClientSession *s,
}
}
-static void nbd_coroutine_start(NBDClientSession *s,
- NBDRequest *request)
-{
- /* Poor man semaphore. The free_sema is locked when no other request
- * can be accepted, and unlocked after receiving one reply. */
- if (s->in_flight == MAX_NBD_REQUESTS) {
- qemu_co_queue_wait(&s->free_sema, NULL);
- assert(s->in_flight < MAX_NBD_REQUESTS);
- }
- s->in_flight++;
-
- /* s->recv_coroutine[i] is set as soon as we get the send_lock. */
-}
-
static void nbd_coroutine_end(BlockDriverState *bs,
NBDRequest *request)
{
@@ -202,13 +192,16 @@ static void nbd_coroutine_end(BlockDriverState *bs,
int i = HANDLE_TO_INDEX(s, request->handle);
s->recv_coroutine[i] = NULL;
- s->in_flight--;
- qemu_co_queue_next(&s->free_sema);
/* Kick the read_reply_co to get the next reply. */
if (s->read_reply_co) {
aio_co_wake(s->read_reply_co);
}
+
+ qemu_co_mutex_lock(&s->send_mutex);
+ s->in_flight--;
+ qemu_co_queue_next(&s->free_sema);
+ qemu_co_mutex_unlock(&s->send_mutex);
}
int nbd_client_co_preadv(BlockDriverState *bs, uint64_t offset,
@@ -226,7 +219,6 @@ int nbd_client_co_preadv(BlockDriverState *bs, uint64_t offset,
assert(bytes <= NBD_MAX_BUFFER_SIZE);
assert(!flags);
- nbd_coroutine_start(client, &request);
ret = nbd_co_send_request(bs, &request, NULL);
if (ret < 0) {
reply.error = -ret;
@@ -256,7 +248,6 @@ int nbd_client_co_pwritev(BlockDriverState *bs, uint64_t offset,
assert(bytes <= NBD_MAX_BUFFER_SIZE);
- nbd_coroutine_start(client, &request);
ret = nbd_co_send_request(bs, &request, qiov);
if (ret < 0) {
reply.error = -ret;
@@ -291,7 +282,6 @@ int nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
request.flags |= NBD_CMD_FLAG_NO_HOLE;
}
- nbd_coroutine_start(client, &request);
ret = nbd_co_send_request(bs, &request, NULL);
if (ret < 0) {
reply.error = -ret;
@@ -316,7 +306,6 @@ int nbd_client_co_flush(BlockDriverState *bs)
request.from = 0;
request.len = 0;
- nbd_coroutine_start(client, &request);
ret = nbd_co_send_request(bs, &request, NULL);
if (ret < 0) {
reply.error = -ret;
@@ -342,7 +331,6 @@ int nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset, int count)
return 0;
}
- nbd_coroutine_start(client, &request);
ret = nbd_co_send_request(bs, &request, NULL);
if (ret < 0) {
reply.error = -ret;
--
1.8.3.1
next prev parent reply other threads:[~2017-06-01 12:42 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-01 12:41 [Qemu-devel] [PULL 00/33] Misc patches for 2017-06-01 Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 01/33] mc146818rtc: update periodic timer only if it is needed Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 02/33] mc146818rtc: precisely count the clock for periodic timer Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 03/33] mc146818rtc: ensure LOST_TICK_POLICY_SLEW is only enabled on TARGET_I386 Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 04/33] mc146818rtc: drop unnecessary '#ifdef TARGET_I386' Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 05/33] mc146818rtc: embrace all x86 specific code Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 06/33] kvm: irqchip: trace changes on msi add/remove Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 07/33] msix: trace control bit write op Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 08/33] kvm: irqchip: skip update msi when disabled Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 09/33] Check the return value of fcntl in qemu_set_cloexec Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 10/33] nbd: strict nbd_wr_syncv Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 11/33] nbd: read_sync and friends: return 0 on success Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 12/33] nbd: add errp parameter to nbd_wr_syncv() Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 13/33] nbd: add errp to read_sync, write_sync and drop_sync Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 14/33] nbd/client.c: use errp instead of LOG Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 15/33] exec: simplify phys_page_find() params Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 16/33] virtio-scsi: Unset hotplug handler when unrealize Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 17/33] vhost-user-scsi: Introduce vhost-user-scsi host device Paolo Bonzini
2017-06-05 16:28 ` Eric Blake
2017-06-05 16:38 ` Felipe Franciosi
2017-06-05 16:41 ` Eric Blake
2017-06-05 16:50 ` Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 18/33] vhost-user-scsi: Introduce a vhost-user-scsi sample application Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 19/33] target/i386: enable A20 automatically in system management mode Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 20/33] target/i386: use multiple CPU AddressSpaces Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 21/33] i386: fix read/write cr with icount option Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 22/33] sockets: improve error reporting if UNIX socket path is too long Paolo Bonzini
2017-06-13 16:10 ` Peter Maydell
2017-06-14 8:05 ` Daniel P. Berrange
2017-06-01 12:41 ` [Qemu-devel] [PULL 23/33] exec: fix address_space_get_iotlb_entry page mask Paolo Bonzini
2017-06-02 9:59 ` Peter Xu
2017-06-01 12:41 ` [Qemu-devel] [PULL 24/33] nbd: Fully initialize client in case of failed negotiation Paolo Bonzini
2017-06-01 15:15 ` Eric Blake
2017-06-01 12:41 ` [Qemu-devel] [PULL 25/33] qtest: add rtc periodic timer test Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 26/33] kvmclock: update system_time_msr address forcibly Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 27/33] linuxboot_dma: compile for i486 Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 28/33] edu: fix memory leak on msi_broken platforms Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 29/33] i386/kvm: do not zero out segment flags if segment is unusable or not present Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 30/33] target/i386: Add GDB XML description for SSE registers Paolo Bonzini
2017-06-01 12:41 ` [Qemu-devel] [PULL 31/33] hw/core: nmi.c can be compiled as common-obj nowadays Paolo Bonzini
2017-06-01 12:41 ` Paolo Bonzini [this message]
2017-06-01 12:41 ` [Qemu-devel] [PULL 33/33] kvm: don't register smram_listener when smm is off Paolo Bonzini
2017-06-01 14:26 ` [Qemu-devel] [PULL 00/33] Misc patches for 2017-06-01 no-reply
2017-06-01 15:09 ` no-reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1496320911-51305-33-git-send-email-pbonzini@redhat.com \
--to=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).