From: Hanna Czenczek <hreitz@redhat.com>
To: qemu-block@nongnu.org
Cc: qemu-devel@nongnu.org, "Hanna Czenczek" <hreitz@redhat.com>,
"Kevin Wolf" <kwolf@redhat.com>,
"Stefan Hajnoczi" <stefanha@redhat.com>,
"Paolo Bonzini" <pbonzini@redhat.com>,
"Richard W . M . Jones" <rjones@redhat.com>,
"Ilya Dryomov" <idryomov@gmail.com>,
"Peter Lieven" <pl@dlhnet.de>,
"Philippe Mathieu-Daudé" <philmd@linaro.org>,
"Alex Bennée" <alex.bennee@linaro.org>,
"Fam Zheng" <fam@euphon.net>,
"Ronnie Sahlberg" <ronniesahlberg@gmail.com>
Subject: [PATCH 03/16] iscsi: Run co BH CB in the coroutine’s AioContext
Date: Tue, 28 Oct 2025 17:33:29 +0100 [thread overview]
Message-ID: <20251028163343.116249-4-hreitz@redhat.com> (raw)
In-Reply-To: <20251028163343.116249-1-hreitz@redhat.com>
For rbd (and others), as described in “rbd: Run co BH CB in the
coroutine’s AioContext”, the pattern of setting a completion flag and
waking a coroutine that yields while the flag is not set can only work
when both run in the same thread.
iscsi has the same pattern, but the details are a bit different:
iscsi_co_generic_cb() can (as far as I understand) only run through
iscsi_service(), not just from a random thread at a random time.
iscsi_service() in turn can only be run after iscsi_set_events() set up
an FD event handler, which is done in iscsi_co_wait_for_task().
As a result, iscsi_co_wait_for_task() will always yield exactly once,
because iscsi_co_generic_cb() can only run after iscsi_set_events(),
after the completion flag has already been checked, and the yielding
coroutine will then be woken only once the completion flag was set to
true. So as far as I can tell, iscsi has no bug and already works fine.
Still, we don’t need the completion flag because we know we have to
yield exactly once, so we can drop it. This simplifies the code and
makes it more obvious that the “rbd bug” isn’t present here.
This makes iscsi_co_generic_bh_cb() and iscsi_retry_timer_expired() a
bit boring, and actually, for the former, we could drop it and run
aio_co_wake() directly from scsi_co_generic_cb() to the same effect; but
that would remove the replay_bh_schedule_oneshot_event(), and I assume
we shouldn’t do that. At least schedule both the BH and the timer in
the coroutine’s AioContext to make them simple wrappers around
qemu_coroutine_enter(), without a further BH indirection.
Finally, remove the iTask->co != NULL checks: This field is set by
iscsi_co_init_iscsitask(), which all users of IscsiTask run before even
setting up iscsi_co_generic_cb() as the callback, and it is never set or
cleared elsewhere, so it is impossible to not be set in
iscsi_co_generic_cb().
Signed-off-by: Hanna Czenczek <hreitz@redhat.com>
---
block/iscsi.c | 39 ++++++++++-----------------------------
1 file changed, 10 insertions(+), 29 deletions(-)
diff --git a/block/iscsi.c b/block/iscsi.c
index 15b96ee880..76c15e20ea 100644
--- a/block/iscsi.c
+++ b/block/iscsi.c
@@ -107,7 +107,6 @@ typedef struct IscsiLun {
typedef struct IscsiTask {
int status;
- int complete;
int retries;
int do_retry;
struct scsi_task *task;
@@ -183,18 +182,13 @@ iscsi_schedule_bh(IscsiAIOCB *acb)
static void iscsi_co_generic_bh_cb(void *opaque)
{
struct IscsiTask *iTask = opaque;
-
- iTask->complete = 1;
aio_co_wake(iTask->co);
}
static void iscsi_retry_timer_expired(void *opaque)
{
struct IscsiTask *iTask = opaque;
- iTask->complete = 1;
- if (iTask->co) {
- aio_co_wake(iTask->co);
- }
+ aio_co_wake(iTask->co);
}
static inline unsigned exp_random(double mean)
@@ -239,6 +233,7 @@ iscsi_co_generic_cb(struct iscsi_context *iscsi, int status,
{
struct IscsiTask *iTask = opaque;
struct scsi_task *task = command_data;
+ AioContext *itask_ctx = qemu_coroutine_get_aio_context(iTask->co);
iTask->status = status;
iTask->do_retry = 0;
@@ -263,9 +258,9 @@ iscsi_co_generic_cb(struct iscsi_context *iscsi, int status,
" (retry #%u in %u ms): %s",
iTask->retries, retry_time,
iscsi_get_error(iscsi));
- aio_timer_init(iTask->iscsilun->aio_context,
- &iTask->retry_timer, QEMU_CLOCK_REALTIME,
- SCALE_MS, iscsi_retry_timer_expired, iTask);
+ aio_timer_init(itask_ctx, &iTask->retry_timer,
+ QEMU_CLOCK_REALTIME, SCALE_MS,
+ iscsi_retry_timer_expired, iTask);
timer_mod(&iTask->retry_timer,
qemu_clock_get_ms(QEMU_CLOCK_REALTIME) + retry_time);
iTask->do_retry = 1;
@@ -284,12 +279,7 @@ iscsi_co_generic_cb(struct iscsi_context *iscsi, int status,
}
}
- if (iTask->co) {
- replay_bh_schedule_oneshot_event(iTask->iscsilun->aio_context,
- iscsi_co_generic_bh_cb, iTask);
- } else {
- iTask->complete = 1;
- }
+ replay_bh_schedule_oneshot_event(itask_ctx, iscsi_co_generic_bh_cb, iTask);
}
static void coroutine_fn
@@ -592,12 +582,10 @@ static inline bool iscsi_allocmap_is_valid(IscsiLun *iscsilun,
static void coroutine_fn iscsi_co_wait_for_task(IscsiTask *iTask,
IscsiLun *iscsilun)
{
- while (!iTask->complete) {
- iscsi_set_events(iscsilun);
- qemu_mutex_unlock(&iscsilun->mutex);
- qemu_coroutine_yield();
- qemu_mutex_lock(&iscsilun->mutex);
- }
+ iscsi_set_events(iscsilun);
+ qemu_mutex_unlock(&iscsilun->mutex);
+ qemu_coroutine_yield();
+ qemu_mutex_lock(&iscsilun->mutex);
}
static int coroutine_fn
@@ -669,7 +657,6 @@ retry:
}
if (iTask.do_retry) {
- iTask.complete = 0;
goto retry;
}
@@ -740,7 +727,6 @@ retry:
scsi_free_scsi_task(iTask.task);
iTask.task = NULL;
}
- iTask.complete = 0;
goto retry;
}
@@ -902,7 +888,6 @@ retry:
}
if (iTask.do_retry) {
- iTask.complete = 0;
goto retry;
}
@@ -940,7 +925,6 @@ retry:
}
if (iTask.do_retry) {
- iTask.complete = 0;
goto retry;
}
@@ -1184,7 +1168,6 @@ retry:
}
if (iTask.do_retry) {
- iTask.complete = 0;
goto retry;
}
@@ -1301,7 +1284,6 @@ retry:
}
if (iTask.do_retry) {
- iTask.complete = 0;
goto retry;
}
@@ -2390,7 +2372,6 @@ retry:
iscsi_co_wait_for_task(&iscsi_task, dst_lun);
if (iscsi_task.do_retry) {
- iscsi_task.complete = 0;
goto retry;
}
--
2.51.0
next prev parent reply other threads:[~2025-10-28 16:39 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-10-28 16:33 [PATCH 00/16] block: Some multi-threading fixes Hanna Czenczek
2025-10-28 16:33 ` [PATCH 01/16] block: Note on aio_co_wake use if not yet yielding Hanna Czenczek
2025-10-28 16:33 ` [PATCH 02/16] rbd: Run co BH CB in the coroutine’s AioContext Hanna Czenczek
2025-10-28 16:33 ` Hanna Czenczek [this message]
2025-10-29 14:27 ` [PATCH 03/16] iscsi: " Kevin Wolf
2025-10-31 9:07 ` Hanna Czenczek
2025-10-31 13:27 ` Kevin Wolf
2025-10-28 16:33 ` [PATCH 04/16] nfs: " Hanna Czenczek
2025-10-28 16:33 ` [PATCH 05/16] curl: Fix coroutine waking Hanna Czenczek
2025-10-29 16:57 ` Kevin Wolf
2025-10-31 9:15 ` Hanna Czenczek
2025-10-31 13:17 ` Kevin Wolf
2025-10-28 16:33 ` [PATCH 06/16] gluster: Do not move coroutine into BDS context Hanna Czenczek
2025-10-29 17:10 ` Kevin Wolf
2025-10-31 9:16 ` Hanna Czenczek
2025-10-28 16:33 ` [PATCH 07/16] nvme: Kick and check completions in " Hanna Czenczek
2025-10-29 17:23 ` Kevin Wolf
2025-10-29 17:39 ` Kevin Wolf
2025-10-31 9:18 ` Hanna Czenczek
2025-10-31 9:19 ` Hanna Czenczek
2025-10-28 16:33 ` [PATCH 08/16] nvme: Fix coroutine waking Hanna Czenczek
2025-10-29 17:43 ` Kevin Wolf
2025-10-28 16:33 ` [PATCH 09/16] block/io: Take reqs_lock for tracked_requests Hanna Czenczek
2025-10-28 16:33 ` [PATCH 10/16] qcow2: Fix cache_clean_timer Hanna Czenczek
2025-10-29 20:23 ` Kevin Wolf
2025-10-31 9:29 ` Hanna Czenczek
2025-10-31 13:03 ` Kevin Wolf
2025-11-06 16:08 ` Hanna Czenczek
2025-10-28 16:33 ` [PATCH 11/16] ssh: Run restart_coroutine in current AioContext Hanna Czenczek
2025-10-28 16:33 ` [PATCH 12/16] blkreplay: Run BH in coroutine’s AioContext Hanna Czenczek
2025-10-28 16:33 ` [PATCH 13/16] block: Note in which AioContext AIO CBs are called Hanna Czenczek
2025-10-28 16:33 ` [PATCH 14/16] iscsi: Create AIO BH in original AioContext Hanna Czenczek
2025-10-28 16:33 ` [PATCH 15/16] null-aio: Run CB " Hanna Czenczek
2025-10-28 16:33 ` [PATCH 16/16] win32-aio: Run CB in original context Hanna Czenczek
2025-10-30 14:12 ` Kevin Wolf
2025-10-31 9:31 ` Hanna Czenczek
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=20251028163343.116249-4-hreitz@redhat.com \
--to=hreitz@redhat.com \
--cc=alex.bennee@linaro.org \
--cc=fam@euphon.net \
--cc=idryomov@gmail.com \
--cc=kwolf@redhat.com \
--cc=pbonzini@redhat.com \
--cc=philmd@linaro.org \
--cc=pl@dlhnet.de \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=rjones@redhat.com \
--cc=ronniesahlberg@gmail.com \
--cc=stefanha@redhat.com \
/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).