| Message ID | 20180907161520.26349-5-kwolf@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Fix some jobs/drain/aio_poll related hangs | expand |
On Fri, 09/07 18:15, Kevin Wolf wrote: > job_finish_sync() needs to release the AioContext lock of the job before > calling aio_poll(). Otherwise, callbacks called by aio_poll() would > possibly take the lock a second time and run into a deadlock with a > nested AIO_WAIT_WHILE() call. > > Also, job_drain() without aio_poll() isn't necessarily enough to make > progress on a job, it could depend on bottom halves to be executed. > > Combine both open-coded while loops into a single AIO_WAIT_WHILE() call > that solves both of these problems. > > Signed-off-by: Kevin Wolf <kwolf@redhat.com> > --- > job.c | 14 ++++++-------- > 1 file changed, 6 insertions(+), 8 deletions(-) > > diff --git a/job.c b/job.c > index 9ad0b7476a..8480eda188 100644 > --- a/job.c > +++ b/job.c > @@ -29,6 +29,7 @@ > #include "qemu/job.h" > #include "qemu/id.h" > #include "qemu/main-loop.h" > +#include "block/aio-wait.h" > #include "trace-root.h" > #include "qapi/qapi-events-job.h" > > @@ -998,6 +999,7 @@ void job_defer_to_main_loop(Job *job, JobDeferToMainLoopFn *fn, void *opaque) > int job_finish_sync(Job *job, void (*finish)(Job *, Error **errp), Error **errp) > { > Error *local_err = NULL; > + AioWait dummy_wait = {}; > int ret; > > job_ref(job); > @@ -1010,14 +1012,10 @@ int job_finish_sync(Job *job, void (*finish)(Job *, Error **errp), Error **errp) > job_unref(job); > return -EBUSY; > } > - /* job_drain calls job_enter, and it should be enough to induce progress > - * until the job completes or moves to the main thread. */ > - while (!job->deferred_to_main_loop && !job_is_completed(job)) { > - job_drain(job); > - } > - while (!job_is_completed(job)) { > - aio_poll(qemu_get_aio_context(), true); > - } > + > + AIO_WAIT_WHILE(&dummy_wait, job->aio_context, > + (job_drain(job), !job_is_completed(job))); The condition expression would read more elegant if job_drain() returns progress. Reviewed-by: Fam Zheng <famz@redhat.com> > + > ret = (job_is_cancelled(job) && job->ret == 0) ? -ECANCELED : job->ret; > job_unref(job); > return ret; > -- > 2.13.6 >
diff --git a/job.c b/job.c index 9ad0b7476a..8480eda188 100644 --- a/job.c +++ b/job.c @@ -29,6 +29,7 @@ #include "qemu/job.h" #include "qemu/id.h" #include "qemu/main-loop.h" +#include "block/aio-wait.h" #include "trace-root.h" #include "qapi/qapi-events-job.h" @@ -998,6 +999,7 @@ void job_defer_to_main_loop(Job *job, JobDeferToMainLoopFn *fn, void *opaque) int job_finish_sync(Job *job, void (*finish)(Job *, Error **errp), Error **errp) { Error *local_err = NULL; + AioWait dummy_wait = {}; int ret; job_ref(job); @@ -1010,14 +1012,10 @@ int job_finish_sync(Job *job, void (*finish)(Job *, Error **errp), Error **errp) job_unref(job); return -EBUSY; } - /* job_drain calls job_enter, and it should be enough to induce progress - * until the job completes or moves to the main thread. */ - while (!job->deferred_to_main_loop && !job_is_completed(job)) { - job_drain(job); - } - while (!job_is_completed(job)) { - aio_poll(qemu_get_aio_context(), true); - } + + AIO_WAIT_WHILE(&dummy_wait, job->aio_context, + (job_drain(job), !job_is_completed(job))); + ret = (job_is_cancelled(job) && job->ret == 0) ? -ECANCELED : job->ret; job_unref(job); return ret;
job_finish_sync() needs to release the AioContext lock of the job before calling aio_poll(). Otherwise, callbacks called by aio_poll() would possibly take the lock a second time and run into a deadlock with a nested AIO_WAIT_WHILE() call. Also, job_drain() without aio_poll() isn't necessarily enough to make progress on a job, it could depend on bottom halves to be executed. Combine both open-coded while loops into a single AIO_WAIT_WHILE() call that solves both of these problems. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- job.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-)