RE: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot
I'm so sorry, forgive my mail client(outlook) I have try your solution, It doesn't work, still cause crash. The reason is: we come to bdrv_mirror_top_pwritev() (which means that mirror-top node is already inserted into block graph), but its bs->opaque->job is not initialized" But the root cause is that in block_job_create() we released(unnecessary) the aio_context, and the iothread get the context. Script has to part, one is run in the VM (to give some workload) we named script A: #!/bin/sh For((i=1;i<=1;i++)); Do dd if=/dev/zero of=./xxx bs=1M count=200 sleep 6 done Another one is in the hypervisor, we named script B: #!/bin/sh for((i=1;i<=1000;i++)); do virsh snapshot-create-as fqtest --atomic --no-metadata --name fq6 --disk-only --diskspec vda,snapshot=external,file=/home/michael/snapshot/fq6.qcow2; virsh blockcommit fqtest /home/michael/snapshot/fq6.qcow2 --shallow --verbose --wait --pivot --top /home/michael/snapshot/fq6.qcow2; rm -r fq6.qcow2 done How to reproduce: 1. start a VM, my case is use libvirt, named fqtest 2. run script B in hypervisor 3. after guest boot up, login and run script A in vda. Make sure, the IO thread enabled for vda. Mostly, just wait for several minutes, it will crash. The whole thread backtrace is: [Switching to Thread 0x7f7c7d91f700 (LWP 99907)] 0x5576d0f65aab in bdrv_mirror_top_pwritev at ../block/mirror.c:1437 1437../block/mirror.c: No such file or directory. (gdb) p s->job $17 = (MirrorBlockJob *) 0x0 (gdb) p s->stop $18 = false (gdb) bt #0 0x5576d0f65aab in bdrv_mirror_top_pwritev at ../block/mirror.c:1437 #1 0x5576d0f7f3ab in bdrv_driver_pwritev at ../block/io.c:1174 #2 0x5576d0f8139d in bdrv_aligned_pwritev at ../block/io.c:1988 #3 0x5576d0f81b65 in bdrv_co_pwritev_part at ../block/io.c:2156 #4 0x5576d0f8e6b7 in blk_do_pwritev_part at ../block/block-backend.c:1260 #5 0x5576d0f8e84d in blk_aio_write_entry at ../block/block-backend.c:1476 #6 0x5576d1060ddb in coroutine_trampoline at ../util/coroutine-ucontext.c:173 #7 0x7f7c8d3be0d0 in __start_context at /lib/../lib64/libc.so.6 #8 0x7f7b52beb1e0 in () #9 0x in () Switch to qemu main thread: #0 0x7f903be704ed in __lll_lock_wait at /lib/../lib64/libpthread.so.0 #1 0x7f903be6bde6 in _L_lock_941 at /lib/../lib64/libpthread.so.0 #2 0x7f903be6bcdf in pthread_mutex_lock at /lib/../lib64/libpthread.so.0 #3 0x564b21456889 in qemu_mutex_lock_impl at ../util/qemu-thread-posix.c:79 #4 0x564b213af8a5 in block_job_add_bdrv at ../blockjob.c:224 #5 0x564b213b00ad in block_job_create at ../blockjob.c:440 #6 0x564b21357c0a in mirror_start_job at ../block/mirror.c:1622 #7 0x564b2135a9af in commit_active_start at ../block/mirror.c:1867 #8 0x564b2133d132 in qmp_block_commit at ../blockdev.c:2768 #9 0x564b2141fef3 in qmp_marshal_block_commit at qapi/qapi-commands-block-core.c:346 #10 0x564b214503c9 in do_qmp_dispatch_bh at ../qapi/qmp-dispatch.c:110 #11 0x564b21451996 in aio_bh_poll at ../util/async.c:164 #12 0x564b2146018e in aio_dispatch at ../util/aio-posix.c:381 #13 0x564b2145187e in aio_ctx_dispatch at ../util/async.c:306 #14 0x7f9040239049 in g_main_context_dispatch at /lib/../lib64/libglib-2.0.so.0 #15 0x564b21447368 in main_loop_wait at ../util/main-loop.c:232 #16 0x564b21447368 in main_loop_wait at ../util/main-loop.c:255 #17 0x564b21447368 in main_loop_wait at ../util/main-loop.c:531 #18 0x564b212304e1 in qemu_main_loop at ../softmmu/runstate.c:721 #19 0x564b20f7975e in main at ../softmmu/main.c:50 Thanks, Michael -Original Message- From: Vladimir Sementsov-Ogievskiy Sent: 2021年2月1日 18:28 To: 08005...@163.com; kw...@redhat.com; mre...@redhat.com; js...@redhat.com Cc: 仇大玉 ; qemu-de...@nongnu.org; qemu-block@nongnu.org Subject: Re: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot Hi! Tanks for fixing and sorry for a delay! Please send each new version of a patch as a separate branch. It's a rule from https://wiki.qemu.org/Contribute/SubmitAPatch and it is more readable and less probable that your patch will be missed. 28.01.2021 04:30, 08005...@163.com wrote: > From: Michael Qiu > > v4: rebase to latest code > > v3: reformat the commit log, remove duplicate content > > v2: modify the coredump backtrace within commit log with the newest > qemu with master branch Such things shouldn't be in a commit message. You may put such comments after --- line[*] in a generated patch email > > Currently, if guest has workloads, IO thread will acquire aio_context > lock before do io_submit, it leads to segmentfault when do block > commit after snapshot. Just like below: Do you have some reproducer script? > > Program received signal SIGSEGV, Segmentation fault. > [Switching to Thread 0x7f7c7d91f700 (LWP 99907)] 0x
RE: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot
Any comments? It's really a bug and can cause the qemu to segmentfault. Thanks, Michael -Original Message- From: 仇大玉 Sent: 2021年1月28日 13:16 To: qemu-block@nongnu.org; qemu-de...@nongnu.org Cc: kw...@redhat.com; mre...@redhat.com; js...@redhat.com; 08005...@163.com Subject: RE: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot Any comments? -Original Message- From: 08005...@163.com <08005...@163.com> Sent: 2021年1月28日 9:31 To: kw...@redhat.com; mre...@redhat.com; js...@redhat.com Cc: qemu-block@nongnu.org; qemu-de...@nongnu.org; 仇大玉 Subject: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot From: Michael Qiu v4: rebase to latest code v3: reformat the commit log, remove duplicate content v2: modify the coredump backtrace within commit log with the newest qemu with master branch Currently, if guest has workloads, IO thread will acquire aio_context lock before do io_submit, it leads to segmentfault when do block commit after snapshot. Just like below: Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0x7f7c7d91f700 (LWP 99907)] 0x5576d0f65aab in bdrv_mirror_top_pwritev at ../block/mirror.c:1437 1437../block/mirror.c: No such file or directory. (gdb) p s->job $17 = (MirrorBlockJob *) 0x0 (gdb) p s->stop $18 = false (gdb) bt Switch to qemu main thread: /lib/../lib64/libpthread.so.0 /lib/../lib64/libpthread.so.0 ../util/qemu-thread-posix.c:79 qapi/qapi-commands-block-core.c:346 ../qapi/qmp-dispatch.c:110 /lib/../lib64/libglib-2.0.so.0 In IO thread when do bdrv_mirror_top_pwritev, the job is NULL, and stop field is false, this means the MirrorBDSOpaque "s" object has not been initialized yet, and this object is initialized by block_job_create(), but the initialize process is stuck in acquiring the lock. The rootcause is that qemu do release/acquire when hold the lock, at the same time, IO thread get the lock after release stage, and the crash occured. Actually, in this situation, job->job.aio_context will not equal to qemu_get_aio_context(), and will be the same as bs->aio_context, thus, no need to release the lock, becasue bdrv_root_attach_child() will not change the context. This patch fix this issue. Signed-off-by: Michael Qiu --- blockjob.c | 6 -- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/blockjob.c b/blockjob.c index 98ac8af982..51a09f3b60 100644 --- a/blockjob.c +++ b/blockjob.c @@ -214,13 +214,15 @@ int block_job_add_bdrv(BlockJob *job, const char *name, BlockDriverState *bs, BdrvChild *c; bdrv_ref(bs); -if (job->job.aio_context != qemu_get_aio_context()) { +if (bdrv_get_aio_context(bs) != job->job.aio_context && +job->job.aio_context != qemu_get_aio_context()) { aio_context_release(job->job.aio_context); } c = bdrv_root_attach_child(bs, name, _job, 0, job->job.aio_context, perm, shared_perm, job, errp); -if (job->job.aio_context != qemu_get_aio_context()) { +if (bdrv_get_aio_context(bs) != job->job.aio_context && +job->job.aio_context != qemu_get_aio_context()) { aio_context_acquire(job->job.aio_context); } if (c == NULL) { -- 2.22.0
RE: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot
Any comments? -Original Message- From: 08005...@163.com <08005...@163.com> Sent: 2021年1月28日 9:31 To: kw...@redhat.com; mre...@redhat.com; js...@redhat.com Cc: qemu-block@nongnu.org; qemu-de...@nongnu.org; 仇大玉 Subject: [PATCH v4] blockjob: Fix crash with IOthread when block commit after snapshot From: Michael Qiu v4: rebase to latest code v3: reformat the commit log, remove duplicate content v2: modify the coredump backtrace within commit log with the newest qemu with master branch Currently, if guest has workloads, IO thread will acquire aio_context lock before do io_submit, it leads to segmentfault when do block commit after snapshot. Just like below: Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0x7f7c7d91f700 (LWP 99907)] 0x5576d0f65aab in bdrv_mirror_top_pwritev at ../block/mirror.c:1437 1437../block/mirror.c: No such file or directory. (gdb) p s->job $17 = (MirrorBlockJob *) 0x0 (gdb) p s->stop $18 = false (gdb) bt Switch to qemu main thread: /lib/../lib64/libpthread.so.0 /lib/../lib64/libpthread.so.0 ../util/qemu-thread-posix.c:79 qapi/qapi-commands-block-core.c:346 ../qapi/qmp-dispatch.c:110 /lib/../lib64/libglib-2.0.so.0 In IO thread when do bdrv_mirror_top_pwritev, the job is NULL, and stop field is false, this means the MirrorBDSOpaque "s" object has not been initialized yet, and this object is initialized by block_job_create(), but the initialize process is stuck in acquiring the lock. The rootcause is that qemu do release/acquire when hold the lock, at the same time, IO thread get the lock after release stage, and the crash occured. Actually, in this situation, job->job.aio_context will not equal to qemu_get_aio_context(), and will be the same as bs->aio_context, thus, no need to release the lock, becasue bdrv_root_attach_child() will not change the context. This patch fix this issue. Signed-off-by: Michael Qiu --- blockjob.c | 6 -- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/blockjob.c b/blockjob.c index 98ac8af982..51a09f3b60 100644 --- a/blockjob.c +++ b/blockjob.c @@ -214,13 +214,15 @@ int block_job_add_bdrv(BlockJob *job, const char *name, BlockDriverState *bs, BdrvChild *c; bdrv_ref(bs); -if (job->job.aio_context != qemu_get_aio_context()) { +if (bdrv_get_aio_context(bs) != job->job.aio_context && +job->job.aio_context != qemu_get_aio_context()) { aio_context_release(job->job.aio_context); } c = bdrv_root_attach_child(bs, name, _job, 0, job->job.aio_context, perm, shared_perm, job, errp); -if (job->job.aio_context != qemu_get_aio_context()) { +if (bdrv_get_aio_context(bs) != job->job.aio_context && +job->job.aio_context != qemu_get_aio_context()) { aio_context_acquire(job->job.aio_context); } if (c == NULL) { -- 2.22.0