block/io: Delay decrementing the quiesce_counter
v2 cover block/io: Delay decrementing the quiesce_counter
See patch 1 for the code-based explanation, and patch 2 for a case where this bites in practice.
v2:
- Remove test artifacts that remained from testing the test (Sorry...)
git backport-diff against v1:
Key:
[----] : patches are identical
[####] : number of functional differences between upstream/downstream patch [down] : patch is downstream-only
The flags [FC] indicate (F)unctional and (C)ontextual differences, respectively
001/2:[----] [--] 'block/io: Delay decrementing the quiesce_counter' 002/2:[0004] [FC] 'iotests: Test cancelling a job and closing the VM'
Max Reitz (2):
block/io: Delay decrementing the quiesce_counter
iotests: Test cancelling a job and closing the VM
block/io.c | 3 ++- tests/qemu-iotests/255 | 52 ++++++++++++++++++++++++++++++++++++++ tests/qemu-iotests/255.out | 24 ++++++++++++++++++ 3 files changed, 78 insertions(+), 1 deletion(-)
Signature · 1 line
2.21.0
PATCH v2 1/2 block/io: Delay decrementing the quiesce_counter
When ending a drained section, bdrv_do_drained_end() currently first decrements the quiesce_counter, and only then actually ends the drain.
The bdrv_drain_invoke(bs, false) call may cause graph changes. Say the graph change involves replacing an existing BB's ("blk") BDS (blk_bs(blk)) by @bs. Let us introducing the following values:
- bs_oqc = old_quiesce_counter
(so bs->quiesce_counter == bs_oqc - 1)
- obs_qc = blk_bs(blk)->quiesce_counter (before bdrv_drain_invoke())
Let us assume there is no blk_pthread_unthrottled() involved, so blk->quiesce_counter == obs_qc (before bdrv_drain_invoke()).
Now replacing blk_bs(blk) by @bs will reduce blk->quiesce_counter by obs_qc (making it 0) and increase it by bs_oqc-1 (making it bs_oqc-1).
bdrv_drain_invoke() returns and we invoke bdrv_parent_drained_end(). This will decrement blk->quiesce_counter by one, so it would be -1 -- were there not an assertion against that in blk_root_drained_end().
We therefore have to keep the quiesce_counter up at least until bdrv_drain_invoke() returns, so that bdrv_parent_drained_end() does the right thing for the parents @bs got during bdrv_drain_invoke().
But let us delay it even further, namely until bdrv_parent_drained_end() returns, because then it mirrors bdrv_do_drained_begin(): There, we first increment the quiesce_counter, then begin draining the parents, and then call bdrv_drain_invoke(). It makes sense to let bdrv_do_drained_end() unravel this exactly in reverse.
--- block/io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
block/io.c +2 -1
@@ -422,11 +422,12 @@ static void bdrv_do_drained_end(BlockDriverState *bs, bool recursive,return ; }assert (bs->quiesce_counter >0 ); - old_quiesce_counter =atomic_fetch_dec (&bs->quiesce_counter);/* Re-enable things in child-to-parent order */ bdrv_drain_invoke (bs, false);bdrv_parent_drained_end (bs, parent, ignore_bds_parents); + + old_quiesce_counter =atomic_fetch_dec (&bs->quiesce_counter);if (old_quiesce_counter ==1 ) {aio_enable_external (bdrv_get_aio_context (bs)); }
Signature · 1 line
2.21.0
PATCH v2 2/2 iotests: Test cancelling a job and closing the VM
In reply to Max ReitzThis patch adds a test where we cancel a throttled mirror job and immediately close the VM before it can be cancelled. Doing so will invoke bdrv_drain_all() while the mirror job tries to drain the throttled node. When bdrv_drain_all_end() tries to lift its drain on the throttle node, the job will exit and replace the current root node of the BB drive0 (which is the job's filter node) by the throttle node. Before the previous patch, this replacement did not increase drive0's quiesce_counter by a sufficient amount, so when bdrv_parent_drained_end() (invoked by bdrv_do_drained_end(), invoked by bdrv_drain_all_end()) tried to end the drain on all of the throttle node's parents, it decreased drive0's quiesce_counter below 0 -- which fails an assertion.
--- tests/qemu-iotests/255 | 52 ++++++++++++++++++++++++++++++++++++++ tests/qemu-iotests/255.out | 24 ++++++++++++++++++ 2 files changed, 76 insertions(+)
tests/qemu-iotests/255 +52 -0
@@ -35,6 +35,10 @@ def blockdev_create(vm, options): vm.run_job('job0') iotests.log("") +iotests.log('Finishing a commit job with background reads') +iotests.log('============================================') +iotests.log('') + with iotests.FilePath('t.qcow2') as disk_path, \ iotests.FilePath('t.qcow2.mid') as mid_path, \ iotests.FilePath('t.qcow2.base') as base_path, \
@@ -81,3 +85,51 @@ with iotests.FilePath('t.qcow2') as disk_path, \ auto_dismiss=True) vm.shutdown() + +iotests.log('') +iotests.log('Closing the VM while a job is being cancelled') +iotests.log('=============================================') +iotests.log('') + +with iotests.FilePath('src.qcow2') as src_path, \ + iotests.FilePath('dst.qcow2') as dst_path, \ + iotests.VM() as vm: + + iotests.log('=== Create images and start VM ===') + iotests.log('') + + size = 128 * 1024 * 1024 + size_str = str(size) + + iotests.qemu_img_log('create', '-f', iotests.imgfmt, src_path, size_str) + iotests.qemu_img_log('create', '-f', iotests.imgfmt, dst_path, size_str) + + iotests.log(iotests.qemu_io('-f', iotests.imgfmt, '-c', 'write 0 1M', + src_path), + filters=[iotests.filter_test_dir, iotests.filter_qemu_io]) + + vm.add_object('throttle-group,x-bps-read=4096,id=throttle0') + + vm.add_blockdev('file,node-name=src-file,filename=%s' % (src_path)) + vm.add_blockdev('%s,node-name=src,file=src-file' % (iotests.imgfmt)) + + vm.add_blockdev('file,node-name=dst-file,filename=%s' % (dst_path)) + vm.add_blockdev('%s,node-name=dst,file=dst-file' % (iotests.imgfmt)) + + vm.add_blockdev('throttle,node-name=src-throttled,' + + 'throttle-group=throttle0,file=src') + + vm.add_device('virtio-blk,drive=src-throttled') + + vm.launch() + + iotests.log('=== Start a mirror job ===') + iotests.log('') + + vm.qmp_log('blockdev-mirror', job_id='job0', device='src-throttled', + target='dst', sync='full') + + vm.qmp_log('block-job-cancel', device='job0') + vm.qmp_log('quit') + + vm.shutdown()
tests/qemu-iotests/255.out +24 -0
@@ -1,3 +1,6 @@ +Finishing a commit job with background reads +============================================ + === Create backing chain and start VM === Formatting 'TEST_DIR/PID-t.qcow2.mid', fmt=qcow2 size=134217728 cluster_size=65536 lazy_refcounts=off refcount_bits=16
@@ -14,3 +17,24 @@ Formatting 'TEST_DIR/PID-t.qcow2', fmt=qcow2 size=134217728 cluster_size=65536 l {"return": {}} {"data": {"id": "job0", "type": "commit"}, "event": "BLOCK_JOB_PENDING", "timestamp": {"microseconds": "USECS", "seconds": "SECS"}} {"data": {"device": "job0", "len": 134217728, "offset": 134217728, "speed": 0, "type": "commit"}, "event": "BLOCK_JOB_COMPLETED", "timestamp": {"microseconds": "USECS", "seconds": "SECS"}} + +Closing the VM while a job is being cancelled +============================================= + +=== Create images and start VM === + +Formatting 'TEST_DIR/PID-src.qcow2', fmt=qcow2 size=134217728 cluster_size=65536 lazy_refcounts=off refcount_bits=16 + +Formatting 'TEST_DIR/PID-dst.qcow2', fmt=qcow2 size=134217728 cluster_size=65536 lazy_refcounts=off refcount_bits=16 + +wrote 1048576/1048576 bytes at offset 0 +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) + +=== Start a mirror job === + +{"execute": "blockdev-mirror", "arguments": {"device": "src-throttled", "job-id": "job0", "sync": "full", "target": "dst"}} +{"return": {}} +{"execute": "block-job-cancel", "arguments": {"device": "job0"}} +{"return": {}} +{"execute": "quit", "arguments": {}} +{"return": {}}
Signature · 1 line
2.21.0
Kevin Wolf
AppliedAm 22.05.2019 um 16:40 hat Max Reitz geschrieben:
Quoted 2 lines
See patch 1 for the code-based explanation, and patch 2 for a case where this bites in practice.
Thanks, applied to the block branch.
Kevin