Re: [PATCH 00/15] iotests: add enhanced debugging info to qemu-io failures

[Hanna Reitz]

On 18.03.22 21:36, John Snow wrote:
> Howdy,

Heya,

[...]

> - Uh, actually, test 040 fails with this patchset and I don't understand
>    if it's intentional, harmless, a test design problem, or worse:
>
> ======================================================================
> ERROR: test_filterless_commit (__main__.TestCommitWithFilters)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>    File "/home/jsnow/src/qemu/tests/qemu-iotests/040", line 822, in tearDown
>      self.do_test_io('read')
>    File "/home/jsnow/src/qemu/tests/qemu-iotests/040", line 751, in do_test_io
>      qemu_io('-f', iotests.imgfmt,
>    File "/home/jsnow/src/qemu/tests/qemu-iotests/iotests.py", line 365, in 
> qemu_io
>      return qemu_tool(*qemu_io_wrap_args(args),
>    File "/home/jsnow/src/qemu/tests/qemu-iotests/iotests.py", line 242, in 
> qemu_tool
>      raise VerboseProcessError(
>
> qemu.utils.VerboseProcessError: Command
>    '('/home/jsnow/src/qemu/bin/git/tests/qemu-iotests/../../qemu-io',
>    '--cache', 'writeback', '--aio', 'threads', '-f', 'qcow2', '-c',
>    'read -P 4 3M 1M',
>    '/home/jsnow/src/qemu/bin/git/tests/qemu-iotests/scratch/3.img')'
>    returned non-zero exit status 1.
>    ┏━ output ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
>    ┃ qemu-io: can't open device
>    ┃ /home/jsnow/src/qemu/bin/git/tests/qemu-iotests/scratch/3.img:
>    ┃ Could not open backing file: Could not open backing file: Throttle
>    ┃ group 'tg' does not exist
>    ┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
>
> It looks like we start with the img chain 3->2->1->0, then we commit 2
> down into 1, but checking '3' fails somewhere in the backing
> chain. Maybe a real bug?

Looks like my hunch was right: The problem is that it’s hard to figure 
out a good backing file string when there are filters involved, and so 
in one test here we generate one that contains a JSON description of the 
backing subgraph including a throttle node. Outside of qemu, that 
doesn’t make much sense, hence the error.

(And because we checked only for “pattern verification failed” 
specifically, that error here never surfaced.)

I think (hope?) we can expect management tools to manually specify 
backing file strings in such cases, like the attached diff does. That 
seems to fix the problem.

Hanna

040.diff
Description: Text Data