From: Max Reitz <mreitz@redhat.com>
To: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>,
qemu-block@nongnu.org
Cc: Kevin Wolf <kwolf@redhat.com>, John Snow <jsnow@redhat.com>,
qemu-devel@nongnu.org
Subject: Re: [PATCH for-6.1? 4/6] job: Add job_cancel_requested()
Date: Mon, 26 Jul 2021 09:09:09 +0200 [thread overview]
Message-ID: <32463d65-834c-64ba-aab4-36f1810ddb43@redhat.com> (raw)
In-Reply-To: <a6a675f1-4ca3-a657-9fe0-35b73a6f441a@virtuozzo.com>
On 22.07.21 19:58, Vladimir Sementsov-Ogievskiy wrote:
> 22.07.2021 15:26, Max Reitz wrote:
>> Most callers of job_is_cancelled() actually want to know whether the job
>> is on its way to immediate termination. For example, we refuse to pause
>> jobs that are cancelled; but this only makes sense for jobs that are
>> really actually cancelled.
>>
>> A mirror job that is cancelled during READY with force=false should
>> absolutely be allowed to pause. This "cancellation" (which is actually
>> a kind of completion)
>
> You have to repeat that this "cancel" is not "cancel".
>
> So, the whole problem is that feature of mirror, on cancel in READY
> state do not cancel but do some specific kind of completion.
>
> You try to make this thing correctly handled on generic layer..
>
> Did you consider instead just drop the feature from generic layer? So
> that all *cancel* functions always do force-cancel. Then the internal
> implementation become a lot clearer.
Yes, I considered that, and I’ve decided against it (for now), because
such a change would obviously be an incompatible change. It would
require a deprecation period, and so we would need to fix this bug now
anyway.
> But we have to support the qmp block-job-cancel of READY mirror (and
> commit) with force=false.
>
> We can do it as an exclusion in qmp_block_job_cancel, something like:
>
> if (job is mirror or commit AND it's ready AND force = false)
> mirror_soft_cancel(...);
> else
> job_cancel(...);
I didn’t consider such a hack, though. I don’t like it. If we think
that we should change our approach because mirror’s soft cancel is
actually a completion mode, and the current situation is too confusing,
such a change should be user-visible, too. (I think there was this idea
of having job-specific flags or parameters you could change at runtime,
and so you’d just change the “pivot” parameter between true or false.)
Also, I don’t know whether this would really make anything “a lot”
easier. After this series job_is_cancelled() already tells the true
story, so all we could really change is to drop force_cancel and unify
the “s->should_complete || job_cancel_requested()” conditions in
block/mirror.c into one variable. So when I considered making cancel
exclusively force-cancel jobs, I thought it wouldn’t actually be worth
it in practice.
>> may take an indefinite amount of time, and so
>> should behave like any job during normal operation. For example, with
>> on-target-error=stop, the job should stop on write errors. (In
>> contrast, force-cancelled jobs should not get write errors, as they
>> should just terminate and not do further I/O.)
>>
>> Therefore, redefine job_is_cancelled() to only return true for jobs that
>> are force-cancelled (which as of HEAD^ means any job that interprets the
>> cancellation request as a request for immediate termination), and add
>> job_cancel_request() as the general variant, which returns true for any
>> jobs which have been requested to be cancelled, whether it be
>> immediately or after an arbitrarily long completion phase.
>>
>> Buglink: https://gitlab.com/qemu-project/qemu/-/issues/462
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>> ---
>
> [..]
>
>> --- a/job.c
>> +++ b/job.c
>> @@ -216,6 +216,11 @@ const char *job_type_str(const Job *job)
>> }
>> bool job_is_cancelled(Job *job)
>> +{
>> + return job->cancelled && job->force_cancel;
>> +}
>> +
>> +bool job_cancel_requested(Job *job)
>> {
>> return job->cancelled;
>> }
>> @@ -650,7 +655,7 @@ static void job_conclude(Job *job)
>> static void job_update_rc(Job *job)
>> {
>> - if (!job->ret && job_is_cancelled(job)) {
>> + if (!job->ret && job_cancel_requested(job)) {
>
> Why not job_is_cancelled() here?
>
> So in case of mirror other kind of completion we set ret to -ECANCELED?
I thought the return value is a user-visible thing, so I left it as-is.
Seems I was wrong, more below.
>> job->ret = -ECANCELED;
>> }
>> if (job->ret) {
>> @@ -704,7 +709,7 @@ static int job_finalize_single(Job *job)
>> /* Emit events only if we actually started */
>> if (job_started(job)) {
>> - if (job_is_cancelled(job)) {
>> + if (job_cancel_requested(job)) {
>> job_event_cancelled(job);
>
> Same question here.. Shouldn't mirror report COMPLETED event in case
> of not-force cancelled in READY state?
Same here, I thought this is user-visible, nothing internal, so I should
leave it as-is.
Now I see that cancelling mirror post-READY indeed should result in a
COMPLETED event. So I’m actually not exactly sure how mirror does that,
despite this code here (which functionally isn’t changed by this patch),
but it’s absolutely true that job_is_cancelled() would be more
appropriate here.
(No iotest failed, so I thought this change was right. Well.)
>> } else {
>> job_event_completed(job);
>> @@ -1015,7 +1020,7 @@ void job_complete(Job *job, Error **errp)
>> if (job_apply_verb(job, JOB_VERB_COMPLETE, errp)) {
>> return;
>> }
>> - if (job_is_cancelled(job) || !job->driver->complete) {
>> + if (job_cancel_requested(job) || !job->driver->complete) {
>> error_setg(errp, "The active block job '%s' cannot be
>> completed",
>> job->id);
>> return;
>> @@ -1043,7 +1048,7 @@ int job_finish_sync(Job *job, void
>> (*finish)(Job *, Error **errp), Error **errp)
>> AIO_WAIT_WHILE(job->aio_context,
>> (job_enter(job), !job_is_completed(job)));
>> - ret = (job_is_cancelled(job) && job->ret == 0) ? -ECANCELED :
>> job->ret;
>> + ret = (job_cancel_requested(job) && job->ret == 0) ? -ECANCELED
>> : job->ret;
So here it should probably stay a job_is_cancelled(), too.
Max
>> job_unref(job);
>> return ret;
>> }
>>
>
>
next prev parent reply other threads:[~2021-07-26 7:13 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-22 12:26 [PATCH for-6.1? 0/6] mirror: Handle errors after READY cancel Max Reitz
2021-07-22 12:26 ` [PATCH for-6.1? 1/6] mirror: Keep s->synced on error Max Reitz
2021-07-22 16:25 ` Vladimir Sementsov-Ogievskiy
2021-07-26 7:11 ` Max Reitz
2021-07-26 10:24 ` Vladimir Sementsov-Ogievskiy
2021-07-22 12:26 ` [PATCH for-6.1? 2/6] job: @force parameter for job_cancel_sync{, _all}() Max Reitz
2021-07-22 17:00 ` [PATCH for-6.1? 2/6] job: @force parameter for job_cancel_sync{,_all}() Vladimir Sementsov-Ogievskiy
2021-07-22 12:26 ` [PATCH for-6.1? 3/6] jobs: Give Job.force_cancel more meaning Max Reitz
2021-07-22 18:16 ` Vladimir Sementsov-Ogievskiy
2021-07-22 12:26 ` [PATCH for-6.1? 4/6] job: Add job_cancel_requested() Max Reitz
2021-07-22 17:58 ` Vladimir Sementsov-Ogievskiy
2021-07-26 7:09 ` Max Reitz [this message]
2021-07-27 14:47 ` Vladimir Sementsov-Ogievskiy
2021-07-27 15:30 ` Max Reitz
2021-07-22 12:26 ` [PATCH for-6.1? 5/6] mirror: Check job_is_cancelled() earlier Max Reitz
2021-07-22 12:26 ` [PATCH for-6.1? 6/6] iotests: Add mirror-ready-cancel-error test Max Reitz
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=32463d65-834c-64ba-aab4-36f1810ddb43@redhat.com \
--to=mreitz@redhat.com \
--cc=jsnow@redhat.com \
--cc=kwolf@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=vsementsov@virtuozzo.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).