Objecter: failed assert(tick_event==NULL) at osdc/Objecter.cc

[wonzhq]

The assert failure is caused by a locking issue. When the Objecter timer
erases the tick_event from its events queue and calls tick() to dispatch
it, if the Objecter::rwlock is held by shutdown(), it waits there to get
the rwlock. However, inside the shutdown function, it checks the
tick_event and tries to cancel it. The cancel_event function returns
false since tick_event is already removed from the events queue. Thus
tick_event is not set to NULL, and leads to this assertion failure.

Fix issue #11183.

Signed-off-by: Zhiqiang Wang zhiqiang.wang@intel.com

[loic-bot]

SUCCESS: the output of run-make-check.sh on centos-7 for a9ecdd6 is http://paste2.org/PfVcHjZZ

Sent from GH.

[tchaikov]

@wonzhq and could you please tag this commit with Fixes: #11183 instead of Fix issue #11183, or if you are sure that the later can be understood by tracker?

[wonzhq]

After digging more into this issue, looks like this is a giant specfic issue, and has been fixed in the current master by removing the following code in Objecter::tick().

  if (!initialized.read())
    return;

[tchaikov]

After digging more into this issue, looks like this is a giant specfic issue

agreed. timer.shutdown() ensures that the callback gets called, and the time_event always reset itself even initialized is set to 0. hence assert(tick_event == NULL) holds.

[tchaikov]

i am labeling this PR with milestone of 'giant', @wonzhq , probably you can create another PR targeting giant instead?

[wonzhq]

You mean create another PR and push it to the giant branch?

[tchaikov]

@wonzhq yes, not sure there is better way to re-target a PR to another branch though...

BTW, i bisected the commits, turns out the fix was introduced in 8253ead and the buggy part was introduced in d790833. we might want to backport (some of) it to giant ?

[wonzhq]

@tchaikov , I created #4175. Pls help take a look. I think simply removing 2 lines of code would be able to fix this problem in giant.