tree 76d6ff5c3fd89d1626dde935e7e695b9124cfc81
parent 8a3e1e7aa85a26c43e4de938ceb40a0a488d761a
author Thomas Gales <tgales@google.com> 1702061908 +0000
committer releases-prod-builder <releases-prod-builder@fuchsia-infra.iam.gserviceaccount.com> 1702314634 +0000

[cobalt] Avoid deadlock on DO_NOT_COLLECT

On deletion, Cobalt gets stuck for 10 minutes until the timeout in
WaitUntilSave gets triggered. The deadlock occurs because initial-thread
is holding a lock on data_mutex_ so no new calls to GetMetricAggregate
will return until after the deletion is written to disk, but
writeback_thread_ also wants to lock data_mutex_ when it saves the
changes to disk.

There's no reason inital-thread needs to wait for the save to disk, so
we can get rid of this problem by not calling WaitUntilSave.

Fixed: b/314009944
Tested: Set policy to DO_NOT_COLLECT on smart display, deadlock not seen
Tested: `./cobaltb.py test`
Change-Id: Ic94c6ea96bcd4325cd35b238b66f005336cb7583
Reviewed-on: https://fuchsia-review.googlesource.com/c/cobalt/+/953958
Commit-Queue: Thomas Gales <tgales@google.com>
Reviewed-by: Cameron Dale <camrdale@google.com>
(cherry picked from commit abd793554459fb58d042ab4fe9e2476338a24475)