Handle an interesting race condition (bug #1747367) #8385
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The static analysis showed that this might be a problem, and we're able to set up a case where it really does fail. Essentially, we have one loop which waits to try and send a notification of an event to users. While waiting for that event to be sent, it allows new requests to be made. So if we start watching for an event, but never handle when it triggers, and then queue up a bunch more watch events that cause us to reallocate the event slice, and *then* unwatch the original event, we will never notice that the original event is no longer watched because we'll be hung on an object that doesn't get updated.
|
!!build!! |
jameinel
force-pushed
the
2.3-presence-1747367
branch
from
49500bf to
ba26f5c
Compare
|
|
rogpeppe
reviewed
Feb 15, 2018
| @@ -353,6 +353,11 @@ func (w *Watcher) flush() { | |||
| return | |||
| case req := <-w.request: | |||
| w.handle(req) | |||
| // handle may append to the w.pending array, and/or it may unwatch something that was previously pending | |||
There was a problem hiding this comment.
ISTM that we're already handling the unwatch case (that's the e.ch != nil loop condition) but that the real issue here is the append, because the append can cause a realloc and thus e might be pointing to the old slice and hence be invalid.
I think I'd probably program this a little more defensively;
for {
// Note: w.pending can be appended to by w.handle so we can't
// safely keep the same e between iterations.
e := &w.pending[i]
if e.ch == nil {
break
}
select {
etc
?
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description of change
The static analysis showed that this might be a problem, and we're able to set
up a case where it really does fail.
Essentially, we have one loop which waits to try and send a notification of an
event to users. While waiting for that event to be sent, it allows new requests
to be made. So if we start watching for an event, but never handle when it
triggers, and then queue up a bunch more watch events that cause us to
reallocate the event slice, and then unwatch the original event, we will
never notice that the original event is no longer watched because we'll be hung
on an object that doesn't get updated.
QA steps
The test properly handles the error conditions. If you uncomment that one line in presence.go, it properly fails the test case (and fails after waiting LongWait).
Documentation changes
None.
Bug reference
lp:1747367