525483 - IPDL check for uses of deleted actors

Status: RESOLVED FIXED

(Core :: IPC, defect)

Description

[Benjamin Smedberg]

Spun off from bug 525454: in that bug, we're returning an invalid actor handle in a result union. IPDL didn't notice this in the callee or the caller, and probably should have noticed in both.

Comment 1

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

What does "invalid" refer to in this bug?

In IPDL, there are two types of bad actor ID:

  (1) id=0 for a non-nullable actor.  That means someone passed a NULL actor in where they shouldn't have (and is evidence of misbehavior on the "other" side, since passing in a NULL actor should have been caught there)
  (2) id != 0, but id is unknown.  Just an invalid handle.

(2) used to silently produce a NULL actor, but that bug was fixed in bug 524220.

The other bad behavior in bug 525454 was passing an already-delete'd actor to an IPDL method.  IPDL doesn't catch this currently, but valgrind would.

Comment 2

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

I'll do this by adding a special "freed actor ID", in addition to id=0, the "NULL actor" ID.

Comment 3

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Attachment: v1

Comment 4

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Comment on attachment 409404 [details] [diff] [review]
v1

Pretty minor, no need to waste review time on

Comment 5

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Pushed http://hg.mozilla.org/projects/electrolysis/rev/82549dbf71d8 to reduce merge conflicts in later patches.  Have a local followup patch, will push that when shmem is ready to land.

Comment 6

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Sorry ^^^ was meant for bug 523175.

Comment 7

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Pushed http://hg.mozilla.org/projects/electrolysis/rev/1da06de15940