Merge lp://staging/~gandelman-a/ubuntu/oneiric/udev/lp818177 into lp://staging/ubuntu/oneiric/udev
Proposed by
Adam Gandelman
Status: | Work in progress |
---|---|
Proposed branch: | lp://staging/~gandelman-a/ubuntu/oneiric/udev/lp818177 |
Merge into: | lp://staging/ubuntu/oneiric/udev |
Diff against target: |
29 lines (+11/-0) 2 files modified
debian/changelog (+7/-0) debian/udev.initramfs-bottom (+4/-0) |
To merge this branch: | bzr merge lp://staging/~gandelman-a/ubuntu/oneiric/udev/lp818177 |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Steve Langasek | Disapprove | ||
Ubuntu branches | Pending | ||
Review via email: mp+73714@code.staging.launchpad.net |
Description of the change
This seems to fix Bug #818177 as well as other possibly related race conditions after init-bottom.
To post a comment you must log in.
Unmerged revisions
- 207. By Adam Gandelman
-
debian/
udev.initramfs- bottom: Allow udev settle before stopping it
(LP: #818177)
This is not a fix for the race conditions. It reduces the race window, but the fact that there's a race at all means that there's a more fundamental problem here. 'udevadm settle' only guarantees that currently- in-process events will be fully handled before it returns, but calling 'udevadm control --exit' should *already* provide this guarantee - and if it doesn't, all you've done is move the race, because when we do call 'udevadm control --exit' a few milliseconds later, any new events that are in process at *that* point will suffer from the same race.
We need to understand why 'udevadm control --exit' isn't sufficient and fix that at its source, because 'udevadm settle' doesn't give us the guarantees we need.