Messages in this thread | | | From | Linus Torvalds <> | Date | Thu, 29 Mar 2012 13:38:23 -0700 | Subject | Re: lockups shortly after booting in current git. |
| |
On Thu, Mar 29, 2012 at 1:26 PM, Dave Jones <davej@redhat.com> wrote: > > I'll go back to trying the bisect now that I know how to reproduce it quickly. > Do you think it might be worth restricting the bisect to fs/ ? > Or shall I just do the whole tree bisect from 3.3 ?
Despite my "it looks like writes only" theory it *might* be a libata issue (I assume this is a SATA drive), so I wouldn't force it to just fs/.
There are also patches in mm/, and who knows if it might be some locking or RCU change that triggers it too.
So I would suggest a whole-tree bisect. At least start out that way for a few rounds. It's painful when there has been 8000+ commits since v3.3, but it may be that as you bisect it becomes clear that it's some particular subsystem, and at *that* point maybe it's worth saying "don't bother bisecting through all the ARM changes", just start a new bisect with the known good/bad points and limit it to subsystem X.
After three bisections, you should be down to a thousand commits. And at that point it becomes worth looking at "ok, which subsystems seem to be represented in these 1000 commits" to see if you can narrow things down.
Linus
| |