Messages in this thread | | | Date | Wed, 26 May 2010 08:41:08 -0700 (PDT) | From | Linus Torvalds <> | Subject | Re: [Regression] Crash in load_module() while freeing args |
| |
On Wed, 26 May 2010, Rusty Russell wrote: > > So then we have the problem that crc32 is finished its init and needs the > lock back, and bnx2x which needs crc32 is waiting for it. We could just > fail bnx2x; and in fact, we did prior to this patch (we timeout) and it breaks > network on booting on some box according to Brandon:
So the sane thing to do seems to simply not get that mutex_lock after a successful init. Which probably implies having a new lock for the actual low-level symbol table stuff.
Linus
| |