Messages in this thread | | | Date | Thu, 7 Jan 2010 17:35:14 -0800 (PST) | From | Linus Torvalds <> | Subject | Re: [GIT PULL] Please pull NFS client bugfixes.... |
| |
On Thu, 7 Jan 2010, Linus Torvalds wrote: > > However, I still suspect that the lock inversion problem can probably be > fixed without any of that at all. Maybe you can just break the chain > somewhere else. I've not actually seen the lockdep chain, so I don't know > the deails. Pointers?
For example, if we're talking about readdir having a lock the other way (NFS lock taken before the mmap_sem), it's entirely possible that there is nothing to "fix" but some lockdep annotation.
You cannot mmap a directory (and you can't readdir a non-directory), so if it's a per-inode NFS lock, then the simplest fix _might_ be to just put directory locks in a different lockdep class from non-directory locks. That might fix it.
Of course, if it's not a per-inode lock, that doesn't help.
And maybe I'm missing something entirely, and such games with lockdep classes are pointless, and I'm a moron. Quite possible.
Linus
| |