nfs mmap vs i_mutex inversion

nfs mmap vs i_mutex inversion

[Peter Zijlstra]

Hi Trond,

About that lock order inversion I remembered, I still have the patches:

http://programming.kicks-ass.net/kernel-patches/mmap-vs-nfs/

Although they would want updating a little.

But looking at a recent (.27-rc5-mm1) the offending code path:

down_write(&mm->mmap_sem)
...
->nfs_file_mmap()
    nfs_revalidate_mapping()
      nfs_invalidate_mapping()
        mutex_lock(&inode->i_mutex);


vs the regular order of

  inode->i_mutex
    mm->mmap_sem

as described in mm/rmap.c

Seems to still exist.

If you agree its a valid concern, I'll brush up the patches and repost.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>