Messages in this thread | | | From | Linus Torvalds <> | Date | Tue, 15 May 2012 18:53:33 -0700 | Subject | Re: [PATCH] vfs: fix IMA lockdep circular locking dependency |
| |
On Tue, May 15, 2012 at 5:45 PM, Linus Torvalds <torvalds@linux-foundation.org> wrote: > > Oops. Err.. I meant to do that. Right. To check that you're awake. > > Here.
Btw, a few notes:
- this adds the new security_mmap_file() calls to the same places that Mimi added the ima_file_mmap() call.
HOWEVER. There are other callers that call do_mmap[_pgoff]. Notably, do_shmat(). IMA probably doesn't care, but other security models might - even if the file in question there will be one just allocated by the shm layer.
So it does change semantics a bit, in that not only does the ->mmap_file() function no longer get called when 'file' is NULL, it also doesn't get called with a couple of special files (the other one I noticed was the i810 driver doing it's do_mmap() dance)
- the fact that we don't call the security function for a NULL-file mmap() is actually consistent with what do_brk() has always done (it used to do "addr_only=1", now it obviously does just security_mmap_addr() instead). So that part looks sane, and no security model can possibly care.
- I booted the patch. Not that I actually *tested* it any real way, but at least it boots and doesn't seem to break anything obvious.
Hmm?
Linus
| |