Re: locking question: do_mmap(), do_munmap()

[Alexander Viro <viro@math.psu.edu>]

On Mon, 11 Oct 1999, Stephen C. Tweedie wrote:

> Hi,
> 
> On Mon, 11 Oct 1999 17:40:52 -0400 (EDT), Alexander Viro
> <viro@math.psu.edu> said:
> 
> > Agreed, but the big lock does not (and IMHO should not) cover the vma list
> > modifications.
> 
> Fine, but as I've said you need _something_.  It doesn't matter what,
> but the fact that the kernel lock is no longer being held for vma
> updates has introduced swapper races.  We can't fix those without either
> restoring or replacing the big lock.

And spinlock being released in the ->swapout() is outright ugly. OK, so
we are adding to mm_struct a new semaphore (vma_sem) and getting it around
the places where the list is modified + in the swapper (for scanning). In
normal situation it will never give us contention - everyone except
swapper uses it with mmap_sem already held. Are there any objections
against it? If it's OK I'll go ahead and do it. Comments?

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/