Re: page fault fastpath patch v2: fix race conditions, stats for8,32 and 512 cpu SMP

[Christoph Lameter]

On Mon, 23 Aug 2004, Rajesh Venkatasubramanian wrote:

> So there are no other active thread (mm_user) other than the current
> exit_mmap() thread. This gives thread exclusion. So we don't need
> mm->mmap_sem.
>
> However, we have to lock out truncate()->zap_pmd_range(), rmap.c
> functions, and other places from walking the page tables while we
> are freeing the page tables in exit_mmap(). The page_table_lock in
> exit_mmap() provides that exclusion.
>
> That's my understanding. Correct me if I am wrong.

Correct. I just looked through the function and it unlinks the pgd before
unlocking page_table_lock. It frees the pgd tree content later.

Since no mm_user is active anymore no pte ops occur and therefore also
atomic pte operations do not need protection.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/