lkml.org 
[lkml]   [2007]   [Apr]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [patch -mm] i386: use pte_update_defer in ptep_test_and_clear_{dirty,young}
Hugh Dickins wrote:
> Zach, while looking at your recent patches, I ran across the comment
> on pte_update_defer, and where it was being used, and now think that
> David's patch is actually incorrect. Previously pte_update_defer
> was being used where a flush_tlb_page followed immediately after
> within the same macro; with David's patch, mm's clear_refs_pte_range
> is calling ptep_test_and_clear_young (including pte_update_defer) on
> several ptes, then unlocking the page table, and later flushing TLB.
> That's exactly wrong for pte_update_defer, isn't it?
>

Ok, disregard most of my last e-mail. It is fine to decouple the flush
from the update, as long as they stay close enough that you can reason
they happen together. I guess I hadn't seen the other parts of the
patch which release the page table spinlock in between the two, and
somehow missed it again when responding to the above as I got too
excited explaining why the decoupling is ok. It is not ok to release
the spinlock when using shadow page tables on SMP. There are some
rather complex races that can result. Here's one case:

CPU-0 CPU-1
----------------------- ---------------------------
test_and_clear_dirty(x)
spin_unlock(ptl)
write address mapped by X
(harware updates dirty bit)
spin_lock(ptl)
set_pte_wrprotect(x)
flush
flush

Now, the write protected pte which maps a dirty page gets broken in two
ways; it is unclear if dirty bit or entiry PTE from CPU-0 is deferred
until flush, so either write protected PTE for modified page loses the
dirty bit (BAD!), or write protected PTE loses both dirty and write
protect bits (VERY BAD!).

To prevent this, we need a flush before dropping the spinlock. If that
gets too complicated, we can drop the defer logic and just use
pte_update instead, which notifies the hypervisor immediately of the
mapping change.

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

\
 
 \ /
  Last update: 2007-04-14 00:29    [W:0.194 / U:0.024 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site