Patchwork can't flush tlb on e500

login
register
mail settings
Submitter Hideo Saito
Date May 25, 2009, 1:33 a.m.
Message ID <20090525.103334.68564567.saito@densan.co.jp>
Download mbox | patch
Permalink /patch/27596/
State Accepted
Headers show

Comments

Hideo Saito - May 25, 2009, 1:33 a.m.
Hi all,

This report came from LKML, although I should post it here first,
the report said that there is a regression in linux-2.6.29 as to flushing a TLB entry that was going to be re-used when a context overflowed from the TLB.

On Fri, May 22, 2009 at 6:27 PM, Benjamin Herrenschmidt wrote:
> Have you verified that this change fixes your problem ?

I tested using "hackbench 100" and it was done successfully on my platform with e500(MPC8548).

>
> Can you re-submit to linuxppc-dev@ozlabs.org mailing list, along with
> proper changeset comment and signed-off-by: line ?

I am sorry I don't know as to the proper changeset comment.

Signed-off-by: Hideo Saito <hsaito.ppc@gmail.com>
---
Benjamin Herrenschmidt - May 25, 2009, 4:20 a.m.
On Mon, 2009-05-25 at 10:33 +0900, Hideo Saito wrote:
> 
> This report came from LKML, although I should post it here first,
> the report said that there is a regression in linux-2.6.29 as to
> flushing a TLB entry that was going to be re-used when a context
> overflowed from the TLB.
> 
> On Fri, May 22, 2009 at 6:27 PM, Benjamin Herrenschmidt wrote:
> > Have you verified that this change fixes your problem ?
> 
> I tested using "hackbench 100" and it was done successfully on my
> platform with e500(MPC8548).
> 
> >
> > Can you re-submit to linuxppc-dev@ozlabs.org mailing list, along
> with
> > proper changeset comment and signed-off-by: line ?
> 
> I am sorry I don't know as to the proper changeset comment.
> 
Well, that simply means a proper description of the bug and
the fix :-)

I'll cook up one.

Cheers,
Ben.

Patch

--- arch/powerpc/mm/mmu_context_nohash.c.orig	2009-03-24 08:12:14.000000000 +0900
+++ arch/powerpc/mm/mmu_context_nohash.c	2009-05-20 18:33:53.000000000 +0900
@@ -122,22 +122,22 @@  static unsigned int steal_context_up(uns
 	struct mm_struct *mm;
 	int cpu = smp_processor_id();
 
 	/* Pick up the victim mm */
 	mm = context_mm[id];
 
 	pr_debug("[%d] steal context %d from mm @%p\n", cpu, id, mm);
 
-	/* Mark this mm has having no context anymore */
-	mm->context.id = MMU_NO_CONTEXT;
-
 	/* Flush the TLB for that context */
 	local_flush_tlb_mm(mm);
 
+	/* Mark this mm has having no context anymore */
+	mm->context.id = MMU_NO_CONTEXT;
+
 	/* XXX This clear should ultimately be part of local_flush_tlb_mm */
 	__clear_bit(id, stale_map[cpu]);
 
 	return id;
 }
 
 #ifdef DEBUG_MAP_CONSISTENCY
 static void context_check_map(void)