Skip to contentSkip to conversation

ttm_get_pages() will OOPS with highmem allocation

PATCH In review Jonathan Morton 2 participants 2 messages Nov 13, 2012 – Nov 14, 2012

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ttm_get_pages() will OOPS with highmem allocation

Reposting from this kernel bug:
https://bugzilla.kernel.org/show_bug.cgi?id=50241

I've tested the patch and it solves a highly repeatable OOPS with the CedarView driver that I'm porting.

Quoted 1 line

From 3552913d3a1c2d3c4ce158e91dc18cd12b522cb2 Mon Sep 17 00:00:00 2001

From: Yakui Zhao <[email protected]>
Date: Tue, 7 Feb 2012 15:33:13 -0500
Subject: [PATCH] Clear the ttm page allocated from high memory zone correctly

The TTM page can be allocated from high memory. In such case it is wrong to use the page_address(page) as the virtual address for the high memory page.

---
 drivers/gpu/drm/ttm/ttm_page_alloc.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)
drivers/gpu/drm/ttm/ttm_page_alloc.c +4 -1
index d948575..df976d9 100644
@@ -708,7 +708,10 @@ int ttm_get_pages(struct list_head *pages, int flags,
 	/* clear the pages coming from the pool if requested */
 	if (flags & TTM_PAGE_FLAG_ZERO_ALLOC) {
 		list_for_each_entry(p, pages, lru) {
-			clear_page(page_address(p));
+			if (PageHighMem(p))
+				clear_highpage(p);
+			else
+				clear_page(page_address(p));
 		}
 	}
 
Signature · 7 lines
1.7.7.6



-- 
From: Jonathan Morton
      [email protected]
Reviewed-by on v1 1/1
#1 lore raw
Quoted 27 lines · Jonathan Morton

Reposting from this kernel bug:
https://bugzilla.kernel.org/show_bug.cgi?id=50241

I've tested the patch and it solves a highly repeatable OOPS with the CedarView driver that I'm porting.

Quoted 1 line

From 3552913d3a1c2d3c4ce158e91dc18cd12b522cb2 Mon Sep 17 00:00:00 2001

From: Yakui Zhao <[email protected]>
Date: Tue, 7 Feb 2012 15:33:13 -0500
Subject: [PATCH] Clear the ttm page allocated from high memory zone correctly

The TTM page can be allocated from high memory. In such case it is wrong to use the page_address(page) as the virtual address for the high memory page.

---
  drivers/gpu/drm/ttm/ttm_page_alloc.c |    5 ++++-
  1 files changed, 4 insertions(+), 1 deletions(-)
drivers/gpu/drm/ttm/ttm_page_alloc.c +0 -0
index d948575..df976d9 100644
@@ -708,7 +708,10 @@ int ttm_get_pages(struct list_head *pages, int flags,
  	/* clear the pages coming from the pool if requested */

drivers/gpu/drm/ttm/ttm_page_alloc.c int ttm_get_pages(struct list_head *pages, int flags, · line 717

Quoted 10 lines
drivers/gpu/drm/ttm/ttm_page_alloc.c +4 -1
  	if (flags & TTM_PAGE_FLAG_ZERO_ALLOC) {
  		list_for_each_entry(p, pages, lru) {
-			clear_page(page_address(p));
+			if (PageHighMem(p))
+				clear_highpage(p);
+			else
+				clear_page(page_address(p));
  		}
  	}
 

This is a pretty bad bug. I think the only reason it hasn't been hit more often is that TTM_PAGE_FLAG_ZERO_ALLOC may not be used that often. Please resend this patch with

Thanks,
Thomas