[RESEND] [PATCH] fix invalidate_inode_pages2_range not to clear ret

[RESEND] [PATCH] fix invalidate_inode_pages2_range not to clear ret

[Hisashi Hifumi]

Hi Andrew.

DIO invalidates page cache through invalidate_inode_pages2_range().
invalidate_inode_pages2_range() sets ret=-EIO when invalidate_complete_page2()
fails, but this ret is cleared if do_launder_page() succeed on a page of next index. 
In this case, dio is carried out even if invalidate_complete_page2() fails on some pages.
This can cause inconsistency between memory and blocks on HDD because the page
cache still exists.

Following patch fixes this issue.

Thanks.

Signed-off-by :Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>

diff -Nrup linux-2.6.25-rc1.org/mm/truncate.c linux-2.6.25-rc1/mm/truncate.c
--- linux-2.6.25-rc1.org/mm/truncate.c	2008-02-12 13:37:41.000000000 +0900
+++ linux-2.6.25-rc1/mm/truncate.c	2008-02-15 11:24:33.000000000 +0900
@@ -392,6 +392,7 @@ int invalidate_inode_pages2_range(struct
 	pgoff_t next;
 	int i;
 	int ret = 0;
+	int ret2 = 0;
 	int did_range_unmap = 0;
 	int wrapped = 0;
 
@@ -439,9 +440,11 @@ int invalidate_inode_pages2_range(struct
 				}
 			}
 			BUG_ON(page_mapped(page));
-			ret = do_launder_page(mapping, page);
-			if (ret == 0 && !invalidate_complete_page2(mapping, page))
-				ret = -EIO;
+			ret2 = do_launder_page(mapping, page);
+			if (ret2 == 0 && !invalidate_complete_page2(mapping, page))
+				ret2 = -EIO;
+			if (ret2 < 0)
+				ret = ret2;
 			unlock_page(page);
 		}
 		pagevec_release(&pvec);

Re: [RESEND] [PATCH] fix invalidate_inode_pages2_range not to clear ret

[Jeff Moyer]

Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp> writes:

> Hi Andrew.
>
> DIO invalidates page cache through invalidate_inode_pages2_range().
> invalidate_inode_pages2_range() sets ret=-EIO when invalidate_complete_page2()
> fails, but this ret is cleared if do_launder_page() succeed on a page of next index. 
> In this case, dio is carried out even if invalidate_complete_page2() fails on some pages.
> This can cause inconsistency between memory and blocks on HDD because the page
> cache still exists.
>
> Following patch fixes this issue.

I agree with the patch (and I see Andrew already accepted it).  Do you
have a reproducer for this problem?

Cheers,

Jeff

Re: [RESEND] [PATCH] fix invalidate_inode_pages2_range not to clear ret

[Hisashi Hifumi]

Hi Jeff.

>
>I agree with the patch (and I see Andrew already accepted it).  Do you
>have a reproducer for this problem?

Do you want to know the way of failing invalidate_complete_page2() ?

Re: [RESEND] [PATCH] fix invalidate_inode_pages2_range not to clear ret

[Jeff Moyer]

Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp> writes:

> Hi Jeff.
>
>>
>>I agree with the patch (and I see Andrew already accepted it).  Do you
>>have a reproducer for this problem?
>
> Do you want to know the way of failing invalidate_complete_page2() ?

I was hoping for a test program that shows the problem.  Looking through
the previous threads on this subject, it seems that Zach may have had a
reproducer.  I tried running the aio-dio-regress tests on the 2.6.25-rc2
kernel, but was unable to trigger this problem.

-Jeff