[PATCH] 2.6.0-test10 O_DIRECT memory leak fix

[PATCH] 2.6.0-test10 O_DIRECT memory leak fix

[Badari Pulavarty]

[-- Attachment #1: Type: text/plain, Size: 431 bytes --]

Hi Andrew,

I found the problem with O_DIRECT memory leak.

The problem is, when we are doing DIO read and crossed the end of file -
we don't release referencess on all the pages we got from get_user_pages().
(since it is a sucess case).

The fix is to call dio_cleanup() even for sucess cases. Can you please
review this ? I tested the patch with fsstress and there are no memory leak 
problems now.

Thanks,
Badari

[-- Attachment #2: dioleak.patch --]
[-- Type: text/x-diff, Size: 550 bytes --]

--- linux-2.6.0-test10/fs/direct-io.c	2003-11-25 12:45:14.000000000 -0800
+++ linux-2.6.0-test10.new/fs/direct-io.c	2003-11-25 15:20:26.554239616 -0800
@@ -946,6 +946,12 @@ direct_io_worker(int rw, struct kiocb *i
 	if (dio->bio)
 		dio_bio_submit(dio);
 
+	/* 
+	 * It is possible that, we return short IO due to end of file.
+	 * In that case, we need to release all the pages we got hold on.
+	 */ 
+	dio_cleanup(dio);
+
 	/*
 	 * OK, all BIOs are submitted, so we can decrement bio_count to truly
 	 * reflect the number of to-be-processed BIOs.

Re: [PATCH] 2.6.0-test10 O_DIRECT memory leak fix

[Andrew Morton]

Badari Pulavarty <pbadari@us.ibm.com> wrote:
>
> Hi Andrew,
> 
> I found the problem with O_DIRECT memory leak.
> 
> The problem is, when we are doing DIO read and crossed the end of file -
> we don't release referencess on all the pages we got from get_user_pages().
> (since it is a sucess case).
> 
> The fix is to call dio_cleanup() even for sucess cases. Can you please
> review this ? I tested the patch with fsstress and there are no memory leak 
> problems now.

Thanks, it looks OK.  Those pages haven't even been put into BIOs so I see
no problem releasing them here even if there is still async I/O in flight.

Re: [PATCH] 2.6.0-test10 O_DIRECT memory leak fix

[IWAMOTO Toshihiro]

At Tue, 25 Nov 2003 14:49:50 -0800,
Badari Pulavarty wrote:
> 
> [1  <text/plain; us-ascii (quoted-printable)>]
> Hi Andrew,
> 
> I found the problem with O_DIRECT memory leak.
> 
> The problem is, when we are doing DIO read and crossed the end of file -
> we don't release referencess on all the pages we got from get_user_pages().
> (since it is a sucess case).
> 
> The fix is to call dio_cleanup() even for sucess cases. Can you please
> review this ? I tested the patch with fsstress and there are no memory leak 
> problems now.
> 
> Thanks,
> Badari
> [2 dioleak.patch <text/x-diff; us-ascii (7bit)>]

I've tested this patch with linux-2.6.0-test11 (plus my memory
hotplug patch) on the machine I observed the memory leak.
The leak seems to be fixed.  Thanks.

--
IWAMOTO Toshihiro