Re: [PATCH 1/1] ext4, dio: Remove overflow for size >2G in aio-dio code.

Re: [PATCH 1/1] ext4, dio: Remove overflow for size >2G in aio-dio code.

[Ted Ts'o]

On Mon, May 21, 2012 at 03:22:28PM -0700, Linus Torvalds wrote:
> 
> So I'm pretty sure this patch is what we want, and rw_verify_area()
> really is required to protect low-level filesystems from these kinds
> of issues. Not just ext4.
> 
> At the same time, I would *really* want somebody who actually uses
> anything AIO to test it out too. Because I want to not only commit it,
> but also mark it for stable - and it would be nice to have some more
> testing than me saying "ok, it passes the one test-case sent to me"
> and "hey, the code looks sane".

Hi Linus,

We do use AIO at Google, and I primarily use fio (usually as part of
xfstests) to test AIO functionality.

If you like I can carry the patch in the ext4 tree and test it as part
of the ext4 commits for the merge window.

I probably won't be able to get to it until a bit later in the week,
though.  Things have been really crazy for me this past two months...

						- Ted

Re: [PATCH 1/1] ext4, dio: Remove overflow for size >2G in aio-dio code.

[Eric Sandeen]

On 5/21/12 6:31 PM, Ted Ts'o wrote:
> On Mon, May 21, 2012 at 03:22:28PM -0700, Linus Torvalds wrote:
>>
>> So I'm pretty sure this patch is what we want, and rw_verify_area()
>> really is required to protect low-level filesystems from these kinds
>> of issues. Not just ext4.
>>
>> At the same time, I would *really* want somebody who actually uses
>> anything AIO to test it out too. Because I want to not only commit it,
>> but also mark it for stable - and it would be nice to have some more
>> testing than me saying "ok, it passes the one test-case sent to me"
>> and "hey, the code looks sane".

Just FWIW, the ext4_file_write() part of this fix was sent by Zheng Liu on 4/12/12:

[PATCH RESEND] ext4: change return value from int to ssize_t in ext4_file_write

I'll see about writing an xfstest for this stuff too so it doesn't regress again.

-Eric

> Hi Linus,
> 
> We do use AIO at Google, and I primarily use fio (usually as part of
> xfstests) to test AIO functionality.
> 
> If you like I can carry the patch in the ext4 tree and test it as part
> of the ext4 commits for the merge window.
> 
> I probably won't be able to get to it until a bit later in the week,
> though.  Things have been really crazy for me this past two months...
> 
> 						- Ted
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/1] ext4, dio: Remove overflow for size >2G in aio-dio code.

[Eric Sandeen]

On 5/22/12 11:11 AM, Eric Sandeen wrote:
> On 5/21/12 6:31 PM, Ted Ts'o wrote:
>> On Mon, May 21, 2012 at 03:22:28PM -0700, Linus Torvalds wrote:
>>>
>>> So I'm pretty sure this patch is what we want, and rw_verify_area()
>>> really is required to protect low-level filesystems from these kinds
>>> of issues. Not just ext4.
>>>
>>> At the same time, I would *really* want somebody who actually uses
>>> anything AIO to test it out too. Because I want to not only commit it,
>>> but also mark it for stable - and it would be nice to have some more
>>> testing than me saying "ok, it passes the one test-case sent to me"
>>> and "hey, the code looks sane".
> 
> Just FWIW, the ext4_file_write() part of this fix was sent by Zheng Liu on 4/12/12:
> 
> [PATCH RESEND] ext4: change return value from int to ssize_t in ext4_file_write

Sorry, I was talking about Manish's original patch, not Linus's followup.

-Eric

> I'll see about writing an xfstest for this stuff too so it doesn't regress again.
> 
> -Eric
> 

Re: [PATCH 1/1] ext4, dio: Remove overflow for size >2G in aio-dio code.

[manish honap]

Hello Linus, Ted



On Mon, May 21, 2012 at 03:22:28PM -0700, Linus Torvalds wrote:
> 
> So I'm pretty sure this patch is what we want, and rw_verify_area()
> really is required to protect low-level filesystems from these kinds
> of issues. Not just ext4.

Patch works on my machine also. 
> 
> At the same time, I would *really* want somebody who actually uses
> anything AIO to test it out too. Because I want to not only commit it,
> but also mark it for stable - and it would be nice to have some more
> testing than me saying "ok, it passes the one test-case sent to me"
> and "hey, the code looks sane".

Used Chris Mason's aio-stress.c

- Manish