Date: Fri, 26 Jan 2024 07:56:39 +0100
From: Hannes Reinecke <hare@...e.de>
To: Damien Le Moal <dlemoal@...nel.org>, Mikulas Patocka
 <mpatocka@...hat.com>, Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Tejun Heo <tj@...nel.org>, Thomas Gleixner <tglx@...utronix.de>,
 linux-kernel@...r.kernel.org, dm-devel@...ts.linux.dev,
 Mike Snitzer <msnitzer@...hat.com>, Ignat Korchagin <ignat@...udflare.com>,
 Damien Le Moal <damien.lemoal@....com>, Hou Tao <houtao1@...wei.com>,
 Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>
Subject: Re: [PATCH] softirq: fix memory corruption when freeing
 tasklet_struct

On 1/26/24 00:15, Damien Le Moal wrote:
> On 1/26/24 07:04, Mikulas Patocka wrote:
>>
>>
>> On Thu, 25 Jan 2024, Linus Torvalds wrote:
>>
>>> On Thu, 25 Jan 2024 at 10:30, Mikulas Patocka <mpatocka@...hat.com> wrote:
>>>>
>>>> There's a problem with the tasklet API - there is no reliable way how to
>>>> free a structure that contains tasklet_struct. The problem is that the
>>>> function tasklet_action_common calls task_unlock(t) after it called the
>>>> callback. If the callback does something that frees tasklet_struct,
>>>> task_unlock(t) would write into free memory.
>>>
>>> Ugh.
>>>
>>> I see what you're doing, but I have to say, I dislike this patch
>>> immensely. It feels like a serious misdesign that is then papered over
>>> with a hack.
>>>
>>> I'd much rather see us trying to move away from tasklets entirely in
>>> cases like this. Just say "you cannot do that".
>>
>> OK. I will delete tasklets from both dm-crypt and dm-verity - it will
>> simplify them quite a bit.
>>
>> BTW. Do you think that we should get rid of request-based device mapper as
>> well? (that's another thing that looks like code bloat to me)
> 
> That would force removing dm-multipath, which is I think the only DM driver
> using requests. But given how widespread the use of dm-multipath is, killing it
> would likely make a lot of people unhappy...
> 
Oh, it's this time of the year again?
(This topic regularly comes up ...)

The reason is not that it will disable dm-multipath (Mike Snitzer put in 
bio-based multipathing as an additional code path); the reason is that 
dm-multipath performance will suffer when you remove request-based DM.

DM-multipath schedules based on request (if you use the request-based 
interface) or bios (if you use the bio-based interface).
Any merge decision is typically done by the block layer when combining 
bios into requests; and you can only merge bios if the bvecs are adjacent.
So if you use bio-based multipathing you will spread sequential bios
across all paths, leaving the block layer unable to merge requests.
For request based multipathing the requests are already fully-formed,
and scheduling across paths does not change them.
Things are slightly better with multi-page bvecs nowadays, but the
overall picture still stands.

Another thing is timeouts; bios don't do timeouts, so a bio can run
for an arbitrary time with no chance of interrupting it.
Requests do have a timeout, and will be aborted from the driver when
the timeout is hit.
Seeing that 99% of all I/O issues I've seen _are_ timeouts it becomes
a crucial feature if you want dm-multipath to control failover time.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@...e.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Ivo Totev, Andrew McDonald,
Werner Knoblich

Powered by blists - more mailing lists