From: Hannes Reinecke <hare@suse.de>
To: Sagi Grimberg <sagi@grimberg.me>, Keith Busch <kbusch@kernel.org>
Cc: "Ewan D. Milne" <emilne@redhat.com>,
Daniel Wagner <dwagner@suse.de>,
linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
Jens Axboe <axboe@fb.com>, Christoph Hellwig <hch@lst.de>
Subject: Re: [PATCH v2] nvme-tcp: Check if request has started before processing it
Date: Thu, 6 May 2021 17:36:32 +0200 [thread overview]
Message-ID: <492b8393-fc35-f58a-3768-94632a083c93@suse.de> (raw)
In-Reply-To: <756aef10-e693-276f-82ac-514a2832b07f@grimberg.me>
On 4/1/21 12:37 AM, Sagi Grimberg wrote:
>
>>>>> It is, but in this situation, the controller is sending a second
>>>>> completion that results in a use-after-free, which makes the
>>>>> transport irrelevant. Unless there is some other flow (which is
>>>>> unclear
>>>>> to me) that causes this which is a bug that needs to be fixed rather
>>>>> than hidden with a safeguard.
>>>>>
>>>>
>>>> The kernel should not crash regardless of any network traffic that is
>>>> sent to the system. It should not be possible to either intentionally
>>>> of mistakenly contruct packets that will deny service in this way.
>>>
>>> This is not specific to nvme-tcp. I can build an rdma or pci controller
>>> that can trigger the same crash... I saw a similar patch from Hannes
>>> implemented in the scsi level, and not the individual scsi transports..
>>
>> If scsi wants this too, this could be made generic at the blk-mq level.
>> We just need to make something like blk_mq_tag_to_rq(), but return NULL
>> if the request isn't started.
>
> Makes sense...
>
>>> I would also mention, that a crash is not even the scariest issue that
>>> we can see here, because if the request happened to be reused we are
>>> in the silent data corruption realm...
>>
>> If this does happen, I think we have to come up with some way to
>> mitigate it. We're not utilizing the full 16 bits of the command_id, so
>> maybe we can append something like a generation sequence number that can
>> be checked for validity.
>
> That's actually a great idea. scsi needs unique tags so it encodes the
> hwq in the upper 16 bits giving the actual tag the lower 16 bits which
> is more than enough for a single queue. We can do the same with
> a gencnt that will increment in both submission and completion and we
> can validate against it.
>
> This will be useful for all transports, so maintaining it in
> nvme_req(rq)->genctr and introducing a helper like:
> rq = nvme_find_tag(tagset, cqe->command_id)
> That will filter genctr, locate the request.
>
> Also:
> nvme_validate_request_gen(rq, cqe->command_id) that would
> compare against it.
>
>
> And then a helper to set the command_id like:
> cmd->common.command_id = nvme_request_command_id(rq)
> that will both increment the genctr and build a command_id
> from it.
>
> Thoughts?
>
Well, that would require a modification to the CQE specification, no?
fmds was not amused when I proposed that :-(
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@suse.de +49 911 74053 688
SUSE Software Solutions Germany GmbH, 90409 Nürnberg
GF: F. Imendörffer, HRB 36809 (AG Nürnberg)
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
next prev parent reply other threads:[~2021-05-06 16:03 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-01 17:56 [PATCH v2] nvme-tcp: Check if request has started before processing it Daniel Wagner
2021-03-05 19:57 ` Sagi Grimberg
2021-03-11 9:43 ` Daniel Wagner
2021-03-15 17:16 ` Sagi Grimberg
2021-03-30 16:19 ` Ewan D. Milne
2021-03-30 17:34 ` Sagi Grimberg
2021-03-30 23:28 ` Keith Busch
2021-03-31 7:11 ` Hannes Reinecke
2021-03-31 21:01 ` Ewan D. Milne
2021-03-31 22:24 ` Sagi Grimberg
2021-04-01 6:20 ` Christoph Hellwig
2021-04-01 8:25 ` Sagi Grimberg
2021-03-31 22:37 ` Sagi Grimberg
2021-05-06 15:36 ` Hannes Reinecke [this message]
2021-05-07 20:26 ` Sagi Grimberg
2021-05-07 20:40 ` Keith Busch
2021-05-07 23:22 ` Sagi Grimberg
2021-05-08 0:03 ` Keith Busch
2021-05-09 11:30 ` Hannes Reinecke
2021-05-11 18:16 ` Sagi Grimberg
2021-05-17 14:58 ` Daniel Wagner
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=492b8393-fc35-f58a-3768-94632a083c93@suse.de \
--to=hare@suse.de \
--cc=axboe@fb.com \
--cc=dwagner@suse.de \
--cc=emilne@redhat.com \
--cc=hch@lst.de \
--cc=kbusch@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nvme@lists.infradead.org \
--cc=sagi@grimberg.me \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).