linux-kernel - Re: [PATCH v2] nvme-tcp: Check if request has started before processing it

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20210517145805.end32zjhqfjh6kga@beryllium.lan>
Date:   Mon, 17 May 2021 16:58:05 +0200
From:   Daniel Wagner <dwagner@...e.de>
To:     Sagi Grimberg <sagi@...mberg.me>
Cc:     Keith Busch <kbusch@...nel.org>, Hannes Reinecke <hare@...e.de>,
        "Ewan D. Milne" <emilne@...hat.com>,
        linux-nvme@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Jens Axboe <axboe@...com>, Christoph Hellwig <hch@....de>
Subject: Re: [PATCH v2] nvme-tcp: Check if request has started before
 processing it

Hi Sagi,

On Fri, May 07, 2021 at 04:22:30PM -0700, Sagi Grimberg wrote:
> Yea, maybe something like this?

I did give this a spin with blktests (loopback device) and on real
hardware (FC). Seems to do work fine. Just two things.

> +/*
> + * nvme command_id is constructed as such:
> + * | xxxx | xxxxxxxxxxxx |
> + *   gen    request tag
> + */
> +#define nvme_cid_install_genctr(gen)           ((gen & 0xf) << 12)
> +#define nvme_genctr_from_cid(cid)              ((cid & 0xf000) >> 12)
> +#define nvme_tag_from_cid(cid)                 (cid & 0xfff)
> +
> +static inline u16 nvme_cid(struct request *rq)
> +{
> +       return nvme_cid_install_genctr(nvme_req(rq)->genctr++) | rq->tag;
> +}

-       return nvme_cid_install_genctr(nvme_req(rq)->genctr++) | rq->tag;
+       nvme_req(rq)->genctr = ++nvme_req(rq)->genctr & 0xf;
+       return nvme_cid_install_genctr(nvme_req(rq)->genctr) | rq->tag;

The first issue, it really needs prefix increment if you want to write
it in one line. And it should store only the first 4 bits.
nvme_find_rq() would complain with 0x0 != 0x10 after the first overflow.

> +static inline struct request *nvme_find_rq(struct blk_mq_tags *tags,
> +               u16 command_id)
> +{
> +       u8 genctr = nvme_genctr_from_cid(command_id);
> +       u16 tag = nvme_tag_from_cid(command_id);
> +       struct request *rq;
> +
> +       rq = blk_mq_tag_to_rq(tags, tag);
> +       if (unlikely(!rq)) {
> +               pr_err("could not locate request for tag %#x\n",
> +                       tag);
> +               return NULL;
> +       }
> +       if (unlikely(nvme_req(rq)->genctr != genctr)) {
> +               dev_err(nvme_req(rq)->ctrl->device,
> +                       "request %#x genctr mismatch (got %#x expected
> %#x)\n",
> +                       tag, nvme_req(rq)->genctr, genctr);

-                       tag, nvme_req(rq)->genctr, genctr);
+                       tag, genctr, nvme_req(rq)->genctr);

The arguments are in the wrong order. Got me a bit confused.

Are you going to send out a proper patch? I'd like to move things
forward and could offer to do some more testing if needed.

Thanks,
Daniel