linux-kernel - Re: [PATCH 025 of 35] Treat rq->hard_nr_sectors as setting an overriding limit in the size of the request

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

Message-ID: <20070801174418.GI13674@htj.dyndns.org>
Date:	Thu, 2 Aug 2007 02:44:18 +0900
From:	Tejun Heo <htejun@...il.com>
To:	NeilBrown <neilb@...e.de>
Cc:	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 025 of 35] Treat rq->hard_nr_sectors as setting an overriding limit in the size of the request

On Tue, Jul 31, 2007 at 12:17:59PM +1000, NeilBrown wrote:
> 
> For a request to be able to refer to part of a bio, we need to be able
> to impose a size limit at the request level.  So allow hard_nr_sectors
> to be less than the size of the bios (and bio_vecs) and interpret it
> such that anything in the last bio beyond that limit is ignored.
> 
> As some bios can be less than one sector - as happens when a SCSI
> sense command is being submitted - we need to set hard_nr_sectors to
> bi_size rounded up in blk_rq_bio_prep, and we need to abort the
> rq_for_each_segment loop if _iter.bio becomes NULL even if _iter.size
> is still non-zero

This is pretty confusing.  In all other places, bi_size -> #sector
conversion is done by rounding down but only in blk_rq_bio_prep() it's
being rounded up.

Is my following reasoning correct?

It was okay till now because unaligned requests don't get merged and
also haven't done partial completions (end_that_request_first with
partial count)?  So till now, hard_nr_sectors and nr_sectors didn't
really matter for unaligned requests but now it matters because it's
considered while iterating over bvecs in rq.

If so, I think the correct thing to do would be changing bio_sectors()
to round up first or let block layer measure transfer in bytes not in
sectors.  I don't think everyone would agree with the latter tho.  I
(tentatively) think it would be better to represent length in bytes
tho.  A lot of requests which aren't aligned to 512 bytes pass through
the block layer and the mismatch can result in subtle bugs.

Thanks.

-- 
tejun
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/