linux-kernel - Re: block: DMA alignment of IO buffer allocated from slab

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: <20180919100256.GD23172@ming.t460p>
Date:   Wed, 19 Sep 2018 18:02:57 +0800
From:   Ming Lei <ming.lei@...hat.com>
To:     Vitaly Kuznetsov <vkuznets@...hat.com>
Cc:     Ming Lei <tom.leiming@...il.com>,
        linux-block <linux-block@...r.kernel.org>,
        linux-mm <linux-mm@...ck.org>,
        Linux FS Devel <linux-fsdevel@...r.kernel.org>,
        "open list:XFS FILESYSTEM" <linux-xfs@...r.kernel.org>,
        Dave Chinner <dchinner@...hat.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Christoph Hellwig <hch@....de>, Jens Axboe <axboe@...nel.dk>
Subject: Re: block: DMA alignment of IO buffer allocated from slab

Hi Vitaly,

On Wed, Sep 19, 2018 at 11:41:07AM +0200, Vitaly Kuznetsov wrote:
> Ming Lei <tom.leiming@...il.com> writes:
> 
> > Hi Guys,
> >
> > Some storage controllers have DMA alignment limit, which is often set via
> > blk_queue_dma_alignment(), such as 512-byte alignment for IO buffer.
> 
> While mostly drivers use 512-byte alignment it is not a rule of thumb,
> 'git grep' tell me we have:
> ide-cd.c with 32-byte alignment
> ps3disk.c and rsxx/dev.c with variable alignment.
> 
> What if our block configuration consists of several devices (in raid
> array, for example) with different requirements, e.g. one requiring
> 512-byte alignment and the other requiring 256?

512-byte alignment is also 256-byte aligned, and the sector size is 512 byte.

> 
> >
> > Block layer now only checks if this limit is respected for buffer of
> > pass-through request,
> > see blk_rq_map_user_iov(), bio_map_user_iov().
> >
> > The userspace buffer for direct IO is checked in dio path, see
> > do_blockdev_direct_IO().
> > IO buffer from page cache should be fine wrt. this limit too.
> >
> > However, some file systems, such as XFS, may allocate single sector IO buffer
> > via slab. Usually I guess kmalloc-512 should be fine to return
> > 512-aligned buffer.
> > But once KASAN or other slab debug options are enabled, looks this
> > isn't true any
> > more, kmalloc-512 may not return 512-aligned buffer. Then data corruption
> > can be observed because the IO buffer from fs layer doesn't respect the DMA
> > alignment limit any more.
> >
> > Follows several related questions:
> >
> > 1) does kmalloc-N slab guarantee to return N-byte aligned buffer?  If
> > yes, is it a stable rule?
> >
> > 2) If it is a rule for kmalloc-N slab to return N-byte aligned buffer,
> > seems KASAN violates this
> > rule?
> 
> (as I was kinda involved in debugging): the issue was observed with SLUB
> allocator KASAN is not to blame, everything wich requires aditional
> metadata space will break this, see e.g. calculate_sizes() in slub.c

Buffer allocated via kmalloc() should be aligned with L1 HW cache size
at least.

I have raised the question: does kmalloc-512 slab guarantee to return
512-byte aligned buffer, let's see what the answer is from MM guys,:-) 

>From the Red Hat BZ, looks I understand this issue is only triggered when
KASAN is enabled, or you have figured out how to reproduce it without
KASAN involved?

> 
> >
> > 3) If slab can't guarantee to return 512-aligned buffer, how to fix
> > this data corruption issue?
> 
> I'm no expert in block layer but in case of complex block device
> configurations when bio submitter can't know all the requirements I see
> no other choice than bouncing.

I guess that might be the last straw, given the current way without
bouncing works for decades, and seems no one complains before.

Thanks,
Ming