linux-kernel - Re: [PATCH] dma-buf: mmap support

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]

Date:	Wed, 18 Apr 2012 09:20:26 -0500
From:	Rob Clark <rob.clark@...aro.org>
To:	Arnd Bergmann <arnd@...db.de>
Cc:	Daniel Vetter <daniel.vetter@...ll.ch>,
	linaro-mm-sig@...ts.linaro.org,
	LKML <linux-kernel@...r.kernel.org>,
	DRI Development <dri-devel@...ts.freedesktop.org>,
	linux-media@...r.kernel.org,
	Rebecca Schultz Zavin <rebecca@...roid.com>
Subject: Re: [PATCH] dma-buf: mmap support

On Wed, Apr 18, 2012 at 9:06 AM, Arnd Bergmann <arnd@...db.de> wrote:
> On Wednesday 18 April 2012, Daniel Vetter wrote:
>> +   Because existing importing subsystems might presume coherent mappings for
>> +   userspace, the exporter needs to set up a coherent mapping. If that's not
>> +   possible, it needs to fake coherency by manually shooting down ptes when
>> +   leaving the cpu domain and flushing caches at fault time. Note that all the
>> +   dma_buf files share the same anon inode, hence the exporter needs to replace
>> +   the dma_buf file stored in vma->vm_file with it's own if pte shootdown is
>> +   requred. This is because the kernel uses the underlying inode's address_space
>> +   for vma tracking (and hence pte tracking at shootdown time with
>> +   unmap_mapping_range).
>> +
>> +   If the above shootdown dance turns out to be too expensive in certain
>> +   scenarios, we can extend dma-buf with a more explicit cache tracking scheme
>> +   for userspace mappings. But the current assumption is that using mmap is
>> +   always a slower path, so some inefficiencies should be acceptable.
>> +
>> +   Exporters that shoot down mappings (for any reasons) shall not do any
>> +   synchronization at fault time with outstanding device operations.
>> +   Synchronization is an orthogonal issue to sharing the backing storage of a
>> +   buffer and hence should not be handled by dma-buf itself. This is explictly
>> +   mentioned here because many people seem to want something like this, but if
>> +   different exporters handle this differently, buffer sharing can fail in
>> +   interesting ways depending upong the exporter (if userspace starts depending
>> +   upon this implicit synchronization).
>
> How do you ensure that no device can do DMA on the buffer while it's mapped
> into user space in a noncoherent manner?

you do unmap_mapping_range() before DMA..

if you have userspace accessing buffer simultaneously with DMA then
the results are undefined, as they always have been (even w/ uncached
mappings)

BR,
-R

>
>        Arnd
> --
> To unsubscribe from this list: send the line "unsubscribe linux-media" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
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/