linux-kernel - Re: [PATCH 0/3] fix GART to respect device's dma

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:	Tue, 16 Sep 2008 18:20:29 +0200
From:	Andi Kleen <andi@...stfloor.org>
To:	Joerg Roedel <joerg.roedel@....com>
Cc:	FUJITA Tomonori <fujita.tomonori@....ntt.co.jp>,
	andi@...stfloor.org, mingo@...e.hu, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 0/3] fix GART to respect device's dma_mask about virtual mappings

On Tue, Sep 16, 2008 at 05:52:11PM +0200, Joerg Roedel wrote:
> On Tue, Sep 16, 2008 at 10:20:40PM +0900, FUJITA Tomonori wrote:
> > On Tue, 16 Sep 2008 02:54:32 +0200
> > Andi Kleen <andi@...stfloor.org> wrote:
> > 
> > > FUJITA Tomonori <fujita.tomonori@....ntt.co.jp> writes:
> > > 
> > > > Currently, gart IOMMU ignores device's dma_mask when it does virtual
> > > > mappings. So it could give a device a virtual address that the device
> > > > can't access to.
> > > 
> > > Huh? That is what the need_iommu() logic in gart_map_sg()
> > > does. An I'm not aware of any bugs in this area.
> > 
> > What the need_iommu() does is seeing if GART needs to do virtual
> > mappings or not.
> > 
> > (After need_iommu() checking) What this patchset does is to guarantee
> > that GART provides a virtual address that a device can access to.
> > 
> > 
> > > Did you actually see that failure in practice? I don't see
> > > how it could happen.
> > 
> > No, I did not. This patchset does the right thing theoretically, I
> > think, but if such problem never happens for GART, I'll drop the patch
> > for GART. Joerg?
> 
> I am not aware of any failures which are fixed by these patches. 

AFAIK all subsystems deal with it on their own. That is because i386
is the same (no remapping pci_map_* at all) and subsystems are usually 
written to i386 semantics.

> But in
> theory there could be failures.

They will stay failures because GFP_DMA bouncing can not be really
done today in the pci_map_* layer. With a lot of effort you could
probably fix all that, but I doubt it would be worth the effort for
the few devices left with DMA masks < 32bit. 

-Andi

> 
> Joerg
> 
> -- 
>            |           AMD Saxony Limited Liability Company & Co. KG
>  Operating |         Wilschdorfer Landstr. 101, 01109 Dresden, Germany
>  System    |                  Register Court Dresden: HRA 4896
>  Research  |              General Partner authorized to represent:
>  Center    |             AMD Saxony LLC (Wilmington, Delaware, US)
>            | General Manager of AMD Saxony LLC: Dr. Hans-R. Deppe, Thomas McCoy
> 

-- 
ak@...ux.intel.com
--
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/