linux-kernel - Re: USB transfer_buffer allocations on 64bit systems

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, 14 Apr 2010 19:36:37 +0100
From:	David Woodhouse <dwmw2@...radead.org>
To:	Alan Stern <stern@...land.harvard.edu>
Cc:	Daniel Mack <daniel@...aq.de>, Pedro Ribeiro <pedrib@...il.com>,
	USB list <linux-usb@...r.kernel.org>,
	Andi Kleen <andi@...stfloor.org>,
	Kernel development list <linux-kernel@...r.kernel.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Greg KH <gregkh@...e.de>, iommu@...ts.linux-foundation.org
Subject: Re: USB transfer_buffer allocations on 64bit systems

On Wed, 2010-04-14 at 14:15 -0400, Alan Stern wrote:
> 
> Since using mem=4096M or GFP_DMA stopped the symptoms, it seems very 
> likely that a buffer is getting allocated above the 4 GB line and not 
> bounced or IOMMU-mapped correctly.
> 
> David, do you have anything to suggest?  Any ways to check for IOMMU
> or related errors?
> 
> The problem, in short, is that USB audio doesn't work properly when
> Pedro boots a 64-bit kernel on his 4-GB machine.  With a 32-bit kernel
> it works okay, and it also works if we use dma_alloc_coherent().  The
> host controller is limited to 32-bit DMA, and the DMA addresses
> generated by dma_map_single() appear to be normal.
> 
> At the moment we don't even know if this is caused by a bug in the 
> kernel or a bug in Pedro's hardware.  But he has observed the same 
> problem on two different machines, both using the ICH9 chipset.

Pedro's dmesg suggests that his machine has an IOMMU, but his kernel
isn't built to support it. So he'll be using swiotlb.

Would be interesting to enable CONFIG_DMAR and check whether the problem
goes away. If so, we can start looking harder at the swiotlb code.

-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse@...el.com                              Intel Corporation

--
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/