Subject: Re: [PATCH 18 of 38] x86: unify pci iommu setup and allow swiotlb
	to	compile for 32 bit
From: Ian Campbell <Ian.Campbell@citrix.com>
To: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: jeremy@goop.org, mingo@elte.hu, linux-kernel@vger.kernel.org,
       xen-devel@lists.xensource.com, x86@kernel.org
In-Reply-To: <20081122104919X.fujita.tomonori@lab.ntt.co.jp>
References: <1226938566.18916.80.camel@zakaz.uk.xensource.com>
	 <20081119112015I.fujita.tomonori@lab.ntt.co.jp>
	 <1227277292.7186.32.camel@zakaz.uk.xensource.com>
	 <20081122104919X.fujita.tomonori@lab.ntt.co.jp>
Content-Type: text/plain
Organization: Citrix Systems, Inc.
Date: Mon, 24 Nov 2008 11:41:37 +0000
Message-Id: <1227526897.9425.38.camel@zakaz.uk.xensource.com>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3194
Lines: 84

On Sat, 2008-11-22 at 10:49 +0900, FUJITA Tomonori wrote:
> On Fri, 21 Nov 2008 14:21:32 +0000
> Ian Campbell <Ian.Campbell@citrix.com> wrote:
> 
> > On Wed, 2008-11-19 at 11:19 +0900, FUJITA Tomonori wrote:
> > > 
> > > The problem that I talked about in the previous mail:
> > > 
> > > >       max_slots = mask + 1
> > > >                   ? ALIGN(mask + 1, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT
> > > >                   : 1UL << (BITS_PER_LONG - IO_TLB_SHIFT);
> > > 
> > > Since the popular value of the mask is 0xffffffff. So the above code
> > > (mask + 1 ?) works wrongly if the size of mask is 32bit (well,
> > > accidentally the result of max_slots is identical though).
> > 
> > I've just been looking at this again and I don't think it is an accident
> > that this evaluates to the correct value when mask + 1 == 0.
> > 
> > The patch which adds the "mask + 1 ? ... : 1UL << ..." stuff is:
> > 
> >         commit b15a3891c916f32a29832886a053a48be2741d4d
> >         Author: Jan Beulich <jbeulich@novell.com>
> >         Date:   Thu Mar 13 09:13:30 2008 +0000
> >         
> >             avoid endless loops in lib/swiotlb.c
> >             
> >             Commit 681cc5cd3efbeafca6386114070e0bfb5012e249 ("iommu sg merging:
> >             swiotlb: respect the segment boundary limits") introduced two
> >             possibilities for entering an endless loop in lib/swiotlb.c:
> >             
> >              - if max_slots is zero (possible if mask is ~0UL)
> >              [...]
> > 
> > I think the existing code is the nicest way to handle this corner case
> > and it is necessary anyway to handle the ~0UL case on 64 bit.
> 
> Ah, I vaguely remember this patch. The ~0ULL mask didn't happen here
> (nobody uses it) so the possibility was false. IMHO, if we use this
> code on 32bit architectures, the mask should be u64 and the overflow
> should be handled explicitly. But as you pointed out, looks like that
> this patch takes account of the overflow.

Something like this?

Ian.
--- 

swiotlb: explicitly handle segment boundary mask overflow.

When swiotlb is used on 32 bit we can overflow mask + 1 in the common
case where mask is 0xffffffffUL. This overflow was previously caught
by the case which attempts to handle a mask of ~0UL on 64 bit.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>

diff -r 5fa30e5284dd lib/swiotlb.c
--- a/lib/swiotlb.c	Mon Nov 24 09:39:50 2008 +0000
+++ b/lib/swiotlb.c	Mon Nov 24 11:37:39 2008 +0000
@@ -303,7 +303,7 @@
 	unsigned int nslots, stride, index, wrap;
 	int i;
 	unsigned long start_dma_addr;
-	unsigned long mask;
+	u64 mask;
 	unsigned long offset_slots;
 	unsigned long max_slots;
 
@@ -314,6 +314,7 @@
 	max_slots = mask + 1
 		    ? ALIGN(mask + 1, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT
 		    : 1UL << (BITS_PER_LONG - IO_TLB_SHIFT);
+	BUG_ON(max_slots > 1UL << (BITS_PER_LONG - IO_TLB_SHIFT));
 
 	/*
 	 * For mappings greater than a page, we limit the stride (and


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