Re: [Qemu-devel] [PATCH 0/2] Fix wide ioport access cracking

From: Avi Kivity <avi@redhat.com>
To: Gerhard Wiesinger <lists@wiesinger.com>
Cc: qemu-devel@nongnu.org, kvm@vger.kernel.org
Subject: Re: [Qemu-devel] [PATCH 0/2] Fix wide ioport access cracking
Date: Thu, 11 Aug 2011 19:20:57 +0300	[thread overview]
Message-ID: <4E440169.8030403@redhat.com> (raw)
In-Reply-To: <alpine.LFD.2.02.1108111807260.27291@bbs.intern>

On 08/11/2011 07:08 PM, Gerhard Wiesinger wrote:
>
> (gdb) frame 4
> #4  0x000000000041eb9b in pci_update_mappings (d=0x1a90bc0)
>     at /root/download/qemu/git/qemu-kvm-test/hw/pci.c:1134
> 1134                memory_region_del_subregion(r->address_space, 
> r->memory);
> (gdb) print i
> $1 = <value optimized out>
> (gdb) print *r
> $2 = {addr = 22058952032257, size = 32, filtered_size = 
> 171717340864446496,
>   type = 1 '\001', memory = 0x1a90000, address_space = 0x200019282f0}
> (gdb) print d->io_regions[0]
> $3 = {addr = 22058952032257, size = 32, filtered_size = 
> 171717340864446496,
>   type = 1 '\001', memory = 0x1a90000, address_space = 0x200019282f0}

Yikes, this looks like corruption, it the leading 0x2000 in 
address_space is out of place.

Can you step through lsi pci bar registration and place a data 
breakpoint on address_space, and see where it gets this value?

'addr' looks bad too.

-- 
error compiling committee.c: too many arguments to function