[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160719162036.GG2356@ZenIV.linux.org.uk>
Date: Tue, 19 Jul 2016 17:20:36 +0100
From: Al Viro <viro@...IV.linux.org.uk>
To: Dave Jones <davej@...emonkey.org.uk>,
Alexey Dobriyan <adobriyan@...il.com>,
Linux Kernel <linux-kernel@...r.kernel.org>
Subject: Re: 4.7-rc7: use-after-free in proc_map_files_readdir
On Tue, Jul 19, 2016 at 11:31:45AM -0400, Dave Jones wrote:
> On Tue, Jul 19, 2016 at 02:16:36PM +0300, Alexey Dobriyan wrote:
> > > BUG: KASAN: use-after-free in proc_map_files_readdir+0x2e3/0x5a0 at addr ffff88044feb2044
> >
> > Just in case can you addr2line this address or post disassembly?
>
> http://codemonkey.org.uk/junk/fs_proc_base.dis.txt
>
> Which by my math, looks to be..
>
> 7253: 41 8b 87 84 00 00 00 mov 0x84(%r15),%eax
> info.len = snprintf(info.name,
The entire expression is
info.len = snprintf(info.name,
sizeof(info.name), "%lx-%lx",
vma->vm_start, vma->vm_end);
and we have
* address of array field in local structure.
* constant
* string literal
* two longs fetched from *vma, that being done under ->mmap_sem
* call of snprintf
* store into a field of local structure.
The only ways to get use-after-free in that would be to have *vma freed
under you or have the same happen to your stack frame.
Could you dump the relevant part of vmlinux objdump, rather than whatever
you've used on base.o? Having relocations resolved makes it much easier
to figure out... Or just dump that vmlinux on anonftp somewhere...
Powered by blists - more mailing lists