Re: net: uninitialized loopback addr leaks to userspace

[Stephen Hemminger <shemminger@vyatta.com>]

On Sat, 30 May 2009 22:23:24 +0200
Vegard Nossum <vegard.nossum@gmail.com> wrote:

> Hi,
> 
> It seems that loopback's hardware address is never initialized by the
> kernel. So if userspace attempts to read this address before it has
> been set, the kernel will return some uninitialized data (only 6
> bytes, though). This can be demonstrated by creating a new network
> namespace (CLONE_NEWNET), which creates a new loopback device, then
> call ioctl() with SIOCGIFHWADDR on "lo". If this is done in a loop,
> with some background load, or by running multiple instances, random
> data will start to show up in the returned address.
> 
> [  406.750329] WARNING: kmemcheck: Caught 16-bit read from
> uninitialized memory (ffff880007220974)
> [  406.753555] 18a2d7060088ffff18a2d7060088ffff00000000010000000100000003000000
> [  406.758862]  i i i i i i i i i i i i i i i i i u u u u u u u u u u u u u u u
> [  406.766224]                                          ^
> [  406.768792] Modules linked in:
> [  406.770416] Pid: 757, comm: ifconfig Not tainted
> 2.6.30-rc7-next-20090529 #404
> [  406.772876] RIP: 0010:[<ffffffff80664789>]  [<ffffffff80664789>]
> dev_ioctl+0x5d9/0x600
> [  406.804677]  [<ffffffff8064ff75>] sock_ioctl+0x95/0x2a0
> [  406.807242]  [<ffffffff802c35eb>] vfs_ioctl+0x1b/0x70
> [  406.809348]  [<ffffffff802c36fa>] do_vfs_ioctl+0x8a/0x570
> [  406.811419]  [<ffffffff802c3c79>] sys_ioctl+0x99/0xa0
> [  406.813400]  [<ffffffff802f3941>] dev_ifsioc+0x81/0x2f0
> [  406.815424]  [<ffffffff802f454d>] compat_sys_ioctl+0xed/0x3c0
> [  406.817596]  [<ffffffff8022d476>] cstar_dispatch+0x7/0x26
> [  406.819978]  [<ffffffffffffffff>] 0xffffffffffffffff
> 
> This is the code that triggers the warning, in net/core/dev.c, around line 4150:
> 
>     memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr,
>         min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len));
> 
> So it's dev->dev_addr that is the pointer to the uninitialized data.
> 
> I didn't know how to fix it.
> 


The whole dev structure is zeroed in alloc_netdev(), kmemcheck
is giving bogus warning.
--