netdev - Re: [PATCH] net: fix /proc/net/snmp as memory corruptor

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]

Message-ID: <4915295B.4050102@cosmosbay.com>
Date:	Sat, 08 Nov 2008 06:53:31 +0100
From:	Eric Dumazet <dada1@...mosbay.com>
To:	Alexey Dobriyan <adobriyan@...il.com>
CC:	Eric Sesterhenn <snakebyte@....de>, davem@...emloft.net,
	netdev@...r.kernel.org, alan@...rguk.ukuu.org.uk
Subject: Re: [PATCH] net: fix /proc/net/snmp as memory corruptor

Alexey Dobriyan a écrit :
> On Sat, Nov 08, 2008 at 05:52:56AM +0300, Alexey Dobriyan wrote:
>> On Sat, Nov 08, 2008 at 04:02:37AM +0300, Alexey Dobriyan wrote:
>>> On Sat, Nov 08, 2008 at 01:22:08AM +0100, Eric Sesterhenn wrote:
>>>> running a bunch of network related stresstests (isic, isicng, ...) 
>>>> and trying to read all files in /proc afterwards gave me two
>>>> oopses. I was able to reproduce them on another box with
>>>> a different config. I was able to reproduce this on 2.6.24 too,
>>>> so this is no regression. The icmpsic is version 0.06. 
>>>> The minimal testcase to trigger this:
>>>>
>>>> ------------8<----------------
>>>> #!/bin/bash
>>>>
>>>> icmpsic -s 127.0.0.1 -d 127.0.0.1 -p 100000
>>>>
>>>> find /proc/net/ | xargs cat > /dev/null
>>>>
>>>> cat /proc/net/ip_mr_cache
>>>> cat /proc/net/ip_mr_vif
>>>> ------------8<----------------
>>>>
>>>>
>>>> root@...puter-desktop:~/testing# cat /proc/338/net/ip_mr_cache
>>>>
>>>> [ 1572.702100] BUG: unable to handle kernel NULL pointer dereferenceat 000001c1
>>>> [ 1572.702588] IP: [<c05942c6>] ipmr_mfc_seq_show+0x26/0xf0
>>> Reproduced.
>> 	icmpsic -s 127.0.0.1 -d 127.0.0.1 -p 100000
>> 	cat /proc/net/snmp				# sic
>> 	cat /proc/net/ip_mr_cache
>>
>> mfc_cache_array is full of small integers
>>
>> 	[0] = 0x1a8
>> 	[1] = 0x1a9
>>
>> and so on.
> 
> OK, this minimally fixes mfc_cache_array corruption.
> 
> Someone was scared of 16 integers on stack. :^)

Good spot Alexey :)

This should be fixed as well, or multiple threads reading /proc/net/snmp
could get mixed results without proper locking.

ICMPMSG_MIB_MAX being 512, "int" can also be changed to "short",
so that "short out[PERLINE]" only use 32 bytes on stack. 

Frankly, snmp_fold_field() results should be cached in a local array too,
because this function can be expensive if machine has a lot of cpus.

Is 128 + 32 bytes on stack considered evil ?
Using a lock here sounds overkill, and dynamic allocation overkill too...

[PATCH] net: fix /proc/net/snmp as memory corruptor

icmpmsg_put() can happily corrupt kernel memory, using a static
table and forgetting to reset an array index in a loop.

Remove the static array since its not safe without proper locking.

Signed-off-by: Alexey Dobriyan <adobriyan@...il.com>
Signed-off-by: Eric Dumazet <dada1@...mosbay.com>
---
 net/ipv4/proc.c |   58 +++++++++++++++++++++++-----------------------
 1 files changed, 30 insertions(+), 28 deletions(-)


View attachment "icmp_snmp.patch" of type "text/plain" (2021 bytes)