netdev - Re: [PATCH] b43: Increase number of RX DMA slots

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] [day] [month] [year] [list]

Message-ID: <5124F085.7010700@lwfinger.net>
Date:	Wed, 20 Feb 2013 09:49:25 -0600
From:	Larry Finger <Larry.Finger@...inger.net>
To:	Gábor Stefanik <netrolller.3d@...il.com>,
	Julian Calaby <julian.calaby@...il.com>,
	David Miller <davem@...emloft.net>, David.Laight@...lab.com,
	linville@...driver.com, linux-wireless@...r.kernel.org,
	netdev@...r.kernel.org, stable@...r.kernel.org,
	openwrt-devel@...ts.openwrt.org
Subject: Re: [PATCH] b43: Increase number of RX DMA slots

On 02/20/2013 02:15 AM, Bastian Bittorf wrote:
> * Larry Finger <Larry.Finger@...inger.net> [20.02.2013 08:32]:
>> On 02/20/2013 12:26 AM, Gábor Stefanik wrote:
>>>
>>> Is this an issue that vendor drivers are also vulnerable to? If it is
>>> a firmware issue, I would certainly think so.
>>
>> I also think so, at least if they are using the firmware version that
>> Bastian is using. His logs don't have that info in them, but I
>> certainly saw the problem on my BCM4312 using firmware 508.154 from
>> 2009.
>
> Another test this morning with heavy downloading (but tcp only)
> show slot usage auf max 204/256. we are using firmware
>
> "version 666.2 (2011-02-23 01:15:07)" which is OpenWrt's default
> for b43. see here the full logs, including minstrel output and dmesg:
>
> http://intercity-vpn.de/files/openwrt/b43test2.dmesg.txt
>
> if a slot needs ~2500 bytes, so 256 slot are only 640kb which seems
> ok to me. ofcourse it raises the memory consumption by 500kb, but now
> the router is useful 8-)

Thanks for the testing and the report. The skb associated with each slot is 
allocated at 2390 bytes, but I think each allocation is a minimum of one page. 
In any case, using extra memory is much better than having the device freeze 
without explanation. I do not think there is any newer firmware for the 4318 
than the version you are using.

I have reworked the patch that resets on overflow, and added the section for 
64-bit DMA. I still need to test that part, but I am sending two patches to you 
for testing on the WRT54G. The first renames a couple of register names to make 
32- and 64-bit naming to only differ in the number. The second is the reset code 
patch.

Larry


View attachment "b43_rename_B43_DMA64_RXSTAT" of type "text/plain" (1968 bytes)

View attachment "b43_workaround_RX_buffer_overflow" of type "text/plain" (4685 bytes)