linux-kernel - Re: sky2 hangs on me again: This time 200 kb/s IPv4 traffic, not easily reproducable

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]

Date:	Sat, 02 Sep 2006 10:49:36 -0700
From:	Stephen Hemminger <shemminger@...l.org>
To:	Matthias Hentges <oe@...tges.net>
CC:	shogunx <shogunx@...ekfreak.ath.cx>,
	Thomas Glanzmann <sithglan@...d.uni-erlangen.de>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: sky2 hangs on me again: This time 200 kb/s IPv4 traffic, not
 easily reproducable

Matthias Hentges wrote:
> Am Freitag, den 01.09.2006, 22:41 -0400 schrieb shogunx:
>   
>>>> Has this not been fixed in the 2.6.18 git?
>>>>         
>>> Good question. I'll try 2.6.18-rc4-mm3 and report back.
>>>       
>> I am having no problems with 2.6.18-rc5, which I just built and tested.
>>     
>
> The NIC is up and running for about 9hrs now w/ -rc4-mm3, thanks for the
> heads up!
>   

My theory still unproven, is that there is a problem with transmit flow 
control and alignment. I have
no direct relation to Marvell, and they tell me nothing about the 
hardware bugs. But it took two
weeks to find a problem where receive flow control was busted when the 
receive buffer was not
aligned on 8 byte boundary. The receiver would stop and not resume. To 
workaround, the driver
ensures alignment of receive buffers. The only clue was that the receive 
DMA FIFO always
had a partial count left in it.

There may well be a similar problem on transmit; since driver has no 
control over transmit buffer
alignment, fixing it would require copying most transmit data, or a hack 
all the way up in protocols.
Maybe if the driver lied about the hard header length, it could fool the 
protocols.

Probably better just to always negotiate no transmit flow control.

-- 
VGER BF report: H 0.232987
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/