netdev - Re: [PATCH v3 1/3] net/usb/r815x: replace USB buffer from stack to DMA-able

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:	Wed, 31 Jul 2013 18:29:47 +0800
From:	Ming Lei <tom.leiming@...il.com>
To:	David Miller <davem@...emloft.net>
Cc:	joe@...ches.com, gregkh@...uxfoundation.org, hayeswang@...ltek.com,
	netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-usb@...r.kernel.org, nic_swsd@...ltek.com
Subject: Re: [PATCH v3 1/3] net/usb/r815x: replace USB buffer from stack to DMA-able

On Wed, Jul 31, 2013 at 2:49 AM, David Miller <davem@...emloft.net> wrote:
> From: Joe Perches <joe@...ches.com>
> Date: Tue, 30 Jul 2013 11:41:17 -0700
>
>> On Tue, 2013-07-30 at 11:33 -0700, David Miller wrote:
>>> From: Greg KH <gregkh@...uxfoundation.org>
>>> Date: Tue, 30 Jul 2013 07:00:59 -0700
>>>
>>> > This call is so slow, you can afford to make a call to kmalloc for the
>>> > data, as it sure just did for other structures it needed :)
>>>
>>> I told him to implement things this way, to avoid calling kmalloc every
>>> single register access.
>>>
>>> Using kmalloc all the time makes the access fragile, since a badly timed
>>> call during high memory pressure can fail.
>>>
>>> I'd rather the potential failure happen at one time, probe time.
>>>
>>> In any event, Ming Lei has suggested using usbnet_{read,write}_cmd()
>>> instead, which sounds like a good solution to this problem.
>>
>> Those do per-call allocs too.
>
> Sigh...  Ok I won't fight this any longer then :-)

It might not be a big deal because most of these commands are sent
during probe()/ethtool/... context, and not in tx/rx path.

Another choice is to reserve a big enough buffer during probe()
for read/write command, but one mutex is needed too for avoiding
concurrent calling.

Anyway, please call the command API to do such thing, so that we
can improve it easily in future.

Thanks,
-- 
Ming Lei
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html