lkml.org 
[lkml]   [2009]   [Feb]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH] net_dma: call dmaengine_get only if NET_DMA enabled
From
On Fri, Feb 6, 2009 at 3:09 PM, David Miller <davem@davemloft.net> wrote:
> From: Dan Williams <dan.j.williams@intel.com>
> Date: Fri, 6 Feb 2009 14:15:02 -0700
>
>> [ please cc netdev on net_dma patches ]
>>
>> On Fri, Feb 6, 2009 at 9:02 AM, Atsushi Nemoto <anemo@mba.ocn.ne.jp> wrote:
>> > The commit 649274d993212e7c23c0cb734572c2311c200872 ("net_dma:
>> > acquire/release dma channels on ifup/ifdown") added unconditional call
>> > of dmaengine_get() to net_dma. The API should be called only if
>> > NET_DMA was enabled.
>> >
>> > Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
>>
>> Acked-by: Dan Williams <dan.j.williams@intel.com>
>>
>> I was looking to avoid ifdefs in this path by making
>> dmaengine_{get,put} a nop in the DMAENGINE=n case. However, the
>> current code with DMAENGINE=y NET_DMA=n will pin channels even though
>> the network stack is not using them.
>
> I don't want to apply this patch at all.
>
> What is the purpose of keeping the ugly ifdefs in dmaengine.h if we're
> just going to pollute the networking code with the ifdefs anyways?
>
> Make the NOP versions in linux/dmaengine.h actually work.
>
> The NET_DMA stuff is the one thing which is polluting up the networking
> stack with ugly ifdefs, I'm not adding new ones.

Yes, it has been on the todo list for a while, but I eventually want
the net case to look more like the raid case. I.e. have one code path
that picks async versus sync at runtime, with the option to compile
out async support with header file ifdefs only.


\
 
 \ /
  Last update: 2009-02-06 23:55    [W:0.036 / U:0.464 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site