Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 21 Aug 2013 08:53:51 -0700
From:      Peter Grehan <grehan@freebsd.org>
To:        Andre Oppermann <andre@freebsd.org>
Cc:        svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org, Navdeep Parhar <np@FreeBSD.org>
Subject:   Re: svn commit: r254520 - in head/sys: kern sys
Message-ID:  <5214E28F.2060703@freebsd.org>
In-Reply-To: <5214DDA4.4010504@freebsd.org>
References:  <201308191116.r7JBGsc6065793@svn.freebsd.org> <521256CE.6070706@FreeBSD.org> <5212870A.50105@freebsd.org> <521291F1.8060500@FreeBSD.org> <5214D5E0.9040002@freebsd.org> <5214D6CA.2040405@freebsd.org> <5214DDA4.4010504@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Hi Andre,

> On 21.08.2013 17:03, Peter Grehan wrote:
>>> The way to go should be 4K clusters as they are native to the
>>> architecture.
>>> IIRC a PCIe DMA can't cross a 4K boundary anyway
>>
>>   That's a 4G boundary, for some devices.
>
> I meant a single PCIe DMA transaction can be at most 4K before it has to
> set up another one?

  4K is the maximum TLP size but that's rarely (ever?) used - 128 or 256 
bytes is more common. A DMA operation is just a sequence of TLPs 
originating from the endpoint.

later,

Peter.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5214E28F.2060703>