From: David Gibson <david@gibson.dropbear.id.au>
To: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: James Bottomley <James.Bottomley@steeleye.com>,
linux-kernel@vger.kernel.org
Subject: Re: [RFC] generic device DMA implementation
Date: Fri, 6 Dec 2002 11:01:16 +1100 [thread overview]
Message-ID: <20021206000116.GR1500@zax.zax> (raw)
In-Reply-To: <1039086496.651.65.camel@zion>
On Thu, Dec 05, 2002 at 12:08:16PM +0100, Benjamin Herrenschmidt wrote:
> On Thu, 2002-12-05 at 01:47, David Gibson wrote:
> > Do you have an example of where the second option is useful? Off hand
> > the only places I can think of where you'd use a consistent_alloc()
> > rather than map_single() and friends is in cases where the hardware's
> > behaviour means you absolutely positively have to have consistent
> > memory.
>
> Looking at our implementation (ppc32 on non-coherent CPUs like 405) of
> pci_map_single, which just flushes the cache, I still feel we need a
> consistent_alloc, that is an implementation that _disables_ caching for
> the area.
No question there: that's James's first option.
> A typical example is an USB OHCI driver. You really don't want to play
> cache tricks with the shared area here. That will happen each time you
> have a shared area in memory in which both the CPU and the device may
> read/write in the same cache line.
>
> For things like ring descriptors of a net driver, I feel it's very much
> simpler (and possibly more efficient too) to also allocate non-cacheable
> space for consistent instead of continuously flushing/invalidating.
> Actually, flush/invalidate here can also have nasty side effects if
> several descriptors fit in the same cache line.
>
> The data buffers, of course (skbuffs typically) would preferably use
> pci_map_* like APIs (hrm... did we ever make sure skbuffs would _not_
> mix the data buffer with control datas in the same cache line ? This
> have been a problem with non-coherent CPUs in the past).
Indeed - the 405GP ethernet driver, which I've worked on, uses exactly
this approach. consistent_alloc() is used for the descriptor ring
buffer, and DMA syncs are used for the data buffers.
--
David Gibson | For every complex problem there is a
david@gibson.dropbear.id.au | solution which is simple, neat and
| wrong.
http://www.ozlabs.org/people/dgibson
next prev parent reply other threads:[~2002-12-06 0:03 UTC|newest]
Thread overview: 119+ messages / expand[flat|nested] mbox.gz Atom feed top
2002-12-04 17:47 [RFC] generic device DMA implementation James Bottomley
2002-12-04 18:27 ` Jeff Garzik
2002-12-04 19:36 ` James Bottomley
2002-12-04 21:19 ` Miles Bader
2002-12-04 21:21 ` Miles Bader
2002-12-04 21:42 ` James Bottomley
2002-12-05 5:44 ` Miles Bader
2002-12-04 21:46 ` James Bottomley
2002-12-05 2:31 ` Miles Bader
2002-12-05 3:06 ` James Bottomley
2002-12-05 5:02 ` David Gibson
2002-12-05 11:15 ` Benjamin Herrenschmidt
2002-12-05 11:16 ` William Lee Irwin III
2002-12-05 15:12 ` James Bottomley
2002-12-05 0:47 ` David Gibson
2002-12-05 0:54 ` Jeff Garzik
2002-12-05 1:44 ` James Bottomley
2002-12-05 2:38 ` David Gibson
2002-12-05 3:13 ` James Bottomley
2002-12-05 5:05 ` David Gibson
2002-12-05 15:03 ` James Bottomley
2002-12-05 23:54 ` David Gibson
2002-12-05 3:17 ` Miles Bader
2002-12-05 6:06 ` David Gibson
2002-12-05 6:43 ` Miles Bader
2002-12-05 23:44 ` David Gibson
2002-12-06 2:23 ` Miles Bader
2002-12-05 3:41 ` Jeff Garzik
2002-12-05 6:04 ` David Gibson
2002-12-05 16:29 ` Jeff Garzik
2002-12-05 23:59 ` David Gibson
2002-12-05 11:08 ` Benjamin Herrenschmidt
2002-12-05 11:35 ` Russell King
2002-12-05 15:24 ` James Bottomley
2002-12-06 0:01 ` David Gibson [this message]
-- strict thread matches above, loose matches on Subject: below --
2002-12-18 3:01 [RFT][PATCH] " James Bottomley
2002-12-18 3:13 ` David Mosberger
2002-12-28 18:14 ` Russell King
2002-12-28 18:19 ` James Bottomley
2002-12-07 14:37 [RFC] " Adam J. Richter
2002-12-07 4:12 Adam J. Richter
2002-12-06 22:52 Adam J. Richter
2002-12-06 22:17 Adam J. Richter
2002-12-06 22:26 ` James Bottomley
2002-12-06 22:29 ` David S. Miller
2002-12-06 22:48 ` James Bottomley
2002-12-06 22:49 ` David S. Miller
2002-12-06 22:32 ` Arjan van de Ven
2002-12-06 17:39 Adam J. Richter
2002-12-06 18:07 ` Matthew Wilcox
2002-12-06 17:07 Adam J. Richter
2002-12-06 16:48 James Bottomley
2002-12-06 16:19 Adam J. Richter
2002-12-06 16:40 ` Matthew Wilcox
2002-12-06 18:17 ` David S. Miller
2002-12-06 18:29 ` James Bottomley
2002-12-06 18:31 ` David S. Miller
2002-12-06 18:40 ` James Bottomley
2002-12-06 18:42 ` David S. Miller
2002-12-06 21:04 ` Oliver Xymoron
2002-12-07 10:19 ` David Gibson
2002-12-06 18:36 ` Matthew Wilcox
2002-12-06 18:38 ` David S. Miller
2002-12-06 7:41 Adam J. Richter
2002-12-06 15:50 ` David S. Miller
2002-12-06 7:14 Adam J. Richter
2002-12-06 16:26 ` James Bottomley
2002-12-06 17:48 ` Miles Bader
2002-12-07 9:56 ` David Gibson
2002-12-07 9:45 ` David Gibson
2002-12-07 11:26 ` Russell King
2002-12-08 5:28 ` David Gibson
2002-12-06 6:15 David Brownell
2002-12-06 2:08 Adam J. Richter
2002-12-06 2:53 ` David Gibson
2002-12-06 4:03 ` David S. Miller
2002-12-05 20:27 Adam J. Richter
2002-12-05 17:49 Manfred Spraul
2002-12-06 0:08 ` David Gibson
2002-12-05 12:21 Adam J. Richter
2002-12-05 12:44 ` Russell King
2002-12-05 12:13 Adam J. Richter
2002-12-05 11:57 Adam J. Richter
2002-12-06 0:06 ` David Gibson
2002-12-05 5:20 Adam J. Richter
2002-12-05 3:02 Adam J. Richter
2002-12-05 6:15 ` David Gibson
2002-12-05 1:21 Adam J. Richter
2002-12-05 2:40 ` David Gibson
2002-12-05 2:49 ` Miles Bader
2002-12-05 6:12 ` David Gibson
2002-12-05 0:43 Adam J. Richter
2002-12-05 0:55 ` Jeff Garzik
2002-12-05 2:02 ` James Bottomley
2002-11-09 4:51 [parisc-linux] Untested port of parisc_device to generic device interface Adam J. Richter
2002-11-09 5:21 ` Matthew Wilcox
2002-11-09 6:03 ` Greg KH
2002-11-09 15:33 ` J.E.J. Bottomley
2002-11-13 6:13 ` Greg KH
2002-11-13 7:46 ` Miles Bader
2002-11-13 7:52 ` Greg KH
2002-11-13 8:02 ` Miles Bader
2002-11-13 8:10 ` Greg KH
2002-11-13 8:26 ` Miles Bader
2002-11-13 8:25 ` Greg KH
2002-11-13 9:05 ` Miles Bader
[not found] ` <miles@lsi.nec.co.jp>
2002-11-13 20:13 ` Grant Grundler
2002-11-13 20:21 ` J.E.J. Bottomley
2002-11-13 20:37 ` Grant Grundler
2002-11-13 11:59 ` Ivan Kokshaysky
2002-11-13 12:36 ` Marc Zyngier
2002-11-13 16:32 ` Bjorn Helgaas
2002-11-13 17:23 ` J.E.J. Bottomley
2002-11-13 20:33 ` Grant Grundler
2002-11-13 20:44 ` J.E.J. Bottomley
2002-11-13 21:42 ` Grant Grundler
2002-11-13 20:12 ` Grant Grundler
2002-11-09 7:58 ` Marc Zyngier
2002-11-09 18:04 ` Grant Grundler
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20021206000116.GR1500@zax.zax \
--to=david@gibson.dropbear.id.au \
--cc=James.Bottomley@steeleye.com \
--cc=benh@kernel.crashing.org \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).