Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 1 Jul 2003 08:22:02 -0400 (EDT)
From:      Andrew Gallatin <gallatin@cs.duke.edu>
To:        ticso@cicely.de
Cc:        freebsd-arch@freebsd.org
Subject:   Re: API change for bus_dma
Message-ID:  <16129.31978.982920.826662@grasshopper.cs.duke.edu>
In-Reply-To: <20030701092944.GE90081@cicely12.cicely.de>
References:  <3EF3C12F.9060303@btc.adaptec.com> <16124.39930.142492.356163@grasshopper.cs.duke.edu> <3EFC9F2D.6020908@btc.adaptec.com> <16124.43999.333761.397624@grasshopper.cs.duke.edu> <3EFCAC7A.6060305@btc.adaptec.com> <16124.45051.919899.414795@grasshopper.cs.duke.edu> <3EFD4755.49BAF150@mindspring.com> <16128.17218.973911.980939@grasshopper.cs.duke.edu> <20030701092944.GE90081@cicely12.cicely.de>

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

Bernd Walter writes:
 > On Mon, Jun 30, 2003 at 10:03:46AM -0400, Andrew Gallatin wrote:
 > > The idea is that I want to establish a mapping that can be used many
 > > times without any driver or kernel attention.  I don't want to do
 > > anything in terms of a system call, or interrupt, etc, to sync the
 > > cache with the state of the DMA'ed page before a DMA read or after a DMA
 > > write.
 > > 
 > > For example, scatter gather mapping on alphas (if FreeBSD supported it
 > > for PCI devices) would be fine with me since its cache-coherent and
 > > doesn't require any ddi_dma_sync() operations.
 > 
 > What makes you shure about alpha dma to be cache-coherent?

7 years of sometimes painful experience with Myrinet on alpha ;)

 > I'm not shure that the current implementation of _bus_dmamap_sync()
 > without barriers is correct.

Barriers, if present, would be used to flush write buffers in the CPU
so they're seen by the cache, not to flush the cache itself.  I agree they
should be present.

Drew




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?16129.31978.982920.826662>