From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Brownell Subject: Re: [patch 2.6.28-rc9] spi: spi_write_then_read() regression fix Date: Wed, 24 Dec 2008 00:39:31 -0800 Message-ID: <200812240039.32008.david-b@pacbell.net> References: <200812202332.36281.david-b@pacbell.net> <200812231953.26407.david-b@pacbell.net> Reply-To: dbrownell-Rn4VEauK+AKRv+LV9MX5uipxlwaOVQ5f@public.gmane.org Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Cc: "spi-devel-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org" , Andrew Morton , Vernon Sauder , Linus Torvalds To: "Tang, Feng" Return-path: In-Reply-To: Content-Disposition: inline List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: spi-devel-general-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org List-Id: linux-spi.vger.kernel.org On Tuesday 23 December 2008, Tang, Feng wrote: > o. After that it fails on 2.6.28, I wrote a new API inside > my driver which is similar to your 2.6.28 spi_write_then_read(), > but modify the buffer layout. And the new API works fine for my > max3111 device. = Sounds to me like the issue is a buggy SPI controller driver ... is this a driver that's in mainline? I've certainly seen a couple drivers that handled the full duplex behavior incorrectly, until they fixed that bug. As in, they wrote the tx_buf *then* read rx_buf (a half-duplex model) instead of doing them concurrently (the full-duplex model a spi_master must implement). > > Before: WWWWWW > > After: =A0RRRRRR That's obviously incorrect, or you're introducing some new notation. See my previous notes. - Dave ---------------------------------------------------------------------------= ---