From mboxrd@z Thu Jan 1 00:00:00 1970 From: tony@atomide.com Subject: Re: DMA support for smc91x Date: Tue, 28 Mar 2006 05:50:28 -0800 Message-ID: <20060328135027.GE28751@atomide.com> References: <20060324002336.GA13667@orphique> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20060324002336.GA13667@orphique> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: linux-omap-open-source-bounces@linux.omap.com Errors-To: linux-omap-open-source-bounces@linux.omap.com To: Ladislav Michl Cc: linux-omap-open-source@linux.omap.com List-Id: linux-omap@vger.kernel.org * Ladislav Michl [060323 16:23]: > Hi, > > this mail contains dirty and for most people completely useless patch. > Here is my motivation for it: I have OMAP based board using LAN91C111, > its interrupt line is routed to GPIO pin. Other GPIO lines are used as > interrupt sources for ISDN chips which needs to be services regulary > (irq can be generated per 125us for each of four chips). All interrupts > from GPIOs are serialized, so if smc91x needs 1.6ms to read 8 full size > packets from memory, communication with ISDN chips gets screwed. Even > decreasing MAX_IRQ_LOOPS to 1 doesn't help too much here, because single > full size packet reading takes 200us. Now programing DMA takes only 6us > and rest of CPU time is left for sometning usesfull :-). Patch is > pre-alpha quality, there are some locking bugs - reading from eeprom > while DMA is progress will result in undefined behaviour (but there > is not eeprom support upstream, so who cares ;-)) - and I just want to > ask if anyone else would need something it and provide some feedback. > Then I may think a bit more, finish it and send it to proper mailing > list. That's cool :) Tony