Message-ID: <467FA93B.6010403@gmail.com> Date: Mon, 25 Jun 2007 13:38:35 +0200 From: Robert de Rooy User-Agent: Thunderbird 2.0.0.4 (Windows/20070604) MIME-Version: 1.0 To: albertl@mail.com Subject: Re: libata and legacy ide pcmcia failure References: <466718B6.4080105@gmail.com> <4667B39D.209@gmail.com> <46685AE5.4010908@gmail.com> <466900F3.7040801@gmail.com> <46694FB2.6040007@pobox.com> <4669D2D8.3020806@gmail.com> <4669D3F5.7040300@rtr.ca> <466B4D86.1020808@gmail.com> <466C631D.2000001@rtr.ca> <466C723C.6000300@gmail.com> <20070610221322.GA18196@flint.arm.linux.org.uk> <466C7A9F.1080700@rtr.ca> <466DCE55.30507@gmail.com> <466DCF43.5060304@pobox.com> <466E11DF.5020901@gmail.com> <466F4A27.8070309@tw.ibm.com> <46710CD7.9070901@gmail.com> <467BB7DB.4090004@gmail.com> <467BCF86.8060508@rtr.ca> <467F3EDC.2060002@tw.ibm.com> <467F8B8B.1000006@gmail.com> <467F95D8.6000805@tw.ibm.com> In-Reply-To: <467F95D8.6000805@tw.ibm.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Albert Lee wrote: > The patch just workarounds the "lost irq" problem by polling; not real > fix. We still need to find out why irq is lost per Mark's comment: > > "This proves that the device does work correctly in most respects > except for interrupt delivery. The status bits are working and > it can be probed for, configured, and used." > > "So, next step might be to try and understand the interrupt mis-delivery > problem some more. I've lost the history of the original issue, > but we now know that everything except the actual interrupt seems good." > > I am not familiar with the PCMCIA interrupt delivery. It seems the > "Dazzle 4in1 Card Adapter" works under windows but somehow lost irq > under both IDE and libata. Maybe Alan/Bart or the PCMCIA developers > know better... > -- > albert I see. So this is not the ideal fix for this problem. Anything else I can do to debug the irq lost issue? perhaps add some debug to the pcmcia code? For what its worth, I compiled a clean 2.6.22-rc5 with just the pio_polling patch added, and that allows me to mount the device without any errors in syslog. So the disable-dev_init_param-and-setxfermode-for-CFA.patch is not needed.