From mboxrd@z Thu Jan 1 00:00:00 1970 From: Konrad Rzeszutek Wilk Subject: Re: floppy not working in Linux Date: Thu, 4 Mar 2010 13:28:49 -0500 Message-ID: <20100304182849.GC20263@phenom.dumpdata.com> References: <4B8FD5370200007800032C05@vpn.id2.novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <4B8FD5370200007800032C05@vpn.id2.novell.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Jan Beulich Cc: xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org On Thu, Mar 04, 2010 at 02:43:51PM +0000, Jan Beulich wrote: > We had a report that the floppy driver isn't working with current Xen. > While certainly not a major concern to most of us, it still seems odd that > this simple a piece of hard-/software wouldn't work under Xen, especially > as it is known that it had been working a couple of years back. The USB one works, thought there the DMA Debug interface throws a nice message about doing DMA from a stack - hadn't looked in details for in this. > > Trying to debug the issue has surfaced quite a perhaps strange mix of > observations: > > - On an old (32-bit) box of mine I found that the driver would work > on Xen 3.1, while the same kernel running on top of 3.2.x would not > work (nor would a reasonably current kernel on a reasonably current > Xen). > - On both this old and not too old a 64-bit box I found that removing > the REALLY_SLOW_IO definition at the top of floppy.c would reduce > the amount of errors encountered quite significantly. Specifically, I > have never seen a mount fail with that change in place (but there > are occasional errors logged by the driver). > - Generally, the problems start with unexpected interrupts or, less > frequently, "-- FDC reply error" messages. > > While the driver certainly didn't change much over the last couple > of years, the effect of the adjustment above suggests that the > driver itself may have problems which are just unnoticeable when > run on bare hardware. > > Given the timing requirements to service the FDC (a maximum of > about 120us when using the FIFO optimally) I wonder whether it > is reasonable to expect this to work reliably at all, especially > when guest domains are consuming non-neglectable amounts of > CPU time. > > Jan > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel