> -----Original Message----- > From: Don Dutile [mailto:ddutile@xxxxxxxxxx] > Sent: Thursday, April 12, 2012 11:20 PM > To: Matthew Wilcox > Cc: Hao, Xudong; Bjorn Helgaas; linux-pci@xxxxxxxxxxxxxxx > Subject: Re: [PATCH V4] Quirk for IVB graphics FLR errata > > On 04/12/2012 12:06 AM, Matthew Wilcox wrote: > > On Wed, Apr 11, 2012 at 10:28:13AM -0400, Don Dutile wrote: > >>> + cycles_t cyc_op_timeout = tsc_khz*op_timeout*1000; > >> Don't we know how to do a 10sec timeout w/o tying it to tsc_khz? > > > > Right, this code should of course be using jiffies and msleep. > > > >> --> other arch compile problem source??? > > > > Well, this device is part of the x86 CPU. It's never going to be > > found as part of any other architecture. Why force other > > architectures to carry this quirk around? > > > > > > Well, the trend to include more IO into chipsets tied to an arch will probably > increase over time, so such conditional quirks will increase as well. > Sounds like the quirk tables need an arch-hook (linked list) to check & traverse. > Then such code can go into arch/<arch>/pci/quirks.c . > If using jiffies instead of tsc, the ifdef <x86> can be removed here, we already have device ID to limit the quirk workaround. So I do want to change this patch file directory, I will change tsc to jiffies. > I was under the impression Linux prefers not to have ifdef <arch> in common > code modules, and to split it out under arch/<> . > Agreed and learned. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
