lkml.org 
[lkml]   [2009]   [Apr]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH] x86/pci: do assign root bus res if _CRS is used
Date
On Monday 27 April 2009 02:15:33 pm Yinghai Lu wrote:
> Bjorn Helgaas wrote:
> > On Monday 20 April 2009 07:35:40 pm Yinghai Lu wrote:
> >> it wil be overwriten later if _CRS is used, so don't bother to set it.
> >>
> >> [ Impact: cleanup ]
> >>
> >> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
> >>
> >> ---
> >> arch/x86/pci/amd_bus.c | 4 ++++
> >> 1 file changed, 4 insertions(+)
> >>
> >> Index: linux-2.6/arch/x86/pci/amd_bus.c
> >> ===================================================================
> >> --- linux-2.6.orig/arch/x86/pci/amd_bus.c
> >> +++ linux-2.6/arch/x86/pci/amd_bus.c
> >> @@ -100,6 +100,10 @@ void x86_pci_root_bus_res_quirks(struct
> >> int j;
> >> struct pci_root_info *info;
> >>
> >> + /* don't go for it if _CRS is used */
> >> + if (pci_probe & PCI_USE__CRS)
> >> + return;
> >> +
> >> /* if only one root bus, don't need to anything */
> >> if (pci_root_num < 2)
> >> return;
> >
> > This isn't a comment on this patch per se.
> >
> > I am concerned about the fact that "pci=use_crs" is not the default.
> > From the changelog of 62f420f8282, it sounds like you have to boot an
> > IBM x3850 with "pci=use_crs" to make hot-plug work, even though ACPI
> > tells us everything we need to know. That's backwards.
> >
> > We shouldn't need an option to tell Linux that the firmware is
> > trustworthy. We should have an option to *ignore* it for the times
> > when we trip over something broken and haven't figured out a way to
> > work around it yet.
>
> other system may have broken _CRS.

Do you have examples of problems here, or are you just worried that
there *may* be problems?

> maybe we could try to use DMI whitelist them?

I don't like a whitelist because it requires ongoing maintenance
for correctly-working machines. A blacklist is nicer because it
only requires maintenance for *broken* machines. A date-based
solution would be better from that point of view.

Bjorn


\
 
 \ /
  Last update: 2009-04-27 22:43    [from the cache]
©2003-2011 Jasper Spaans