Re: [RFC v2 PATCH 1/1] PCI: override BIOS/firmware resource allocation

[Ram Pai <linuxram@us.ibm.com>]

On Fri, Oct 22, 2010 at 11:55:18AM -0600, Bjorn Helgaas wrote:
> On Thursday, October 21, 2010 06:28:51 pm Ram Pai wrote:
> > On Tue, Oct 19, 2010 at 11:24:39AM -0700, Jesse Barnes wrote:
> 
> > Ok. After further investigation, I find that the BIOS has not allocated any
> > resource to hotplug bridges that have no devices behind them. However
> > the OS tries to allocate some minimal resources, 4096 I/O ports and 2M mem
> > window, to the these hotplug bridges but fails because there are not enough
> > resources to satisfy all the hotplug bridges.
> > 
> > This issue exists even today with the latest mainline kernel but is masked
> > because no devices are really effected. However Yinghai's reallocation patch
> > exposed the issue, since it released the BIOS allocated window and tried to
> > reallocate. Unfortunately it ended up allocating in the wrong order. The
> > bridges with real devices got no resources, where as the hotplug bridges with
> > no devices got some mimimal resources.
> > 
> > The details are captured at: https://bugzilla.kernel.org/show_bug.cgi?id=15960
> > 
> > I suppose the solution is to not pre-allocate resources to hotplug bridges that
> > have no devices behind them, and then bring back Yinghai's reallocation patch.  
> 
> If we assign resources to bridges with no devices behind them while
> starving bridges that DO have devices behind them, I think that's a bug.
> It'd be great if you could isolate and fix that before we complicate
> things by throwing Yinghai's patch into the mix.

Yes. I sent out a patch an hour back with a fix for that.

> 
> I think it'd interesting to have a debug parameter like "pci=assign-all"
> that meant "ignore all BIOS PCI config and start from scratch."  I'm
> sure we'd trip over lots of issues and it would be easier to resolve them
> before trying to make things fancier.

The patch that I had sent 3-weeks back had that feature. It can be triggered
by pci=override=always.

However, your arguement earlier was that it would make no difference because
no one will enable that parameter by default. This means hardly anyone will
report any bugs. 

Moreover I am sure we will encounter lots of bugs if we enable that
parameter, and that would mean we will be fixing bugs for a long long time.

Gating "Yinghai's patch or any other approach" on fixing all the bugs in the
pci subsystem, implicitly means 'go away' ;). I hope you dont mean that.

I propose to bring Yinghai's patch, for that will enable us to expose bugs and
at the same time it will enable SRIOV on *all* the platforms that don't have it.

RP

> 
> > > So on that note, does Windows on these machines support allocation of
> > > SRIOV resources?  If so, how is it handled?  Which resource ranges are
> > > used for the extra BARs?
> > 
> > No. I have not tried windows on these boxes. But last I heard windows
> > did not support SRIOV. Does it?
> 
> I've heard anecdotally that Windows supports SRIOV much better than
> Linux does, but I have no evidence.  I would guess that it depends on
> driver support in addition to Windows itself.
> 
> Bjorn