2.6.31-rc1: parisc: multiple PCI BAR collisions

2.6.31-rc1: parisc: multiple PCI BAR collisions

[Alexander Beregalov]

Hi

Hardware is HP j6000.
It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
usb(onci)) and cannot boot (no root device).
Messages like this:
sym53c8xx 0:0:0f.0: device not available because of BAR 1 [0xf4005000
- 0xf40053ff] collisions.

Re: 2.6.31-rc1: parisc: multiple PCI BAR collisions

[James Bottomley]

On Fri, 2009-06-26 at 12:58 +0400, Alexander Beregalov wrote:
> Hardware is HP j6000.
> It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
> usb(onci)) and cannot boot (no root device).
> Messages like this:
> sym53c8xx 0:0:0f.0: device not available because of BAR 1 [0xf4005000
> - 0xf40053ff] collisions.

This tends to indicate a problem with resource parenting ... could you
post the full boot output?  That might indicate why.

James

Re: 2.6.31-rc1: parisc: multiple PCI BAR collisions

[Jeroen Roovers]

[-- Attachment #1: Type: text/plain, Size: 603 bytes --]

On Fri, 26 Jun 2009 14:37:40 +0000
James Bottomley <James.Bottomley@HansenPartnership.com> wrote:

> On Fri, 2009-06-26 at 12:58 +0400, Alexander Beregalov wrote:
> > Hardware is HP j6000.
> > It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
> > usb(onci)) and cannot boot (no root device).
> > Messages like this:
> > sym53c8xx 0:0:0f.0: device not available because of BAR 1
> > [0xf4005000
> > - 0xf40053ff] collisions.
> 
> This tends to indicate a problem with resource parenting ... could you
> post the full boot output?  That might indicate why.

Here goes.


Kind regards,
     jer

[-- Attachment #2: minicom.cap.bz2 --]
[-- Type: application/x-bzip, Size: 5195 bytes --]

Re: 2.6.31-rc1: parisc: multiple PCI BAR collisions

[James Bottomley]

On Fri, 2009-06-26 at 17:08 +0200, Jeroen Roovers wrote:
> On Fri, 26 Jun 2009 14:37:40 +0000
> James Bottomley <James.Bottomley@HansenPartnership.com> wrote:
> 
> > On Fri, 2009-06-26 at 12:58 +0400, Alexander Beregalov wrote:
> > > Hardware is HP j6000.
> > > It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
> > > usb(onci)) and cannot boot (no root device).
> > > Messages like this:
> > > sym53c8xx 0:0:0f.0: device not available because of BAR 1
> > > [0xf4005000
> > > - 0xf40053ff] collisions.
> > 
> > This tends to indicate a problem with resource parenting ... could you
> > post the full boot output?  That might indicate why.
> 
> Here goes.

It looks like there's some sort of screw up in the LBA resource
allocation on 32 bits.  The slight problem with this is that I don't
have an LBA 32 bit system to debug this; I've only got a C360.

How recent is the failure? Could you bisect it back to the failing
commit?

Thanks,

James

Re: 2.6.31-rc1: parisc: multiple PCI BAR collisions

[Grant Grundler]

On Fri, Jun 26, 2009 at 12:58:22PM +0400, Alexander Beregalov wrote:
> Hi
> 
> Hardware is HP j6000.
> It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
> usb(onci)) and cannot boot (no root device).
> Messages like this:
> sym53c8xx 0:0:0f.0: device not available because of BAR 1 [0xf4005000
> - 0xf40053ff] collisions.

Yup - I found that out yesterday too. Console output is here:
http://iou.parisc-linux.org/~grundler/console/j6000-2.6.30-parisc_2.6_parenting_failed-01
http://iou.parisc-linux.org/~grundler/console/j6000-2.6.30-parisc_2.6_parenting_failed-02

but in an IRC conversation it' clear I only dumped the "elmmio" and not
the regular "lmmio" resource request. Need to dump both.

I'm pretty sure this is a problem of the root bus resources not getting
setup correctly. I'm not clear on what's wrong. jejb and willy are looking
at it now as well.

thanks,
grant

Re: 2.6.31-rc1: parisc: multiple PCI BAR collisions

[Matthew Wilcox]

On Fri, Jun 26, 2009 at 11:03:04AM -0600, Grant Grundler wrote:
> On Fri, Jun 26, 2009 at 12:58:22PM +0400, Alexander Beregalov wrote:
> > Hi
> > 
> > Hardware is HP j6000.
> > It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
> > usb(onci)) and cannot boot (no root device).
> > Messages like this:
> > sym53c8xx 0:0:0f.0: device not available because of BAR 1 [0xf4005000
> > - 0xf40053ff] collisions.
> 
> I'm pretty sure this is a problem of the root bus resources not getting
> setup correctly. I'm not clear on what's wrong. jejb and willy are looking
> at it now as well.

This patch fixes it for me.

----

Fix PCI resource allocation on non-PAT SBA machines

We weren't marking the resources as memory resources, so they weren't
being found by pci_claim_resource().

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>

diff --git a/drivers/parisc/sba_iommu.c b/drivers/parisc/sba_iommu.c
index d46dd57..123d8fe 100644
--- a/drivers/parisc/sba_iommu.c
+++ b/drivers/parisc/sba_iommu.c
@@ -2057,6 +2057,7 @@ void sba_directed_lmmio(struct parisc_device *pci_hba, struct resource *r)
 		r->start = (base & ~1UL) | PCI_F_EXTEND;
 		size = ~ READ_REG32(reg + LMMIO_DIRECT0_MASK);
 		r->end = r->start + size;
+		r->flags = IORESOURCE_MEM;
 	}
 }
 
@@ -2093,4 +2094,5 @@ void sba_distributed_lmmio(struct parisc_device *pci_hba, struct resource *r )
 	size = (~READ_REG32(sba->sba_hpa + LMMIO_DIST_MASK)) / ROPES_PER_IOC;
 	r->start += rope * (size + 1);	/* adjust base for this rope */
 	r->end = r->start + size;
+	r->flags = IORESOURCE_MEM;
 }

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."

Re: 2.6.31-rc1: parisc: multiple PCI BAR collisions

[Kyle McMartin]

On Fri, Jun 26, 2009 at 11:44:18AM -0600, Matthew Wilcox wrote:
> > > Hardware is HP j6000.
> > > It cannot initialize many PCI devices (sym53c8xx, tulip, STI,
> > > usb(onci)) and cannot boot (no root device).
> > > Messages like this:
> > > sym53c8xx 0:0:0f.0: device not available because of BAR 1 [0xf4005000
> > > - 0xf40053ff] collisions.
> > 
> > I'm pretty sure this is a problem of the root bus resources not getting
> > setup correctly. I'm not clear on what's wrong. jejb and willy are looking
> > at it now as well.
> 
> This patch fixes it for me.
> 

obviously correct, applied and pushed out.

thanks, kyle