On Wed, Aug 19, 2015 at 04:06:38PM +0200, Helge Deller wrote:
> > On 2015-08-19, at 1:30 AM, Yinghai Lu wrote:
> > > On Tue, Aug 18, 2015 at 9:48 PM, Meelis Roos <mroos@xxxxxxxx> wrote:
> > >>> On Tue, Aug 18, 2015 at 12:47 PM, Helge Deller <deller@xxxxxx> wrote:
> > >>>
> > >>> Then we should change to
> > >>>
> > >>> config PCI_BUS_ADDR_T_64BIT
> > >>> def_bool y if (ARCH_DMA_ADDR_T_64BIT || SPARC64)
> > >>> depends on PCI
> > >>
> > >> Why SPARC64? The problem happened on parisc.
> > > so will not set PCI_BUS_ADDR_T_64BIT for parisc.
>
> I think given the current time frame it's probably the best approach to fix this problem for kernel 4.2.
> It reverts the behaviour back to how it was before (for all arches beside SPARC64).
> I'm still wondering if/why parisc is the only arch (in the sym53c8xx driver only!) which broke by this change...
>
> > I'm not sure this is optimal. While the A500 may only have 32-bit PCI, c8000 appears to have a mix
> > of 32 and 64-bit, and rp34XX is all 64-bit.
>
> True, but probably nobody of us noticed that we only used the 32-bit PCI interface even with 64bit kernel on parisc up to now?
> pci_bus_alloc_resource() in drivers/pci/bus.c just disabled (flag=0) all 64bit resources for us.
> But I agree, I think we need to fix drivers/parisc/lba_pci.c to correctly cope with 64bit pci addresses.
I doubt we can fix the underlying issue before v4.2, but I'd at least
like to avoid it. I applied the patch below to for-linus.
Yinghai proposed this:
> > >>> def_bool y if (ARCH_DMA_ADDR_T_64BIT || SPARC64)
I did this instead:
+ def_bool y if (ARCH_DMA_ADDR_T_64BIT || (64BIT && !PARISC))
because (a) Yinghai's proposal suggests this change is related to
SPARC64, which is misleading, and (b) I want to keep 64-bit bus
addresses for all 64-bit kernels *except* PA-RISC.
I didn't mark it as Tested-by Meelis because I think he tested
Yinghai's proposal. But I can add that back, given confirmation.
commit 51a660a2732c
Author: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
Date: Thu Aug 20 00:08:15 2015 -0500
PCI: Don't use 64-bit bus addresses on PA-RISC
Meelis and Helge reported that 3a9ad0b4fdcd ("PCI: Add pci_bus_addr_t")
caused HPMCs on A500 and hangs on rp5470.
PA-RISC does not set ARCH_DMA_ADDR_T_64BIT, even for 64-bit kernels, so
prior to 3a9ad0b4fdcd, we always used 32-bit PCI addresses. After
3a9ad0b4fdcd, we do use 64-bit PCI addresses in 64-bit kernels, and
apparently there's some PA-RISC problem related to them.
Fixes: 3a9ad0b4fdcd ("PCI: Add pci_bus_addr_t")
Link: http://lkml.kernel.org/r/alpine.LRH.2.11.1507260929000.30065@xxxxxxxxxx
Reported-by: Meelis Roos <mroos@xxxxxxxx>
Reported-by: Helge Deller <deller@xxxxxx>
Based-on-idea-by: Yinghai Lu <yinghai@xxxxxxxxxx>
Signed-off-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
index 73de4ef..944f500 100644
--- a/drivers/pci/Kconfig
+++ b/drivers/pci/Kconfig
@@ -2,7 +2,7 @@
# PCI configuration
#
config PCI_BUS_ADDR_T_64BIT
- def_bool y if (ARCH_DMA_ADDR_T_64BIT || 64BIT)
+ def_bool y if (ARCH_DMA_ADDR_T_64BIT || (64BIT && !PARISC))
depends on PCI
config PCI_MSI
--
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
