Re: AMD-Vi regression with PCIe/PCI bridges in 4.4

[Clemens Ladisch <clemens-P6GI/4k7KOmELgA04lAiVw@public.gmane.org>]

Joerg Roedel wrote:
> On Tue, Apr 05, 2016 at 12:02:07PM +0200, Clemens Ladisch wrote:
>> Joerg Roedel wrote:
>>> I have seen that before, can you please boot with 'amd_iommu_dump' on
>>> the kernel command line and send me dmesg after boot?
>>
>> Full dmesg: <https://bugzilla.kernel.org/attachment.cgi?id=211771>:
>> ...
>> [    0.179504] pci 0000:04:00.0: [10b5:8112] type 01 class 0x060400
>> [    0.179681] pci 0000:04:00.0: disabling ASPM on pre-1.1 PCIe device.  You can enable it with 'pcie_aspm=force'
>> [    0.179966] pci 0000:00:15.0: PCI bridge to [bus 04-05]
>> [    0.180128] pci 0000:00:15.0:   bridge window [io  0xc000-0xcfff]
>> [    0.180216] pci 0000:05:04.0: [13f6:8788] type 00 class 0x040100
>> [    0.180245] pci 0000:05:04.0: reg 0x10: [io  0xc000-0xc0ff]
>> [    0.180390] pci 0000:05:04.0: supports D1 D2
>> [    0.180478] pci 0000:04:00.0: PCI bridge to [bus 05]
>> [    0.180641] pci 0000:04:00.0:   bridge window [io  0xc000-0xcfff]
>> ...
>> [    0.459076] AMD-Vi: device: 00:00.2 cap: 0040 seg: 0 flags: fe info 1300
>> [    0.459238] AMD-Vi:        mmio-addr: 00000000feb80000
>> [    0.459430] AMD-Vi:   DEV_SELECT_RANGE_START	 devid: 00:01.0 flags: 00
>> [    0.459592] AMD-Vi:   DEV_RANGE_END		 devid: ff:1f.6
>> [    0.460268] AMD-Vi:   DEV_ALIAS_RANGE		 devid: 03:00.0 flags: 00 devid_to: 00:14.4
>> [    0.460555] AMD-Vi:   DEV_RANGE_END		 devid: 03:1f.7
>> [    0.460717] AMD-Vi:   DEV_SPECIAL(HPET[0])		devid: 00:14.0
>> [    0.460876] AMD-Vi:   DEV_SPECIAL(IOAPIC[5])		devid: 00:14.0
>> [    0.542806] iommu: Adding device 0000:00:02.0 to group 0
>> ...
>> [    0.547889] iommu: Adding device 0000:00:15.0 to group 9
>> ...
>> [    0.549907] iommu: Adding device 0000:04:00.0 to group 9
>> [    0.550071] iommu: Adding device 0000:05:04.0 to group 9
>> [    0.550229] AMD-Vi: Found IOMMU at 0000:00:00.2 cap 0x40
>> [    0.550387] AMD-Vi:  Extended features:  PreF PPR GT IA
>> [    0.550804] AMD-Vi: Lazy IO/TLB flushing enabled
>> ...
>> [  124.474883] AMD-Vi: Event logged [IO_PAGE_FAULT device=05:00.0 domain=0x0000 address=0x0000000000100000 flags=0x0050]
>
> Hmm, with this IVRS table it should've never worked.

It did work before 4.4.  Is it possible that the IOMMU, or some feature,
got enabled by default?

Martin/Tomasz, could we get the amd_iommu_dump dmesg output for an
earlier kernel?

> Did you do a BIOS update in-between?

None of the three reporters reported this.

Switching between 4.4 and 4.3 makes the problem (dis)appear.
The problem must be connected with the kernel code.

> Or maybe some change to the sound driver makes it misbehave and use
> a wrong pci request-id?

The driver did not change at all, and the sound device is a plain PCI
chip.  The various sound card models use completely different PCIe/PCI
bridges (PEX8112 and ASM1083).


Regards,
Clemens