On Wed, Feb 24, 2016 at 06:11:58PM -0800, Yinghai Lu wrote:
> Meelis reported strange conflicts on sparc v210:
> pci 0001:00:07.0: can't claim BAR 2 [mem 0x7ff00000000-0x7ff000fffff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
> pci 0001:00:06.0: can't claim BAR 0 [io 0x7fe01000000-0x7fe0100000f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
> pci 0001:00:0d.0: can't claim BAR 0 [io 0x7fe01000900-0x7fe01000907]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
> pci 0001:00:0d.0: can't claim BAR 1 [io 0x7fe01000918-0x7fe0100091f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
> pci 0001:00:0d.0: can't claim BAR 2 [io 0x7fe01000910-0x7fe01000917]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
> pci 0001:00:0d.0: can't claim BAR 3 [io 0x7fe01000908-0x7fe0100090f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
> pci 0001:00:0d.0: can't claim BAR 4 [io 0x7fe01000920-0x7fe0100092f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
> PCI: /pci@1e,600000 can't claim Video RAM area [mem 0x7ff000a0000-0x7ff000bffff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
> PCI: /pci@1e,600000 can't claim System ROM [mem 0x7ff000f0000-0x7ff000fffff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
> PCI: /pci@1e,600000 can't claim Video ROM [mem 0x7ff000c0000-0x7ff000c7fff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
>
> we have
> pci_bus 0001:00: root bus resource [io 0x7fe01000000-0x7fe01ffffff] (bus address [0x0000-0xffffff])
> pci_bus 0001:00: root bus resource [mem 0x7ff00000000-0x7ffffffffff] (bus address [0x00000000-0xffffffff])
>
> so pci bus address in the bars for 0001:00:07.0 has all 0s.
>
> 0001:00:07.0 ISA bridge: ULi Electronics Inc. M1533/M1535/M1543 PCI to ISA Bridge [Aladdin IV/V/V+]
> Region 0: [virtual] I/O ports at 0000 [size=64K]
> Region 1: Memory at <unassigned> (32-bit, non-prefetchable) [size=1M]
> Region 2: Memory at <unassigned> (32-bit, non-prefetchable) [size=1M]
> 00: b9 10 33 15 0f 00 10 02 00 00 01 06 00 00 00 00
> 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 30: 00 00 00 00 a0 00 00 00 00 00 00 00 00 00 00 00
>
> According to http://www.versalogic.com/Support/Downloads/pdf/ali1543.pdf
> page 28: The indices before 40h are read-only.
> and we have all 0 from 0x10-0x2f, according to lspci.
> So those BAR do not work as regular BAR, just clean flags, and ignore them all
> the way include claim and sizing and alloc etc.
A BAR that is read-only zeroes is perfectly legal (PCI spec r3.0, sec
6.2.5.1) and __pci_read_base() handles it correctly.
Maybe you're using something other than __pci_read_base() and the problem
is there?
The following dmesg logs show the [10b9:1533] device handled
correctly, with no implemented BARs:
https://bugs.freedesktop.org/attachment.cgi?id=118855
http://d11mgdpsdcgrvc.cloudfront.net/dmesg.txt
http://www.hivmr.com/db/sasj1xfk98zsz9amfmpp1mdkzfk3pjk3
> Reported-by: Meelis Roos <[email protected]>
> Signed-off-by: Yinghai Lu <[email protected]>
> ---
> drivers/pci/quirks.c | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index 051999e..2361a62 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -445,6 +445,21 @@ static void quirk_amd_nl_class(struct pci_dev *pdev)
> DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_NL_USB,
> quirk_amd_nl_class);
>
> +
> +/*
> + * ALi m1533 pci to isa bridge does not have BAR according to datasheet,
> + * clear the flags, so we will try to claim them or reallocate res.
> + */
> +static void quirk_ali1533(struct pci_dev *dev)
> +{
> + int i;
> +
> + for (i = 0; i < PCI_ROM_RESOURCE; i++)
> + dev->resource[i].flags = 0;
> +}
> +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_AL, PCI_DEVICE_ID_AL_M1533,
> + quirk_ali1533);
> +
> static void piix4_io_quirk(struct pci_dev *dev, const char *name, unsigned int port, unsigned int enable)
> {
> u32 devres;
> --
> 1.8.4.5
>
On Thu, Mar 10, 2016 at 9:54 AM, Bjorn Helgaas <[email protected]> wrote:
> On Wed, Feb 24, 2016 at 06:11:58PM -0800, Yinghai Lu wrote:
>> Meelis reported strange conflicts on sparc v210:
>> pci 0001:00:07.0: can't claim BAR 2 [mem 0x7ff00000000-0x7ff000fffff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
>> pci 0001:00:06.0: can't claim BAR 0 [io 0x7fe01000000-0x7fe0100000f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
>> pci 0001:00:0d.0: can't claim BAR 0 [io 0x7fe01000900-0x7fe01000907]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
>> pci 0001:00:0d.0: can't claim BAR 1 [io 0x7fe01000918-0x7fe0100091f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
>> pci 0001:00:0d.0: can't claim BAR 2 [io 0x7fe01000910-0x7fe01000917]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
>> pci 0001:00:0d.0: can't claim BAR 3 [io 0x7fe01000908-0x7fe0100090f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
>> pci 0001:00:0d.0: can't claim BAR 4 [io 0x7fe01000920-0x7fe0100092f]: address conflict with 0001:00:07.0 [io 0x7fe01000000-0x7fe0100ffff]
>> PCI: /pci@1e,600000 can't claim Video RAM area [mem 0x7ff000a0000-0x7ff000bffff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
>> PCI: /pci@1e,600000 can't claim System ROM [mem 0x7ff000f0000-0x7ff000fffff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
>> PCI: /pci@1e,600000 can't claim Video ROM [mem 0x7ff000c0000-0x7ff000c7fff]: address conflict with 0001:00:07.0 [mem 0x7ff00000000-0x7ff000fffff]
>>
>> we have
>> pci_bus 0001:00: root bus resource [io 0x7fe01000000-0x7fe01ffffff] (bus address [0x0000-0xffffff])
>> pci_bus 0001:00: root bus resource [mem 0x7ff00000000-0x7ffffffffff] (bus address [0x00000000-0xffffffff])
>>
>> so pci bus address in the bars for 0001:00:07.0 has all 0s.
>>
>> 0001:00:07.0 ISA bridge: ULi Electronics Inc. M1533/M1535/M1543 PCI to ISA Bridge [Aladdin IV/V/V+]
>> Region 0: [virtual] I/O ports at 0000 [size=64K]
>> Region 1: Memory at <unassigned> (32-bit, non-prefetchable) [size=1M]
>> Region 2: Memory at <unassigned> (32-bit, non-prefetchable) [size=1M]
>> 00: b9 10 33 15 0f 00 10 02 00 00 01 06 00 00 00 00
>> 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> 20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> 30: 00 00 00 00 a0 00 00 00 00 00 00 00 00 00 00 00
>>
>> According to http://www.versalogic.com/Support/Downloads/pdf/ali1543.pdf
>> page 28: The indices before 40h are read-only.
>> and we have all 0 from 0x10-0x2f, according to lspci.
>> So those BAR do not work as regular BAR, just clean flags, and ignore them all
>> the way include claim and sizing and alloc etc.
>
> A BAR that is read-only zeroes is perfectly legal (PCI spec r3.0, sec
> 6.2.5.1) and __pci_read_base() handles it correctly.
>
> Maybe you're using something other than __pci_read_base() and the problem
> is there?
>
> The following dmesg logs show the [10b9:1533] device handled
> correctly, with no implemented BARs:
>
> https://bugs.freedesktop.org/attachment.cgi?id=118855
> http://d11mgdpsdcgrvc.cloudfront.net/dmesg.txt
> http://www.hivmr.com/db/sasj1xfk98zsz9amfmpp1mdkzfk3pjk3
ok, then those resource info should come from OF instead of __pci_read_base().
will drop this patch.
Thanks
Yinghai