PCIe IOMEM Resource Allocation

DISCUSSION John 2 participants 3 messages Jul 24, 2014 1 message not in the archive PCI

Share & tools

John

author

2014-07-24 06:40 UTC #0 lore raw

Hi
I am trying to write a PCIe linux device driver to handle PCIe switches and endpoints that show up after the Redhat 6.4 linux is already up and running.

If I try to allocate the resources of the PCIe switches & endpoints between 0xF4000000 and 0xF47FFFFF everything works fine .. But this range is quite full with other devices, so I tried to extend this range to start from 0xF0000000, or from 0xD0000000 the allocation works fine, but I am unable to access devices in the extended range

Anybody has idea why I am unable to use the complete 3G - 4G address range ? Is it related to some ACPI boot configuration that need to be changed, I can see in my dmesg

pci_root PNP0A08:00: host bridge window [mem 0xf4000000-0xf7ffffff] pci_root PNP0A08:00: host bridge window [io 0x1000-0x7fff] pci_root PNP0A08:00: host bridge window [io 0x0000-0x03af] pci_root PNP0A08:00: host bridge window [io 0x03e0-0x0cf7] pci_root PNP0A08:00: host bridge window [io 0x0d00-0x0fff] pci_root PNP0A08:00: host bridge window [mem 0xfed00000-0xfed03fff] pci_root PNP0A08:00: host bridge window [mem 0xfed40000-0xfed44fff] pci_root PNP0A08:00: host bridge window [io 0x03b0-0x03bb] pci_root PNP0A08:00: host bridge window [io 0x03c0-0x03df] pci_root PNP0A08:00: host bridge window [mem 0x000a0000-0x000bffff] pci_root PNP0A08:01: host bridge window [mem 0xf8000000-0xfbffffff] pci_root PNP0A08:01: host bridge window [io 0x8000-0xffff]

Is there a reason for this limited range ? Is it processor dependant ?

Thanks

Yijing Wang

2014-07-24 07:14 UTC #1 lore raw

Quoted 27 lines · John

Hi
I am trying to write a PCIe linux device driver to handle PCIe switches and endpoints that show up after the Redhat 6.4 linux is already up and running.
If I try to allocate the resources of the PCIe switches & endpoints between 0xF4000000 and 0xF47FFFFF everything works fine .. But this range is quite full with other devices, so I tried to extend this range to start from 0xF0000000, or from 0xD0000000 the allocation works fine, but I am unable to access devices in the extended range
Anybody has idea why I am unable to use the complete 3G - 4G address range ? Is it related to some ACPI boot configuration that need to be changed, I can see in my dmesg
pci_root PNP0A08:00: host bridge window [mem 0xf4000000-0xf7ffffff] pci_root PNP0A08:00: host bridge window [io 0x1000-0x7fff] pci_root PNP0A08:00: host bridge window [io 0x0000-0x03af] pci_root PNP0A08:00: host bridge window [io 0x03e0-0x0cf7] pci_root PNP0A08:00: host bridge window [io 0x0d00-0x0fff] pci_root PNP0A08:00: host bridge window [mem 0xfed00000-0xfed03fff] pci_root PNP0A08:00: host bridge window [mem 0xfed40000-0xfed44fff] pci_root PNP0A08:00: host bridge window [io 0x03b0-0x03bb] pci_root PNP0A08:00: host bridge window [io 0x03c0-0x03df] pci_root PNP0A08:00: host bridge window [mem 0x000a0000-0x000bffff] pci_root PNP0A08:01: host bridge window [mem 0xf8000000-0xfbffffff] pci_root PNP0A08:01: host bridge window [io 0x8000-0xffff]

Quoted 1 line

Is there a reason for this limited range ? Is it processor dependant ?

Yes, you should use the resource range in host bridge window, or you won't request resource success. PCI hostbridge retrieve the host bridge window from ACPI _CRS method, if _CRS is not exist, PCI may use global resource iomem_resource and ioport_resource as its hostbridge window.

BIOS is responsible for the _CRS resource range report.

BTW, you can append the pci=nocrs in command line and try again.

Thanks!
Yijing.

Quoted 9 lines

Thanks
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 2 lines

Thanks!
Yijing

A message that is not in the archive (sent off-list or not yet received): CAEYyDAt-PHsB9wrfCiNnd1_RFq3xUQNXWeh5QR9nJVZKho_0JQ@mail.gmail.com

Yijing Wang

2014-07-24 08:03 UTC #3 lore raw

Quoted 6 lines · John Fadel

I already tried pci=nocrs and I have exactly the same behaviour ..
Is there any way to extend this range after linux is up ? How can I access _CRS method to extend this range ? Is there a way to force larger range during the boot of linux ?

First you should guarantee the extend resource range is free in your system,
you can check this from cat /proc/iomem;
Then if you add the pci=nocrs and still fail to use the extend resource range,
this is not a OS problem, and maybe BIOS does not allow OS to use extend resource out of default hostbridge window.

There is no other way to extend the _CRS from OS.

Thanks!
Yijing.

Quoted 65 lines · On Thu, Jul 24, 2014 at 9:14 AM, Yijing Wang <[email protected]…

    On 2014/7/24 14:35, John wrote:
    > Hi
    > I am trying to write a PCIe linux device driver to handle PCIe switches and
    > endpoints that show up after the Redhat 6.4 linux is already up and running.
    >
    > If I try to allocate the resources of the PCIe switches & endpoints between
    > 0xF4000000 and 0xF47FFFFF everything works fine .. But this range is quite
    > full with other devices, so I tried to extend this range to start from
    > 0xF0000000, or from 0xD0000000 the allocation works fine, but I am unable to
    > access devices in the extended range
    >
    > Anybody has idea why I am unable to use the complete 3G - 4G address range ?
    > Is it related to some ACPI boot configuration that need to be changed, I can
    > see in my dmesg
    >
    > pci_root PNP0A08:00: host bridge window [mem 0xf4000000-0xf7ffffff]
    > pci_root PNP0A08:00: host bridge window [io 0x1000-0x7fff]
    > pci_root PNP0A08:00: host bridge window [io 0x0000-0x03af]
    > pci_root PNP0A08:00: host bridge window [io 0x03e0-0x0cf7]
    > pci_root PNP0A08:00: host bridge window [io 0x0d00-0x0fff]
    > pci_root PNP0A08:00: host bridge window [mem 0xfed00000-0xfed03fff]
    > pci_root PNP0A08:00: host bridge window [mem 0xfed40000-0xfed44fff]
    > pci_root PNP0A08:00: host bridge window [io 0x03b0-0x03bb]
    > pci_root PNP0A08:00: host bridge window [io 0x03c0-0x03df]
    > pci_root PNP0A08:00: host bridge window [mem 0x000a0000-0x000bffff]
    > pci_root PNP0A08:01: host bridge window [mem 0xf8000000-0xfbffffff]
    > pci_root PNP0A08:01: host bridge window [io 0x8000-0xffff]
    >
    > Is there a reason for this limited range ? Is it processor dependant ?

    Yes, you should use the resource range in host bridge window, or you won't
    request resource success. PCI hostbridge retrieve the host bridge window
    from ACPI _CRS method, if _CRS is not exist, PCI may use global resource
    iomem_resource and ioport_resource as its hostbridge window.

    BIOS is responsible for the _CRS resource range report.

    BTW, you can append the pci=nocrs in command line and try again.


    Thanks!
    Yijing.


    >
    > Thanks
    >
    > --
    > To unsubscribe from this list: send the line "unsubscribe linux-pci" in
    > the body of a message to [email protected] <mailto:[email protected]>
    > More majordomo info at  http://vger.kernel.org/majordomo-info.html
    >
    >


    --
    Thanks!
    Yijing

Signature · 2 lines

Thanks!
Yijing