linux/drivers/pci
Johannes Thumshirn e42010d820 PCI: Set Read Completion Boundary to 128 iff Root Port supports it (_HPX)
Per PCIe spec r3.0, sec 2.3.1.1, the Read Completion Boundary (RCB)
determines the naturally aligned address boundaries on which a Read Request
may be serviced with multiple Completions:

  - For a Root Complex, RCB is 64 bytes or 128 bytes
    This value is reported in the Link Control Register

    Note: Bridges and Endpoints may implement a corresponding command bit
    which may be set by system software to indicate the RCB value for the
    Root Complex, allowing the Bridge/Endpoint to optimize its behavior
    when the Root Complex’s RCB is 128 bytes.

  - For all other system elements, RCB is 128 bytes

Per sec 7.8.7, if a Root Port only supports a 64-byte RCB, the RCB of all
downstream devices must be clear, indicating an RCB of 64 bytes.  If the
Root Port supports a 128-byte RCB, we may optionally set the RCB of
downstream devices so they know they can generate larger Completions.

Some BIOSes supply an _HPX that tells us to set RCB, even though the Root
Port doesn't have RCB set, which may lead to Malformed TLP errors if the
Endpoint generates completions larger than the Root Port can handle.

The IBM x3850 X6 with BIOS version -[A8E120CUS-1.30]- 08/22/2016 supplies
such an _HPX and a Mellanox MT27500 ConnectX-3 device fails to initialize:

  mlx4_core 0000:41:00.0: command 0xfff timed out (go bit not cleared)
  mlx4_core 0000:41:00.0: device is going to be reset
  mlx4_core 0000:41:00.0: Failed to obtain HW semaphore, aborting
  mlx4_core 0000:41:00.0: Fail to reset HCA
  ------------[ cut here ]------------
  kernel BUG at drivers/net/ethernet/mellanox/mlx4/catas.c:193!

After 6cd33649fa ("PCI: Add pci_configure_device() during enumeration")
and 7a1562d4f2 ("PCI: Apply _HPX Link Control settings to all devices
with a link"), we apply _HPX settings to *all* devices, not just those
hot-added after boot.

Before 7a1562d4f2, we didn't touch the Mellanox RCB, and the device
worked.  After 7a1562d4f2, we set its RCB to 128, and it failed.

Set the RCB to 128 iff the Root Port supports a 128-byte RCB.  Otherwise,
set RCB to 64 bytes.  This effectively ignores what _HPX tells us about
RCB.

Note that this change only affects _HPX handling.  If we have no _HPX, this
does nothing with RCB.

[bhelgaas: changelog, clear RCB if not set for Root Port]
Fixes: 6cd33649fa ("PCI: Add pci_configure_device() during enumeration")
Fixes: 7a1562d4f2 ("PCI: Apply _HPX Link Control settings to all devices with a link")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=187781
Tested-by: Frank Danapfel <fdanapfe@redhat.com>
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Myron Stowe <myron.stowe@redhat.com>
CC: stable@vger.kernel.org	# v3.18+
2016-11-23 16:23:55 -06:00
..
host PCI: designware-plat: Update author email 2016-11-16 16:16:40 -06:00
hotplug powerpc updates for 4.9 2016-10-07 20:19:31 -07:00
pcie PCI: Export pcie_find_root_port 2016-11-23 16:23:23 -06:00
access.c PCI: Add pci_set_vpd_size() to set VPD size 2016-04-15 13:00:11 -05:00
ats.c PCI: Remove pci_ats_enabled() 2015-08-13 15:59:59 -05:00
bus.c Merge branches 'pci/demodularize-hosts' and 'pci/host-request-windows' into next 2016-08-01 12:23:57 -05:00
ecam.c tree-wide: replace config_enabled() with IS_ENABLED() 2016-08-04 08:50:07 -04:00
host-bridge.c cxl: use pcibios_free_controller_deferred() when removing vPHBs 2016-08-22 11:09:33 +10:00
hotplug-pci.c
htirq.c x86/htirq: Use hierarchical irqdomain to manage Hypertransport interrupts 2015-04-24 15:36:50 +02:00
iov.c PCI: Check for pci_setup_device() failure in pci_iov_add_virtfn() 2016-09-12 16:23:38 -05:00
irq.c
Kconfig PCI/MSI: Enable PCI_MSI_IRQ_DOMAIN support for ARC 2016-09-06 10:53:32 -05:00
Makefile x86/platform/intel-mid: Add Power Management Unit driver 2016-06-15 10:10:49 +02:00
msi.c Merge branch 'irq/for-block' into irq/core 2016-09-15 20:54:40 +02:00
of.c PCI/MSI: Use of_msi_get_domain instead of open-coded "msi-parent" parsing 2015-10-16 13:07:14 +01:00
pci-acpi.c PCI: Query platform firmware for device power state 2016-09-28 11:46:51 -05:00
pci-driver.c Merge branch 'pci/pm' into next 2016-10-03 09:43:21 -05:00
pci-label.c PCI: Fix broken URL for Dell biosdevname 2016-02-29 12:03:19 -06:00
pci-mid.c x86/platform/intel-mid: Add Intel Penwell to ID table 2016-09-08 14:07:53 +02:00
pci-stub.c
pci-sysfs.c PCI: Put PCIe ports into D3 during suspend 2016-06-13 14:57:36 -05:00
pci.c PCI changes for the v4.9 merge window: 2016-10-07 11:46:37 -07:00
pci.h Merge branch 'pci/pm' into next 2016-10-03 09:43:21 -05:00
probe.c PCI: Set Read Completion Boundary to 128 iff Root Port supports it (_HPX) 2016-11-23 16:23:55 -06:00
proc.c PCI: Ignore write combining when mapping I/O port space 2016-06-17 14:43:33 -05:00
quirks.c Merge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6 2016-10-10 14:04:16 -07:00
remove.c PCI: Fix bridge_d3 update on device removal 2016-09-13 16:00:18 -05:00
rom.c PCI: Remove unused IORESOURCE_ROM_COPY and IORESOURCE_ROM_BIOS_COPY 2016-03-12 06:00:29 -06:00
search.c PCI: Add support for multiple DMA aliases 2016-04-11 14:34:32 -05:00
setup-bus.c x86/ioapic: Ignore root bridges without a companion ACPI device 2016-09-10 20:30:31 +02:00
setup-irq.c
setup-res.c PCI: Don't attempt to claim shadow copies of ROM 2016-11-08 14:25:24 -06:00
slot.c PCI: Hold pci_slot_mutex while searching bus->slots list 2015-07-30 16:19:53 -05:00
syscall.c
vc.c PCI: Fix unaligned accesses in VC code 2016-06-20 13:24:20 -05:00
vpd.c
xen-pcifront.c treewide: replace obsolete _refok by __ref 2016-08-02 17:31:41 -04:00