linux.git
23 months agoMerge branch 'pci/controller/endpoint'
Bjorn Helgaas [Mon, 26 Jun 2023 18:00:00 +0000 (13:00 -0500)]
Merge branch 'pci/controller/endpoint'

- Change "PCI Endpoint Virtual NTB driver" Kconfig prompt to be different
  from "PCI Endpoint NTB driver" (Shunsuke Mie)

- Automatically create a function specific attributes group for endpoint
  drivers to avoid reference counting issues (Damien Le Moal)

- Move and unexport pci_epf_type_add_cfs() (Damien Le Moal)

- Reinitialize EPF test DMA transfer completion before submitting it to
  avoid losing the completion notification (Damien Le Moal)

- Fix EPF test DMA transfer completion detection (Damien Le Moal)

- Submit EPF test DMA transfers with dmaengine_submit(), not tx_submit()
  (Damien Le Moal)

- Simplify EPF test read/write/copy functions (Damien Le Moal)

- Simplify EPF test "raise IRQ" interface (Damien Le Moal)

- Simplify EPF test IRQ command execution (Damien Le Moal)

- Improve EPF test command/status register handling (Damien Le Moal)

- Free IRQs before removing device (Damien Le Moal)

- Reinitialize IRQ completions for every test (Damien Le Moal)

- Don't write status in IRQ handler to avoid race (Damien Le Moal)

- Fix dma_chan direction in data transfer test (Yoshihiro Shimoda)

- Return pci_epf_type_add_cfs() error if EPF has no driver (Damien Le Moal)

- Add kernel-doc for pci_epc_raise_irq() and pci_epc_map_msi_irq() MSI
  vector parameters (Manivannan Sadhasivam)

- Pass EPF device ID to driver probe functions (Manivannan Sadhasivam)

- Return -EALREADY if EPC has already been started/stopped (Manivannan
  Sadhasivam)

- Add linkdown notifier support and use it in qcom-ep (Manivannan
  Sadhasivam)

- Add Bus Master Enable event support and use it in qcom-ep (Manivannan
  Sadhasivam)

- Add Qualcomm Modem Host Interface (MHI) endpoint driver (Manivannan
  Sadhasivam)

- Add Layerscape PME interrupt handling to manage link-up notification
  (Frank Li)

* pci/controller/endpoint:
  PCI: layerscape: Add the endpoint linkup notifier support
  PCI: endpoint: pci-epf-vntb: Fix typo in comments
  MAINTAINERS: Add PCI MHI endpoint function driver under MHI bus
  PCI: endpoint: Add PCI Endpoint function driver for MHI bus
  PCI: qcom-ep: Add support for BME notification
  PCI: qcom-ep: Add support for Link down notification
  PCI: endpoint: Add BME notifier support
  PCI: endpoint: Add linkdown notifier support
  PCI: endpoint: Return error if EPC is started/stopped multiple times
  PCI: endpoint: Pass EPF device ID to the probe function
  PCI: endpoint: Add missing documentation about the MSI/MSI-X range
  PCI: endpoint: Improve pci_epf_type_add_cfs()
  PCI: endpoint: functions/pci-epf-test: Fix dma_chan direction
  misc: pci_endpoint_test: Simplify pci_endpoint_test_msi_irq()
  misc: pci_endpoint_test: Do not write status in IRQ handler
  misc: pci_endpoint_test: Re-init completion for every test
  misc: pci_endpoint_test: Free IRQs before removing the device
  PCI: epf-test: Simplify transfers result print
  PCI: epf-test: Simplify DMA support checks
  PCI: epf-test: Cleanup request result handling
  PCI: epf-test: Cleanup pci_epf_test_cmd_handler()
  PCI: epf-test: Improve handling of command and status registers
  PCI: epf-test: Simplify IRQ test commands execution
  PCI: epf-test: Simplify pci_epf_test_raise_irq()
  PCI: epf-test: Simplify read/write/copy test functions
  PCI: epf-test: Use dmaengine_submit() to initiate DMA transfer
  PCI: epf-test: Fix DMA transfer completion detection
  PCI: epf-test: Fix DMA transfer completion initialization
  PCI: endpoint: Move pci_epf_type_add_cfs() code
  PCI: endpoint: Automatically create a function specific attributes group
  PCI: endpoint: Fix a Kconfig prompt of vNTB driver

23 months agoMerge branch 'pci/controller/vmd'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:59 +0000 (12:59 -0500)]
Merge branch 'pci/controller/vmd'

- Reset VMD config register between soft reboots (Nirmal Patel)

- Capture pci_reset_bus() return value instead of printing junk when it
  fails (Xinghui Li)

* pci/controller/vmd:
  PCI: vmd: Fix uninitialized variable usage in vmd_enable_domain()
  PCI: vmd: Reset VMD config register between soft reboots

23 months agoMerge branch 'pci/controller/rockchip'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:59 +0000 (12:59 -0500)]
Merge branch 'pci/controller/rockchip'

- Remove writes to unused registers (Rick Wertenbroek)

- Write endpoint Device ID using correct register (Rick Wertenbroek)

- Assert PCI Configuration Enable bit after probe so endpoint responds
  instead of generating Request Retry Status messages (Rick Wertenbroek)

- Poll waiting for PHY PLLs to lock (Rick Wertenbroek)

- Update RK3399 example DT binding to be valid (Rick Wertenbroek)

- Use RK3399 PCIE_CLIENT_LEGACY_INT_CTRL to generate INTx instead of
  manually generating PCIe message (Rick Wertenbroek)

- Use multiple windows to avoid address translation conflicts (Rick
  Wertenbroek)

- Use u32 (not u16) when accessing 32-bit registers (Rick Wertenbroek)

- Hide MSI-X Capability, since RK3399 can't generate MSI-X (Rick
  Wertenbroek)

- Set endpoint controller required alignment to 256 (Damien Le Moal)

* pci/controller/rockchip:
  PCI: rockchip: Set address alignment for endpoint mode
  PCI: rockchip: Don't advertise MSI-X in PCIe capabilities
  PCI: rockchip: Use u32 variable to access 32-bit registers
  PCI: rockchip: Fix window mapping and address translation for endpoint
  PCI: rockchip: Fix legacy IRQ generation for RK3399 PCIe endpoint core
  dt-bindings: PCI: Update the RK3399 example to a valid one
  PCI: rockchip: Add poll and timeout to wait for PHY PLLs to be locked
  PCI: rockchip: Assert PCI Configuration Enable bit after probe
  PCI: rockchip: Write PCI Device ID to correct register
  PCI: rockchip: Remove writes to unused registers

23 months agoMerge branch 'pci/controller/rcar'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:59 +0000 (12:59 -0500)]
Merge branch 'pci/controller/rcar'

- Remove unused static pcie_base and pcie_dev (Geert Uytterhoeven)

* pci/controller/rcar:
  PCI: rcar: Use correct product family name for Renesas R-Car
  PCI: rcar-host: Remove unused static pcie_base and pcie_dev

23 months agoMerge branch 'pci/controller/qcom'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:59 +0000 (12:59 -0500)]
Merge branch 'pci/controller/qcom'

- Disable register write access after init for IP v2.3.3, v2.9.0
  (Manivannan Sadhasivam)

- Use DWC helpers for enabling/disabling writes to DBI registers
  (Manivannan Sadhasivam)

- Hide slot hotplug capability for IP v1.0.0, v1.9.0, v2.1.0, v2.3.2,
  v2.3.3, v2.7.0, v2.9.0 (Manivannan Sadhasivam)

- Reuse v2.3.2 post-init sequence for v2.4.0 (Manivannan Sadhasivam)

-

* pci/controller/qcom:
  PCI: qcom: Do not advertise hotplug capability for IP v2.1.0
  PCI: qcom: Do not advertise hotplug capability for IP v1.0.0
  PCI: qcom: Use post init sequence of IP v2.3.2 for v2.4.0
  PCI: qcom: Do not advertise hotplug capability for IP v2.3.2
  PCI: qcom: Do not advertise hotplug capability for IPs v2.3.3 and v2.9.0
  PCI: qcom: Do not advertise hotplug capability for IPs v2.7.0 and v1.9.0
  PCI: qcom: Disable write access to read only registers for IP v2.9.0
  PCI: qcom: Use DWC helpers for modifying the read-only DBI registers
  PCI: qcom: Disable write access to read only registers for IP v2.3.3

23 months agoMerge branch 'pci/pci/ftpci100'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:58 +0000 (12:59 -0500)]
Merge branch 'pci/pci/ftpci100'

- Release clock resources on error paths (Junyan Ye)

* pci/pci/ftpci100:
  PCI: ftpci100: Release the clock resources

23 months agoMerge branch 'pci/controller/dwc'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:58 +0000 (12:59 -0500)]
Merge branch 'pci/controller/dwc'

- Wait for link to come up only if we've initiated link training (Ajay
  Agarwal)

- Save and restore imx6 Root Port MSI control to work around hardware
  defect (Richard Zhu)

* pci/controller/dwc:
  PCI: imx6: Save and restore root port MSI control in suspend and resume
  PCI: dwc: Wait for link up only if link is started

23 months agoMerge branch 'pci/controller/cadence'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:58 +0000 (12:59 -0500)]
Merge branch 'pci/controller/cadence'

- Wait for link retrain to complete when working around the J721E i2085
  erratum with Gen2 mode (Siddharth Vadapalli)

* pci/controller/cadence:
  PCI: cadence: Fix Gen2 Link Retraining process

23 months agoMerge branch 'pci/controller/dt'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:57 +0000 (12:59 -0500)]
Merge branch 'pci/controller/dt'

- Add Qualcomm SDX65 endpoint DT compatible string (Rohit Agarwal)

* pci/controller/dt:
  dt-bindings: PCI: qcom: Add SDX65 SoC

23 months agoMerge branch 'pci/misc'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:57 +0000 (12:59 -0500)]
Merge branch 'pci/misc'

- Add pci_clear_master() stub for non-CONFIG_PCI (Sui Jingfeng)

- Correct documentation typos (Randy Dunlap)

* pci/misc:
  Documentation: PCI: correct spelling
  PCI: Add pci_clear_master() stub for non-CONFIG_PCI
  PCI: Expand comment about sorting pci_ids.h entries

23 months agoMerge branch 'pci/virtualization'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:57 +0000 (12:59 -0500)]
Merge branch 'pci/virtualization'

- Delay extra 250ms after FLR of Solidigm P44 Pro NVMe to avoid KVM hang
  when guest is rebooted (Mike Pastore)

- Add function 1 DMA alias quirk for Marvell 88SE9235 (Robin Murphy)

* pci/virtualization:
  PCI: Add function 1 DMA alias quirk for Marvell 88SE9235
  PCI: Delay after FLR of Solidigm P44 Pro NVMe

23 months agoMerge branch 'pci/resource'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:57 +0000 (12:59 -0500)]
Merge branch 'pci/resource'

- When we coalesce host bridge windows, remove invalidated resources from
  the resource tree so future allocations work correctly (Ross Lagerwall)

* pci/resource:
  PCI: Release resource invalidated by coalescing

23 months agoMerge branch 'pci/pm'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:56 +0000 (12:59 -0500)]
Merge branch 'pci/pm'

- Reduce wait time for secondary bus to be ready to speed up resume (Mika
  Westerberg)

- Avoid putting EloPOS E2/S2/H2 (as well as Elo i2) PCIe Ports in D3cold
  (Ondrej Zary)

- Call _REG when transitioning D-states so AML that uses the PCI config
  space OpRegion works, which fixes some ASMedia GPIO controllers (Mario
  Limonciello)

* pci/pm:
  PCI/ACPI: Call _REG when transitioning D-states
  PCI/ACPI: Validate acpi_pci_set_power_state() parameter
  PCI/PM: Avoid putting EloPOS E2/S2/H2 PCIe Ports in D3cold
  PCI/PM: Shorten pci_bridge_wait_for_secondary_bus() wait time for slow links

23 months agoMerge branch 'pci/hotplug'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:56 +0000 (12:59 -0500)]
Merge branch 'pci/hotplug'

- Simplify Attention Button logging (Bjorn Helgaas)

- Cancel bringup sequence if card is not present, to keep from blinking
  Power Indicator indefinitely (Rongguang Wei)

- Reassign bridge resources if necessary for ACPI hotplug (Igor Mammedov)

* pci/hotplug:
  PCI: acpiphp: Reassign resources on bridge if necessary
  PCI: pciehp: Cancel bringup sequence if card is not present
  PCI: pciehp: Simplify Attention Button logging

23 months agoMerge branch 'pci/enumeration'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:56 +0000 (12:59 -0500)]
Merge branch 'pci/enumeration'

- Add PCI_EXT_CAP_ID_PL_32GT define (Ben Dooks)

- Propagate firmware node by calling device_set_node() for better
  modularity (Andy Shevchenko)

- Discover Data Link Layer Link Active Reporting earlier so quirks can take
  advantage of it (Maciej W. Rozycki)

- Use cached Data Link Layer Link Active Reporting capability in pciehp,
  powerpc/eeh, and mlx5 (Maciej W. Rozycki)

- Run quirk for devices that require OS to clear Retrain Link earlier, so
  later quirks can rely on it (Maciej W. Rozycki)

- Export pcie_retrain_link() for use outside ASPM (Maciej W. Rozycki)

- Add Data Link Layer Link Active Reporting as another way for
  pcie_retrain_link() to determine the link is up (Maciej W. Rozycki)

- Work around link training failures (especially on the ASMedia ASM2824
  switch) by training first at 2.5GT/s and then attempting higher rates
  (Maciej W. Rozycki)

* pci/enumeration:
  PCI: Add failed link recovery for device reset events
  PCI: Work around PCIe link training failures
  PCI: Use pcie_wait_for_link_status() in pcie_wait_for_link_delay()
  PCI: Add support for polling DLLLA to pcie_retrain_link()
  PCI: Export pcie_retrain_link() for use outside ASPM
  PCI: Export PCIe link retrain timeout
  PCI: Execute quirk_enable_clear_retrain_link() earlier
  PCI/ASPM: Factor out waiting for link training to complete
  PCI/ASPM: Avoid unnecessary pcie_link_state use
  PCI/ASPM: Use distinct local vars in pcie_retrain_link()
  net/mlx5: Rely on dev->link_active_reporting
  powerpc/eeh: Rely on dev->link_active_reporting
  PCI: pciehp: Rely on dev->link_active_reporting
  PCI: Initialize dev->link_active_reporting earlier
  PCI: of: Propagate firmware node by calling device_set_node()
  PCI: Add PCI_EXT_CAP_ID_PL_32GT define

# Conflicts:
# drivers/pci/pcie/aspm.c

23 months agoMerge branch 'pci/aspm'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:55 +0000 (12:59 -0500)]
Merge branch 'pci/aspm'

- Disable ASPM on MFD function removal to avoid use-after-free (Ding Hui)

- Tighten up pci_enable_link_state() and pci_disable_link_state()
  interfaces so they don't enable/disable states the driver didn't specify
  (Ajay Agarwal)

- Avoid link retraining race that can happen if ASPM sets link control
  parameters while the link is in the midst of training for some other
  reason (Ilpo Järvinen)

* pci/aspm:
  PCI/ASPM: Avoid link retraining race
  PCI/ASPM: Factor out pcie_wait_for_retrain()
  PCI/ASPM: Return 0 or -ETIMEDOUT from  pcie_retrain_link()
  PCI/ASPM: Remove unnecessary ASPM_STATE_L1SS check
  PCI/ASPM: Rename L1.2-specific functions from 'l1ss' to 'l12'
  PCI/ASPM: Set ASPM_STATE_L1 when driver enables L1.1 or L1.2
  PCI/ASPM: Set only ASPM_STATE_L1 when driver enables L1
  PCI/ASPM: Disable only ASPM_STATE_L1 when driver disables L1
  PCI/ASPM: Disable ASPM on MFD function removal to avoid use-after-free

23 months agoMerge branch 'pci/aer'
Bjorn Helgaas [Mon, 26 Jun 2023 17:59:55 +0000 (12:59 -0500)]
Merge branch 'pci/aer'

- Unexport pci_save_aer_state() since it's only used in drivers/pci/ (Bjorn
  Helgaas)

- Drop recommendation for drivers to configure AER Capability, since the
  PCI core does this for all devices (Dave Jiang, Bjorn Helgaas)

* pci/aer:
  Documentation: PCI: Tidy AER documentation
  Documentation: PCI: Update cross references to .rst files
  Documentation: PCI: Drop recommendation to configure AER Capability
  PCI: Unexport pci_save_aer_state()

23 months agoDocumentation: PCI: correct spelling
Randy Dunlap [Thu, 9 Feb 2023 07:13:49 +0000 (23:13 -0800)]
Documentation: PCI: correct spelling

Correct spelling problems for Documentation/PCI/ as reported
by codespell.

Link: https://lore.kernel.org/linux-pci/20230209071400.31476-14-rdunlap@infradead.org
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: vmd: Fix uninitialized variable usage in vmd_enable_domain()
Xinghui Li [Thu, 20 Apr 2023 09:43:31 +0000 (17:43 +0800)]
PCI: vmd: Fix uninitialized variable usage in vmd_enable_domain()

The ret variable in the vmd_enable_domain() function was used
uninitialized when printing a warning message upon failure of
the pci_reset_bus() function.

Thus, fix the issue by assigning ret with the value returned from
pci_reset_bus() before referencing it in the warning message.

This was detected by Smatch:

  drivers/pci/controller/vmd.c:931 vmd_enable_domain() error: uninitialized symbol 'ret'.

[kwilczynski: drop the second patch from the series, add missing reported
by tag, commit log]
Fixes: 0a584655ef89 ("PCI: vmd: Fix secondary bus reset for Intel bridges")
Link: https://lore.kernel.org/all/202305270219.B96IiIfv-lkp@intel.com
Link: https://lore.kernel.org/linux-pci/20230420094332.1507900-2-korantwork@gmail.com
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Xinghui Li <korantli@tencent.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Nirmal Patel <nirmal.patel@linux.intel.com>
23 months agoPCI: rcar: Use correct product family name for Renesas R-Car
Wolfram Sang [Wed, 7 Jun 2023 20:47:50 +0000 (22:47 +0200)]
PCI: rcar: Use correct product family name for Renesas R-Car

Renesas uses "R-Car" as the name for their product family and development
platform. Thus, correct other variants such as "rcar", "RCar", "Rcar",
etc., to the preferred spelling.

[kwilczynski: commit log]
Link: https://lore.kernel.org/linux-pci/20230607204750.27837-1-wsa+renesas@sang-engineering.com
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
23 months agoPCI: layerscape: Add the endpoint linkup notifier support
Frank Li [Mon, 15 May 2023 15:10:49 +0000 (11:10 -0400)]
PCI: layerscape: Add the endpoint linkup notifier support

Layerscape has PME interrupt, which can be used as linkup notifier.  Set
CFG_READY bit of PEX_PF0_CONFIG to enable accesses from root complex when
linkup detected.

Link: https://lore.kernel.org/r/20230515151049.2797105-1-Frank.Li@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
23 months agoPCI: endpoint: pci-epf-vntb: Fix typo in comments
Frank Li [Wed, 14 Dec 2022 17:22:54 +0000 (12:22 -0500)]
PCI: endpoint: pci-epf-vntb: Fix typo in comments

Replace "Span" with "Spad".

Link: https://lore.kernel.org/r/20221214172254.668282-1-Frank.Li@nxp.com
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoMAINTAINERS: Add PCI MHI endpoint function driver under MHI bus
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:56 +0000 (17:17 +0530)]
MAINTAINERS: Add PCI MHI endpoint function driver under MHI bus

Add PCI endpoint driver for MHI bus under the MHI bus entry in MAINTAINERS
file.

Link: https://lore.kernel.org/r/20230602114756.36586-10-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Add PCI Endpoint function driver for MHI bus
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:55 +0000 (17:17 +0530)]
PCI: endpoint: Add PCI Endpoint function driver for MHI bus

Add PCI Endpoint driver for the Qualcomm MHI (Modem Host Interface) bus.

The driver implements the MHI function over PCI in the endpoint device such
as SDX55 modem. The MHI endpoint function driver acts as a controller
driver for the MHI Endpoint stack and carries out all PCI related
functionality.

Link: https://lore.kernel.org/r/20230602114756.36586-9-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: qcom-ep: Add support for BME notification
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:54 +0000 (17:17 +0530)]
PCI: qcom-ep: Add support for BME notification

Add support to pass BME (Bus Master Enable) notification to Endpoint
function driver so that the BME event can be processed by the function.

Link: https://lore.kernel.org/r/20230602114756.36586-8-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: qcom-ep: Add support for Link down notification
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:53 +0000 (17:17 +0530)]
PCI: qcom-ep: Add support for Link down notification

Add support to pass Link down notification to Endpoint function driver so
that the LINK_DOWN event can be processed by the function.

Link: https://lore.kernel.org/r/20230602114756.36586-7-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Add BME notifier support
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:52 +0000 (17:17 +0530)]
PCI: endpoint: Add BME notifier support

Add support to notify the EPF device about the Bus Master Enable (BME)
event received by the EPC device from the Root complex.

Link: https://lore.kernel.org/r/20230602114756.36586-6-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Add linkdown notifier support
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:51 +0000 (17:17 +0530)]
PCI: endpoint: Add linkdown notifier support

Add support to notify the EPF device about the linkdown event from the EPC
device.

Link: https://lore.kernel.org/r/20230602114756.36586-5-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Return error if EPC is started/stopped multiple times
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:50 +0000 (17:17 +0530)]
PCI: endpoint: Return error if EPC is started/stopped multiple times

When the EPC is started or stopped multiple times from configfs, just
return -EALREADY. There is no need to call the EPC start/stop functions
in those cases.

Link: https://lore.kernel.org/r/20230602114756.36586-4-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Pass EPF device ID to the probe function
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:49 +0000 (17:17 +0530)]
PCI: endpoint: Pass EPF device ID to the probe function

Currently, the EPF probe function doesn't get the device ID argument needed
to correctly identify the device table ID of the EPF device.

When multiple entries are added to the "struct pci_epf_device_id" table,
the probe function needs to identify the correct one. This is achieved by
modifying the pci_epf_match_id() function to return the match ID pointer
and passing it to the driver's probe function.

pci_epf_device_match() function can return bool based on the return value
of pci_epf_match_id().

Link: https://lore.kernel.org/r/20230602114756.36586-3-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kishon Vijay Abraham I <kishon@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Add missing documentation about the MSI/MSI-X range
Manivannan Sadhasivam [Fri, 2 Jun 2023 11:47:48 +0000 (17:17 +0530)]
PCI: endpoint: Add missing documentation about the MSI/MSI-X range

Both pci_epc_raise_irq() and pci_epc_map_msi_irq() APIs expect the
MSI/MSI-X vectors to start from 1 but it is not documented. Add the range
info to the kdoc of the APIs to make it clear.

Link: https://lore.kernel.org/r/20230602114756.36586-2-manivannan.sadhasivam@linaro.org
Fixes: 5e8cb4033807 ("PCI: endpoint: Add EP core layer to enable EP controller and EP functions")
Fixes: 87d5972e476f ("PCI: endpoint: Add pci_epc_ops to map MSI IRQ")
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI: endpoint: Improve pci_epf_type_add_cfs()
Damien Le Moal [Mon, 15 May 2023 07:43:47 +0000 (16:43 +0900)]
PCI: endpoint: Improve pci_epf_type_add_cfs()

pci_epf_type_add_cfs() should not be called with an unbound EPF device,
that is, an epf device with epf->driver not set. For such case, replace the
NULL return in pci_epf_type_add_cfs() with a clear ERR_PTR(-ENODEV) pointer
error return.

Link: https://lore.kernel.org/r/20230515074348.595704-2-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivami <manivannan.sadhasivam@linaro.org>
23 months agoPCI: endpoint: functions/pci-epf-test: Fix dma_chan direction
Yoshihiro Shimoda [Wed, 12 Apr 2023 06:34:47 +0000 (15:34 +0900)]
PCI: endpoint: functions/pci-epf-test: Fix dma_chan direction

In pci_epf_test_init_dma_chan() epf_test->dma_chan_rx is assigned from
dma_request_channel() with DMA_DEV_TO_MEM as filter.dma_mask.

However, in pci_epf_test_data_transfer() if the dir is DMA_DEV_TO_MEM,
epf->dma_chan_rx should be used but instead we are using
epf_test->dma_chan_tx.

Fix it.

Link: https://lore.kernel.org/r/20230412063447.2841177-1-yoshihiro.shimoda.uh@renesas.com
Fixes: 8353813c88ef ("PCI: endpoint: Enable DMA tests for endpoints with DMA capabilities")
Tested-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
23 months agomisc: pci_endpoint_test: Simplify pci_endpoint_test_msi_irq()
Damien Le Moal [Sat, 15 Apr 2023 02:35:42 +0000 (11:35 +0900)]
misc: pci_endpoint_test: Simplify pci_endpoint_test_msi_irq()

Simplify the code of pci_endpoint_test_msi_irq() by correctly using
booleans: remove the msix comparison to false as that variable is already a
boolean, and directly return the result of the comparison of the raised
interrupt number.

Link: https://lore.kernel.org/r/20230415023542.77601-18-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agomisc: pci_endpoint_test: Do not write status in IRQ handler
Damien Le Moal [Sat, 15 Apr 2023 02:35:41 +0000 (11:35 +0900)]
misc: pci_endpoint_test: Do not write status in IRQ handler

pci_endpoint_test_irqhandler() always rewrites the status register when an
IRQ is raised, either as-is if STATUS_IRQ_RAISED is not set, or with
STATUS_IRQ_RAISED cleared if that flag is set. The first case creates a
race window with the endpoint side, meaning that the host side test driver
may end up reading what it just wrote, thus losing the real status as set
by the endpoint side before raising the next interrupt.  This can prevent
detecting that the STATUS_IRQ_RAISED flag was set by the endpoint.

Remove this race window by not clearing the STATUS_IRQ_RAISED status flag
and not rewriting that register for every IRQ received.

Link: https://lore.kernel.org/r/20230415023542.77601-17-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agomisc: pci_endpoint_test: Re-init completion for every test
Damien Le Moal [Sat, 15 Apr 2023 02:35:40 +0000 (11:35 +0900)]
misc: pci_endpoint_test: Re-init completion for every test

The irq_raised completion used to detect the end of a test case is
initialized when the test device is probed, but never reinitialized again
before a test case. As a result, the irq_raised completion synchronization
is effective only for the first ioctl test case executed. Any subsequent
call to wait_for_completion() by another ioctl() call will immediately
return, potentially too early, leading to false positive failures.

Fix this by reinitializing the irq_raised completion before starting a new
ioctl() test command.

Link: https://lore.kernel.org/r/20230415023542.77601-16-dlemoal@kernel.org
Fixes: 2c156ac71c6b ("misc: Add host side PCI driver for PCI test function device")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Cc: stable@vger.kernel.org
23 months agomisc: pci_endpoint_test: Free IRQs before removing the device
Damien Le Moal [Sat, 15 Apr 2023 02:35:39 +0000 (11:35 +0900)]
misc: pci_endpoint_test: Free IRQs before removing the device

In pci_endpoint_test_remove(), freeing the IRQs after removing the device
creates a small race window for IRQs to be received with the test device
memory already released, causing the IRQ handler to access invalid memory,
resulting in an oops.

Free the device IRQs before removing the device to avoid this issue.

Link: https://lore.kernel.org/r/20230415023542.77601-15-dlemoal@kernel.org
Fixes: e03327122e2c ("pci_endpoint_test: Add 2 ioctl commands")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: epf-test: Simplify transfers result print
Damien Le Moal [Sat, 15 Apr 2023 02:35:38 +0000 (11:35 +0900)]
PCI: epf-test: Simplify transfers result print

In pci_epf_test_print_rate(), instead of open coding a reduction loop to
allow for a division by a 32-bits ns value, simply use div64_u64() to
calculate the transfer rate. To match the printed unit of KB/s, this
calculation divides the rate by 1000 instead of 1024 (that would be KiB/s
unit).

Change the format of the results printed by pci_epf_test_print_rate() to be
more compact without the double new line. Also use dev_info() instead of
pr_info().

Link: https://lore.kernel.org/r/20230415023542.77601-14-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Simplify DMA support checks
Damien Le Moal [Sat, 15 Apr 2023 02:35:37 +0000 (11:35 +0900)]
PCI: epf-test: Simplify DMA support checks

There is no need to have each read, write and copy test functions check
for the FLAG_USE_DMA flag against the DMA support status indicated by
epf_test->dma_supported. Move this test to the command handler function
pci_epf_test_cmd_handler() to check once for all cases.

Link: https://lore.kernel.org/r/20230415023542.77601-13-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Cleanup request result handling
Damien Le Moal [Sat, 15 Apr 2023 02:35:36 +0000 (11:35 +0900)]
PCI: epf-test: Cleanup request result handling

Each of the test functions pci_epf_test_write(), pci_epf_test_read() and
pci_epf_test_copy() return an int result which is used by
pci_epf_test_cmd_handler() to set a success or error bit in the request
status.

In the spirit of keeping the processing of each test case self-contained
within its own test function, move the request status field update from
pci_epf_test_cmd_handler() to each of these test functions and change these
functions declaration to returning void.

Link: https://lore.kernel.org/r/20230415023542.77601-12-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Cleanup pci_epf_test_cmd_handler()
Damien Le Moal [Sat, 15 Apr 2023 02:35:35 +0000 (11:35 +0900)]
PCI: epf-test: Cleanup pci_epf_test_cmd_handler()

Command codes are never combined together as flags into a single value.
Thus we can replace the series of "if" tests in pci_epf_test_cmd_handler()
with a cleaner switch-case statement.  This also allows checking that we
got a valid command and print an error message if we did not.

Link: https://lore.kernel.org/r/20230415023542.77601-11-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Improve handling of command and status registers
Damien Le Moal [Sat, 15 Apr 2023 02:35:34 +0000 (11:35 +0900)]
PCI: epf-test: Improve handling of command and status registers

The pci-epf-test driver uses the test register BAR memory directly to get
and execute a test registers set by the RC side and defined using a struct
pci_epf_test_reg. This direct use relies on using the register BAR address
as a pointer to a struct pci_epf_test_reg to execute the test case and to
send back the test result through the status field of struct
pci_epf_test_reg. In practice, the status field is always updated before an
interrupt is raised in pci_epf_test_raise_irq(), to ensure that the RC side
sees the updated status when receiving an interrupt.

However, such assignment direct access does not ensure that changes to the
status register make it to memory, and so visible to the host, before an
interrupt is raised, thus potentially resulting in the RC host not seeing
the correct status result for a test.

Avoid this potential problem by using READ_ONCE()/WRITE_ONCE() when
accessing the command and status fields of a pci_epf_test_reg structure.
This ensure that a test start (pci_epf_test_cmd_handler() function) and
completion (with the function pci_epf_test_raise_irq()) achieve a correct
synchronization with the MMIO register accesses on the RC host.

Link: https://lore.kernel.org/r/20230415023542.77601-10-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Simplify IRQ test commands execution
Damien Le Moal [Sat, 15 Apr 2023 02:35:33 +0000 (11:35 +0900)]
PCI: epf-test: Simplify IRQ test commands execution

For the commands COMMAND_RAISE_LEGACY_IRQ, COMMAND_RAISE_MSI_IRQ and
COMMAND_RAISE_MSIX_IRQ, the function pci_epf_test_cmd_handler()
sets the STATUS_IRQ_RAISED status flag and calls the epc function
pci_epc_raise_irq() directly. However, this is also exactly what the
pci_epf_test_raise_irq() function does. Avoid duplicating these
operations by directly using pci_epf_test_raise_irq() for the IRQ test
commands. It is OK to do so as the host side endpoint test driver always
set the correct IRQ type for the IRQ test commands.

At the same time, move the IRQ number check done for the
COMMAND_RAISE_MSI_IRQ and COMMAND_RAISE_MSIX_IRQ commands
to pci_epf_test_raise_irq(), to also check the IRQ number requested
by the host for other test commands.

This significantly simplifies pci_epf_test_cmd_handler().

Link: https://lore.kernel.org/r/20230415023542.77601-9-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Simplify pci_epf_test_raise_irq()
Damien Le Moal [Sat, 15 Apr 2023 02:35:32 +0000 (11:35 +0900)]
PCI: epf-test: Simplify pci_epf_test_raise_irq()

Change the interface of the function pci_epf_test_raise_irq() to directly
pass a pointer to the struct pci_epf_test_reg defining the test being
executed. This avoids the need for grabbing this pointer using the register
BAR address and simplifies the call sites as the IRQ type and IRQ numbers
do not have to be passed as arguments.

Link: https://lore.kernel.org/r/20230415023542.77601-8-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Simplify read/write/copy test functions
Damien Le Moal [Sat, 15 Apr 2023 02:35:31 +0000 (11:35 +0900)]
PCI: epf-test: Simplify read/write/copy test functions

The function pci_epf_test_cmd_handler() uses the register BAR address as a
pointer to a struct pci_epf_test_reg to determine the command sent by the
host and to execute the test function accordingly. There is no need for
doing this assignment again in each of the read, write and copy test
functions. We can simply pass the reg pointer as an argument to the
functions pci_epf_test_write(), pci_epf_test_read() and
pci_epf_test_copy().

Link: https://lore.kernel.org/r/20230415023542.77601-7-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Use dmaengine_submit() to initiate DMA transfer
Damien Le Moal [Sat, 15 Apr 2023 02:35:30 +0000 (11:35 +0900)]
PCI: epf-test: Use dmaengine_submit() to initiate DMA transfer

Instead of an open coded call to the tx_submit() operation of struct
dma_async_tx_descriptor, use the helper function dmaengine_submit().
No functional change is introduced with this.

Link: https://lore.kernel.org/r/20230415023542.77601-6-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: epf-test: Fix DMA transfer completion detection
Damien Le Moal [Sat, 15 Apr 2023 02:35:29 +0000 (11:35 +0900)]
PCI: epf-test: Fix DMA transfer completion detection

pci_epf_test_data_transfer() and pci_epf_test_dma_callback() are not
handling DMA transfer completion correctly, leading to completion
notifications to the RC side that are too early. This problem can be
detected when the RC side is running an IOMMU with messages such as:

  pci-endpoint-test 0000:0b:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x001c address=0xfff00000 flags=0x0000]

When running the pcitest.sh tests: the address used for a previous
test transfer generates the above error while the next test transfer is
running.

Fix this by testing the DMA transfer status in pci_epf_test_dma_callback()
and notifying the completion only when the transfer status is DMA_COMPLETE
or DMA_ERROR. Furthermore, in pci_epf_test_data_transfer(), be paranoid and
check again the transfer status and always call dmaengine_terminate_sync()
before returning.

Link: https://lore.kernel.org/r/20230415023542.77601-5-dlemoal@kernel.org
Fixes: 8353813c88ef ("PCI: endpoint: Enable DMA tests for endpoints with DMA capabilities")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: epf-test: Fix DMA transfer completion initialization
Damien Le Moal [Sat, 15 Apr 2023 02:35:28 +0000 (11:35 +0900)]
PCI: epf-test: Fix DMA transfer completion initialization

Reinitialize the transfer_complete DMA transfer completion before calling
tx_submit(), to avoid seeing the DMA transfer complete before the
completion is initialized, thus potentially losing the completion
notification.

Link: https://lore.kernel.org/r/20230415023542.77601-4-dlemoal@kernel.org
Fixes: 8353813c88ef ("PCI: endpoint: Enable DMA tests for endpoints with DMA capabilities")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: endpoint: Move pci_epf_type_add_cfs() code
Damien Le Moal [Sat, 15 Apr 2023 02:35:27 +0000 (11:35 +0900)]
PCI: endpoint: Move pci_epf_type_add_cfs() code

pci_epf_type_add_cfs() is called only from pci_ep_cfs_add_type_group() in
drivers/pci/endpoint/pci-ep-cfs.c, so there is no need to export this
function.  Move its code from pci-epf-core.c to pci-ep-cfs.c as a static
function.

Link: https://lore.kernel.org/r/20230415023542.77601-3-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: endpoint: Automatically create a function specific attributes group
Damien Le Moal [Sat, 15 Apr 2023 02:35:26 +0000 (11:35 +0900)]
PCI: endpoint: Automatically create a function specific attributes group

A PCI endpoint function driver can define function specific attributes
under its function configfs directory using the add_cfs() endpoint driver
operation. This is done by tying up the mkdir operation for the function
configfs directory to a call to the add_cfs() operation.  However, there
are no checks preventing the user from repeatedly creating function
specific attribute directories with different names, resulting in the same
endpoint specific attributes group being added multiple times, which also
result in an invalid reference counting for the attribute groups. E.g.,
using the pci-epf-ntb function driver as an example, the user creates the
function as follows:

  $ modprobe pci-epf-ntb
  $ cd /sys/kernel/config/pci_ep/functions/pci_epf_ntb
  $ mkdir func0
  $ tree func0
  func0/
  |-- baseclass_code
  |-- cache_line_size
  |-- ...
  `-- vendorid

  $ mkdir func0/attrs
  $ tree func0
  func0/
  |-- attrs
  |   |-- db_count
  |   |-- mw1
  |   |-- mw2
  |   |-- mw3
  |   |-- mw4
  |   |-- num_mws
  |   `-- spad_count
  |-- baseclass_code
  |-- cache_line_size
  |-- ...
  `-- vendorid

At this point, the function can be started by linking the EP controller.
However, if the user mistakenly creates again a directory:

  $ mkdir func0/attrs2
  $ tree func0
  func0/
  |-- attrs
  |   |-- db_count
  |   |-- mw1
  |   |-- mw2
  |   |-- mw3
  |   |-- mw4
  |   |-- num_mws
  |   `-- spad_count
  |-- attrs2
  |   |-- db_count
  |   |-- mw1
  |   |-- mw2
  |   |-- mw3
  |   |-- mw4
  |   |-- num_mws
  |   `-- spad_count
  |-- baseclass_code
  |-- cache_line_size
  |-- ...
  `-- vendorid

The endpoint function specific attributes are duplicated and cause a crash
when the endpoint function device is torn down:

  refcount_t: addition on 0; use-after-free.
  WARNING: CPU: 2 PID: 834 at lib/refcount.c:25 refcount_warn_saturate+0xc8/0x144
  CPU: 2 PID: 834 Comm: rmdir Not tainted 6.3.0-rc1 #1
  Hardware name: Pine64 RockPro64 v2.1 (DT)
  pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
  ...
  Call trace:
  refcount_warn_saturate+0xc8/0x144
  config_item_get+0x7c/0x80
  configfs_rmdir+0x17c/0x30c
  vfs_rmdir+0x8c/0x204
  do_rmdir+0x158/0x184
  __arm64_sys_unlinkat+0x64/0x80
  invoke_syscall+0x48/0x114
  ...

Fix this by modifying pci_epf_cfs_work() to execute the new function
pci_ep_cfs_add_type_group() which itself calls pci_epf_type_add_cfs() to
obtain the function specific attribute group and the group name (directory
name) from the endpoint function driver. If the function driver defines an
attribute group, pci_ep_cfs_add_type_group() then proceeds to register this
group using configfs_register_group(), thus automatically exposing the
function type specific configfs attributes to the user. E.g.:

  $ modprobe pci-epf-ntb
  $ cd /sys/kernel/config/pci_ep/functions/pci_epf_ntb
  $ mkdir func0
  $ tree func0
  func0/
  |-- baseclass_code
  |-- cache_line_size
  |-- ...
  |-- pci_epf_ntb.0
  |   |-- db_count
  |   |-- mw1
  |   |-- mw2
  |   |-- mw3
  |   |-- mw4
  |   |-- num_mws
  |   `-- spad_count
  |-- primary
  |-- ...
  `-- vendorid

With this change, there is no need for the user to create or delete
directories in the endpoint function attributes directory. The
pci_epf_type_group_ops group operations are thus removed.

Also update the documentation for the pci-epf-ntb and pci-epf-vntb function
drivers to reflect this change, removing the explanations showing the need
to manually create the sub-directory for the function specific attributes.

Link: https://lore.kernel.org/r/20230415023542.77601-2-dlemoal@kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI: endpoint: Fix a Kconfig prompt of vNTB driver
Shunsuke Mie [Thu, 2 Feb 2023 10:38:32 +0000 (19:38 +0900)]
PCI: endpoint: Fix a Kconfig prompt of vNTB driver

vNTB driver and NTB driver have same Kconfig prompt. Changed to make it
distinguishable.

Link: https://lore.kernel.org/r/20230202103832.2038286-1-mie@igel.co.jp
Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP")
Signed-off-by: Shunsuke Mie <mie@igel.co.jp>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
23 months agoPCI/ACPI: Call _REG when transitioning D-states
Mario Limonciello [Tue, 20 Jun 2023 14:04:51 +0000 (09:04 -0500)]
PCI/ACPI: Call _REG when transitioning D-states

ACPI r6.5, sec 6.5.4, describes how AML is unable to access an
OperationRegion unless _REG has been called to connect a handler:

  The OS runs _REG control methods to inform AML code of a change in the
  availability of an operation region. When an operation region handler is
  unavailable, AML cannot access data fields in that region.  (Operation
  region writes will be ignored and reads will return indeterminate data.)

The PCI core does not call _REG at any time, leading to the undefined
behavior mentioned in the spec.

The spec explains that _REG should be executed to indicate whether a
given region can be accessed:

  Once _REG has been executed for a particular operation region, indicating
  that the operation region handler is ready, a control method can access
  fields in the operation region. Conversely, control methods must not
  access fields in operation regions when _REG method execution has not
  indicated that the operation region handler is ready.

An example included in the spec demonstrates calling _REG when devices are
turned off: "when the host controller or bridge controller is turned off
or disabled, PCI Config Space Operation Regions for child devices are
no longer available. As such, ETH0’s _REG method will be run when it
is turned off and will again be run when PCI1 is turned off."

It is reported that ASMedia PCIe GPIO controllers fail functional tests
after the system has returning from suspend (S3 or s2idle). This is because
the BIOS checks whether the OSPM has called the _REG method to determine
whether it can interact with the OperationRegion assigned to the device as
part of the other AML called for the device.

To fix this issue, call acpi_evaluate_reg() when devices are transitioning
to D3cold or D0.

[bhelgaas: split pci_power_t checking to preliminary patch]
Link: https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#reg-region
Link: https://lore.kernel.org/r/20230620140451.21007-1-mario.limonciello@amd.com
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael@kernel.org>
23 months agoPCI/ACPI: Validate acpi_pci_set_power_state() parameter
Bjorn Helgaas [Wed, 21 Jun 2023 21:36:12 +0000 (16:36 -0500)]
PCI/ACPI: Validate acpi_pci_set_power_state() parameter

Previously acpi_pci_set_power_state() assumed the requested power state was
valid (PCI_D0 ... PCI_D3cold).  If a caller supplied something else, we
could index outside the state_conv[] array and pass junk to
acpi_device_set_power().

Validate the pci_power_t parameter and return -EINVAL if it's invalid.

Link: https://lore.kernel.org/r/20230621222857.GA122930@bhelgaas
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
23 months agoPCI/PM: Avoid putting EloPOS E2/S2/H2 PCIe Ports in D3cold
Ondrej Zary [Wed, 14 Jun 2023 07:42:53 +0000 (09:42 +0200)]
PCI/PM: Avoid putting EloPOS E2/S2/H2 PCIe Ports in D3cold

The quirk for Elo i2 introduced in commit 92597f97a40b ("PCI/PM: Avoid
putting Elo i2 PCIe Ports in D3cold") is also needed by EloPOS E2/S2/H2
which uses the same Continental Z2 board.

Change the quirk to match the board instead of system.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=215715
Link: https://lore.kernel.org/r/20230614074253.22318-1-linux@zary.sk
Signed-off-by: Ondrej Zary <linux@zary.sk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Set address alignment for endpoint mode
Damien Le Moal [Tue, 18 Apr 2023 07:46:58 +0000 (09:46 +0200)]
PCI: rockchip: Set address alignment for endpoint mode

The address translation unit of the rockchip EP controller does not use
the lower 8 bits of a PCIe-space address to map local memory. Thus we
must set the align feature field to 256 to let the user know about this
constraint.

Link: https://lore.kernel.org/r/20230418074700.1083505-12-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Don't advertise MSI-X in PCIe capabilities
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:57 +0000 (09:46 +0200)]
PCI: rockchip: Don't advertise MSI-X in PCIe capabilities

The RK3399 PCIe endpoint controller cannot generate MSI-X IRQs.
This is documented in the RK3399 technical reference manual (TRM)
section 17.5.9 "Interrupt Support".

MSI-X capability should therefore not be advertised. Remove the
MSI-X capability by editing the capability linked-list. The
previous entry is the MSI capability, therefore get the next
entry from the MSI-X capability entry and set it as next entry
for the MSI capability. This in effect removes MSI-X from the list.

Linked list before : MSI cap -> MSI-X cap -> PCIe Device cap -> ...
Linked list now : MSI cap -> PCIe Device cap -> ...

Link: https://lore.kernel.org/r/20230418074700.1083505-11-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Use u32 variable to access 32-bit registers
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:56 +0000 (09:46 +0200)]
PCI: rockchip: Use u32 variable to access 32-bit registers

Previously u16 variables were used to access 32-bit registers, this
resulted in not all of the data being read from the registers. Also
the left shift of more than 16-bits would result in moving data out
of the variable. Use u32 variables to access 32-bit registers

Link: https://lore.kernel.org/r/20230418074700.1083505-10-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Fix window mapping and address translation for endpoint
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:55 +0000 (09:46 +0200)]
PCI: rockchip: Fix window mapping and address translation for endpoint

The RK3399 PCI endpoint core has 33 windows for PCIe space, now in the
driver up to 32 fixed size (1M) windows are used and pages are allocated
and mapped accordingly. The driver first used a single window and allocated
space inside which caused translation issues (between CPU space and PCI
space) because a window can only have a single translation at a given
time, which if multiple pages are allocated inside will cause conflicts.
Now each window is a single region of 1M which will always guarantee that
the translation is not in conflict.

Set the translation register addresses for physical function. As documented
in the technical reference manual (TRM) section 17.5.5 "PCIe Address
Translation" and section 17.6.8 "Address Translation Registers Description"

Link: https://lore.kernel.org/r/20230418074700.1083505-9-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Fix legacy IRQ generation for RK3399 PCIe endpoint core
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:54 +0000 (09:46 +0200)]
PCI: rockchip: Fix legacy IRQ generation for RK3399 PCIe endpoint core

Fix legacy IRQ generation for RK3399 PCIe endpoint core according to
the technical reference manual (TRM). Assert and deassert legacy
interrupt (INTx) through the legacy interrupt control register
("PCIE_CLIENT_LEGACY_INT_CTRL") instead of manually generating a PCIe
message. The generation of the legacy interrupt was tested and validated
with the PCIe endpoint test driver.

Link: https://lore.kernel.org/r/20230418074700.1083505-8-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agodt-bindings: PCI: Update the RK3399 example to a valid one
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:53 +0000 (09:46 +0200)]
dt-bindings: PCI: Update the RK3399 example to a valid one

Update the example in the documentation to a valid example.
Address for mem-base was invalid, it pointed to address
0x8000'0000 which is the upper region of the DDR which
is not necessarily populated depending on the board.
This address should point to the base of the memory
window region of the controller which is 0xfa00'0000.
Add missing pinctrl.

Link: https://lore.kernel.org/r/20230418074700.1083505-7-rick.wertenbroek@gmail.com
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
23 months agoPCI: rockchip: Add poll and timeout to wait for PHY PLLs to be locked
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:51 +0000 (09:46 +0200)]
PCI: rockchip: Add poll and timeout to wait for PHY PLLs to be locked

The RK3399 PCIe controller should wait until the PHY PLLs are locked.
Add poll and timeout to wait for PHY PLLs to be locked. If they cannot
be locked generate error message and jump to error handler. Accessing
registers in the PHY clock domain when PLLs are not locked causes hang
The PHY PLLs status is checked through a side channel register.
This is documented in the TRM section 17.5.8.1 "PCIe Initialization
Sequence".

Link: https://lore.kernel.org/r/20230418074700.1083505-5-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Assert PCI Configuration Enable bit after probe
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:50 +0000 (09:46 +0200)]
PCI: rockchip: Assert PCI Configuration Enable bit after probe

Assert PCI Configuration Enable bit after probe. When this bit is left to
0 in the endpoint mode, the RK3399 PCIe endpoint core will generate
configuration request retry status (CRS) messages back to the root complex.
Assert this bit after probe to allow the RK3399 PCIe endpoint core to reply
to configuration requests from the root complex.
This is documented in section 17.5.8.1.2 of the RK3399 TRM.

Link: https://lore.kernel.org/r/20230418074700.1083505-4-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Write PCI Device ID to correct register
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:49 +0000 (09:46 +0200)]
PCI: rockchip: Write PCI Device ID to correct register

Write PCI Device ID (DID) to the correct register. The Device ID was not
updated through the correct register. Device ID was written to a read-only
register and therefore did not work. The Device ID is now set through the
correct register. This is documented in the RK3399 TRM section 17.6.6.1.1

Link: https://lore.kernel.org/r/20230418074700.1083505-3-rick.wertenbroek@gmail.com
Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Cc: stable@vger.kernel.org
23 months agoPCI: rockchip: Remove writes to unused registers
Rick Wertenbroek [Tue, 18 Apr 2023 07:46:48 +0000 (09:46 +0200)]
PCI: rockchip: Remove writes to unused registers

Remove write accesses to registers that are marked "unused" (and
therefore read-only) in the technical reference manual (TRM)
(see RK3399 TRM 17.6.8.1)

Link: https://lore.kernel.org/r/20230418074700.1083505-2-rick.wertenbroek@gmail.com
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
23 months agoPCI/ASPM: Avoid link retraining race
Ilpo Järvinen [Tue, 2 May 2023 08:39:23 +0000 (11:39 +0300)]
PCI/ASPM: Avoid link retraining race

PCIe r6.0.1, sec 7.5.3.7, recommends setting the link control parameters,
then waiting for the Link Training bit to be clear before setting the
Retrain Link bit.

This avoids a race where the LTSSM may not use the updated parameters if it
is already in the midst of link training because of other normal link
activity.

Wait for the Link Training bit to be clear before toggling the Retrain Link
bit to ensure that the LTSSM uses the updated link control parameters.

[bhelgaas: commit log, return 0 (success)/-ETIMEDOUT instead of bool for
both pcie_wait_for_retrain() and the existing pcie_retrain_link()]
Suggested-by: Lukas Wunner <lukas@wunner.de>
Fixes: 7d715a6c1ae5 ("PCI: add PCI Express ASPM support")
Link: https://lore.kernel.org/r/20230502083923.34562-1-ilpo.jarvinen@linux.intel.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org
23 months agoPCI/ASPM: Factor out pcie_wait_for_retrain()
Ilpo Järvinen [Tue, 20 Jun 2023 19:49:33 +0000 (14:49 -0500)]
PCI/ASPM: Factor out pcie_wait_for_retrain()

Factor pcie_wait_for_retrain() out from pcie_retrain_link().  No functional
change intended.

[bhelgaas: split out from
https://lore.kernel.org/r/20230502083923.34562-1-ilpo.jarvinen@linux.intel.com]
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI/ASPM: Return 0 or -ETIMEDOUT from pcie_retrain_link()
Bjorn Helgaas [Tue, 20 Jun 2023 19:44:55 +0000 (14:44 -0500)]
PCI/ASPM: Return 0 or -ETIMEDOUT from  pcie_retrain_link()

"pcie_retrain_link" is not a question with a true/false answer, so "bool"
isn't quite the right return type.  Return 0 for success or -ETIMEDOUT if
the retrain failed.  No functional change intended.

[bhelgaas: based on Ilpo's patch below]
Link: https://lore.kernel.org/r/20230502083923.34562-1-ilpo.jarvinen@linux.intel.com
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Add failed link recovery for device reset events
Maciej W. Rozycki [Sun, 11 Jun 2023 17:20:06 +0000 (18:20 +0100)]
PCI: Add failed link recovery for device reset events

Request failed link recovery with any upstream PCIe bridge where a device
has not come back after reset within PCI_RESET_WAIT time.  Reset the
polling interval if recovery succeeded, otherwise continue as usual.

[bhelgaas: inline pcie_parent_link_retrain()]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306111631050.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Work around PCIe link training failures
Maciej W. Rozycki [Sun, 11 Jun 2023 17:20:10 +0000 (18:20 +0100)]
PCI: Work around PCIe link training failures

Attempt to handle cases such as with a downstream port of the ASMedia
ASM2824 PCIe switch where link training never completes and the link
continues switching between speeds indefinitely with the data link layer
never reaching the active state.

It has been observed with a downstream port of the ASMedia ASM2824 Gen 3
switch wired to the upstream port of the Pericom PI7C9X2G304 Gen 2 switch,
using a Delock Riser Card PCI Express x1 > 2 x PCIe x1 device, P/N 41433,
wired to a SiFive HiFive Unmatched board.  In this setup the switches
should negotiate a link speed of 5.0GT/s, falling back to 2.5GT/s if
necessary.

Instead the link continues oscillating between the two speeds, at the rate
of 34-35 times per second, with link training reported repeatedly active
~84% of the time.  Limiting the target link speed to 2.5GT/s with the
upstream ASM2824 device makes the two switches communicate correctly.
Removing the speed restriction afterwards makes the two devices switch to
5.0GT/s then.

Make use of these observations and detect the inability to train the link
by checking for the Data Link Layer Link Active status bit being off while
the Link Bandwidth Management Status indicating that hardware has changed
the link speed or width in an attempt to correct unreliable link operation.

Restrict the speed to 2.5GT/s then with the Target Link Speed field,
request a retrain and wait 200ms for the data link to go up.  If this is
successful, lift the restriction, letting the devices negotiate a higher
speed.

Also check for a 2.5GT/s speed restriction the firmware may have already
arranged and lift it too with ports of devices known to continue working
afterwards (currently only ASM2824), that already report their data link
being up.

[bhelgaas: reorder and squash stubs from
https://lore.kernel.org/r/alpine.DEB.2.21.2306111619570.64925@angie.orcam.me.uk
to avoid adding stubs that do nothing]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2203022037020.56670@angie.orcam.me.uk/
Link: https://source.denx.de/u-boot/u-boot/-/commit/a398a51ccc68
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310038540.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Use pcie_wait_for_link_status() in pcie_wait_for_link_delay()
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:57 +0000 (18:19 +0100)]
PCI: Use pcie_wait_for_link_status() in pcie_wait_for_link_delay()

Remove a DLLLA status bit polling loop from pcie_wait_for_link_delay() and
call almost identical code in pcie_wait_for_link_status() instead.  This
reduces the lower bound on the polling interval from 10ms to 1ms, possibly
increasing the CPU load on the system in favour to reducing the wait time.

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306111611170.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Add support for polling DLLLA to pcie_retrain_link()
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:53 +0000 (18:19 +0100)]
PCI: Add support for polling DLLLA to pcie_retrain_link()

Let the caller of pcie_retrain_link() specify whether they want to use the
LT bit or the DLLLA bit of the Link Status Register to determine if link
training has completed.  It is up to the caller to verify whether the use
of the DLLLA bit, the implementation of which is optional, is valid for the
device requested.

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306110310540.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Export pcie_retrain_link() for use outside ASPM
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:41 +0000 (18:19 +0100)]
PCI: Export pcie_retrain_link() for use outside ASPM

Export pcie_retrain_link() for link retrain needs outside ASPM.  Struct
pcie_link_state is local to ASPM and only used by pcie_retrain_link() to
get at the associated PCI device, so change the operand and adjust the lone
call site accordingly.  Document the interface.  No functional change at
this point.

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306110229010.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Export PCIe link retrain timeout
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:19 +0000 (18:19 +0100)]
PCI: Export PCIe link retrain timeout

Convert LINK_RETRAIN_TIMEOUT from jiffies to milliseconds, accordingly
rename to PCIE_LINK_RETRAIN_TIMEOUT_MS, and make available via "pci.h" for
the PCI core to use.  Use in pcie_wait_for_link_delay().

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310030280.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: Execute quirk_enable_clear_retrain_link() earlier
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:23 +0000 (18:19 +0100)]
PCI: Execute quirk_enable_clear_retrain_link() earlier

Make quirk_enable_clear_retrain_link() an early quirk so that any later
fixups can rely on dev->clear_retrain_link to have been already
initialised.

[bhelgaas: reorder to just before it becomes possible to call
pcie_retrain_link() earlier]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310049000.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI/ASPM: Factor out waiting for link training to complete
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:49 +0000 (18:19 +0100)]
PCI/ASPM: Factor out waiting for link training to complete

Move code polling for the Link Training bit to clear into a function of its
own.

[bhelgaas: reorder to clean up before exposing to PCI core]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306111605060.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI/ASPM: Avoid unnecessary pcie_link_state use
Maciej W. Rozycki [Wed, 14 Jun 2023 20:09:06 +0000 (15:09 -0500)]
PCI/ASPM: Avoid unnecessary pcie_link_state use

[bhelgaas: extract from expose patch, reorder to clean up before exposing]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306110229010.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: qcom: Do not advertise hotplug capability for IP v2.1.0
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:08 +0000 (20:34 +0530)]
PCI: qcom: Do not advertise hotplug capability for IP v2.1.0

SoCs making use of Qcom PCIe controller IP v2.1.0 do not support hotplug
functionality. But the hotplug capability bit is set by default in the
hardware. This causes the kernel PCI core to register hotplug service for
the controller and send hotplug commands to it. But those commands will
timeout generating messages as below during boot and suspend/resume.

[    5.782159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2020 msec ago)
[    5.810161] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2048 msec ago)
[    7.838162] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2020 msec ago)
[    7.870159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2052 msec ago)

This not only spams the console output but also induces a delay of a
couple of seconds. To fix this issue, let's clear the HPC bit in
PCI_EXP_SLTCAP register as a part of the post init sequence to not
advertise the hotplug capability for the controller.

Link: https://lore.kernel.org/r/20230619150408.8468-10-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
23 months agoPCI: qcom: Do not advertise hotplug capability for IP v1.0.0
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:07 +0000 (20:34 +0530)]
PCI: qcom: Do not advertise hotplug capability for IP v1.0.0

SoCs making use of Qcom PCIe controller IP v1.0.0 do not support hotplug
functionality. But the hotplug capability bit is set by default in the
hardware. This causes the kernel PCI core to register hotplug service for
the controller and send hotplug commands to it. But those commands will
timeout generating messages as below during boot and suspend/resume.

[    5.782159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2020 msec ago)
[    5.810161] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2048 msec ago)
[    7.838162] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2020 msec ago)
[    7.870159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2052 msec ago)

This not only spams the console output but also induces a delay of a
couple of seconds. To fix this issue, let's clear the HPC bit in
PCI_EXP_SLTCAP register as a part of the post init sequence to not
advertise the hotplug capability for the controller.

Link: https://lore.kernel.org/r/20230619150408.8468-9-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
23 months agoPCI: qcom: Use post init sequence of IP v2.3.2 for v2.4.0
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:06 +0000 (20:34 +0530)]
PCI: qcom: Use post init sequence of IP v2.3.2 for v2.4.0

The post init sequence of IP v2.4.0 is same as v2.3.2. So let's reuse the
v2.3.2 sequence which now also disables hotplug capability of the
controller as it is not at all supported on any SoCs making use of this IP.

Link: https://lore.kernel.org/r/20230619150408.8468-8-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
23 months agoPCI: qcom: Do not advertise hotplug capability for IP v2.3.2
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:05 +0000 (20:34 +0530)]
PCI: qcom: Do not advertise hotplug capability for IP v2.3.2

SoCs making use of Qcom PCIe controller IP v2.3.2 do not support hotplug
functionality. But the hotplug capability bit is set by default in the
hardware. This causes the kernel PCI core to register hotplug service for
the controller and send hotplug commands to it. But those commands will
timeout generating messages as below during boot and suspend/resume.

[    5.782159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2020 msec ago)
[    5.810161] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2048 msec ago)
[    7.838162] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2020 msec ago)
[    7.870159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2052 msec ago)

This not only spams the console output but also induces a delay of a
couple of seconds. To fix this issue, let's clear the HPC bit in
PCI_EXP_SLTCAP register as a part of the post init sequence to not
advertise the hotplug capability for the controller.

Link: https://lore.kernel.org/r/20230619150408.8468-7-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
23 months agoPCI: qcom: Do not advertise hotplug capability for IPs v2.3.3 and v2.9.0
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:04 +0000 (20:34 +0530)]
PCI: qcom: Do not advertise hotplug capability for IPs v2.3.3 and v2.9.0

SoCs making use of Qcom PCIe controller IPs v2.3.3 and v2.9.0 do not
support hotplug functionality. But the hotplug capability bit is set by
default in the hardware. This causes the kernel PCI core to register
hotplug service for the controller and send hotplug commands to it. But
those commands will timeout generating messages as below during boot
and suspend/resume.

[    5.782159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2020 msec ago)
[    5.810161] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2048 msec ago)
[    7.838162] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2020 msec ago)
[    7.870159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2052 msec ago)

This not only spams the console output but also induces a delay of a
couple of seconds. To fix this issue, let's not set the HPC bit in
PCI_EXP_SLTCAP register as a part of the post init sequence to not
advertise the hotplug capability for the controller.

Link: https://lore.kernel.org/r/20230619150408.8468-6-manivannan.sadhasivam@linaro.org
Tested-by: Sricharan Ramabadhran <quic_srichara@quicinc.com>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
23 months agoPCI: qcom: Do not advertise hotplug capability for IPs v2.7.0 and v1.9.0
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:03 +0000 (20:34 +0530)]
PCI: qcom: Do not advertise hotplug capability for IPs v2.7.0 and v1.9.0

SoCs making use of Qcom PCIe controller IPs v2.7.0 and v1.9.0 do not
support hotplug functionality. But the hotplug capability bit is set by
default in the hardware. This causes the kernel PCI core to register
hotplug service for the controller and send hotplug commands to it. But
those commands will timeout generating messages as below during boot and
suspend/resume.

[    5.782159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2020 msec ago)
[    5.810161] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2048 msec ago)
[    7.838162] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2020 msec ago)
[    7.870159] pcieport 0001:00:00.0: pciehp: Timeout on hotplug command 0x07c0 (issued 2052 msec ago)

This not only spams the console output but also induces a delay of a
couple of seconds. To fix this issue, let's clear the HPC bit in
PCI_EXP_SLTCAP register as a part of the post init sequence to not
advertise the hotplug capability for the controller.

Link: https://lore.kernel.org/r/20230619150408.8468-5-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
23 months agoPCI: qcom: Disable write access to read only registers for IP v2.9.0
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:02 +0000 (20:34 +0530)]
PCI: qcom: Disable write access to read only registers for IP v2.9.0

In the post init sequence of v2.9.0, write access to read only registers
are not disabled after updating the registers. Fix it by disabling the
access after register update.

While at it, let's also add a newline after existing dw_pcie_dbi_ro_wr_en()
guard function to align with rest of the driver.

Link: https://lore.kernel.org/r/20230619150408.8468-4-manivannan.sadhasivam@linaro.org
Fixes: 0cf7c2efe8ac ("PCI: qcom: Add IPQ60xx support")
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
23 months agoPCI: qcom: Use DWC helpers for modifying the read-only DBI registers
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:01 +0000 (20:34 +0530)]
PCI: qcom: Use DWC helpers for modifying the read-only DBI registers

DWC core already exposes dw_pcie_dbi_ro_wr_{en/dis} helper APIs for
enabling and disabling the write access to read only DBI registers. So
let's use them instead of doing it manually.

Also, the existing code doesn't disable the write access when it's done.
This is also fixed now.

Link: https://lore.kernel.org/r/20230619150408.8468-3-manivannan.sadhasivam@linaro.org
Fixes: 5d76117f070d ("PCI: qcom: Add support for IPQ8074 PCIe controller")
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
23 months agoPCI: qcom: Disable write access to read only registers for IP v2.3.3
Manivannan Sadhasivam [Mon, 19 Jun 2023 15:04:00 +0000 (20:34 +0530)]
PCI: qcom: Disable write access to read only registers for IP v2.3.3

In the post init sequence of v2.9.0, write access to read only registers
are not disabled after updating the registers. Fix it by disabling the
access after register update.

Link: https://lore.kernel.org/r/20230619150408.8468-2-manivannan.sadhasivam@linaro.org
Fixes: 5d76117f070d ("PCI: qcom: Add support for IPQ8074 PCIe controller")
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: <stable@vger.kernel.org>
23 months agoPCI: imx6: Save and restore root port MSI control in suspend and resume
Richard Zhu [Thu, 8 Dec 2022 06:05:34 +0000 (14:05 +0800)]
PCI: imx6: Save and restore root port MSI control in suspend and resume

The imx6 PCI host controller suffers from a HW integration bug whereby
the MSI enable bit in the root port MSI capability enables/disables MSIs
interrupts for all downstream components in the PCI tree.

This requires, as implemented in

75cb8d20c112 ("PCI: imx: Enable MSI from downstream components")

that the root port MSI enable bit should be set in order for downstream
PCI devices MSIs to function.

The MSI enable bit programming might be lost during the suspend and
should be re-stored during resume.

Save the MSI control during suspend and restore it in resume.

Link: https://lore.kernel.org/r/1670479534-22154-1-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
[lpieralisi@kernel.org: commit log]
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
23 months agoPCI/ASPM: Use distinct local vars in pcie_retrain_link()
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:45 +0000 (18:19 +0100)]
PCI/ASPM: Use distinct local vars in pcie_retrain_link()

Use separate local variables to hold the respective values retrieved from
the Link Control Register and the Link Status Register.  Improves
readability and it makes it possible for the compiler to detect actual
uninitialised use should this code change in the future.

[bhelgaas: reorder to clean up before exposing to PCI core]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2306110252260.64925@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agonet/mlx5: Rely on dev->link_active_reporting
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:36 +0000 (18:19 +0100)]
net/mlx5: Rely on dev->link_active_reporting

Use dev->link_active_reporting to determine whether Data Link Layer Link
Active Reporting is available rather than re-retrieving the capability.

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310125370.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agopowerpc/eeh: Rely on dev->link_active_reporting
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:32 +0000 (18:19 +0100)]
powerpc/eeh: Rely on dev->link_active_reporting

Use dev->link_active_reporting to determine whether Data Link Layer Link
Active Reporting is available rather than re-retrieving the capability.

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310124100.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoPCI: pciehp: Rely on dev->link_active_reporting
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:14 +0000 (18:19 +0100)]
PCI: pciehp: Rely on dev->link_active_reporting

Use dev->link_active_reporting to determine whether Data Link Layer Link
Active Reporting is available rather than re-retrieving the capability.

Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310028150.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lukas Wunner <lukas@wunner.de>
23 months agoPCI: Initialize dev->link_active_reporting earlier
Maciej W. Rozycki [Sun, 11 Jun 2023 17:19:27 +0000 (18:19 +0100)]
PCI: Initialize dev->link_active_reporting earlier

Determine whether Data Link Layer Link Active Reporting is available before
calling any fixups so that the cached value can be used there and later on.

[bhelgaas: move to set_pcie_port_type() where other PCIe init is done]
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2305310122210.59226@angie.orcam.me.uk
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
23 months agoDocumentation: PCI: Tidy AER documentation
Bjorn Helgaas [Fri, 9 Jun 2023 22:25:00 +0000 (17:25 -0500)]
Documentation: PCI: Tidy AER documentation

Consistently use:

  PCIe          previously PCIe, PCI Express, or pci express
  Root Port     previously Root Port or root port
  Endpoint      previously EndPoint or endpoint
  AER           previously AER or aer
  please        previously pls

Also update a few awkward wordings.

Link: https://lore.kernel.org/r/20230609222500.1267795-5-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
23 months agoDocumentation: PCI: Update cross references to .rst files
Bjorn Helgaas [Fri, 9 Jun 2023 22:24:59 +0000 (17:24 -0500)]
Documentation: PCI: Update cross references to .rst files

Change references to *.txt to *.rst to match the current filenames.

Link: https://lore.kernel.org/r/20230609222500.1267795-4-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
23 months agoDocumentation: PCI: Drop recommendation to configure AER Capability
Bjorn Helgaas [Fri, 9 Jun 2023 22:24:58 +0000 (17:24 -0500)]
Documentation: PCI: Drop recommendation to configure AER Capability

Since f26e58bf6f54 ("PCI/AER: Enable error reporting when AER is native"),
the PCI core enables PCIe device error reporting for all devices during
enumeration, so drivers don't need to do it.

Remove the recommendation for drivers to configure AER and call
pci_enable_pcie_error_reporting() themselves.

Also remove the suggestion that drivers may change AER mask and severity
registers.  Ownership of these registers is negotiated between the OS and
platform firmware.  If firmware owns these registers, the OS must not
change them.

Link: https://lore.kernel.org/r/20230609222500.1267795-3-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
23 months agoPCI: Unexport pci_save_aer_state()
Bjorn Helgaas [Fri, 9 Jun 2023 22:24:57 +0000 (17:24 -0500)]
PCI: Unexport pci_save_aer_state()

pci_save_aer_state() and pci_restore_aer_state() are only used in
drivers/pci, so don't expose them to the rest of the kernel.  No functional
change intended.

Link: https://lore.kernel.org/r/20230609222500.1267795-2-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
23 months agoPCI: Release resource invalidated by coalescing
Ross Lagerwall [Thu, 25 May 2023 15:32:48 +0000 (16:32 +0100)]
PCI: Release resource invalidated by coalescing

When contiguous windows are coalesced by pci_register_host_bridge(), the
second resource is expanded to include the first, and the first is
invalidated and consequently not added to the bus. However, it remains in
the resource hierarchy.  For example, these windows:

  fec00000-fec7ffff : PCI Bus 0000:00
  fec80000-fecbffff : PCI Bus 0000:00

are coalesced into this, where the first resource remains in the tree with
start/end zeroed out:

  00000000-00000000 : PCI Bus 0000:00
  fec00000-fecbffff : PCI Bus 0000:00

In some cases (e.g. the Xen scratch region), this causes future calls to
allocate_resource() to choose an inappropriate location which the caller
cannot handle.

Fix by releasing the zeroed-out resource and removing it from the resource
hierarchy.

[bhelgaas: commit log]
Fixes: 7c3855c423b1 ("PCI: Coalesce host bridge contiguous apertures")
Link: https://lore.kernel.org/r/20230525153248.712779-1-ross.lagerwall@citrix.com
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org # v5.16+
23 months agoPCI: Add function 1 DMA alias quirk for Marvell 88SE9235
Robin Murphy [Wed, 7 Jun 2023 17:18:47 +0000 (18:18 +0100)]
PCI: Add function 1 DMA alias quirk for Marvell 88SE9235

Marvell's own product brief implies the 92xx series are a closely related
family, and sure enough it turns out that 9235 seems to need the same quirk
as the other three, although possibly only when certain ports are used.

Link: https://lore.kernel.org/linux-iommu/2a699a99-545c-1324-e052-7d2f41fed1ae@yahoo.co.uk/
Link: https://lore.kernel.org/r/731507e05d70239aec96fcbfab6e65d8ce00edd2.1686157165.git.robin.murphy@arm.com
Reported-by: Jason Adriaanse <jason_a69@yahoo.co.uk>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: stable@vger.kernel.org
23 months agoPCI/PM: Shorten pci_bridge_wait_for_secondary_bus() wait time for slow links
Mika Westerberg [Tue, 25 Apr 2023 06:47:51 +0000 (09:47 +0300)]
PCI/PM: Shorten pci_bridge_wait_for_secondary_bus() wait time for slow links

With slow links (<= 5GT/s) active link reporting is not mandatory, so if a
device is disconnected during system sleep we might end up waiting for it
to respond for ~60s, which slows down resume time.

PCIe r6.0, sec 6.6.1, mandates that software must wait for at least 1s
before it can assume a device is broken, so use that minimum requirement
for slow links and bail out if the device doesn't respond within 1s.
However, if the port supports active link reporting we can wait longer as
we do with the fast links.

This should make system resume time faster for slow links as well while
still following the PCIe spec.

While there move the PCI_RESET_WAIT constant into pci.c because it is
not used outside of that file anymore.

Link: https://lore.kernel.org/r/20230425064751.24951-1-mika.westerberg@linux.intel.com
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
23 months agoPCI: of: Propagate firmware node by calling device_set_node()
Andy Shevchenko [Fri, 21 Apr 2023 10:09:39 +0000 (13:09 +0300)]
PCI: of: Propagate firmware node by calling device_set_node()

Insulate pci_set_of_node() and pci_set_bus_of_node() from possible
changes to fwnode_handle implementation by using device_set_node()
instead of open-coding dev->dev.fwnode assignments.

Link: https://lore.kernel.org/r/20230421100939.68225-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: Add PCI_EXT_CAP_ID_PL_32GT define
Ben Dooks [Wed, 31 May 2023 09:57:13 +0000 (10:57 +0100)]
PCI: Add PCI_EXT_CAP_ID_PL_32GT define

Add the define for PCI_EXT_CAP_ID_PL_32GT for drivers that will want this
whilst doing Gen5/Gen6 accesses.

Link: https://lore.kernel.org/r/20230531095713.293229-1-ben.dooks@codethink.co.uk
Signed-off-by: Ben Dooks <ben.dooks@sifive.com>
Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>