linux-2.6-microblaze.git
3 years agoMerge branch 'remotes/lorenzo/pci/mvebu'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:42 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/mvebu'

- Remove useless msi_controller pointer allocation (Lorenzo Pieralisi)

* remotes/lorenzo/pci/mvebu:
  PCI: mvebu: Remove useless msi_controller pointer allocation

3 years agoMerge branch 'remotes/lorenzo/pci/mobiveil'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:42 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/mobiveil'

- Simplify mobiveil_pcie_init_irq_domain() (Liu Shixin)

* remotes/lorenzo/pci/mobiveil:
  PCI: mobiveil: Simplify mobiveil_pcie_init_irq_domain() return expression

3 years agoMerge branch 'remotes/lorenzo/pci/meson'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:41 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/meson'

- Add pci-meson module support and enable by default on ARCH_MESON (Kevin
  Hilman)

* remotes/lorenzo/pci/meson:
  PCI: meson: Build as module by default

3 years agoMerge branch 'remotes/lorenzo/pci/loongson'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:41 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/loongson'

* remotes/lorenzo/pci/loongson:
  PCI: loongson: Simplify loongson_pci_probe() return expression

3 years agoMerge branch 'remotes/lorenzo/pci/kirin'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:41 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/kirin'

- Return -EPROBE_DEFER in case the gpio isn't ready (Bean Huo)

* remotes/lorenzo/pci/kirin:
  PCI: kirin: Return -EPROBE_DEFER in case the gpio isn't ready

3 years agoMerge branch 'remotes/lorenzo/pci/iproc'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:40 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/iproc'

- Set affinity mask on MSI interrupts (Mark Tomlinson)

- Simplify by using module_bcma_driver (Liu Shixin)

- Fix 'using integer as NULL pointer' warning (Krzysztof Wilczyński)

* remotes/lorenzo/pci/iproc:
  PCI: iproc: Fix using plain integer as NULL pointer in iproc_pcie_pltfm_probe
  PCI: iproc: Use module_bcma_driver to simplify the code
  PCI: iproc: Set affinity mask on MSI interrupts

3 years agoMerge branch 'remotes/lorenzo/pci/imx6'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:40 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/imx6'

- Use "fallthrough" pseudo-keyword (Gustavo A. R. Silva)

- Drop redundant error messages after devm_clk_get() (Anson Huang)

* remotes/lorenzo/pci/imx6:
  PCI: imx6: Do not output error message when devm_clk_get() failed with -EPROBE_DEFER
  PCI: imx6: Use fallthrough pseudo-keyword

3 years agoMerge branch 'remotes/lorenzo/pci/hv'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:39 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/hv'

- Fix hibernation in case interrupts are not re-created (Dexuan Cui)

* remotes/lorenzo/pci/hv:
  PCI: hv: Fix hibernation in case interrupts are not re-created

3 years agoMerge branch 'remotes/lorenzo/pci/dwc'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:39 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/dwc'

- Fix designware-ep Header Type check (Hou Zhiqiang)

- Use DBI accessors instead of own config accessors (Rob Herring)

- Allow overriding bridge pci_ops (Rob Herring)

- Allow root and child buses to have different pci_ops (Rob Herring)

- Add default dwc pci_ops.map_bus (Rob Herring)

- Use pci_ops for root config space accessors in al, exynos, histb,
  keystone, kirin, meson, tegra (Rob Herring)

- Remove dwc own/other config accessor ops (Rob Herring)

- Use generic config accessors in dwc (Rob Herring)

- Also call .add_bus() callback for root bus (Rob Herring)

- Convert keystone .scan_bus() callback to use pci_ops.add_bus (Rob
  Herring)

- Convert dwc to use pci_host_probe() (Rob Herring)

- Remove dwc root_bus pointer (Rob Herring)

- Remove storing of PCI resources in dwc-specific structs (Rob Herring)

- Simplify config space handling (Rob Herring)

- Drop keystone duplicated DT num-viewport handling (Rob Herring)

- Check CONFIG_PCI_MSI in dw_pcie_msi_init() instead of duplicating it in
  all the drivers (Rob Herring)

- Remove imx6 duplicate PCIE_LINK_WIDTH_SPEED_CONTROL definition (Rob
  Herring)

- Add dwc num_lanes for use when it's lacking from DT (Rob Herring)

- Ensure "Fast Link Mode" simulation environment setting is cleared (Rob
  Herring)

- Drop meson duplicate number of lanes setup (Rob Herring)

- Drop meson unnecessary RC config space init (Rob Herring)

- Rework meson config and dwc port logic register accesses (Rob Herring)

- Use common PCI register definitions in imx6 and qcom (Rob Herring)

- Search for DesignWare PCIe Capability instead of hard-coding its location
  (Rob Herring)

- Use common DesignWare register definitions in tegra (Rob Herring)

- Drop keystone unused DBI2 code (Rob Herring)

- Make dwc ATU accessors private (Rob Herring)

- Centralize link gen setting in dwc (Rob Herring)

- Set PORT_LINK_DLL_LINK_EN in common dwc setup code (Rob Herring)

- Drop intel-gw unnecessary DT 'device_type' checking (Rob Herring)

- Move intel-gw PCI_CAP_ID_EXP discovery to the single place it's used (Rob
  Herring)

- Drop intel-gw unused max_width (Rob Herring)

- Move N_FTS (fast training sequence) setup to common dwc setup (Rob
  Herring)

- Convert spear13xx, tegra194 to use DBI accessors (Rob Herring)

- Add multiple PFs support for DWC (Xiaowei Bao)

- Add MSI-X doorbell mode for endpoint mode (Xiaowei Bao)

- Update MSI/MSI-X capability management for endpoints (Xiaowei Bao)

- Add layerscape ls1088a and ls2088a compatible strings (Xiaowei Bao)

- Update layerscape MSI/MSI-X management (Xiaowei Bao)

- Use doorbell to support MSI-X on layerscape (Xiaowei Bao)

- Add layerscape endpoint mode support for ls1088a and ls2088a (Xiaowei
  Bao)

- Add layerscape ls1088a node to DT (Xiaowei Bao)

- Add Freescale/Layerscape ls1088a to endpoint test (Xiaowei Bao)

- Add endpoint test driver data for Layerscape PCIe controllers (Hou
  Zhiqiang)

- Fix 'cast truncates bits from constant value' warning (Gustavo Pimentel)

- Add uniphier iATU register description (Kunihiko Hayashi)

- Add common iATU register support (Kunihiko Hayashi)

- Remove keystone iATU register mapping in favor of generic dwc support
  (Kunihiko Hayashi)

- Skip PCIE_MSI_INTR0* programming if MSI is disabled (Jisheng Zhang)

- Fix MSI page leakage in suspend/resume (Jisheng Zhang)

- Check whether link is up before attempting config access (best-effort fix
  even though it's racy) (Hou Zhiqiang)

* remotes/lorenzo/pci/dwc:
  PCI: dwc: Add link up check in dw_child_pcie_ops.map_bus()
  PCI: dwc: Fix MSI page leakage in suspend/resume
  PCI: dwc: Skip PCIE_MSI_INTR0* programming if MSI is disabled
  PCI: keystone: Remove iATU register mapping
  PCI: dwc: Add common iATU register support
  dt-bindings: PCI: uniphier-ep: Add iATU register description
  dt-bindings: PCI: uniphier: Add iATU register description
  PCI: dwc: Fix 'cast truncates bits from constant value'
  misc: pci_endpoint_test: Add driver data for Layerscape PCIe controllers
  misc: pci_endpoint_test: Add LS1088a in pci_device_id table
  PCI: layerscape: Add EP mode support for ls1088a and ls2088a
  PCI: layerscape: Modify the MSIX to the doorbell mode
  PCI: layerscape: Modify the way of getting capability with different PEX
  PCI: layerscape: Fix some format issue of the code
  dt-bindings: pci: layerscape-pci: Add compatible strings for ls1088a and ls2088a
  PCI: designware-ep: Modify MSI and MSIX CAP way of finding
  PCI: designware-ep: Move the function of getting MSI capability forward
  PCI: designware-ep: Add the doorbell mode of MSI-X in EP mode
  PCI: designware-ep: Add multiple PFs support for DWC
  PCI: dwc: Use DBI accessors
  PCI: dwc: Move N_FTS setup to common setup
  PCI: dwc/intel-gw: Drop unused max_width
  PCI: dwc/intel-gw: Move getting PCI_CAP_ID_EXP offset to intel_pcie_link_setup()
  PCI: dwc/intel-gw: Drop unnecessary checking of DT 'device_type' property
  PCI: dwc: Set PORT_LINK_DLL_LINK_EN in common setup code
  PCI: dwc: Centralize link gen setting
  PCI: dwc: Make ATU accessors private
  PCI: dwc: Remove read_dbi2 code
  PCI: dwc/tegra: Use common Designware port logic register definitions
  PCI: dwc: Remove hardcoded PCI_CAP_ID_EXP offset
  PCI: dwc/qcom: Use common PCI register definitions
  PCI: dwc/imx6: Use common PCI register definitions
  PCI: dwc/meson: Rework PCI config and DW port logic register accesses
  PCI: dwc/meson: Drop unnecessary RC config space initialization
  PCI: dwc/meson: Drop the duplicate number of lanes setup
  PCI: dwc: Ensure FAST_LINK_MODE is cleared
  PCI: dwc: Add a 'num_lanes' field to struct dw_pcie
  PCI: dwc/imx6: Remove duplicate define PCIE_LINK_WIDTH_SPEED_CONTROL
  PCI: dwc: Check CONFIG_PCI_MSI inside dw_pcie_msi_init()
  PCI: dwc/keystone: Drop duplicated 'num-viewport'
  PCI: dwc: Simplify config space handling
  PCI: dwc: Remove storing of PCI resources
  PCI: dwc: Remove root_bus pointer
  PCI: dwc: Convert to use pci_host_probe()
  PCI: dwc: keystone: Convert .scan_bus() callback to use add_bus
  PCI: Also call .add_bus() callback for root bus
  PCI: dwc: Use generic config accessors
  PCI: dwc: Remove dwc specific config accessor ops
  PCI: dwc: histb: Use pci_ops for root config space accessors
  PCI: dwc: exynos: Use pci_ops for root config space accessors
  PCI: dwc: kirin: Use pci_ops for root config space accessors
  PCI: dwc: meson: Use pci_ops for root config space accessors
  PCI: dwc: tegra: Use pci_ops for root config space accessors
  PCI: dwc: keystone: Use pci_ops for config space accessors
  PCI: dwc: al: Use pci_ops for child config space accessors
  PCI: dwc: Add a default pci_ops.map_bus for root port
  PCI: dwc: Allow overriding bridge pci_ops
  PCI: dwc: Use DBI accessors instead of own config accessors
  PCI: Allow root and child buses to have different pci_ops
  PCI: designware-ep: Fix the Header Type check

3 years agoMerge branch 'remotes/lorenzo/pci/cadence'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:38 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/cadence'

- Remove obsolete path from comment (Flavio Suligoi)

- Simplify cdns_pcie_host_init_address_translation() (Qinglang Miao)

* remotes/lorenzo/pci/cadence:
  PCI: cadence: Simplify cdns_pcie_host_init_address_translation() return expression
  PCI: cadence-ep: Remove obsolete path from comment

3 years agoMerge branch 'remotes/lorenzo/pci/brcmstb'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:38 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/brcmstb'

- Make PCIE_BRCMSTB depend on and default to ARCH_BRCMSTB (Jim Quinlan)

- Add DT bindings for 7278, 7216, 7211, and new properties (Jim Quinlan)

- Add bcm7278 register info (Jim Quinlan)

- Add suspend and resume pm_ops (Jim Quinlan)

- Add bcm7278 PERST# support (Jim Quinlan)

- Add control of RESCAL reset (Jim Quinlan)

- Set additional internal memory DMA viewport sizes (Jim Quinlan)

- Accommodate MSI for older chips (Jim Quinlan)

- Set bus max burst size by chip type (Jim Quinlan)

- Add bcm7211, bcm7216, bcm7445, bcm7278 to match list (Jim Quinlan)

* remotes/lorenzo/pci/brcmstb:
  PCI: brcmstb: Add bcm7211, bcm7216, bcm7445, bcm7278 to match list
  PCI: brcmstb: Set bus max burst size by chip type
  PCI: brcmstb: Accommodate MSI for older chips
  PCI: brcmstb: Set additional internal memory DMA viewport sizes
  PCI: brcmstb: Add control of rescal reset
  PCI: brcmstb: Add bcm7278 PERST# support
  PCI: brcmstb: Add suspend and resume pm_ops
  PCI: brcmstb: Add bcm7278 register info
  dt-bindings: PCI: Add bindings for more Brcmstb chips
  PCI: brcmstb: PCIE_BRCMSTB depends on ARCH_BRCMSTB

3 years agoMerge branch 'remotes/lorenzo/pci/aardvark'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:37 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/aardvark'

- Fix s390 build error (Pali Rohár)

- Check for errors from pci_bridge_emul_init() (Pali Rohár)

- Export pci-bridge-emul functions for use by modules (Pali Rohár)

- Make aardvark driver modular (Pali Rohár)

- Move PCIe reset code to advk_pcie_train_link() (Pali Rohár)

- Convert internal SMCC firmware return codes to errno (Pali Rohár)

- Fix initialization with old Marvell's Arm Trusted Firmware (Pali Rohár)

* remotes/lorenzo/pci/aardvark:
  PCI: aardvark: Fix initialization with old Marvell's Arm Trusted Firmware
  phy: marvell: comphy: Convert internal SMCC firmware return codes to errno
  PCI: aardvark: Move PCIe reset card code to advk_pcie_train_link()
  PCI: aardvark: Implement driver 'remove' function and allow to build it as module
  PCI: pci-bridge-emul: Export API functions
  PCI: aardvark: Check for errors from pci_bridge_emul_init() call
  PCI: aardvark: Fix compilation on s390

3 years agoMerge branch 'remotes/lorenzo/pci/arm'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:37 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/arm'

- Remove unused msi_ctrl, io_optional and align_resource fields from ARM
  struct hw_pci (Lorenzo Pieralisi)

* remotes/lorenzo/pci/arm:
  ARM/PCI: Remove unused fields from struct hw_pci

3 years agoMerge branch 'remotes/lorenzo/pci/pci-iomap'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:37 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/pci-iomap'

- Remove useless __KERNEL__ preprocessor guard in sparc io_32.h (Lorenzo
  Pieralisi)

- Move ioremap/iounmap declaration so it's visible in asm-generic/io.h
  (Lorenzo Pieralisi)

- Fix memory leak in generic !CONFIG_GENERIC_IOMAP pci_iounmap()
  implementation (Lorenzo Pieralisi)

* remotes/lorenzo/pci/pci-iomap:
  asm-generic/io.h: Fix !CONFIG_GENERIC_IOMAP pci_iounmap() implementation
  sparc32: Move ioremap/iounmap declaration before asm-generic/io.h include
  sparc32: Remove useless io_32.h __KERNEL__ preprocessor guard

3 years agoMerge branch 'remotes/lorenzo/pci/apei'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:36 +0000 (09:58 -0500)]
Merge branch 'remotes/lorenzo/pci/apei'

- Add ACPI APEI notifier chain for unknown (vendor) CPER records (Shiju
  Jose)

- Add handling of HiSilicon HIP PCIe controller errors (Yicong Yang)

* remotes/lorenzo/pci/apei:
  PCI: hip: Add handling of HiSilicon HIP PCIe controller errors
  ACPI / APEI: Add a notifier chain for unknown (vendor) CPER records

3 years agoMerge branch 'pci/misc'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:36 +0000 (09:58 -0500)]
Merge branch 'pci/misc'

- Remove unnecessary #includes (Gustavo Pimentel)

- Fix intel_mid_pci.c build error when !CONFIG_ACPI (Randy Dunlap)

- Use scnprintf(), not snprintf(), in sysfs "show" functions (Krzysztof
  Wilczyński)

- Simplify pci-pf-stub by using module_pci_driver() (Liu Shixin)

- Print IRQ used by Link Bandwidth Notification (Dongdong Liu)

- Update sysfs mmap-related #ifdef comments (Clint Sbisa)

- Simplify pci_dev_reset_slot_function() (Lukas Wunner)

- Use "NULL" instead of "0" to fix sparse warnings (Gustavo Pimentel)

- Simplify bool comparisons (Krzysztof Wilczyński)

- Drop double zeroing for P2PDMA sg_init_table() (Julia Lawall)

* pci/misc:
  PCI: v3-semi: Remove unneeded break
  PCI/P2PDMA: Drop double zeroing for sg_init_table()
  PCI: Simplify bool comparisons
  PCI: endpoint: Use "NULL" instead of "0" as a NULL pointer
  PCI: Simplify pci_dev_reset_slot_function()
  PCI: Update mmap-related #ifdef comments
  PCI/LINK: Print IRQ number used by port
  PCI/IOV: Simplify pci-pf-stub with module_pci_driver()
  PCI: Use scnprintf(), not snprintf(), in sysfs "show" functions
  x86/PCI: Fix intel_mid_pci.c build error when ACPI is not enabled
  PCI: Remove unnecessary header includes

3 years agoMerge branch 'pci/pm'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:35 +0000 (09:58 -0500)]
Merge branch 'pci/pm'

- Remove unused pcibios_pm_ops (Vaibhav Gupta)

- Rename pci_dev.d3_delay to d3hot_delay (Krzysztof Wilczyński)

- Apply D2 transition delay as microseconds, not milliseconds (Bjorn
  Helgaas)

* pci/pm:
  PCI/PM: Revert "PCI/PM: Apply D2 delay as milliseconds, not microseconds"
  PCI/PM: Remove unused PCI_PM_BUS_WAIT
  PCI/PM: Rename pci_dev.d3_delay to d3hot_delay
  PCI/PM: Remove unused pcibios_pm_ops

3 years agoMerge branch 'pci/hotplug'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:35 +0000 (09:58 -0500)]
Merge branch 'pci/hotplug'

- Use for_each_child_of_node() and for_each_node_by_name() instead of
  open-coding them (Qinglang Miao)

- Reduce pciehp noisiness on hot removal (Lukas Wunner)

- Remove unused assignment in shpchp (Krzysztof Wilczyński)

* pci/hotplug:
  PCI: shpchp: Remove unused 'rc' assignment
  PCI: pciehp: Reduce noisiness on hot removal
  PCI: rpadlpar: Use for_each_child_of_node() and for_each_node_by_name()

3 years agoMerge branch 'pci/enumeration'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:34 +0000 (09:58 -0500)]
Merge branch 'pci/enumeration'

- Tone down message about missing optional MCFG (Jeremy Linton)

- Add schedule point in pci_read_config() (Jiang Biao)

- Add Ampere Altra SOC MCFG quirk (Tuan Phan)

- Add Kconfig options for MPS/MRRS strategy (Jim Quinlan)

* pci/enumeration:
  PCI: Add Kconfig options for MPS/MRRS strategy
  PCI/ACPI: Add Ampere Altra SOC MCFG quirk
  PCI: Add schedule point in pci_read_config()
  PCI/ACPI: Tone down missing MCFG message

3 years agoMerge branch 'pci/aspm'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:34 +0000 (09:58 -0500)]
Merge branch 'pci/aspm'

- Remove struct aspm_register_info (Saheed O. Bolarinwa)

- Remove struct pcie_link_state.l1ss (Saheed O. Bolarinwa)

* pci/aspm:
  PCI/ASPM: Remove struct pcie_link_state.l1ss
  PCI/ASPM: Remove struct aspm_register_info.l1ss_cap
  PCI/ASPM: Pass L1SS Capabilities value, not struct aspm_register_info
  PCI/ASPM: Remove struct aspm_register_info.l1ss_ctl1
  PCI/ASPM: Remove struct aspm_register_info.l1ss_ctl2 (unused)
  PCI/ASPM: Remove struct aspm_register_info.l1ss_cap_ptr
  PCI/ASPM: Remove struct aspm_register_info.latency_encoding
  PCI/ASPM: Remove struct aspm_register_info.enabled
  PCI/ASPM: Remove struct aspm_register_info.support
  PCI/ASPM: Use 'parent' and 'child' for readability
  PCI/ASPM: Move LTR path check to where it's used
  PCI/ASPM: Move pci_clear_and_set_dword() earlier

3 years agoMerge branch 'pci/acs'
Bjorn Helgaas [Wed, 21 Oct 2020 14:58:34 +0000 (09:58 -0500)]
Merge branch 'pci/acs'

- Enable Translation Blocking for external devices (Rajat Jain)

* pci/acs:
  PCI/ACS: Enable Translation Blocking for external devices

3 years agoPCI: v3-semi: Remove unneeded break
Tom Rix [Mon, 19 Oct 2020 19:02:49 +0000 (12:02 -0700)]
PCI: v3-semi: Remove unneeded break

A break is not needed if it is preceded by a return

Link: https://lore.kernel.org/r/20201019190249.7825-1-trix@redhat.com
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
3 years agoPCI: dwc: Add link up check in dw_child_pcie_ops.map_bus()
Hou Zhiqiang [Wed, 16 Sep 2020 05:41:30 +0000 (13:41 +0800)]
PCI: dwc: Add link up check in dw_child_pcie_ops.map_bus()

NXP Layerscape (ls1028a, ls2088a), dra7xxx and imx6 platforms are either
programmed or statically configured to forward the error triggered by a
link-down state (eg no connected endpoint device) on the system bus for
PCI configuration transactions; these errors are reported as an SError
at system level, which is fatal.

Enumerating a PCI tree when the PCIe link is down is not sensible
either, so even if the link-up check is racy (link can go down after
map_bus() is called) add a link-up check in map_bus() to prevent issuing
configuration transactions when the link is down.

SError report:

 SError Interrupt on CPU2, code 0xbf000002 -- SError
 CPU: 2 PID: 1 Comm: swapper/0 Not tainted 5.9.0-rc5-next-20200914-00001-gf965d3ec86fa #67
 Hardware name: LS1046A RDB Board (DT)
 pstate: 20000085 (nzCv daIf -PAN -UAO BTYPE=--)
 pc : pci_generic_config_read+0x3c/0xe0
 lr : pci_generic_config_read+0x24/0xe0
 sp : ffff80001003b7b0
 x29: ffff80001003b7b0 x28: ffff80001003ba74
 x27: ffff000971d96800 x26: ffff00096e77e0a8
 x25: ffff80001003b874 x24: ffff80001003b924
 x23: 0000000000000004 x22: 0000000000000000
 x21: 0000000000000000 x20: ffff80001003b874
 x19: 0000000000000004 x18: ffffffffffffffff
 x17: 00000000000000c0 x16: fffffe0025981840
 x15: ffffb94c75b69948 x14: 62203a383634203a
 x13: 666e6f635f726568 x12: 202c31203d207265
 x11: 626d756e3e2d7375 x10: 656877202c307830
 x9 : 203d206e66766564 x8 : 0000000000000908
 x7 : 0000000000000908 x6 : ffff800010900000
 x5 : ffff00096e77e080 x4 : 0000000000000000
 x3 : 0000000000000003 x2 : 84fa3440ff7e7000
 x1 : 0000000000000000 x0 : ffff800010034000
 Kernel panic - not syncing: Asynchronous SError Interrupt
 CPU: 2 PID: 1 Comm: swapper/0 Not tainted 5.9.0-rc5-next-20200914-00001-gf965d3ec86fa #67
 Hardware name: LS1046A RDB Board (DT)
 Call trace:
  dump_backtrace+0x0/0x1c0
  show_stack+0x18/0x28
  dump_stack+0xd8/0x134
  panic+0x180/0x398
  add_taint+0x0/0xb0
  arm64_serror_panic+0x78/0x88
  do_serror+0x68/0x180
  el1_error+0x84/0x100
  pci_generic_config_read+0x3c/0xe0
  dw_pcie_rd_other_conf+0x78/0x110
  pci_bus_read_config_dword+0x88/0xe8
  pci_bus_generic_read_dev_vendor_id+0x30/0x1b0
  pci_bus_read_dev_vendor_id+0x4c/0x78
  pci_scan_single_device+0x80/0x100

Link: https://lore.kernel.org/r/20200916054130.8685-1-Zhiqiang.Hou@nxp.com
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
[lorenzo.pieralisi@arm.com: rewrote the commit log, remove Fixes tag]
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
3 years agoPCI/ASPM: Remove struct pcie_link_state.l1ss
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:39 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct pcie_link_state.l1ss

Previously we computed L1.2 parameters in the enumeration path, saved them
in struct pcie_link_state.l1ss, and programmed them into the devices
whenever we enabled or disabled L1.2 on the link.  But these parameters are
constant and don't need to be updated when enabling/disabling L1.2.

Compute and program the L1.2 parameters once during enumeration and remove
the struct pcie_link_state.l1ss member.  No functional change intended.

[bhelgaas: rework to program L1.2 parameters during enumeration]
Link: https://lore.kernel.org/r/20201015193039.12585-13-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.l1ss_cap
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:38 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.l1ss_cap

Previously we stored the L1SS Capabilities value in the struct
aspm_register_info.

We only need this information in one place, so read it there and remove
struct aspm_register_info completely, since it's now empty.  No functional
change intended.

[bhelgaas: split up, don't cache l1ss_cap in pci_dev]
Link: https://lore.kernel.org/r/20201015193039.12585-12-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Pass L1SS Capabilities value, not struct aspm_register_info
Bjorn Helgaas [Thu, 15 Oct 2020 19:30:37 +0000 (14:30 -0500)]
PCI/ASPM: Pass L1SS Capabilities value, not struct aspm_register_info

aspm_calc_l1ss_info() needs only the L1SS Capabilities.  It doesn't need
anything else from struct aspm_register_info, so pass only the Capabilities
value.  No functional change intended.

Link: https://lore.kernel.org/r/20201015193039.12585-11-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.l1ss_ctl1
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:36 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.l1ss_ctl1

Previously we stored the L1SS Control 1 register in the struct
aspm_register_info.

We only need this information in one place, so read it there and remove it
from struct aspm_register_info.  No functional change intended.

[bhelgaas: split ctl1/ctl2]
Link: https://lore.kernel.org/r/20201015193039.12585-10-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.l1ss_ctl2 (unused)
Bjorn Helgaas [Thu, 15 Oct 2020 19:30:35 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.l1ss_ctl2 (unused)

We never use the aspm_register_info.l1ss_ctl2 value, so remove it.  No
functional change intended.

Link: https://lore.kernel.org/r/20201015193039.12585-9-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.l1ss_cap_ptr
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:34 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.l1ss_cap_ptr

Save the L1 Substates Capability pointer in struct pci_dev.  Then we don't
have to keep track of it in the struct aspm_register_info and struct
pcie_link_state, which makes the code easier to read.  No functional change
intended.

[bhelgaas: split to a separate patch]
Link: https://lore.kernel.org/r/20201015193039.12585-8-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.latency_encoding
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:33 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.latency_encoding

Previously we stored L0s and L1 Exit Latency information from the Link
Capabilities register in the struct aspm_register_info.

We only need these latencies when we already have the Link Capabilities
values, so use those directly and remove the latencies from struct
aspm_register_info.  No functional change intended.

Link: https://lore.kernel.org/r/20201015193039.12585-7-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.enabled
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:32 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.enabled

Previously we stored the "ASPM Control" bits from the Link Control register
in the struct aspm_register_info.

Read PCI_EXP_LNKCTL directly when needed.  This means we can use the
PCI_EXP_LNKCTL_ASPM_* bits directly instead of the similar but different
PCIE_LINK_STATE_* bits.  No functional change intended.

[bhelgaas: drop get_aspm_enable() and read LNKCTL once directly]
Link: https://lore.kernel.org/r/20201015193039.12585-6-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Remove struct aspm_register_info.support
Saheed O. Bolarinwa [Thu, 15 Oct 2020 19:30:31 +0000 (14:30 -0500)]
PCI/ASPM: Remove struct aspm_register_info.support

Previously we stored the "ASPM Support" field from the Link Capabilities
register in the struct aspm_register_info.

Read the Link Capabilities directly when needed and remove it from the
struct aspm_register_info.  No functional change intended.

[bhelgaas: remove pci_dev cached copy since LNKCAP isn't truly read-only,
add PCI_EXP_LNKCAP_ASPM_L0S & PCI_EXP_LNKCAP_ASPM_L1, check them directly
instead of adding aspm_support()]
Link: https://lore.kernel.org/r/20201015193039.12585-5-helgaas@kernel.org
Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Use 'parent' and 'child' for readability
Bjorn Helgaas [Thu, 15 Oct 2020 19:30:30 +0000 (14:30 -0500)]
PCI/ASPM: Use 'parent' and 'child' for readability

Other users of link->pdev and link->downstream, e.g., pcie_aspm_cap_init(),
pcie_config_aspm_l1ss(), and pcie_config_aspm_link(), use "parent" and
"child" as local names.

Do the same in aspm_calc_l1ss_info() for readability.  No functional change
intended.

Link: https://lore.kernel.org/r/20201015193039.12585-4-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Move LTR path check to where it's used
Bjorn Helgaas [Thu, 15 Oct 2020 19:30:29 +0000 (14:30 -0500)]
PCI/ASPM: Move LTR path check to where it's used

pcie_get_aspm_reg() mostly reads ASPM-related registers, but in some cases
it also updates the value read from PCI_L1SS_CAP based on LTR properties.

Move this update to the point where the value is used to make the code more
readable.

No functional change intended, although previously we could clear
PCI_L1SS_CAP_ASPM_L1_2 for both ends of the link, and now we'll only do it
for the downstream end of a link.  This shouldn't matter because we always
test that bit by ANDing l1ss_cap for the upstream and downstream ends.

Link: https://lore.kernel.org/r/20201015193039.12585-3-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/ASPM: Move pci_clear_and_set_dword() earlier
Bjorn Helgaas [Thu, 15 Oct 2020 19:30:28 +0000 (14:30 -0500)]
PCI/ASPM: Move pci_clear_and_set_dword() earlier

Move pci_clear_and_set_dword() earlier in file to prepare for future patch.
No functional change intended.

Link: https://lore.kernel.org/r/20201015193039.12585-2-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: dwc: Fix MSI page leakage in suspend/resume
Jisheng Zhang [Fri, 9 Oct 2020 07:55:05 +0000 (15:55 +0800)]
PCI: dwc: Fix MSI page leakage in suspend/resume

Currently, dw_pcie_msi_init() allocates and maps page for msi, then
program the PCIE_MSI_ADDR_LO and PCIE_MSI_ADDR_HI. The Root Complex
may lose power during suspend-to-RAM, so when we resume, we want to
redo the latter but not the former. If designware based driver (for
example, pcie-tegra194.c) calls dw_pcie_msi_init() in resume path, the
msi page will be leaked.

As pointed out by Rob and Ard, there's no need to allocate a page for
the MSI address, we could use an address in the driver data.

To avoid map the MSI msg again during resume, we move the map MSI msg
from dw_pcie_msi_init() to dw_pcie_host_init().

Suggested-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20201009155505.5a580ef5@xhacker.debian
Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: dwc: Skip PCIE_MSI_INTR0* programming if MSI is disabled
Jisheng Zhang [Fri, 9 Oct 2020 07:54:36 +0000 (15:54 +0800)]
PCI: dwc: Skip PCIE_MSI_INTR0* programming if MSI is disabled

If MSI is disabled, there's no need to program PCIE_MSI_INTR0_MASK
and PCIE_MSI_INTR0_ENABLE registers.

Link: https://lore.kernel.org/r/20201009155436.27e67238@xhacker.debian
Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
3 years agoPCI: keystone: Remove iATU register mapping
Kunihiko Hayashi [Wed, 30 Sep 2020 05:36:07 +0000 (14:36 +0900)]
PCI: keystone: Remove iATU register mapping

After applying "PCI: dwc: Add common iATU register support",
there is no need to set own iATU in the Keystone driver itself.

Suggested-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/1601444167-11316-5-git-send-email-hayashi.kunihiko@socionext.com
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: Murali Karicheri <m-karicheri2@ti.com>
Cc: Jingoo Han <jingoohan1@gmail.com>
Cc: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
3 years agoPCI: dwc: Add common iATU register support
Kunihiko Hayashi [Wed, 30 Sep 2020 05:36:06 +0000 (14:36 +0900)]
PCI: dwc: Add common iATU register support

This gets iATU register area from reg property that has reg-names "atu".
In Synopsys DWC version 4.80 or later, since iATU register area is
separated from core register area, this area is necessary to get from
DT independently.

Suggested-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/1601444167-11316-4-git-send-email-hayashi.kunihiko@socionext.com
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: Murali Karicheri <m-karicheri2@ti.com>
Cc: Jingoo Han <jingoohan1@gmail.com>
Cc: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
3 years agodt-bindings: PCI: uniphier-ep: Add iATU register description
Kunihiko Hayashi [Wed, 30 Sep 2020 05:36:05 +0000 (14:36 +0900)]
dt-bindings: PCI: uniphier-ep: Add iATU register description

In the dt-bindings, "atu" reg-names is required to get the register space
for iATU in Synopsis DWC version 4.80 or later.

Link: https://lore.kernel.org/r/1601444167-11316-3-git-send-email-hayashi.kunihiko@socionext.com
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agodt-bindings: PCI: uniphier: Add iATU register description
Kunihiko Hayashi [Wed, 30 Sep 2020 05:36:04 +0000 (14:36 +0900)]
dt-bindings: PCI: uniphier: Add iATU register description

In the dt-bindings, "atu" reg-names is required to get the register space
for iATU in Synopsys DWC version 4.80 or later.

Link: https://lore.kernel.org/r/1601444167-11316-2-git-send-email-hayashi.kunihiko@socionext.com
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
3 years agoPCI: iproc: Fix using plain integer as NULL pointer in iproc_pcie_pltfm_probe
Krzysztof Wilczyński [Tue, 22 Sep 2020 19:49:32 +0000 (19:49 +0000)]
PCI: iproc: Fix using plain integer as NULL pointer in iproc_pcie_pltfm_probe

Fix sparse build warning:

  drivers/pci/controller/pcie-iproc-platform.c:102:33: warning: Using plain integer as NULL pointer

The map_irq member of the struct iproc_pcie takes a function pointer
serving as a callback to map interrupts, therefore we should pass a NULL
pointer to it rather than a integer in the iproc_pcie_pltfm_probe()
function.

Related:
  commit b64aa11eb2dd ("PCI: Set bridge map_irq and swizzle_irq to
  default functions")

Link: https://lore.kernel.org/r/20200922194932.465925-1-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: meson: Build as module by default
Kevin Hilman [Fri, 18 Sep 2020 18:12:51 +0000 (11:12 -0700)]
PCI: meson: Build as module by default

Enable pci-meson to build as a module whenever ARCH_MESON is enabled.

Link: https://lore.kernel.org/r/20200918181251.32423-1-khilman@baylibre.com
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Cc: Yue Wang <yue.wang@amlogic.com>
3 years agoasm-generic/io.h: Fix !CONFIG_GENERIC_IOMAP pci_iounmap() implementation
Lorenzo Pieralisi [Wed, 16 Sep 2020 11:06:58 +0000 (12:06 +0100)]
asm-generic/io.h: Fix !CONFIG_GENERIC_IOMAP pci_iounmap() implementation

For arches that do not select CONFIG_GENERIC_IOMAP, the current
pci_iounmap() function does nothing causing obvious memory leaks
for mapped regions that are backed by MMIO physical space.

In order to detect if a mapped pointer is IO vs MMIO, a check must made
available to the pci_iounmap() function so that it can actually detect
whether the pointer has to be unmapped.

In configurations where CONFIG_HAS_IOPORT_MAP && !CONFIG_GENERIC_IOMAP,
a mapped port is detected using an ioport_map() stub defined in
asm-generic/io.h.

Use the same logic to implement a stub (ie __pci_ioport_unmap()) that
detects if the passed in pointer in pci_iounmap() is IO vs MMIO to
iounmap conditionally and call it in pci_iounmap() fixing the issue.

Leave __pci_ioport_unmap() as a NOP for all other config options.

Tested-by: George Cherian <george.cherian@marvell.com>
Link: https://lore.kernel.org/lkml/20200905024811.74701-1-yangyingliang@huawei.com
Link: https://lore.kernel.org/lkml/20200824132046.3114383-1-george.cherian@marvell.com
Link: https://lore.kernel.org/r/a9daf8d8444d0ebd00bc6d64e336ec49dbb50784.1600254147.git.lorenzo.pieralisi@arm.com
Reported-by: George Cherian <george.cherian@marvell.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: George Cherian <george.cherian@marvell.com>
Cc: Will Deacon <will@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Yang Yingliang <yangyingliang@huawei.com>
3 years agosparc32: Move ioremap/iounmap declaration before asm-generic/io.h include
Lorenzo Pieralisi [Wed, 16 Sep 2020 11:06:57 +0000 (12:06 +0100)]
sparc32: Move ioremap/iounmap declaration before asm-generic/io.h include

Move the ioremap/iounmap declaration before asm-generic/io.h is
included so that it is visible within it.

Link: https://lore.kernel.org/r/93e2f23cda474a92a4708d4c50c9c359426a2162.1600254147.git.lorenzo.pieralisi@arm.com
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: "David S. Miller" <davem@davemloft.net>
Cc: "David S. Miller" <davem@davemloft.net>
3 years agosparc32: Remove useless io_32.h __KERNEL__ preprocessor guard
Lorenzo Pieralisi [Wed, 16 Sep 2020 11:06:56 +0000 (12:06 +0100)]
sparc32: Remove useless io_32.h __KERNEL__ preprocessor guard

The __KERNEL_ preprocessor guard is useless in non-uapi headers.

Remove it.

Link: https://lore.kernel.org/r/084753d3064fe946ff1963eda2eb425cfd7daa7b.1600254147.git.lorenzo.pieralisi@arm.com
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: David S. Miller <davem@davemloft.net>
3 years agoPCI: aardvark: Fix initialization with old Marvell's Arm Trusted Firmware
Pali Rohár [Wed, 2 Sep 2020 14:43:44 +0000 (16:43 +0200)]
PCI: aardvark: Fix initialization with old Marvell's Arm Trusted Firmware

Old ATF automatically power on pcie phy and does not provide SMC call for
phy power on functionality which leads to aardvark initialization failure:

[    0.330134] mvebu-a3700-comphy d0018300.phy: unsupported SMC call, try updating your firmware
[    0.338846] phy phy-d0018300.phy.1: phy poweron failed --> -95
[    0.344753] advk-pcie d0070000.pcie: Failed to initialize PHY (-95)
[    0.351160] advk-pcie: probe of d0070000.pcie failed with error -95

This patch fixes above failure by ignoring 'not supported' error in
aardvark driver. In this case it is expected that phy is already power on.

Tested-by: Tomasz Maciej Nowak <tmn505@gmail.com>
Link: https://lore.kernel.org/r/20200902144344.16684-3-pali@kernel.org
Fixes: 366697018c9a ("PCI: aardvark: Add PHY support")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: <stable@vger.kernel.org> # 5.8+: ea17a0f153af: phy: marvell: comphy: Convert internal SMCC firmware return codes to errno
3 years agophy: marvell: comphy: Convert internal SMCC firmware return codes to errno
Pali Rohár [Wed, 2 Sep 2020 14:43:43 +0000 (16:43 +0200)]
phy: marvell: comphy: Convert internal SMCC firmware return codes to errno

Driver ->power_on and ->power_off callbacks leaks internal SMCC firmware
return codes to phy caller. This patch converts SMCC error codes to
standard linux errno codes. Include file linux/arm-smccc.h already provides
defines for SMCC error codes, so use them instead of custom driver defines.
Note that return value is signed 32bit, but stored in unsigned long type
with zero padding.

Tested-by: Tomasz Maciej Nowak <tmn505@gmail.com>
Link: https://lore.kernel.org/r/20200902144344.16684-2-pali@kernel.org
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: loongson: Simplify loongson_pci_probe() return expression
Qinglang Miao [Mon, 21 Sep 2020 13:10:54 +0000 (21:10 +0800)]
PCI: loongson: Simplify loongson_pci_probe() return expression

Simplify the return expression.

Link: https://lore.kernel.org/r/20200921131054.92797-1-miaoqinglang@huawei.com
Signed-off-by: Qinglang Miao <miaoqinglang@huawei.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: cadence: Simplify cdns_pcie_host_init_address_translation() return expression
Qinglang Miao [Mon, 21 Sep 2020 13:10:53 +0000 (21:10 +0800)]
PCI: cadence: Simplify cdns_pcie_host_init_address_translation() return expression

Simplify the return expression.

Link: https://lore.kernel.org/r/20200921131053.92752-1-miaoqinglang@huawei.com
Signed-off-by: Qinglang Miao <miaoqinglang@huawei.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: mobiveil: Simplify mobiveil_pcie_init_irq_domain() return expression
Liu Shixin [Mon, 21 Sep 2020 08:24:47 +0000 (16:24 +0800)]
PCI: mobiveil: Simplify mobiveil_pcie_init_irq_domain() return expression

Simplify the return expression by removing useless code.

Link: https://lore.kernel.org/r/20200921082447.2591877-1-liushixin2@huawei.com
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: iproc: Use module_bcma_driver to simplify the code
Liu Shixin [Fri, 18 Sep 2020 03:08:29 +0000 (11:08 +0800)]
PCI: iproc: Use module_bcma_driver to simplify the code

module_bcma_driver() makes the code simpler by eliminating
boilerplate code.

Link: https://lore.kernel.org/r/20200918030829.3946025-1-liushixin2@huawei.com
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Ray Jui <ray.jui@broadcom.com>
3 years agoPCI: brcmstb: Add bcm7211, bcm7216, bcm7445, bcm7278 to match list
Jim Quinlan [Fri, 11 Sep 2020 17:52:30 +0000 (13:52 -0400)]
PCI: brcmstb: Add bcm7211, bcm7216, bcm7445, bcm7278 to match list

Now that the support is in place with previous commits, we add several
chips that use the BrcmSTB driver.

Link: https://lore.kernel.org/r/20200911175232.19016-11-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <james.quinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
3 years agoPCI: brcmstb: Set bus max burst size by chip type
Jim Quinlan [Fri, 11 Sep 2020 17:52:29 +0000 (13:52 -0400)]
PCI: brcmstb: Set bus max burst size by chip type

The proper value of the parameter SCB_MAX_BURST_SIZE varies per chip.  The
2711 family requires 128B whereas other devices can employ 512.  The
assignment is complicated by the fact that the values for this two-bit
field have different meanings;

  Value   Type_Generic    Type_7278

     00       Reserved         128B
     01           128B         256B
     10           256B         512B
     11           512B     Reserved

Link: https://lore.kernel.org/r/20200911175232.19016-10-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI: brcmstb: Accommodate MSI for older chips
Jim Quinlan [Fri, 11 Sep 2020 17:52:28 +0000 (13:52 -0400)]
PCI: brcmstb: Accommodate MSI for older chips

Older BrcmSTB chips do not have a separate register for MSI interrupts; the
MSIs are in a register that also contains unrelated interrupts.  In
addition, the interrupts lie in bits [31..24] for these legacy chips.  This
commit provides common code for both legacy and non-legacy MSI interrupt
registers.

Link: https://lore.kernel.org/r/20200911175232.19016-9-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI: brcmstb: Set additional internal memory DMA viewport sizes
Jim Quinlan [Fri, 11 Sep 2020 17:52:27 +0000 (13:52 -0400)]
PCI: brcmstb: Set additional internal memory DMA viewport sizes

The Raspberry Pi (RPI) is currently the only chip using this driver
(pcie-brcmstb.c).  There, only one memory controller is used, without an
extension region, and the SCB0 viewport size is set to the size of the
first and only dma-range region.  Other BrcmSTB SOCs have more complicated
memory configurations that require setting additional viewport sizes.

BrcmSTB PCIe controllers are intimately connected to the memory
controller(s) on the SOC.  The SOC may have one to three memory
controllers; they are indicated by the term SCBi.  Each controller has a
base region and an optional extension region.  In physical memory, the base
and extension regions of a controller are not adjacent, but in PCIe-space
they are.

There is a "viewport" for each memory controller that allows DMA from
endpoint devices.  Each viewport's size must be set to a power of two, and
that size must be equal to or larger than the amount of memory each
controller supports which is the sum of base region and its optional
extension.  Further, the 1-3 viewports are also adjacent in PCIe-space.

Unfortunately the viewport sizes cannot be ascertained from the
"dma-ranges" property so they have their own property, "brcm,scb-sizes".
This is because dma-range information does not indicate what memory
controller it is associated.  For example, consider the following case
where the size of one dma-range is 2GB and the second dma-range is 1GB:

    /* Case 1: SCB0 size set to 4GB */
    dma-range0: 2GB (from memc0-base)
    dma-range1: 1GB (from memc0-extension)

    /* Case 2: SCB0 size set to 2GB, SCB1 size set to 1GB */
    dma-range0: 2GB (from memc0-base)
    dma-range1: 1GB (from memc0-extension)

By just looking at the dma-ranges information, one cannot tell which
situation applies. That is why an additional property is needed.  Its
length indicates the number of memory controllers being used and each value
indicates the viewport size.

Note that the RPI DT does not have a "brcm,scb-sizes" property value,
as it is assumed that it only requires one memory controller and no
extension.  So the optional use of "brcm,scb-sizes" will be backwards
compatible.

One last layer of complexity exists: all of the viewports sizes must be
added and rounded up to a power of two to determine what the "BAR" size is.
Further, an offset must be given that indicates the base PCIe address of
this "BAR".  The use of the term BAR is typically associated with endpoint
devices, and the term is used here because the PCIe HW may be used as an RC
or an EP.  In the former case, all of the system memory appears in a single
"BAR" region in PCIe memory.  As it turns out, BrcmSTB PCIe HW is rarely
used in the EP role and its system of mapping memory is an artifact that
requires multiple dma-ranges regions.

Link: https://lore.kernel.org/r/20200911175232.19016-8-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <james.quinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI: brcmstb: Add control of rescal reset
Jim Quinlan [Fri, 11 Sep 2020 17:52:26 +0000 (13:52 -0400)]
PCI: brcmstb: Add control of rescal reset

Some STB chips have a special purpose reset controller named RESCAL (reset
calibration).  The PCIe HW can now control RESCAL to start and stop its
operation.  On probe(), the RESCAL is deasserted and the driver goes
through the sequence of setting registers and reading status in order to
start the internal PHY that is required for the PCIe.

Link: https://lore.kernel.org/r/20200911175232.19016-7-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI: hv: Fix hibernation in case interrupts are not re-created
Dexuan Cui [Fri, 2 Oct 2020 08:51:58 +0000 (01:51 -0700)]
PCI: hv: Fix hibernation in case interrupts are not re-created

pci_restore_msi_state() directly writes the MSI/MSI-X related registers
via MMIO. On a physical machine, this works perfectly; for a Linux VM
running on a hypervisor, which typically enables IOMMU interrupt remapping,
the hypervisor usually should trap and emulate the MMIO accesses in order
to re-create the necessary interrupt remapping table entries in the IOMMU,
otherwise the interrupts can not work in the VM after hibernation.

Hyper-V is different from other hypervisors in that it does not trap and
emulate the MMIO accesses, and instead it uses a para-virtualized method,
which requires the VM to call hv_compose_msi_msg() to notify the hypervisor
of the info that would be passed to the hypervisor in the case of the
trap-and-emulate method. This is not an issue to a lot of PCI device
drivers, which destroy and re-create the interrupts across hibernation, so
hv_compose_msi_msg() is called automatically. However, some PCI device
drivers (e.g. the in-tree GPU driver nouveau and the out-of-tree Nvidia
proprietary GPU driver) do not destroy and re-create MSI/MSI-X interrupts
across hibernation, so hv_pci_resume() has to call hv_compose_msi_msg(),
otherwise the PCI device drivers can no longer receive interrupts after
the VM resumes from hibernation.

Hyper-V is also different in that chip->irq_unmask() may fail in a
Linux VM running on Hyper-V (on a physical machine, chip->irq_unmask()
can not fail because unmasking an MSI/MSI-X register just means an MMIO
write): during hibernation, when a CPU is offlined, the kernel tries
to move the interrupt to the remaining CPUs that haven't been offlined
yet. In this case, hv_irq_unmask() -> hv_do_hypercall() always fails
because the vmbus channel has been closed: here the early "return" in
hv_irq_unmask() means the pci_msi_unmask_irq() is not called, i.e. the
desc->masked remains "true", so later after hibernation, the MSI interrupt
always remains masked, which is incorrect. Refer to cpu_disable_common()
-> fixup_irqs() -> irq_migrate_all_off_this_cpu() -> migrate_one_irq():

static bool migrate_one_irq(struct irq_desc *desc)
{
...
        if (maskchip && chip->irq_mask)
                chip->irq_mask(d);
...
        err = irq_do_set_affinity(d, affinity, false);
...
        if (maskchip && chip->irq_unmask)
                chip->irq_unmask(d);

Fix the issue by calling pci_msi_unmask_irq() unconditionally in
hv_irq_unmask(). Also suppress the error message for hibernation because
the hypercall failure during hibernation does not matter (at this time
all the devices have been frozen). Note: the correct affinity info is
still updated into the irqdata data structure in migrate_one_irq() ->
irq_do_set_affinity() -> hv_set_affinity(), so later when the VM
resumes, hv_pci_restore_msi_state() is able to correctly restore
the interrupt with the correct affinity.

Link: https://lore.kernel.org/r/20201002085158.9168-1-decui@microsoft.com
Fixes: ac82fc832708 ("PCI: hv: Add hibernation support")
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Jake Oshins <jakeo@microsoft.com>
3 years agoPCI: Add Kconfig options for MPS/MRRS strategy
Jim Quinlan [Mon, 28 Sep 2020 19:46:51 +0000 (15:46 -0400)]
PCI: Add Kconfig options for MPS/MRRS strategy

Add Kconfig options for changing the default pcie_bus_config, i.e., the
strategy for configuration MPS and MRRS, in the same manner as the
CONFIG_PCIEASPM_XXXX choice.  The pci_bus_config setting may still be
overridden by kernel command-line parameters, e.g.,
"pci=pcie_bus_tune_off".

[bhelgaas: depend on EXPERT, tweak help texts]
Link: https://lore.kernel.org/r/20200928194651.5393-2-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <james.quinlan@broadcom.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/PM: Revert "PCI/PM: Apply D2 delay as milliseconds, not microseconds"
Bjorn Helgaas [Tue, 29 Sep 2020 19:24:11 +0000 (14:24 -0500)]
PCI/PM: Revert "PCI/PM: Apply D2 delay as milliseconds, not microseconds"

This reverts commit 7e24bc347e57992d532bc2ed700209b0fc0a4bf5.

7e24bc347e57 was based on PCIe r5.0, sec 5.9, which claims we need a 200 ms
delay when transitioning to or from D2.  However, sec 5.3.1.3 states the
delay as 200 μs (microseconds), as does the table in PCIe r4.0, sec 5.9.1.

This looks like a typo in the r5.0 spec, so revert back to a 200 μs delay
instead of a 200 ms delay.

Fixes: 7e24bc347e57 ("PCI/PM: Apply D2 delay as milliseconds, not microseconds")
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
3 years agoPCI/PM: Remove unused PCI_PM_BUS_WAIT
Bjorn Helgaas [Tue, 29 Sep 2020 19:20:15 +0000 (14:20 -0500)]
PCI/PM: Remove unused PCI_PM_BUS_WAIT

476e7faefc43 ("PCI PM: Do not wait for buses in B2 or B3 during resume")
removed the last use of PCI_PM_BUS_WAIT.  Remove the definition as well.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
3 years agoPCI/P2PDMA: Drop double zeroing for sg_init_table()
Julia Lawall [Sun, 20 Sep 2020 11:26:26 +0000 (13:26 +0200)]
PCI/P2PDMA: Drop double zeroing for sg_init_table()

sg_init_table() zeroes its first argument, so the allocation of that
argument doesn't have to.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

  // <smpl>
  @@
  expression x;
  @@

  x =
  - kzalloc
  + kmalloc
   (...)
  ...
  sg_init_table(x,...)
  // </smpl>

Link: https://lore.kernel.org/r/1600601186-7420-15-git-send-email-Julia.Lawall@inria.fr
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: Simplify bool comparisons
Krzysztof Wilczyński [Fri, 25 Sep 2020 22:45:55 +0000 (22:45 +0000)]
PCI: Simplify bool comparisons

Take care about Coccinelle warnings:

  drivers/pci/pci.c:6008:6-12: WARNING: Comparison to bool
  drivers/pci/pci.c:6024:7-13: WARNING: Comparison to bool

No change to functionality intended.

Link: https://lore.kernel.org/r/20200925224555.1752460-1-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/PM: Rename pci_dev.d3_delay to d3hot_delay
Krzysztof Wilczyński [Thu, 30 Jul 2020 21:08:48 +0000 (21:08 +0000)]
PCI/PM: Rename pci_dev.d3_delay to d3hot_delay

PCI devices support two variants of the D3 power state: D3hot (main power
present) D3cold (main power removed).  Previously struct pci_dev contained:

  unsigned int    d3_delay;       /* D3->D0 transition time in ms */
  unsigned int    d3cold_delay;   /* D3cold->D0 transition time in ms */

"d3_delay" refers specifically to the D3hot state.  Rename it to
"d3hot_delay" to avoid ambiguity and align with the ACPI "_DSM for
Specifying Device Readiness Durations" in the PCI Firmware spec r3.2,
sec 4.6.9.

There is no change to the functionality.

Link: https://lore.kernel.org/r/20200730210848.1578826-1-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/PM: Remove unused pcibios_pm_ops
Vaibhav Gupta [Thu, 30 Jul 2020 19:44:16 +0000 (01:14 +0530)]
PCI/PM: Remove unused pcibios_pm_ops

The "struct dev_pm_ops pcibios_pm_ops", declared in include/linux/pci.h and
defined in drivers/pci/pci-driver.c, provided arch-specific hooks when a
PCI device was doing a hibernate transition.

394216275c7d ("s390: remove broken hibernate / power management support")
removed the last use of pcibios_pm_ops, so remove it completely.

[bhelgaas: drop unused "error"]
Link: https://lore.kernel.org/r/20200730194416.1029509-1-vaibhavgupta40@gmail.com
Reported-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Vaibhav Gupta <vaibhavgupta40@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: shpchp: Remove unused 'rc' assignment
Krzysztof Wilczyński [Wed, 23 Sep 2020 02:52:25 +0000 (02:52 +0000)]
PCI: shpchp: Remove unused 'rc' assignment

The value of the constant POWER_FAILURE assigned to the variable rc
after the power fault check is never used for anything, so remove it.

Addresses-Coverity-ID: 1226899 ("Unused value")
Link: https://lore.kernel.org/r/20200923025225.471459-1-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: kirin: Return -EPROBE_DEFER in case the gpio isn't ready
Bean Huo [Fri, 18 Sep 2020 12:38:00 +0000 (14:38 +0200)]
PCI: kirin: Return -EPROBE_DEFER in case the gpio isn't ready

PCI host bridge driver can be probed before the gpiochip it requires,
so, of_get_named_gpio() can return -EPROBE_DEFER. Current code lets the
kirin_pcie_probe() directly return -ENODEV, which results in the PCI
host controller driver probe failure; with this error code the PCI host
controller driver will not be probed again when the gpiochip driver is
loaded.

Fix the above issue by letting kirin_pcie_probe() return -EPROBE_DEFER in
such a case.

Link: https://lore.kernel.org/r/20200918123800.19983-1-huobean@gmail.com
Signed-off-by: Bean Huo <beanhuo@micron.com>
[lorenzo.pieralisi@arm.com: commit log]
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
3 years agoPCI: dwc: Fix 'cast truncates bits from constant value'
Gustavo Pimentel [Tue, 22 Sep 2020 09:59:10 +0000 (11:59 +0200)]
PCI: dwc: Fix 'cast truncates bits from constant value'

Fixes warning given by executing "make C=2 drivers/pci/"

Sparse output:
CHECK drivers/pci/controller/dwc/pcie-designware.c
 drivers/pci/controller/dwc/pcie-designware.c:432:52: warning:
 cast truncates bits from constant value (ffffffff7fffffff becomes
 7fffffff)

Link: https://lore.kernel.org/r/7ea7f7d342f97c758949a17b870012f52ce5b3f5.1600767645.git.gustavo.pimentel@synopsys.com
Reported-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Joao Pinto <jpinto@synopsys.com>
3 years agomisc: pci_endpoint_test: Add driver data for Layerscape PCIe controllers
Hou Zhiqiang [Fri, 18 Sep 2020 08:00:24 +0000 (16:00 +0800)]
misc: pci_endpoint_test: Add driver data for Layerscape PCIe controllers

The commit 0a121f9bc3f5 ("misc: pci_endpoint_test: Use streaming DMA
APIs for buffer allocation") changed to use streaming DMA APIs, however,
dma_map_single() might not return a 4KB aligned address, so add the
default_data as driver data for Layerscape PCIe controllers to make it
4KB aligned.

Link: https://lore.kernel.org/r/20200918080024.13639-13-Zhiqiang.Hou@nxp.com
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
3 years agomisc: pci_endpoint_test: Add LS1088a in pci_device_id table
Xiaowei Bao [Fri, 18 Sep 2020 08:00:23 +0000 (16:00 +0800)]
misc: pci_endpoint_test: Add LS1088a in pci_device_id table

Add LS1088a in pci_device_id table so that pci-epf-test can be used
for testing PCIe EP in LS1088a.

Link: https://lore.kernel.org/r/20200918080024.13639-12-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
3 years agoPCI: layerscape: Add EP mode support for ls1088a and ls2088a
Xiaowei Bao [Fri, 18 Sep 2020 08:00:21 +0000 (16:00 +0800)]
PCI: layerscape: Add EP mode support for ls1088a and ls2088a

Add PCIe EP mode support for ls1088a and ls2088a, there are some
difference between LS1 and LS2 platform, so refactor the code of
the EP driver.

Link: https://lore.kernel.org/r/20200918080024.13639-10-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: layerscape: Modify the MSIX to the doorbell mode
Xiaowei Bao [Fri, 18 Sep 2020 08:00:20 +0000 (16:00 +0800)]
PCI: layerscape: Modify the MSIX to the doorbell mode

dw_pcie_ep_raise_msix_irq was never called in the exisitng driver
before, because the ls1046a platform don't support the MSIX feature
and msix_capable was always set to false.
Now that add the ls1088a platform with MSIX support, use the doorbell
method to support the MSIX feature.

Link: https://lore.kernel.org/r/20200918080024.13639-9-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
3 years agoPCI: layerscape: Modify the way of getting capability with different PEX
Xiaowei Bao [Fri, 18 Sep 2020 08:00:19 +0000 (16:00 +0800)]
PCI: layerscape: Modify the way of getting capability with different PEX

The different PCIe controller in one board may be have different
capability of MSI or MSIX, so change the way of getting the MSI
capability, make it more flexible.

Link: https://lore.kernel.org/r/20200918080024.13639-8-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: layerscape: Fix some format issue of the code
Xiaowei Bao [Fri, 18 Sep 2020 08:00:18 +0000 (16:00 +0800)]
PCI: layerscape: Fix some format issue of the code

Fix some format issue of the code in EP driver.

Link: https://lore.kernel.org/r/20200918080024.13639-7-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
3 years agodt-bindings: pci: layerscape-pci: Add compatible strings for ls1088a and ls2088a
Xiaowei Bao [Fri, 18 Sep 2020 08:00:17 +0000 (16:00 +0800)]
dt-bindings: pci: layerscape-pci: Add compatible strings for ls1088a and ls2088a

Add compatible strings for ls1088a and ls2088a.

Link: https://lore.kernel.org/r/20200918080024.13639-6-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
3 years agoPCI: designware-ep: Modify MSI and MSIX CAP way of finding
Xiaowei Bao [Fri, 18 Sep 2020 08:00:16 +0000 (16:00 +0800)]
PCI: designware-ep: Modify MSI and MSIX CAP way of finding

Each PF of EP device should have its own MSI or MSIX capabitily
struct, so create a dw_pcie_ep_func struct and move the msi_cap
and msix_cap to this struct from dw_pcie_ep, and manage the PFs
via a list.

Link: https://lore.kernel.org/r/20200918080024.13639-5-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
3 years agoPCI: designware-ep: Move the function of getting MSI capability forward
Xiaowei Bao [Fri, 18 Sep 2020 08:00:15 +0000 (16:00 +0800)]
PCI: designware-ep: Move the function of getting MSI capability forward

Move the function of getting MSI capability to the front of init
function, because the init function of the EP platform driver will use
the return value by the function of getting MSI capability.

Link: https://lore.kernel.org/r/20200918080024.13639-4-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
3 years agoPCI: designware-ep: Add the doorbell mode of MSI-X in EP mode
Xiaowei Bao [Fri, 18 Sep 2020 08:00:14 +0000 (16:00 +0800)]
PCI: designware-ep: Add the doorbell mode of MSI-X in EP mode

Add the doorbell mode of MSI-X in DWC EP driver.

Link: https://lore.kernel.org/r/20200918080024.13639-3-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
3 years agoPCI: designware-ep: Add multiple PFs support for DWC
Xiaowei Bao [Fri, 18 Sep 2020 08:00:13 +0000 (16:00 +0800)]
PCI: designware-ep: Add multiple PFs support for DWC

Add multiple PFs support for DWC, due to different PF have different
config space, we use func_conf_select callback function to access
the different PF's config space, the different chip company need to
implement this callback function when use the DWC IP core and intend
to support multiple PFs feature.

Link: https://lore.kernel.org/r/20200918080024.13639-2-Zhiqiang.Hou@nxp.com
Signed-off-by: Xiaowei Bao <xiaowei.bao@nxp.com>
Signed-off-by: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
3 years agoARM/PCI: Remove unused fields from struct hw_pci
Lorenzo Pieralisi [Wed, 16 Sep 2020 10:30:45 +0000 (11:30 +0100)]
ARM/PCI: Remove unused fields from struct hw_pci

The msi_ctrl, io_optional and align_resource fields in struct hw_pci are
currently unused by arm/mach PCI host controller drivers and we won't
be adding any new users.

Remove them and related code.

Link: https://lore.kernel.org/r/20200904141607.4066-1-lorenzo.pieralisi@arm.com
Link: https://lore.kernel.org/r/20200916103045.28651-1-lorenzo.pieralisi@arm.com
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: Rob Herring <robh@kernel.org>
Cc: Russell King <linux@armlinux.org.uk>
3 years agoPCI: endpoint: Use "NULL" instead of "0" as a NULL pointer
Gustavo Pimentel [Thu, 17 Sep 2020 21:30:27 +0000 (23:30 +0200)]
PCI: endpoint: Use "NULL" instead of "0" as a NULL pointer

When returning a NULL pointer, use "NULL" instead of "0".

Fixes sparse warning given by executing "make C=2 drivers/pci/":

  CHECK   drivers/pci/endpoint/pci-epc-core.c
   drivers/pci/endpoint/pci-epc-core.c: note: in included file:
   ./include/linux/pci-ep-cfs.h:22:16: warning:
   Using plain integer as NULL pointer
  CHECK   drivers/pci/endpoint/pci-epf-core.c
   drivers/pci/endpoint/pci-epf-core.c: note: in included file:
   ./include/linux/pci-ep-cfs.h:31:16: warning:
   Using plain integer as NULL pointer

Link: https://lore.kernel.org/r/80895f7465719edb3aa259e907acc4bc3217945c.1600378209.git.gustavo.pimentel@synopsys.com
Reported-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Kishon Vijay Abraham I <kishon@ti.com>
Cc: Joao Pinto <jpinto@synopsys.com>
3 years agoPCI: Simplify pci_dev_reset_slot_function()
Lukas Wunner [Tue, 21 Jul 2020 11:24:51 +0000 (13:24 +0200)]
PCI: Simplify pci_dev_reset_slot_function()

pci_dev_reset_slot_function() refuses to reset a hotplug slot if it is
shared by multiple pci_devs.  That's the case if and only if the slot is
occupied by a multifunction device.

Simplify the function to check the device's multifunction flag instead
of iterating over the devices on the bus.  (Iterating over the devices
requires holding pci_bus_sem, which the function erroneously does not
acquire.)

Link: https://lore.kernel.org/r/c6aab5af096f7b1b3db57f6335cebba8f0fcca89.1595330431.git.lukas@wunner.de
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
3 years agoPCI: pciehp: Reduce noisiness on hot removal
Lukas Wunner [Thu, 17 Sep 2020 21:13:20 +0000 (16:13 -0500)]
PCI: pciehp: Reduce noisiness on hot removal

When a PCIe card is hot-removed, the Presence Detect State and Data Link
Layer Link Active bits often do not clear simultaneously.  I've seen delays
of up to 244 msec between the two events with Thunderbolt.

After pciehp has brought down the slot in response to the first event, the
other bit may still be set.  It's not discernible whether it's set because
a new card is already in the slot or if it will soon clear.  So pciehp
tries to bring up the slot and in the latter case fails with a bunch of
messages, some of them at KERN_ERR severity.  If the slot is no longer
occupied, the messages are false positives and annoy users.

Stuart Hayes reports the following splat on hot removal:

  KERN_INFO pcieport 0000:3c:06.0: pciehp: Slot(180): Link Up
  KERN_INFO pcieport 0000:3c:06.0: pciehp: Timeout waiting for Presence Detect
  KERN_ERR  pcieport 0000:3c:06.0: pciehp: link training error: status 0x0001
  KERN_ERR  pcieport 0000:3c:06.0: pciehp: Failed to check link status

Dongdong Liu complains about a similar splat:

  KERN_INFO pciehp 0000:80:10.0:pcie004: Slot(36): Link Down
  KERN_INFO iommu: Removing device 0000:87:00.0 from group 12
  KERN_INFO pciehp 0000:80:10.0:pcie004: Slot(36): Card present
  KERN_INFO pcieport 0000:80:10.0: Data Link Layer Link Active not set in 1000 msec
  KERN_ERR  pciehp 0000:80:10.0:pcie004: Failed to check link status

Users are particularly irritated to see a bringup attempt even though the
slot was explicitly brought down via sysfs.  In a perfect world, we could
avoid this by setting Link Disable on slot bringdown and re-enabling it
upon a Presence Detect State change.  In reality however, there are broken
hotplug ports which hardwire Presence Detect to zero, see 80696f991424
("PCI: pciehp: Tolerate Presence Detect hardwired to zero").  Conversely,
PCIe r1.0 hotplug ports hardwire Link Active to zero because Link Active
Reporting wasn't specified before PCIe r1.1.  On unplug, some ports first
clear Presence then Link (see Stuart Hayes' splat) whereas others use the
inverse order (see Dongdong Liu's splat).  To top it off, there are hotplug
ports which flap the Presence and Link bits on slot bringup, see
6c35a1ac3da6 ("PCI: pciehp: Tolerate initially unstable link").

pciehp is designed to work with all of these variants.  Surplus attempts at
slot bringup are a lesser evil than not being able to bring up slots at
all.  Although we could try to perfect the behavior for specific hotplug
controllers, we'd risk breaking others or increasing code complexity.

But we can certainly minimize annoyance by emitting only a single message
with KERN_INFO severity if bringup is unsuccessful:

* Drop the "Timeout waiting for Presence Detect" message in
  pcie_wait_for_presence().  The sole caller of that function,
  pciehp_check_link_status(), ignores the timeout and carries on.  It emits
  error messages of its own and I don't think this particular message adds
  much value.

* There's a single error condition in pciehp_check_link_status() which
  does not emit a message.  Adding one allows dropping the "Failed to check
  link status" message emitted by board_added() if
  pciehp_check_link_status() returns a non-zero integer.

* Tone down all messages in pciehp_check_link_status() to KERN_INFO
  severity and rephrase them to look as innocuous as possible.  To this
  end, move the message emitted by pcie_wait_for_link_delay() to its
  callers.

As a result, Stuart Hayes' splat becomes:

  KERN_INFO pcieport 0000:3c:06.0: pciehp: Slot(180): Link Up
  KERN_INFO pcieport 0000:3c:06.0: pciehp: Slot(180): Cannot train link: status 0x0001

Dongdong Liu's splat becomes:

  KERN_INFO pciehp 0000:80:10.0:pcie004: Slot(36): Card present
  KERN_INFO pciehp 0000:80:10.0:pcie004: Slot(36): No link

The messages now merely serve as information that presence or link bits
were set a little longer than expected.  Bringup failures which are not
false positives are still reported, albeit no longer at KERN_ERR severity.

Link: https://lore.kernel.org/linux-pci/20200310182100.102987-1-stuart.w.hayes@gmail.com/
Link: https://lore.kernel.org/linux-pci/1547649064-19019-1-git-send-email-liudongdong3@huawei.com/
Link: https://lore.kernel.org/r/b45e46fd8a6aa6930aaac9d7718c2e4b787a4e5e.1595935071.git.lukas@wunner.de
Reported-by: Stuart Hayes <stuart.w.hayes@gmail.com>
Reported-by: Dongdong Liu <liudongdong3@huawei.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
3 years agoPCI: rpadlpar: Use for_each_child_of_node() and for_each_node_by_name()
Qinglang Miao [Wed, 16 Sep 2020 06:21:28 +0000 (14:21 +0800)]
PCI: rpadlpar: Use for_each_child_of_node() and for_each_node_by_name()

Use for_each_child_of_node() and for_each_node_by_name() macros instead of
open coding them.

Link: https://lore.kernel.org/r/20200916062128.190819-1-miaoqinglang@huawei.com
Signed-off-by: Qinglang Miao <miaoqinglang@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: Update mmap-related #ifdef comments
Clint Sbisa [Fri, 21 Aug 2020 15:51:21 +0000 (15:51 +0000)]
PCI: Update mmap-related #ifdef comments

f719582435af ("PCI: Add pci_mmap_resource_range() and use it for ARM64")
changed the #ifdef condition around pci_create_resource_files(),
pci_remove_resource_files(), and related functions, but did not update
comments at the #else and #ifdef.

Update the comments to match the #ifdef.

[bhelgaas: commit log, drop #endif comment since it's close to the #else]
Link: https://lore.kernel.org/r/20200821155121.nzxjeeoze4h5pone@amazon.com
Signed-off-by: Clint Sbisa <csbisa@amazon.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/LINK: Print IRQ number used by port
Dongdong Liu [Thu, 10 Sep 2020 11:24:15 +0000 (19:24 +0800)]
PCI/LINK: Print IRQ number used by port

Print the IRQ used by PCIe Link Bandwidth Notification services port as
AER, PME and DPC do.  It provides convenience to track PCIe BW notification
interrupt counts of certain port from /proc/interrupts.

The dmesg log is as below:

  pcieport 0000:00:00.0: bw_notification: enabled with IRQ 1166

Link: https://lore.kernel.org/r/1599737055-73624-1-git-send-email-liudongdong3@huawei.com
Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI/IOV: Simplify pci-pf-stub with module_pci_driver()
Liu Shixin [Thu, 17 Sep 2020 07:10:42 +0000 (15:10 +0800)]
PCI/IOV: Simplify pci-pf-stub with module_pci_driver()

Use the module_pci_driver() macro to make the code simpler by eliminating
module_init() and module_exit() calls.

Link: https://lore.kernel.org/r/20200917071042.1909191-1-liushixin2@huawei.com
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
3 years agoPCI/ACPI: Add Ampere Altra SOC MCFG quirk
Tuan Phan [Thu, 6 Aug 2020 21:57:34 +0000 (14:57 -0700)]
PCI/ACPI: Add Ampere Altra SOC MCFG quirk

Ampere Altra SOC supports only 32-bit ECAM reads.  Add an MCFG quirk for
the platform.

Link: https://lore.kernel.org/r/1596751055-12316-1-git-send-email-tuanphan@os.amperecomputing.com
Signed-off-by: Tuan Phan <tuanphan@os.amperecomputing.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: Add schedule point in pci_read_config()
Jiang Biao [Mon, 24 Aug 2020 05:20:25 +0000 (13:20 +0800)]
PCI: Add schedule point in pci_read_config()

The PCI sysfs "config" file allows large reads, and the resulting PCI
config reads can take several milliseconds to complete.  Testing with the
cyclictest [1] benchmark showed 5ms+ latencies.

Add a schedule point in pci_read_config() to reduce the maximum latency.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/clrkwllms/rt-tests.git/

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20200824052025.48362-1-benbjiang@tencent.com
Reported-by: Bin Lai <robinlai@tencent.com>
Signed-off-by: Jiang Biao <benbjiang@tencent.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: brcmstb: Add bcm7278 PERST# support
Jim Quinlan [Fri, 11 Sep 2020 17:52:25 +0000 (13:52 -0400)]
PCI: brcmstb: Add bcm7278 PERST# support

The PERST# bit was moved to a different register in 7278-type STB chips.
In addition, the polarity of the bit was also changed; for other chips
writing a 1 specified assert; for 7278-type chips, writing a 0 specifies
assert.  Of course, PERST# is a PCIe asserted-low signal.

While we are here, also change the bridge_sw_init_set() functions so like
the perst_set() functions they are chip specific and we no longer rely on
data wrt chip specific field mask and shift values.

Link: https://lore.kernel.org/r/20200911175232.19016-6-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI: brcmstb: Add suspend and resume pm_ops
Jim Quinlan [Fri, 11 Sep 2020 17:52:24 +0000 (13:52 -0400)]
PCI: brcmstb: Add suspend and resume pm_ops

Broadcom Set-top (BrcmSTB) boards typically support S2, S3, and S5 suspend
and resume.  Now the PCIe driver may do so as well.

Link: https://lore.kernel.org/r/20200911175232.19016-5-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI: brcmstb: Add bcm7278 register info
Jim Quinlan [Fri, 11 Sep 2020 17:52:23 +0000 (13:52 -0400)]
PCI: brcmstb: Add bcm7278 register info

Add in compatibility strings and code for three Broadcom STB chips.  Some
of the register locations, shifts, and masks are different for certain
chips, requiring the use of different constants based on of_id.

We would like to add the following at this time to the match list but we
need to wait until the end of this patchset so that everything works.

    { .compatible = "brcm,bcm7211-pcie", .data = &generic_cfg },
    { .compatible = "brcm,bcm7278-pcie", .data = &bcm7278_cfg },
    { .compatible = "brcm,bcm7216-pcie", .data = &bcm7278_cfg },
    { .compatible = "brcm,bcm7445-pcie", .data = &generic_cfg },

Link: https://lore.kernel.org/r/20200911175232.19016-4-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
3 years agodt-bindings: PCI: Add bindings for more Brcmstb chips
Jim Quinlan [Fri, 11 Sep 2020 17:52:22 +0000 (13:52 -0400)]
dt-bindings: PCI: Add bindings for more Brcmstb chips

- Add compatible strings for three more Broadcom STB chips: 7278, 7216,
  7211 (STB version of RPi4).
- Add new property 'brcm,scb-sizes'.
- Add new property 'resets'.
- Add new property 'reset-names' for 7216 only.
- Allow 'ranges' and 'dma-ranges' to have more than one item and update
  the example to show this.

Link: https://lore.kernel.org/r/20200911175232.19016-3-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agoPCI: brcmstb: PCIE_BRCMSTB depends on ARCH_BRCMSTB
Jim Quinlan [Fri, 11 Sep 2020 17:52:21 +0000 (13:52 -0400)]
PCI: brcmstb: PCIE_BRCMSTB depends on ARCH_BRCMSTB

Have PCIE_BRCMSTB depend on ARCH_BRCMSTB.  Also set the default value to
ARCH_BRCMSTB.

Link: https://lore.kernel.org/r/20200911175232.19016-2-james.quinlan@broadcom.com
Signed-off-by: Jim Quinlan <jquinlan@broadcom.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
3 years agoPCI/ACS: Enable Translation Blocking for external devices
Rajat Jain [Tue, 7 Jul 2020 22:46:04 +0000 (15:46 -0700)]
PCI/ACS: Enable Translation Blocking for external devices

Translation Blocking is a required feature for Downstream Ports (Root
Ports or Switch Downstream Ports) that implement ACS.  When enabled, the
Port checks the Address Type (AT) of each upstream Memory Request it
receives.

The default AT (00b) means "untranslated" and the IOMMU can decide whether
to treat the address as I/O virtual or physical.

If AT is not the default, i.e., if the Memory Request contains an
already-translated (physical) address, the Port blocks the request and
reports an ACS error.

When enabling ACS, enable Translation Blocking for external-facing ports
and untrusted (external) devices.  This is to help prevent attacks from
external devices that initiate DMA with physical addresses that bypass the
IOMMU.

[bhelgaas: commit log, simplify setting bit and drop warning; TB is
required for Downstream Ports with ACS, so we should never see the warning]
Link: https://lore.kernel.org/r/20200707224604.3737893-4-rajatja@google.com
Signed-off-by: Rajat Jain <rajatja@google.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoPCI: hip: Add handling of HiSilicon HIP PCIe controller errors
Yicong Yang [Thu, 3 Sep 2020 12:34:56 +0000 (13:34 +0100)]
PCI: hip: Add handling of HiSilicon HIP PCIe controller errors

The HiSilicon HIP PCIe controller is capable of handling errors
on root port and performing port reset separately at each root port.

Add error handling driver for HIP PCIe controller to log
and report recoverable errors. Perform root port reset and restore
link status after the recovery.

Following are some of the PCIe controller's recoverable errors
1. completion transmission timeout error.
2. CRS retry counter over the threshold error.
3. ECC 2 bit errors
4. AXI bresponse/rresponse errors etc.

The driver placed in the drivers/pci/controller/ because the
HIP PCIe controller does not use DWC IP.

Link: https://lore.kernel.org/r/20200903123456.1823-3-shiju.jose@huawei.com
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
3 years agoACPI / APEI: Add a notifier chain for unknown (vendor) CPER records
Shiju Jose [Thu, 3 Sep 2020 12:34:55 +0000 (13:34 +0100)]
ACPI / APEI: Add a notifier chain for unknown (vendor) CPER records

CPER records describing a firmware-first error are identified by GUID.
The ghes driver currently logs, but ignores any unknown CPER records.
This prevents describing errors that can't be represented by a standard
entry, that would otherwise allow a driver to recover from an error.
The UEFI spec calls these 'Non-standard Section Body' (N.2.3 of
version 2.8).

Add a notifier chain for these non-standard/vendor-records. Callers
must identify their type of records by GUID.

Record data is copied to memory from the ghes_estatus_pool to allow
us to keep it until after the notifier has run.

Co-developed-by: James Morse <james.morse@arm.com>
Link: https://lore.kernel.org/r/20200903123456.1823-2-shiju.jose@huawei.com
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: "Rafael J. Wysocki" <rjw@rjwysocki.net>
3 years agoPCI/ACPI: Tone down missing MCFG message
Jeremy Linton [Tue, 8 Sep 2020 21:03:59 +0000 (16:03 -0500)]
PCI/ACPI: Tone down missing MCFG message

MCFG is an optional ACPI table.  Given there are machines without PCI (or
it is hidden) we have been receiving queries/complaints about what this
message means given it's being presented as an error.

Reduce the message severity.  The ACPI table list printed at boot will
continue to provide another way to detect when the table is missing.

Link: https://lore.kernel.org/r/20200908210359.569294-1-jeremy.linton@arm.com
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
3 years agoPCI: dwc: Use DBI accessors
Rob Herring [Fri, 21 Aug 2020 03:54:20 +0000 (21:54 -0600)]
PCI: dwc: Use DBI accessors

Convert the remaining cases of register accesses using dbi_base rather
than dw_pcie_(read|write)[bwl]_dbi accessors.

Link: https://lore.kernel.org/r/20200821035420.380495-41-robh@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Pratyush Anand <pratyush.anand@gmail.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Jonathan Hunter <jonathanh@nvidia.com>
Cc: linux-tegra@vger.kernel.org
3 years agoPCI: dwc: Move N_FTS setup to common setup
Rob Herring [Fri, 21 Aug 2020 03:54:19 +0000 (21:54 -0600)]
PCI: dwc: Move N_FTS setup to common setup

The Designware controller has common registers to set number of fast
training sequence ordered sets. The Artpec6, Intel, and Tegra driver
initialize these register fields. Let's move the initialization to the
common setup code and drivers just have to provide the value.

There's a slight change in that the common clock mode N_FTS field is
now initialized. Previously only the Intel driver set this. It's not
clear from the code if common clock mode is used in the Artpec6 or Tegra
driver. It depends on the DWC configuration. Given the field is not
initialized while the others are, it seems unlikely common clock mode
is used.

Link: https://lore.kernel.org/r/20200821035420.380495-40-robh@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Jingoo Han <jingoohan1@gmail.com>
Cc: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Jonathan Hunter <jonathanh@nvidia.com>
Cc: linux-tegra@vger.kernel.org