platform/kernel/linux-starfive.git
2 years agoMerge branch 'pci/ctrl/iproc'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:56 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/iproc'

- Use bitmap API to allocate bitmaps instead of doing it by hand
  (Christophe JAILLET)

* pci/ctrl/iproc:
  PCI: iproc: Use bitmap API to allocate bitmaps

2 years agoMerge branch 'pci/ctrl/imx6'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:55 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/imx6'

- Factor out ref clock disables to match enables (Bjorn Helgaas)

- Collect clock enables in imx6_pcie_clk_enable() (Richard Zhu)

- Propagate regulator and clock errors back to .host_init() caller (Richard
  Zhu)

- Disable i.MX6QDL clock when disabling ref clocks (Richard Zhu)

- Call host init function directly in resume instead of duplicating the
  code (Richard Zhu)

- Turn off regulators when suspending (Richard Zhu)

- Make link being down a non-fatal error so probe doesn't fail (Richard
  Zhu)

- Start link in resume only if it was up before suspend to reduce resume
  time (Richard Zhu)

- Move PHY init and power-on out of clock- and reset-related functions
  (Richard Zhu)

- Rework suspend callback to be more symmetric with resume (Richard Zhu)

- Set PCIE_DBI_RO_WR_EN before writing DBI registers (Richard Zhu)

- Allow speeds faster than Gen2 (Richard Zhu)

* pci/ctrl/imx6:
  PCI: imx6: Support more than Gen2 speed link mode
  PCI: imx6: Set PCIE_DBI_RO_WR_EN before writing DBI registers
  PCI: imx6: Reformat suspend callback to keep symmetric with resume
  PCI: imx6: Move the imx6_pcie_ltssm_disable() earlier
  PCI: imx6: Disable clocks in reverse order of enable
  PCI: imx6: Do not hide PHY driver callbacks and refine the error handling
  PCI: imx6: Reduce resume time by only starting link if it was up before suspend
  PCI: imx6: Mark the link down as non-fatal error
  PCI: imx6: Move regulator enable out of imx6_pcie_deassert_core_reset()
  PCI: imx6: Turn off regulator when system is in suspend mode
  PCI: imx6: Call host init function directly in resume
  PCI: imx6: Disable i.MX6QDL clock when disabling ref clocks
  PCI: imx6: Propagate .host_init() errors to caller
  PCI: imx6: Collect clock enables in imx6_pcie_clk_enable()
  PCI: imx6: Factor out ref clock disable to match enable
  PCI: imx6: Move imx6_pcie_clk_disable() earlier
  PCI: imx6: Move imx6_pcie_enable_ref_clk() earlier
  PCI: imx6: Move PHY management functions together
  PCI: imx6: Move imx6_pcie_grp_offset(), imx6_pcie_configure_type() earlier
  PCI: imx6: Convert to NOIRQ_SYSTEM_SLEEP_PM_OPS()

2 years agoMerge branch 'pci/ctrl/fu740'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:55 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/fu740'

- Remove unnecessary include files (Bjorn Helgaas)

* pci/ctrl/fu740:
  PCI: fu740: Remove unnecessary include files

2 years agoMerge branch 'pci/ctrl/exynos'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:55 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/exynos'

- Move samsung phy-exynos-pcie init all to exynos5433_pcie_phy_init()
  instead of splitting across phy_init() and phy_power_on() (Marek
  Szyprowski)

- Call phy_init() before phy_power_on() for samsung phy-exynos-pcie, as
  required by the PHY programming model (Marek Szyprowski)

* pci/ctrl/exynos:
  PCI: exynos: Correct generic PHY usage
  phy: samsung: phy-exynos-pcie: sanitize init/power_on callbacks

2 years agoMerge branch 'pci/ctrl/dwc-edma'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:54 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/dwc-edma'

- Remove unused struct dw_edma_chip.irq (Frank Li)

- Move eDMA private data from struct dw_edma to struct dw_edma_chip (Frank
  Li)

- Convert "struct dw_edma_region rg_region" to "void __iomem *reg_base"
  since only the virtual address (not physical address or size) is used
  (Frank Li)

- Rename "*_ch_cnt" to "ll_*_cnt" to reflect actual usage (Frank Li)

- Drop dma_slave_config.direction field usage (Serge Semin)

- Fix eDMA Rd/Wr-channels and DMA-direction semantics (Serge Semin)

- Add chip-specific DW_EDMA_CHIP_LOCAL flag to indicate that local eDMA
  doesn't require generating MSIs to remote (Frank Li)

- Enable DMA tests for endpoints that support it (Frank Li)

* pci/ctrl/dwc-edma:
  PCI: endpoint: Enable DMA tests for endpoints with DMA capabilities
  dmaengine: dw-edma: Add support for chip-specific flags
  dmaengine: dw-edma: Fix eDMA Rd/Wr-channels and DMA-direction semantics
  dmaengine: dw-edma: Drop dma_slave_config.direction field usage
  dmaengine: dw-edma: Rename wr(rd)_ch_cnt to ll_wr(rd)_cnt in struct dw_edma_chip
  dmaengine: dw-edma: Change rg_region to reg_base in struct dw_edma_chip
  dmaengine: dw-edma: Detach the private data and chip info structures
  dmaengine: dw-edma: Remove unused irq field in struct dw_edma_chip

2 years agoMerge branch 'pci/ctrl/dwc'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:54 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/dwc'

- Stop link on host_init errors and de-initialization (Serge Semin)

- Add support for unrolled iATU register space in dw_pcie_disable_atu()
  (Serge Semin)

- Disable outbound windows only for controllers that use iATU (Serge Semin)

- Set INCREASE_REGION_SIZE flag based on limit address, not on the size,
  since even a small size may cross a 4GB boundary (Serge Semin)

- Deallocate EPC memory on dw_pcie_ep_init() errors to avoid a leak (Serge
  Semin)

- Always enable CDM check if "snps,enable-cdm-check" exists instead of
  exiting early if the optional "num-lanes" was absent (Serge Semin)

- Simplify detection of whether we're using unrolled iATU registers (Serge
  Semin)

- Make dw_pcie_link_up() more generic by using dw_pcie_readl_dbi() instead
  of readl() (Serge Semin)

- Add dw_pcie_start_link() and dw_pcie_stop_link() wrappers to factor out
  checks for ops being implemented (Serge Semin)

- Move io_cfg_atu_shared to struct pcie_port and rename to cfg0_io_shared,
  since it's not used by dwc common code or dwc endpoint code (Serge Semin)

- Rename struct pcie_port to dw_pcie_rp to indicate that it's
  DesignWare-specific (Serge Semin)

- Drop unused struct dw_plat_pcie regmap pointer (Serge Semin)

- Fix some coding style issues (Serge Semin)

- Log link speed and width if it comes up (Serge Semin)

- Save DWC IP core version in native format as read from
  PORT_LOGIC.PCIE_VERSION_OFF register (Serge Semin)

- Read DWC IP core version from PORT_LOGIC.PCIE_VERSION_OFF (Serge Semin)

- Add macros to compare Synopsys IP core versions (Serge Semin)

- Drop manual DWC IP core version setup from intel-gw and tegra194 (Serge
  Semin)

- Add dw_pcie_ops.host_deinit() callback (Serge Semin)

- Drop enum dw_pcie_as_type in favor of PCIE_ATU_TYPE_MEM/IO (Serge Semin)

- Drop enum dw_pcie_region_type in favor of PCIE_ATU_REGION_DIR_IB/OB
  (Serge Semin)

- Simplify in/outbound iATU setup methods and reduce duplicated code (Serge
  Semin)

- Detect iATU region size from hardware (Serge Semin)

- Validate iATU outbound mappings against hardware constraints (Serge
  Semin)

- Check for errors in iATU setup (Serge Semin)

- Allocate a 32-bit DMA-able page to be MSI target instead of using a
  driver data structure that may not be addressable with 32-bit address
  (Will McVicker)

- Use the bitmap API to allocate bitmaps instead of open-coding it
  (Christophe JAILLET)

- Correct dw_pcie_free_msi() checking for when to remove IRQ handler and
  data (Dmitry Baryshkov)

- Split MSI init to new dw_pcie_msi_host_init() function (Dmitry Baryshkov)

- Convert struct pcie_port.msi_irq to an array so we can support more than
  32 MSI interrupts (Dmitry Baryshkov)

- Handle MSIs routed to multiple GIC interrupts for Qualcomm platforms with
  groups of 32 MSI vectors (Dmitry Baryshkov)

- Add additional MSI interrupts to qcom DT (Dmitry Baryshkov)

* pci/ctrl/dwc:
  dt-bindings: PCI: qcom: Support additional MSI vectors
  PCI: dwc: Handle MSIs routed to multiple GIC interrupts
  PCI: dwc: Convert struct pcie_port.msi_irq to an array
  PCI: dwc: Split MSI IRQ parsing/allocation to a separate function
  PCI: dwc: Correct msi_irq condition in dw_pcie_free_msi()
  PCI: dwc: Use the bitmap API to allocate bitmaps
  PCI: dwc: Fix MSI msi_msg DMA mapping
  PCI: dwc: Check iATU in/outbound range setup status
  PCI: dwc: Validate iATU outbound mappings against hardware constraints
  PCI: dwc: Add iATU regions size detection procedure
  PCI: dwc: Simplify in/outbound iATU setup methods
  PCI: dwc: Drop enum dw_pcie_region_type in favor of PCIE_ATU_REGION_DIR_IB/OB
  PCI: dwc: Drop enum dw_pcie_as_type in favor of PCIE_ATU_TYPE_MEM/IO
  PCI: dwc: Add dw_pcie_ops.host_deinit() callback
  PCI: tegra194: Drop manual DW PCIe controller version setup
  PCI: intel-gw: Drop manual DW PCIe controller version setup
  PCI: dwc: Add macros to compare Synopsys IP core versions
  PCI: dwc: Read DWC IP core version from register
  PCI: dwc: Use native DWC IP core version representation
  PCI: dwc: Detect iATU settings after getting "addr_space" resource
  PCI: dwc: Log link speed and width if it comes up
  PCI: dwc-plat: Drop dw_plat_pcie_of_match[] forward declaration
  PCI: dwc-plat: Drop unused regmap pointer
  PCI: dwc-plat: Simplify dw_plat_pcie_probe() return values
  PCI: dwc: Rename struct pcie_port to dw_pcie_rp
  PCI: dwc: Move io_cfg_atu_shared to struct pcie_port
  PCI: dwc: Add start_link/stop_link inlines
  PCI: dwc: Reuse local pointer to the resource data
  PCI: dwc: Organize local variable usage
  PCI: dwc: Convert dw_pcie_link_up() to use dw_pcie_readl_dbi()
  PCI: dwc: Simplify unrolled iATU detection
  PCI: dwc: Add newlines to log messages
  PCI: dwc: Add braces to multi-line if-else statements
  PCI: dwc: Always enable CDM check if "snps,enable-cdm-check" exists
  PCI: dwc: Deallocate EPC memory on dw_pcie_ep_init() errors
  PCI: dwc: Set INCREASE_REGION_SIZE flag based on limit address
  PCI: dwc: Disable outbound windows only for controllers using iATU
  PCI: dwc: Add unroll iATU space support to dw_pcie_disable_atu()
  PCI: dwc: Stop link on host_init errors and de-initialization

2 years agoMerge branch 'pci/ctrl/brcmstb'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:53 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/brcmstb'

- Remove unnecessary forward declarations (Jim Quinlan)

- Prevent config space access when link is down (Jim Quinlan)

- Split post-link up initialization to brcm_pcie_start_link() (Jim Quinlan)

- Enable child bus device regulators described under Root Ports in DT (Jim
  Quinlan)

- Disable/enable regulators in suspend/resume (Jim Quinlan)

- Rename .map_bus() functions to end with 'map_bus' as they do in other
  drivers (Jim Quinlan)

* pci/ctrl/brcmstb:
  PCI: brcmstb: Rename .map_bus() functions to end with 'map_bus'
  PCI: brcmstb: Disable/enable regulators in suspend/resume
  PCI: brcmstb: Enable child bus device regulators from DT
  PCI: brcmstb: Split post-link up initialization to brcm_pcie_start_link()
  PCI: brcmstb: Prevent config space access when link is down
  PCI: brcmstb: Remove unnecessary forward declarations

2 years agoMerge branch 'pci/ctrl/aardvark'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:53 +0000 (11:41 -0500)]
Merge branch 'pci/ctrl/aardvark'

- Add support for AER capability on emulated bridge (Pali Rohár)

- Add support for Slot capability on emulated bridge (Pali Rohár)

* pci/ctrl/aardvark:
  PCI: aardvark: Fix reporting Slot capabilities on emulated bridge
  PCI: aardvark: Add support for AER registers on emulated bridge

2 years agoMerge branch 'pci/virtualization'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:53 +0000 (11:41 -0500)]
Merge branch 'pci/virtualization'

- Add ACS quirk for Broadcom BCM5750x multifunction NICs that isolate the
  functions but don't advertise an ACS capability (Pavan Chebbi)

* pci/virtualization:
  PCI: Add ACS quirk for Broadcom BCM5750x NICs

2 years agoMerge branch 'pci/resource'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:52 +0000 (11:41 -0500)]
Merge branch 'pci/resource'

- Replace sparc pci_mmap_page_range() wrapper.  This still leaves a
  sparc-specific pci_mmap_resource_range(), but it's only one interface
  instead of two (Arnd Bergmann)

- Remove sparc-specific pci_mmap_resource_range() by implementing
  pci_iobar_pfn().  This removes the ability to map the entire PCI I/O
  space using /proc/bus/pci, but we believe that's already been broken
  since v2.6.28 (Arnd Bergmann)

* pci/resource:
  sparc: Use generic pci_mmap_resource_range()
  PCI: Remove pci_mmap_page_range() wrapper

2 years agoMerge branch 'pci/pm'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:52 +0000 (11:41 -0500)]
Merge branch 'pci/pm'

- Convert documentation examples to generic power management (Bjorn
  Helgaas)

* pci/pm:
  PCI/doc: Convert examples to generic power management

2 years agoMerge branch 'pci/err'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:52 +0000 (11:41 -0500)]
Merge branch 'pci/err'

- Recognize disconnected devices so we don't bother trying to set them to
  "frozen" or "normal" state (Christoph Hellwig)

- Clear PCI Status register during enumeration in case firmware left errors
  logged (Kai-Heng Feng)

- Configure ECRC for every device, including hot-added ones (Stefan Roese)

- Keep AER error reporting enabled for switches (Stefan Roese)

- Enable error reporting for all devices that support AER (Stefan Roese)

- Iterate over error counters instead of error strings to avoid printing
  junk in AER sysfs counters (Mohamed Khalfella)

* pci/err:
  PCI/AER: Iterate over error counters instead of error strings
  PCI/AER: Enable error reporting when AER is native
  PCI/portdrv: Don't disable AER reporting in get_port_device_capability()
  PCI/AER: Configure ECRC for every device
  PCI: Clear PCI_STATUS when setting up device
  PCI/ERR: Recognize disconnected devices in report_error_detected()

2 years agoMerge branch 'pci/enumeration'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:51 +0000 (11:41 -0500)]
Merge branch 'pci/enumeration'

- Split out ARI "next function" handling from the traditional one (Niklas
  Schnelle)

- Move jailhouse "isolated function" (non-zero functions where function 0
  doesn't exist) handling to pci_scan_slot() to avoid duplicating
  multi-function scanning in pci_scan_child_bus_extend() (Niklas Schnelle)

- Extend "isolated function" probing to s390 (Niklas Schnelle).

- Allow s390 zPCI zbus without a function 0 (Niklas Schnelle)

* pci/enumeration:
  s390/pci: allow zPCI zbus without a function zero
  PCI: Extend isolated function probing to s390
  PCI: Move jailhouse's isolated function handling to pci_scan_slot()
  PCI: Split out next_ari_fn() from next_fn()
  PCI: Clean up pci_scan_slot()

2 years agoMerge branch 'pci/endpoint'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:51 +0000 (11:41 -0500)]
Merge branch 'pci/endpoint'

- Don't stop an EPC when unbinding an EPF from it (Shunsuke Mie)

* pci/endpoint:
  PCI: endpoint: Don't stop controller when unbinding endpoint function

2 years agoMerge branch 'pci/aspm'
Bjorn Helgaas [Thu, 4 Aug 2022 16:41:50 +0000 (11:41 -0500)]
Merge branch 'pci/aspm'

- Remove pcie_aspm_pm_state_change(), which overwrites ASPM config on power
  state changes (Kai-Heng Feng)

* pci/aspm:
  PCI/ASPM: Unexport pcie_aspm_support_enabled()
  PCI/ASPM: Remove pcie_aspm_pm_state_change()

2 years agoPCI: imx6: Support more than Gen2 speed link mode
Richard Zhu [Wed, 20 Jul 2022 03:26:16 +0000 (11:26 +0800)]
PCI: imx6: Support more than Gen2 speed link mode

Support more than Gen2 speed link mode, since i.MX8MP PCIe supports up to
Gen3 link speed.

Link: https://lore.kernel.org/r/1658287576-26908-1-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Set PCIE_DBI_RO_WR_EN before writing DBI registers
Richard Zhu [Wed, 18 May 2022 09:35:27 +0000 (17:35 +0800)]
PCI: imx6: Set PCIE_DBI_RO_WR_EN before writing DBI registers

The PCIE_DBI_RO_WR_EN bit should be set when write some DBI registers.  To
make sure that the DBI registers are writable, set the PCIE_DBI_RO_WR_EN
properly when writing the DBI registers.

Link: https://lore.kernel.org/r/1652866528-13220-1-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Reformat suspend callback to keep symmetric with resume
Richard Zhu [Thu, 14 Jul 2022 07:31:09 +0000 (15:31 +0800)]
PCI: imx6: Reformat suspend callback to keep symmetric with resume

Create imx6_pcie_stop_link() and imx6_pcie_host_exit() functions.
Encapsulate clocks, regulators disables and PHY uninitialization into
imx6_pcie_host_exit().

To keep suspend/resume symmetric as much as possible, invoke these two
new created functions in suspend callback.

To be symmetric with imx6_pcie_host_exit(), move imx6_pcie_clk_enable()
to imx6_pcie_host_init() from imx6_pcie_deassert_core_reset().

Link: https://lore.kernel.org/r/1657783869-19194-18-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Move the imx6_pcie_ltssm_disable() earlier
Richard Zhu [Thu, 14 Jul 2022 07:31:08 +0000 (15:31 +0800)]
PCI: imx6: Move the imx6_pcie_ltssm_disable() earlier

Move the imx6_pcie_ltssm_disable() earlier and place it just behind the
imx6_pcie_ltssm_enable(), since it might not be only used by suspend
callback directly.

To be symmetric with imx6_pcie_ltssm_enable(), add the IMX6Q and IMX8MQ
switch cases in imx6_pcie_ltssm_disable().

Link: https://lore.kernel.org/r/1657783869-19194-17-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Disable clocks in reverse order of enable
Bjorn Helgaas [Thu, 14 Jul 2022 07:31:07 +0000 (15:31 +0800)]
PCI: imx6: Disable clocks in reverse order of enable

imx6_pcie_clk_enable() enables clocks in the order:

  pcie_phy
  pcie_bus
  pcie
  imx6_pcie_enable_ref_clk

Change imx6_pcie_clk_disable() to disable them in the reverse order.

Link: https://lore.kernel.org/r/1657783869-19194-16-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Richard Zhu <hongxing.zhu@nxp.com>
2 years agoPCI: imx6: Do not hide PHY driver callbacks and refine the error handling
Richard Zhu [Thu, 14 Jul 2022 07:31:06 +0000 (15:31 +0800)]
PCI: imx6: Do not hide PHY driver callbacks and refine the error handling

Move the phy_power_on() to host_init from imx6_pcie_clk_enable().

Move the phy_init() to host_init from imx6_pcie_deassert_core_reset().

Refine the error handling in imx6_pcie_host_init() accordingly.

Link: https://lore.kernel.org/r/1657783869-19194-15-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Reduce resume time by only starting link if it was up before suspend
Richard Zhu [Thu, 14 Jul 2022 07:31:05 +0000 (15:31 +0800)]
PCI: imx6: Reduce resume time by only starting link if it was up before suspend

i.MX PCIe doesn't support hotplug. During resume, only start PCIe link
training when the link was up before system suspend to avoid the long
latency in the link training period.

Link: https://lore.kernel.org/r/1657783869-19194-14-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Mark the link down as non-fatal error
Richard Zhu [Thu, 14 Jul 2022 07:31:04 +0000 (15:31 +0800)]
PCI: imx6: Mark the link down as non-fatal error

If the PCIe link is down, return zero from imx6_pcie_start_link() so the
driver will probe successfully.

Link: https://lore.kernel.org/r/1657783869-19194-13-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Move regulator enable out of imx6_pcie_deassert_core_reset()
Richard Zhu [Thu, 14 Jul 2022 07:31:03 +0000 (15:31 +0800)]
PCI: imx6: Move regulator enable out of imx6_pcie_deassert_core_reset()

Move regulator enable out of imx6_pcie_deassert_core_reset(), since the
regulator_enable() has nothing to do with imx6_pcie_deassert_core_reset().

Link: https://lore.kernel.org/r/1657783869-19194-12-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Turn off regulator when system is in suspend mode
Richard Zhu [Thu, 14 Jul 2022 07:31:02 +0000 (15:31 +0800)]
PCI: imx6: Turn off regulator when system is in suspend mode

The driver should undo any enables it did itself. The regulator disable
shouldn't be basing decisions on regulator_is_enabled().

Move the regulator_disable to the suspend function, turn off regulator when
the system is in suspend mode.

Link: https://lore.kernel.org/r/1657783869-19194-11-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: imx6: Call host init function directly in resume
Richard Zhu [Thu, 14 Jul 2022 07:31:01 +0000 (15:31 +0800)]
PCI: imx6: Call host init function directly in resume

Call imx6_pcie_host_init() instead of duplicating codes in resume.  Note
that this also means we do MPLL setup again during resume, which we didn't
do before.

[bhelgaas: add MPLL setup note, pointed out by Lucas]
Link: https://lore.kernel.org/r/1657783869-19194-10-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Disable i.MX6QDL clock when disabling ref clocks
Richard Zhu [Thu, 14 Jul 2022 07:31:00 +0000 (15:31 +0800)]
PCI: imx6: Disable i.MX6QDL clock when disabling ref clocks

When disabling PCIe clocks, disable i.MX6QDL ref clock too.

Link: https://lore.kernel.org/r/1657783869-19194-9-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Propagate .host_init() errors to caller
Richard Zhu [Thu, 14 Jul 2022 07:30:59 +0000 (15:30 +0800)]
PCI: imx6: Propagate .host_init() errors to caller

Since dw_pcie_host_init() checks for errors from ops->host_init(),
check for errors when enabling power regulators and clocks and return them.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/1657783869-19194-8-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Collect clock enables in imx6_pcie_clk_enable()
Richard Zhu [Thu, 14 Jul 2022 07:30:58 +0000 (15:30 +0800)]
PCI: imx6: Collect clock enables in imx6_pcie_clk_enable()

Encapsulate the i.MX PCIe clock enable operations into one standalone
function, imx6_pcie_clk_enable().  No functional change intended.

[bhelgaas: split pure code moves into separate patches]
Link: https://lore.kernel.org/r/1657783869-19194-7-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Factor out ref clock disable to match enable
Bjorn Helgaas [Thu, 14 Jul 2022 07:30:57 +0000 (15:30 +0800)]
PCI: imx6: Factor out ref clock disable to match enable

The PCIe ref clocks are specific to different variants.  The enables are
already split out into imx6_pcie_enable_ref_clk(), but the disables were
combined with the more generic bus/phy/pcie clock disables in
imx6_pcie_clk_disable().

Split out the variant-specific disables into imx6_pcie_disable_ref_clk() to
match imx6_pcie_enable_ref_clk().

No functional change intended.

Link: https://lore.kernel.org/r/1657783869-19194-6-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Richard Zhu <hongxing.zhu@nxp.com>
2 years agoPCI: imx6: Move imx6_pcie_clk_disable() earlier
Richard Zhu [Thu, 14 Jul 2022 07:30:56 +0000 (15:30 +0800)]
PCI: imx6: Move imx6_pcie_clk_disable() earlier

Move imx6_pcie_clk_disable() earlier to be near other clock-related
functions.  No functional change intended.

[bhelgaas: reorder patch so pure moves are earlier]
Link: https://lore.kernel.org/r/1657783869-19194-5-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
2 years agoPCI: imx6: Move imx6_pcie_enable_ref_clk() earlier
Bjorn Helgaas [Thu, 14 Jul 2022 07:30:55 +0000 (15:30 +0800)]
PCI: imx6: Move imx6_pcie_enable_ref_clk() earlier

Move imx6_pcie_enable_ref_clk() earlier so it's not in the middle between
imx6_pcie_assert_core_reset() and imx6_pcie_deassert_core_reset().  No
functional change intended.

Link: https://lore.kernel.org/r/1657783869-19194-4-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Richard Zhu <hongxing.zhu@nxp.com>
2 years agoPCI: imx6: Move PHY management functions together
Bjorn Helgaas [Thu, 14 Jul 2022 07:30:54 +0000 (15:30 +0800)]
PCI: imx6: Move PHY management functions together

Collect imx6_pcie_init_phy(), imx7d_pcie_wait_for_phy_pll_lock(), and
imx6_setup_phy_mpll() earlier with other PHY-related code.  No functional
change intended.

Link: https://lore.kernel.org/r/1657783869-19194-3-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Richard Zhu <hongxing.zhu@nxp.com>
2 years agoPCI: imx6: Move imx6_pcie_grp_offset(), imx6_pcie_configure_type() earlier
Bjorn Helgaas [Thu, 14 Jul 2022 07:30:53 +0000 (15:30 +0800)]
PCI: imx6: Move imx6_pcie_grp_offset(), imx6_pcie_configure_type() earlier

Move imx6_pcie_grp_offset() and imx6_pcie_configure_type() earlier in the
file since they depend on nothing and are used by several other functions
that will be moved earlier.  No functional change intended.

Link: https://lore.kernel.org/r/1657783869-19194-2-git-send-email-hongxing.zhu@nxp.com
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Richard Zhu <hongxing.zhu@nxp.com>
2 years agoPCI: imx6: Convert to NOIRQ_SYSTEM_SLEEP_PM_OPS()
Bjorn Helgaas [Tue, 19 Jul 2022 20:50:18 +0000 (15:50 -0500)]
PCI: imx6: Convert to NOIRQ_SYSTEM_SLEEP_PM_OPS()

Replace SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() with NOIRQ_SYSTEM_SLEEP_PM_OPS(),
which has the advantage that the compiler always sees the PM callbacks as
referenced, so they don't need to be wrapped with "#ifdef CONFIG_PM_SLEEP"
or tagged with "__maybe_unused" to avoid "defined but not used" warnings.

See 1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate old ones").

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
2 years agoPCI: exynos: Correct generic PHY usage
Marek Szyprowski [Tue, 28 Jun 2022 22:04:09 +0000 (00:04 +0200)]
PCI: exynos: Correct generic PHY usage

The proper initialization for generic PHYs is to call first phy_init(),
then phy_power_on().

While touching this, remove the phy_reset() call. It is just a left-over
from the obsoleted Exynos5440 support and the current exynos-pcie PHY
driver doesn't even support this function. It is also rarely used by other
drivers.

Link: https://lore.kernel.org/r/20220628220409.26545-2-m.szyprowski@samsung.com
Reported-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Chanho Park <chanho61.park@samsung.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
2 years agophy: samsung: phy-exynos-pcie: sanitize init/power_on callbacks
Marek Szyprowski [Tue, 28 Jun 2022 22:04:08 +0000 (00:04 +0200)]
phy: samsung: phy-exynos-pcie: sanitize init/power_on callbacks

The exynos-pcie driver called phy_power_on() before phy_init() for some
historical reasons. However the generic PHY framework assumes that the
proper sequence is to call phy_init() first, then phy_power_on(). The
operations done by both functions should be considered as one action and as
such they are called by the exynos-pcie driver (without doing anything
between them). The initialization is just a sequence of register writes,
which cannot be altered without breaking the hardware operation.

To match the generic PHY framework requirement, simply move all register
writes to the phy_init()/phy_exit() and drop power_on()/power_off()
callbacks. This way the driver will also work with the old (incorrect)
PHY initialization call sequence.

Link: https://lore.kernel.org/r/20220628220409.26545-1-m.szyprowski@samsung.com
Reported-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Chanho Park <chanho61.park@samsung.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Acked-By: Vinod Koul <vkoul@kernel.org>
2 years agodt-bindings: PCI: qcom: Support additional MSI vectors
Dmitry Baryshkov [Thu, 7 Jul 2022 13:47:32 +0000 (16:47 +0300)]
dt-bindings: PCI: qcom: Support additional MSI vectors

On Qualcomm platforms each group of 32 MSI vectors is routed to the
separate GIC interrupt. Document mapping of additional interrupts.

Link: https://lore.kernel.org/r/20220707134733.2436629-6-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Stanimir Varbanov <svarbanov@mm-sol.com>
2 years agoPCI: dwc: Handle MSIs routed to multiple GIC interrupts
Dmitry Baryshkov [Thu, 7 Jul 2022 13:47:31 +0000 (16:47 +0300)]
PCI: dwc: Handle MSIs routed to multiple GIC interrupts

On some Qualcomm platforms each group of 32 MSI vectors is routed to a
separate GIC interrupt. Implement support for such configurations by
parsing "msi0" ... "msiX" interrupts and attaching them to the chained
handler.

Note that if DT doesn't list an array of MSI interrupts and uses a single
"msi" IRQ, the driver will limit the number of supported MSI vectors to 32.

Link: https://lore.kernel.org/r/20220707134733.2436629-5-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Convert struct pcie_port.msi_irq to an array
Dmitry Baryshkov [Fri, 22 Jul 2022 16:34:12 +0000 (11:34 -0500)]
PCI: dwc: Convert struct pcie_port.msi_irq to an array

The Qualcomm DWC PCIe controller supports more than 32 MSI interrupts, but
they are routed to separate interrupts in groups of 32 vectors. To support
this configuration, change the msi_irq field to an array. Let the DWC core
handle all interrupts that were set in this array.

[bhelgaas: reorder, drop "irq" temporary to make patch cleaner]
Link: https://lore.kernel.org/r/20220707134733.2436629-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Split MSI IRQ parsing/allocation to a separate function
Dmitry Baryshkov [Fri, 22 Jul 2022 16:28:28 +0000 (11:28 -0500)]
PCI: dwc: Split MSI IRQ parsing/allocation to a separate function

Split handling of MSI host IRQs to a separate dw_pcie_msi_host_init()
function. The code is complex enough to warrant a separate function.

[bhelgaas: reorder patch earlier]
Link: https://lore.kernel.org/r/20220707134733.2436629-4-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Correct msi_irq condition in dw_pcie_free_msi()
Dmitry Baryshkov [Thu, 7 Jul 2022 13:47:28 +0000 (16:47 +0300)]
PCI: dwc: Correct msi_irq condition in dw_pcie_free_msi()

The dwc-based drivers set pp->msi_irq to -ENODEV if they do not want the
dwc core to do anything with pp->msi_irq.

dw_pcie_host_init() sets the handler and data when "pp->msi_irq > 0", so
use the same condition when removing the handler and data in
dw_pcie_free_msi().

Link: https://lore.kernel.org/r/20220707134733.2436629-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Use the bitmap API to allocate bitmaps
Christophe JAILLET [Sat, 9 Jul 2022 14:10:52 +0000 (16:10 +0200)]
PCI: dwc: Use the bitmap API to allocate bitmaps

Use devm_bitmap_zalloc() instead of hand-writing them.

It is less verbose and it improves the semantic.

Link: https://lore.kernel.org/r/bc6586a603abc0db7d4531308b698fbe7a6d7083.1657375829.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: dwc: Fix MSI msi_msg DMA mapping
Will McVicker [Mon, 13 Jun 2022 22:26:11 +0000 (17:26 -0500)]
PCI: dwc: Fix MSI msi_msg DMA mapping

As of 07940c369a6b ("PCI: dwc: Fix MSI page leakage in suspend/resume"),
the PCIe designware host driver has been using the driver data allocation
for the msi_msg DMA mapping which can result in a DMA_MAPPING_ERROR due to
the DMA overflow check in dma_direct_map_page() when the address is greater
than 32 bits (reported in [1]). The commit was trying to address a memory
leak on suspend/resume by moving the MSI mapping to dw_pcie_host_init(),
but subsequently dropped the page allocation thinking it wasn't needed.

To fix the DMA mapping issue as well as make msi_msg DMA'able, switch back
to allocating a 32-bit page for the msi_msg. To avoid the suspend/resume
leak, allocate the page in dw_pcie_host_init() since that shouldn't be
called during suspend/resume.

[1] https://lore.kernel.org/all/Yo0soniFborDl7+C@google.com/

Signed-off-by: Will McVicker <willmcvicker@google.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Check iATU in/outbound range setup status
Serge Semin [Fri, 24 Jun 2022 14:39:46 +0000 (17:39 +0300)]
PCI: dwc: Check iATU in/outbound range setup status

Make the DWC PCIe RC/EP safer and more verbose for invalid or failed
inbound and outbound iATU window setups.  Silently ignoring iATU regions
setup errors may cause unpredictable errors. For instance if a cfg or IO
window fails to be activated, then any CFG/IO requested won't reach target
PCIe devices and the corresponding accessors will return platform-specific
random values.

[bhelgaas: trim commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-15-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Validate iATU outbound mappings against hardware constraints
Serge Semin [Fri, 24 Jun 2022 14:39:45 +0000 (17:39 +0300)]
PCI: dwc: Validate iATU outbound mappings against hardware constraints

Make __dw_pcie_prog_outbound_atu() check the requested region base and size
against what the hardware can support.  Return error if the region is not
correctly aligned or of a supported size.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-14-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Add iATU regions size detection procedure
Serge Semin [Fri, 24 Jun 2022 14:39:44 +0000 (17:39 +0300)]
PCI: dwc: Add iATU regions size detection procedure

The DWC PCIe RC/EP/DM IP core configuration parameters determine the number
of inbound and outbound iATU windows, alignment requirements (which is also
the minimum window size), minimum and maximum sizes.  If internal ATU is
enabled, the former settings are determined by CX_ATU_MIN_REGION_SIZE; the
latter are determined by CX_ATU_MAX_REGION_SIZE.

Determine the required alignment and maximum size supported by the
controller and log it to help verify whether the requested inbound or
outbound memory mappings can be fully created.

Note 1. The extended iATU regions have been supported since DWC PCIe
v4.60a. There is no need in testing the upper limit register availability
for the older cores.

Note 2. The regions alignment is determined with using the fls() method
since the lower four bits of the ATU Limit register can be occupied with
the Circular Buffer Increment setting, which can be initialized with zeros.

Link: https://lore.kernel.org/r/20220624143947.8991-13-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Simplify in/outbound iATU setup methods
Serge Semin [Fri, 24 Jun 2022 14:39:43 +0000 (17:39 +0300)]
PCI: dwc: Simplify in/outbound iATU setup methods

Previously __dw_pcie_prog_outbound_atu() duplicated a lot of code between
the iatu_unroll_enabled version and the PCIE_ATU_VIEWPORT version:

  __dw_pcie_prog_outbound_atu
    if (iatu_unroll_enabled)
      dw_pcie_prog_outbound_atu_unroll
        dw_pcie_writel_ob_unroll(PCIE_ATU_UNR_LOWER_BASE, ...)
        dw_pcie_writel_ob_unroll(PCIE_ATU_UNR_UPPER_BASE, ...)
        ...
      return
    dw_pcie_writel_dbi(PCIE_ATU_VIEWPORT, ...)
    dw_pcie_writel_dbi(PCIE_ATU_LOWER_BASE, ...)
    dw_pcie_writel_dbi(PCIE_ATU_UPPER_BASE, ...)
    ...

Unify those by pushing the unroll address computation and viewport
selection down into dw_pcie_writel_atu() so we can use the same
dw_pcie_writel_atu_ob() accessor for both paths:

  __dw_pcie_prog_outbound_atu
    dw_pcie_writel_atu_ob(PCIE_ATU_LOWER_BASE, ...)
      dw_pcie_writel_atu
        dw_pcie_select_atu                      # new
          if (iatu_unroll_enabled)
            return pci->atu_base + PCIE_ATU_UNROLL_BASE(...)
          dw_pcie_writel_dbi(PCIE_ATU_VIEWPORT, ...)
          return pci->atu_base
        dw_pcie_write(base + reg)
      dw_pcie_writel_atu_ob(PCIE_ATU_UPPER_BASE, ...)
      ...

In the non-unroll case, this does involve more MMIO writes to
PCIE_ATU_VIEWPORT, but it's mainly in initialization paths and the code
simplification is significant.

[bhelgaas: commit log, simplify dw_pcie_select_atu()]
Link: https://lore.kernel.org/r/20220624143947.8991-12-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: dwc: Drop enum dw_pcie_region_type in favor of PCIE_ATU_REGION_DIR_IB/OB
Serge Semin [Fri, 24 Jun 2022 14:39:42 +0000 (17:39 +0300)]
PCI: dwc: Drop enum dw_pcie_region_type in favor of PCIE_ATU_REGION_DIR_IB/OB

Previously callers of dw_pcie_disable_atu() supplied enum
dw_pcie_region_type (DW_PCIE_REGION_INBOUND, DW_PCIE_REGION_OUTBOUND),
which dw_pcie_disable_atu() converted to the PCIE_ATU_REGION_DIR_IB or
PCIE_ATU_REGION_DIR_OB values needed to program the ATU registers.

Simplify the code by dropping the dw_pcie_region_type enum and passing
PCIE_ATU_REGION_DIR_IB or PCIE_ATU_REGION_DIR_OB directly.

Reorder dw_pcie_disable_atu() arguments to (dir, index) since "index"
indicates an ATU window in the regions of the corresponding direction.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-11-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Drop enum dw_pcie_as_type in favor of PCIE_ATU_TYPE_MEM/IO
Serge Semin [Fri, 24 Jun 2022 14:39:41 +0000 (17:39 +0300)]
PCI: dwc: Drop enum dw_pcie_as_type in favor of PCIE_ATU_TYPE_MEM/IO

Previously dw_pcie_ep_set_bar() converted the BAR PCI_BASE_ADDRESS_SPACE
bit to the internal dw_pcie_as_type enum (DW_PCIE_AS_MEM, DW_PCIE_AS_IO)
and passed it down to dw_pcie_prog_inbound_atu(), which converted the enum
to the PCIE_ATU_TYPE_MEM/PCIE_ATU_TYPE_IO values needed to program the ATU
registers.

Simplify the code by dropping the dw_pcie_as_type enum and passing
PCIE_ATU_TYPE_MEM or PCIE_ATU_TYPE_IO directly.

Reorder inbound ATU function arguments to match the outbound functions,
with address-related parameters at the end.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-10-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Add dw_pcie_ops.host_deinit() callback
Serge Semin [Fri, 24 Jun 2022 14:39:40 +0000 (17:39 +0300)]
PCI: dwc: Add dw_pcie_ops.host_deinit() callback

dw_pcie_host_init() calls the dw_pcie_ops.host_init() callback to do
platform-specific host initialization.

Add a dw_pcie_ops.host_deinit() callback to perform the corresponding
cleanups in dw_pcie_host_deinit() and in dw_pcie_host_init() failure paths.

Link: https://lore.kernel.org/r/20220624143947.8991-9-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: tegra194: Drop manual DW PCIe controller version setup
Serge Semin [Fri, 24 Jun 2022 14:39:39 +0000 (17:39 +0300)]
PCI: tegra194: Drop manual DW PCIe controller version setup

Since the DW PCIe common code (dw_pcie_version_detect()) now reads the IP
core version directly from the hardware, there is no point manually setting
the version for controllers newer than v4.70a.

Tegra194 only supports v4.90a, so remove the now-superfluous code that sets
struct dw_pcie.version.

Suggested-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220624143947.8991-8-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: intel-gw: Drop manual DW PCIe controller version setup
Serge Semin [Fri, 24 Jun 2022 14:39:38 +0000 (17:39 +0300)]
PCI: intel-gw: Drop manual DW PCIe controller version setup

Since the DW PCIe common code (dw_pcie_version_detect()) now reads the IP
core version directly from the hardware, there is no point manually setting
the version for controllers newer than v4.70a.

Remove the now-superfluous intel-gw code that sets struct dw_pcie.version.

Suggested-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220624143947.8991-7-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Add macros to compare Synopsys IP core versions
Serge Semin [Fri, 24 Jun 2022 14:39:37 +0000 (17:39 +0300)]
PCI: dwc: Add macros to compare Synopsys IP core versions

Add macros to compare DWC IP core versions:

  dw_pcie_ver_is()
  dw_pcie_ver_is_ge()
  dw_pcie_ver_type_is()
  dw_pcie_ver_type_is_ge()

These are along the lines of DWC3_VER_IS() and dw_spi_ver_is().

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-6-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Read DWC IP core version from register
Serge Semin [Fri, 24 Jun 2022 14:39:36 +0000 (17:39 +0300)]
PCI: dwc: Read DWC IP core version from register

Since DWC PCIe v4.70a, the controller version and version type can be read
from the PORT_LOGIC.PCIE_VERSION_OFF and PORT_LOGIC.PCIE_VERSION_TYPE_OFF
registers respectively.

Read the version from those registers and warn if if's different from the
version we got from the device tree.

We can only read the version after platform-specific drivers have done any
DBI-related initialization, such as reference clock activation.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-5-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Use native DWC IP core version representation
Serge Semin [Fri, 24 Jun 2022 14:39:35 +0000 (17:39 +0300)]
PCI: dwc: Use native DWC IP core version representation

Save the DWC IP core version in the same format as the
PORT_LOGIC.PCIE_VERSION_OFF register, similar to what other drivers for DWC
IP do (dw_spi_hw_init(), dwc3_core_is_valid(), stmmac_hwif_init()).

[bhelgaas: trim commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-4-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Detect iATU settings after getting "addr_space" resource
Serge Semin [Fri, 24 Jun 2022 14:39:34 +0000 (17:39 +0300)]
PCI: dwc: Detect iATU settings after getting "addr_space" resource

Previously, dw_pcie_ep_init() did:

  dw_pcie_iatu_detect(pci);
  res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
  if (!res)
    return -EINVAL;

The platform_get_resource_byname() can fail, and dw_pcie_iatu_detect()
doesn't depend on the "addr_space" resource, so delay it until afterwards,
i.e.,

  platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
  dw_pcie_iatu_detect(pci);

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143947.8991-3-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Log link speed and width if it comes up
Serge Semin [Fri, 24 Jun 2022 14:39:33 +0000 (17:39 +0300)]
PCI: dwc: Log link speed and width if it comes up

Printing just "link up" isn't very informative for PCI Express. Even if the
link is up, bus performance can degrade to slower speeds or to narrower
width than both Root Port and its partner is capable of. In that case it
would be handy to know the link specifications as early as possible.

If the link comes up, log the link speed (PCIe generation) and width.

Link: https://lore.kernel.org/r/20220624143947.8991-2-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agosparc: Use generic pci_mmap_resource_range()
Arnd Bergmann [Fri, 15 Jul 2022 15:36:17 +0000 (17:36 +0200)]
sparc: Use generic pci_mmap_resource_range()

The main feature of the sparc-specific implementation of
pci_mmap_resource_range() is that it allows mapping the entire PCI I/O
space for a PCI host bridge using the /proc/bus/pci interface on a bridge
device.

The generic implementation cannot do this, but it also appears that this
got broken for sparc by commit 9eff02e2042f ("PCI: check mmap range of
/proc/bus/pci files too"), which enforces that each address is part of a
BAR for kernels after 2.6.28.

Remove it all, assuming that the corresponding user space code has already
been changed to access /dev/ioport instead a long time ago.  Add
pci_iobar_pfn() to make it possible to map I/O resources. This is adapted
from the powerpc version.

Link: https://lore.kernel.org/lkml/1519887203.622.3.camel@infradead.org/t/
Link: https://lore.kernel.org/r/20220715153617.3393420-2-arnd@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI: Remove pci_mmap_page_range() wrapper
Arnd Bergmann [Fri, 15 Jul 2022 15:36:16 +0000 (17:36 +0200)]
PCI: Remove pci_mmap_page_range() wrapper

The ARCH_GENERIC_PCI_MMAP_RESOURCE symbol came up in a recent discussion,
and I noticed that this was left behind by an unfinished cleanup from 2017.

The only architecture that still relies on providing its own
pci_mmap_page_range() helper instead of using the generic
pci_mmap_resource_range() is sparc. Presumably the reasons for this have
not changed, but at least this can be simplified by converting sparc to use
the same interface as the others.

The only difference between the two is the device-specific offset that gets
added to or subtracted from vma->vm_pgoff.

Change the only caller of pci_mmap_page_range() in common code to subtract
this offset and call the modern interface, while adding it back in the
sparc implementation to preserve the existing behavior.

This removes the complexities of the dual interfaces from the common code,
and keeps it all specific to the sparc architecture code. According to
David Miller, the sparc code lets user space poke into the VGA I/O port
registers by mmapping the I/O space of the parent bridge device, which is
something that the generic pci_mmap_resource_range() code apparently does
not.

Link: https://lore.kernel.org/lkml/1519887203.622.3.camel@infradead.org/t/
Link: https://lore.kernel.org/lkml/20220714214657.2402250-3-shorne@gmail.com/
Link: https://lore.kernel.org/r/20220715153617.3393420-1-arnd@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Stafford Horne <shorne@gmail.com>
2 years agoPCI: brcmstb: Rename .map_bus() functions to end with 'map_bus'
Jim Quinlan [Tue, 26 Jul 2022 17:39:11 +0000 (12:39 -0500)]
PCI: brcmstb: Rename .map_bus() functions to end with 'map_bus'

Rename the .map_bus() functions to end with 'map_bus' so they're easy to
find with, e.g., 'git grep "^static.*_map_bus" drivers/pci/'.

[bhelgaas: rename brcm_pcie_map_bus32() to brcm7425_pcie_map_bus() for
better cscope-ability (".*_map_bus" is not the same as ".*_map_bus.*")]
Link: https://lore.kernel.org/r/20220725151258.42574-8-jim2101024@gmail.com
Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
2 years agoPCI: brcmstb: Disable/enable regulators in suspend/resume
Jim Quinlan [Mon, 25 Jul 2022 15:12:54 +0000 (11:12 -0400)]
PCI: brcmstb: Disable/enable regulators in suspend/resume

If we found power regulators for a device below the Root Port, disable them
during suspend and re-enable them during resume.

If any downstream device can be a wakeup device, do not turn off the
regulators as the device will need them on.

[bhelgaas: drop unused regulator_oops, skip wrapping of .add_bus()/
.remove_bus(), move brcm_pcie_start_link() to .add_bus() in previous patch,
squash WOL checking into this patch]
Link: https://lore.kernel.org/r/20220725151258.42574-6-jim2101024@gmail.com
Link: https://lore.kernel.org/r/20220725151258.42574-7-jim2101024@gmail.com
Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
2 years agoPCI: brcmstb: Enable child bus device regulators from DT
Jim Quinlan [Mon, 25 Jul 2022 15:12:53 +0000 (11:12 -0400)]
PCI: brcmstb: Enable child bus device regulators from DT

Some platforms have power regulators for slots or devices below Root Ports.
On platforms like Raspberry Pi 4, these regulators are described in the
Root Port device tree node, since they logically belong to the Root Port,
not to the host bridge itself.

Add an .add_bus() hook (called when pci_alloc_child_bus() allocates the
secondary ("child") bus for a bridge), and look for such regulators.  If we
find some, enable them before bringing up the link and enumerating devices
on the child bus.

Similarly, when pci_remove_bus() calls the ops->remove_bus() hook, disable
the regulators.

The regulators that may be described in a Root Port DT device are:

  vpcie3v3
  vpcie3v3aux
  vpcie12v

These control power to the device downstream from the Root Port.

[bhelgaas: commit log, name hooks brcm_pcie_add_bus(), etc, since we only
support one set of subregulator info, save info in struct brcm_pcie instead
of dev->driver_data, move brcm_pcie_start_link() from probe to .add_bus()
(from subsequent patch)]
Link: https://lore.kernel.org/r/20220725151258.42574-5-jim2101024@gmail.com
Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
2 years agoPCI: brcmstb: Split post-link up initialization to brcm_pcie_start_link()
Jim Quinlan [Mon, 25 Jul 2022 15:12:51 +0000 (11:12 -0400)]
PCI: brcmstb: Split post-link up initialization to brcm_pcie_start_link()

Previously brcm_pcie_setup() initialized the Root Port itself as well as
doing the actual link-up.  Split brcm_pcie_setup() into two functions:

  - brcm_pcie_setup(), which initializes everything that does not require
    the link itself to be up, and

  - brcm_pcie_start_link(), which brings up the link and initializes things
    that depend on the link being up.

[bhelgaas: condense commit log, deferring details for future changes]
Link: https://lore.kernel.org/r/20220725151258.42574-3-jim2101024@gmail.com
Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
2 years agoPCI: brcmstb: Prevent config space access when link is down
Jim Quinlan [Mon, 25 Jul 2022 15:12:52 +0000 (11:12 -0400)]
PCI: brcmstb: Prevent config space access when link is down

When the link is down, config accesses to downstream devices cause CPU
aborts.  Allow config accesses only when the link is up.

As the following scenario shows, this check is racy and cannot completely
avoid CPU aborts, but it makes them less likely:

  pci_generic_config_read
    addr = brcm_pcie_map_conf            # bus->ops->map_bus()
      brcm_pcie_link_up                  # returns "true"; link is up
    <link goes down>
    *val = readb(addr)                   # link is now down
    <CPU abort>

Note that config space accesses to the Root Port are not affected by link
status.

[bhelgaas: commit log, use PCIE_ECAM_REG() instead of magic 0xfff masks;
note that pci_generic_config_read32() masks low two bits already]
Link: https://lore.kernel.org/r/20220725151258.42574-4-jim2101024@gmail.com
Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
2 years agoPCI: brcmstb: Remove unnecessary forward declarations
Jim Quinlan [Mon, 25 Jul 2022 15:12:50 +0000 (11:12 -0400)]
PCI: brcmstb: Remove unnecessary forward declarations

Remove forward function declarations in this driver.  Also move some
constant structure definitions lower in the file.  There are no changes to
the code that has been moved.

Link: https://lore.kernel.org/r/20220725151258.42574-2-jim2101024@gmail.com
Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
2 years agos390/pci: allow zPCI zbus without a function zero
Niklas Schnelle [Tue, 28 Jun 2022 14:31:00 +0000 (16:31 +0200)]
s390/pci: allow zPCI zbus without a function zero

Currently the zPCI code blocks PCI bus creation and probing of a zPCI zbus
unless there is a PCI function with devfn 0. This is always the case for
the PCI functions with hidden RID, but may keep PCI functions from a
multi-function PCI device with RID information invisible until the function
0 becomes visible. Worse, as a PCI bus is necessary to even present a PCI
hotplug slot, even that remains invisible.

With the probing of these so-called isolated PCI functions enabled for s390
in common code, this restriction is no longer necessary. On network cards
with multiple ports and a PF per port this also allows using each port on
its own while still providing the physical PCI topology information in the
devfn needed to associate VFs with their parent PF.

Link: https://lore.kernel.org/r/20220628143100.3228092-6-schnelle@linux.ibm.com
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
2 years agoPCI: Extend isolated function probing to s390
Niklas Schnelle [Tue, 28 Jun 2022 14:30:59 +0000 (16:30 +0200)]
PCI: Extend isolated function probing to s390

Like the jailhouse hypervisor, s390's PCI architecture allows passing
isolated PCI functions to a guest OS instance. As of now this is was not
utilized even with multi-function support as the s390 PCI code makes sure
that only virtual PCI busses including a function with devfn 0 are
presented to the PCI subsystem. A subsequent change will remove this
restriction.

Allow probing such functions by replacing the existing check for
jailhouse_paravirt() with a new hypervisor_isolated_pci_functions() helper.

Link: https://lore.kernel.org/r/20220628143100.3228092-5-schnelle@linux.ibm.com
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
2 years agoPCI: Move jailhouse's isolated function handling to pci_scan_slot()
Niklas Schnelle [Tue, 28 Jun 2022 14:30:58 +0000 (16:30 +0200)]
PCI: Move jailhouse's isolated function handling to pci_scan_slot()

The special case of the jailhouse hypervisor passing through individual PCI
functions handles scanning for PCI functions even if function 0 does not
exist. Previously this was done with an extra loop duplicating the one in
pci_scan_slot(). By incorporating the check for jailhouse_paravirt() into
pci_scan_slot() we can instead do this as part of the normal slot scan.
Note that with the assignment of dev->multifunction gated by fn > 0 we set
dev->multifunction unconditionally for all functions if function 0 is
missing just as in the existing jailhouse loop.

The only functional change is that we now call pcie_aspm_init_link_state()
for these functions, but this already happened if function 0 was passed
through and should not be a problem.

Link: https://lore.kernel.org/linux-pci/20220408224514.GA353445@bhelgaas/
Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20220628143100.3228092-4-schnelle@linux.ibm.com
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
2 years agoPCI: Split out next_ari_fn() from next_fn()
Niklas Schnelle [Tue, 28 Jun 2022 14:30:57 +0000 (16:30 +0200)]
PCI: Split out next_ari_fn() from next_fn()

In commit b1bd58e448f2 ("PCI: Consolidate "next-function" functions") the
next_fn() function subsumed the traditional and ARI-based next function
determination. This got rid of some needlessly complex function pointer
handling but also reduced the separation between these very different
methods of finding the next function. With the next_fn() cleaned up a bit
we can re-introduce this separation by moving out the ARI handling while
sticking with direct function calls.

Link: https://lore.kernel.org/r/20220628143100.3228092-3-schnelle@linux.ibm.com
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
2 years agoPCI: Clean up pci_scan_slot()
Niklas Schnelle [Tue, 28 Jun 2022 14:30:56 +0000 (16:30 +0200)]
PCI: Clean up pci_scan_slot()

While determining the next PCI function is factored out of pci_scan_slot()
into next_fn(), the former still handles the first function as a special
case, which duplicates the code from the scan loop.

Furthermore the non-ARI branch of next_fn() is generally hard to understand
and especially the check for multifunction devices is hidden in the
handling of NULL devices for non-contiguous multifunction. It also signals
that no further functions need to be scanned by returning 0 via wraparound
and this is a valid function number.

Improve upon this by transforming the conditions in next_fn() to be easier
to understand.

By changing next_fn() to return -ENODEV instead of 0 when there is no next
function we can then handle the initial function inside the loop and
deduplicate the shared handling. This also makes it more explicit that only
function 0 must exist.

No functional change is intended.

Link: https://lore.kernel.org/r/20220628143100.3228092-2-schnelle@linux.ibm.com
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
2 years agoPCI: fu740: Remove unnecessary include files
Bjorn Helgaas [Tue, 19 Jul 2022 17:28:05 +0000 (12:28 -0500)]
PCI: fu740: Remove unnecessary include files

fu740 uses no syscon or regman interfaces, so it doesn't need to include
mfs/syscon.h.  It uses no regulator interfaces, so it doesn't need to
include regulator/consumer.h either.

Remove both unnecessary includes.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI/AER: Iterate over error counters instead of error strings
Mohamed Khalfella [Mon, 9 May 2022 18:14:41 +0000 (18:14 +0000)]
PCI/AER: Iterate over error counters instead of error strings

Previously we iterated over AER stat *names*, e.g.,
aer_correctable_error_string[32], but the actual stat *counters* may not be
that large, e.g., pdev->aer_stats->dev_cor_errs[16], which means that we
printed junk in the sysfs stats files.

Iterate over the stat counter arrays instead of the names to avoid this
junk.

Also, added a build time check to make sure all
counters have entries in strings array.

Fixes: 0678e3109a3c ("PCI/AER: Simplify __aer_print_error()")
Link: https://lore.kernel.org/r/20220509181441.31884-1-mkhalfella@purestorage.com
Reported-by: Meeta Saggi <msaggi@purestorage.com>
Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Meeta Saggi <msaggi@purestorage.com>
Reviewed-by: Eric Badger <ebadger@purestorage.com>
Cc: stable@vger.kernel.org
2 years agoPCI/AER: Enable error reporting when AER is native
Stefan Roese [Tue, 25 Jan 2022 07:18:20 +0000 (08:18 +0100)]
PCI/AER: Enable error reporting when AER is native

If we have native control of AER, set the following error reporting enable
bits:

  - Correctable Error Reporting Enable
  - Non-Fatal Error Reporting Enable
  - Fatal Error Reporting Enable
  - Unsupported Request Reporting Enable

Note that these bits are all in the Device Control register and are not
AER-specific.

This affects all devices with an AER capability, including hot-added
devices.

Please note that this change is quite invasive, as error reporting now will
be enabled for all available PCIe Endpoints, which was previously not the
case.

When "pci=noaer" is selected, error reporting stays disabled of course.

[bhelgaas: commit log, note error reporting is not AER-specific]
Link: https://lore.kernel.org/r/20220125071820.2247260-4-sr@denx.de
Signed-off-by: Stefan Roese <sr@denx.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pali Rohár <pali@kernel.org>
Cc: Bharat Kumar Gogada <bharat.kumar.gogada@xilinx.com>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: Yao Hongbo <yaohongbo@linux.alibaba.com>
Cc: Naveen Naidu <naveennaidu479@gmail.com>
2 years agoPCI/portdrv: Don't disable AER reporting in get_port_device_capability()
Stefan Roese [Tue, 25 Jan 2022 07:18:19 +0000 (08:18 +0100)]
PCI/portdrv: Don't disable AER reporting in get_port_device_capability()

AER reporting is currently disabled in the DevCtl registers of all non Root
Port PCIe devices on systems using pcie_ports_native || host->native_aer,
disabling AER completely in such systems. This is because 2bd50dd800b5
("PCI: PCIe: Disable PCIe port services during port initialization"), added
a call to pci_disable_pcie_error_reporting() *after* the AER setup was
completed for the PCIe device tree.

Here a longer analysis about the current status of AER enabling /
disabling upon bootup provided by Bjorn:

  pcie_portdrv_probe
    pcie_port_device_register
      get_port_device_capability
        pci_disable_pcie_error_reporting
          clear CERE NFERE FERE URRE               # <-- disable for RP USP DSP
      pcie_device_init
        device_register                            # new AER service device
          aer_probe
            aer_enable_rootport                    # RP only
              set_downstream_devices_error_reporting
                set_device_error_reporting         # self (RP)
                  if (RP || USP || DSP)
                    pci_enable_pcie_error_reporting
                      set CERE NFERE FERE URRE     # <-- enable for RP
                pci_walk_bus
                  set_device_error_reporting
                    if (RP || USP || DSP)
                      pci_enable_pcie_error_reporting
                        set CERE NFERE FERE URRE   # <-- enable for USP DSP

In a typical Root Port -> Endpoint hierarchy, the above:
  - Disables Error Reporting for the Root Port,
  - Enables Error Reporting for the Root Port,
  - Does NOT enable Error Reporting for the Endpoint because it is not a
    Root Port or Switch Port.

In a deeper Root Port -> Upstream Switch Port -> Downstream Switch
Port -> Endpoint hierarchy:
  - Disables Error Reporting for the Root Port,
  - Enables Error Reporting for the Root Port,
  - Enables Error Reporting for both Switch Ports,
  - Does NOT enable Error Reporting for the Endpoint because it is not a
    Root Port or Switch Port,
  - Disables Error Reporting for the Switch Ports when pcie_portdrv_probe()
    claims them.  AER does not re-enable it because these are not Root
    Ports.

Remove this call to pci_disable_pcie_error_reporting() from
get_port_device_capability(), leaving the already enabled AER configuration
intact. With this change, AER is enabled in the Root Port and the PCIe
switch upstream and downstream ports. Only the PCIe Endpoints don't have
AER enabled yet. A follow-up patch will take care of this Endpoint
enabling.

Fixes: 2bd50dd800b5 ("PCI: PCIe: Disable PCIe port services during port initialization")
Link: https://lore.kernel.org/r/20220125071820.2247260-3-sr@denx.de
Signed-off-by: Stefan Roese <sr@denx.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pali Rohár <pali@kernel.org>
Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Cc: Bharat Kumar Gogada <bharat.kumar.gogada@xilinx.com>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: Yao Hongbo <yaohongbo@linux.alibaba.com>
Cc: Naveen Naidu <naveennaidu479@gmail.com>
2 years agoPCI/ASPM: Unexport pcie_aspm_support_enabled()
Bjorn Helgaas [Fri, 24 Jun 2022 21:27:19 +0000 (16:27 -0500)]
PCI/ASPM: Unexport pcie_aspm_support_enabled()

pcie_aspm_support_enabled() is used only by the acpi/pci_root.c driver,
which cannot be built as a module, so it does not need to be exported.
Unexport it.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2 years agoPCI: endpoint: Enable DMA tests for endpoints with DMA capabilities
Frank Li [Tue, 24 May 2022 15:21:59 +0000 (10:21 -0500)]
PCI: endpoint: Enable DMA tests for endpoints with DMA capabilities

Some PCI Endpoint controllers integrate an eDMA (embedded DMA).  eDMA can
bypass the outbound memory address translation unit to access all RC memory
space.

Add eDMA support for pci-epf-test.

Depending on HW availability, the EPF test can use either eDMA or general
system DMA controllers to perform DMA. The test tries to use eDMA first and
falls back to general system DMA controllers if there's no eDMA

Separate dma_chan to dma_chan_tx and dma_chan_rx. Search for an eDMA
channel first, then search for a memory-to-memory DMA channel.  If general
memory to memory channels are used, dma_chan_rx = dma_chan_tx.

Add dma_addr_t dma_remote in pci_epf_test_data_transfer() because eDMA uses
remote RC physical address directly.

Add enum dma_transfer_direction dir in pci_epf_test_data_transfer() because
eDMA chooses the correct RX/TX channel by dir.

The overall steps are:

  1. Execute dma_request_channel() and filter function to find correct eDMA
     RX and TX Channel. If a channel does not exist, fallback to try to
     allocate general memory to memory DMA channel.

  2. Execute dmaengine_slave_config() to configure remote side physical
     address.

  3. Execute dmaengine_prep_slave_single() to create transfer descriptor.

  4. Execute tx_submit().

  5. Execute dma_async_issue_pending()

[bhelgaas: squash in fix from Dan Carpenter <dan.carpenter@oracle.com>:
https://lore.kernel.org/r/Ys2GSTnZhuLzzQG5@kili, also previously posted by
Peng Wu <wupeng58@huawei.com>:
https://lore.kernel.org/all/CANXvt5rK98-cEMgpzopY9POOK8a5=VDib8uKPLgJakOG=hRfwQ@mail.gmail.com/]
Link: https://lore.kernel.org/r/20220524152159.2370739-9-Frank.Li@nxp.com
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
Acked-By: Vinod Koul <vkoul@kernel.org>
2 years agoPCI/ASPM: Remove pcie_aspm_pm_state_change()
Kai-Heng Feng [Mon, 11 Jul 2022 23:07:01 +0000 (18:07 -0500)]
PCI/ASPM: Remove pcie_aspm_pm_state_change()

pcie_aspm_pm_state_change() was introduced at the inception of PCIe ASPM
code, but it can cause some issues. For instance, when ASPM config is
changed via sysfs, those changes won't persist across power state change
because pcie_aspm_pm_state_change() overwrites them.

Also, if the driver restores L1SS [1] after system resume, the restored
state will also be overwritten by pcie_aspm_pm_state_change().

Remove pcie_aspm_pm_state_change().  If there's any hardware that really
needs it to function, a quirk can be used instead.

[1] https://lore.kernel.org/linux-pci/20220201123536.12962-1-vidyas@nvidia.com/
Link: https://lore.kernel.org/r/20220509073639.2048236-1-kai.heng.feng@canonical.com
[bhelgaas: remove additional pcie_aspm_pm_state_change() call in
pci_set_low_power_state(), added by
10aa5377fc8a ("PCI/PM: Split pci_raw_set_power_state()") and moved by
7957d201456f ("PCI/PM: Relocate pci_set_low_power_state()")]
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI/doc: Convert examples to generic power management
Bjorn Helgaas [Tue, 7 Jun 2022 23:29:46 +0000 (18:29 -0500)]
PCI/doc: Convert examples to generic power management

PCI-specific power management (pci_driver.suspend and pci_driver.resume) is
deprecated.  Convert sample code to the generic power management framework.

Link: https://lore.kernel.org/r/20220607232946.355987-1-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoPCI/AER: Configure ECRC for every device
Stefan Roese [Tue, 25 Jan 2022 07:18:18 +0000 (08:18 +0100)]
PCI/AER: Configure ECRC for every device

Move pcie_set_ecrc_checking() to pci_aer_init() to make sure that
pcie_set_ecrc_checking() is called for each PCIe device, including
hot-added devices.

Link: https://lore.kernel.org/r/20220125071820.2247260-2-sr@denx.de
Signed-off-by: Stefan Roese <sr@denx.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pali Rohár <pali@kernel.org>
Cc: Bharat Kumar Gogada <bharat.kumar.gogada@xilinx.com>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: Yao Hongbo <yaohongbo@linux.alibaba.com>
Cc: Naveen Naidu <naveennaidu479@gmail.com>
2 years agoPCI: endpoint: Don't stop controller when unbinding endpoint function
Shunsuke Mie [Wed, 22 Jun 2022 04:09:24 +0000 (13:09 +0900)]
PCI: endpoint: Don't stop controller when unbinding endpoint function

Unbinding an endpoint function from the endpoint controller shouldn't stop
the controller.  This is especially a problem for multi-function endpoints
where other endpoints may still be active.

Don't stop the controller when unbinding one of its endpoints.  Normally
the controller is stopped via configfs.

Fixes: 349e7a85b25f ("PCI: endpoint: functions: Add an EP function to test PCI")
Link: https://lore.kernel.org/r/20220622040924.113279-1-mie@igel.co.jp
Signed-off-by: Shunsuke Mie <mie@igel.co.jp>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
2 years agoPCI: dwc-plat: Drop dw_plat_pcie_of_match[] forward declaration
Serge Semin [Fri, 24 Jun 2022 14:34:28 +0000 (17:34 +0300)]
PCI: dwc-plat: Drop dw_plat_pcie_of_match[] forward declaration

The dw_plat_pcie_of_match[] forward declaration was required when
dw_plat_pcie_probe() called of_match_device().  5c204204cf24 ("PCI:
designware-plat: Prefer of_device_get_match_data()") replaced that with
of_device_get_match_data(), which no longer needs the declaration.

Drop the unnecessary forward declaration.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-19-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc-plat: Drop unused regmap pointer
Serge Semin [Fri, 24 Jun 2022 14:34:27 +0000 (17:34 +0300)]
PCI: dwc-plat: Drop unused regmap pointer

1d906b22076e ("PCI: dwc: Add support for EP mode") added the struct
dw_plat_pcie regmap pointer, but it has never been used.  Remove it.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-18-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc-plat: Simplify dw_plat_pcie_probe() return values
Serge Semin [Fri, 24 Jun 2022 14:34:26 +0000 (17:34 +0300)]
PCI: dwc-plat: Simplify dw_plat_pcie_probe() return values

Save the return value in "ret" for all three cases (DW_PCIE_RC_TYPE,
DW_PCIE_EP_TYPE, default) handled by dw_plat_pcie_probe() and return from a
single place.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-17-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Rename struct pcie_port to dw_pcie_rp
Serge Semin [Fri, 24 Jun 2022 14:34:25 +0000 (17:34 +0300)]
PCI: dwc: Rename struct pcie_port to dw_pcie_rp

All of the DW PCIe core driver entities except the pcie_port struct have
names with the "dw_" prefix to distinguish local and common PCIe name
spaces, and endpoint-related entities have an "_ep" suffix.

Rename struct pcie_port to dw_pcie_rp to make it more consistent with other
names.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-16-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Jesper Nilsson <jesper.nilsson@axis.com>
2 years agoPCI: dwc: Move io_cfg_atu_shared to struct pcie_port
Serge Semin [Fri, 24 Jun 2022 14:34:24 +0000 (17:34 +0300)]
PCI: dwc: Move io_cfg_atu_shared to struct pcie_port

The io_cfg_atu_shared flag is set if there is an outbound iATU window used
for both config space accesses and IO port transfers.

Since the flag semantic is purely Root Port specific, it's not used in
either the DW PCIe common code or in the DW PCIe Endpoint driver. Move it
to the struct pcie_port and rename to cfg0_io_shared.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-15-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Add start_link/stop_link inlines
Serge Semin [Fri, 24 Jun 2022 14:34:23 +0000 (17:34 +0300)]
PCI: dwc: Add start_link/stop_link inlines

Factor out this pattern:

  if (!pci->ops || !pci->ops->start_link)
    return -EINVAL;

  return pci->ops->start_link(pci);

into a new dw_pcie_start_link() wrapper and do the same for the stop_link()
method.

Note that dw_pcie_ep_start() previously returned -EINVAL if there was no
platform start_link() method, which didn't make much sense since that is
not an error.  It will now return 0 in that case.

As a side-effect, drop the empty start_link() and dummy dw_pcie_ops
instances from the generic DW PCIe and Layerscape EP platform drivers.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-14-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Reuse local pointer to the resource data
Serge Semin [Fri, 24 Jun 2022 14:34:22 +0000 (17:34 +0300)]
PCI: dwc: Reuse local pointer to the resource data

dw_pcie_host_init() has two instances of the resource structure pointers
used in unrelated places. It's pointless to have two different local
storages for them since the corresponding code is small and having
resource-specific names doesn't make it more readable.

Convert these parts of the function to use a common pointer to the
resource structure instance.

Link: https://lore.kernel.org/r/20220624143428.8334-13-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Organize local variable usage
Serge Semin [Fri, 24 Jun 2022 14:34:21 +0000 (17:34 +0300)]
PCI: dwc: Organize local variable usage

There are several places in the common DW PCIe code with incoherent local
variable usage: a variable is defined and initialized with a structure
field, but the structure pointer is dereferenced to access that field
anyway; the local variable is defined and initialized but either used just
once or not used afterwards in the main part of the subsequent method.  It
mainly concerns the pcie_port.dev field. Fix that in the relevant places.

Link: https://lore.kernel.org/r/20220624143428.8334-12-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Convert dw_pcie_link_up() to use dw_pcie_readl_dbi()
Serge Semin [Fri, 24 Jun 2022 14:34:20 +0000 (17:34 +0300)]
PCI: dwc: Convert dw_pcie_link_up() to use dw_pcie_readl_dbi()

While the rest of the generic DWC PCIe code uses the dedicated IO-mem
accessors, the dw_pcie_link_up() method for some unobvious reason directly
calls readl() to get PortLogic.DEBUG1 register content. Since the way the
DBI bus is accessed can be platform-specific, use dw_pcie_readl_dbi()
instead so dw_pcie_link_up() is slightly more generic.

Link: https://lore.kernel.org/r/20220624143428.8334-11-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Simplify unrolled iATU detection
Serge Semin [Fri, 24 Jun 2022 14:34:19 +0000 (17:34 +0300)]
PCI: dwc: Simplify unrolled iATU detection

The unrolled version of the internal ATU has been available since the DWC
PCIe v4.80a IP core, but it may not be enabled.  Per [1], if unrolled ATU
is enabled, the PCIE_ATU_VIEWPORT does not exist and reads as 0xffffffff;
while if unrolled ATU is disabled, PCIE_ATU_VIEWPORT will contain some
zeros.

Simplify dw_pcie_iatu_unroll_enabled() by checking the value of
PCIE_ATU_VIEWPORT.

[1] DesignWare Cores, PCI Express Controller, Register Desciptions,
v.4.90a, December 2016, p.855

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-10-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Add newlines to log messages
Serge Semin [Fri, 24 Jun 2022 14:34:18 +0000 (17:34 +0300)]
PCI: dwc: Add newlines to log messages

Add newlines to log messages that are missing them.

[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20220624143428.8334-9-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Add braces to multi-line if-else statements
Serge Semin [Fri, 24 Jun 2022 14:34:17 +0000 (17:34 +0300)]
PCI: dwc: Add braces to multi-line if-else statements

Add braces around single-line if-else statements when the opposite case
requires them.

Link: https://lore.kernel.org/r/20220624143428.8334-8-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Always enable CDM check if "snps,enable-cdm-check" exists
Serge Semin [Fri, 24 Jun 2022 14:34:16 +0000 (17:34 +0300)]
PCI: dwc: Always enable CDM check if "snps,enable-cdm-check" exists

If the "snps,enable-cdm-check" property exists, we should enable the CDM
check.  But previously dw_pcie_setup() could exit before doing so if the
"num-lanes" property was absent or invalid.

Move the CDM enable earlier so we do it regardless of whether "num-lanes"
is present.

[bhelgaas: commit log]
Fixes: 07f123def73e ("PCI: dwc: Add support to enable CDM register check")
Link: https://lore.kernel.org/r/20220624143428.8334-7-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Deallocate EPC memory on dw_pcie_ep_init() errors
Serge Semin [Fri, 24 Jun 2022 14:34:15 +0000 (17:34 +0300)]
PCI: dwc: Deallocate EPC memory on dw_pcie_ep_init() errors

If dw_pcie_ep_init() fails to perform any action after the EPC memory is
initialized and the MSI memory region is allocated, the latter parts won't
be undone thus causing a memory leak.  Add a cleanup-on-error path to fix
these leaks.

[bhelgaas: commit log]
Fixes: 2fd0c9d966cc ("PCI: designware-ep: Pre-allocate memory for MSI in dw_pcie_ep_init")
Link: https://lore.kernel.org/r/20220624143428.8334-6-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
2 years agoPCI: dwc: Set INCREASE_REGION_SIZE flag based on limit address
Serge Semin [Fri, 24 Jun 2022 14:34:14 +0000 (17:34 +0300)]
PCI: dwc: Set INCREASE_REGION_SIZE flag based on limit address

We program the 64-bit ATU limit address (in PCIE_ATU_LIMIT/
PCIE_ATU_UPPER_LIMIT or PCIE_ATU_UNR_LOWER_LIMIT/PCIE_ATU_UNR_UPPER_LIMIT),
but in addition, the PCIE_ATU_INCREASE_REGION_SIZE bit must be set if the
upper 32 bits of the limit address differ from the upper 32 bits of the
base address (see [1,2]).

5b4cf0f65324 ("PCI: dwc: Add upper limit address for outbound iATU") set
PCIE_ATU_INCREASE_REGION_SIZE, but only when the *size* was greater than
4GB.  It did not set it when a smaller region crossed a 4GB boundary, e.g.,
[mem 0x0_f0000000-0x1_0fffffff].

Set PCIE_ATU_INCREASE_REGION_SIZE whenever PCIE_ATU_UPPER_LIMIT is
greater than PCIE_ATU_UPPER_BASE.

[1] DesignWare Cores PCI Express Controller Databook - DWC PCIe Root Port,
    v5.40a, March 2019, fig.3-36, p.175
[2] DesignWare Cores PCI Express Controller Databook - DWC PCIe Root Port,
    v5.40a, March 2019, fig.3-37, p.176

[bhelgaas: commit log]
Fixes: 5b4cf0f65324 ("PCI: dwc: Add upper limit address for outbound iATU")
Link: https://lore.kernel.org/r/20220624143428.8334-5-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Disable outbound windows only for controllers using iATU
Serge Semin [Fri, 24 Jun 2022 14:34:13 +0000 (17:34 +0300)]
PCI: dwc: Disable outbound windows only for controllers using iATU

Some DWC-based controllers (e.g., pcie-al.c and pci-keystone.c, identified
by the fact that they override the default dw_child_pcie_ops) use their own
address translation approach instead of the DWC internal ATU (iATU).  For
those controllers, skip disabling the iATU outbound windows.

[bhelgaas: commit log, update multiple window comment]
Fixes: 458ad06c4cdd ("PCI: dwc: Ensure all outbound ATU windows are reset")
Link: https://lore.kernel.org/r/20220624143428.8334-4-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Add unroll iATU space support to dw_pcie_disable_atu()
Serge Semin [Fri, 24 Jun 2022 14:34:12 +0000 (17:34 +0300)]
PCI: dwc: Add unroll iATU space support to dw_pcie_disable_atu()

dw_pcie_disable_atu() was introduced by f8aed6ec624f ("PCI: dwc:
designware: Add EP mode support") and supported only the viewport version
of the iATU CSRs.

DW PCIe IP cores v4.80a and newer also support unrolled iATU/eDMA space.
Callers of dw_pcie_disable_atu(), including pci_epc_ops.clear_bar(),
pci_epc_ops.unmap_addr(), and dw_pcie_setup_rc(), don't work correctly when
it is enabled.

Add dw_pcie_disable_atu() support for controllers with unrolled iATU CSRs
enabled.

[bhelgaas: commit log]
Fixes: f8aed6ec624f ("PCI: dwc: designware: Add EP mode support")
Link: https://lore.kernel.org/r/20220624143428.8334-3-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: dwc: Stop link on host_init errors and de-initialization
Serge Semin [Fri, 24 Jun 2022 14:34:11 +0000 (17:34 +0300)]
PCI: dwc: Stop link on host_init errors and de-initialization

It's logically correct to undo everything that was done when an error is
discovered or in the corresponding cleanup counterpart. Otherwise the host
controller will be left in an undetermined state. Since the link is set up
in the host_init method, deactivate it there in the cleanup-on-error block
and stop the link in the antagonistic routine - dw_pcie_host_deinit(). Link
deactivation is platform-specific and should be implemented in
dw_pcie_ops.stop_link().

Fixes: 886a9c134755 ("PCI: dwc: Move link handling into common code")
Link: https://lore.kernel.org/r/20220624143428.8334-2-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
2 years agoPCI: Clear PCI_STATUS when setting up device
Kai-Heng Feng [Tue, 17 May 2022 04:37:38 +0000 (12:37 +0800)]
PCI: Clear PCI_STATUS when setting up device

We are seeing Master Abort bit is set on Intel I350 ethernet device and its
root port right after boot, probably happened during BIOS phase:

  00:06.0 PCI bridge [0604]: Intel Corporation Device [8086:464d] (rev 05) (prog-if 00 [Normal decode])
    Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort+ >SERR- <PERR- INTx-

  6e:00.0 Ethernet controller [0200]: Intel Corporation I350 Gigabit Network Connection [8086:1521] (rev 01)
    Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort+ >SERR- <PERR- INTx-

The Master Abort bit is cleared after S3.

Since there's no functional impact found, clear the PCI_STATUS to treat it
anew at setting up.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=215989
Link: https://lore.kernel.org/r/20220517043738.2308499-1-kai.heng.feng@canonical.com
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>