Bjorn Helgaas [Tue, 4 May 2021 15:43:28 +0000 (10:43 -0500)]
Merge branch 'remotes/lorenzo/pci/microchip'
- Make several microchip symbols static (Wei Yongjun)
- Drop redundant dev_err() for platform_get_irq() errors (Krzysztof
Wilczyński)
* remotes/lorenzo/pci/microchip:
PCI: microchip: Remove dev_err() when handing an error from platform_get_irq()
PCI: microchip: Make some symbols static
- Add MediaTek MT8192 system power management support (Jianjun Wang)
* remotes/lorenzo/pci/mediatek:
MAINTAINERS: Add Jianjun Wang as MediaTek PCI co-maintainer
PCI: mediatek-gen3: Add system PM support
PCI: mediatek-gen3: Add MSI support
PCI: mediatek-gen3: Add INTx support
PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192
PCI: Export pci_pio_to_address() for module use
dt-bindings: PCI: mediatek-gen3: Add YAML schema
PCI: mediatek: Add missing MODULE_DEVICE_TABLE
PCI: mediatek: Configure FC and FTS for functions other than 0
Bjorn Helgaas [Tue, 4 May 2021 15:43:26 +0000 (10:43 -0500)]
Merge branch 'remotes/lorenzo/pci/dwc'
- Use generic config accessors for TI AM65x (K3) to fix regression (Kishon
Vijay Abraham I)
- Move MSI Receiver init to dw_pcie_host_init() so it is re-initialized
along with the RC in resume (Jisheng Zhang)
- Remove unused pcie_app_rd() (Jiapeng Chong)
- Move iATU detection earlier to fix regression (Hou Zhiqiang)
* remotes/lorenzo/pci/dwc:
PCI: dwc: Move iATU detection earlier
PCI: dwc/intel-gw: Remove unused function
PCI: dwc: Move dw_pcie_msi_init() to dw_pcie_setup_rc()
PCI: keystone: Let AM65 use the pci_ops defined in pcie-designware-host.c
Bjorn Helgaas [Tue, 4 May 2021 15:43:26 +0000 (10:43 -0500)]
Merge branch 'remotes/lorenzo/pci/cadence'
- Add DT binding and TI j721e support for refclk to PCIe connector (Kishon
Vijay Abraham I)
- Add host mode and endpoint mode DT bindings for TI AM64 SoC (Kishon Vijay
Abraham I)
* remotes/lorenzo/pci/cadence:
PCI: j721e: Add support to provide refclk to PCIe connector
dt-bindings: PCI: ti,j721e: Add endpoint mode dt-bindings for TI's AM64 SoC
dt-bindings: PCI: ti,j721e: Add host mode dt-bindings for TI's AM64 SoC
dt-bindings: PCI: ti,j721e: Add binding to represent refclk to the connector
Add driver for the SiFive FU740 PCIe host controller.
This controller is based on the DesignWare PCIe core.
Co-developed-by: Henry Styles <hes@sifive.com> Co-developed-by: Erik Danie <erik.danie@sifive.com> Co-developed-by: Greentime Hu <greentime.hu@sifive.com> Link: https://lore.kernel.org/r/20210504105940.100004-6-greentime.hu@sifive.com Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com> Signed-off-by: Henry Styles <hes@sifive.com> Signed-off-by: Erik Danie <erik.danie@sifive.com> Signed-off-by: Greentime Hu <greentime.hu@sifive.com> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Greentime Hu [Tue, 4 May 2021 10:59:36 +0000 (18:59 +0800)]
clk: sifive: Use reset-simple in prci driver for PCIe driver
We use reset-simple in this patch so that pcie driver can use
devm_reset_control_get() to get this reset data structure and use
reset_control_deassert() to deassert pcie_power_up_rst_n.
0790ae0564ab ("PCI: Prevent VPD access for QLogic ISP2722") disabled access
to VPD of the ISP2722-based 16/32Gb Fibre Channel to PCIe Adapter because
reading past the end of the VPD caused NMIs.
d3afef2eae86 ("PCI: Determine actual VPD size on first access") limits
reads to the actual size of VPD, which should prevent these NMIs.
d3afef2eae86 was merged *before* 0790ae0564ab, but we think the testing
that prompted 0790ae0564ab ("PCI: Prevent VPD access for QLogic ISP2722")
was done with a kernel that lacked d3afef2eae86. See [1, 2].
Remove the quirk added by 0790ae0564ab ("PCI: Prevent VPD access for QLogic
ISP2722") so customers can read the HBA VPD.
Realtek RTL8169/8168/8125 NIC families indicate VPD capability and an
optional VPD EEPROM can be connected via I2C/SPI. However I haven't seen
any card or system with such a VPD EEPROM yet. The missing EEPROM causes
the following warning whenever e.g. lscpi -vv is executed.
invalid short VPD tag 00 at offset 01
The warning confuses users, and I think we should handle the situation more
gently. Therefore, if first VPD byte is read as 0x00, assume a missing
optional VPD PROM and replace the warning with a more descriptive message
at info level.
Heiner Kallweit [Fri, 12 Feb 2021 10:02:47 +0000 (11:02 +0100)]
PCI/VPD: Remove pci_set_vpd_size()
3ea9a4374dc4 ("cxgb4: collect serial config version from register") removed
the only usage of pci_set_vpd_size(). If a device needs to override the
auto-detected VPD size, then this can be done with a PCI quirk, as is done
for Chelsio devices. There's no need to allow drivers to change the VPD
size.
Remove pci_set_vpd_size().
[bhelgaas: squash in Arnd's fix for "'pci_vpd_set_size' defined but not
used" from
https://lore.kernel.org/r/20210421140334.3847155-1-arnd@kernel.org] Link: https://lore.kernel.org/r/47d86e52-9bcf-7da7-1edb-0d988a7a82ab@gmail.com Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Krzysztof Wilczyński <kw@linux.com>
Jianjun Wang [Tue, 20 Apr 2021 06:17:22 +0000 (14:17 +0800)]
PCI: mediatek-gen3: Add system PM support
Add suspend_noirq and resume_noirq callback functions to implement PM
system suspend and resume hooks for the MediaTek Gen3 PCIe controller.
When the system suspends, trigger the PCIe link to enter the L2 state
and pull down the PERST# pin, gating the clocks of the MAC layer, and
then power-off the physical layer to provide power-saving.
When the system resumes, the PCIe link should be re-established and the
related control register values should be restored.
With 256 MSI vectors supported, the MSI vectors are composed of 8 sets,
each set has its own address for MSI message, and supports 32 MSI vectors
to generate interrupt.
Jianjun Wang [Tue, 20 Apr 2021 06:17:19 +0000 (14:17 +0800)]
PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192
MediaTek's PCIe host controller has three generation HWs, the new
generation HW is an individual bridge, it supports Gen3 speed and
compatible with Gen2, Gen1 speed.
Add support for new Gen3 controller which can be found on MT8192.
dw_pcie_ep_init() depends on the detected iATU region numbers to allocate
the in/outbound window management bitmap. It fails after 3d69e247270c
("PCI: dwc: Detect number of iATU windows").
Move the iATU region detection into a new function, move the detection to
the very beginning of dw_pcie_host_init() and dw_pcie_ep_init(). Also
remove it from the dw_pcie_setup(), since it's more like a software
initialization step than hardware setup.
Jisheng Zhang [Thu, 25 Mar 2021 07:26:04 +0000 (15:26 +0800)]
PCI: dwc: Move dw_pcie_msi_init() to dw_pcie_setup_rc()
If the host which makes use of IP's integrated MSI Receiver losts
power during suspend, we need to reinit the RC and MSI Receiver in
resume. But after we move dw_pcie_msi_init() into the core, we have no
API to do so. Usually the dwc users need to call dw_pcie_setup_rc() to
reinit the RC, we can solve this problem by moving dw_pcie_msi_init()
to dw_pcie_setup_rc().
PCI/sysfs: Use sysfs_emit() and sysfs_emit_at() in "show" functions
The sysfs_emit() and sysfs_emit_at() functions were introduced to make it
less ambiguous which function is preferred when writing to the output
buffer in a device attribute's "show" callback [1].
Convert the PCI sysfs object "show" functions from sprintf(), snprintf()
and scnprintf() to sysfs_emit() and sysfs_emit_at() accordingly, as the
latter is aware of the PAGE_SIZE buffer and correctly returns the number of
bytes written into the buffer.
PCI/sysfs: Convert "index", "acpi_index", "label" to static attributes
The "label", "index", and "acpi_index" sysfs attributes show firmware label
information about the device. If the ACPI Device Name _DSM is implemented
for the device, we have:
label Device name (optional, may be null)
acpi_index Instance number (unique under \_SB scope)
When there is no ACPI _DSM and SMBIOS provides an Onboard Devices structure
for the device, we have:
label Reference Designation, e.g., a silkscreen label
index Device Type Instance
Previously these attributes were dynamically created either by
pci_bus_add_device() or the pci_sysfs_init() initcall, but since they don't
need to be created or removed dynamically, we can use a static attribute so
the device model takes care of addition and removal automatically.
Convert "label", "index", and "acpi_index" to static attributes.
Presence of the ACPI _DSM (device_has_acpi_name()) determines whether the
ACPI information (label, acpi_index) or the SMBIOS information (label,
index) is visible.
[bhelgaas: commit log, split to separate patch, add "pci_dev_" prefix] Suggested-by: Oliver O'Halloran <oohall@gmail.com> Link: https://lore.kernel.org/r/20210416205856.3234481-6-kw@linux.com Signed-off-by: Krzysztof Wilczyński <kw@linux.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
PCI/sysfs: Define SMBIOS label attributes with DEVICE_ATTR*()
Use DEVICE_ATTR*() to simplify definition of the SMBIOS label attributes.
No functional change intended.
Note that dev_attr_smbios_label requires __ATTR() because the "label"
attribute can be exposed via either ACPI or SMBIOS, and we already have the
ACPI label_show() function in this file.
PCI/sysfs: Rename device_has_dsm() to device_has_acpi_name()
Rename device_has_dsm() to device_has_acpi_name() to better reflect its
purpose and move it earlier so it's available for a future SMBIOS
.is_visible() function. No functional change intended.
The "vpd" sysfs attribute allows access to Vital Product Data (VPD).
Previously it was dynamically created either by pci_bus_add_device() or the
pci_sysfs_init() initcall, but since it doesn't need to be created or
removed dynamically, we can use a static attribute so the device model
takes care of addition and removal automatically.
Convert "vpd" to a static attribute and use the .is_bin_visible() callback
to check whether the device supports VPD.
Remove pcie_vpd_create_sysfs_dev_files(),
pcie_vpd_remove_sysfs_dev_files(), pci_create_capabilities_sysfs(), and
pci_create_capabilities_sysfs(), which are no longer needed.
[bhelgaas: This is substantially the same as the earlier patch from Heiner
Kallweit <hkallweit1@gmail.com>. I included Krzysztof's change here so all
the "convert to static attribute" changes are together.]
[bhelgaas: rename to vpd_read()/vpd_write() and pci_dev_vpd_attr_group] Suggested-by: Oliver O'Halloran <oohall@gmail.com>
Based-on: https://lore.kernel.org/r/7703024f-8882-9eec-a122-599871728a89@gmail.com Based-on-patch-by: Heiner Kallweit <hkallweit1@gmail.com> Link: https://lore.kernel.org/r/20210416205856.3234481-5-kw@linux.com Signed-off-by: Krzysztof Wilczyński <kw@linux.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Rename "vpd" attribute accessors so they fit with the BIN_ATTR_RW() macro
usage. Currently there is no BIN_ATTR_ADMIN_RW() that uses 0600
permissions, but if there were, it would likely use "vpd_read()" and
"vpd_write()". No functional change intended.
Extracted from the patch mentioned below by Heiner Kallweit
<hkallweit1@gmail.com>.
The "reset" sysfs attribute allows for resetting a PCI function.
Previously it was dynamically created either by pci_bus_add_device() or
the pci_sysfs_init() initcall, but since it doesn't need to be created or
removed dynamically, we can use a static attribute so the device model
takes care of addition and removal automatically.
Convert "reset" to a static attribute and use the .is_visible() callback to
check whether the device supports reset.
Clear reset_fn in pci_stop_dev() instead of pci_remove_capabilities_sysfs()
since we no longer explicitly remove the "reset" sysfs file.
The "rom" sysfs attribute allows access to the PCI Option ROM. Previously
it was dynamically created either by pci_bus_add_device() or the
pci_sysfs_init() initcall, but since it doesn't need to be created or
removed dynamically, we can use a static attribute so the device model
takes care of addition and removal automatically.
Convert "rom" to a static attribute and use the .is_bin_visible() callback
to set the correct object size based on the ROM size.
Remove "rom_attr" from the struct pci_dev since it is no longer needed.
This attribute was added in the pre-git era by https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/drivers/pci/pci-sysfs.c?id=f6d553444da2
The "config" sysfs attribute allows access to either the legacy (PCI and
PCI-X Mode 1) or the extended (PCI-X Mode 2 and PCIe) device configuration
space. Previously it was dynamically created either when a device was
added (for hot-added devices) or via a late_initcall (for devices present
at boot):
pci_bus_add_devices
pci_bus_add_device
pci_create_sysfs_dev_files
if (!sysfs_initialized)
return
sysfs_create_bin_file # for hot-added devices
pci_sysfs_init # late_initcall
sysfs_initialized = 1
for_each_pci_dev(pdev)
pci_create_sysfs_dev_files(pdev) # for devices present at boot
And dynamically removed when the PCI device is stopped and removed:
This attribute does not need to be created or removed dynamically, so we
can use a static attribute so the device model takes care of addition and
removal automatically.
Convert "config" to a static attribute and use the .is_bin_visible()
callback to set the correct object size (either 256 bytes or 4 KiB) at
runtime.
The pci_sysfs_init() scheme was added in the pre-git era by
https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/drivers/pci/pci-sysfs.c?id=f6d553444da2
Qiheng Lin [Wed, 31 Mar 2021 08:59:38 +0000 (16:59 +0800)]
PCI: mediatek: Add missing MODULE_DEVICE_TABLE
This patch adds missing MODULE_DEVICE_TABLE definition which generates
correct modalias for automatic loading of this driver when it is built
as an external module.
Yang Yingliang [Wed, 31 Mar 2021 08:40:12 +0000 (16:40 +0800)]
PCI: endpoint: Fix missing destroy_workqueue()
Add the missing destroy_workqueue() before return from
pci_epf_test_init() in the error handling case and add
destroy_workqueue() in pci_epf_test_exit().
Link: https://lore.kernel.org/r/20210331084012.2091010-1-yangyingliang@huawei.com Fixes: dbbde090d453c ("PCI: endpoint: functions: Add an EP function to test PCI") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Colin Ian King [Fri, 26 Mar 2021 19:09:09 +0000 (19:09 +0000)]
PCI: endpoint: Remove redundant initialization of pointer dev
The pointer dev is being initialized with a value that is
never read and it is being updated later with a new value. The
initialization is redundant and can be removed.
Dinghao Liu [Thu, 8 Apr 2021 07:26:58 +0000 (15:26 +0800)]
PCI: tegra: Fix runtime PM imbalance in pex_ep_event_pex_rst_deassert()
pm_runtime_get_sync() will increase the runtime PM counter
even it returns an error. Thus a pairing decrement is needed
to prevent refcount leak. Fix this by replacing this API with
pm_runtime_resume_and_get(), which will not change the runtime
PM counter on error.
In enable_slot(), if pci_get_slot() returns NULL, we clear the SLOT_ENABLED
flag. When pci_get_slot() finds a device, it increments the device's
reference count. In this case, we did not call pci_dev_put() to decrement
the reference count, so the memory of the device (struct pci_dev type) will
eventually leak.
Call pci_dev_put() to decrement its reference count when pci_get_slot()
returns a PCI device.
Link: https://lore.kernel.org/r/b411af88-5049-a1c6-83ac-d104a1f429be@huawei.com Signed-off-by: Feilong Lin <linfeilong@huawei.com> Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If PCI_D3cold is passed to acpi_pci_set_power_state() as the second
argument and there is no ACPI D3cold support for the given device, the
debug message will state that the device power state has been changed to
D3cold, while in fact it will be D3hot, because acpi_device_set_power()
falls back to D3hot automatically if D3cold is not supported without
returning an error.
To address this issue, modify the debug message to print the current power
state of the target PCI device's ACPI companion instead of printing the
target power state, which may not reflect the real final power state of the
device.
Link: https://lore.kernel.org/r/4319486.LvFx2qVVIh@kreacher Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Krzysztof Wilczyński <kw@linux.com>
Heiner Kallweit [Tue, 30 Mar 2021 17:43:17 +0000 (12:43 -0500)]
IB/mthca: Disable parity reporting
For Mellanox Tavor devices, we previously set dev->broken_parity_status,
which does not change the device's behavior; it merely prevents the EDAC
PCI error reporting from warning about Master Data Parity Error, Signaled
System Error, or Detected Parity Error for this device.
Instead, disable Parity Error Response so the device doesn't report
parity errors in the first place.
Bjorn Helgaas [Tue, 30 Mar 2021 17:43:16 +0000 (12:43 -0500)]
PCI: Add pci_disable_parity()
Add pci_disable_parity() to disable reporting of parity errors for a
device by clearing PCI_COMMAND_PARITY.
The device will still set PCI_STATUS_DETECTED_PARITY when it detects
a parity error or receives a Poisoned TLP, but it will not set
PCI_STATUS_PARITY, which means it will not assert PERR#
(conventional PCI) or report Poisoned TLPs (PCIe).
Dejin Zheng [Sun, 28 Mar 2021 14:41:18 +0000 (22:41 +0800)]
PCI: xgene: Fix cfg resource mapping
In commit 69c2bd0dbc09 a change was made to use
devm_platform_ioremap_resource_byname() to simplify code and remove
the res variable; this was wrong since the res variable is still needed
and as an outcome the port->cfg_addr gets an erroneous address.
Revert the change going back to original behaviour.
Link: https://lore.kernel.org/r/20210328144118.305074-1-zhengdejin5@gmail.com Fixes: 69c2bd0dbc094 ("PCI: controller: Convert to devm_platform_ioremap_resource_byname()") Reported-by: dann.frazier@canonical.com Tested-by: dann frazier <dann.frazier@canonical.com> Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> Cc: stable@vger.kernel.org # v5.9+
Heiner Kallweit [Thu, 17 Dec 2020 20:59:03 +0000 (21:59 +0100)]
PCI/VPD: Remove obsolete Broadcom NIC quirk
quirk_brcm_570x_limit_vpd() was added in 2008 [0] when we didn't have the
logic to determine VPD size based on checking for the VPD end tag. Now
that we do have this logic [1] and don't read beyond the end tag, this
quirk can be removed.
Shradha Todi [Wed, 24 Mar 2021 10:16:09 +0000 (15:46 +0530)]
PCI: endpoint: Fix NULL pointer dereference for ->get_features()
get_features ops of pci_epc_ops may return NULL, causing NULL pointer
dereference in pci_epf_test_alloc_space function. Let us add a check for
pci_epc_feature pointer in pci_epf_test_bind before we access it to avoid
any such NULL pointer dereference and return -ENOTSUPP in case
pci_epc_feature is not found.
When the patch is not applied and EPC features is not implemented in the
platform driver, we see the following dump due to kernel NULL pointer
dereference.
Link: https://lore.kernel.org/r/20210324101609.79278-1-shradha.t@samsung.com Fixes: 1d23170c629c4 ("PCI: pci-epf-test: Use pci_epc_get_features() to get EPC features") Signed-off-by: Sriram Dash <dash.sriram@gmail.com> Signed-off-by: Shradha Todi <shradha.t@samsung.com> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> Reviewed-by: Pankaj Dubey <pankaj.dubey@samsung.com> Reviewed-by: Kishon Vijay Abraham I <kishon@ti.com>
Dmitry Baryshkov [Sun, 24 Jan 2021 23:28:26 +0000 (02:28 +0300)]
PCI: Release OF node in pci_scan_device()'s error path
In pci_scan_device(), if pci_setup_device() fails for any reason, the code
will not release device's of_node by calling pci_release_of_node(). Fix
that by calling the release function.
Fixes: e9743ceae099 ("pci/of: Match PCI devices to OF nodes dynamically") Link: https://lore.kernel.org/r/20210124232826.1879-1-dmitry.baryshkov@linaro.org Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
PCI: keystone: Let AM65 use the pci_ops defined in pcie-designware-host.c
Both TI's AM65x (K3) and TI's K2 PCIe driver are implemented in
pci-keystone. However Only K2 PCIe driver should use it's own pci_ops
for configuration space accesses. But commit 6513ff7e95c1
("PCI: dwc: keystone: Use pci_ops for config space accessors") used
custom pci_ops for both AM65x and K2. This breaks configuration space
access for AM65x platform. Fix it here.
Link: https://lore.kernel.org/r/20210317131518.11040-1-kishon@ti.com Fixes: 6513ff7e95c1 ("PCI: dwc: keystone: Use pci_ops for config space accessors") Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> Reviewed-by: Krzysztof Wilczyński <kw@linux.com> Cc: <stable@vger.kernel.org> # v5.10
PCI: microchip: Remove dev_err() when handing an error from platform_get_irq()
There is no need to call the dev_err() function directly to print a
custom message when handling an error from either the platform_get_irq()
or platform_get_irq_byname() functions as both are going to display an
appropriate error message in case of a failure.
This change is as per suggestions from Coccinelle, e.g.,
drivers/pci/controller/pcie-microchip-host.c:1027:2-9: line 1027 is
redundant because platform_get_irq() already prints an error
Related commit 9284bc49b6bd ("PCI: Remove dev_err() when handing an
error from platform_get_irq()").
Wei Yongjun [Mon, 8 Mar 2021 09:48:42 +0000 (09:48 +0000)]
PCI: microchip: Make some symbols static
The sparse tool complains as follows:
drivers/pci/controller/pcie-microchip-host.c:304:18: warning:
symbol 'pcie_event_to_event' was not declared. Should it be static?
drivers/pci/controller/pcie-microchip-host.c:310:18: warning:
symbol 'sec_error_to_event' was not declared. Should it be static?
drivers/pci/controller/pcie-microchip-host.c:317:18: warning:
symbol 'ded_error_to_event' was not declared. Should it be static?
drivers/pci/controller/pcie-microchip-host.c:324:18: warning:
symbol 'local_status_to_event' was not declared. Should it be static?
Those symbols are not used outside of pcie-microchip-host.c, so this
commit marks them static.
Jon Derrick [Wed, 10 Feb 2021 16:13:15 +0000 (09:13 -0700)]
PCI: vmd: Disable MSI-X remapping when possible
VMD will retransmit child device MSI-X using its own MSI-X table and
requester-id. This limits the number of MSI-X available to the whole
child device domain to the number of VMD MSI-X interrupts.
Some VMD devices have a mode where this remapping can be disabled,
allowing child device interrupts to bypass processing with the VMD MSI-X
domain interrupt handler and going straight the child device interrupt
handler, allowing for better performance and scaling. The requester-id
still gets changed to the VMD endpoint's requester-id, and the interrupt
remapping handlers have been updated to properly set IRTE for child
device interrupts to the VMD endpoint's context.
Some VMD platforms have existing production BIOS which rely on MSI-X
remapping and won't explicitly program the MSI-X remapping bit. This
re-enables MSI-X remapping on unload.
Jon Derrick [Wed, 10 Feb 2021 16:13:14 +0000 (09:13 -0700)]
iommu/vt-d: Use Real PCI DMA device for IRTE
VMD retransmits child device MSI-X with the VMD endpoint's requester-id.
In order to support direct interrupt remapping of VMD child devices,
ensure that the IRTE is programmed with the VMD endpoint's requester-id
using pci_real_dma_dev().
The only usage of them is to assign their address to the 'ops' field in
the pcie_port and the dw_pcie_ep structs, both which are pointers to
const. Make them const to allow the compiler to put them in read-only
memory.
Arnd Bergmann [Mon, 8 Mar 2021 15:24:48 +0000 (16:24 +0100)]
PCI: Avoid building empty drivers
There are harmless warnings when compile testing the kernel with
CONFIG_TRIM_UNUSED_KSYMS:
drivers/pci/controller/dwc/pcie-al.o: no symbols
drivers/pci/controller/pci-thunder-ecam.o: no symbols
drivers/pci/controller/pci-thunder-pem.o: no symbols
The problem here is that the host drivers get built even when the
configuration symbols are all disabled, as they pretend to not be drivers
but are silently enabled because of the promise that ACPI-based systems
need no drivers.
Add back the normal symbols to have these drivers built, and change the
logic to otherwise only build them when both CONFIG_PCI_QUIRKS and
CONFIG_ACPI are enabled.
As a side-effect, this enables compile-testing the drivers on other
architectures, which in turn needs the acpi_get_rc_resources() function to
be defined.
Arnd Bergmann [Mon, 8 Mar 2021 15:24:47 +0000 (16:24 +0100)]
PCI: thunder: Fix compile testing
Compile-testing these drivers is currently broken. Enabling it causes a
couple of build failures though:
drivers/pci/controller/pci-thunder-ecam.c:119:30: error: shift count >= width of type [-Werror,-Wshift-count-overflow]
drivers/pci/controller/pci-thunder-pem.c:54:2: error: implicit declaration of function 'writeq' [-Werror,-Wimplicit-function-declaration]
drivers/pci/controller/pci-thunder-pem.c:392:8: error: implicit declaration of function 'acpi_get_rc_resources' [-Werror,-Wimplicit-function-declaration]
Qiuxu Zhuo [Mon, 22 Feb 2021 01:17:17 +0000 (09:17 +0800)]
PCI/RCEC: Fix RCiEP device to RCEC association
rcec_assoc_rciep() used "rciep->devfn" (a single byte encoding both the
device and function number) as the device number to check whether the
corresponding bit was set in the RCEC's Association Bitmap for RCiEPs.
But per PCIe r5.0, sec 7.9.10.2, "Association Bitmap for RCiEPs", the
32-bit bitmap contains one bit per device. That bit applies to all
functions of the device.
Fix rcec_assoc_rciep() to convert the value of "rciep->devfn" to a device
number to ensure that RCiEP devices are correctly associated with the RCEC.
Reported-and-tested-by: Wen Jin <wen.jin@intel.com> Fixes: 7d7fb714f52e ("PCI/ERR: Add pcie_link_rcec() to associate RCiEPs") Link: https://lore.kernel.org/r/20210222011717.43266-1-qiuxu.zhuo@intel.com Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Sean V Kelley <sean.v.kelley@intel.com>
Linus Torvalds [Sat, 6 Mar 2021 01:27:59 +0000 (17:27 -0800)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Pull rdma fixes from Jason Gunthorpe:
"Nothing special here, though Bob's regression fixes for rxe would have
made it before the rc cycle had there not been such strong winter
weather!
- Fix corner cases in the rxe reference counting cleanup that are
causing regressions in blktests for SRP
- Two kdoc fixes so W=1 is clean
- Missing error return in error unwind for mlx5
- Wrong lock type nesting in IB CM"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
RDMA/rxe: Fix errant WARN_ONCE in rxe_completer()
RDMA/rxe: Fix extra deref in rxe_rcv_mcast_pkt()
RDMA/rxe: Fix missed IB reference counting in loopback
RDMA/uverbs: Fix kernel-doc warning of _uverbs_alloc
RDMA/mlx5: Set correct kernel-doc identifier
IB/mlx5: Add missing error code
RDMA/rxe: Fix missing kconfig dependency on CRYPTO
RDMA/cm: Fix IRQ restore in ib_send_cm_sidr_rep
projects / kernel.git / log