- Delay PHY initialization to make boots reliable for ZBT WE1326 and ZBT
WF3526-P and some Netgear models (Sergio Paracuellos)
* pci/controller/mt7621:
PCI: mt7621: Delay phy ports initialization
- Release previously-requested DW eDMA IRQs if request_irq() fails (Serge
Semin)
- Convert DW eDMA linked-list (ll) and data target (dt) from CPU-relative
addresses to PCI bus addresses (Serge Semin)
- Fix missing src/dst address for interleaved transfers (Serge Semin)
- Enforce the DW eDMA restriction that interleaved transfers must increment
src and dst addresses (Serge Semin)
- Fix some invalid interleaved transfer semantics (Serge Semin)
- Convert CPU-relative addresses to PCI bus addresses for eDMA engine
(Serge Semin)
- Drop chancnt initialization from dw-edma-core, since it is managed by the
dmaengine core, e.g., in dma_async_device_channel_register() (Serge Semin)
- Clean up bogus casting of debugfs_entries.reg addresses (Serge Semin)
- Ignore debugfs file/directory creation errors (Serge Semin)
- Allocate debugfs entries from the heap to prepare for multi-eDMA
platforms (Serge Semin)
- Simplify and rework register accessors to remove another obstacle to
multi-eDMA platforms (Serge Semin)
- Consolidate eDMA read/write channels in a single dma_device to simplify,
better reflect the hardware design, and avoid a debugfs complaint (Serge
Semin)
- Move eDMA-specific debugfs nodes into existing dmaengine subdirectory
(Serge Semin)
- Fix a readq_ch() truncation from 64 to 32 bits (Serge Semin)
- Use existing readq()/writeq rather than hand-coding new ones (Serge
Semin)
- Drop unnecessary data target region allocation in favor of existing
dw_edma_chip members (Serge Semin)
- Use parent device in eDMA controller name to prepare for multi-eDMA
platforms (Serge Semin)
- In addition to the existing MMIO accessors for linked list entries, add
support for ioremapped entries for use by eDMA in Root Ports or local
Endpoints (Serge Semin)
- Convert DW_EDMA_PCIE so it depends on DW_EDMA instead of selecting it
(Serge Semin)
- Allow DWC drivers to set streaming DMA masks larger than 32 bits;
previously both streaming and coherent DMA were limited to 32 bits
because some PCI devices only support coherent 32-bit DMA for MSI (Serge
Semin)
- Set 64-bit streaming and coherent DMA mask for the bt1 driver (Serge
Semin)
- Add DW Root Port and Endpoint controller support for eDMA (Serge Semin)
* pci/controller/dwc:
PCI: dwc: Add Root Port and Endpoint controller eDMA engine support
PCI: bt1: Set 64-bit DMA mask
PCI: dwc: Restrict only coherent DMA mask for MSI address allocation
dmaengine: dw-edma: Prepare dw_edma_probe() for builtin callers
dmaengine: dw-edma: Depend on DW_EDMA instead of selecting it
dmaengine: dw-edma: Add mem-mapped LL-entries support
dmaengine: dw-edma: Skip cleanup procedure if no private data found
dmaengine: dw-edma: Replace chip ID number with device name
dmaengine: dw-edma: Drop DT-region allocation
dmaengine: dw-edma: Use non-atomic io-64 methods
dmaengine: dw-edma: Fix readq_ch() return value truncation
dmaengine: dw-edma: Use DMA engine device debugfs subdirectory
dmaengine: dw-edma: Join read/write channels into a single device
dmaengine: dw-edma: Move eDMA data pointer to debugfs node descriptor
dmaengine: dw-edma: Simplify debugfs context CSRs init procedure
dmaengine: dw-edma: Rename debugfs dentry variables to 'dent'
dmaengine: dw-edma: Convert debugfs descs to being heap-allocated
dmaengine: dw-edma: Add dw_edma prefix to debugfs nodes descriptor
dmaengine: dw-edma: Stop checking debugfs_create_*() return value
dmaengine: dw-edma: Drop unnecessary debugfs reg casts
dmaengine: dw-edma: Drop chancnt initialization
dmaengine: dw-edma: Add PCI bus address getter to the remote EP glue driver
dmaengine: dw-edma: Add CPU to PCI bus address translation
dmaengine: dw-edma: Fix invalid interleaved xfers semantics
dmaengine: dw-edma: Don't permit non-inc interleaved xfers
dmaengine: dw-edma: Fix missing src/dst address of interleaved xfers
dmaengine: dw-edma: Convert ll/dt phys address to PCI bus/DMA address
dmaengine: dw-edma: Release requested IRQs on failure
dmaengine: Fix dma_slave_config.dst_addr description
- Convert dra7xx to threaded IRQ handler (Manivannan Sadhasivam)
- Move tegra194 dw_pcie_ep_linkup() to threaded IRQ handler (Manivannan
Sadhasivam)
- Add a separate lock for the endpoint pci_epf list to avoid deadlock
while running callbacks (Manivannan Sadhasivam)
- Use callbacks instead of notifier chains to signal events from EPC to EPF
drivers (Manivannan Sadhasivam)
- Use link_up() callback in place of LINK_UP notifier (Manivannan
Sadhasivam)
* pci/endpoint:
PCI: endpoint: Use link_up() callback in place of LINK_UP notifier
PCI: endpoint: Use callback mechanism for passing events from EPC to EPF
PCI: endpoint: Use a separate lock for protecting epc->pci_epf list
PCI: tegra194: Move dw_pcie_ep_linkup() to threaded IRQ handler
PCI: dra7xx: Use threaded IRQ handler for "dra7xx-pcie-main" IRQ
- Avoid FLR for AMD FCH AHCI adapters to avoid a hardware defect (Damien Le
Moal)
- Add ACS quirk for Wangxun NICs that don't allow peer-to-peer between
functions, but don't advertise an ACS Capability (Mengyuan Lou)
* pci/virtualization:
PCI: Add ACS quirk for Wangxun NICs
PCI: Avoid FLR for AMD FCH AHCI adapters
- Realign space as required by bridge windows after dividing it up (Mika
Westerberg)
- Account for space required by other devices on the bus before
distributing it all to bridges (Mika Westerberg)
- Distribute spare resources to root bus devices as well as to other
hotplug bridges (Mika Westerberg)
- Fix bug that dropped root bus resources that end at zero, e.g., a host
bridge that leads only to bus 00 (Geert Uytterhoeven)
* pci/resource:
PCI: Fix dropping valid root bus resources with .end = zero
PCI: Distribute available resources for root buses, too
PCI: Take other bus devices into account when distributing resources
PCI: Align extra resources for hotplug bridges properly
- Always observe reset delay when waking devices from D3cold, e.g., after
system sleep, regardless of whether we're allowed to runtime-suspend to
D3cold (Lukas Wunner)
- Unify reset and resume delays to wait for downstream devices after a
bridge reset (Lukas Wunner)
- Wait for downstream devices after a DPC-induced bridge reset (Lukas
Wunner)
* pci/reset:
PCI/DPC: Await readiness of secondary bus after reset
PCI: Unify delay handling for reset and resume
PCI/PM: Observe reset delay irrespective of bridge_d3
- Account for _S0W when deciding whether to put bridges in D3 to avoid
missing hotplug events (Rafael J. Wysocki)
* pci/pm:
PCI/ACPI: Account for _S0W of the target bridge in acpi_pci_bridge_d3()
- Remove MODULE_LICENSE from boolean drivers so they don't look like
modules so modprobe will complain about them (Nick Alcock)
* pci/kbuild:
PCI: Remove MODULE_LICENSE so boolean drivers don't look like modules
- Add quirk to work around Qualcomm hardware defect in Command Completed
signaling (Manivannan Sadhasivam)
- Remove locking to allow devices to be marked as disconnected immediately
instead of waiting for concurrent bind/unbind to complete (Lukas Wunner)
* pci/hotplug:
PCI: hotplug: Allow marking devices as disconnected during bind/unbind
PCI: pciehp: Add Qualcomm quirk for Command Completed erratum
- Implement portdrv .shutdown() method that calls service driver .remove()
methods (which disables interrupt generation as required by .shutdown()),
but doesn't disable bus mastering (which hangs on Loongson LS7A because
of a hardware defect) (Huacai Chen)
- Prevent MRRS increases for devices below Loongson LS7A to avoid hardware
limitations (Huacai Chen)
- Ignore devices with a firmware (DT/ACPI) node that says the device is
disabled (Rob Herring)
* pci/enumeration:
PCI: Honor firmware's device disabled status
PCI: loongson: Add more devices that need MRRS quirk
PCI: loongson: Prevent LS7A MRRS increases
PCI/portdrv: Prevent LS7A Bus Master clearing on shutdown
- Configure ECRC only if AER is native (Vidya Sagar)
- Stop enabling device error reporting for the downstream hierarchy when
the AER service driver probes a Root Port because we've already done that
when enumerating those downstream devices (Bjorn Helgaas)
* pci/aer:
PCI/AER: Remove redundant Device Control Error Reporting Enable
PCI/AER: Configure ECRC only if AER is native
Since the DW eDMA core now supports eDMA controllers embedded in locally
accessible DW PCIe Root Ports and Endpoints, register these controllers
when possible.
To do that the DW PCIe core driver needs to perform some preparations
first. First of all, it needs to find the eDMA controller CSRs base
address, whether they are accessible over the Port Logic or iATU unrolled
space. Afterwards it can try to auto-detect the eDMA controller
availability and number of read/write channels. If none are found the
procedure silently returns without error.
Secondly, the platform is supposed to provide either combined or
per-channel IRQ signals. If no valid IRQs set is found, the procedure
returns without error to be backward compatible with platforms where DW
PCIe controllers have eDMA but lack the IRQ description.
Finally, before actually probing the eDMA device we need to allocate LLP
items buffers. After that the DW eDMA can be registered. If registration is
successful, a message regarding the number of detected Read/Write eDMA
channels will be printed to the system as is done for the iATU settings.
Note: the DW PCI controller driver (either host or endpoint mode) is
currently always built-in, so if the DW eDMA core is built as a module
(CONFIG_DW_EDMA=m), eDMA controllers will not be registered even if the
dw-edma module is later loaded.
Link: https://lore.kernel.org/r/20230113171409.30470-28-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
The DW PCIe Root Port IP core is synthesized with the 64-bit AXI address
bus. Since the device is also equipped with the eDMA engine, explicitly
set the device DMA mask so DMA engine clients can allocate data buffers
anywhere in the 64-bit memory space.
Link: https://lore.kernel.org/r/20230113171409.30470-27-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The MSI target address must be in the lowest 4GB memory to support PCI
peripherals without 64-bit MSI support. Since the allocation is done from
DMA coherent memory, set only the coherent DMA mask, leaving the streaming
DMA mask alone.
Thus streaming DMA operations will work with no artificial limitations. It
will be specifically useful for the eDMA-capable controllers so the
corresponding DMA engine clients would map the DMA buffers with no need for
SWIOTLB for buffers allocated above 4GB.
Add a brief comment about the reason allocating the MSI target address
below 4GB.
Link: https://lore.kernel.org/r/20230113171409.30470-26-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
When CONFIG_DW_EDMA=m, dw_edma_probe() is built as a module. Previously
edma.h declared it as extern, but the implementation isn't available for
builtin callers. A subsequent commit will add calls from
dw_pcie_host_init() and dw_pcie_ep_init(), which can only be built-in.
Make it safe for such builtin callers to call dw_edma_probe() by using
IS_REACHABLE() to define a stub when CONFIG_DW_EDMA=m.
When CONFIG_DW_EDMA=m, these builtin callers will fail to detect and
register eDMA devices, so eDMA won't be usable even if the dw-edma module
is loaded.
[bhelgaas: split to separate patch, commit log]
Link: https://lore.kernel.org/r/20230113171409.30470-25-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Vinod Koul <vkoul@kernel.org>
Kconfig "select" is discouraged for visible symbols like DW_EDMA because it
makes it possible to set DW_EDMA even if DW_EDMA depends on things that are
not set (see Documentation/kbuild/kconfig-language.txt).
Convert DW_EDMA_PCIE so it depends on DW_EDMA instead of selecting it.
There will likely be several future drivers that depend on DW_EDMA, so this
uses "if DW_EDMA" to enclose them all rather than repeating "depends on
DW_EDMA" for each.
[bhelgaas: split to separate patch, commit log]
Link: https://lore.kernel.org/r/20230113171409.30470-25-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Vinod Koul <vkoul@kernel.org>
Currently the DW eDMA driver only supports the linked lists memory
allocated locally with respect to the remote eDMA engine setup. It means
the linked lists will be accessible by the CPU via the MMIO space only. If
eDMA is embedded into the DW PCIe Root Ports or local Endpoints (which
support will be added in subsequent commits) the linked lists are supposed
to be allocated in the CPU memory. In that case the LL-entries can be
directly accessed, while the former case implies using the MMIO accessors
for that.
In order to have both cases supported by the driver, the dw_edma_region
descriptor should be fixed to contain the MMIO-backed and just memory-based
virtual addresses. The linked lists initialization procedure will use one
of them depending on the eDMA device nature. If the eDMA engine is embedded
into the local DW PCIe Root Port/Endpoint controllers, the list entries
will be directly accessed by referencing the corresponding structure
fields. Otherwise the MMIO accessors usage will be preserved.
Link: https://lore.kernel.org/r/20230113171409.30470-24-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Vinod Koul <vkoul@kernel.org>
Since 8b41fc4454 ("kbuild: create modules.builtin without
Makefile.modbuiltin or tristate.conf"), MODULE_LICENSE declarations are
used to identify modules. As a consequence, MODULE_LICENSE in non-modules
causes modprobe to misidentify the object file as a module when it is not,
and modprobe might succeed rather than failing with a suitable error
message.
For tristate modules that can be either built-in or loaded at runtime,
modprobe succeeds in both cases:
# modprobe ext4
[exit status zero if CONFIG_EXT4_FS=y or =m]
For boolean modules like the Standard Hot Plug Controller driver (shpchp)
that cannot be loaded at runtime, modprobe should always fail like this:
# modprobe shpchp
modprobe: FATAL: Module shpchp not found in directory /lib/modules/...
[exit status non-zero regardless of CONFIG_HOTPLUG_PCI_SHPC]
but prior to this commit, shpchp_core.c contained MODULE_LICENSE, so
"modprobe shpchp" silently succeeded when it should have failed.
Remove MODULE_LICENSE in files that cannot be built as modules.
[bhelgaas: commit log, squash]
Suggested-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230216152410.4312-1-nick.alcock@oracle.com/
Signed-off-by: Nick Alcock <nick.alcock@oracle.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Cc: Luis Chamberlain <mcgrof@kernel.org>
Cc: Hitomi Hasegawa <hasegawa-hitomi@fujitsu.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Lorenzo Pieralisi <lpieralisi@kernel.org>
A dereference of the __rcu pointer was noticed by sparse:
drivers/pci/p2pdma.c:199:44: sparse: sparse: dereference of noderef expression
Dereference the __rcu pointer using rcu_dereference_protected() instead of
accessing it directly. It's safe to use rcu_dereference_protected() because
a reference is held on the pgmap's percpu reference counter and thus it
cannot disappear.
Link: https://lore.kernel.org/r/20230209172953.4597-1-logang@deltatee.com
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
On surprise removal, pciehp_unconfigure_device() and acpiphp's
trim_stale_devices() call pci_dev_set_disconnected() to mark removed
devices as permanently offline. Thereby, the PCI core and drivers know
to skip device accesses.
However pci_dev_set_disconnected() takes the device_lock and thus waits for
a concurrent driver bind or unbind to complete. As a result, the driver's
->probe and ->remove hooks have no chance to learn that the device is gone.
That doesn't make any sense, so drop the device_lock and instead use atomic
xchg() and cmpxchg() operations to update the device state.
As a byproduct, an AB-BA deadlock reported by Anatoli is fixed which occurs
on surprise removal with AER concurrently performing a bus reset.
AER bus reset:
INFO: task irq/26-aerdrv:95 blocked for more than 120 seconds.
Tainted: G W 6.2.0-rc3-custom-norework-jan11+
schedule
rwsem_down_write_slowpath
down_write_nested
pciehp_reset_slot # acquires reset_lock
pci_reset_hotplug_slot
pci_slot_reset # acquires device_lock
pci_bus_error_reset
aer_root_reset
pcie_do_recovery
aer_process_err_devices
aer_isr
pciehp surprise removal:
INFO: task irq/26-pciehp:96 blocked for more than 120 seconds.
Tainted: G W 6.2.0-rc3-custom-norework-jan11+
schedule_preempt_disabled
__mutex_lock
mutex_lock_nested
pci_dev_set_disconnected # acquires device_lock
pci_walk_bus
pciehp_unconfigure_device
pciehp_disable_slot
pciehp_handle_presence_or_link_change
pciehp_ist # acquires reset_lock
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215590
Fixes: a6bd101b8f ("PCI: Unify device inaccessible")
Link: https://lore.kernel.org/r/3dc88ea82bdc0e37d9000e413d5ebce481cbd629.1674205689.git.lukas@wunner.de
Reported-by: Anatoli Antonovitch <anatoli.antonovitch@amd.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org # v4.20+
Cc: Keith Busch <kbusch@kernel.org>
The Qualcomm PCI bridge device (Device ID 0x010e) found in chipsets such as
SC8280XP used in Lenovo Thinkpad X13s, does not set the Command Completed
bit unless writes to the Slot Command register change "Control" bits.
This results in timeouts like below during boot and resume from suspend:
pcieport 0002:00:00.0: pciehp: Timeout on hotplug command 0x03c0 (issued 2020 msec ago)
...
pcieport 0002:00:00.0: pciehp: Timeout on hotplug command 0x13f1 (issued 107724 msec ago)
Add the device to the Command Completed quirk to mark commands "completed"
immediately unless they change the "Control" bits.
Link: https://lore.kernel.org/r/20230213144922.89982-1-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Wangxun has verified there is no peer-to-peer between functions for the
below selection of SFxxx, RP1000 and RP2000 NICS. They may be
multi-function devices, but the hardware does not advertise ACS capability.
Add an ACS quirk for these devices so the functions can be in independent
IOMMU groups.
Link: https://lore.kernel.org/r/20230207102419.44326-1-mengyuanlou@net-swift.com
Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
On r8a7791/koelsch:
kmemleak: 1 new suspected memory leaks (see /sys/kernel/debug/kmemleak)
# cat /sys/kernel/debug/kmemleak
unreferenced object 0xc3a34e00 (size 64):
comm "swapper/0", pid 1, jiffies 4294937460 (age 199.080s)
hex dump (first 32 bytes):
b4 5d 81 f0 b4 5d 81 f0 c0 b0 a2 c3 00 00 00 00 .]...]..........
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<fe3aa979>] __kmalloc+0xf0/0x140
[<34bd6bc0>] resource_list_create_entry+0x18/0x38
[<767046bc>] pci_add_resource_offset+0x20/0x68
[<b3f3edf2>] devm_of_pci_get_host_bridge_resources.constprop.0+0xb0/0x390
When coalescing two resources for a contiguous aperture, the second
resource is enlarged to cover the full contiguous range, while the first
resource is marked invalid. This invalidation is done by clearing the
flags, start, and end members.
When adding the initial resources to the bus later, invalid resources are
skipped. Unfortunately, the check for an invalid resource considers only
the end member, causing false positives.
E.g. on r8a7791/koelsch, root bus resource 0 ("bus 00") is skipped, and no
longer registered with pci_bus_insert_busn_res() (causing the memory leak),
nor printed:
pci-rcar-gen2 ee090000.pci: host bridge /soc/pci@ee090000 ranges:
pci-rcar-gen2 ee090000.pci: MEM 0x00ee080000..0x00ee08ffff -> 0x00ee080000
pci-rcar-gen2 ee090000.pci: PCI: revision 11
pci-rcar-gen2 ee090000.pci: PCI host bridge to bus 0000:00
-pci_bus 0000:00: root bus resource [bus 00]
pci_bus 0000:00: root bus resource [mem 0xee080000-0xee08ffff]
Fix this by only skipping resources where all of the flags, start, and end
members are zero.
Fixes: 7c3855c423 ("PCI: Coalesce host bridge contiguous apertures")
Link: https://lore.kernel.org/r/da0fcd5e86c74239be79c7cb03651c0fce31b515.1676036673.git.geert+renesas@glider.be
Tested-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
As a part of the transition towards callback mechanism for signalling the
events from EPC to EPF, let's use the link_up() callback in the place of
the LINK_UP notifier. This also removes the notifier support completely
from the PCI endpoint framework.
Link: https://lore.kernel.org/linux-pci/20230124071158.5503-6-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Acked-by: Kishon Vijay Abraham I <kishon@kernel.org>
Instead of using the notifiers for passing the events from EPC to EPF,
let's introduce a callback based mechanism where the EPF drivers can
populate relevant callbacks for EPC events they want to subscribe.
The use of notifiers in kernel is not recommended if there is a real link
between the sender and receiver, like in this case. Also, the existing
atomic notifier forces the notification functions to be in atomic context
while the caller may be in non-atomic context. For instance, the two
in-kernel users of the notifiers, pcie-qcom and pcie-tegra194, both are
calling the notifier functions in non-atomic context (from threaded IRQ
handlers). This creates a sleeping in atomic context issue with the
existing EPF_TEST driver that calls the EPC APIs that may sleep.
For all these reasons, let's get rid of the notifier chains and use the
simple callback mechanism for signalling the events from EPC to EPF
drivers. This preserves the context of the caller and avoids the latency
of going through a separate interface for triggering the notifications.
As a first step of the transition, the core_init() callback is introduced
in this commit, that'll replace the existing CORE_INIT notifier used for
signalling the init complete event from EPC.
During the occurrence of the event, EPC will go over the list of EPF
drivers attached to it and will call the core_init() callback if available.
Link: https://lore.kernel.org/linux-pci/20230124071158.5503-5-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Acked-by: Kishon Vijay Abraham I <kishon@kernel.org>
The EPC controller maintains a list of EPF drivers added to it. For
protecting this list against the concurrent accesses, the epc->lock
(used for protecting epc_ops) has been used so far. Since there were
no users trying to use epc_ops and modify the pci_epf list simultaneously,
this was not an issue.
But with the addition of callback mechanism for passing the events, this
will be a problem. Because the pci_epf list needs to be iterated first
for getting hold of the EPF driver and then the relevant event specific
callback needs to be called for the driver.
If the same epc->lock is used, then it will result in a deadlock scenario.
For instance,
...
mutex_lock(&epc->lock);
list_for_each_entry(epf, &epc->pci_epf, list) {
epf->event_ops->core_init(epf);
|
|-> pci_epc_set_bar();
|
|-> mutex_lock(&epc->lock) # DEADLOCK
...
So to fix this issue, use a separate lock called "list_lock" for
protecting the pci_epf list against the concurrent accesses. This lock
will also be used by the callback mechanism.
Link: https://lore.kernel.org/linux-pci/20230124071158.5503-4-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
dw_pcie_ep_linkup() may take more time to execute depending on the EPF
driver implementation. Calling this API in the hard IRQ handler is not
encouraged since the hard IRQ handlers are supposed to complete quickly.
So move the dw_pcie_ep_linkup() call to threaded IRQ handler.
Link: https://lore.kernel.org/linux-pci/20230124071158.5503-3-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
The "dra7xx-pcie-main" hard IRQ handler is just printing the IRQ status
and calling the dw_pcie_ep_linkup() API if LINK_UP status is set. But the
execution of dw_pcie_ep_linkup() depends on the EPF driver and may take
more time depending on the EPF implementation.
In general, hard IRQ handlers are supposed to return quickly and not block
for so long. Moreover, there is no real need of the current IRQ handler to
be a hard IRQ handler. So switch to the threaded IRQ handler for the
"dra7xx-pcie-main" IRQ.
Link: https://lore.kernel.org/linux-pci/20230124071158.5503-2-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
If a device has a firmware node (DT/ACPI), and the device is marked
disabled, that is currently ignored. Add a check for this condition and
bail out creating the pci_dev.
This assumes the config space for the device can still be accessed because
they already have by this point in order to identify the device.
Link: https://lore.kernel.org/r/20230210164351.2687475-1-robh@kernel.org
Tested-by: Binbin Zhou <zhoubinbin@loongson.cn>
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Liu Peibao <liupeibao@loongson.cn>
Cc: Huacai Chen <chenhuacai@loongson.cn>
DW eDMA driver private data is preserved in the passed DW eDMA chip info
structure. If the probe fails or for some reason the passed info object
doesn't have the private data pointer initialized, halt the DMA device
cleanup procedure to prevent system crashes.
Link: https://lore.kernel.org/r/20230113171409.30470-23-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
Using an abstract number as the DW eDMA chip identifier isn't practical
because there can be more than one DW eDMA controller on the platform. Some
may be detected as the PCIe Endpoints, and others may be embedded in DW
PCIe Root Port/Endpoint controllers. An abstract number in, for instance,
the IRQ handlers list, doesn't give a notion regarding their reference to
the particular DMA controller.
To preserve the code simplicity and support multi-eDMA platforms, use the
parental device name to create the DW eDMA controller name.
Link: https://lore.kernel.org/r/20230113171409.30470-22-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
There is no point in allocating additional memory for the data target
regions passed to the client drivers. Use the already available structures
defined in the dw_edma_chip instance.
Note: these regions are unused in normal circumstances since they are
specific to the case of eDMA being embedded into the DW PCIe Endpoint and
having its CSRs accessible via an Endpoint BAR. This case is only known to
be implemented as a part of the Synopsys PCIe Endpoint IP prototype kit.
Link: https://lore.kernel.org/r/20230113171409.30470-21-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
Instead of splitting 64-bits IOs up into two 32-bits ones, use the existing
non-atomic readq()/writeq() functions. By doing so we can discard
CONFIG_64BIT #ifdefs from the code.
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>
Acked-by: Vinod Koul <vkoul@kernel.org>
Previously, readq_ch() did a 64-bit readq(), but truncated the result by
storing it in the u32 "value". Change "value" to u64 to avoid the
truncation.
Note: the method is currently unused, so the bug hasn't caused any problem
so far.
Fixes: 04e0a39fc1 ("dmaengine: dw-edma: Add writeq() and readq() for 64 bits architectures")
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Since all DW eDMA read and write channels are now installed in a framework
of a single DMA engine device, move all the DW eDMA-specific debugfs nodes
into a ready-to-use DMA-engine debugfs subdirectory. It's created during
the DMA-device registration and can be found in the dma_device.dbg_dev_root
field.
Link: https://lore.kernel.org/r/20230113171409.30470-19-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
There is no point in splitting read/write channels. First of all, eDMA
read and write channels belong to one physical controller. Secondly,
channel differentiation can be done by filtering and dma_get_slave_caps().
Finally, having these channels handled separately needlessly complicates
the code and causes this debugfs warning:
debugfs: Directory '1f052000.pcie' with parent 'dmaengine' already present!
Join the read/write channels into a single DMA device. Client drivers can
choose the correct channel via the DMA slave direction setting. The default
value is overridden by the dw_edma_device_caps() callback in accordance
with the channel type.
Link: https://lore.kernel.org/r/20230113171409.30470-18-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
The last thing that stops the debugfs part of the eDMA driver from
supporting multi-eDMA platforms is keeping the eDMA private data pointer in
the static area of the debugfs module. Since the debugfs node descriptors
are now heap-allocated, we can freely move that pointer to being preserved
in the descriptors. After the debugfs initialization procedure, that
pointer will be used in the debugfs files getter to access the common CSRs
space and the context CSRs spinlock. So the main part of this change is
connected with the debugfs nodes descriptors initialization macros, which
aside with already defined prototypes now require to have the DW eDMA
private data pointer passed.
[bhelgaas: squash in https://lore.kernel.org/r/20230130185101.2883245-1-arnd@kernel.org]
Link: https://lore.kernel.org/r/20230113171409.30470-17-Sergey.Semin@baikalelectronics.ru
Tested-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
pci_bridge_wait_for_secondary_bus() is called after a Secondary Bus
Reset, but not after a DPC-induced Hot Reset.
As a result, the delays prescribed by PCIe r6.0 sec 6.6.1 are not
observed and devices on the secondary bus may be accessed before
they're ready.
One affected device is Intel's Ponte Vecchio HPC GPU. It comprises a
PCIe switch whose upstream port is not immediately ready after reset.
Because its config space is restored too early, it remains in
D0uninitialized, its subordinate devices remain inaccessible and DPC
recovery fails with messages such as:
i915 0000:8c:00.0: can't change power state from D3cold to D0 (config space inaccessible)
intel_vsec 0000:8e:00.1: can't change power state from D3cold to D0 (config space inaccessible)
pcieport 0000:89:02.0: AER: device recovery failed
Fix it.
Link: https://lore.kernel.org/r/9f5ff00e1593d8d9a4b452398b98aa14d23fca11.1673769517.git.lukas@wunner.de
Tested-by: Ravi Kishore Koppuravuri <ravi.kishore.koppuravuri@intel.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: stable@vger.kernel.org
Sheng Bi reports that pci_bridge_secondary_bus_reset() may fail to wait
for devices on the secondary bus to become accessible after reset:
Although it does call pci_dev_wait(), it erroneously passes the bridge's
pci_dev rather than that of a child. The bridge of course is always
accessible while its secondary bus is reset, so pci_dev_wait() returns
immediately.
Sheng Bi proposes introducing a new pci_bridge_secondary_bus_wait()
function which is called from pci_bridge_secondary_bus_reset():
https://lore.kernel.org/linux-pci/20220523171517.32407-1-windy.bi.enflame@gmail.com/
However we already have pci_bridge_wait_for_secondary_bus() which does
almost exactly what we need. So far it's only called on resume from
D3cold (which implies a Fundamental Reset per PCIe r6.0 sec 5.8).
Re-using it for Secondary Bus Resets is a leaner and more rational
approach than introducing a new function.
That only requires a few minor tweaks:
- Amend pci_bridge_wait_for_secondary_bus() to await accessibility of
the first device on the secondary bus by calling pci_dev_wait() after
performing the prescribed delays. pci_dev_wait() needs two parameters,
a reset reason and a timeout, which callers must now pass to
pci_bridge_wait_for_secondary_bus(). The timeout is 1 sec for resume
(PCIe r6.0 sec 6.6.1) and 60 sec for reset (commit 821cdad5c4 ("PCI:
Wait up to 60 seconds for device to become ready after FLR")).
Introduce a PCI_RESET_WAIT macro for the 1 sec timeout.
- Amend pci_bridge_wait_for_secondary_bus() to return 0 on success or
-ENOTTY on error for consumption by pci_bridge_secondary_bus_reset().
- Drop an unnecessary 1 sec delay from pci_reset_secondary_bus() which
is now performed by pci_bridge_wait_for_secondary_bus(). A static
delay this long is only necessary for Conventional PCI, so modern
PCIe systems benefit from shorter reset times as a side effect.
Fixes: 6b2f1351af ("PCI: Wait for device to become ready after secondary bus reset")
Link: https://lore.kernel.org/r/da77c92796b99ec568bd070cbe4725074a117038.1673769517.git.lukas@wunner.de
Reported-by: Sheng Bi <windy.bi.enflame@gmail.com>
Tested-by: Ravi Kishore Koppuravuri <ravi.kishore.koppuravuri@intel.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Cc: stable@vger.kernel.org # v4.17+
If a PCI bridge is suspended to D3cold upon entering system sleep,
resuming it entails a Fundamental Reset per PCIe r6.0 sec 5.8.
The delay prescribed after a Fundamental Reset in PCIe r6.0 sec 6.6.1
is sought to be observed by:
pci_pm_resume_noirq()
pci_pm_bridge_power_up_actions()
pci_bridge_wait_for_secondary_bus()
However, pci_bridge_wait_for_secondary_bus() bails out if the bridge_d3
flag is not set. That flag indicates whether a bridge is allowed to
suspend to D3cold at *runtime*.
Hence *no* delay is observed on resume from system sleep if runtime
D3cold is forbidden. That doesn't make any sense, so drop the bridge_d3
check from pci_bridge_wait_for_secondary_bus().
The purpose of the bridge_d3 check was probably to avoid delays if a
bridge remained in D0 during suspend. However the sole caller of
pci_bridge_wait_for_secondary_bus(), pci_pm_bridge_power_up_actions(),
is only invoked if the previous power state was D3cold. Hence the
additional bridge_d3 check seems superfluous.
Fixes: ad9001f2f4 ("PCI/PM: Add missing link delays required by the PCIe spec")
Link: https://lore.kernel.org/r/eb37fa345285ec8bacabbf06b020b803f77bdd3d.1673769517.git.lukas@wunner.de
Tested-by: Ravi Kishore Koppuravuri <ravi.kishore.koppuravuri@intel.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Cc: stable@vger.kernel.org # v5.5+
Previously we distributed spare resources only upon hot-add, so if the
initial root bus scan found devices that had not been fully configured by
the BIOS, we allocated only enough resources to cover what was then
present. If some of those devices were hotplug bridges, we did not leave
any additional resource space for future expansion.
Distribute the available resources for root buses, too, to make this work
the same way as the normal hotplug case.
A previous commit to do this was reverted due to a regression reported by
Jonathan Cameron:
e96e27fc6f ("PCI: Distribute available resources for root buses, too")
5632e2beaf ("Revert "PCI: Distribute available resources for root buses, too"")
This commit changes pci_bridge_resources_not_assigned() to work with
bridges that do not have all the resource windows programmed by the boot
firmware (previously we expected all I/O, memory and prefetchable memory
were programmed).
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216000
Link: https://lore.kernel.org/r/20220905080232.36087-5-mika.westerberg@linux.intel.com
Link: https://lore.kernel.org/r/20230131092405.29121-4-mika.westerberg@linux.intel.com
Reported-by: Chris Chiu <chris.chiu@canonical.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
A PCI bridge may reside on a bus with other devices as well. The resource
distribution code does not take this into account and therefore it expands
the bridge resource windows too much, not leaving space for the other
devices (or functions of a multifunction device). This leads to an issue
that Jonathan reported when running QEMU with the following topology (QEMU
parameters):
-device pcie-root-port,port=0,id=root_port13,chassis=0,slot=2 \
-device x3130-upstream,id=sw1,bus=root_port13,multifunction=on \
-device e1000,bus=root_port13,addr=0.1 \
-device xio3130-downstream,id=fun1,bus=sw1,chassis=0,slot=3 \
-device e1000,bus=fun1
The first e1000 NIC here is another function in the switch upstream port.
This leads to following errors:
pci 0000:00:04.0: bridge window [mem 0x10200000-0x103fffff] to [bus 02-04]
pci 0000:02:00.0: bridge window [mem 0x10200000-0x103fffff] to [bus 03-04]
pci 0000:02:00.1: BAR 0: failed to assign [mem size 0x00020000]
e1000 0000:02:00.1: can't ioremap BAR 0: [??? 0x00000000 flags 0x0]
Fix this by taking into account bridge windows, device BARs and SR-IOV PF
BARs on the bus (PF BARs include space for VF BARS so only account PF
BARs), including the ones belonging to bridges themselves if it has any.
Link: https://lore.kernel.org/linux-pci/20221014124553.0000696f@huawei.com/
Link: https://lore.kernel.org/linux-pci/6053736d-1923-41e7-def9-7585ce1772d9@ixsystems.com/
Link: https://lore.kernel.org/r/20230131092405.29121-3-mika.westerberg@linux.intel.com
Reported-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reported-by: Alexander Motin <mav@ixsystems.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
After division the extra resource space per hotplug bridge may not be
aligned according to the window alignment, so align it before passing it
down for further distribution.
Link: https://lore.kernel.org/r/20230131092405.29121-2-mika.westerberg@linux.intel.com
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Some devices like ZBT WE1326 and ZBT WF3526-P and some Netgear models need
to delay phy port initialization after calling the mt7621_pcie_init_port()
driver function to get into reliable boots for both warm and hard resets.
The delay required to detect the ports seems to be in the range [75-100]
milliseconds.
If the ports are not detected the controller is not functional.
There is no datasheet or something similar to really understand why this
extra delay is needed only for these devices and it is not for most of
the boards that are built on mt7621 SoC.
This issue has been reported by openWRT community and the complete
discussion is in [0]. The 100 milliseconds delay has been tested in all
devices to validate it.
Add the extra 100 milliseconds delay to fix the issue.
[0]: https://github.com/openwrt/openwrt/pull/11220
Link: https://lore.kernel.org/r/20221231074041.264738-1-sergio.paracuellos@gmail.com
Fixes: 2bdd5238e7 ("PCI: mt7621: Add MediaTek MT7621 PCIe host controller driver")
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Except for isochronous-configured devices, software may set
Max_Read_Request_Size (MRRS) to any value up to 4096. If a device issues a
read request with size greater than the completer's Max_Payload_Size (MPS),
the completer is required to break the response into multiple completions.
Instead of correctly responding with multiple completions to a large read
request, some LS7A Root Ports respond with a Completer Abort. To prevent
this, the MRRS must be limited to an implementation-specific value.
The OS cannot detect that value, so rely on BIOS to configure MRRS before
booting, and quirk the Root Ports so we never set an MRRS larger than that
BIOS value for any downstream device.
N.B. Hot-added devices are not configured by BIOS, and they power up with
MRRS = 512 bytes, so these devices will be limited to 512 bytes. If the
LS7A limit is smaller, those hot-added devices may not work correctly, but
per [1], hotplug is not supported with this chipset revision.
[1] https://lore.kernel.org/r/073638a7-ae68-2847-ac3d-29e5e760d6af@loongson.cn
[bhelgaas: commit log]
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216884
Link: https://lore.kernel.org/r/20230201043018.778499-3-chenhuacai@loongson.cn
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
After cc27b735ad ("PCI/portdrv: Turn off PCIe services during shutdown")
we observe hangs during poweroff/reboot on systems with LS7A chipset.
This happens because the portdrv .shutdown() method (pcie_portdrv_remove())
clears PCI_COMMAND_MASTER via pci_disable_device(), which prevents bridges
from forwarding memory or I/O Requests in the upstream direction (PCIe
r6.0, sec 7.5.1.1.3).
LS7A Root Ports have a hardware defect: clearing PCI_COMMAND_MASTER *also*
prevents the bridge from forwarding CPU MMIO requests in the downstream
direction, and these MMIO accesses to devices below the bridge happen even
after .shutdown(), e.g., to print console messages. LS7A neither forwards
the requests nor sends an unsuccessful completion to the CPU, so the CPU
waits forever, resulting in the hang.
The purpose of .shutdown() is to disable interrupts and DMA from the
device. PCIe ports may generate interrupts (either MSI/MSI-X or INTx) for
AER, DPC, PME, hotplug, etc., but they never perform DMA except MSI/MSI-X.
Clearing PCI_COMMAND_MASTER effectively disables MSI/MSI-X, but not INTx.
The port service driver .remove() methods clear the interrupt enables in
PCI_ERR_ROOT_COMMAND, PCI_EXP_DPC_CTL, PCI_EXP_SLTCTL, and PCI_EXP_RTCTL,
etc., which disables interrupts regardless of whether they are MSI/MSI-X or
INTx.
Add a pcie_portdrv_shutdown() method that calls all the port service driver
.remove() methods to clear the interrupt enables for each service but does
not clear Bus Mastering on the port itself.
[bhelgaas: commit log]
Link: https://lore.kernel.org/r/20230201043018.778499-2-chenhuacai@loongson.cn
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
