linux.git/drivers/pci/bus.c, branch for-next

Merge branch 'pci/pwrctl'

2024-11-25T19:40:54+00:00

- Use of_platform_device_create() instead of of_platform_populate() to
  create pwrctl platform devices so we can control it based on the child
  nodes (Manivannan Sadhasivam)

- Create pwrctrl platform devices only if there's a relevant power supply
  property (Manivannan Sadhasivam)

- Add device link from the pwrctl supplier to the PCI dev to ensure pwrctl
  drivers are probed before the PCI dev driver; this avoids a race where
  pwrctl could change device power state while the PCI driver was active
  (Manivannan Sadhasivam)

- Find pwrctl device for removal with of_find_device_by_node() instead of
  searching all children of the parent (Manivannan Sadhasivam)

- Rename 'pwrctl' to 'pwrctrl' to use the same 'ctrl' suffix as 'bwctrl'
  and other PCI files to reduce confusion (Bjorn Helgaas)

* pci/pwrctl:
  PCI/pwrctrl: Rename pwrctrl functions and structures
  PCI/pwrctrl: Rename pwrctl files to pwrctrl
  PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
  PCI/pwrctl: Ensure that pwrctl drivers are probed before PCI client drivers
  PCI/pwrctl: Create pwrctl device only if at least one power supply is present
  PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices

# Conflicts:
#	drivers/pci/bus.c
#	drivers/pci/remove.c

Merge branch 'pci/locking'

2024-11-25T19:40:45+00:00

- Make pci_stop_dev() and pci_destroy_dev() concurrent safe (Keith Busch)

- Move __pci_walk_bus() mutex up into the caller, which avoids the need for
  a parameter to control locking (Keith Busch)

- Simplify __pci_walk_bus() by making it recursive (Keith Busch)

- Unexport pci_walk_bus_locked(), which is only used internally by the PCI
  core (Keith Busch)

* pci/locking:
  PCI: Unexport pci_walk_bus_locked()
  PCI: Convert __pci_walk_bus() to be recursive
  PCI: Move __pci_walk_bus() mutex to where we need it
  PCI: Make pci_destroy_dev() concurrent safe
  PCI: Make pci_stop_dev() concurrent safe

PCI/pwrctrl: Rename pwrctrl functions and structures

2024-11-21T16:01:27+00:00

Rename pwrctrl functions and structures from "pwrctl" to "pwrctrl" to match
the similar file renames.

Link: https://lore.kernel.org/r/20241115214428.2061153-3-helgaas@kernel.org
Signed-off-by: Bjorn Helgaas 
Signed-off-by: Krzysztof Wilczyński 
Acked-by: Krzysztof Wilczyński

PCI/pwrctl: Ensure that pwrctl drivers are probed before PCI client drivers

2024-11-21T16:01:13+00:00

As per the kernel device driver model, a pwrctl device is the supplier for
the PCI device, but the device link that enforces the supplier-consumer
relationship was previously created by the pwrctl driver. Therefore, the
driver model didn't prevent probing PCI client drivers before probing the
corresponding pwrctl drivers. This may lead to a race condition if the PCI
device was already powered on by the bootloader (before the pwrctl driver).

If the bootloader did not power on the PCI device, this wouldn't create any
problem as the pwrctl driver will be the one powering on the device, so the
PCI client driver always gets probed afterward. But if the device was
already powered on, then the device will be seen by the PCI core and the
PCI client driver may get probed before its pwrctl driver. This creates a
race condition as the pwrctl driver may change the device power state while
the device is being accessed by the client driver.

One such issue was already reported on the Qcom X13s platform with the WLAN
device and fixed with a hack in the WCN pwrseq driver by a9aaf1ff88a8
("power: sequencing: request the WLAN enable GPIO as-is").

A cleaner way to fix the above mentioned race condition is to ensure that
the pwrctl drivers are always probed before the client drivers.

If the PCI device is associated with a pwrctl platform device with a power
supply, add a device link between the PCI device and the pwrctl device
before device_attach() in pci_bus_add_device().

Note that there is no need to explicitly remove the device link as that
will be taken care of by the driver core when the PCI device gets removed.

Fixes: 4565d2652a37 ("PCI/pwrctl: Add PCI power control core code")
Fixes: 8fb18619d910 ("PCI/pwrctl: Create platform devices for child OF nodes of the port node")
Link: https://lore.kernel.org/r/20241025-pci-pwrctl-rework-v2-3-568756156cbe@linaro.org
Tested-by: Bartosz Golaszewski 
Tested-by: Krishna chaitanya chundru 
Signed-off-by: Manivannan Sadhasivam 
[bhelgaas: squash fix from
https://lore.kernel.org/r/20241120062459.6371-1-manivannan.sadhasivam@linaro.org
for SPARCv9 issue reported by Jonathan Currier ]
Signed-off-by: Bjorn Helgaas 
[kwilczynski: wrap code to 80 columns]
Signed-off-by: Krzysztof Wilczyński 
Reviewed-by: Bartosz Golaszewski 
Cc: stable+noautosel@kernel.org		# Depends on power supply check

PCI/pwrctl: Create pwrctl device only if at least one power supply is present

2024-11-21T16:01:04+00:00

Currently, pwrctl devices are created if the corresponding PCI nodes are
defined in devicetree. But this is not correct, because not all PCI nodes
require pwrctl support. Pwrctl comes into the picture only when the device
requires kernel to manage its power state. This can be determined using the
power supply properties present in the devicetree node of the device.

Add of_pci_supply_present() to check whether the devicetree contains at
least one power supply property for a device. If one is present, create a
pwrctl device for that PCI node.

Suggested-by: Dmitry Baryshkov 
Fixes: 8fb18619d910 ("PCI/pwrctl: Create platform devices for child OF nodes of the port node")
Link: https://lore.kernel.org/r/20241025-pci-pwrctl-rework-v2-2-568756156cbe@linaro.org
Tested-by: Bartosz Golaszewski 
Tested-by: Krishna chaitanya chundru 
Signed-off-by: Manivannan Sadhasivam 
[bhelgaas: rename of_pci_is_supply_present() to of_pci_supply_present() for
readability]
Signed-off-by: Bjorn Helgaas 
Signed-off-by: Krzysztof Wilczyński 
Reviewed-by: Bartosz Golaszewski 
Cc: stable+noautosel@kernel.org		# Depends on of_platform_device_create() rework

PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices

2024-11-21T16:01:00+00:00

The of_platform_populate() API creates platform devices by descending
through the children of the parent node. But it provides no control over
the child nodes, which makes it difficult to add checks for the child
nodes in the future.

Use of_platform_device_create() and for_each_child_of_node_scoped() to make
it possible to add checks for each node before creating the platform
device.

Link: https://lore.kernel.org/r/20241025-pci-pwrctl-rework-v2-1-568756156cbe@linaro.org
Tested-by: Bartosz Golaszewski 
Tested-by: Krishna chaitanya chundru 
Signed-off-by: Manivannan Sadhasivam 
Signed-off-by: Bjorn Helgaas 
Signed-off-by: Krzysztof Wilczyński 
Reviewed-by: Bartosz Golaszewski

PCI: Unexport pci_walk_bus_locked()

2024-11-11T20:32:12+00:00

There's only one user of pci_walk_bus_locked(), and it's internal to the
PCI core.  Unexport it and make it private to drivers/pci/.

Link: https://lore.kernel.org/r/20241022224851.340648-6-kbusch@meta.com
Signed-off-by: Keith Busch 
[bhelgaas: move decl to drivers/pci/pci.h]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Jonathan Cameron 
Reviewed-by: Davidlohr Bueso 
Reviewed-by: Ilpo Järvinen

PCI: Convert __pci_walk_bus() to be recursive

2024-11-11T19:06:02+00:00

The original implementation of __pci_walk_bus() chose a non-recursive walk,
presumably as a precaution on stack use. We do recursive bus walking in
other places though. For example:

  pci_bus_resettable()
  pci_stop_bus_device()
  pci_remove_bus_device()
  pci_bus_allocate_dev_resources()

So recursive pci bus walking is well tested and safe, and is easier to
follow.

Convert __pci_walk_bus() to be recursive to make it easier to introduce
finer grain locking in the future.

Link: https://lore.kernel.org/r/20241022224851.340648-5-kbusch@meta.com
Signed-off-by: Keith Busch 
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Jonathan Cameron 
Reviewed-by: Ilpo Järvinen

PCI: Move __pci_walk_bus() mutex to where we need it

2024-11-11T19:06:02+00:00

Simplify __pci_walk_bus() by moving the pci_bus_sem mutex into
pci_walk_bus(), the only place it is needed, and removing the parameter
that told __pci_walk_bus() whether to acquire the mutex.

Link: https://lore.kernel.org/r/20241022224851.340648-4-kbusch@meta.com
Signed-off-by: Keith Busch 
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Jonathan Cameron 
Reviewed-by: Davidlohr Bueso 
Reviewed-by: Ilpo Järvinen

PCI: Make pci_stop_dev() concurrent safe

2024-11-11T19:05:50+00:00

Use the atomic ADDED flag to ensure concurrent callers can't attempt to
stop the device multiple times. Callers should currently all be holding the
pci_rescan_remove_lock, so there shouldn't be an existing race. But that
global lock can cause lock dependency issues, so this is preparing to
reduce reliance on that lock by using the existing existing atomic bit ops.

Link: https://lore.kernel.org/r/20241022224851.340648-2-kbusch@meta.com
Signed-off-by: Keith Busch 
[bhelgaas: squash https://lore.kernel.org/r/20241111180659.3321671-1-kbusch@meta.com]
Signed-off-by: Bjorn Helgaas 
Reviewed-by: Jonathan Cameron