drivers/pci/bus.c | 64 +++++++++++++++++++++++++++++++++++++++++------ drivers/pci/of.c | 27 ++++++++++++++++++++ drivers/pci/pci.h | 5 ++++ drivers/pci/pwrctl/core.c | 10 -------- drivers/pci/remove.c | 17 ++++++------- 5 files changed, 96 insertions(+), 27 deletions(-)
Hi,
This series reworks the PCI/pwrctl integration to ensure that the pwrctl drivers
are always probed before the PCI client drivers. This series addresses a race
condition when both pwrctl and pwrctl/pwrseq drivers probe parallely (even when
the later one probes last). One such issue was reported for the Qcom X13s
platform with WLAN module and fixed with 'commit a9aaf1ff88a8 ("power:
sequencing: request the WLAN enable GPIO as-is")'.
Though the issue was fixed with a hack in the pwrseq driver, it was clear that
the issue is applicable to all pwrctl drivers. Hence, this series tries to
address the issue in the PCI/pwrctl integration.
- Mani
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
Manivannan Sadhasivam (5):
PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices
PCI/pwrctl: Create pwrctl devices only if at least one power supply is present
PCI/pwrctl: Ensure that the pwrctl drivers are probed before the PCI client drivers
PCI/pwrctl: Move pwrctl device creation to its own helper function
PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
drivers/pci/bus.c | 64 +++++++++++++++++++++++++++++++++++++++++------
drivers/pci/of.c | 27 ++++++++++++++++++++
drivers/pci/pci.h | 5 ++++
drivers/pci/pwrctl/core.c | 10 --------
drivers/pci/remove.c | 17 ++++++-------
5 files changed, 96 insertions(+), 27 deletions(-)
---
base-commit: 48dc7986beb60522eb217c0016f999cc7afaf0b7
change-id: 20241022-pci-pwrctl-rework-a1b024158555
Best regards,
--
Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Hello,
> This series reworks the PCI/pwrctl integration to ensure that the pwrctl drivers
> are always probed before the PCI client drivers. This series addresses a race
> condition when both pwrctl and pwrctl/pwrseq drivers probe parallely (even when
> the later one probes last). One such issue was reported for the Qcom X13s
> platform with WLAN module and fixed with 'commit a9aaf1ff88a8 ("power:
> sequencing: request the WLAN enable GPIO as-is")'.
>
> Though the issue was fixed with a hack in the pwrseq driver, it was clear that
> the issue is applicable to all pwrctl drivers. Hence, this series tries to
> address the issue in the PCI/pwrctl integration.
Applied to bwctrl, thank you!
[01/05] PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices
https://git.kernel.org/pci/pci/c/d2b6619e7419
[02/05] PCI/pwrctl: Create pwrctl devices only if at least one power supply is present
https://git.kernel.org/pci/pci/c/5f2710a4c275
[03/05] PCI/pwrctl: Ensure that the pwrctl drivers are probed before the PCI client drivers
https://git.kernel.org/pci/pci/c/4c963d4c13b9
[04/05] PCI/pwrctl: Move pwrctl device creation to its own helper function
https://git.kernel.org/pci/pci/c/73ae23a6af78
[05/05] PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
https://git.kernel.org/pci/pci/c/5ccc52fd1e5a
Krzysztof
On Mon, Nov 04, 2024 at 05:31:07AM +0900, Krzysztof Wilczyński wrote:
> Hello,
>
> > This series reworks the PCI/pwrctl integration to ensure that the pwrctl drivers
> > are always probed before the PCI client drivers. This series addresses a race
> > condition when both pwrctl and pwrctl/pwrseq drivers probe parallely (even when
> > the later one probes last). One such issue was reported for the Qcom X13s
> > platform with WLAN module and fixed with 'commit a9aaf1ff88a8 ("power:
> > sequencing: request the WLAN enable GPIO as-is")'.
> >
> > Though the issue was fixed with a hack in the pwrseq driver, it was clear that
> > the issue is applicable to all pwrctl drivers. Hence, this series tries to
> > address the issue in the PCI/pwrctl integration.
>
> Applied to bwctrl, thank you!
Should be pci/pwrctl. bwctrl (bandwidth control) and pwrctl (power
control) are quite different despite the confusingly similar names.
> [01/05] PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices
> https://git.kernel.org/pci/pci/c/d2b6619e7419
>
> [02/05] PCI/pwrctl: Create pwrctl devices only if at least one power supply is present
> https://git.kernel.org/pci/pci/c/5f2710a4c275
>
> [03/05] PCI/pwrctl: Ensure that the pwrctl drivers are probed before the PCI client drivers
> https://git.kernel.org/pci/pci/c/4c963d4c13b9
>
> [04/05] PCI/pwrctl: Move pwrctl device creation to its own helper function
> https://git.kernel.org/pci/pci/c/73ae23a6af78
>
> [05/05] PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
> https://git.kernel.org/pci/pci/c/5ccc52fd1e5a
>
> Krzysztof
Hello,
> > > This series reworks the PCI/pwrctl integration to ensure that the pwrctl drivers
> > > are always probed before the PCI client drivers. This series addresses a race
> > > condition when both pwrctl and pwrctl/pwrseq drivers probe parallely (even when
> > > the later one probes last). One such issue was reported for the Qcom X13s
> > > platform with WLAN module and fixed with 'commit a9aaf1ff88a8 ("power:
> > > sequencing: request the WLAN enable GPIO as-is")'.
> > >
> > > Though the issue was fixed with a hack in the pwrseq driver, it was clear that
> > > the issue is applicable to all pwrctl drivers. Hence, this series tries to
> > > address the issue in the PCI/pwrctl integration.
> >
> > Applied to bwctrl, thank you!
>
> Should be pci/pwrctl. bwctrl (bandwidth control) and pwrctl (power
> control) are quite different despite the confusingly similar names.
Correct. I moved patches to the correct branch. Sorry about that!
Krzysztof
On Tue, 22 Oct 2024 at 12:28, Manivannan Sadhasivam via B4 Relay
<devnull+manivannan.sadhasivam.linaro.org@kernel.org> wrote:
>
> Hi,
>
> This series reworks the PCI/pwrctl integration to ensure that the pwrctl drivers
> are always probed before the PCI client drivers. This series addresses a race
> condition when both pwrctl and pwrctl/pwrseq drivers probe parallely (even when
> the later one probes last). One such issue was reported for the Qcom X13s
> platform with WLAN module and fixed with 'commit a9aaf1ff88a8 ("power:
> sequencing: request the WLAN enable GPIO as-is")'.
>
> Though the issue was fixed with a hack in the pwrseq driver, it was clear that
> the issue is applicable to all pwrctl drivers. Hence, this series tries to
> address the issue in the PCI/pwrctl integration.
>
> - Mani
>
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> ---
> Manivannan Sadhasivam (5):
> PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices
> PCI/pwrctl: Create pwrctl devices only if at least one power supply is present
> PCI/pwrctl: Ensure that the pwrctl drivers are probed before the PCI client drivers
> PCI/pwrctl: Move pwrctl device creation to its own helper function
> PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
>
> drivers/pci/bus.c | 64 +++++++++++++++++++++++++++++++++++++++++------
> drivers/pci/of.c | 27 ++++++++++++++++++++
> drivers/pci/pci.h | 5 ++++
> drivers/pci/pwrctl/core.c | 10 --------
> drivers/pci/remove.c | 17 ++++++-------
> 5 files changed, 96 insertions(+), 27 deletions(-)
> ---
> base-commit: 48dc7986beb60522eb217c0016f999cc7afaf0b7
> change-id: 20241022-pci-pwrctl-rework-a1b024158555
>
> Best regards,
> --
> Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
>
>
Excellent work, thanks for doing this.
Tested on: sc8280xp-crd, RB5 and sm8450-hdk.
Tested-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Just a couple nits from my side under respective patches.
Bart
On 10/23/2024 4:00 PM, Bartosz Golaszewski wrote:
> On Tue, 22 Oct 2024 at 12:28, Manivannan Sadhasivam via B4 Relay
> <devnull+manivannan.sadhasivam.linaro.org@kernel.org> wrote:
>>
>> Hi,
>>
>> This series reworks the PCI/pwrctl integration to ensure that the pwrctl drivers
>> are always probed before the PCI client drivers. This series addresses a race
>> condition when both pwrctl and pwrctl/pwrseq drivers probe parallely (even when
>> the later one probes last). One such issue was reported for the Qcom X13s
>> platform with WLAN module and fixed with 'commit a9aaf1ff88a8 ("power:
>> sequencing: request the WLAN enable GPIO as-is")'.
>>
>> Though the issue was fixed with a hack in the pwrseq driver, it was clear that
>> the issue is applicable to all pwrctl drivers. Hence, this series tries to
>> address the issue in the PCI/pwrctl integration.
>>
>> - Mani
>>
>> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
>> ---
>> Manivannan Sadhasivam (5):
>> PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices
>> PCI/pwrctl: Create pwrctl devices only if at least one power supply is present
>> PCI/pwrctl: Ensure that the pwrctl drivers are probed before the PCI client drivers
>> PCI/pwrctl: Move pwrctl device creation to its own helper function
>> PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
>>
>> drivers/pci/bus.c | 64 +++++++++++++++++++++++++++++++++++++++++------
>> drivers/pci/of.c | 27 ++++++++++++++++++++
>> drivers/pci/pci.h | 5 ++++
>> drivers/pci/pwrctl/core.c | 10 --------
>> drivers/pci/remove.c | 17 ++++++-------
>> 5 files changed, 96 insertions(+), 27 deletions(-)
>> ---
>> base-commit: 48dc7986beb60522eb217c0016f999cc7afaf0b7
>> change-id: 20241022-pci-pwrctl-rework-a1b024158555
>>
>> Best regards,
>> --
>> Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
>>
>>
>
> Excellent work, thanks for doing this.
>
> Tested on: sc8280xp-crd, RB5 and sm8450-hdk.
>
> Tested-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
>
> Just a couple nits from my side under respective patches.
>
> Bart
Tested on: qcs6490-rb3gen board with work in progress qps615 pcie switch
Tested-by: Krishna chaitanya chundru <quic_krichai@quicinc.com>
- Krishna Chaitanya.
>
© 2016 - 2024 Red Hat, Inc.