[PATCH 0/4] staging: mt7621-pci: Use gpio perst instead builtin behaviour

Greg Ungerer gerg at kernel.org
Thu Jun 20 00:40:33 UTC 2019

Hi Sergio,

On 19/6/19 5:44 pm, Sergio Paracuellos wrote:
> Some boards seems to ignore builtin perst configuration and use gpio
> instead. This approach seems to be more common. Hence, update the driver
> to properly use gpio perst via gpio descriptor's API.
> For more information refer to [1].
> Even with this set of patches applied, there still seems to have some issues
> with a non stable pci link through the boot process. This will be resolved
> in nest patch series when the problem is find.
> Patches are only compile-tested. It would be good to test them before being
> applied.
> [1]: http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2019-June/134947.html
> Sergio Paracuellos (4):
>    staging: mt7621-pci: fix two messages in driver code
>    staging: mt7621-pci: use gpio perst instead of builtin behaviour
>    staging: mt7621-dts: add gpio perst to pcie bindings node
>    staging: mt7621-pci: dt-bindings: add perst-gpio to sample bindings
>   drivers/staging/mt7621-dts/mt7621.dtsi        |   2 +
>   .../mt7621-pci/mediatek,mt7621-pci.txt        |   2 +
>   drivers/staging/mt7621-pci/pci-mt7621.c       | 108 ++++++++++--------
>   3 files changed, 63 insertions(+), 49 deletions(-)

Thanks for putting this together.

I tried a quick test, applying this onto a linux-5.2.0-rc5 kernel and
saw this in the boot trace:

   rt2880-pinmux pinctrl: pcie is already enabled
   mt7621-pci 1e140000.pcie: Error applying setting, reverse things back

The system went on to boot successfully, with PCI working.

Testing across multiple boots, it sometimes hangs - but we know about that
and it is what we are still looking into.


More information about the devel mailing list