Re: [PATCH v1] serdev: Set fwnode for serdev devices

[Saravana Kannan]

On Fri, Mar 3, 2023 at 9:22 AM Florian Fainelli <f.fainelli@xxxxxxxxx> wrote:
>
> On 3/3/23 03:57, Stefan Wahren wrote:
> > Hi,
> >
> > Am 02.03.23 um 18:51 schrieb Florian Fainelli:
> >>
> >>
> >> On 3/2/2023 9:20 AM, Saravana Kannan wrote:
> >>> On Thu, Mar 2, 2023 at 9:01 AM Stefan Wahren <stefan.wahren@xxxxxxxx>
> >>> wrote:
> >>>>
> >>>> Hi Saravana,
> >>>>
> >>>> Am 02.03.23 um 03:35 schrieb Saravana Kannan:
> >>>>> This allow fw_devlink to do dependency tracking for serdev devices.
> >>>>>
> >>>>> Reported-by: Florian Fainelli <f.fainelli@xxxxxxxxx>
> >>>>> Link:
> >>>>> https://lore.kernel.org/lkml/03b70a8a-0591-f28b-a567-9d2f736f17e5@xxxxxxxxx/
> >>>>> Cc: Stefan Wahren <stefan.wahren@xxxxxxxx>
> >>>>> Signed-off-by: Saravana Kannan <saravanak@xxxxxxxxxx>
> >>>>
> >>>> since this fixes an issue on Raspberry Pi 4, shouldn't this be
> >>>> mentioned
> >>>> in the commit message and providing a Fixes tag?
> >>>
> >>> So RPi 4 was never creating a device links between serdev devices and
> >>> their consumers. The error message was just a new one I added and we
> >>> are noticing and catching the fact that serdev wasn't setting fwnode
> >>> for a device.
> >>>
> >>> I'm also not sure if I can say this commit "Fixes" an issue in serdev
> >>> core because when serdev core was written, fw_devlink wasn't a thing.
> >>> Once I add Fixes, people will start pulling this into stable
> >>> branches/other trees where I don't think this should be pulled into
> >>> older stable branches.
> >>
> >> That is kind of the point of Fixes: tag, is not it? It is appropriate
> >> to list a commit that is not specific to serdev, but maybe a
> >> particular point into the fw_devlink history. Given this did not
> >> appear to have a functional impact, we could go without one.
> >
> > i was under the impression that this issue breaks at least Bluetooth on
> > Raspberry Pi 4 because the driver is never probed. I cannot see the
> > success output in Florian's trace. Something like this:
> >
> > [    7.124879] hci_uart_bcm serial0-0: supply vbat not found, using
> > dummy regulator
> > [    7.131743] hci_uart_bcm serial0-0: supply vddio not found, using
> > dummy regulator
> > ...
> > [    7.517249] Bluetooth: hci0: BCM: chip id 107
> > [    7.517499] Bluetooth: hci0: BCM: features 0x2f
> > [    7.519757] Bluetooth: hci0: BCM4345C0
> > [    7.519768] Bluetooth: hci0: BCM4345C0 (003.001.025) build 0000
> > [    7.539495] Bluetooth: hci0: BCM4345C0 'brcm/BCM4345C0.hcd' Patch
> > ...
> > [    8.348831] Bluetooth: hci0: BCM43455 37.4MHz Raspberry Pi 3+
> > [    8.348845] Bluetooth: hci0: BCM4345C0 (003.001.025) build 0342
> >
> > I just want to make sure that 6.2 doesn't have a regression.
>
> My configuration uses hci_uart as a module, and it would always load
> fine, but I suppose I can make sure that even built-in this works
> properly. Give me a day or two to test that.

Thanks Stefan and Florian! I'll wait to see the results.

But based on my mental model of fw_devlink I don't expect BT to be
broken without this patch. If a device doesn't have fwnode set, it's
effectively invisible to fw_devlink. That could only affect consumers
of the device and not the device itself.

-Saravana