Re: [PATCH v1 0/4] Remove use of fw_devlink_purge_absent_suppliers()

[Martin Kepplinger <martin.kepplinger@puri.sm>]

Am Freitag, dem 10.03.2023 um 14:18 -0800 schrieb Saravana Kannan:
> On Fri, Mar 10, 2023 at 2:07 AM Martin Kepplinger
> <martin.kepplinger@puri.sm> wrote:
> > 
> > Am Donnerstag, dem 09.03.2023 um 16:24 -0800 schrieb Saravana
> > Kannan:
> > > On Thu, Mar 2, 2023 at 1:41 AM Martin Kepplinger
> > > <martin.kepplinger@puri.sm> wrote:
> > > > 
> > > > Am Donnerstag, dem 02.03.2023 um 10:12 +0100 schrieb Martin
> > > > Kepplinger:
> > > > > Am Mittwoch, dem 01.03.2023 um 13:49 -0800 schrieb Saravana
> > > > > Kannan:
> > > > > > Yongqin, Martin, Amelie,
> > > > > > 
> > > > > > We recent refactor of fw_devlink that ends with commit
> > > > > > fb42378dcc7f
> > > > > > ("mtd: mtdpart: Don't create platform device that'll never
> > > > > > probe"),
> > > > > > fw_devlink is smarter and doesn't depend on compatible
> > > > > > property.
> > > > > > So,
> > > > > > I
> > > > > > don't think these calls are needed anymore. But I don't
> > > > > > have
> > > > > > these
> > > > > > devices to test on and be sure and the hardware I use to
> > > > > > test
> > > > > > changes
> > > > > > doesn't have this issue either.
> > > > > > 
> > > > > > Can you please test these changes on the hardware where you
> > > > > > hit
> > > > > > the
> > > > > > issue to make sure things work as expected?
> > > > > > 
> > > > > > Yongqin, If you didn't have the context, this affected
> > > > > > hikey960.
> > > > > > 
> > > > > > Greg,
> > > > > > 
> > > > > > Let's wait for some tests before we land these.
> > > > > > 
> > > > > > Thanks,
> > > > > > Saravana
> > > > > 
> > > > > hi Sravana,
> > > > > 
> > > > > I picked the 12 commits leading up to commit fb42378dcc7f
> > > > > ("mtd:
> > > > > mtdpart: Don't create platform device that'll never probe") (
> > > > > https://source.puri.sm/martin.kepplinger/linux-next/-/commits/test_fw_devlink
> > > > > ) and included the tipd patch below to test it.
> > > > > 
> > > > > With that, I get the following errors:
> > > > > 
> > > > > [    0.237931] imx-uart 30890000.serial: Failed to create
> > > > > device
> > > > > link
> > > > > with regulator-gnss
> > > > > [    0.334054] nwl-dsi 30a00000.mipi-dsi: Failed to create
> > > > > device
> > > > > link
> > > > > with regulator-lcd-1v8
> > > > > [    0.346964] nwl-dsi 30a00000.mipi-dsi: Failed to create
> > > > > device
> > > > > link
> > > > > with backlight-dsi
> > > > > 
> > > > > but they are independent of this final tipd patch below. I'll
> > > > > test a
> > > > > real linux-next tree soon, for completeness, maybe I missed
> > > > > something?
> > > > > 
> > > > > Anyways, on that tree, your tipd removal patch breaks type-c
> > > > > still
> > > > > for
> > > > > me, imx8mq-librem5.dtsi
> > > > > 
> > > > > just to give a first reply quickly... thanks,
> > > > > 
> > > > >                              martin
> > > > > 
> > > > 
> > > > just confirming: it's the same as above on next-20230302 + this
> > > > patch (
> > > > https://source.puri.sm/martin.kepplinger/linux-next/-/commits/test_fw_devlink_next-20230302
> > > > ) with the errors already independent from the patch. I should
> > > > have
> > > > tested earlier patches -.-
> > > 
> > > Thanks a lot for testing Martin!
> > > 
> > > Your email is a little ambiguous to me. With the 12 refactor
> > > commits
> > > +
> > > the 4 patches in this series, things are breaking for you. But if
> > > you
> > > drop the 4 patches in this series, things work again. Is that
> > > right?
> > 
> > no. Sorry if I wasn't clear. I can't justify to block these 4
> > patches.
> > they *themselves* don't break anything.
> > 
> > Something broke *earlier* than these 4 patches in one of the other
> > 12.
> 
> If you find out it's one of the other 12 patches in the refactor that
> broke things for you, can you please reply to the right email in that
> series[1] and let me know which patch broke things for you and
> provide
> the debug details there? I don't want to mix issues with unrelated
> threads -- I want them to be easy to find in the future.
> 
> [1] -  
> https://lore.kernel.org/lkml/20230207014207.1678715-1-saravanak@google.com/
> 
> For all my questions below, you don't need to reply here. Just reply
> to the right thread.

Thanks. I'll have to reply here though - I'm puzzled how, but I got it
wrong - I must have seen the "Failed to create device link" messages
without checking broken drivers: The 12 patches you linked above are
fine. (In a way that's good as I saw them in a stable kernel already).

commit ("usb: typec: tipd: Remove use of
fw_devlink_purge_absent_suppliers()") breaks things for me. That is
patch 2 of this series. That's for sure now.

> 
> > As to *what* broke: The error messages during boot,
> 
> The error messages, is it anything other than the "Failed to create
> device link"?
> 
> > and the charger and
> > fuel gauge don't work anymore, I see:
> > 
> > # cat /sys/kernel/debug/devices_deferred
> > 3-0036  max17042: failed: power supply register
> > 0-003f
> 
> This seems to be the main culprit. That 0-003f isn't probing. It's
> surprising that the reason is empty too. IIRC that means it didn't do
> deferred probing, but rather failed with an error. Wild guess ....
> Try
> increasing the deferred_probe_timeout to see if it helps?
> 
> > 38100000.usb    platform: wait for supplier endpoint
> > 32c00000.hdmi   platform: supplier 0-003f not ready
> > 3-006a  bq25890-charger: registering power supply
> > 
> > 
> > I haven't had time to track it down and see where the issue is.
> > Could
> > be one of the 12 patches, could be a wrong description for my
> > board:
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi
> > or something else (an i2c issue?).
> > 
> > (for completeness, the exact tree I ran:
> > https://source.puri.sm/martin.kepplinger/linux-next/-/commits/test_fw_devlink
> > )
> > 
> > > 
> > > Let's ignore the "Failed to create device link" errors for now --
> > > it's
> > > not related to this usb-c-connector series. It's basically
> > > pointing
> > > out issues that we ignored silently in the past -- it's basically
> > > pointing out holes in fw_devlink's visibility of devices. I'll
> > > get to
> > > them later.
> > 
> > ok. good to know. please do.
> 
> Here's a fix for the Failed to create device link issues.
> https://lore.kernel.org/lkml/20230310063910.2474472-1-saravanak@google.com/
> 
> > And since I'm the only one seeing an issue, please move the patches
> > forward. I'll need to fix this after the fact and hope to have time
> > for
> > it next week.
> 
> I'd rather wait for you than rush these patches in. The patches are
> just removing what I expect to be cruft. I'd rather confirm that
> first
> than break anything.
> 
> -Saravana