regression from commit "usb: host: xhci-plat: Prevent an abnormally restrictive PHY init skipping"

regression from commit "usb: host: xhci-plat: Prevent an abnormally restrictive PHY init skipping"

[Marek Behun]

Hi Miquel,

I am encountering a regression caused by your commit eb6c2eb6c7fb
"usb: host: xhci-plat: Prevent an abnormally restrictive PHY init
skipping" [1]

In the Turris Mox device tree, we use both a comphy and a usb-phy, see
[2]. I am not sure how to solve this now. You write in your commit
message that "While there is not users of both PHY types at the same
time, drop this limitation from the xhci-plat.c driver."

Should I get this patch reverted or try to solve it another way?

Thank you.

Marek

[1]
https://lore.kernel.org/linux-usb/20190731121150.2253-1-miquel.raynal@bootlin.com/
[2]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/marvell/armada-3720-turris-mox.dts#n258

Re: regression from commit "usb: host: xhci-plat: Prevent an abnormally restrictive PHY init skipping"

[Marek Behun]

On Sat, 28 Sep 2019 04:29:50 +0200
Marek Behun <marek.behun@nic.cz> wrote:

> Hi Miquel,
> 
> I am encountering a regression caused by your commit eb6c2eb6c7fb
> "usb: host: xhci-plat: Prevent an abnormally restrictive PHY init
> skipping" [1]

I forgot to explain how the regression presents itself

  rcu: INFO: rcu_sched detected stalls on CPUs/tasks:

The kernel than hangs.

Marek

Re: regression from commit "usb: host: xhci-plat: Prevent an abnormally restrictive PHY init skipping"

[Miquel Raynal]

Hi Marek,

Marek Behun <marek.behun@nic.cz> wrote on Sat, 28 Sep 2019 04:29:50
+0200:

> Hi Miquel,
> 
> I am encountering a regression caused by your commit eb6c2eb6c7fb
> "usb: host: xhci-plat: Prevent an abnormally restrictive PHY init
> skipping" [1]

Sorry for the regression.

> 
> In the Turris Mox device tree, we use both a comphy and a usb-phy, see
> [2]. I am not sure how to solve this now. You write in your commit
> message that "While there is not users of both PHY types at the same
> time, drop this limitation from the xhci-plat.c driver."

After checking, this statement was true at the time I wrote it but
with the game of patches waiting on ML before being applied, both
our changes got merged almost at the same time.

> Should I get this patch reverted or try to solve it another way?

This patch [1] should enter during the next release (part of a bigger
series), it shows how to avoid this harsh situation where we handle two
(very different) PHY types in the same way by describing better the
hardware in the DT (discussed previously with Martin).

[1] https://patchwork.kernel.org/patch/11067687/

Hope that will help?

Thanks,
Miquèl