Message-ID: <cb081d0f-3a53-4b5e-9f14-92745820fc75@bootlin.com>
Date: Thu, 12 Jun 2025 15:23:14 +0200
From: Louis Chauvet <louis.chauvet@...tlin.com>
To: Luca Ceresoli <luca.ceresoli@...tlin.com>,
 Minas Harutyunyan <Minas.Harutyunyan@...opsys.com>
Cc: Alan Stern <stern@...land.harvard.edu>,
 "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
 Kever Yang <kever.yang@...k-chips.com>,
 Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
 Hervé Codina <herve.codina@...tlin.com>,
 Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
 Stefan Wahren <wahrenst@....net>,
 Fabrice Gasnier <fabrice.gasnier@...s.st.com>,
 linux-arm-kernel@...ts.infradead.org, linux-phy@...ts.infradead.org,
 heiko@...ech.de, vkoul@...nel.org, kishon@...nel.org,
 linux-rockchip@...ts.infradead.org
Subject: Re: DWC2 gadget: unexpected device reenumeration on Rockchip RK3308



Le 20/05/2025 à 14:09, Luca Ceresoli a écrit :
> Hello Minas,

[...]

> Minas, do you have any hints or advice to understand why there is a
> disconnect about ~6 seconds after a successful enumeration in gadget
> mode?

+CC: vkoul@...nel.org, kishon@...nel.org, 
linux-rockchip@...ts.infradead.org, 
linux-arm-kernel@...ts.infradead.org, linux-phy@...ts.infradead.org, 
heiko@...ech.de

Hello,

I am a colleague of Luca and have been working on an issue related to 
USB peripheral disconnections over the past few days. I believe I have 
found some interesting insights that explain the disconnection.

To provide some context, I have added traces in relevant functions to 
understand the sequence of events (filtered on interesting logs, see [1] 
for full logs):


[... boot proccess ...]

[    0.628744] dwc2 ff400000.usb: mapped PA ff400000 to VA (____ptrval____)
[    0.629358] dwc2 ff400000.usb: supply vusb_d not found, using dummy 
regulator
[    0.630478] dwc2 ff400000.usb: supply vusb_a not found, using dummy 
regulator
[    0.631297] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/usb/dwc2/platform.c:157:dwc2_lowlevel_hw_enable
[    0.632406] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/phy-core.c:230:phy_init
[    0.633233] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:491:rockchip_usb2phy_init
[    0.634262] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:539:rockchip_usb2phy_init 
(before schedule_delayed_work otg_sm_work)
[    0.635602] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/phy-core.c:309:phy_power_on
[    0.636379] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:577:rockchip_usb2phy_power_on
[    0.637496] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:606:rockchip_usb2phy_power_on 
(before rockchip_usb2phy_reset)

[... bunch of dwc2 ff400000.usb logs ...]

[    0.688708] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/usb/dwc2/platform.c:200:dwc2_lowlevel_hw_disable
[    0.689661] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/phy-core.c:360:phy_power_off
[    0.690442] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:620:rockchip_usb2phy_power_off
[    0.691522] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/phy-core.c:271:phy_exit
[    0.692261] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:648:rockchip_usb2phy_exit 
(before cancel_delayed_work_sync otg_sm_work/chg_work)
[    0.698909] dwmmc_rockchip ff490000.mmc: IDMAC supports 32-bit 
address mode.

[... boot finished ...]

[... call to my peripheral setup script ...]

[   44.493316] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/usb/dwc2/platform.c:157:dwc2_lowlevel_hw_enable
[   44.495078] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/phy-core.c:230:phy_init
[   44.495837] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:491:rockchip_usb2phy_init
[   44.496960] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:539:rockchip_usb2phy_init 
(before schedule_delayed_work otg_sm_work)
[   44.498317] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/phy-core.c:309:phy_power_on
[   44.499107] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:577:rockchip_usb2phy_power_on
[   44.500160] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:606:rockchip_usb2phy_power_on 
(before rockchip_usb2phy_reset)


[... bunch of dwc2 ff400000.usb logs ...]
[   44.976346] dwc2 ff400000.usb: dwc2_hsotg_start_req: DXEPCTL=0x80028000

[... ~5 seconds delay ...]

[   50.660998] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:670:rockchip_usb2phy_otg_sm_work
[   50.662128] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:685:rockchip_usb2phy_otg_sm_work 
(rport->state=0 vbus_attach=1 rphy->chg_state=0 rphy->chg_type=0 
extcon_get_state=0)
[   50.663862] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:831:rockchip_chg_detect_work
[   50.664973] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:839:rockchip_chg_detect_work 
(rphy->chg_state=0 rport->suspended=0)
[   50.666316] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:620:rockchip_usb2phy_power_off
[   50.769027] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:831:rockchip_chg_detect_work
[   50.770093] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:839:rockchip_chg_detect_work 
(rphy->chg_state=1 rport->suspended=1)
[   50.813030] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:831:rockchip_chg_detect_work
[   50.814083] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:839:rockchip_chg_detect_work 
(rphy->chg_state=2 rport->suspended=1)
[   50.857013] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:831:rockchip_chg_detect_work
[   50.858067] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:839:rockchip_chg_detect_work 
(rphy->chg_state=3 rport->suspended=1)
[   50.859409] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:670:rockchip_usb2phy_otg_sm_work
[   50.860475] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:685:rockchip_usb2phy_otg_sm_work 
(rport->state=1 vbus_attach=1 rphy->chg_state=5 rphy->chg_type=6 
extcon_get_state=0)
[   50.862267] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:577:rockchip_usb2phy_power_on
[   50.863326] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:606:rockchip_usb2phy_power_on 
(before rockchip_usb2phy_reset)

[... bunch of dwc2 ff400000.usb logs ...]

[   52.868992] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:670:rockchip_usb2phy_otg_sm_work
[   52.870108] phy phy-ff008000.syscon:usb2phy@....0: 
drivers/phy/rockchip/phy-rockchip-inno-usb2.c:685:rockchip_usb2phy_otg_sm_work 
(rport->state=3 vbus_attach=1 rphy->chg_state=5 rphy->chg_type=6 
extcon_get_state=0)

[... repeated every 2 seconds ...]



I found two potential issues that may be relevant:

1 - The workers rockchip_usb2phy_otg_sm_work and 
rockchip_chg_detect_work call rockchip_usb2phy_power_off/on directly, 
bypassing phy_power_off/on. This means the state of the PHY after a call 
to phy_power_on may not be the one expected by users of the struct phy api.
2 -  The worker rockchip_chg_detect_work always disables the PHY at its 
first call, which creates the disconnection.


For issue 1, here is the phy_power_on/off reference count:

[    0.635602] phy_power_on	count = 1
[    0.689661] phy_power_off	count = 0
[   44.498317] phy_power_on	count = 1 // ./setup-cdc-serial
[   50.666316] rockchip_usb2phy_power_off => direct call to power_off! 
This causes issues in the dwc2 driver because the PHY is powered off 
even if the phy_power_on counter is not zero, leading to the disconnection
[   50.862267] rockchip_usb2phy_power_on => direct call to power_on and 
reset! This will reset the PHY even if the phy_power_on counter is not zero.

This is my first time digging into the USB/PHY subsystem, so I might be 
wrong, but this behavior seems very strange. Once phy_power_on is 
called, I expect the PHY to never go off until phy_power_off is called.


For issue 2, this is also a strange behavior. I understand that 
resetting the PHY before its first use might be necessary, but I don't 
understand why this must be delayed by 6 seconds. In my opinion, this 
reset sequence should occur synchronously during the phy_init call.

I quickly looked at other PHY drivers, and none of them seem to have 
asynchronous behavior or any internal calls to power_on/power_off.

Could you confirm if what I found is indeed the issue and if the 
rockchip_usb2phy driver needs to be fixed?

Does anyone know why rockchip_usb2phy needs to perform these 
asynchronous tasks and these power_on/off calls?

Let me know if you need more logs to understand the issue.

Thanks a lot,
Louis Chauvet


[1]:https://paste.sr.ht/~fomys/26e0d03d03537a6eb3757fb709b6d5a81484eee3

-- 
Louis Chauvet, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Powered by blists - more mailing lists