usb: ehci: Fix db410c usb reset not working

[Caleb Connolly]

On 19/08/2024 13:21, Marek Vasut wrote:
> On 8/19/24 4:05 AM, JianfengA.Zhu at sony.com wrote:
>> usb reset not working error log
>> ====================================================================
>> dragonboard410c => usb reset
>> resetting USB...
>> Bus usb at 78d9000: Port not available.
>> dragonboard410c =>
>> ====================================================================
>>
>> After commit ed8fbd2889fc ("dts: msm8916: replace with upstream DTS")
>> msm8916_usbphy will be defined as a child device of usb at 78d9000. usb 
>> reset
>> will first call usb_stop, and msm8916_usbphy as child dev will be unbind
>> in usb_stop, and there is no bind operation in do_usb_start 
>> afterwards, so
>> the msm8916_usbphy uclass cannot be found. This causes ehci_msm probe to
>> fail.
>>
>> Detailed calling process
>> ====================================================================
>> usb reset:
>> |-> usb_stop (drivers/usb/host/usb-uclass.c)
>> .|-> device_find_first_child
>> . |-> device_unbind(rh); <== (unbind msm8916_usbphy)
>> |-> do_usb_start
>>   |-> usb_init (drivers/usb/host/usb-uclass.c)
>>    |-> device_probe(bus);
>>     |-> ehci_usb_probe
>>      |-> generic_setup_phy <== err couldn't find msm8916_usbphy
>> ====================================================================
>>
>> Fix: rebind the msm8916_usbphy driver during ehci_usb_probe if the
>> msm8916_usbphy uclass has been removed.
>>
>> Signed-off-by: Jianfeng Zhu <JianfengA.Zhu at sony.com>
>> Reviewed-by: Jacky Cao <Jacky.Cao at sony.com>
>> Reviewed-by: Toyama, Yoshihiro <Yoshihiro.Toyama at sony.com>
>> ---
>>   drivers/usb/host/ehci-msm.c | 7 +++++++
>>   1 file changed, 7 insertions(+)
>>
>> diff --git a/drivers/usb/host/ehci-msm.c b/drivers/usb/host/ehci-msm.c
>> index ff336082e3..7cb3ae0849 100644
>> --- a/drivers/usb/host/ehci-msm.c
>> +++ b/drivers/usb/host/ehci-msm.c
>> @@ -47,6 +47,8 @@ static const struct ehci_ops msm_ehci_ops = {
>>       .init_after_reset = msm_init_after_reset
>>   };
>> +static int ehci_usb_of_bind(struct udevice *dev);
>> +
>>   static int ehci_usb_probe(struct udevice *dev)
>>   {
>>       struct msm_ehci_priv *p = dev_get_priv(dev);
>> @@ -54,6 +56,7 @@ static int ehci_usb_probe(struct udevice *dev)
>>       struct usb_plat *plat = dev_get_plat(dev);
>>       struct ehci_hccr *hccr;
>>       struct ehci_hcor *hcor;
>> +    struct udevice *phydev;
>>       int ret;
>>       ret = clk_get_by_name(dev, "core", &p->core_clk);
>> @@ -80,6 +83,10 @@ static int ehci_usb_probe(struct udevice *dev)
>>       hcor = (struct ehci_hcor *)((phys_addr_t)hccr +
>>               HC_LENGTH(ehci_readl(&(hccr)->cr_capbase)));
>> +    ret = uclass_get_device_by_name(UCLASS_PHY, "msm8916_usbphy", 
>> &phydev);
> 
> Thanks for looking into this.
> 
> Caleb, do you know why generic_setup_phy() started failing now ? Can it 
> be fixed to handle this platform instead ?

I'm not sure, but I do agree that it doesn't seem like this patch 
addresses the root cause.

Jianfeng: could you run the usb reset command with log level 8 and paste 
a log? That might reveal where the FDT lookup goes wrong.

Kind regards,


-- 
// Caleb (they/them)