[1/1] usb: gadget: g_dnl: Use serial# variable to set g_dnl_serial

Message ID 20170831182004.24946-1-semen.protsenko@linaro.org
State New
Headers show
Series
  • [1/1] usb: gadget: g_dnl: Use serial# variable to set g_dnl_serial
Related show

Commit Message

Sam Protsenko Aug. 31, 2017, 6:20 p.m.
Since 842778a09104 commit, "fastboot devices" stopped to show correct
device serial number for TI boards, showing this line instead:

    ????????????	fastboot

This is because serial# env variable could be set after g_dnl gadget was
initialized (e.g. by using env_set() in the board file).

To fix this, before checking g_dnl_serial, let's re-check if we have
valid serial# value. And if so, let's set it as g_dnl_serial value.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
---
 drivers/usb/gadget/g_dnl.c | 9 +++++++++
 1 file changed, 9 insertions(+)

Comments

Sam Protsenko Aug. 31, 2017, 6:45 p.m. | #1
On 31 August 2017 at 21:20, Sam Protsenko <semen.protsenko@linaro.org> wrote:
> Since 842778a09104 commit, "fastboot devices" stopped to show correct
> device serial number for TI boards, showing this line instead:
>
>     ????????????        fastboot
>
> This is because serial# env variable could be set after g_dnl gadget was
> initialized (e.g. by using env_set() in the board file).
>
> To fix this, before checking g_dnl_serial, let's re-check if we have
> valid serial# value. And if so, let's set it as g_dnl_serial value.
>
> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
> ---
>  drivers/usb/gadget/g_dnl.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
>
> diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
> index 0491a0eea9..e4d0289757 100644
> --- a/drivers/usb/gadget/g_dnl.c
> +++ b/drivers/usb/gadget/g_dnl.c
> @@ -226,6 +226,15 @@ static int g_dnl_bind(struct usb_composite_dev *cdev)
>
>         g_dnl_bind_fixup(&device_desc, cdev->driver->name);
>
> +       /* First try to obtain serial number from serial# variable */
> +       if (strlen(g_dnl_serial) == 0) {
> +               const char *s;
> +
> +               s = env_get("serial#");
> +               if (s)
> +                       g_dnl_set_serialnumber((char *)s);
> +       }
> +
>         if (strlen(g_dnl_serial)) {
>                 id = usb_string_id(cdev);
>                 if (id < 0)
> --
> 2.14.1
>

Similar commit: [1]. But I'm not sure that implementing
g_dnl_bind_fixup() in board file is the best way to fix it.

[1] http://git.denx.de/?p=u-boot.git;a=commit;h=e91ead868b536d0a0dc49aa48f4d8c80c1a94b79
Lukasz Majewski Aug. 31, 2017, 8:24 p.m. | #2
On 08/31/2017 08:45 PM, Sam Protsenko wrote:
> On 31 August 2017 at 21:20, Sam Protsenko <semen.protsenko@linaro.org> wrote:
>> Since 842778a09104 commit, "fastboot devices" stopped to show correct
>> device serial number for TI boards, showing this line instead:
>>
>>     ????????????        fastboot
>>
>> This is because serial# env variable could be set after g_dnl gadget was
>> initialized (e.g. by using env_set() in the board file).
>>
>> To fix this, before checking g_dnl_serial, let's re-check if we have
>> valid serial# value. And if so, let's set it as g_dnl_serial value.
>>
>> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
>> ---
>>  drivers/usb/gadget/g_dnl.c | 9 +++++++++
>>  1 file changed, 9 insertions(+)
>>
>> diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
>> index 0491a0eea9..e4d0289757 100644
>> --- a/drivers/usb/gadget/g_dnl.c
>> +++ b/drivers/usb/gadget/g_dnl.c
>> @@ -226,6 +226,15 @@ static int g_dnl_bind(struct usb_composite_dev *cdev)
>>
>>         g_dnl_bind_fixup(&device_desc, cdev->driver->name);
>>
>> +       /* First try to obtain serial number from serial# variable */
>> +       if (strlen(g_dnl_serial) == 0) {
>> +               const char *s;
>> +
>> +               s = env_get("serial#");
>> +               if (s)
>> +                       g_dnl_set_serialnumber((char *)s);
>> +       }
>> +
>>         if (strlen(g_dnl_serial)) {
>>                 id = usb_string_id(cdev);
>>                 if (id < 0)
>> --
>> 2.14.1
>>
>
> Similar commit: [1]. But I'm not sure that implementing
> g_dnl_bind_fixup() in board file is the best way to fix it.
>
> [1] http://git.denx.de/?p=u-boot.git;a=commit;h=e91ead868b536d0a0dc49aa48f4d8c80c1a94b79
>

I would opt for having g_dnl_bind_fixup() since it is generally used for 
tweaking board specific data to gadget subsystem (like vid, pid, serial 
number).
Marek Vasut Sept. 1, 2017, 8:19 a.m. | #3
On 08/31/2017 08:20 PM, Sam Protsenko wrote:
> Since 842778a09104 commit, "fastboot devices" stopped to show correct
> device serial number for TI boards, showing this line instead:
> 
>     ????????????	fastboot
> 
> This is because serial# env variable could be set after g_dnl gadget was
> initialized (e.g. by using env_set() in the board file).
> 
> To fix this, before checking g_dnl_serial, let's re-check if we have
> valid serial# value. And if so, let's set it as g_dnl_serial value.
> 
> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>

Can't we use U_BOOT_ENV_CALLBACK() here instead and set the serial
number when the environment variable is set ?

> ---
>  drivers/usb/gadget/g_dnl.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
> index 0491a0eea9..e4d0289757 100644
> --- a/drivers/usb/gadget/g_dnl.c
> +++ b/drivers/usb/gadget/g_dnl.c
> @@ -226,6 +226,15 @@ static int g_dnl_bind(struct usb_composite_dev *cdev)
>  
>  	g_dnl_bind_fixup(&device_desc, cdev->driver->name);
>  
> +	/* First try to obtain serial number from serial# variable */
> +	if (strlen(g_dnl_serial) == 0) {
> +		const char *s;
> +
> +		s = env_get("serial#");
> +		if (s)
> +			g_dnl_set_serialnumber((char *)s);
> +	}
> +
>  	if (strlen(g_dnl_serial)) {
>  		id = usb_string_id(cdev);
>  		if (id < 0)
>
Lukasz Majewski Sept. 1, 2017, 8:36 a.m. | #4
On 09/01/2017 10:19 AM, Marek Vasut wrote:
> On 08/31/2017 08:20 PM, Sam Protsenko wrote:
>> Since 842778a09104 commit, "fastboot devices" stopped to show correct
>> device serial number for TI boards, showing this line instead:
>>
>>     ????????????	fastboot
>>
>> This is because serial# env variable could be set after g_dnl gadget was
>> initialized (e.g. by using env_set() in the board file).
>>
>> To fix this, before checking g_dnl_serial, let's re-check if we have
>> valid serial# value. And if so, let's set it as g_dnl_serial value.
>>
>> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
>
> Can't we use U_BOOT_ENV_CALLBACK() here instead and set the serial
> number when the environment variable is set ?

Hmm... This might work.

However, I'm a bit concerned with having "bound" the one particular env 
variable ("serial#" in this case) to generic g_dnl.c code.


>
>> ---
>>  drivers/usb/gadget/g_dnl.c | 9 +++++++++
>>  1 file changed, 9 insertions(+)
>>
>> diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
>> index 0491a0eea9..e4d0289757 100644
>> --- a/drivers/usb/gadget/g_dnl.c
>> +++ b/drivers/usb/gadget/g_dnl.c
>> @@ -226,6 +226,15 @@ static int g_dnl_bind(struct usb_composite_dev *cdev)
>>
>>  	g_dnl_bind_fixup(&device_desc, cdev->driver->name);
>>
>> +	/* First try to obtain serial number from serial# variable */
>> +	if (strlen(g_dnl_serial) == 0) {
>> +		const char *s;
>> +
>> +		s = env_get("serial#");
>> +		if (s)
>> +			g_dnl_set_serialnumber((char *)s);
>> +	}
>> +
>>  	if (strlen(g_dnl_serial)) {
>>  		id = usb_string_id(cdev);
>>  		if (id < 0)
>>
>
>
Marek Vasut Sept. 1, 2017, 9:51 a.m. | #5
On 09/01/2017 10:36 AM, Łukasz Majewski wrote:
> On 09/01/2017 10:19 AM, Marek Vasut wrote:
>> On 08/31/2017 08:20 PM, Sam Protsenko wrote:
>>> Since 842778a09104 commit, "fastboot devices" stopped to show correct
>>> device serial number for TI boards, showing this line instead:
>>>
>>>     ????????????    fastboot
>>>
>>> This is because serial# env variable could be set after g_dnl gadget was
>>> initialized (e.g. by using env_set() in the board file).
>>>
>>> To fix this, before checking g_dnl_serial, let's re-check if we have
>>> valid serial# value. And if so, let's set it as g_dnl_serial value.
>>>
>>> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
>>
>> Can't we use U_BOOT_ENV_CALLBACK() here instead and set the serial
>> number when the environment variable is set ?
> 
> Hmm... This might work.
> 
> However, I'm a bit concerned with having "bound" the one particular env
> variable ("serial#" in this case) to generic g_dnl.c code.
That's a separate problem though, we already have this serial# variable
in place.
Lukasz Majewski Sept. 1, 2017, 10:09 a.m. | #6
On 09/01/2017 11:51 AM, Marek Vasut wrote:
> On 09/01/2017 10:36 AM, Łukasz Majewski wrote:
>> On 09/01/2017 10:19 AM, Marek Vasut wrote:
>>> On 08/31/2017 08:20 PM, Sam Protsenko wrote:
>>>> Since 842778a09104 commit, "fastboot devices" stopped to show correct
>>>> device serial number for TI boards, showing this line instead:
>>>>
>>>>     ????????????    fastboot
>>>>
>>>> This is because serial# env variable could be set after g_dnl gadget was
>>>> initialized (e.g. by using env_set() in the board file).
>>>>
>>>> To fix this, before checking g_dnl_serial, let's re-check if we have
>>>> valid serial# value. And if so, let's set it as g_dnl_serial value.
>>>>
>>>> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
>>>
>>> Can't we use U_BOOT_ENV_CALLBACK() here instead and set the serial
>>> number when the environment variable is set ?
>>
>> Hmm... This might work.
>>
>> However, I'm a bit concerned with having "bound" the one particular env
>> variable ("serial#" in this case) to generic g_dnl.c code.
> That's a separate problem though, we already have this serial# variable
> in place.

Do you mean that it is the same of "fixed" variable, as "ipaddr" is for 
network subsytem?

>
Sam Protsenko Sept. 1, 2017, 11:58 a.m. | #7
On 1 September 2017 at 11:19, Marek Vasut <marex@denx.de> wrote:
> On 08/31/2017 08:20 PM, Sam Protsenko wrote:
>> Since 842778a09104 commit, "fastboot devices" stopped to show correct
>> device serial number for TI boards, showing this line instead:
>>
>>     ????????????      fastboot
>>
>> This is because serial# env variable could be set after g_dnl gadget was
>> initialized (e.g. by using env_set() in the board file).
>>
>> To fix this, before checking g_dnl_serial, let's re-check if we have
>> valid serial# value. And if so, let's set it as g_dnl_serial value.
>>
>> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
>
> Can't we use U_BOOT_ENV_CALLBACK() here instead and set the serial
> number when the environment variable is set ?
>

Good point. I've tested this approach and it works fine. Will resend
v2 shortly. Thanks for the hint!

>> ---
>>  drivers/usb/gadget/g_dnl.c | 9 +++++++++
>>  1 file changed, 9 insertions(+)
>>
>> diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
>> index 0491a0eea9..e4d0289757 100644
>> --- a/drivers/usb/gadget/g_dnl.c
>> +++ b/drivers/usb/gadget/g_dnl.c
>> @@ -226,6 +226,15 @@ static int g_dnl_bind(struct usb_composite_dev *cdev)
>>
>>       g_dnl_bind_fixup(&device_desc, cdev->driver->name);
>>
>> +     /* First try to obtain serial number from serial# variable */
>> +     if (strlen(g_dnl_serial) == 0) {
>> +             const char *s;
>> +
>> +             s = env_get("serial#");
>> +             if (s)
>> +                     g_dnl_set_serialnumber((char *)s);
>> +     }
>> +
>>       if (strlen(g_dnl_serial)) {
>>               id = usb_string_id(cdev);
>>               if (id < 0)
>>
>
>
> --
> Best regards,
> Marek Vasut
Sam Protsenko Sept. 1, 2017, 12:02 p.m. | #8
On 1 September 2017 at 14:58, Sam Protsenko <semen.protsenko@linaro.org> wrote:
> On 1 September 2017 at 11:19, Marek Vasut <marex@denx.de> wrote:
>> On 08/31/2017 08:20 PM, Sam Protsenko wrote:
>>> Since 842778a09104 commit, "fastboot devices" stopped to show correct
>>> device serial number for TI boards, showing this line instead:
>>>
>>>     ????????????      fastboot
>>>
>>> This is because serial# env variable could be set after g_dnl gadget was
>>> initialized (e.g. by using env_set() in the board file).
>>>
>>> To fix this, before checking g_dnl_serial, let's re-check if we have
>>> valid serial# value. And if so, let's set it as g_dnl_serial value.
>>>
>>> Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
>>
>> Can't we use U_BOOT_ENV_CALLBACK() here instead and set the serial
>> number when the environment variable is set ?
>>
>
> Good point. I've tested this approach and it works fine. Will resend
> v2 shortly. Thanks for the hint!
>

Sent v2, using U_BOOT_ENV_CALLBACK: [1]. Guys, please review.

[1] https://lists.denx.de/pipermail/u-boot/2017-September/304460.html

>>> ---
>>>  drivers/usb/gadget/g_dnl.c | 9 +++++++++
>>>  1 file changed, 9 insertions(+)
>>>
>>> diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
>>> index 0491a0eea9..e4d0289757 100644
>>> --- a/drivers/usb/gadget/g_dnl.c
>>> +++ b/drivers/usb/gadget/g_dnl.c
>>> @@ -226,6 +226,15 @@ static int g_dnl_bind(struct usb_composite_dev *cdev)
>>>
>>>       g_dnl_bind_fixup(&device_desc, cdev->driver->name);
>>>
>>> +     /* First try to obtain serial number from serial# variable */
>>> +     if (strlen(g_dnl_serial) == 0) {
>>> +             const char *s;
>>> +
>>> +             s = env_get("serial#");
>>> +             if (s)
>>> +                     g_dnl_set_serialnumber((char *)s);
>>> +     }
>>> +
>>>       if (strlen(g_dnl_serial)) {
>>>               id = usb_string_id(cdev);
>>>               if (id < 0)
>>>
>>
>>
>> --
>> Best regards,
>> Marek Vasut

Patch

diff --git a/drivers/usb/gadget/g_dnl.c b/drivers/usb/gadget/g_dnl.c
index 0491a0eea9..e4d0289757 100644
--- a/drivers/usb/gadget/g_dnl.c
+++ b/drivers/usb/gadget/g_dnl.c
@@ -226,6 +226,15 @@  static int g_dnl_bind(struct usb_composite_dev *cdev)
 
 	g_dnl_bind_fixup(&device_desc, cdev->driver->name);
 
+	/* First try to obtain serial number from serial# variable */
+	if (strlen(g_dnl_serial) == 0) {
+		const char *s;
+
+		s = env_get("serial#");
+		if (s)
+			g_dnl_set_serialnumber((char *)s);
+	}
+
 	if (strlen(g_dnl_serial)) {
 		id = usb_string_id(cdev);
 		if (id < 0)