| Message ID | 20190217175901.3986-1-pados@pados.hu (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] USB: serial: cp210x: Fix GPIO in autosuspend | expand |
On So, 2019-02-17 at 18:59 +0100, Karoly Pados wrote: > Current GPIO code in cp210x fails to take USB autosuspend into account, > making it practically impossible to use GPIOs with autosuspend enabled > without user configuration. Fix this like for ftdi_sio in a previous patch. > Tested on a CP2102N. Hi, your patch is looking good to me, but I am afraid there are issues. How do the GPIO lines on the device interact with USB reset and system suspend? Regards Oliver
On Sun, Feb 17, 2019 at 06:59:01PM +0100, Karoly Pados wrote: > Current GPIO code in cp210x fails to take USB autosuspend into account, > making it practically impossible to use GPIOs with autosuspend enabled > without user configuration. Fix this like for ftdi_sio in a previous patch. > Tested on a CP2102N. > > Signed-off-by: Karoly Pados <pados@pados.hu> > --- > Changelog: > - v2: Restrict new autopm calls to GPIO paths. > Always check result of usb_autopm_get. > Rebased to current usb_serial upstream. Thanks for the v2. Now applied after removing the unmotivated white space change. I also added a Fixes and CC-stable tag as this seems annoying enough to warrant a backport even if autosuspend was never actually supported. Johan
On Mon, Feb 18, 2019 at 10:17:12AM +0100, Oliver Neukum wrote: > On So, 2019-02-17 at 18:59 +0100, Karoly Pados wrote: > > Current GPIO code in cp210x fails to take USB autosuspend into account, > > making it practically impossible to use GPIOs with autosuspend enabled > > without user configuration. Fix this like for ftdi_sio in a previous patch. > > Tested on a CP2102N. > your patch is looking good to me, but I am afraid there are issues. > How do the GPIO lines on the device interact with USB reset and system > suspend? What was your concern here, Oliver? Johan
On Di, 2019-03-19 at 11:36 +0100, Johan Hovold wrote: > On Mon, Feb 18, 2019 at 10:17:12AM +0100, Oliver Neukum wrote: > > On So, 2019-02-17 at 18:59 +0100, Karoly Pados wrote: > > > Current GPIO code in cp210x fails to take USB autosuspend into account, > > > making it practically impossible to use GPIOs with autosuspend enabled > > > without user configuration. Fix this like for ftdi_sio in a previous patch. > > > Tested on a CP2102N. > > your patch is looking good to me, but I am afraid there are issues. > > How do the GPIO lines on the device interact with USB reset and system > > suspend? > > What was your concern here, Oliver? If you have a device resetting or losing power (for reset_resume) the GPIO lines will revert to the default. But that change is not reported to user space, is it? So the original patch is correct, but there are more situations rather than suspend which could trigger the problem. Regards Oliver
On Tue, Mar 19, 2019 at 12:15:06PM +0100, Oliver Neukum wrote: > On Di, 2019-03-19 at 11:36 +0100, Johan Hovold wrote: > > On Mon, Feb 18, 2019 at 10:17:12AM +0100, Oliver Neukum wrote: > > > On So, 2019-02-17 at 18:59 +0100, Karoly Pados wrote: > > > > Current GPIO code in cp210x fails to take USB autosuspend into account, > > > > making it practically impossible to use GPIOs with autosuspend enabled > > > > without user configuration. Fix this like for ftdi_sio in a previous patch. > > > > Tested on a CP2102N. > > > your patch is looking good to me, but I am afraid there are issues. > > > How do the GPIO lines on the device interact with USB reset and system > > > suspend? > > > > What was your concern here, Oliver? > > If you have a device resetting or losing power (for reset_resume) the > GPIO lines will revert to the default. But that change is not reported > to user space, is it? This driver doesn't support reset_resume() so that shouldn't be an issue, right? We'd disconnect (deregister the gpiochip) and re-probe instead. > So the original patch is correct, but there are more situations rather > than suspend which could trigger the problem. This patch just added the missing auto-resume handling, but, yeah, there may be devices out there for which things may get out of sync if they lose state over suspend. I think Karoly confirmed this wasn't the case with cp210x. Johan
Hi Oliver, USB suspend does not turn off power. Your concerns about the situations you list are valid, but my patch is for another problem. Not for what happens when power is lost, but because the user couldn't use the GPIO lines even if power was on and continous, due to the kernel automatically putting the device into suspend and then all usb messages failed in this case. Karoly March 19, 2019 12:26 PM, "Oliver Neukum" <oneukum@suse.com> wrote: > On Di, 2019-03-19 at 11:36 +0100, Johan Hovold wrote: > >> On Mon, Feb 18, 2019 at 10:17:12AM +0100, Oliver Neukum wrote: >> On So, 2019-02-17 at 18:59 +0100, Karoly Pados wrote: >>> Current GPIO code in cp210x fails to take USB autosuspend into account, >>> making it practically impossible to use GPIOs with autosuspend enabled >>> without user configuration. Fix this like for ftdi_sio in a previous patch. >>> Tested on a CP2102N. >> your patch is looking good to me, but I am afraid there are issues. >> How do the GPIO lines on the device interact with USB reset and system >> suspend? >> >> What was your concern here, Oliver? > > If you have a device resetting or losing power (for reset_resume) the > GPIO lines will revert to the default. But that change is not reported > to user space, is it? > So the original patch is correct, but there are more situations rather > than suspend which could trigger the problem. > > Regards > Oliver
>> On Di, 2019-03-19 at 11:36 +0100, Johan Hovold wrote: >> On Mon, Feb 18, 2019 at 10:17:12AM +0100, Oliver Neukum wrote: >>> On So, 2019-02-17 at 18:59 +0100, Karoly Pados wrote: >>>> Current GPIO code in cp210x fails to take USB autosuspend into account, >>>> making it practically impossible to use GPIOs with autosuspend enabled >>>> without user configuration. Fix this like for ftdi_sio in a previous patch. >>>> Tested on a CP2102N. >>> your patch is looking good to me, but I am afraid there are issues. >>> How do the GPIO lines on the device interact with USB reset and system >>> suspend? >> >> What was your concern here, Oliver? >> >> If you have a device resetting or losing power (for reset_resume) the >> GPIO lines will revert to the default. But that change is not reported >> to user space, is it? > > This driver doesn't support reset_resume() so that shouldn't be an > issue, right? We'd disconnect (deregister the gpiochip) and re-probe > instead. > >> So the original patch is correct, but there are more situations rather >> than suspend which could trigger the problem. > > This patch just added the missing auto-resume handling, but, yeah, > there may be devices out there for which things may get out of sync if > they lose state over suspend. I think Karoly confirmed this wasn't the > case with cp210x. > The way I tested this is very simple: I just toggled an output pin. If the device would lose its state during suspend, then the pin state would return to its factory default as soon the it went into suspend. Since the LED stayed stable over suspend (I connected an LED), it seems these devices don't lose state. I did not, however, test all device state in detail to see if every setting remained. I cannot vouch for whether *all* state is retained.
On Tue, Mar 19, 2019 at 05:19:24PM +0000, Karoly Pados wrote: > >> On Di, 2019-03-19 at 11:36 +0100, Johan Hovold wrote: > > This patch just added the missing auto-resume handling, but, yeah, > > there may be devices out there for which things may get out of sync if > > they lose state over suspend. I think Karoly confirmed this wasn't the > > case with cp210x. > > The way I tested this is very simple: I just toggled an output pin. If > the device would lose its state during suspend, then the pin state > would return to its factory default as soon the it went into suspend. > Since the LED stayed stable over suspend (I connected an LED), it > seems these devices don't lose state. I did not, however, test all > device state in detail to see if every setting remained. I cannot > vouch for whether *all* state is retained. I think your test should be sufficient. Thanks, Johan
diff --git a/drivers/usb/serial/cp210x.c b/drivers/usb/serial/cp210x.c index fac7a4547523..f7aaecad6e21 100644 --- a/drivers/usb/serial/cp210x.c +++ b/drivers/usb/serial/cp210x.c @@ -1370,8 +1370,13 @@ static int cp210x_gpio_get(struct gpio_chip *gc, unsigned int gpio) if (priv->partnum == CP210X_PARTNUM_CP2105) req_type = REQTYPE_INTERFACE_TO_HOST; + result = usb_autopm_get_interface(serial->interface); + if (result) + return result; + result = cp210x_read_vendor_block(serial, req_type, CP210X_READ_LATCH, &buf, sizeof(buf)); + usb_autopm_put_interface(serial->interface); if (result < 0) return result; @@ -1392,6 +1397,10 @@ static void cp210x_gpio_set(struct gpio_chip *gc, unsigned int gpio, int value) buf.mask = BIT(gpio); + result = usb_autopm_get_interface(serial->interface); + if (result) + goto out; + if (priv->partnum == CP210X_PARTNUM_CP2105) { result = cp210x_write_vendor_block(serial, REQTYPE_HOST_TO_INTERFACE, @@ -1399,7 +1408,6 @@ static void cp210x_gpio_set(struct gpio_chip *gc, unsigned int gpio, int value) sizeof(buf)); } else { u16 wIndex = buf.state << 8 | buf.mask; - result = usb_control_msg(serial->dev, usb_sndctrlpipe(serial->dev, 0), CP210X_VENDOR_SPECIFIC, @@ -1409,6 +1417,9 @@ static void cp210x_gpio_set(struct gpio_chip *gc, unsigned int gpio, int value) NULL, 0, USB_CTRL_SET_TIMEOUT); } + usb_autopm_put_interface(serial->interface); + +out: if (result < 0) { dev_err(&serial->interface->dev, "failed to set GPIO value: %d\n", result);
Current GPIO code in cp210x fails to take USB autosuspend into account, making it practically impossible to use GPIOs with autosuspend enabled without user configuration. Fix this like for ftdi_sio in a previous patch. Tested on a CP2102N. Signed-off-by: Karoly Pados <pados@pados.hu> --- Changelog: - v2: Restrict new autopm calls to GPIO paths. Always check result of usb_autopm_get. Rebased to current usb_serial upstream. drivers/usb/serial/cp210x.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-)