| Message ID | 1388833760-23260-1-git-send-email-m.chehab@samsung.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Did you trace the i2c messages on the bus? This seems like papering the actual bug. USB 3.0 is a disaster with Linux, maybe your hardware or your controller driver is not okay? There are other bugreports out there which are USB 3.0 related, some of our customers reported that since 3.6.0 is okay while 3.7.10 give them a complete system lock up also with the driver in question here. On Sat, Jan 4, 2014 at 12:09 PM, Mauro Carvalho Chehab <m.chehab@samsung.com> wrote: > At least on HVR-950, sometimes an I2C operation fails. > > This seems to be more frequent when the device is connected > into an USB 3.0 port. > > Instead of report an error, try to repeat it, for up to > 20 ms. That makes the code more reliable. > > Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> > --- > drivers/media/usb/em28xx/em28xx-i2c.c | 23 +++++++++++------------ > 1 file changed, 11 insertions(+), 12 deletions(-) > > diff --git a/drivers/media/usb/em28xx/em28xx-i2c.c b/drivers/media/usb/em28xx/em28xx-i2c.c > index 6cd3d909bb3a..35d6808aa9ff 100644 > --- a/drivers/media/usb/em28xx/em28xx-i2c.c > +++ b/drivers/media/usb/em28xx/em28xx-i2c.c > @@ -189,6 +189,7 @@ static int em28xx_i2c_send_bytes(struct em28xx *dev, u16 addr, u8 *buf, > * Zero length reads always succeed, even if no device is connected > */ > > +retry: > /* Write to i2c device */ > ret = dev->em28xx_write_regs_req(dev, stop ? 2 : 3, addr, buf, len); > if (ret != len) { > @@ -208,26 +209,24 @@ static int em28xx_i2c_send_bytes(struct em28xx *dev, u16 addr, u8 *buf, > ret = dev->em28xx_read_reg(dev, 0x05); > if (ret == 0) /* success */ > return len; > - if (ret == 0x10) { > - em28xx_warn("I2C transfer timeout on writing to addr 0x%02x", > - addr); > - return -EREMOTEIO; > - } > + if (ret == 0x10) > + goto retry; > if (ret < 0) { > em28xx_warn("failed to get i2c transfer status from bridge register (error=%i)\n", > ret); > return ret; > } > msleep(5); > - /* > - * NOTE: do we really have to wait for success ? > - * Never seen anything else than 0x00 or 0x10 > - * (even with high payload) ... > - */ > } > > - if (i2c_debug) > - em28xx_warn("write to i2c device at 0x%x timed out\n", addr); > + if (ret == 0x10) { > + if (i2c_debug) > + em28xx_warn("I2C transfer timeout on writing to addr 0x%02x", > + addr); > + } else { > + em28xx_warn("write to i2c device at 0x%x timed out (ret=0x%02x)\n", > + addr, ret); > + } > return -EREMOTEIO; > } > > -- > 1.8.3.1 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-media" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Am 04.01.2014 18:58, schrieb Markus Rechberger: > Did you trace the i2c messages on the bus? This seems like papering > the actual bug. The USB traces are clear: i2c status 0x10 is treated as _final_ i2c transfer status and the driver does _not_ retry. Yes, it's papering over the actual bug. We need to stop this crap and fix the actual bugs instead. And it's wrong i2c adapter driver design. Whether or retrying makes sense (and how often, with which delay) depends on the context the i2c operation is used in. That's why this decision is up to the i2c client driver. Regards, Frank > USB 3.0 is a disaster with Linux, maybe your hardware or your > controller driver is not okay? > There are other bugreports out there which are USB 3.0 related, some > of our customers reported that since 3.6.0 is okay while 3.7.10 give > them a complete system lock up also with the driver in question here. > > > On Sat, Jan 4, 2014 at 12:09 PM, Mauro Carvalho Chehab > <m.chehab@samsung.com> wrote: >> At least on HVR-950, sometimes an I2C operation fails. >> >> This seems to be more frequent when the device is connected >> into an USB 3.0 port. >> >> Instead of report an error, try to repeat it, for up to >> 20 ms. That makes the code more reliable. >> >> Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> >> --- >> drivers/media/usb/em28xx/em28xx-i2c.c | 23 +++++++++++------------ >> 1 file changed, 11 insertions(+), 12 deletions(-) >> >> diff --git a/drivers/media/usb/em28xx/em28xx-i2c.c b/drivers/media/usb/em28xx/em28xx-i2c.c >> index 6cd3d909bb3a..35d6808aa9ff 100644 >> --- a/drivers/media/usb/em28xx/em28xx-i2c.c >> +++ b/drivers/media/usb/em28xx/em28xx-i2c.c >> @@ -189,6 +189,7 @@ static int em28xx_i2c_send_bytes(struct em28xx *dev, u16 addr, u8 *buf, >> * Zero length reads always succeed, even if no device is connected >> */ >> >> +retry: >> /* Write to i2c device */ >> ret = dev->em28xx_write_regs_req(dev, stop ? 2 : 3, addr, buf, len); >> if (ret != len) { >> @@ -208,26 +209,24 @@ static int em28xx_i2c_send_bytes(struct em28xx *dev, u16 addr, u8 *buf, >> ret = dev->em28xx_read_reg(dev, 0x05); >> if (ret == 0) /* success */ >> return len; >> - if (ret == 0x10) { >> - em28xx_warn("I2C transfer timeout on writing to addr 0x%02x", >> - addr); >> - return -EREMOTEIO; >> - } >> + if (ret == 0x10) >> + goto retry; >> if (ret < 0) { >> em28xx_warn("failed to get i2c transfer status from bridge register (error=%i)\n", >> ret); >> return ret; >> } >> msleep(5); >> - /* >> - * NOTE: do we really have to wait for success ? >> - * Never seen anything else than 0x00 or 0x10 >> - * (even with high payload) ... >> - */ >> } >> >> - if (i2c_debug) >> - em28xx_warn("write to i2c device at 0x%x timed out\n", addr); >> + if (ret == 0x10) { >> + if (i2c_debug) >> + em28xx_warn("I2C transfer timeout on writing to addr 0x%02x", >> + addr); >> + } else { >> + em28xx_warn("write to i2c device at 0x%x timed out (ret=0x%02x)\n", >> + addr, ret); >> + } >> return -EREMOTEIO; >> } >> >> -- >> 1.8.3.1 >> >> -- >> To unsubscribe from this list: send the line "unsubscribe linux-media" in >> the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html > -- > To unsubscribe from this list: send the line "unsubscribe linux-usb" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/media/usb/em28xx/em28xx-i2c.c b/drivers/media/usb/em28xx/em28xx-i2c.c index 6cd3d909bb3a..35d6808aa9ff 100644 --- a/drivers/media/usb/em28xx/em28xx-i2c.c +++ b/drivers/media/usb/em28xx/em28xx-i2c.c @@ -189,6 +189,7 @@ static int em28xx_i2c_send_bytes(struct em28xx *dev, u16 addr, u8 *buf, * Zero length reads always succeed, even if no device is connected */ +retry: /* Write to i2c device */ ret = dev->em28xx_write_regs_req(dev, stop ? 2 : 3, addr, buf, len); if (ret != len) { @@ -208,26 +209,24 @@ static int em28xx_i2c_send_bytes(struct em28xx *dev, u16 addr, u8 *buf, ret = dev->em28xx_read_reg(dev, 0x05); if (ret == 0) /* success */ return len; - if (ret == 0x10) { - em28xx_warn("I2C transfer timeout on writing to addr 0x%02x", - addr); - return -EREMOTEIO; - } + if (ret == 0x10) + goto retry; if (ret < 0) { em28xx_warn("failed to get i2c transfer status from bridge register (error=%i)\n", ret); return ret; } msleep(5); - /* - * NOTE: do we really have to wait for success ? - * Never seen anything else than 0x00 or 0x10 - * (even with high payload) ... - */ } - if (i2c_debug) - em28xx_warn("write to i2c device at 0x%x timed out\n", addr); + if (ret == 0x10) { + if (i2c_debug) + em28xx_warn("I2C transfer timeout on writing to addr 0x%02x", + addr); + } else { + em28xx_warn("write to i2c device at 0x%x timed out (ret=0x%02x)\n", + addr, ret); + } return -EREMOTEIO; }
At least on HVR-950, sometimes an I2C operation fails. This seems to be more frequent when the device is connected into an USB 3.0 port. Instead of report an error, try to repeat it, for up to 20 ms. That makes the code more reliable. Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com> --- drivers/media/usb/em28xx/em28xx-i2c.c | 23 +++++++++++------------ 1 file changed, 11 insertions(+), 12 deletions(-)