castet.matthieu at free.fr
Sat Mar 19 13:38:12 EST 2005
Roman Kagan wrote:
> On Mon, Mar 14, 2005 at 03:01:05PM +0100, castet.matthieu at free.fr wrote:
>>>I'm a bit confused here. Do you mean that "failed" is considered a
>>>transient error happening every now and then during modem operation,
>>>which needs graceful recovery (reboot)?
>>Yes failed can happen in case the modem lost synchronisation or other problems.
>>There also other stat info (los,crc error) which need to reboot the modem.
> Is crc error or synchronization loss is such a disaster that requires
> complete reboot of the modem?
That's how it is implemented, and without the spec I can't tell more.
>>>If the modem has to be rebooted from time to time during normal
>>>operation (grrrr...) I suggest not to use heavy_init, as you need a
>>>reboot hook all the time.
>>No need to have a reboot hook, because we don't need to stop usbatm when
>>rebooting : it is independant : there is a part of the driver for atm/usb and a
>>part for the state machine of the driver (firmware + stats)
> That's exactly what I mean: if you expect a need to reboot the modem
> during normal operation, you can run the booting routine at any time.
> This means that you can consider your modem ready when the usb probe
> succeeded, and don't need heavy_init at all. Your booting routine will
> run as part of the normal modem operation, and ATM device will be
> available at that time.
No you were right : the modem sometimes lock if we sumit data urb when
(re)booting :(, and there is no way to detect this lock in the driver as
everything seem normal. The solution could to not submit urb when
I don't know if this only apply for the send endpoint which is share
with the firmware uploading endpoint or for both.
More information about the Usbatm