Resume issue after a longer sleep

Andy Green andy at
Mon May 19 11:10:49 CEST 2008

Hash: SHA1

Somebody in the thread at some point said:
| On Sunday 18 May 2008 14:13:33 Andy Green wrote:
|> Holger I found there are some funny wake sources enabled for 50633, I
|> will send a patch to reduce them to those that are useful.
|> Are you testing with USB cable plugged in there are powered?  If so it
|> can maybe allow some unusual wake stimulus like BATFULL, which occur on
|> say 30s period and you are guaranteed to get one if you suspend > 30s.
|> Batfull doesn't crash me here but maybe it is something else along those
|> lines since we die in pcf50633.
| Ah interesting. I have only seen wakeup on USB cable removal so maybe
| the "modem wakeup interrupt" in dmesg is just a red herring?

GSM bits and bobs are unique in they are powered pretty direct from the
battery.  You can't run GSM logic without a battery in GTA02, USB cable
isn't enough.  But I guess you must run with battery in if you wake on
USB cable removal.  I run here without battery most of the time because
the thing is out of its case most of the time :-)  I push a battery on
there to get around some U-Boot issue where at one point early in boot
it seems to require a battery around.

| Regarding my suspend/resume issue:
| 	- Your addition of "suspend" to the i2c s3c driver has a possible NULL
| dereference but we don't hit that.

Don't be shy, where is it?

| 	- the i2c s3c driver is resumed before the pcf50633

Has to be according to the suspend / resume rules since it is on that bus.

| 	- It is hanging in __reg_write even before the _xfer of the i2c s3c is
| getting called (which means i2c_transfer is not called) which can't be...

How would we know that _xfer is called?  Maybe the notification is
asynchronous somehow, when we panic it is never issued then even though
it happened.

Whose __reg_write() is it?  Definitely the one at the i2c layer itself?
~ Because it is a popular name.

| 	- I enabled the CRC on the RAM and know I stopped crashing on me...
so it
| might be a timing issue?

Wah, lovely some horrible race somewhere no doubt (like the others that
were found in this area).

- -Andy
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora -


More information about the openmoko-kernel mailing list