Re: Seemingly random hangs...

[Robert Erlich <erlichr@ping.ca>]

On Fri, Nov 30, 2001 at 12:03:41AM -0500, erlichr wrote:

> I've been tring to get the knack of the posix API.

Still trying.  Had a pretty good grasp of the old v1 API.  Happy to say I'm making progress though.

> The attached code is influenced by "Frank" example.  The idea is to make a four channel pwm generator to drive motors hung off the port.

The code has evolved a bit since then, and getting an oscilliscope on the outputs helped in debugging timing issues.

This is actually the first step in an attempt to create a closed loop, PID filter servo motor controller.  I'm currently gathering parts for a shaft encoder interface (looking at a pair of LSI LS7266).

> Testing with rtl_printk showed the fifos processed writes correctly and the correct bit patterns were being generated and written to the printer port.

Yup, the rtl_printk's were fine, and it was a stretch to say the patterns were being written to the printer port; I had the arguments in the rtl_outb reversed.  So, the LSB of the printer port address got written to addresses created by correctly generated bit pattern.  D'oh.

>  Where this all seems to fall down is after addition of the code to suspend a thread.  Sporadic hard lockups when writing from user space to the fifos (usually 3, but not allways) and sometimes during rmmod.

Yea, this makes sense now.

> Again, the posix API is still new so I may not be grokking where and how to use mutex locks or thread suspention/wake_up.

Still applies, but I no longer have to deal with nasty fscks increasing my test/debug cycle time.

BTW, is there a patch set for a more recent kernel, I'd really like to be using the Archangeli VM, as well as ext3 or preferably xfs as a filesystem (see above).

> There is no contention for the io space of the printer port, parport_pc and parport modules are unloaded.

Yea, 'cause I'm wasn't actually writing to the printer port ;)

> Any help, advice or research direction would be greatly appreciated.

Thanks to those who replied.  BTW, is there a problem with the list?  I'm not getting traffic from it, or is my subscription messed up.

-R.
-- 
GPG-fingerprint:
6CE9 8B6E 8089 FECC 4AA6  9CC9 F2B3 70A6 E98A 8B77

PGP signature