Real Time in 2.4.18

[Amosov Evgeniy]

Good day.
I am working on improved Real-Time support in Linux for kernel 2.4.18 ( i388, sparc).
I have  written  my own POSIX  pthreads support in kernel, library as an interface for use the new systemcall (like libpthread)  and  measure  the time of switching two pthreads ( pthread_bcast - pthread_wait  circle in two threads)
I have done some changes in schedule,  disabled BH and  intercepted  all interrupts and exeptions in entry.S.

All looks fine, the time of switching is about 3mks, the maximum time is 20 mks,  but sometimes ...., after about 20 seconds, I see a 
single surge about 300mks.
I have no idea what is a source of this surge.
There are only my two switching SHED_FIFO tasks  and standart kernel threads  in the system(keventd, kswapd, bdflush etc...),  but during the test in RUNNING state were only my two pthreads; there are no interrupts or exeptions during the test,
all BHs is disabled, but the measurements show that it is an interrupt becouse  my tasks preempt in random places   for a long time.
It is not a timer_interrupt.

It is strange:  no interrupts(exept timer ofcouse, but it is not it) but it looks like an interrupt :)

I intercept interrupts in irq.c:do_IRQ(),   entry.S: system_call() and exeptions in entry.S: error_code.
May be I something lose ?
Have anybody some ideas?

P.S.
Sorry for my english :)


-- 
Amosov Evgeniy (aka Efreet)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/