workqueues and schedule()
From: Davide Rossetti
Date: Wed Feb 16 2005 - 11:05:48 EST
<environment>
I'm authoring a (GPL) driver for a custom 3D network card.
</environment>
on calling some polling functions from within a workqueue, I'm getting
lockups... sysrq-p got it clear I have a workqueue thread down into
schedule_timeout().
is considered polite to call schedule ? is in_softirq() capable to diff
workqueue/normal case ?
static int __apedev_hdrring_poll_end_dma(ApeDev* apedev, struct
ApeHdrRing* ring)
{
int ret = 0;
int counter;
APE_BUG_ON(0 == ring);
APE_BUG_ON(0 == apedev);
// here I poll on the rwbuf memory content till dma is done
ndelay(5*HZPCI);
APEDEV_COUNTER_INC(apedev, hdrring_poll_ndelay_cnt);
counter = 0;
while(1) {
if(0 != __apedev_hdrring_check_magic(apedev, ring))
break;
counter++;
ndelay(20*HZPCI);
if(0 == counter % 512) {
APEDEV_COUNTER_INC(apedev, hdrring_poll_ndelay_cnt);
ndelay(50*HZPCI);
}
if(counter==2048*16) { // time past is 2048*16/512*50HZPCI
PDEBUG("counter overflows 2048*16, delaying 1 jiffy\n");
set_current_state(TASK_UNINTERRUPTIBLE);
schedule_timeout(MAX(HZ/100,1)); //<----
counter = 0;
APEDEV_COUNTER_INC(apedev, hdrring_poll_resched_cnt);
}
if(signal_pending(current)) {
PERROR("GOT SIGNAL!!!\n");
//apedev_v3_dump_regs(apedev);
ret = -EINTR;
goto err;
}
}
err:
return ret;
}
// fast path
static inline int apedev_hdrring_poll_end_dma(ApeDev* apedev, struct
ApeHdrRing* ring)
{
int ret = 0;
int retcode;
APE_BUG_ON(0 == ring);
APE_BUG_ON(0 == apedev);
retcode = __apedev_hdrring_check_magic(apedev, ring);
if(retcode < 0) {
PERROR("error in check_magic\n");
ret = retcode;
// exit with true...
} else if(ret==0) {
ret = __apedev_hdrring_poll_end_dma(apedev, ring);
PDEBUG("ret=%d\n", ret);
}
return ret;
}
-
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/