hrtimers on arm5vtel poss bug

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Anyone played with hrtimers?

I setup a hrtimer to expire in 50,000,000 ns (20 hz), but it returns
much faster than specified.  Changing sample_ns_interval (say to
80,000,000ns or 1,000,000,000) does not scale delta_time as expected.

So I think I may have found a bug in the clock/hrtimers code.

Taking the last pair of OSCR values, (8992-4107)/3,686,400=0.001325141
sec.  This number is the same regardless of sample_ns_interval.  Too
slow to be expiring immediately, but it is consistent.

I have not tested this code against the latest kernel version.

Pertinent info: kernel is stock 2.6.18, with minor patches to work
with the gumstix platform, hrtimers on, HZ=1000, and processor is a
pxa255.

enclosed: sample dmesg output, relevant code from module

cheng tsai


My output from dmesg:
testing timing
system hrtimer resolution is: 0 s, 999891 ns
t:-1059931735
t:-1059926802
t:-1059921915
t:-1059917039
t:-1059912174
t:-1059907294
t:-1059902416
t:-1059897549
t:-1059892635
t:-1059887760
t:-1059882895
t:-1059878015
t:-1059873133
t:-1059868270
t:-1059863385
t:-1059858512
t:-1059853645
t:-1059848732
t:-1059843856
t:-1059838992
t:-1059834107
target sample_ns_interval was 50000000
hrtimertest: driver seemed to register correctly
hrtimertest: driver unregistered



/* This module tests hrtimers and shows a possible unit conversion
* problem; target requested rate is different from the rate calculated
* by using the free running counter.
*
* problem exists on 2.6.18. Need to test against the newest version of
* the kernel.
*/
#include <linux/init.h>
#include <linux/module.h>
#include <linux/errno.h>
#include <asm/arch/hardware.h>  //needed for __reg
#include <asm/arch/pxa-regs.h>  //needed for OSCR
#include <linux/hrtimer.h>
#include <linux/ktime.h>
#include <linux/time.h>
#define MODULE_NAME "hrtimertest"
#define DRVNAME  "hrtimertest"
#define MAX 20
struct hrtimer timer;
unsigned char cnt;
static unsigned long sample_ns_interval = 50000000;  //50000000=20hz
module_param( sample_ns_interval, long, S_IRUGO);
MODULE_PARM_DESC( sample_ns_interval, "Desired sample rate.");
int recall_me(struct hrtimer* hrtimer_p){
static int32_t temp;
// OSCR is the free running counter on ARM5vtel, actual time elapsed
// can be calculated from this by taking the delta, and dividing by
// 3.6864 Mhz
// I know this is a hackish way to check actual time.
temp=OSCR;
printk("t:%d\n",temp);
if(cnt<MAX){
  hrtimer_restart(hrtimer_p);
  cnt++;
 }
return 0;
}
static int __init hrtimertest_init(void){
struct timespec tp;
ktime_t mytime;

mytime.tv.sec=0;
mytime.tv.nsec=sample_ns_interval;
cnt=0;
printk("testing timing\n");

hrtimer_init(&timer,CLOCK_MONOTONIC,HRTIMER_ABS);
timer.function=recall_me;
hrtimer_start(&timer, mytime, HRTIMER_ABS);

hrtimer_get_res(CLOCK_MONOTONIC, &tp);

printk( "system hrtimer resolution is: %d s, %d ns\n", tp.tv_sec,tp.tv_nsec);
printk( "target sample_ns_interval was %d\n", sample_ns_interval);
printk( KERN_ERR MODULE_NAME ": driver seemed to register correctly\n");

return 0;
}
module_init(hrtimertest_init);
static void __exit hrtimertest_exit(void){
hrtimer_cancel(&timer);

printk( KERN_ERR MODULE_NAME ": driver unregistered \n");
return;
}
module_exit(hrtimertest_exit);
MODULE_DESCRIPTION("hrtimertest strap");
MODULE_AUTHOR("Cheng Tsai, <cheng705 at gmail dot com>");
MODULE_LICENSE("GPL v2");
MODULE_VERSION("$Date: 2006/11/22 21:43:03 $");
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[Index of Archives]     [Kernel Newbies]     [Netfilter]     [Bugtraq]     [Photo]     [Stuff]     [Gimp]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Video 4 Linux]     [Linux for the blind]     [Linux Resources]
  Powered by Linux