On Tuesday 28 April 2015 18:39:07 Baolin Wang wrote:
On 28 April 2015 at 17:42, Arnd Bergmann arnd@linaro.org wrote:
if (!new_clock->clock_get) {printk(KERN_WARNING "POSIX clock id %d lacksclock_get()\n",
if (!new_clock->clock_get && !new_clock->clock_get64) {printk(KERN_WARNING "POSIX clock id %d lacks clock_get()and clock_get64()\n",
clock_id); return; }Here you are missing a step that Thomas suggested in the previous review:
add a default clock_get64() implementation that calls clock_get()
But the default_timer_get64((struct k_clock *kc, struct k_itimer *timr, struct itimerspec64 *cur_setting64) function he suggested can't get the "kc" parameter, cause the "timer_get" point prototype is "void (*timer_get64) (struct k_itimer *timr, struct itimerspec64 *cur_setting);".
static int default_timer_get64(struct k_clock *kc, struct k_itimer *timr, struct itimerspec64 *cur_setting64) { struct itimerspec cur_setting;
kc->timer_get(timer, &cur_setting); return 0;}
This is something you should have asked in reply to Thomas' suggestion when you noticed this was a problem.
I think the answer is that you must use clockid_to_kclock(timr->it_clock) to convert from a k_itimer structure to a k_clock structure.
This is slightly inefficient, but it is only done in an intermediate step and gets removed in the final patch/
Arnd