In message <[email protected]>, Mar
tin Blix Grydeland writes:
I may have asked this before, but why not simply make a
local VTAILQ_HEAD on the stack and move everything there in one
iteration ?
Is it to not block the mutex'es too long ?
Poul-Henning
PS: This email is _also_ a sort of test of patchwork :-)
>+#define NUKEBUF 10
>+
>+void
>+EXP_NukeLRU(struct worker *wrk, struct vsl_log *vsl, struct lru *lru)
>+{
>+ struct objcore *oc;
>+ struct objcore *oc_array[NUKEBUF];
>+ struct object *o;
>+ int i, n;
>+ double t;
>+
>+ CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
>+ CHECK_OBJ_NOTNULL(lru, LRU_MAGIC);
>+
>+ t = VTIM_real();
>+ Lck_Lock(&lru->mtx);
>+ while (!VTAILQ_EMPTY(&lru->lru_head)) {
>+ Lck_Lock(&exp_mtx);
>+ n = 0;
>+ while (n < NUKEBUF) {
>+ oc = VTAILQ_FIRST(&lru->lru_head);
>+ if (oc == NULL)
>+ break;
>+ CHECK_OBJ_NOTNULL(oc, OBJCORE_MAGIC);
>+ assert(oc_getlru(oc) == lru);
>+
>+ /* Remove from the LRU and binheap */
>+ VTAILQ_REMOVE(&lru->lru_head, oc, lru_list);
>+ assert(oc->timer_idx != BINHEAP_NOIDX);
>+ binheap_delete(exp_heap, oc->timer_idx);
>+ assert(oc->timer_idx == BINHEAP_NOIDX);
>+
>+ oc_array[n++] = oc;
>+ VSC_C_main->n_lru_nuked++;
>+ }
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
[email protected] | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
_______________________________________________
varnish-dev mailing list
[email protected]
https://www.varnish-cache.org/lists/mailman/listinfo/varnish-dev
