Ping, this patch is still good for most recent api-next branch
Best Regards, Yi
On 23 December 2016 at 10:32, Yi He <yi...@linaro.org> wrote:
> SP scheduler hangs in process mode performance test
> due to global data structure were not created in shared
> memory region.
>
> Signed-off-by: Yi He <yi...@linaro.org>
> ---
> since v1: rebased upon Petri's linux-gen: schedule_sp: use ring as
> priority queue
>
> platform/linux-generic/odp_schedule_sp.c | 100
> ++++++++++++++++++-------------
> 1 file changed, 60 insertions(+), 40 deletions(-)
>
> diff --git a/platform/linux-generic/odp_schedule_sp.c
> b/platform/linux-generic/odp_schedule_sp.c
> index 5150d28..bb7416a 100644
> --- a/platform/linux-generic/odp_schedule_sp.c
> +++ b/platform/linux-generic/odp_schedule_sp.c
> @@ -9,6 +9,7 @@
> #include <odp/api/thread.h>
> #include <odp/api/time.h>
> #include <odp/api/schedule.h>
> +#include <odp/api/shared_memory.h>
> #include <odp_schedule_if.h>
> #include <odp_debug_internal.h>
> #include <odp_align_internal.h>
> @@ -108,6 +109,7 @@ typedef struct {
> sched_cmd_t pktio_cmd[NUM_PKTIO];
> prio_queue_t prio_queue[NUM_GROUP][NUM_PRIO];
> sched_group_t sched_group;
> + odp_shm_t shm;
> } sched_global_t;
>
> typedef struct {
> @@ -119,7 +121,7 @@ typedef struct {
> int group[NUM_GROUP];
> } sched_local_t;
>
> -static sched_global_t sched_global;
> +static sched_global_t *sched_global;
> static __thread sched_local_t sched_local;
>
> static inline uint32_t index_to_ring_idx(int pktio, uint32_t index)
> @@ -145,30 +147,44 @@ static inline uint32_t index_from_ring_idx(uint32_t
> *index, uint32_t ring_idx)
> static int init_global(void)
> {
> int i, j;
> - sched_group_t *sched_group = &sched_global.sched_group;
> + odp_shm_t shm;
> + sched_group_t *sched_group = NULL;
>
> ODP_DBG("Using SP scheduler\n");
>
> - memset(&sched_global, 0, sizeof(sched_global_t));
> + shm = odp_shm_reserve("sp_scheduler",
> + sizeof(sched_global_t),
> + ODP_CACHE_LINE_SIZE, 0);
> +
> + sched_global = odp_shm_addr(shm);
> +
> + if (sched_global == NULL) {
> + ODP_ERR("Schedule init: Shm reserve failed.\n");
> + return -1;
> + }
> +
> + memset(sched_global, 0, sizeof(sched_global_t));
> + sched_global->shm = shm;
>
> for (i = 0; i < NUM_QUEUE; i++) {
> - sched_global.queue_cmd[i].s.type = CMD_QUEUE;
> - sched_global.queue_cmd[i].s.index = i;
> - sched_global.queue_cmd[i].s.ring_idx =
> index_to_ring_idx(0, i);
> + sched_global->queue_cmd[i].s.type = CMD_QUEUE;
> + sched_global->queue_cmd[i].s.index = i;
> + sched_global->queue_cmd[i].s.ring_idx =
> index_to_ring_idx(0, i);
> }
>
> for (i = 0; i < NUM_PKTIO; i++) {
> - sched_global.pktio_cmd[i].s.type = CMD_PKTIO;
> - sched_global.pktio_cmd[i].s.index = i;
> - sched_global.pktio_cmd[i].s.ring_idx =
> index_to_ring_idx(1, i);
> - sched_global.pktio_cmd[i].s.prio = PKTIN_PRIO;
> - sched_global.pktio_cmd[i].s.group = GROUP_PKTIN;
> + sched_global->pktio_cmd[i].s.type = CMD_PKTIO;
> + sched_global->pktio_cmd[i].s.index = i;
> + sched_global->pktio_cmd[i].s.ring_idx =
> index_to_ring_idx(1, i);
> + sched_global->pktio_cmd[i].s.prio = PKTIN_PRIO;
> + sched_global->pktio_cmd[i].s.group = GROUP_PKTIN;
> }
>
> for (i = 0; i < NUM_GROUP; i++)
> for (j = 0; j < NUM_PRIO; j++)
> - ring_init(&sched_global.prio_queue[i][j].ring);
> + ring_init(&sched_global->prio_queue[i][j].ring);
>
> + sched_group = &sched_global->sched_group;
> odp_ticketlock_init(&sched_group->s.lock);
>
> for (i = 0; i < NUM_THREAD; i++)
> @@ -202,16 +218,22 @@ static int init_local(void)
>
> static int term_global(void)
> {
> - int qi;
> + int qi, ret = 0;
>
> for (qi = 0; qi < NUM_QUEUE; qi++) {
> - if (sched_global.queue_cmd[qi].s.init) {
> + if (sched_global->queue_cmd[qi].s.init) {
> /* todo: dequeue until empty ? */
> sched_cb_queue_destroy_finalize(qi);
> }
> }
>
> - return 0;
> + ret = odp_shm_free(sched_global->shm);
> + if (ret < 0) {
> + ODP_ERR("Shm free failed for sp_scheduler");
> + ret = -1;
> + }
> +
> + return ret;
> }
>
> static int term_local(void)
> @@ -267,7 +289,7 @@ static void remove_group(sched_group_t *sched_group,
> int thr, int group)
>
> static int thr_add(odp_schedule_group_t group, int thr)
> {
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
>
> if (group < 0 || group >= NUM_GROUP)
> return -1;
> @@ -292,7 +314,7 @@ static int thr_add(odp_schedule_group_t group, int thr)
>
> static int thr_rem(odp_schedule_group_t group, int thr)
> {
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
>
> if (group < 0 || group >= NUM_GROUP)
> return -1;
> @@ -320,7 +342,7 @@ static int num_grps(void)
>
> static int init_queue(uint32_t qi, const odp_schedule_param_t
> *sched_param)
> {
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
> odp_schedule_group_t group = sched_param->group;
> int prio = 0;
>
> @@ -333,18 +355,18 @@ static int init_queue(uint32_t qi, const
> odp_schedule_param_t *sched_param)
> if (sched_param->prio > 0)
> prio = LOWEST_QUEUE_PRIO;
>
> - sched_global.queue_cmd[qi].s.prio = prio;
> - sched_global.queue_cmd[qi].s.group = group;
> - sched_global.queue_cmd[qi].s.init = 1;
> + sched_global->queue_cmd[qi].s.prio = prio;
> + sched_global->queue_cmd[qi].s.group = group;
> + sched_global->queue_cmd[qi].s.init = 1;
>
> return 0;
> }
>
> static void destroy_queue(uint32_t qi)
> {
> - sched_global.queue_cmd[qi].s.prio = 0;
> - sched_global.queue_cmd[qi].s.group = 0;
> - sched_global.queue_cmd[qi].s.init = 0;
> + sched_global->queue_cmd[qi].s.prio = 0;
> + sched_global->queue_cmd[qi].s.group = 0;
> + sched_global->queue_cmd[qi].s.init = 0;
> }
>
> static inline void add_tail(sched_cmd_t *cmd)
> @@ -354,8 +376,7 @@ static inline void add_tail(sched_cmd_t *cmd)
> int prio = cmd->s.prio;
> uint32_t idx = cmd->s.ring_idx;
>
> - prio_queue = &sched_global.prio_queue[group][prio];
> -
> + prio_queue = &sched_global->prio_queue[group][prio];
> ring_enq(&prio_queue->ring, RING_MASK, idx);
> }
>
> @@ -365,8 +386,7 @@ static inline sched_cmd_t *rem_head(int group, int
> prio)
> uint32_t ring_idx, index;
> int pktio;
>
> - prio_queue = &sched_global.prio_queue[group][prio];
> -
> + prio_queue = &sched_global->prio_queue[group][prio];
> ring_idx = ring_deq(&prio_queue->ring, RING_MASK);
>
> if (ring_idx == RING_EMPTY)
> @@ -375,16 +395,16 @@ static inline sched_cmd_t *rem_head(int group, int
> prio)
> pktio = index_from_ring_idx(&index, ring_idx);
>
> if (pktio)
> - return &sched_global.pktio_cmd[index];
> + return &sched_global->pktio_cmd[index];
>
> - return &sched_global.queue_cmd[index];
> + return &sched_global->queue_cmd[index];
> }
>
> static int sched_queue(uint32_t qi)
> {
> sched_cmd_t *cmd;
>
> - cmd = &sched_global.queue_cmd[qi];
> + cmd = &sched_global->queue_cmd[qi];
> add_tail(cmd);
>
> return 0;
> @@ -410,7 +430,7 @@ static void pktio_start(int pktio_index, int num, int
> pktin_idx[])
> ODP_DBG("pktio index: %i, %i pktin queues %i\n",
> pktio_index, num, pktin_idx[0]);
>
> - cmd = &sched_global.pktio_cmd[pktio_index];
> + cmd = &sched_global->pktio_cmd[pktio_index];
>
> if (num > NUM_PKTIN)
> ODP_ABORT("Supports only %i pktin queues per interface\n",
> @@ -428,7 +448,7 @@ static inline sched_cmd_t *sched_cmd(void)
> {
> int prio, i;
> int thr = sched_local.thr_id;
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
> thr_group_t *thr_group = &sched_group->s.thr[thr];
> uint32_t gen_cnt;
>
> @@ -602,7 +622,7 @@ static odp_schedule_group_t
> schedule_group_create(const char *name,
> const odp_thrmask_t
> *thrmask)
> {
> odp_schedule_group_t group = ODP_SCHED_GROUP_INVALID;
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
> int i;
>
> odp_ticketlock_lock(&sched_group->s.lock);
> @@ -633,7 +653,7 @@ static odp_schedule_group_t
> schedule_group_create(const char *name,
>
> static int schedule_group_destroy(odp_schedule_group_t group)
> {
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
>
> if (group < NUM_STATIC_GROUP || group >= NUM_GROUP)
> return -1;
> @@ -656,7 +676,7 @@ static int schedule_group_destroy(odp_schedule_group_t
> group)
> static odp_schedule_group_t schedule_group_lookup(const char *name)
> {
> odp_schedule_group_t group = ODP_SCHED_GROUP_INVALID;
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
> int i;
>
> odp_ticketlock_lock(&sched_group->s.lock);
> @@ -677,7 +697,7 @@ static int schedule_group_join(odp_schedule_group_t
> group,
> const odp_thrmask_t *thrmask)
> {
> int thr;
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
>
> if (group < 0 || group >= NUM_GROUP)
> return -1;
> @@ -709,7 +729,7 @@ static int schedule_group_leave(odp_schedule_group_t
> group,
> const odp_thrmask_t *thrmask)
> {
> int thr;
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
> odp_thrmask_t *all = &sched_group->s.group[GROUP_ALL].mask;
> odp_thrmask_t not;
>
> @@ -743,7 +763,7 @@ static int schedule_group_leave(odp_schedule_group_t
> group,
> static int schedule_group_thrmask(odp_schedule_group_t group,
> odp_thrmask_t *thrmask)
> {
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
>
> if (group < 0 || group >= NUM_GROUP)
> return -1;
> @@ -765,7 +785,7 @@ static int schedule_group_thrmask(odp_schedule_group_t
> group,
> static int schedule_group_info(odp_schedule_group_t group,
> odp_schedule_group_info_t *info)
> {
> - sched_group_t *sched_group = &sched_global.sched_group;
> + sched_group_t *sched_group = &sched_global->sched_group;
>
> if (group < 0 || group >= NUM_GROUP)
> return -1;
> --
> 2.7.4
>
>
