Currently the performance application creates a device session per lcore.
This was not reflected in how many session objects were available
in the mempool, when using a scheduler device.
The number of lcores is now taken into consideration when calculating
how many sessions are needed for the scheduler.
Fixes: 09fcf99dcad2 ("test/crypto-perf: fix number of scheduler sessions")
Reported-by: Kevin O'Sullivan <kevin.osulli...@intel.com>
Signed-off-by: Ciara Power <ciara.po...@intel.com>
---
app/test-crypto-perf/main.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/app/test-crypto-perf/main.c b/app/test-crypto-perf/main.c
index 0fe47c6caa..22f6b2b3ba 100644
--- a/app/test-crypto-perf/main.c
+++ b/app/test-crypto-perf/main.c
@@ -269,8 +269,8 @@ cperf_initialize_cryptodev(struct cperf_options *opts,
uint8_t *enabled_cdevs)
uint32_t nb_slaves =
rte_cryptodev_scheduler_workers_get(cdev_id,
NULL);
- /* scheduler session header + 1 session per worker qp */
- sessions_needed = 1 + enabled_cdev_count *
+ /* scheduler session header per lcore + 1 session per
worker qp */
+ sessions_needed = nb_lcores + enabled_cdev_count *
opts->nb_qps * nb_slaves;
#endif
} else
--
2.34.1
