Skip to contentSkip to conversation

pcrypt_aead01: break early when approaching timeout

PATCH New Jan Stancek 2 participants 3 messages Oct 15, 2019 – Oct 18, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 pcrypt_aead01: break early when approaching timeout

PowerNV (P9 witherspoon, 176CPUs) systems running latest upstream kernel (5.3.6) have been observed to come very close to test timeout and also sporadically fail, because they couldn't complete 10000 iterations in time.

Each iteration of test leads to modprobe of cryptomgr and crypto framework running number of tests, taking up to ~70ms per iteration.

Looking at traces, a significant contributor is wakeup time. After one crypto test is done, it takes ~0.5ms for other tests to resume:
(gettimeofday_us)
1571141551769329 7 kworker/1:1(897):<-crypto_req_done
1571141551769843 1603 cryptomgr_test(3813):

Booting with powersave=off cuts this latency significantly, single iteration of test completes in ~10ms.

But changing kernel defaults isn't very practical solution, hence patch is watching for timeout approaching and breaks the loop early.

---
 testcases/kernel/crypto/pcrypt_aead01.c | 6 ++++++
 1 file changed, 6 insertions(+)
testcases/kernel/crypto/pcrypt_aead01.c +6 -0
index d9c27f69dce7..3dc8c58c2817 100644
@@ -54,6 +54,12 @@ void run(void)
 		TEST(tst_crypto_del_alg(&ses, &a));
 		if (TST_RET)
 			tst_brk(TBROK | TRERRNO, "del_alg");
+
+		if (tst_timeout_remaining() < 10) {
+			tst_res(TINFO, "Time limit reached, stopping@"
+				"%d iterations", i);
+			break;
+		}
 	}
 
 	tst_res(TPASS, "Nothing bad appears to have happened");
Signature · 1 line
1.8.3.1
#2 lore raw

PATCH 1/1 pcrypt_aead01: break early when approaching timeout

----- Original Message -----

Quoted 3 lines

Hi!
Acked.

Pushed.