[next-20200621] LTP tests af_alg02/05 failure on POWER9 PowerVM LPAR

With recent next(next-20200621) af_alg02/05 tests fail while running on POWER9
PowerVM LPAR.

Results from  5.8.0-rc1-next-20200622
# ./af_alg02
tst_test.c:1096: INFO: Timeout per run is 0h 00m 20s
af_alg02.c:52: BROK: Timed out while reading from request socket.
#

5.8.0-rc1-next-20200618 was good. The test case ran fine.

Root cause analysis point to following commit:

commit f3c802a1f30013f8f723b62d7fa49eb9e991da23
    crypto: algif_aead - Only wake up when ctx->more is zero

Reverting this commit allows the test to PASS.

Results after reverting the mentioned commit:

# uname -r
5.8.0-rc1-next-20200622-dirty
# ./af_alg02 
tst_test.c:1096: INFO: Timeout per run is 0h 00m 20s
af_alg02.c:33: PASS: Successfully "encrypted" an empty message
#
# ./af_alg05
tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
af_alg05.c:34: PASS: read() expectedly failed with EINVAL
# 

Thanks
-Sachin

On Mon, Jun 22, 2020 at 05:55:29PM +0530, Sachin Sant wrote:

With recent next(next-20200621) af_alg02/05 tests fail while running on POWER9
PowerVM LPAR.

Results from  5.8.0-rc1-next-20200622
# ./af_alg02
tst_test.c:1096: INFO: Timeout per run is 0h 00m 20s
af_alg02.c:52: BROK: Timed out while reading from request socket.
#

5.8.0-rc1-next-20200618 was good. The test case ran fine.

Root cause analysis point to following commit:

commit f3c802a1f30013f8f723b62d7fa49eb9e991da23
    crypto: algif_aead - Only wake up when ctx->more is zero

Reverting this commit allows the test to PASS.

Yes that commit does change the behaviour if you don't terminate
your AEAD request by clearing the MSG_MORE flag.  AEAD does not
support chaining so each request must be sent in full before it
can be processed through recvmsg(2).  Setting the MSG_MORE flag
indicates that your request has not been sent in full.

Cheers,
-- 
Email: Herbert Xu [off-list ref]
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt