I've observed that the tls multi_chunk_sendfile selftest hangs during recv() and ultimately times out, and it seems to have done so even when the test was first introduced. Reading through the commit message when it was added (0e6fbe39bdf7 "net/tls(TLS_SW): Add selftest for 'chunked' sendfile test") I get the impression that the test is meant to demonstrate a problem with ktls, but there's no indication that the problem has been fixed.
Am I right that the expectation is that this test will fail? If that's the case, shouldn't this test be removed until the problem is fixed?
Thanks, Seth
On Thu, 17 Jun 2021 14:46:11 -0500 Seth Forshee wrote:
I've observed that the tls multi_chunk_sendfile selftest hangs during recv() and ultimately times out, and it seems to have done so even when the test was first introduced.
It hangs yet it passes? I lost track of this issue because the test does pass on my system:
# PASSED: 183 / 183 tests passed. # Totals: pass:183 fail:0 xfail:0 xpass:0 skip:0 error:0
$ uname -r 5.12.9-300.fc34.x86_64
Reading through the commit message when it was added (0e6fbe39bdf7 "net/tls(TLS_SW): Add selftest for 'chunked' sendfile test") I get the impression that the test is meant to demonstrate a problem with ktls, but there's no indication that the problem has been fixed.
Yeah, the fix was discussed here:
https://lore.kernel.org/netdev/1591392508-14592-1-git-send-email-pooja.trive...
IDK why it stalled to be honest :S
Am I right that the expectation is that this test will fail? If that's the case, shouldn't this test be removed until the problem is fixed?
On Thu, Jun 17, 2021 at 02:22:34PM -0700, Jakub Kicinski wrote:
On Thu, 17 Jun 2021 14:46:11 -0500 Seth Forshee wrote:
I've observed that the tls multi_chunk_sendfile selftest hangs during recv() and ultimately times out, and it seems to have done so even when the test was first introduced.
It hangs yet it passes? I lost track of this issue because the test does pass on my system:
# PASSED: 183 / 183 tests passed. # Totals: pass:183 fail:0 xfail:0 xpass:0 skip:0 error:0
$ uname -r 5.12.9-300.fc34.x86_64
It doesn't pass with Ubuntu kernels:
# RUN tls.12_gcm.multi_chunk_sendfile ... # multi_chunk_sendfile: Test terminated by timeout # FAIL tls.12_gcm.multi_chunk_sendfile not ok 6 tls.12_gcm.multi_chunk_sendfile ... # RUN tls.13_gcm.multi_chunk_sendfile ... # multi_chunk_sendfile: Test terminated by timeout # FAIL tls.13_gcm.multi_chunk_sendfile not ok 51 tls.13_gcm.multi_chunk_sendfile ... # RUN tls.12_chacha.multi_chunk_sendfile ... # multi_chunk_sendfile: Test terminated by timeout # FAIL tls.12_chacha.multi_chunk_sendfile not ok 96 tls.12_chacha.multi_chunk_sendfile ... # RUN tls.13_chacha.multi_chunk_sendfile ... # multi_chunk_sendfile: Test terminated by timeout # FAIL tls.13_chacha.multi_chunk_sendfile not ok 141 tls.13_chacha.multi_chunk_sendfile ... # FAILED: 177 / 183 tests passed. # Totals: pass:177 fail:6 xfail:0 xpass:0 skip:0 error:0
$ uname -r 5.13.0-7-generic
The results are the same with 5.12, etc. Maybe some difference in configs.
Reading through the commit message when it was added (0e6fbe39bdf7 "net/tls(TLS_SW): Add selftest for 'chunked' sendfile test") I get the impression that the test is meant to demonstrate a problem with ktls, but there's no indication that the problem has been fixed.
Yeah, the fix was discussed here:
https://lore.kernel.org/netdev/1591392508-14592-1-git-send-email-pooja.trive...
IDK why it stalled to be honest :S
Hopefully it can make some progress now.
Thanks, Seth
linux-kselftest-mirror@lists.linaro.org
-
Jakub Kicinski -
Seth Forshee