Graphene-SGX: Regression Failures on Kernel 5.12 upstream kernel

[anjalirai-intel]

Config

Graphene :- Commit: f9ed396
Linux :- 5.12.0 upstream kernel

Issues

PAL Failures

TC-1: test_200_event
The test expect to return “TEST OK” but it returned “Error at line 71, pal_errno: -1, debug: DkProcessExit: Returning exit code 1”

TC-2: test_900_misc
Expected Result
Query System Time OK
Delay Execution for 10000 Microseconds OK
Delay Execution for 3 Seconds OK
Generate Random Bits OK

Actual Result: (Missing Delay Execution for 3 Seconds OK)
Query System Time OK
Sleeped 10078 Microseconds
Delay Execution for 10000 Microseconds OK
Sleeped 2993086 Microseconds
Generate Random Bits OK

LibOS Failures

TC-1: test_041_futex_timeout
Expected Result:
futex correctly timed out

Actual Result:
invoke futex syscall with a 1 second timeout
Slept for 997926 microseconds, which is less than 1 seconds

LTP Test Failures:

TC-1: alarm05
Expected Result:
TINFO: Timeout per run is 0h 05m 00s
TPASS: alarm() returned zero
TPASS: alarm() returned remainder correctly
TPASS: alarm handler fired once

Summary:
passed 0
failed 0
skipped 0
warnings 0

Actual Result:
TINFO: Timeout per run is 0h 05m 00s
TPASS: alarm() returned zero
TFAIL: alarm() returned wrong remained 10
TPASS: alarm handler fired once

Summary:
passed 0
failed 0
skipped 0
warnings 0

TC-2: alarm06
Expected Result:
TINFO: Timeout per run is 0h 05m 00s
TPASS: Received 0 alarms
TPASS: alarm(0) returned 1

Summary:
passed 0
failed 0
skipped 0
warnings 0

Actual Result:
TINFO: Timeout per run is 0h 05m 00s
TPASS: Received 0 alarms
TFAIL: alarm(0) returned 2, expected 1

Summary:
passed 0
failed 0
skipped 0
warnings 0

TC-3: sigtimedwait01
Expected Result:
TINFO: Timeout per run is 0h 05m 00s
TPASS: Wait interrupted by expected signal
TPASS: struct siginfo is correct
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask
TPASS: Wait interrupted by expected signal
TPASS: Wait interrupted by expected signal
TPASS: sigwaitinfo restored the original mask
TPASS: Fault occurred while accessing the buffers
TPASS: Child exited with expected code
TPASS: Fault occurred while accessing the buffers
TPASS: Wait interrupted by timeout

Summary:
passed 0
failed 0
skipped 0
warnings 0

Actual Result:
TINFO: Timeout per run is 0h 05m 00s
TPASS: Wait interrupted by expected signal
TPASS: struct siginfo is correct
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask
TPASS: Wait interrupted by expected signal
TPASS: Wait interrupted by expected signal
TPASS: sigwaitinfo restored the original mask
TPASS: Fault occurred while accessing the buffers
graphene/LibOS/shim/test/ltp/ltp_src/lib/tst_test.c:1300: TBROK: Test killed by SIGPWR!

Summary:
passed 0
failed 0
skipped 0
warnings 0

error: Using insecure argv source. Graphene will continue application execution, but this configuration must not be used in production!
error: Failed to initialize secure pipe 152772: -6

TC-4: rt_sigtimedwait01
Expected Result:
TINFO: Timeout per run is 0h 05m 00s
TINFO: Testing variant: syscall with old kernel spec
TPASS: Wait interrupted by expected signal
TPASS: struct siginfo is correct
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask
TPASS: Wait interrupted by expected signal
TPASS: Wait interrupted by expected signal
TPASS: sigwaitinfo restored the original mask
TPASS: Fault occurred while accessing the buffers
TPASS: Child exited with expected code
TPASS: Fault occurred while accessing the buffers
TPASS: Wait interrupted by timeout
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask
TFAIL: Expected error number EAGAIN, got: EINTR (4)
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask
TPASS: struct siginfo is correct
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask

Summary:
passed 0
failed 0
skipped 0
warnings 0

Actual Result:
TINFO: Timeout per run is 0h 05m 00s
TINFO: Testing variant: syscall with old kernel spec
TPASS: Wait interrupted by expected signal
TPASS: struct siginfo is correct
TPASS: struct siginfo is correct
TPASS: sigwaitinfo restored the original mask
TPASS: Wait interrupted by expected signal
TPASS: Wait interrupted by expected signal
TPASS: sigwaitinfo restored the original mask
TPASS: Fault occurred while accessing the buffers
TBROK: Test killed by SIGPWR!

Summary:
passed 0
failed 0
skipped 0
warnings 0

TC-5: futex_wait_bitset01
Actual Result:
TINFO: Timeout per run is 0h 05m 00s
TINFO: Testing variant: syscall with old kernel spec
TINFO: testing futex_wait_bitset() timeout with CLOCK_MONOTONIC
TFAIL: futex_wait_bitset() woken up prematurely 99918us, expected 100010us
TINFO: testing futex_wait_bitset() timeout with CLOCK_REALTIME
TFAIL: futex_wait_bitset() woken up prematurely 99933us, expected 100010us

Summary:
passed 0
failed 0
skipped 0
warnings 0

TC-6: alarm03
Actual Result:
TINFO: Timeout per run is 0h 05m 00s
TFAIL: alarm(100), fork, alarm(0) parent's alarm returned 99
TPASS: alarm(100), fork, alarm(0) child's alarm returned 0

Summary:
passed 0
failed 0
skipped 0
warnings 0

Steps to reproduce

PAL Failures
cd Pal/regression
make SGX=1 regression

LibOS Failures
cd LibOS/shim/test/regression
sudo make SGX=1 regression
Failure Trace Logs.zip

LTP Failures:
cd LibOS/shim/test/ltp
make -j8 SGX=1 all sgx-tokens
make -j8 ltp-sgx.xml

[dimakuv]

So is this regression because of:

• moving to the new Linux kernel, or
• moving to a new Graphene commit?

[dimakuv]

There seems to be something wrong with timings in this new Linux 5.12 kernel. Looks like all new regressions are about timings that are very slightly off.

[anjalirai-intel]

@dimakuv This regression is because moving to new Linux Kernel

[anjalirai-intel]

Updated ltp test failures

[dimakuv]

@anjalirx-intel I don't remember if we agreed on some way of solving this issue... Is anyone looking at it? I think someone volunteered to check this. I will assign priority P1 for now and mark it as a bug.

[anjalirai-intel]

No @dimakuv , we haven't agreed on any way for solving this issue

[Satya1493]

@dimakuv, please assign this to me. I've started to have a look.

[anjalirai-intel]

@Satya1493 This issue is also seen on RHEL setup Kernel 5.12