Message ID | 20200806095944.43862-1-colin.king@canonical.com |
---|---|
Headers | show |
Series | selftests/net: relax cpu affinity requirement in msg_zerocopy test | expand |
Applied to Focal/msater-next. thank you! -Kelsey On 2020-08-06 10:59:43 , Colin King wrote: > From: Colin Ian King <colin.king@canonical.com> > > BugLink: https://bugs.launchpad.net/bugs/1812620 > > == SRU Justification [ FOCAL ] == > > The msg_zerocopy.sh kernel self test will fail on machines that don't > have 2 or 3 CPUs such as 1 CPU cloud instances since the C test program > tries to set CPU affinity to CPUs 2 and 3 and bails out if it fails. > > == Fix == > > Upstream linux-next commit > > commit 16f6458f2478b55e2b628797bc81a4455045c74e > Author: Willem de Bruijn <willemb@google.com> > Date: Wed Aug 5 04:40:45 2020 -0400 > > selftests/net: relax cpu affinity requirement in msg_zerocopy test > > The fix now just emits a warning that CPU affinity can't be set > rather than cause an exit(1) termination. > > == Test cast == > > Run the msg_zerocopy.sh test from the kernel net selftest on a 1 CPU > system or a system with CPUs 2-3 offline'd. Without the fix the test > fails. With the fix it runs successfully as expected. > > == Regression Potential == > > The original test pinned the CPUs for a benchmarking metric, for our > testing we are using this to test to see if the operations in the tes > work successfully. There is a potential that users using this test > will not notice the warning if they are using this test as a benchmark > on a 1 CPU system and may get more jittery timing in their benchmarks > rather than a test failing and complaining they are not running it on > a suitable multi-CPU system. However, the likelyhood of a user using > this test on a single CPU system for benchmarking is small and as it > stands the test will now run and produce potentially jittery benchmarks > on a 1 CPU system compared to previously where it never ran. > > Willem de Bruijn (1): > selftests/net: relax cpu affinity requirement in msg_zerocopy test > > tools/testing/selftests/net/msg_zerocopy.c | 5 ++--- > 1 file changed, 2 insertions(+), 3 deletions(-) > > -- > 2.27.0 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team
On Thu, Aug 06, 2020 at 10:59:43AM +0100, Colin King wrote: > From: Colin Ian King <colin.king@canonical.com> > > BugLink: https://bugs.launchpad.net/bugs/1812620 > > == SRU Justification [ FOCAL ] == > > The msg_zerocopy.sh kernel self test will fail on machines that don't > have 2 or 3 CPUs such as 1 CPU cloud instances since the C test program > tries to set CPU affinity to CPUs 2 and 3 and bails out if it fails. > > == Fix == > > Upstream linux-next commit > > commit 16f6458f2478b55e2b628797bc81a4455045c74e > Author: Willem de Bruijn <willemb@google.com> > Date: Wed Aug 5 04:40:45 2020 -0400 > > selftests/net: relax cpu affinity requirement in msg_zerocopy test > > The fix now just emits a warning that CPU affinity can't be set > rather than cause an exit(1) termination. > > == Test cast == > > Run the msg_zerocopy.sh test from the kernel net selftest on a 1 CPU > system or a system with CPUs 2-3 offline'd. Without the fix the test > fails. With the fix it runs successfully as expected. > > == Regression Potential == > > The original test pinned the CPUs for a benchmarking metric, for our > testing we are using this to test to see if the operations in the tes > work successfully. There is a potential that users using this test > will not notice the warning if they are using this test as a benchmark > on a 1 CPU system and may get more jittery timing in their benchmarks > rather than a test failing and complaining they are not running it on > a suitable multi-CPU system. However, the likelyhood of a user using > this test on a single CPU system for benchmarking is small and as it > stands the test will now run and produce potentially jittery benchmarks > on a 1 CPU system compared to previously where it never ran. Applied to groovy/master-next, thanks!
Hello folks, I just found that this issue is affecting the B/hwe 5.3 as well. Can we add this patch the B/hwe too? (It's also a cherry-pick, and I have the bug nominated for linux-hwe) Thanks Sam On Thu, Aug 6, 2020 at 6:00 PM Colin King <colin.king@canonical.com> wrote: > > From: Colin Ian King <colin.king@canonical.com> > > BugLink: https://bugs.launchpad.net/bugs/1812620 > > == SRU Justification [ FOCAL ] == > > The msg_zerocopy.sh kernel self test will fail on machines that don't > have 2 or 3 CPUs such as 1 CPU cloud instances since the C test program > tries to set CPU affinity to CPUs 2 and 3 and bails out if it fails. > > == Fix == > > Upstream linux-next commit > > commit 16f6458f2478b55e2b628797bc81a4455045c74e > Author: Willem de Bruijn <willemb@google.com> > Date: Wed Aug 5 04:40:45 2020 -0400 > > selftests/net: relax cpu affinity requirement in msg_zerocopy test > > The fix now just emits a warning that CPU affinity can't be set > rather than cause an exit(1) termination. > > == Test cast == > > Run the msg_zerocopy.sh test from the kernel net selftest on a 1 CPU > system or a system with CPUs 2-3 offline'd. Without the fix the test > fails. With the fix it runs successfully as expected. > > == Regression Potential == > > The original test pinned the CPUs for a benchmarking metric, for our > testing we are using this to test to see if the operations in the tes > work successfully. There is a potential that users using this test > will not notice the warning if they are using this test as a benchmark > on a 1 CPU system and may get more jittery timing in their benchmarks > rather than a test failing and complaining they are not running it on > a suitable multi-CPU system. However, the likelyhood of a user using > this test on a single CPU system for benchmarking is small and as it > stands the test will now run and produce potentially jittery benchmarks > on a 1 CPU system compared to previously where it never ran. > > Willem de Bruijn (1): > selftests/net: relax cpu affinity requirement in msg_zerocopy test > > tools/testing/selftests/net/msg_zerocopy.c | 5 ++--- > 1 file changed, 2 insertions(+), 3 deletions(-) > > -- > 2.27.0 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team
On 31.08.20 10:06, Po-Hsu Lin wrote: > Hello folks, > > I just found that this issue is affecting the B/hwe 5.3 as well. > Can we add this patch the B/hwe too? > (It's also a cherry-pick, and I have the bug nominated for linux-hwe) Ok, since selftest only affects our side and this helps to have less troubles with the testing, we can pick it up while doing other work. Releasing might be delayed until there is a required change to be done. -Stefan > > Thanks > Sam > > On Thu, Aug 6, 2020 at 6:00 PM Colin King <colin.king@canonical.com> wrote: >> >> From: Colin Ian King <colin.king@canonical.com> >> >> BugLink: https://bugs.launchpad.net/bugs/1812620 >> >> == SRU Justification [ FOCAL ] == >> >> The msg_zerocopy.sh kernel self test will fail on machines that don't >> have 2 or 3 CPUs such as 1 CPU cloud instances since the C test program >> tries to set CPU affinity to CPUs 2 and 3 and bails out if it fails. >> >> == Fix == >> >> Upstream linux-next commit >> >> commit 16f6458f2478b55e2b628797bc81a4455045c74e >> Author: Willem de Bruijn <willemb@google.com> >> Date: Wed Aug 5 04:40:45 2020 -0400 >> >> selftests/net: relax cpu affinity requirement in msg_zerocopy test >> >> The fix now just emits a warning that CPU affinity can't be set >> rather than cause an exit(1) termination. >> >> == Test cast == >> >> Run the msg_zerocopy.sh test from the kernel net selftest on a 1 CPU >> system or a system with CPUs 2-3 offline'd. Without the fix the test >> fails. With the fix it runs successfully as expected. >> >> == Regression Potential == >> >> The original test pinned the CPUs for a benchmarking metric, for our >> testing we are using this to test to see if the operations in the tes >> work successfully. There is a potential that users using this test >> will not notice the warning if they are using this test as a benchmark >> on a 1 CPU system and may get more jittery timing in their benchmarks >> rather than a test failing and complaining they are not running it on >> a suitable multi-CPU system. However, the likelyhood of a user using >> this test on a single CPU system for benchmarking is small and as it >> stands the test will now run and produce potentially jittery benchmarks >> on a 1 CPU system compared to previously where it never ran. >> >> Willem de Bruijn (1): >> selftests/net: relax cpu affinity requirement in msg_zerocopy test >> >> tools/testing/selftests/net/msg_zerocopy.c | 5 ++--- >> 1 file changed, 2 insertions(+), 3 deletions(-) >> >> -- >> 2.27.0 >> >> >> -- >> kernel-team mailing list >> kernel-team@lists.ubuntu.com >> https://lists.ubuntu.com/mailman/listinfo/kernel-team >
From: Colin Ian King <colin.king@canonical.com> BugLink: https://bugs.launchpad.net/bugs/1812620 == SRU Justification [ FOCAL ] == The msg_zerocopy.sh kernel self test will fail on machines that don't have 2 or 3 CPUs such as 1 CPU cloud instances since the C test program tries to set CPU affinity to CPUs 2 and 3 and bails out if it fails. == Fix == Upstream linux-next commit commit 16f6458f2478b55e2b628797bc81a4455045c74e Author: Willem de Bruijn <willemb@google.com> Date: Wed Aug 5 04:40:45 2020 -0400 selftests/net: relax cpu affinity requirement in msg_zerocopy test The fix now just emits a warning that CPU affinity can't be set rather than cause an exit(1) termination. == Test cast == Run the msg_zerocopy.sh test from the kernel net selftest on a 1 CPU system or a system with CPUs 2-3 offline'd. Without the fix the test fails. With the fix it runs successfully as expected. == Regression Potential == The original test pinned the CPUs for a benchmarking metric, for our testing we are using this to test to see if the operations in the tes work successfully. There is a potential that users using this test will not notice the warning if they are using this test as a benchmark on a 1 CPU system and may get more jittery timing in their benchmarks rather than a test failing and complaining they are not running it on a suitable multi-CPU system. However, the likelyhood of a user using this test on a single CPU system for benchmarking is small and as it stands the test will now run and produce potentially jittery benchmarks on a 1 CPU system compared to previously where it never ran. Willem de Bruijn (1): selftests/net: relax cpu affinity requirement in msg_zerocopy test tools/testing/selftests/net/msg_zerocopy.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)