[PATCH v4 0/2] rtla: handle SIGTERM and test abnormal exits
From: Chaosheng Qu
Date: Sat Nov 21 2026 - 07:03:33 EST
v3 split the test changes into their own patch and fixed the commit
messages, but the test still signalled as soon as the instance directory
appeared plus one second. On a slow guest that can beat set_signals(),
which is the startup race sashiko described. v4 waits until
/proc/<pid>/status reports SIGTERM in SigCgt before signalling.
Patch 1 is unchanged: it adds the SIGTERM handler that was missing.
SIGINT, the -d/--duration timer and a failing --on-threshold action
already reach the cleanup path; SIGTERM terminated the process and left
its tracefs instance behind.
Patch 2 is the test, now with the readiness check. With it, cleanup.t
passes all five cases on a 7.3-rc7 guest, including SIGTERM; without the
handler, SIGTERM leaves the instance behind and the test fails.
v3: https://lore.kernel.org/linux-trace-kernel/cover.1795258353.git.chaosheng.qu.linux@xxxxxxxxx/
v2: https://lore.kernel.org/linux-trace-kernel/1794300001.11111.1234567890@xxxxxxxxx/
v1: https://lore.kernel.org/linux-trace-kernel/20261017122359.402696-1-chaosheng.qu.linux@xxxxxxxxx/
This replaces the v1 submission, the v2 series and v3.
Changes since v3:
- wait for the SIGTERM handler instead of a fixed sleep in the test
- no change to patch 1
Testing: cleanup.t on 7.3-rc7, 4 vCPU QEMU guest -- SIGINT, SIGTERM,
-d timer, action failure and repeated interrupts all pass. With the
handler removed, the SIGTERM case fails as expected.
Chaosheng Qu (2):
rtla: handle SIGTERM during a run
rtla/tests: cover cleanup on abnormal exits
tools/tracing/rtla/src/common.c | 2 +
tools/tracing/rtla/tests/cleanup.t | 70 ++++++
.../rtla/tests/scripts/check-cleanup.sh | 219 ++++++++++++++++++
3 files changed, 291 insertions(+)
create mode 100644 tools/tracing/rtla/tests/cleanup.t
create mode 100755 tools/tracing/rtla/tests/scripts/check-cleanup.sh
--
2.43.0