From 79ddf3c6f364d5c8b94aa11955c108628813d540 Mon Sep 17 00:00:00 2001
From: Luca Boccassi <luca.boccassi@gmail.com>
Date: Mon, 25 May 2026 11:28:36 +0100
Subject: [PATCH 1/3] Revert "test: switch TEST-55-OOMD stress-ng --vm-method
 to lfsr32"

This reverts commit 4ac23697280bf54bb768f0aa7a5c7d7d0bcf3f6b.
---
 .../TEST-55-OOMD-testbloat.service                |  7 ++++---
 .../TEST-55-OOMD-testmunch.service                |  7 ++++---
 test/units/TEST-55-OOMD.sh                        | 15 +++++++++------
 3 files changed, 17 insertions(+), 12 deletions(-)

diff --git a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service
index e4109fd60d..22bbd210e9 100644
--- a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service
+++ b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service
@@ -7,6 +7,7 @@ Description=Create a lot of memory pressure
 # to throttle and be put under heavy pressure.
 MemoryHigh=3M
 Slice=TEST-55-OOMD-workload.slice
-# Pin --vm-method=lfsr32: the only stress-ng vm method without TARGET_CLONES,
-# so it can't dispatch to an AVX-512 clone and SIGILL on CPUs lacking it.
-ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep --vm-method=lfsr32
+# Pin --vm-method to a portable method (zero-one): the default 'all' cycles
+# through methods, including newer ones using AVX-512 instructions that SIGILL
+# on CPUs without AVX-512 (e.g. AMD Zen 1-3), making the test flaky.
+ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep --vm-method=zero-one
diff --git a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service
index 81305c7e97..06eea10b79 100644
--- a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service
+++ b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service
@@ -5,6 +5,7 @@ Description=Create some memory pressure
 [Service]
 MemoryHigh=12M
 Slice=TEST-55-OOMD-workload.slice
-# Pin --vm-method=lfsr32: the only stress-ng vm method without TARGET_CLONES,
-# so it can't dispatch to an AVX-512 clone and SIGILL on CPUs lacking it.
-ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep --vm-method=lfsr32
+# Pin --vm-method to a portable method (zero-one): the default 'all' cycles
+# through methods, including newer ones using AVX-512 instructions that SIGILL
+# on CPUs without AVX-512 (e.g. AMD Zen 1-3), making the test flaky.
+ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep --vm-method=zero-one
diff --git a/test/units/TEST-55-OOMD.sh b/test/units/TEST-55-OOMD.sh
index 84baee12a4..b7311e83dc 100755
--- a/test/units/TEST-55-OOMD.sh
+++ b/test/units/TEST-55-OOMD.sh
@@ -366,12 +366,13 @@ EOF
     systemctl reload systemd-oomd.service
 
     # Run a transient service with OOMRules=testrule that generates memory pressure.
-    # Pin --vm-method=lfsr32: the only stress-ng vm method without TARGET_CLONES,
-    # so it can't dispatch to an AVX-512 clone and SIGILL on CPUs lacking it.
+    # Pin --vm-method to a portable method (zero-one): the default 'all' cycles
+    # through every method, including newer ones using AVX-512 instructions that
+    # SIGILL on CPUs without AVX-512 (e.g. AMD Zen 1-3), making the test flaky.
     (! systemd-run --wait --unit=TEST-55-OOMD-testrules \
         -p MemoryHigh=3M \
         -p OOMRules=testrule \
-        stress-ng --timeout 3m --vm 10 --vm-bytes 50M --vm-keep --vm-method=lfsr32)
+        stress-ng --timeout 3m --vm 10 --vm-bytes 50M --vm-keep --vm-method=zero-one)
 
     # Verify in the journal that the rule triggered
     journalctl --sync
@@ -456,12 +457,14 @@ EOF
 
     # Start the unit without --wait so we can check mid-run state. The
     # stress-ng timeout bounds the test if anything goes wrong.
-    # Pin --vm-method=lfsr32: the only stress-ng vm method without TARGET_CLONES,
-    # so it can't SIGILL on AVX-512-less CPUs and exit before the 6 s wait below.
+    # Pin --vm-method to a portable method (zero-one): the default 'all' cycles
+    # through every method, including newer ones using AVX-512 instructions that
+    # SIGILL on CPUs without AVX-512 (e.g. AMD Zen 1-3) and would cause stress-ng
+    # to exit before the 6 s wait below elapses, failing the ActiveState check.
     systemd-run --unit=TEST-55-OOMD-slowrule \
         -p MemoryHigh=3M \
         -p OOMRules=slowrule \
-        stress-ng --timeout 15s --vm 10 --vm-bytes 50M --vm-keep --vm-method=lfsr32
+        stress-ng --timeout 15s --vm 10 --vm-bytes 50M --vm-keep --vm-method=zero-one
 
     # Wait long enough for oomd's 1s rule-check loop to evaluate the condition
     # many times. With LastingSec=1h the kill must not fire.

From aac19fe4685906a2166464cb0e9e600ebfe850a1 Mon Sep 17 00:00:00 2001
From: Luca Boccassi <luca.boccassi@gmail.com>
Date: Mon, 25 May 2026 11:28:37 +0100
Subject: [PATCH 2/3] Revert "test: pin stress-ng --vm-method to a portable
 scalar method in TEST-55-OOMD"

This reverts commit 881e4717c7981b274853309e68b39153e3b292f4.
---
 .../TEST-55-OOMD-testbloat.service                  |  5 +----
 .../TEST-55-OOMD-testmunch.service                  |  5 +----
 test/units/TEST-55-OOMD.sh                          | 13 +++----------
 3 files changed, 5 insertions(+), 18 deletions(-)

diff --git a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service
index 22bbd210e9..70c87727c8 100644
--- a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service
+++ b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testbloat.service
@@ -7,7 +7,4 @@ Description=Create a lot of memory pressure
 # to throttle and be put under heavy pressure.
 MemoryHigh=3M
 Slice=TEST-55-OOMD-workload.slice
-# Pin --vm-method to a portable method (zero-one): the default 'all' cycles
-# through methods, including newer ones using AVX-512 instructions that SIGILL
-# on CPUs without AVX-512 (e.g. AMD Zen 1-3), making the test flaky.
-ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep --vm-method=zero-one
+ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep
diff --git a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service
index 06eea10b79..79bd01838e 100644
--- a/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service
+++ b/test/integration-tests/TEST-55-OOMD/TEST-55-OOMD.units/TEST-55-OOMD-testmunch.service
@@ -5,7 +5,4 @@ Description=Create some memory pressure
 [Service]
 MemoryHigh=12M
 Slice=TEST-55-OOMD-workload.slice
-# Pin --vm-method to a portable method (zero-one): the default 'all' cycles
-# through methods, including newer ones using AVX-512 instructions that SIGILL
-# on CPUs without AVX-512 (e.g. AMD Zen 1-3), making the test flaky.
-ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep --vm-method=zero-one
+ExecStart=stress-ng --timeout 3m --vm 10 --vm-bytes 200M --vm-keep
diff --git a/test/units/TEST-55-OOMD.sh b/test/units/TEST-55-OOMD.sh
index b7311e83dc..6689bbdd73 100755
--- a/test/units/TEST-55-OOMD.sh
+++ b/test/units/TEST-55-OOMD.sh
@@ -365,14 +365,11 @@ EOF
 
     systemctl reload systemd-oomd.service
 
-    # Run a transient service with OOMRules=testrule that generates memory pressure.
-    # Pin --vm-method to a portable method (zero-one): the default 'all' cycles
-    # through every method, including newer ones using AVX-512 instructions that
-    # SIGILL on CPUs without AVX-512 (e.g. AMD Zen 1-3), making the test flaky.
+    # Run a transient service with OOMRules=testrule that generates memory pressure
     (! systemd-run --wait --unit=TEST-55-OOMD-testrules \
         -p MemoryHigh=3M \
         -p OOMRules=testrule \
-        stress-ng --timeout 3m --vm 10 --vm-bytes 50M --vm-keep --vm-method=zero-one)
+        stress-ng --timeout 3m --vm 10 --vm-bytes 50M --vm-keep)
 
     # Verify in the journal that the rule triggered
     journalctl --sync
@@ -457,14 +454,10 @@ EOF
 
     # Start the unit without --wait so we can check mid-run state. The
     # stress-ng timeout bounds the test if anything goes wrong.
-    # Pin --vm-method to a portable method (zero-one): the default 'all' cycles
-    # through every method, including newer ones using AVX-512 instructions that
-    # SIGILL on CPUs without AVX-512 (e.g. AMD Zen 1-3) and would cause stress-ng
-    # to exit before the 6 s wait below elapses, failing the ActiveState check.
     systemd-run --unit=TEST-55-OOMD-slowrule \
         -p MemoryHigh=3M \
         -p OOMRules=slowrule \
-        stress-ng --timeout 15s --vm 10 --vm-bytes 50M --vm-keep --vm-method=zero-one
+        stress-ng --timeout 15s --vm 10 --vm-bytes 50M --vm-keep
 
     # Wait long enough for oomd's 1s rule-check loop to evaluate the condition
     # many times. With LastingSec=1h the kill must not fire.

From a17efef1375c07df89e7f02e87fb881dfced53fb Mon Sep 17 00:00:00 2001
From: Luca Boccassi <luca.boccassi@gmail.com>
Date: Mon, 25 May 2026 12:11:23 +0100
Subject: [PATCH 3/3] test: try to detect SIGILL in stress-ng and skip
 TEST-55-OOMD gracefully

The attempt to fix SIGILL in TEST-55-OOMD actually made the test more
flaky, as the chosen stress-ng method generates less pressure, so it
often fails.

Try a different approach: try to detect that SIGILL caused the unit
calling stress-ng to fail, and skip gracefully in that case.
---
 test/units/TEST-55-OOMD.sh | 54 +++++++++++++++++++++++++++++++-------
 1 file changed, 44 insertions(+), 10 deletions(-)

diff --git a/test/units/TEST-55-OOMD.sh b/test/units/TEST-55-OOMD.sh
index 6689bbdd73..9993e98937 100755
--- a/test/units/TEST-55-OOMD.sh
+++ b/test/units/TEST-55-OOMD.sh
@@ -94,6 +94,19 @@ else
     systemd-run -t -p MemoryMax=10M -p MemorySwapMax=0 -p MemoryZSwapMax=0 true
 fi
 
+# stress-ng can fail with SIGILL due to trying to use AVX-512 on older CPUs, try to detect and avoid failing
+stress_ng_sigilled() {
+    local result status sigill
+    local unit="${1:?}"
+    shift
+
+    result=$(systemctl "$@" show "$unit" -P Result)
+    status=$(systemctl "$@" show "$unit" -P ExecMainStatus)
+    sigill=$(kill -l ILL)
+
+    [[ "$status" == "$sigill" && ( "$result" == "signal" || "$result" == "core-dump" ) ]]
+}
+
 test_basic() {
     local cgroup_path="${1:?}"
     shift
@@ -123,7 +136,11 @@ test_basic() {
     if systemctl "$@" status TEST-55-OOMD-testbloat.service; then exit 42; fi
     if ! systemctl "$@" status TEST-55-OOMD-testchill.service; then exit 24; fi
 
-    assert_eq "$(systemctl "$@" show TEST-55-OOMD-testbloat.service -P ManagedOOMKills)" "1"
+    if stress_ng_sigilled TEST-55-OOMD-testbloat.service "$@"; then
+        echo "stress-ng died with SIGILL, skipping ManagedOOMKills assertion"
+    else
+        assert_eq "$(systemctl "$@" show TEST-55-OOMD-testbloat.service -P ManagedOOMKills)" "1"
+    fi
 
     systemctl "$@" kill --signal=KILL TEST-55-OOMD-testbloat.service || :
     systemctl "$@" stop TEST-55-OOMD-testbloat.service
@@ -171,10 +188,14 @@ EOF
         sleep 2
     done
 
-    # testmunch should be killed since testbloat had the avoid xattr on it
-    if ! systemctl status TEST-55-OOMD-testbloat.service; then exit 25; fi
-    if systemctl status TEST-55-OOMD-testmunch.service; then exit 43; fi
-    if ! systemctl status TEST-55-OOMD-testchill.service; then exit 24; fi
+    if stress_ng_sigilled TEST-55-OOMD-testbloat.service || stress_ng_sigilled TEST-55-OOMD-testmunch.service; then
+        echo "stress-ng died with SIGILL, skipping testcase_preference_avoid assertions"
+    else
+        # testmunch should be killed since testbloat had the avoid xattr on it
+        if ! systemctl status TEST-55-OOMD-testbloat.service; then exit 25; fi
+        if systemctl status TEST-55-OOMD-testmunch.service; then exit 43; fi
+        if ! systemctl status TEST-55-OOMD-testchill.service; then exit 24; fi
+    fi
 
     systemctl kill --signal=KILL TEST-55-OOMD-testbloat.service || :
     systemctl kill --signal=KILL TEST-55-OOMD-testmunch.service || :
@@ -249,8 +270,12 @@ EOF
         sleep 2
     done
 
-    if systemctl status TEST-55-OOMD-testmunch.service; then exit 44; fi
-    if ! systemctl status TEST-55-OOMD-testchill.service; then exit 23; fi
+    if stress_ng_sigilled TEST-55-OOMD-testmunch.service; then
+        echo "stress-ng died with SIGILL, skipping testcase_duration_override assertions"
+    else
+        if systemctl status TEST-55-OOMD-testmunch.service; then exit 44; fi
+        if ! systemctl status TEST-55-OOMD-testchill.service; then exit 23; fi
+    fi
 
     systemctl kill --signal=KILL TEST-55-OOMD-testmunch.service || :
     systemctl stop TEST-55-OOMD-testmunch.service
@@ -463,9 +488,13 @@ EOF
     # many times. With LastingSec=1h the kill must not fire.
     sleep 6
 
-    # Unit must still be active — if it were killed, Result= would be oom-kill.
-    assert_eq "$(systemctl show TEST-55-OOMD-slowrule.service -P ActiveState)" "active"
-    assert_eq "$(systemctl show TEST-55-OOMD-slowrule.service -P Result)" "success"
+    if stress_ng_sigilled TEST-55-OOMD-slowrule.service; then
+        echo "stress-ng died with SIGILL, skipping testcase_oom_rulesets_lasting_sec assertions"
+    else
+        # Unit must still be active. If it were killed, Result= would be oom-kill.
+        assert_eq "$(systemctl show TEST-55-OOMD-slowrule.service -P ActiveState)" "active"
+        assert_eq "$(systemctl show TEST-55-OOMD-slowrule.service -P Result)" "success"
+    fi
 
     systemctl stop TEST-55-OOMD-slowrule.service 2>/dev/null || true
 
@@ -484,6 +513,11 @@ EOF
     systemctl reload systemd-oomd.service
     ! systemctl start --wait TEST-55-OOMD-testbloat.service || exit 1
 
+    if stress_ng_sigilled TEST-55-OOMD-testbloat.service; then
+        echo "stress-ng died with SIGILL, skipping testcase_prekill_hook"
+        return 0
+    fi
+
     # one hook
     mkdir -p /run/systemd/oomd.prekill.hook/
     ncat --recv-only -kUl /run/systemd/oomd.prekill.hook/althook >/tmp/oomd_event.json &