guest-diag: cheaper trace read, optional push sampler, bigger pagemap batches

Under load the semfix poll grepped the whole (growing) trace file and the
track_waits push block spawned three guest processes per sample, both of
which slowed the pin path into the client's failure window. The trace read
now tails 128 KB; track_waits.sh gates the push dump behind PUSH=0; the
pagemap helper reads 32 KB chunks.

3 files changed, 14 insertions, 10 deletions

diff --git a/tools/guest-diag/guest_pmap_win.py b/tools/guest-diag/guest_pmap_win.py
index f6ca85c..501e7bc 100644
--- a/tools/guest-diag/guest_pmap_win.py
+++ b/tools/guest-diag/guest_pmap_win.py
@@ -16,7 +16,7 @@ hi = int(sys.argv[3], 16)
 maxp = int(sys.argv[4]) if len(sys.argv) > 4 else 20000
 pa_max = int(sys.argv[5], 16) if len(sys.argv) > 5 else 0
 
-BATCH = 512
+BATCH = 4096
 out = []
 run_va = run_pa = None
 run_n = 0
diff --git a/tools/guest-diag/semfix_watch.py b/tools/guest-diag/semfix_watch.py
index 400fafe..347da5e 100644
--- a/tools/guest-diag/semfix_watch.py
+++ b/tools/guest-diag/semfix_watch.py
@@ -145,7 +145,8 @@ def guest_trace():
     Read *before* clearing: the create-time probes fire once per client and
     would be wiped by a clear that races them.
     """
-    inner = ("grep -E \"uvmt|uvch|uvbp|uvpbf\" /sys/kernel/tracing/trace "
+    inner = ("tail -c 131072 /sys/kernel/tracing/trace "
+             "| grep -E \"uvmt|uvch|uvbp|uvpbf\" "
              "| tail -400")
     if not os.environ.get("KEEP_TRACE"):
         inner += "; echo > /sys/kernel/tracing/trace"
diff --git a/tools/guest-diag/track_waits.sh b/tools/guest-diag/track_waits.sh
index 45a9a66..6ae9154 100644
--- a/tools/guest-diag/track_waits.sh
+++ b/tools/guest-diag/track_waits.sh
@@ -16,6 +16,7 @@ PAGE_VA="${PAGE_VA:-0x20400f000}"
 COUNT="${COUNT:-20}"
 STRIDE="${STRIDE:-16}"
 OFFSET="${OFFSET:-0xec0}"
+PUSH="${PUSH:-1}"
 
 : > /tmp/trk.log
 : > /tmp/cnt.log
@@ -58,14 +59,16 @@ for i in $(seq 1 "$N"); do
             | grep -vE "^\[|Reading symbols|^0x[0-9a-f]+ in |warning:" >> /tmp/trk.log
     fi
 
-    {
-        echo "=== $TS push pages ==="
-        python3 /tmp/guest_pmap.py "$PID" 0x201000000 0x203400000 0x203400020 2>&1
-        for va in 0x201000000 0x203400000 0x203400020; do
-            echo "--- $va"
-            timeout 5 python3 /tmp/read_va.py "$PID" "$va" 64 --u32 2>&1
-        done
-    } >> /tmp/push.log
+    if [ "$PUSH" = "1" ]; then
+        {
+            echo "=== $TS push pages ==="
+            python3 /tmp/guest_pmap.py "$PID" 0x201000000 0x203400000 0x203400020 2>&1
+            for va in 0x201000000 0x203400000 0x203400020; do
+                echo "--- $va"
+                timeout 5 python3 /tmp/read_va.py "$PID" "$va" 64 --u32 2>&1
+            done
+        } >> /tmp/push.log
+    fi
 
     sleep "$SLEEP"
 done