summaryrefslogtreecommitdiffstats
path: root/tools
diff options
context:
space:
mode:
Diffstat (limited to 'tools')
-rw-r--r--tools/guest-diag/guest_pin_stream.py60
-rw-r--r--tools/guest-diag/pin_gr_stream.py15
-rwxr-xr-xtools/guest-diag/semfix_uvmt.py89
3 files changed, 144 insertions, 20 deletions
diff --git a/tools/guest-diag/guest_pin_stream.py b/tools/guest-diag/guest_pin_stream.py
index bb3f99a..1ff0d3f 100644
--- a/tools/guest-diag/guest_pin_stream.py
+++ b/tools/guest-diag/guest_pin_stream.py
@@ -11,7 +11,9 @@
#
# Usage: sudo python3 guest_pin_stream.py [host-cid] [port] [lo-hex] [hi-hex] [pa-max-hex]
import os
+import re
import socket
+import subprocess
import sys
import time
@@ -24,6 +26,30 @@ INTERVAL = float(os.environ.get("STREAM_INTERVAL", "0.25"))
BATCH = 4096
REGION = 0x200000
SPAN = int(os.environ.get("STREAM_SPAN", "0x10000"), 16)
+TRACE = ("/sys/kernel/tracing/trace"
+ if os.path.exists("/sys/kernel/tracing/trace")
+ else "/sys/kernel/debug/tracing/trace")
+SEM_EVERY = int(os.environ.get("STREAM_SEM_EVERY", "4"))
+
+
+def new_sems(seen):
+ """New tracking-sem kptrs (uvmt lines) in the trace tail.
+
+ Read in the guest so the host does not need a guest-agent round trip for
+ every semaphore discovery; the trace is only tailed while a client runs.
+ """
+ try:
+ out = subprocess.run(["tail", "-c", "262144", TRACE],
+ capture_output=True, timeout=10).stdout
+ except Exception:
+ return []
+ found = []
+ for m in re.finditer(rb"sem=0x([0-9a-fA-F]+)", out):
+ v = int(m.group(1), 16)
+ if v not in seen:
+ seen.add(v)
+ found.append(v)
+ return found
def cdt_pid():
@@ -82,18 +108,23 @@ def main():
f.write('%d\n' % os.getpid())
sock = None
seen = {}
+ seen_sems = set()
last_pid = None
last_log = 0.0
+ loop = 0
while True:
pid = cdt_pid()
+ loop += 1
if pid is None:
last_pid = None
seen = {}
+ seen_sems = set()
time.sleep(INTERVAL)
continue
if pid != last_pid:
last_pid = pid
seen = {}
+ seen_sems = set()
if sock is not None:
try:
sock.sendall(("#pid %d\n" % pid).encode())
@@ -109,28 +140,33 @@ def main():
dt = time.time() - t0
new = [(va, pa) for va, pa in cur.items() if seen.get(va) != pa]
seen = cur
+ lines = []
+ if loop % SEM_EVERY == 0:
+ for sem in new_sems(seen_sems):
+ lines.append('#sem 0x%x\n' % sem)
now = time.time()
if now - last_log > 1.0:
last_log = now
try:
with open('/tmp/guest_pin_stream.log', 'a') as f:
- f.write('%s pid=%s present=%d new=%d scan=%.3fs\n'
- % (time.strftime('%T'), pid, len(cur), len(new), dt))
+ f.write('%s pid=%s present=%d new=%d sem=%d scan=%.3fs\n'
+ % (time.strftime('%T'), pid, len(cur), len(new),
+ len(seen_sems), dt))
except OSError:
pass
- if not new:
+ if not new and not lines:
time.sleep(INTERVAL)
continue
- new.sort()
- lines = []
- va0, pa0, n = new[0][0], new[0][1], 1
- for va, pa in new[1:]:
- if va == va0 + n * 0x1000 and pa == pa0 + n * 0x1000:
- n += 1
- continue
+ if new:
+ new.sort()
+ va0, pa0, n = new[0][0], new[0][1], 1
+ for va, pa in new[1:]:
+ if va == va0 + n * 0x1000 and pa == pa0 + n * 0x1000:
+ n += 1
+ continue
+ lines.append('0x%x,0x%x,%d\n' % (va0, pa0, n))
+ va0, pa0, n = va, pa, 1
lines.append('0x%x,0x%x,%d\n' % (va0, pa0, n))
- va0, pa0, n = va, pa, 1
- lines.append('0x%x,0x%x,%d\n' % (va0, pa0, n))
payload = ''.join(lines).encode()
if sock is None:
try:
diff --git a/tools/guest-diag/pin_gr_stream.py b/tools/guest-diag/pin_gr_stream.py
index c1ef7b2..44f8681 100644
--- a/tools/guest-diag/pin_gr_stream.py
+++ b/tools/guest-diag/pin_gr_stream.py
@@ -25,12 +25,23 @@ PORT = int(os.environ.get("PINWIN_PORT", "45678"))
LO = os.environ.get("PINWIN_LO", "0x200400000")
HI = os.environ.get("PINWIN_HI", "0x206000000")
PAMAX = os.environ.get("PINWIN_PA_MAX", "0x180000000")
+SEMSTREAM = os.environ.get("SEM_STREAM_FILE", "/tmp/opencode/sem_stream.txt")
def log(msg):
print("[%s] %s" % (time.strftime("%H:%M:%S"), msg), flush=True)
+def _mirror(line):
+ """Append guest control lines (#pid/#sem) to the shared host file so the
+ single-QMP watcher can consume them without guest-agent round trips."""
+ try:
+ with open(SEMSTREAM, "a") as f:
+ f.write(line + "\n")
+ except OSError:
+ pass
+
+
def vexec(cmd, timeout=40):
try:
r = subprocess.run([VMEXEC, "exec", "--timeout", str(timeout - 5), cmd],
@@ -115,6 +126,10 @@ def main():
m = re.search(r"(\d+)", text)
seen = {}
log("guest reports cdt pid %s" % (m.group(1) if m else "?"))
+ _mirror(text)
+ continue
+ if text.startswith("#sem"):
+ _mirror(text)
continue
m = re.match(r"^(0x[0-9a-f]+),(0x[0-9a-f]+)(?:,(\d+))?$", text)
if m:
diff --git a/tools/guest-diag/semfix_uvmt.py b/tools/guest-diag/semfix_uvmt.py
index 0d31394..fb0e92e 100755
--- a/tools/guest-diag/semfix_uvmt.py
+++ b/tools/guest-diag/semfix_uvmt.py
@@ -41,6 +41,53 @@ DLOG = os.environ.get("UVMT_DMESG", "1") == "1"
VMEXEC = os.path.normpath(os.path.join(HERE, "..", "..", "scripts", "vgpu_vm.py"))
COUNTER_LO = int(os.environ.get("COUNTER_LO", "0x20400f000"), 16)
COUNTER_HI = int(os.environ.get("COUNTER_HI", "0x204010000"), 16)
+SEMSTREAM = os.environ.get("SEM_STREAM_FILE", "/tmp/opencode/sem_stream.txt")
+TRACE_EVERY = int(os.environ.get("UVMT_TRACE_EVERY", "5"))
+
+
+class Stream:
+ """Tail the guest scanner's control lines mirrored by pin_gr_stream.py.
+
+ The guest emits "#pid <n>" and "#sem 0x<k>" over vsock; the listener
+ appends them here. This replaces per-loop guest-agent trace reads, which
+ were too slow to catch short-lived tracking semaphores.
+ """
+
+ def __init__(self, path):
+ self.path = path
+ self.off = 0
+ self.sems = []
+ self.pids = []
+
+ def poll(self):
+ try:
+ with open(self.path) as f:
+ f.seek(0, os.SEEK_END)
+ size = f.tell()
+ if size < self.off:
+ self.off = 0
+ f.seek(self.off)
+ data = f.read()
+ self.off = f.tell()
+ except OSError:
+ return [], []
+ new_sems, new_pids = [], []
+ for line in data.splitlines():
+ if line.startswith("#sem"):
+ m = re.search(r"0x([0-9a-fA-F]+)", line)
+ if m:
+ v = int(m.group(1), 16)
+ if v not in self.sems:
+ self.sems.append(v)
+ new_sems.append(v)
+ elif line.startswith("#pid"):
+ m = re.search(r"(\d+)", line)
+ if m:
+ v = int(m.group(1))
+ if v not in self.pids:
+ self.pids.append(v)
+ new_pids.append(v)
+ return new_sems, new_pids
def log(msg):
@@ -191,6 +238,8 @@ def main():
dseen = set()
counter_pids = set()
last_pid = None
+ stream = Stream(SEMSTREAM)
+ loop = 0
def pin_range_once(gpu_vas, cpu, gpa, size=0):
key = tuple(gpu_vas), cpu, size
@@ -200,7 +249,30 @@ def main():
pin_pushbuffer(q, gpu_vas, cpu, gpa, size)
while True:
- sems, pbs, chans = guest_trace()
+ loop += 1
+ sems = set()
+ pbs = set()
+ chans = set()
+
+ new_sems, new_pids = stream.poll()
+ for s in new_sems:
+ log("stream: new uvmt sem 0x%x" % s)
+ for p in new_pids:
+ log("stream: client pid %d" % p)
+ last_pid = p
+ sems.update(stream.sems)
+
+ # The guest-agent trace read is slow under client load; use it only
+ # as a fallback for channel/pushbuffer discovery and to recover sems
+ # the scanner may have missed.
+ if loop % TRACE_EVERY == 0:
+ try:
+ t_sems, t_pbs, t_chans = guest_trace()
+ except Exception:
+ t_sems, t_pbs, t_chans = set(), set(), set()
+ sems.update(t_sems)
+ pbs.update(t_pbs)
+ chans.update(t_chans)
# Poll every sem ever seen: the client's final wait may only appear in
# the trace once, and the trace ring wraps while it spins.
@@ -293,14 +365,15 @@ def main():
dmesg_releases(dseen)
# Pin the completion-counter page of a newly started client, before its
- # first GR pushes decode their releases.
- pid = cdt_pid()
- if pid and pid != last_pid:
- last_pid = pid
- log("counter: client pid %d" % pid)
- if pid and pid not in counter_pids:
+ # first GR pushes decode their releases. The pid comes from the guest
+ # scanner; the guest-agent fallback is only a safety net.
+ if last_pid is None and loop % TRACE_EVERY == 0:
+ last_pid = cdt_pid()
+ if last_pid:
+ log("counter: client pid %d (agent)" % last_pid)
+ if last_pid and last_pid not in counter_pids:
try:
- pin_counter_page(pid, counter_pids)
+ pin_counter_page(last_pid, counter_pids)
except Exception as e:
log("counter: pin failed: %s" % e)
generated by cgit v1.2.3 (git 2.39.1) at 2026-10-12 22:40:33 +0000