diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-23 21:32:31 -0400 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-23 21:32:31 -0400 |
| commit | d4cc5f989943f01875dea86b834d90538a748112 (patch) | |
| tree | dd45f6cb5a91a747e19a72192e0a43f64001d3f6 /tools/guest-diag/pin_gr_stream.py | |
| parent | 4f648aba4706af53e746e093ed8137771d947fe1 (diff) | |
| download | open-iov-d4cc5f989943f01875dea86b834d90538a748112.tar.gz open-iov-d4cc5f989943f01875dea86b834d90538a748112.zip | |
tools: transport-proof demand writes + tolerant demand parsing (the pool-pin latency root cause)
The r581 boot's guest scanner died mid-boot with ValueError in main():
"invalid literal for int() with base 16:
'0x20400f000\\n0x121000000\\n0x12000ab4c'". A guest-agent transport-escaped
append merged demand VAs onto one line with literal backslash-n, and the
old parser crashed on it. The crash stopped ALL new pins; the host only
re-asserted known pairs, so the UVM pool VA 0x121000000 never arrived
while 185 other injections flowed - the pool-pin latency that blocked
every validation boot.
* pin_gr_stream._demand_send/_demand_loop now write one "echo <va> >>"
per VA (no newlines, no backslashes in the transported command; immune
to transport escaping).
* guest_pin_stream's demand parser extracts every 0x token with re and
skips malformed ones instead of raising out of main().
Diffstat (limited to 'tools/guest-diag/pin_gr_stream.py')
| -rw-r--r-- | tools/guest-diag/pin_gr_stream.py | 26 |
1 files changed, 22 insertions, 4 deletions
diff --git a/tools/guest-diag/pin_gr_stream.py b/tools/guest-diag/pin_gr_stream.py index 6eb5be7..ff7e299 100644 --- a/tools/guest-diag/pin_gr_stream.py +++ b/tools/guest-diag/pin_gr_stream.py @@ -100,15 +100,33 @@ def _scanner_alive(): return "ALIVE" in (out or "") +def _demand_write_cmd(vas): + """Build a transport-proof append command for the guest demand file. + + r582: the old `printf '%s' '<lines with real LF>'` form was mangled by + the guest-agent transport under load - the newlines arrived as LITERAL + backslash-n text and merged VAs onto one line, and the guest scanner's + parser then crashed on `int('0x20400f000\\\\n0x121000000...', 16)` + (ValueError in main; r581 boot scanner log). The scanner's death stopped + ALL new pins while the host only re-asserted known pairs - the observed + pool-pin-latency (0x121000000 never arrived while other injections + flowed). One `echo <va> >> file` per VA carries no newlines and no + backslashes at all, so no transport layer can escape them.""" + return "; ".join("echo 0x%x >> /tmp/guest_want_vas.txt" % va + for va in vas) + + def _demand_send(vas): """Append VA hints to the guest scanner's demand file (best-effort). Same channel pin_gr_stream._demand_loop uses for host-dmesg stalls; the stall watchdog uses it to ask the scanner to re-resolve the fixed pre-capture VAs (counter page / UVM pool) by content/pagemap.""" - lines = "".join("0x%x\n" % va for va in vas) + cmd = _demand_write_cmd(vas) + if not cmd: + return try: - vexec("printf '%%s' '%s' >> /tmp/guest_want_vas.txt" % lines, 30) + vexec(cmd, 30) except Exception: pass @@ -222,9 +240,9 @@ def _demand_loop(): todo = [va for va, (f, l) in track.items() if now - l > 20.0] if not todo: continue - lines = "".join("0x%x\n" % va for va in todo) + cmd = _demand_write_cmd(todo) try: - vexec("printf '%%s' '%s' >> /tmp/guest_want_vas.txt" % lines, 30) + vexec(cmd, 30) except Exception: continue for va in todo: |
