guest-diag: cycle monitor CPUs on kernel-VA translation failure

HMP x/gva2gpa translate with the selected vCPU's CR3; while that CPU runs
user code under PTI the kernel half is hidden, so kernel-VA reads (the
tracking-sem chain and the CE pushbuffer CPU alias) fail and the pin pass
aborts. Try the other vCPUs before giving up: during an active CUDA client
at least one CPU is in kernel mode, so the chain/pin work no longer require
the client to be parked in sleep().

1 files changed, 39 insertions, 4 deletions

diff --git a/tools/guest-diag/semfix_watch.py b/tools/guest-diag/semfix_watch.py
index fa435ea..1bb4b39 100644
--- a/tools/guest-diag/semfix_watch.py
+++ b/tools/guest-diag/semfix_watch.py
@@ -96,11 +96,42 @@ class Qmp:
     def hmp(self, line):
         return self.cmd("human-monitor-command", **{"command-line": line}).get("return", "")
 
+    def ncpu(self):
+        if getattr(self, "_ncpu", 0):
+            return self._ncpu
+        n = 0
+        for line in self.hmp("info cpus").splitlines():
+            if line.strip().startswith("CPU #") or line.strip().startswith("* CPU #"):
+                n += 1
+        self._ncpu = n or 4
+        return self._ncpu
+
+    def _set_cpu(self, n):
+        """Select the monitor's current CPU. HMP `x`/`gva2gpa` translate with
+        the selected CPU's CR3: while that CPU runs user code under PTI the
+        kernel half is hidden, so a translation of a kernel VA fails. Cycling
+        CPUs until one is in kernel mode makes the reads work during active
+        clients (the client spins/sleeps in kernel code on at least one CPU)."""
+        if getattr(self, "_cpu", 0) == n:
+            return
+        self.hmp("cpu %d" % n)
+        self._cpu = n
+
+    def _cpu_order(self):
+        cur = getattr(self, "_cpu", 0)
+        return [cur] + [c for c in range(self.ncpu()) if c != cur]
+
     def read64(self, addr, n=1, cache=True):
         key = (addr, n)
         if cache and key in self._cache:
             return self._cache[key]
-        out = self.hmp("x/%dgx 0x%x" % (n, addr))
+        out = ""
+        for c in self._cpu_order():
+            self._set_cpu(c)
+            out = self.hmp("x/%dgx 0x%x" % (n, addr))
+            m = re.search(r":(.*)$", out, re.S)
+            if m and re.findall(r"0x([0-9a-fA-F]+)", m.group(1))[:n]:
+                break
         m = re.search(r":(.*)$", out, re.S)
         if not m:
             return []
@@ -110,9 +141,13 @@ class Qmp:
         return vals
 
     def gva2gpa(self, va):
-        out = self.hmp("gva2gpa 0x%x" % va)
-        m = re.search(r"gpa: 0x([0-9a-fA-F]+)", out)
-        return int(m.group(1), 16) if m else None
+        for c in self._cpu_order():
+            self._set_cpu(c)
+            out = self.hmp("gva2gpa 0x%x" % va)
+            m = re.search(r"gpa: 0x([0-9a-fA-F]+)", out)
+            if m:
+                return int(m.group(1), 16)
+        return None
 
 
 def is_kern(v):