tools/guest-diag: add UVM wait diagnostics, tracker poke and VA->PA map injector

1 files changed, 36 insertions, 0 deletions

diff --git a/tools/guest-diag/bp999.py b/tools/guest-diag/bp999.py
new file mode 100644
index 0000000..db83d7d
--- /dev/null
+++ b/tools/guest-diag/bp999.py
@@ -0,0 +1,36 @@
+# gdb script: break on the libcuda instructions that produce CUDA errors
+# 719 (0x2cf) and 999 (0x3e7), then print the stack that led there.
+# Run: sudo gdb -p <cdt-pid> -batch -x /tmp/bp999.py
+import gdb
+
+out = gdb.execute("info proc mappings", to_string=True)
+base = None
+for line in out.splitlines():
+    if "libcuda" in line and "r-xp" in line:
+        f = line.split()
+        base = int(f[0], 16) - int(f[3], 16)
+        break
+print("BASE=0x%x" % (base or 0))
+if base is None:
+    raise SystemExit
+for off in (0x44fb80, 0x44fb6c):
+    gdb.execute("break *0x%x" % (base + off))
+gdb.execute("set pagination off")
+gdb.execute("set confirm off")
+try:
+    gdb.execute("continue")
+except gdb.error as e:
+    print("CONTINUE_ERR %s" % e)
+print("HIT_OR_EXIT")
+try:
+    print(gdb.execute("bt 16", to_string=True))
+except Exception as e:
+    print("bt_fail %s" % e)
+try:
+    print(gdb.execute("info registers rdi rsi rdx rax rbx rbp rip", to_string=True))
+except Exception as e:
+    print("regs_fail %s" % e)
+try:
+    gdb.execute("detach")
+except Exception:
+    pass