gh-138122: Skip threads on EPERM in blocking mode profiler

pablogsal · pablogsal · commit fc2283b32872 · 2026-01-02T17:27:36.000Z
When using blocking mode in the remote debugging profiler, ptrace calls
to seize threads can fail with EPERM if the thread has exited between
listing and attaching, is in a special kernel state, or is already being
traced. Previously this raised a RuntimeError that was caught by the
Python sampling loop,and retried indefinitely since EPERM is
a persistent condition that will not resolve on its own.

Treat EPERM the same as ESRCH by returning 1 (skip this thread) instead
of -1 (fatal error). This allows profiling to continue with the threads
that can be traced rather than entering an endless retry loop printing
the same error message repeatedly.
diff --git a/Misc/NEWS.d/next/Library/2026-01-02-17-26-33.gh-issue-143368.m3EF9E.rst b/Misc/NEWS.d/next/Library/2026-01-02-17-26-33.gh-issue-143368.m3EF9E.rst
@@ -0,0 +1,3 @@
+Fix endless retry loop in :mod:`profiling.sampling` blocking mode when
+threads cannot be seized due to ``EPERM``. Such threads are now skipped
+instead of causing repeated error messages. Patch by Pablo Galindo.
diff --git a/Modules/_remote_debugging/threads.c b/Modules/_remote_debugging/threads.c
@@ -632,15 +632,21 @@ seize_thread(pid_t tid)
     if (errno == ESRCH) {
         return 1;  // Thread gone, skip
     }
+    if (errno == EPERM) {
+        // Thread may have exited, be in a special state, or already be traced.
+        // Skip rather than fail - this avoids endless retry loops when
+        // threads transiently become inaccessible.
+        return 1;
+    }
     if (errno == EINVAL || errno == EIO) {
         // Fallback for older kernels
         if (ptrace(PTRACE_ATTACH, tid, NULL, NULL) == 0) {
             int status;
             waitpid(tid, &status, __WALL);
             return 0;
         }
-        if (errno == ESRCH) {
-            return 1;  // Thread gone
+        if (errno == ESRCH || errno == EPERM) {
+            return 1;  // Thread gone or inaccessible
         }
     }
     return -1;  // Real error

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,3 @@`
	`1`	+Fix endless retry loop in :mod:`profiling.sampling` blocking mode when
	`2`	+threads cannot be seized due to ``EPERM``. Such threads are now skipped
	`3`	`+instead of causing repeated error messages. Patch by Pablo Galindo.`
Original file line number	Diff line number	Diff line change
`@@ -632,15 +632,21 @@ seize_thread(pid_t tid)`
`632`	`632`	`if (errno == ESRCH) {`
`633`	`633`	`return 1; // Thread gone, skip`
`634`	`634`	`}`
	`635`	`+ if (errno == EPERM) {`
	`636`	`+ // Thread may have exited, be in a special state, or already be traced.`
	`637`	`+ // Skip rather than fail - this avoids endless retry loops when`
	`638`	`+ // threads transiently become inaccessible.`
	`639`	`+ return 1;`
	`640`	`+ }`
`635`	`641`	`if (errno == EINVAL \|\| errno == EIO) {`
`636`	`642`	`// Fallback for older kernels`
`637`	`643`	`if (ptrace(PTRACE_ATTACH, tid, NULL, NULL) == 0) {`
`638`	`644`	`int status;`
`639`	`645`	`waitpid(tid, &status, __WALL);`
`640`	`646`	`return 0;`
`641`	`647`	`}`
`642`		`- if (errno == ESRCH) {`
`643`		`- return 1; // Thread gone`
	`648`	`+ if (errno == ESRCH \|\| errno == EPERM) {`
	`649`	`+ return 1; // Thread gone or inaccessible`
`644`	`650`	`}`
`645`	`651`	`}`
`646`	`652`	`return -1; // Real error`