Thread (62 messages) flat view 62 messages, 2 authors, 15h ago

Re: [PATCH v8 29/43] dyndbg: detect class_id reservation conflicts

From: sashiko-bot@kernel.org
Date: 2026-09-05 18:30:58
Also in: dri-devel

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] A classmap conflict in a single built-in module causes dynamic_debug_init() to fail, completely disabling dynamic debugging system-wide.
--

commit 10601ffba9f53c0ad961efc6f72432bdf809f6d1
Author: Jim Cromie [off-list ref]

dyndbg: detect class_id reservation conflicts

This commit detects if a module defines multiple classmaps that overlap in the
0..62 class-id space. It adds overlap checking in ddebug_add_module() and
returns -EINVAL when conflicts are detected, ensuring that class_id to
classname lookup works properly.
quoted hunk ↗ jump to hunk
diff --git a/lib/dynamic_debug.c b/lib/dynamic_debug.c
index ba618dcf9677f..0f89b6784ab26 100644
--- a/lib/dynamic_debug.c
+++ b/lib/dynamic_debug.c
[ ... ]
quoted hunk ↗ jump to hunk
@@ -1409,16 +1425,24 @@ static int ddebug_add_module(struct _ddebug_info *di)
 	dd_set_module_subrange(i, cm, &dt->info, maps);
 	dd_set_module_subrange(i, cli, &dt->info, users);
 
+	/* insure 2+ classmaps share the per-module 0..62 class_id space */
+	for_subvec(i, cm, &dt->info, maps)
+		if (ddebug_class_range_overlap(cm, &reserved_ids))
+			goto cleanup;
+
 	mutex_lock(&ddebug_lock);
 	list_add_tail(&dt->link, &ddebug_tables);
 	mutex_unlock(&ddebug_lock);
-
 	if (dt->info.users.len)
 		ddebug_apply_class_users(&dt->info);
 
 	vpr_info("%3u debug prints in module %s\n",
 		 dt->info.descs.len, dt->info.mod_name);
 	return 0;
+cleanup:
+	WARN_ONCE(1, "dyndbg multi-classmap conflict in %s\n", di->mod_name);
+	kfree(dt);
+	return -EINVAL;
 }
[Severity: High]
Does returning an error here cause a systemic failure for built-in modules?

Looking at dynamic_debug_init(), it iterates over all built-in modules
and calls ddebug_add_module() for each one. If any single module fails,
it aborts initialization:

lib/dynamic_debug.c:dynamic_debug_init() {
    ...
		if (strcmp(modname, iter->modname)) {
			...
			ret = ddebug_add_module(&di);
			if (ret)
				goto out_err;
    ...
out_err:
	ddebug_remove_all_tables();
	return 0;
}

If a single built-in module has an overlapping classmap, the new -EINVAL return
bubbles up and triggers the out_err path. This destroys all previously loaded
tables and completely disables dynamic debugging system-wide, rather than just
skipping the misconfigured module. Should ddebug_add_module() failure be
non-fatal for built-in modules, or perhaps handled more gracefully?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260905-dd-cmap-part2-clean-v8-0-a4cc0674f6fd@gmail.com?part=29
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help