Thread (50 messages) flat view 50 messages, 2 authors, 2d ago

Re: [PATCH v10 32/38] dyndbg: resolve "protection" of class'd pr_debug

From: sashiko-bot@kernel.org
Date: 2026-09-16 16:56:22
Also in: dri-devel

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] dyndbg: module's own classmaps bypass protection due to missing controlling_param init
- [Medium] dyndbg: data race on map->controlling_param due to lockless initialization
--

commit 5ffc9f5551765ac7c4413f320b8e9d1977dbe99a
Author: Jim Cromie [off-list ref]

dyndbg: resolve "protection" of class'd pr_debug

This commit updates the protection mechanism for class'd pr_debug callsites
against unclassed queries. It introduces a requirement that a sysfs parameter
(controlling_param) must be present for a classmap to be protected.
quoted hunk ↗ jump to hunk
diff --git a/lib/dynamic_debug.c b/lib/dynamic_debug.c
index c023d07623f29..ac54a0499369c 100644
--- a/lib/dynamic_debug.c
+++ b/lib/dynamic_debug.c
[ ... ]
quoted hunk ↗ jump to hunk
@@ -257,7 +300,28 @@ static bool ddebug_match_desc(const struct ddebug_query *query,
 	    dp->lineno > query->last_lineno)
 		return false;
 
-	return true;
+	/*
+	 * above are all satisfied, so we can make final decisions:
+	 * 1- class FOO or implied class __DEFAULT__
+	 * 2- site.is_classed or not
+	 */
+	if (query->class_string) {
+		/* class FOO given, exact match required */
+		return (dp->class_id == selected_class);
+	}
+	/* query class __DEFAULT__ by omission. */
+	if (dp->class_id == _DPRINTK_CLASS_DFLT) {
+		/* un-classed site */
+		return true;
+	}
+	/* site is class'd */
+	site_map = ddebug_find_map_by_class_id(di, dp->class_id);
+	if (!site_map) {
+		WARN_ONCE(1, "unknown class_id %d, check %s's CLASSMAP definitions", dp->class_id, di->mod_name);
+		return false;
+	}
+	/* module(-param) decides protection */
+	return !ddebug_class_wants_protection(site_map);
[Severity: High]
Does this code inadvertently bypass ABI protection for a module's own
class'd pr_debug sites?

Since the ddebug_apply_class_maps() block was removed in v10, it appears
ddebug_add_module() only initializes user maps (by calling
ddebug_apply_class_users() when dt->info.users.len > 0), and misses
initializing the defining module's own native classmaps.

Consequently, if a module defines a classmap and parameter but has no
loaded dependent modules, map->controlling_param remains NULL. This causes
ddebug_class_wants_protection() to default to false, allowing unclassed
legacy queries to incorrectly modify these protected sites.
 }
 
[ ... ]
+static void ddebug_match_apply_kparam(const struct kernel_param *kp,
+				      struct ddebug_class_map *map,
+				      const char *mod_name)
+{
+	struct ddebug_class_param *dcp = ddebug_get_classmap_kparam(kp, map);
 
 	if (dcp && dcp->map == map) {
+		map->controlling_param = dcp;
[Severity: Medium]
Can this assignment introduce a data race?

In ddebug_add_module(), the new debug table is published to the globally
visible ddebug_tables list under ddebug_lock, but the lock is released
before calling ddebug_apply_class_users().

Since ddebug_match_apply_kparam() is subsequently called without locks,
concurrent threads processing ddebug_change() could hold ddebug_lock
and read map->controlling_param while it is being updated here, leading
to a data race.
 		v2pr_info(" kp:%s.%s =0x%x", mod_name, kp->name, *dcp->bits);
 		vpr_cm_info(map, " %s maps ", mod_name);
 		ddebug_sync_classbits(kp, mod_name);
 	}
 }
-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260916-dd-cmap-part2-clean-v10-0-af4cf4767707@gmail.com?part=32
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help