Skip to contentSkip to conversation

mm/vmscan: avoid false-positive -Wuninitialized warning

PATCH Superseded Arnd Bergmann 5 participants 7 messages Feb 13 – Mar 27

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm/vmscan: avoid false-positive -Wuninitialized warning

From: Arnd Bergmann <[email protected]>

When the -fsanitize=bounds sanitizer is enabled, gcc-16 sometimes runs into a corner case in the read_ctrl_pos() pos function, where it sees possible undefined behavior from the 'tier' index overflowing, presumably in the case that this was called with a negative tier:

In function 'get_tier_idx',

    inlined from 'isolate_folios' at mm/vmscan.c:4671:14:

mm/vmscan.c: In function 'isolate_folios':
mm/vmscan.c:4645:29: error: 'pv.refaulted' is used uninitialized [-Werror=uninitialized]

Part of the problem seems to be that read_ctrl_pos() has unusual calling conventions since commit 37a260870f2c ("mm/mglru: rework type selection") where passing MAX_NR_TIERS makes it accumulate all tiers but passing a smaller positive number makes it read a single tier instead.

Avoid this case by splitting read_ctrl_pos() into two separate helpers that each only do one of the two cases. This avoids the warning as far as I can tell, and seems a bit easier to understand to me.

---
This is currently the only such warning I get from gcc-16.0.1, and none from any other version.

I'm not overly happy about having to work around it with a random code chance, but hopefully the version I ended up with makes sense regardless.

---
 mm/vmscan.c | 37 +++++++++++++++++++++++++------------
 1 file changed, 25 insertions(+), 12 deletions(-)
mm/vmscan.c +25 -12
index eaf795c1cfb3..602c955d1f30 100644
@@ -3159,20 +3159,15 @@ struct ctrl_pos {
 static void read_ctrl_pos(struct lruvec *lruvec, int type, int tier, int gain,
 			  struct ctrl_pos *pos)
 {
-	int i;
 	struct lru_gen_folio *lrugen = &lruvec->lrugen;
 	int hist = lru_hist_from_seq(lrugen->min_seq[type]);
 
 	pos->gain = gain;
-	pos->refaulted = pos->total = 0;
-
-	for (i = tier % MAX_NR_TIERS; i <= min(tier, MAX_NR_TIERS - 1); i++) {
-		pos->refaulted += lrugen->avg_refaulted[type][i] +
-				  atomic_long_read(&lrugen->refaulted[hist][type][i]);
-		pos->total += lrugen->avg_total[type][i] +
-			      lrugen->protected[hist][type][i] +
-			      atomic_long_read(&lrugen->evicted[hist][type][i]);
-	}
+	pos->refaulted = lrugen->avg_refaulted[type][tier] +
+			 atomic_long_read(&lrugen->refaulted[hist][type][tier]);
+	pos->total = lrugen->avg_total[type][tier] +
+		     lrugen->protected[hist][type][tier] +
+		     atomic_long_read(&lrugen->evicted[hist][type][tier]);
 }
 
 static void reset_ctrl_pos(struct lruvec *lruvec, int type, bool carryover)
@@ -4640,6 +4635,24 @@ static int get_tier_idx(struct lruvec *lruvec, int type)
 	return tier - 1;
 }
 
+static void aggregate_ctrl_pos(struct lruvec *lruvec, int type, int gain,
+			       struct ctrl_pos *pos)
+{
+	struct lru_gen_folio *lrugen = &lruvec->lrugen;
+	int hist = lru_hist_from_seq(lrugen->min_seq[type]);
+
+	pos->gain = gain;
+	pos->refaulted = pos->total = 0;
+
+	for (int i = 0; i < MAX_NR_TIERS; i++) {
+		pos->refaulted += lrugen->avg_refaulted[type][i] +
+				  atomic_long_read(&lrugen->refaulted[hist][type][i]);
+		pos->total += lrugen->avg_total[type][i] +
+			      lrugen->protected[hist][type][i] +
+			      atomic_long_read(&lrugen->evicted[hist][type][i]);
+	}
+}
+
 static int get_type_to_scan(struct lruvec *lruvec, int swappiness)
 {
 	struct ctrl_pos sp, pv;
@@ -4653,8 +4666,8 @@ static int get_type_to_scan(struct lruvec *lruvec, int swappiness)
 	 * Compare the sum of all tiers of anon with that of file to determine
 	 * which type to scan.
 	 */
-	read_ctrl_pos(lruvec, LRU_GEN_ANON, MAX_NR_TIERS, swappiness, &sp);
-	read_ctrl_pos(lruvec, LRU_GEN_FILE, MAX_NR_TIERS, MAX_SWAPPINESS - swappiness, &pv);
+	aggregate_ctrl_pos(lruvec, LRU_GEN_ANON, swappiness, &sp);
+	aggregate_ctrl_pos(lruvec, LRU_GEN_FILE, MAX_SWAPPINESS - swappiness, &pv);
 
 	return positive_ctrl_err(&sp, &pv);
 }
Signature · 1 line
2.39.5
maintainer
#1 lore raw
Quoted 3 lines · On Fri, 13 Feb 2026 13:38:56 +0100 Arnd Bergmann <[email protected]> wrote:

From: Arnd Bergmann <[email protected]>

When the -fsanitize=bounds sanitizer is enabled,

Is this an option in current kernels?

Quoted 19 lines

gcc-16 sometimes runs
into a corner case in the read_ctrl_pos() pos function, where it sees possible undefined behavior from the 'tier' index overflowing, presumably in the case that this was called with a negative tier:

In function 'get_tier_idx',

    inlined from 'isolate_folios' at mm/vmscan.c:4671:14:

mm/vmscan.c: In function 'isolate_folios':
mm/vmscan.c:4645:29: error: 'pv.refaulted' is used uninitialized [-Werror=uninitialized]

Part of the problem seems to be that read_ctrl_pos() has unusual calling conventions since commit 37a260870f2c ("mm/mglru: rework type selection") where passing MAX_NR_TIERS makes it accumulate all tiers but passing a smaller positive number makes it read a single tier instead.

Avoid this case by splitting read_ctrl_pos() into two separate helpers that each only do one of the two cases. This avoids the warning as far as I can tell, and seems a bit easier to understand to me.

Quoted 8 lines

---
This is currently the only such warning I get from gcc-16.0.1, and none from any other version.

I'm not overly happy about having to work around it with a random code chance, but hopefully the version I ended up with makes sense regardless.

Seems a large change just to squish a compiler warning. People might prefer a simple

- struct ctrl_pos sp, pv;
+ struct ctrl_pos sp, pv = {};

?

#2 lore raw
Quoted 7 lines · Andrew Morton
Quoted 3 lines · On Fri, 13 Feb 2026 13:38:56 +0100 Arnd Bergmann <[email protected]> wrote:

From: Arnd Bergmann <[email protected]>

When the -fsanitize=bounds sanitizer is enabled,

Is this an option in current kernels?

Yes, this is CONFIG_UBSAN_ARRAY_BOUNDS. The actual warning only shows up in some configurations with that, so either there is some other dependency, or an element of chance based on gcc optimizations.

Review

Quoted 5 lines

Seems a large change just to squish a compiler warning. People might prefer a simple

(quoted diff) +1 -1
-	struct ctrl_pos sp, pv;
+	struct ctrl_pos sp, pv = {};

Right, that would clearly also shut up the warning.

To me this seems less intuitive without an extra comment, since read_ctrl_pos() is meant to initialize the entire struct, but please pick whichever you find most readable here.

Arnd

maintainer
#3 lore raw
Quoted 5 lines · On Fri, 13 Feb 2026 18:07:04 +0100 "Arnd Bergmann" <[email protected]> wrote:
Quoted 4 lines · Andrew Morton
Quoted 2 lines · On Fri, 13 Feb 2026 13:38:56 +0100 Arnd Bergmann <[email protected]> wrote:

From: Arnd Bergmann <[email protected]>

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

When the -fsanitize=bounds sanitizer is enabled,

Is this an option in current kernels?

Yes, this is CONFIG_UBSAN_ARRAY_BOUNDS. The actual warning only shows up in some configurations with that, so either there is some other dependency, or an element of chance based on gcc optimizations.

OK, I'll put a cc:stable on this, as people will want to compile older kernels with gcc-16.

Aiming for upstreaming into 7.1-rc1 unless it's more urgent than I think.

Quoted 8 lines

Review

Quoted 5 lines

Seems a large change just to squish a compiler warning. People might prefer a simple

(quoted diff) +1 -1
-	struct ctrl_pos sp, pv;
+	struct ctrl_pos sp, pv = {};

Right, that would clearly also shut up the warning.

Quoted 4 lines

To me this seems less intuitive without an extra comment, since read_ctrl_pos() is meant to initialize the entire struct, but please pick whichever you find most readable here.

Let's see what the MGLRU maintainers have to say.

reviewer
#4 lore raw

Hi Andrew and Arnd,

Quoted 29 lines · Andrew Morton
Quoted 5 lines · On Fri, 13 Feb 2026 18:07:04 +0100 "Arnd Bergmann" <[email protected]> wrote:
Quoted 4 lines · Andrew Morton
Quoted 2 lines · On Fri, 13 Feb 2026 13:38:56 +0100 Arnd Bergmann <[email protected]> wrote:

From: Arnd Bergmann <[email protected]>

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

When the -fsanitize=bounds sanitizer is enabled,

Is this an option in current kernels?

Yes, this is CONFIG_UBSAN_ARRAY_BOUNDS. The actual warning only shows up in some configurations with that, so either there is some other dependency, or an element of chance based on gcc optimizations.

OK, I'll put a cc:stable on this, as people will want to compile older kernels with gcc-16.

Aiming for upstreaming into 7.1-rc1 unless it's more urgent than I think.

Quoted 6 lines
Quoted 5 lines

Seems a large change just to squish a compiler warning. People might prefer a simple

(quoted diff) +1 -1
-   struct ctrl_pos sp, pv;
+   struct ctrl_pos sp, pv = {};
Quoted 8 lines
Quoted 6 lines

Right, that would clearly also shut up the warning.

To me this seems less intuitive without an extra comment, since read_ctrl_pos() is meant to initialize the entire struct, but please pick whichever you find most readable here.

Let's see what the MGLRU maintainers have to say.

I went over the various cases of read_ctrl_pos and couldn't find anything wrong (let me know if I'm mistaken), so this seems like a mild compiler bug we're trying to work around. Given it's one single version of gcc, I'm a fan of the simple struct ctrl_pos sp, pv = {}; with a comment.

Thanks,
Yuanchu

#5 lore raw
Quoted 41 lines · Yuanchu Xie

Hi Andrew and Arnd,

Quoted 36 lines · Andrew Morton
Quoted 5 lines · On Fri, 13 Feb 2026 18:07:04 +0100 "Arnd Bergmann" <[email protected]> wrote:
Quoted 4 lines · Andrew Morton
Quoted 2 lines · On Fri, 13 Feb 2026 13:38:56 +0100 Arnd Bergmann <[email protected]> wrote:

From: Arnd Bergmann <[email protected]>

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

When the -fsanitize=bounds sanitizer is enabled,

Is this an option in current kernels?

Yes, this is CONFIG_UBSAN_ARRAY_BOUNDS. The actual warning only shows up in some configurations with that, so either there is some other dependency, or an element of chance based on gcc optimizations.

OK, I'll put a cc:stable on this, as people will want to compile older kernels with gcc-16.

Aiming for upstreaming into 7.1-rc1 unless it's more urgent than I think.

Quoted 12 lines

Review

Quoted 5 lines

Seems a large change just to squish a compiler warning. People might prefer a simple

(quoted diff) +1 -1
-   struct ctrl_pos sp, pv;
+   struct ctrl_pos sp, pv = {};

Right, that would clearly also shut up the warning.

To me this seems less intuitive without an extra comment, since read_ctrl_pos() is meant to initialize the entire struct, but please pick whichever you find most readable here.

Quoted 7 lines
Quoted 1 line

Let's see what the MGLRU maintainers have to say.

I went over the various cases of read_ctrl_pos and couldn't find anything wrong (let me know if I'm mistaken), so this seems like a mild compiler bug we're trying to work around. Given it's one single version of gcc, I'm a fan of the simple struct ctrl_pos sp, pv = {}; with a comment.

Agreed, that simpler change with a comment seems better to me.

I don't necessarily hate the refactor to split up the single / aggregate cases, but if we were going to do that I'd probably have aggregate_ctrl_pos call read_ctrl_pos to avoid duplicating the code, and I'd want to consider marking read_ctrl_pos inline.

But, it does seem like a large delta just to work around a false positive warning. The simpler change would be easier to backport to stable too for example. So on balance I'd prefer that.

Quoted 3 lines

Thanks,
Yuanchu

maintainer
#6 lore raw
Quoted 10 lines · On Tue, 17 Feb 2026 13:22:01 -0800 Axel Rasmussen <[email protected]>…
Quoted 7 lines
Quoted 1 line

Let's see what the MGLRU maintainers have to say.

I went over the various cases of read_ctrl_pos and couldn't find anything wrong (let me know if I'm mistaken), so this seems like a mild compiler bug we're trying to work around. Given it's one single version of gcc, I'm a fan of the simple struct ctrl_pos sp, pv = {}; with a comment.

Agreed, that simpler change with a comment seems better to me.

Quoted 8 lines

I don't necessarily hate the refactor to split up the single / aggregate cases, but if we were going to do that I'd probably have aggregate_ctrl_pos call read_ctrl_pos to avoid duplicating the code, and I'd want to consider marking read_ctrl_pos inline.

But, it does seem like a large delta just to work around a false positive warning. The simpler change would be easier to backport to stable too for example. So on balance I'd prefer that.

Thanks. Did anyone send a v2? If so, I misplaced it!