fix unexpected type conversions and potential overflows
Subsystems (3)
v2 cover fix unexpected type conversions and potential overflows
From: Qi Zheng <[email protected]>
Hi all,
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
If the parameter types are not large enough (like 'int') or are handled incorrectly, it can lead to severe truncation, potential overflow issues, and unexpected type conversion bugs.
This series aims to address these issues by correcting the parameter types in the relevant functions, and fixing an implicit conversion bug in memcg_state_val_in_pages().
This series is based on the next-20260323.
Comments and suggestions are welcome!
Thanks,
Qi
[1]. https://lore.kernel.org/all/acDxaEgnqPI-Z4be@hyeyoo/
Qi Zheng (4):
mm: memcontrol: correct the type of stats_updates to unsigned long
mm: memcontrol: change val type to long in
__mod_memcg_{lruvec_}state() mm: memcontrol: correct the nr_pages parameter type of
mem_cgroup_update_lru_size()
mm: memcontrol: fix unexpected massive positive number in
memcg_state_val_in_pages()
include/linux/memcontrol.h | 2 +- include/trace/events/memcg.h | 10 +++++----- mm/memcontrol.c | 37 +++++++++++++++++++----------------- 3 files changed, 26 insertions(+), 23 deletions(-)
Signature · 1 line
2.20.1
PATCH v2 1/4 mm: memcontrol: correct the type of stats_updates to unsigned long
From: Qi Zheng <[email protected]>
The memcg_rstat_updated() tracks updates for vmstats_percpu->state and lruvec_stats_percpu->state. Since these state values are of type long, change the val parameter passed to memcg_rstat_updated() to long as well.
Correspondingly, change the type of stats_updates in struct memcg_vmstats_percpu and struct memcg_vmstats from unsigned int and atomic_t to unsigned long and atomic_long_t respectively to prevent potential overflow when handling large state updates during the reparenting of LRU folios.
--- mm/memcontrol.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)
mm/memcontrol.c +9 -9
@@ -608,7 +608,7 @@ static inline int memcg_events_index(enum vm_event_item idx)struct memcg_vmstats_percpu {/* Stats updates since the last flush */ -unsigned int stats_updates; +unsigned long stats_updates;/* Cached pointers for fast iteration in memcg_rstat_updated() */ struct memcg_vmstats_percpu __percpu *parent_pcpu;
@@ -639,7 +639,7 @@ struct memcg_vmstats {unsigned long events_pending[NR_MEMCG_EVENTS ];/* Stats updates since the last flush */ -atomic_t stats_updates; +atomic_long_t stats_updates; };/*
@@ -665,16 +665,16 @@ static u64 flush_last_time;static bool memcg_vmstats_needs_flush (struct memcg_vmstats *vmstats) { -return atomic_read (&vmstats->stats_updates) > +return atomic_long_read (&vmstats->stats_updates) >MEMCG_CHARGE_BATCH *num_online_cpus (); } -static inline void memcg_rstat_updated (struct mem_cgroup *memcg,int val, +static inline void memcg_rstat_updated (struct mem_cgroup *memcg,long val,int cpu) {struct memcg_vmstats_percpu __percpu *statc_pcpu;struct memcg_vmstats_percpu *statc; -unsigned int stats_updates; +unsigned long stats_updates;if (!val)return ;
@@ -697,7 +697,7 @@ static inline void memcg_rstat_updated(struct mem_cgroup *memcg, int val,continue ; stats_updates =this_cpu_xchg (statc_pcpu->stats_updates,0 ); -atomic_add (stats_updates, &statc->vmstats->stats_updates); +atomic_long_add (stats_updates, &statc->vmstats->stats_updates); } }
@@ -705,7 +705,7 @@ static void __mem_cgroup_flush_stats(struct mem_cgroup *memcg, bool force) {bool needs_flush =memcg_vmstats_needs_flush (memcg->vmstats); -trace_memcg_flush_stats (memcg,atomic_read (&memcg->vmstats->stats_updates), +trace_memcg_flush_stats (memcg,atomic_long_read (&memcg->vmstats->stats_updates), force, needs_flush);if (!force && !needs_flush)
@@ -4406,8 +4406,8 @@ static void mem_cgroup_css_rstat_flush(struct cgroup_subsys_state *css, int cpu) }WRITE_ONCE (statc->stats_updates,0 );/* We are in a per-cpu loop here, only do the atomic write once */ -if (atomic_read (&memcg->vmstats->stats_updates)) -atomic_set (&memcg->vmstats->stats_updates,0 ); +if (atomic_long_read (&memcg->vmstats->stats_updates)) +atomic_long_set (&memcg->vmstats->stats_updates,0 ); }static void mem_cgroup_fork (struct task_struct *task)
Signature · 1 line
2.20.1
Re: [PATCH v2 1/4] mm: memcontrol: correct the type of stats_updates to unsigned long
Quoted 11 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
memcg_rstat_updated()tracks updates forvmstats_percpu->stateandlruvec_stats_percpu->state. Since these state values are of type long, change the val parameter passed tomemcg_rstat_updated()to long as well.Correspondingly, change the type of
stats_updatesin structmemcg_vmstats_percpuand structmemcg_vmstatsfrom unsigned int andatomic_tto unsigned long andatomic_long_trespectively to prevent potential overflow when handling large state updates during the reparenting of LRU folios.
Do we need a Fixes, possibly cc: stable for that? Apologies if already asked + answered.
Quoted 2 lines
Signed-off-by Qi Zheng <[email protected]>
Anyway logic seems fine to me, so:
Quoted 78 lines
--- mm/memcontrol.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)mm/memcontrol.c +9
-10@@ -608,7 +608,7 @@ static inline int memcg_events_index(enum vm_event_item idx)struct memcg_vmstats_percpu {/* Stats updates since the last flush */ -unsigned int stats_updates; +unsigned long stats_updates;/* Cached pointers for fast iteration in memcg_rstat_updated() */ struct memcg_vmstats_percpu __percpu *parent_pcpu;@@ -639,7 +639,7 @@ struct memcg_vmstats {unsigned long events_pending[NR_MEMCG_EVENTS ];/* Stats updates since the last flush */ -atomic_t stats_updates; +atomic_long_t stats_updates; };/* @@ -665,16 +665,16 @@ static u64 flush_last_time;static bool memcg_vmstats_needs_flush (struct memcg_vmstats *vmstats) { -return atomic_read (&vmstats->stats_updates) > +return atomic_long_read (&vmstats->stats_updates) >MEMCG_CHARGE_BATCH *num_online_cpus (); } -static inline void memcg_rstat_updated (struct mem_cgroup *memcg,int val, +static inline void memcg_rstat_updated (struct mem_cgroup *memcg,long val,int cpu) {struct memcg_vmstats_percpu __percpu *statc_pcpu;struct memcg_vmstats_percpu *statc; -unsigned int stats_updates; +unsigned long stats_updates;if (!val)return ;@@ -697,7 +697,7 @@ static inline void memcg_rstat_updated(struct mem_cgroup *memcg, int val,continue ; stats_updates =this_cpu_xchg (statc_pcpu->stats_updates,0 ); -atomic_add (stats_updates, &statc->vmstats->stats_updates); +atomic_long_add (stats_updates, &statc->vmstats->stats_updates); } }@@ -705,7 +705,7 @@ static void __mem_cgroup_flush_stats(struct mem_cgroup *memcg, bool force) {bool needs_flush =memcg_vmstats_needs_flush (memcg->vmstats); -trace_memcg_flush_stats (memcg,atomic_read (&memcg->vmstats->stats_updates), +trace_memcg_flush_stats (memcg,atomic_long_read (&memcg->vmstats->stats_updates), force, needs_flush);if (!force && !needs_flush)@@ -4406,8 +4406,8 @@ static void mem_cgroup_css_rstat_flush(struct cgroup_subsys_state *css, int cpu) }WRITE_ONCE (statc->stats_updates,0 );/* We are in a per-cpu loop here, only do the atomic write once */ -if (atomic_read (&memcg->vmstats->stats_updates)) -atomic_set (&memcg->vmstats->stats_updates,0 ); +if (atomic_long_read (&memcg->vmstats->stats_updates)) +atomic_long_set (&memcg->vmstats->stats_updates,0 ); }static void mem_cgroup_fork (struct task_struct *task) --2.20.1
Re: [PATCH v2 1/4] mm: memcontrol: correct the type of stats_updates to unsigned long
Quoted 7 lines · Lorenzo Stoakes (Oracle)
Quoted 6 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
memcg_rstat_updated()tracks updates forvmstats_percpu->stateandlruvec_stats_percpu->state. Since these state values are of type long, change the val parameter passed tomemcg_rstat_updated()to long as well.
Quoted 8 lines
Quoted 5 lines
Correspondingly, change the type of
stats_updatesin structmemcg_vmstats_percpuand structmemcg_vmstatsfrom unsigned int andatomic_tto unsigned long andatomic_long_trespectively to prevent potential overflow when handling large state updates during the reparenting of LRU folios.Do we need a Fixes, possibly cc: stable for that? Apologies if already asked + answered.
Before LRU folio reparenting was introduced, we wouldn’t pass in such a large value, so this wasn’t a problem. Since LRU folio reparenting is still in mm-unstable, so I didn't add a Fixes tag in [4/4].
Quoted 7 lines
Quoted 3 lines
Signed-off-by Qi Zheng <[email protected]>Anyway logic seems fine to me, so:
Reviewed-by Lorenzo Stoakes (Oracle) <[email protected]>
Thanks!
Re: [PATCH v2 1/4] mm: memcontrol: correct the type of stats_updates to unsigned long
Quoted 16 lines · Qi Zheng
Quoted 13 lines · Lorenzo Stoakes (Oracle)
Quoted 11 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
memcg_rstat_updated()tracks updates forvmstats_percpu->stateandlruvec_stats_percpu->state. Since these state values are of type long, change the val parameter passed tomemcg_rstat_updated()to long as well.Correspondingly, change the type of
stats_updatesin structmemcg_vmstats_percpuand structmemcg_vmstatsfrom unsigned int andatomic_tto unsigned long andatomic_long_trespectively to prevent potential overflow when handling large state updates during the reparenting of LRU folios.
Quoted 6 lines
Quoted 2 lines
Do we need a Fixes, possibly cc: stable for that? Apologies if already asked + answered.
Before LRU folio reparenting was introduced, we wouldn’t pass in such a large value, so this wasn’t a problem. Since LRU folio reparenting is still in mm-unstable, so I didn't add a Fixes tag in [4/4].
Ah, well these patches should be _before_ the LRU folio reparenting then?
Andrew - can we ensure correct ordering here?
Quoted 12 lines
Quoted 8 lines
Quoted 4 lines
Signed-off-by Qi Zheng <[email protected]>Anyway logic seems fine to me, so:
Reviewed-by Lorenzo Stoakes (Oracle) <[email protected]>Thanks!
Thanks, Lorenzo
Re: [PATCH v2 1/4] mm: memcontrol: correct the type of stats_updates to unsigned long
Quoted 17 lines · Lorenzo Stoakes (Oracle)
Quoted 16 lines · Qi Zheng
Quoted 13 lines · Lorenzo Stoakes (Oracle)
Quoted 11 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
memcg_rstat_updated()tracks updates forvmstats_percpu->stateandlruvec_stats_percpu->state. Since these state values are of type long, change the val parameter passed tomemcg_rstat_updated()to long as well.Correspondingly, change the type of
stats_updatesin structmemcg_vmstats_percpuand structmemcg_vmstatsfrom unsigned int andatomic_tto unsigned long andatomic_long_trespectively to prevent potential overflow when handling large state updates during the reparenting of LRU folios.
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
Do we need a Fixes, possibly cc: stable for that? Apologies if already asked + answered.
Before LRU folio reparenting was introduced, we wouldn’t pass in such a large value, so this wasn’t a problem. Since LRU folio reparenting is still in mm-unstable, so I didn't add a Fixes tag in [4/4].
Ah, well these patches should be _before_ the LRU folio reparenting then?
Yes. I think that'll be the best option.
If that's too much of a headache at this point (I'm not sure), it should be at least part of 7.1-rcX, given that it's quite unlikely that people will notice it during bisection anyway...
Signature · 2 lines
Cheers, Harry / Hyeonggon
Re: [PATCH v2 1/4] mm: memcontrol: correct the type of stats_updates to unsigned long
In reply to Lorenzo Stoakes (Oracle)Quoted 20 lines · Lorenzo Stoakes (Oracle)
Quoted 19 lines · Qi Zheng
Quoted 15 lines · Lorenzo Stoakes (Oracle)
Quoted 11 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
memcg_rstat_updated()tracks updates forvmstats_percpu->stateandlruvec_stats_percpu->state. Since these state values are of type long, change the val parameter passed tomemcg_rstat_updated()to long as well.Correspondingly, change the type of
stats_updatesin structmemcg_vmstats_percpuand structmemcg_vmstatsfrom unsigned int andatomic_tto unsigned long andatomic_long_trespectively to prevent potential overflow when handling large state updates during the reparenting of LRU folios.Do we need a Fixes, possibly cc: stable for that? Apologies if already asked + answered.
Quoted 7 lines
Quoted 3 lines
Before LRU folio reparenting was introduced, we wouldn’t pass in such a large value, so this wasn’t a problem. Since LRU folio reparenting is still in mm-unstable, so I didn't add a Fixes tag in [4/4].
Ah, well these patches should be _before_ the LRU folio reparenting then?
Andrew - can we ensure correct ordering here?
There are some dependencies for this.
To be precise, it should be applied after:
[PATCH v6 29/33] mm: memcontrol: refactor mod_memcg_state() and mod_memcg_lruvec_state()
and before:
[PATCH v6 32/33] mm: memcontrol: eliminate the problem of dying memory cgroup for LRU folios
and there might be conflicts.
Would it be okay to merge them together into v7.1-rcX? Otherwise, perhaps updating to v7 would be more convenient for Andrew.
Hi Andrew, what do you think?
Thanks,
Qi
Quoted 15 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 9 lines
Quoted 5 lines
Quoted 1 line
Signed-off-by Qi Zheng <[email protected]>Anyway logic seems fine to me, so:
Reviewed-by Lorenzo Stoakes (Oracle) <[email protected]>Thanks!
Thanks, Lorenzo
PATCH v2 2/4 mm: memcontrol: change val type to long in __mod_memcg_{lruvec_}state()
In reply to Qi ZhengFrom: Qi Zheng <[email protected]>
The __mod_memcg_state() and __mod_memcg_lruvec_state() functions are also used to reparent non-hierarchical stats. In this scenario, the values passed to them are accumulated statistics that might be extremely large and exceed the upper limit of a 32-bit integer.
Change the val parameter type from int to long in these functions and their corresponding tracepoints (memcg_rstat_stats) to prevent potential overflow issues.
--- include/trace/events/memcg.h | 10 +++++----- mm/memcontrol.c | 8 ++++---- 2 files changed, 9 insertions(+), 9 deletions(-)
include/trace/events/memcg.h +5 -5
@@ -11,14 +11,14 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val), TP_ARGS(memcg, item, val), TP_STRUCT__entry( __field(u64, id) __field(int, item) - __field(int, val) + __field(long, val) ), TP_fast_assign(
@@ -27,20 +27,20 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, __entry->val = val; ), -TP_printk ("memcg_id=%llu item=%d val=%d" , +TP_printk ("memcg_id=%llu item=%d val=%ld" , __entry->id, __entry->item, __entry->val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_lruvec_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );
mm/memcontrol.c +4 -4
@@ -527,7 +527,7 @@ unsigned long lruvec_page_state_local(struct lruvec *lruvec,#ifdef CONFIG_MEMCG_V1 static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val); +enum node_stat_item idx,long val);void reparent_memcg_lruvec_state_local (struct mem_cgroup *memcg,struct mem_cgroup *parent,int idx)
@@ -784,7 +784,7 @@ static int memcg_page_state_unit(int item);* Normalize the value passed into memcg_rstat_updated() to be in pages. Round * up non-zero sub-page updates to 1 page as zero page updates are ignored. */ -static int memcg_state_val_in_pages (int idx,int val) +static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx);
@@ -831,7 +831,7 @@ static inline void get_non_dying_memcg_end(void)#endif static void __mod_memcg_state (struct mem_cgroup *memcg, -enum memcg_stat_item idx,int val) +enum memcg_stat_item idx,long val) {int i =memcg_stats_index (idx);int cpu;
@@ -896,7 +896,7 @@ void reparent_memcg_state_local(struct mem_cgroup *memcg,#endif static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val) +enum node_stat_item idx,long val) {struct mem_cgroup *memcg = pn->memcg;int i =memcg_stats_index (idx);
Signature · 1 line
2.20.1
Re: [PATCH v2 2/4] mm: memcontrol: change val type to long in __mod_memcg_{lruvec_}state()
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
__mod_memcg_state()and__mod_memcg_lruvec_state()functions are also used to reparent non-hierarchical stats. In this scenario, the values passed to them are accumulated statistics that might be extremely large and exceed the upper limit of a 32-bit integer.
Quoted 5 lines
Change the val parameter type from int to long in these functions and their corresponding tracepoints (
memcg_rstat_stats) to prevent potential overflow issues.Signed-off-by Qi Zheng <[email protected]>
LGTM, so:
Quoted 93 lines
--- include/trace/events/memcg.h | 10 +++++----- mm/memcontrol.c | 8 ++++---- 2 files changed, 9 insertions(+), 9 deletions(-)include/trace/events/memcg.h +5
-5@@ -11,14 +11,14 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val), TP_ARGS(memcg, item, val), TP_STRUCT__entry( __field(u64, id) __field(int, item) - __field(int, val) + __field(long, val) ), TP_fast_assign(@@ -27,20 +27,20 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, __entry->val = val; ), -TP_printk ("memcg_id=%llu item=%d val=%d" , +TP_printk ("memcg_id=%llu item=%d val=%ld" , __entry->id, __entry->item, __entry->val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_lruvec_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );mm/memcontrol.c +4
-5@@ -527,7 +527,7 @@ unsigned long lruvec_page_state_local(struct lruvec *lruvec,#ifdef CONFIG_MEMCG_V1 static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val); +enum node_stat_item idx,long val);void reparent_memcg_lruvec_state_local (struct mem_cgroup *memcg,struct mem_cgroup *parent,int idx)@@ -784,7 +784,7 @@ static int memcg_page_state_unit(int item);* Normalize the value passed into memcg_rstat_updated() to be in pages. Round * up non-zero sub-page updates to 1 page as zero page updates are ignored. */ -static int memcg_state_val_in_pages (int idx,int val) +static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx);@@ -831,7 +831,7 @@ static inline void get_non_dying_memcg_end(void)#endif static void __mod_memcg_state (struct mem_cgroup *memcg, -enum memcg_stat_item idx,int val) +enum memcg_stat_item idx,long val) {int i =memcg_stats_index (idx);int cpu;@@ -896,7 +896,7 @@ void reparent_memcg_state_local(struct mem_cgroup *memcg,#endif static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val) +enum node_stat_item idx,long val) {struct mem_cgroup *memcg = pn->memcg;int i =memcg_stats_index (idx); --2.20.1
Re: [PATCH v2 2/4] mm: memcontrol: change val type to long in __mod_memcg_{lruvec_}state()
Quoted 11 lines · On Thu, 26 Mar 2026 09:19:29 +0000 "Lorenzo Stoakes (Oracle)" <[email protected]>…
Quoted 10 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
__mod_memcg_state()and__mod_memcg_lruvec_state()functions are also used to reparent non-hierarchical stats. In this scenario, the values passed to them are accumulated statistics that might be extremely large and exceed the upper limit of a 32-bit integer.Change the val parameter type from int to long in these functions and their corresponding tracepoints (
memcg_rstat_stats) to prevent potential overflow issues.
Won't that are be true on 32bit systems?
Which means the underlying values need to be u64 not long.
David
Quoted 101 lines
Quoted 2 lines
Signed-off-by Qi Zheng <[email protected]>LGTM, so:
Reviewed-by Lorenzo Stoakes (Oracle) <[email protected]>Quoted 93 lines
--- include/trace/events/memcg.h | 10 +++++----- mm/memcontrol.c | 8 ++++---- 2 files changed, 9 insertions(+), 9 deletions(-)include/trace/events/memcg.h +5
-5@@ -11,14 +11,14 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val), TP_ARGS(memcg, item, val), TP_STRUCT__entry( __field(u64, id) __field(int, item) - __field(int, val) + __field(long, val) ), TP_fast_assign(@@ -27,20 +27,20 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, __entry->val = val; ), -TP_printk ("memcg_id=%llu item=%d val=%d" , +TP_printk ("memcg_id=%llu item=%d val=%ld" , __entry->id, __entry->item, __entry->val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_lruvec_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );mm/memcontrol.c +4
-5@@ -527,7 +527,7 @@ unsigned long lruvec_page_state_local(struct lruvec *lruvec,#ifdef CONFIG_MEMCG_V1 static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val); +enum node_stat_item idx,long val);void reparent_memcg_lruvec_state_local (struct mem_cgroup *memcg,struct mem_cgroup *parent,int idx)@@ -784,7 +784,7 @@ static int memcg_page_state_unit(int item);* Normalize the value passed into memcg_rstat_updated() to be in pages. Round * up non-zero sub-page updates to 1 page as zero page updates are ignored. */ -static int memcg_state_val_in_pages (int idx,int val) +static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx);@@ -831,7 +831,7 @@ static inline void get_non_dying_memcg_end(void)#endif static void __mod_memcg_state (struct mem_cgroup *memcg, -enum memcg_stat_item idx,int val) +enum memcg_stat_item idx,long val) {int i =memcg_stats_index (idx);int cpu;@@ -896,7 +896,7 @@ void reparent_memcg_state_local(struct mem_cgroup *memcg,#endif static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val) +enum node_stat_item idx,long val) {struct mem_cgroup *memcg = pn->memcg;int i =memcg_stats_index (idx); --2.20.1
Re: [PATCH v2 2/4] mm: memcontrol: change val type to long in __mod_memcg_{lruvec_}state()
Quoted 11 lines · David Laight
Quoted 8 lines · On Thu, 26 Mar 2026 09:19:29 +0000 "Lorenzo Stoakes (Oracle)" <[email protected]>…
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
The
__mod_memcg_state()and__mod_memcg_lruvec_state()functions are also used to reparent non-hierarchical stats. In this scenario, the values passed to them are accumulated statistics that might be extremely large and exceed the upper limit of a 32-bit integer.
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
Change the val parameter type from int to long in these functions and their corresponding tracepoints (
memcg_rstat_stats) to prevent potential overflow issues.Won't that are be true on 32bit systems?
Which means the underlying values need to be u64 not long.
On a 32-bit system, such a large number of values should not accumulate.
Furthermore, changing it to u64/s64 would mean modifying memcg_vmstats_percpu->state and lruvec_stats_percpu->state, which involves significant changes, so I think it might be unnecessary.
Thanks,
Qi
Quoted 105 lines
David
Quoted 101 lines
Quoted 2 lines
Signed-off-by Qi Zheng <[email protected]>LGTM, so:
Reviewed-by Lorenzo Stoakes (Oracle) <[email protected]>Quoted 93 lines
--- include/trace/events/memcg.h | 10 +++++----- mm/memcontrol.c | 8 ++++---- 2 files changed, 9 insertions(+), 9 deletions(-)include/trace/events/memcg.h +5
-5@@ -11,14 +11,14 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val), TP_ARGS(memcg, item, val), TP_STRUCT__entry( __field(u64, id) __field(int, item) - __field(int, val) + __field(long, val) ), TP_fast_assign(@@ -27,20 +27,20 @@ DECLARE_EVENT_CLASS(memcg_rstat_stats, __entry->val = val; ), -TP_printk ("memcg_id=%llu item=%d val=%d" , +TP_printk ("memcg_id=%llu item=%d val=%ld" , __entry->id, __entry->item, __entry->val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );DEFINE_EVENT (memcg_rstat_stats, mod_memcg_lruvec_state, - TP_PROTO(struct mem_cgroup *memcg, int item, int val), + TP_PROTO(struct mem_cgroup *memcg, int item, long val),TP_ARGS (memcg, item, val) );mm/memcontrol.c +4
-5@@ -527,7 +527,7 @@ unsigned long lruvec_page_state_local(struct lruvec *lruvec,#ifdef CONFIG_MEMCG_V1 static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val); +enum node_stat_item idx,long val);void reparent_memcg_lruvec_state_local (struct mem_cgroup *memcg,struct mem_cgroup *parent,int idx)@@ -784,7 +784,7 @@ static int memcg_page_state_unit(int item);* Normalize the value passed into memcg_rstat_updated() to be in pages. Round * up non-zero sub-page updates to 1 page as zero page updates are ignored. */ -static int memcg_state_val_in_pages (int idx,int val) +static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx);@@ -831,7 +831,7 @@ static inline void get_non_dying_memcg_end(void)#endif static void __mod_memcg_state (struct mem_cgroup *memcg, -enum memcg_stat_item idx,int val) +enum memcg_stat_item idx,long val) {int i =memcg_stats_index (idx);int cpu;@@ -896,7 +896,7 @@ void reparent_memcg_state_local(struct mem_cgroup *memcg,#endif static void __mod_memcg_lruvec_state (struct mem_cgroup_per_node *pn, -enum node_stat_item idx,int val) +enum node_stat_item idx,long val) {struct mem_cgroup *memcg = pn->memcg;int i =memcg_stats_index (idx); --2.20.1
PATCH v2 3/4 mm: memcontrol: correct the nr_pages parameter type of mem_cgroup_update_lru_size()
In reply to Qi ZhengFrom: Qi Zheng <[email protected]>
The nr_pages parameter of mem_cgroup_update_lru_size() represents a page count. During the reparenting of LRU folios, the value passed to it can potentially exceed the maximum value of a 32-bit integer. It should be declared as long instead of int to match the types used in lruvec size accounting and to prevent possible overflow.
Update the parameter type to long to ensure correctness.
--- include/linux/memcontrol.h | 2 +- mm/memcontrol.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-)
include/linux/memcontrol.h +1 -1
@@ -878,7 +878,7 @@ static inline bool mem_cgroup_online(struct mem_cgroup *memcg) }void mem_cgroup_update_lru_size (struct lruvec *lruvec,enum lru_list lru, -int zid,int nr_pages); +int zid,long nr_pages); static inline unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
mm/memcontrol.c +2 -2
@@ -1466,7 +1466,7 @@ struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,* to or just after a page is removed from an lru list. */ void mem_cgroup_update_lru_size (struct lruvec *lruvec,enum lru_list lru, -int zid,int nr_pages) +int zid,long nr_pages) {struct mem_cgroup_per_node *mz;unsigned long *lru_size;
@@ -1483,7 +1483,7 @@ void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, size = *lru_size;if (WARN_ONCE (size <0 , -"%s(%p, %d, %d): lru_size %ld\n" , +"%s(%p, %d, %ld): lru_size %ld\n" , __func__, lruvec, lru, nr_pages, size)) {VM_BUG_ON (1 ); *lru_size =0 ;
Signature · 1 line
2.20.1
PATCH v2 4/4 mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
In reply to Qi ZhengFrom: Qi Zheng <[email protected]>
In memcg_state_val_in_pages(), if the passed val is negative, the expression val * unit / PAGE_SIZE could be implicitly converted to a massive positive number when compared with 1UL in the max() macro. This leads to returning an incorrect massive positive value.
Fix this by using abs(val) to calculate the magnitude first, and then restoring the sign of the value before returning the result. Additionally, use mult_frac() to prevent potential overflow during the multiplication of val and unit.
--- mm/memcontrol.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
mm/memcontrol.c +5 -2
@@ -787,11 +787,14 @@ static int memcg_page_state_unit(int item);static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx); +long res;if (!val || unit ==PAGE_SIZE )return val; -else -return max (val * unit /PAGE_SIZE ,1UL ); + + res =max (mult_frac (abs (val), unit,PAGE_SIZE ),1UL ); + +return val <0 ? -res : res; }#ifdef CONFIG_MEMCG_V1
Signature · 1 line
2.20.1
Re: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
In
memcg_state_val_in_pages(), if the passed val is negative, the expression val * unit /PAGE_SIZEcould be implicitly converted to a massive positive number when compared with 1UL in themax()macro. This leads to returning an incorrect massive positive value.
Quoted 7 lines
Fix this by using abs(val) to calculate the magnitude first, and then restoring the sign of the value before returning the result. Additionally, use
mult_frac()to prevent potential overflow during the multiplication of val and unit.Reported-by Harry Yoo (Oracle) <[email protected]>Signed-off-by Qi Zheng <[email protected]>
The logic is correct, but I think this needs rework for better understanding, and obviously this should be squashed into 2/4 as per Andrew.
With the below change applied:
Quoted 9 lines
--- mm/memcontrol.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)mm/memcontrol.c +0
-0
mm/memcontrol.c static int memcg_page_state_unit(int item); · line 794
Quoted 9 lines
mm/memcontrol.c +1
-2@@ -787,11 +787,14 @@ static int memcg_page_state_unit(int item);static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx); +long res;if (!val || unit ==PAGE_SIZE )return val; -else -return max (val * unit /PAGE_SIZE ,1UL );
Hm this was already fairly horrid, because we're comparing an unsigned long value of 1 vs. a ULONG_MAX - abs(val) so this was intended to make 0 -> 1UL but not what you'd mathematically think this was which was to make negative values (logically < 1) -> 1.
Of course before it was just broken and would promote (val * unit / PAGE_SIZE) to unsigned long first (thus massive number) and return that :)
mm/memcontrol.c static int memcg_page_state_unit(int item); · line 795
Quoted 2 lines
mm/memcontrol.c +2
-0+ + res =max (mult_frac (abs (val), unit,PAGE_SIZE ),1UL );
This is way too compressed into one line and retains the confusing behaviour.
Could we split this out and explain what we're doing (sign-extension, integer promotion and all of this stuff is confusing - so let's just accept that and spell it out):
/* Get the absolute value of (val * unit / PAGE_SIZE). */ res = mult_frac(abs(val), unit, PAGE_SIZE); /* Round up zero values. */ res = res ?: 1; /* Retain sign. */ return val < 0 ? -res : res;
This is functionally identical, but a lot more readable, I think.
Quoted 8 lines
mm/memcontrol.c +2
-1+ +return val <0 ? -res : res; }#ifdef CONFIG_MEMCG_V1 --2.20.1
Cheers, Lorenzo
Re: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 6 lines · Lorenzo Stoakes (Oracle)
/* Get the absolute value of (val * unit / PAGE_SIZE). */ res = mult_frac(abs(val), unit, PAGE_SIZE); /* Round up zero values. */ res = res ?: 1; /* Retain sign. */ return val < 0 ? -res : res;
(If you think the comments are too much, you can drop them, and the code essentially spells it out anyway)
Cheers, Lorenzo
Re: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
In reply to Lorenzo Stoakes (Oracle)Quoted 61 lines · Lorenzo Stoakes (Oracle)
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
In
memcg_state_val_in_pages(), if the passed val is negative, the expression val * unit /PAGE_SIZEcould be implicitly converted to a massive positive number when compared with 1UL in themax()macro. This leads to returning an incorrect massive positive value.Quoted 7 lines
Fix this by using abs(val) to calculate the magnitude first, and then restoring the sign of the value before returning the result. Additionally, use
mult_frac()to prevent potential overflow during the multiplication of val and unit.Reported-by Harry Yoo (Oracle) <[email protected]>Signed-off-by Qi Zheng <[email protected]>The logic is correct, but I think this needs rework for better understanding, and obviously this should be squashed into 2/4 as per Andrew.
With the below change applied:
Reviewed-by Lorenzo Stoakes (Oracle) <[email protected]>Quoted 9 lines
--- mm/memcontrol.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)mm/memcontrol.c +0
-0mm/memcontrol.c static int memcg_page_state_unit(int item); · line 794
Quoted 9 lines
mm/memcontrol.c +1
-2@@ -787,11 +787,14 @@ static int memcg_page_state_unit(int item);static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx); +long res;if (!val || unit ==PAGE_SIZE )return val; -else -return max (val * unit /PAGE_SIZE ,1UL );Hm this was already fairly horrid, because we're comparing an unsigned long value of 1 vs. a
ULONG_MAX- abs(val) so this was intended to make 0 -> 1UL but not what you'd mathematically think this was which was to make negative values (logically < 1) -> 1.Of course before it was just broken and would promote (val * unit /
PAGE_SIZE) to unsigned long first (thus massive number) and return that :)mm/memcontrol.c static int memcg_page_state_unit(int item); · line 795
Quoted 2 lines
mm/memcontrol.c +2
-0+ + res =max (mult_frac (abs (val), unit,PAGE_SIZE ),1UL );This is way too compressed into one line and retains the confusing behaviour.
Could we split this out and explain what we're doing (sign-extension, integer promotion and all of this stuff is confusing - so let's just accept that and spell it out):
Quoted 8 lines
/* Get the absolute value of (val * unit / PAGE_SIZE). */ res = mult_frac(abs(val), unit, PAGE_SIZE); /* Round up zero values. */ res = res ?: 1; /* Retain sign. */ return val < 0 ? -res : res;This is functionally identical, but a lot more readable, I think.
Make sense, I will update to v3.
If Andrew needs me to merge this patchset into "[PATCH v6 00/33] Eliminate Dying Memory Cgroup" [1], then I will develop and send v7.
[1]. https://lore.kernel.org/all/[email protected]/
Thanks,
Qi
Quoted 11 lines
Quoted 9 lines
mm/memcontrol.c +2
-1+ +return val <0 ? -res : res; }#ifdef CONFIG_MEMCG_V1 --2.20.1
Cheers, Lorenzo
Re: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 48 lines · Qi Zheng
Quoted 45 lines · Lorenzo Stoakes (Oracle)
Quoted 9 lines · Qi Zheng
--- mm/memcontrol.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)mm/memcontrol.c +0
-0mm/memcontrol.c static int memcg_page_state_unit(int item); · line 794
Quoted 9 lines
mm/memcontrol.c +1
-2@@ -787,11 +787,14 @@ static int memcg_page_state_unit(int item);static long memcg_state_val_in_pages (int idx,long val) {int unit =memcg_page_state_unit (idx); +long res;if (!val || unit ==PAGE_SIZE )return val; -else -return max (val * unit /PAGE_SIZE ,1UL );Hm this was already fairly horrid, because we're comparing an unsigned long value of 1 vs. a
ULONG_MAX- abs(val) so this was intended to make 0 -> 1UL but not what you'd mathematically think this was which was to make negative values (logically < 1) -> 1.Of course before it was just broken and would promote (val * unit /
PAGE_SIZE) to unsigned long first (thus massive number) and return that :)mm/memcontrol.c static int memcg_page_state_unit(int item); · line 795
Quoted 2 lines
mm/memcontrol.c +2
-0+ + res =max (mult_frac (abs (val), unit,PAGE_SIZE ),1UL );This is way too compressed into one line and retains the confusing behaviour.
Could we split this out and explain what we're doing (sign-extension, integer promotion and all of this stuff is confusing - so let's just accept that and spell it out):
/* Get the absolute value of (val * unit / PAGE_SIZE). */ res = mult_frac(abs(val), unit, PAGE_SIZE); /* Round up zero values. */ res = res ?: 1; /* Retain sign. */ return val < 0 ? -res : res;
Quoted 3 lines
Quoted 1 line
This is functionally identical, but a lot more readable, I think.
Make sense, I will update to v3.
Thanks!
Quoted 6 lines
If Andrew needs me to merge this patchset into "[PATCH v6 00/33] Eliminate Dying Memory Cgroup" [1], then I will develop and send v7.
Oh that's your series too :)
That would be ideal unless that's already in mm-stable, as the series ordering will give us strict ordering on patches.
Anyway let's wait for Andrew on that!
Cheers, Lorenzo
Andrew Morton
maintainerRe: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 7 lines · On Thu, 26 Mar 2026 09:38:02 +0000 "Lorenzo Stoakes (Oracle)" <[email protected]>…
Quoted 6 lines
If Andrew needs me to merge this patchset into "[PATCH v6 00/33] Eliminate Dying Memory Cgroup" [1], then I will develop and send v7.
Quoted 6 lines
Oh that's your series too :)
That would be ideal unless that's already in mm-stable, as the series ordering will give us strict ordering on patches.
Anyway let's wait for Andrew on that!
<trying to catch up here>
Gee, I'd rather not churn that 33 patch series. Could of course do so in the normal fashion, if that's considered particularly desirable.
As I understand it, Qi will be preparing a v3 and I should stage that ahead of "Eliminate ..." to avoid a bisection hole?
If so, that works.
<checks>
Well dang, this series ("fix unexpected type conversions and potential overflows") is at least textually dependent on "Eliminate ...".
Options:
a) Redo this "fix unexpected ..." series on top of "Eliminate ..." and tolerate the bisection hole (easiest).
Am I correct in believing that the concern here is a runtime bisection hole? And that the bug is pretty unlikely to hit even if our unlucky bisector happens to hit that hole? If so, we can live with that, surely. Every darn hotfix we do creates a runtime bisecton hole!
b) Redo this series on top of mm-stable or mainline or whatever then I stage this series ahead of "Eliminate ..." and fix up the merge issues (probably OK)
c) Qi redoes everything as a single series. That's OK.
If we choose c) then please lmk and I'll drop both series to give Qi a clean run at mm-unstable.
Re: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 11 lines · Andrew Morton
Quoted 9 lines · On Thu, 26 Mar 2026 09:38:02 +0000 "Lorenzo Stoakes (Oracle)" <[email protected]>…
Quoted 6 lines
If Andrew needs me to merge this patchset into "[PATCH v6 00/33] Eliminate Dying Memory Cgroup" [1], then I will develop and send v7.
Oh that's your series too :)
Quoted 8 lines
Quoted 4 lines
That would be ideal unless that's already in mm-stable, as the series ordering will give us strict ordering on patches.
Anyway let's wait for Andrew on that!
<trying to catch up here>
Gee, I'd rather not churn that 33 patch series. Could of course do so
Agree.
Quoted 9 lines
in the normal fashion, if that's considered particularly desirable.
As I understand it, Qi will be preparing a v3 and I should stage that ahead of "Eliminate ..." to avoid a bisection hole?
If so, that works.
<checks>
Quoted 7 lines
Well dang, this series ("fix unexpected type conversions and potential overflows") is at least textually dependent on "Eliminate ...".
Options:
a) Redo this "fix unexpected ..." series on top of "Eliminate ..." and tolerate the bisection hole (easiest).
From my personal perspective, I prefer this approach. The issues fixed by this patchset aren't too critical; it's just that the counter might overflow (and only with CONFIG_MEMCG_V1). If it can be included in v7.1-rcX along with "Eliminate ...", I think it's acceptable.
Thanks,
Qi
Quoted 16 lines
Am I correct in believing that the concern here is a runtime bisection hole? And that the bug is pretty unlikely to hit even if our unlucky bisector happens to hit that hole? If so, we can live with that, surely. Every darn hotfix we do creates a runtime bisecton hole!
b) Redo this series on top of mm-stable or mainline or whatever then I stage this series ahead of "Eliminate ..." and fix up the merge issues (probably OK)
c) Qi redoes everything as a single series. That's OK.
If we choose c) then please lmk and I'll drop both series to give Qi a clean run at mm-unstable.
Andrew Morton
maintainerRe: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 7 lines · On Fri, 27 Mar 2026 10:42:52 +0800 Qi Zheng <[email protected]> wrote:
Quoted 2 lines
a) Redo this "fix unexpected ..." series on top of "Eliminate ..." and tolerate the bisection hole (easiest).
From my personal perspective, I prefer this approach. The issues fixed
by this patchset aren't too critical; it's just that the counter might overflow (and only withCONFIG_MEMCG_V1). If it can be included in v7.1-rcX along with "Eliminate ...", I think it's acceptable.
Cool, thanks.
I've dropped "fix unexpected type conversions and potential overflows v2" from mm.git to give you a clean mm-unstable to work against. I expect to be pushing this a couple of hours from now.
Re: [PATCH v2 4/4] mm: memcontrol: fix unexpected massive positive number in memcg_state_val_in_pages()
Quoted 10 lines · Andrew Morton
Quoted 7 lines · On Fri, 27 Mar 2026 10:42:52 +0800 Qi Zheng <[email protected]> wrote:
Quoted 2 lines
a) Redo this "fix unexpected ..." series on top of "Eliminate ..." and tolerate the bisection hole (easiest).
From my personal perspective, I prefer this approach. The issues fixed
by this patchset aren't too critical; it's just that the counter might overflow (and only withCONFIG_MEMCG_V1). If it can be included in v7.1-rcX along with "Eliminate ...", I think it's acceptable.
Quoted 6 lines
Cool, thanks.
I've dropped "fix unexpected type conversions and potential overflows v2" from mm.git to give you a clean mm-unstable to work against. I expect to be pushing this a couple of hours from now.
Yeah this works for me too, I think given low chance of hitting we can tolerate a small bisection hole!
Forgot to include the changelog:
Changes in v2:
- modified all commit messages. (suggested-by Lorenzo Stoakes)
- added a fix patch to resolve the unexpected massive positive number (pointed-by Harry Yoo and sashiko)
- fix the print type mismatch in [PATCH 3/3]
- collect Reviewed-by
Quoted 39 lines · Qi Zheng
From: Qi Zheng <[email protected]>
Hi all,
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
If the parameter types are not large enough (like 'int') or are handled incorrectly, it can lead to severe truncation, potential overflow issues, and unexpected type conversion bugs.
This series aims to address these issues by correcting the parameter types in the relevant functions, and fixing an implicit conversion bug in
memcg_state_val_in_pages().This series is based on the next-20260323.
Comments and suggestions are welcome!
Thanks,
Qi[1]. https://lore.kernel.org/all/acDxaEgnqPI-Z4be@hyeyoo/
Qi Zheng (4):
mm: memcontrol: correct the type ofstats_updatesto unsigned long
mm: memcontrol: change val type to long in__mod_memcg_{lruvec_}state()mm: memcontrol: correct the
nr_pagesparameter type of
mem_cgroup_update_lru_size()
mm: memcontrol: fix unexpected massive positive number in
memcg_state_val_in_pages()include/linux/memcontrol.h | 2 +- include/trace/events/memcg.h | 10 +++++----- mm/memcontrol.c | 37 +++++++++++++++++++----------------- 3 files changed, 26 insertions(+), 23 deletions(-)
Andrew Morton
maintainerQuoted 11 lines · On Wed, 25 Mar 2026 22:13:21 +0800 Qi Zheng <[email protected]> wrote:
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
If the parameter types are not large enough (like 'int') or are handled incorrectly, it can lead to severe truncation, potential overflow issues, and unexpected type conversion bugs.
This series aims to address these issues by correcting the parameter types in the relevant functions, and fixing an implicit conversion bug in
memcg_state_val_in_pages().
Thanks. I'll add this to mm.git's mm-new branch.
AI review
(https://sashiko.dev/#/patchset/cover.1774447069.git.zhengqi.arch%40bytedance.com)
still points at the problem in [2/4], now describing it as a bisection hole.
In
https://lkml.kernel.org/r/[email protected]
you said you were going to address this by using abs(), and I see that being done in [4/4] so yup, runtime bisection hole.
I'm inclined to mark this "don't care". But if we decide to backport [2/4] ("to prevent potential overflow issues") then we might have a problem.
Also, if some downstream person decides to backport [2/4] into their kernel without [4/4] then they'll have a bad day.
So perhaps this issue should be addressed within [2/4]?
Andrew Morton
maintainerQuoted 8 lines · On Wed, 25 Mar 2026 16:57:16 -0700 Andrew Morton <[email protected]>…
I'm inclined to mark this "don't care". But if we decide to backport [2/4] ("to prevent potential overflow issues") then we might have a problem.
Also, if some downstream person decides to backport [2/4] into their kernel without [4/4] then they'll have a bad day.
So perhaps this issue should be addressed within [2/4]?
Actually I can trivially do this locally if you like - just turn [4/4] into a -fix patch against [2/4], squash them together later on.
Please lmk if you'd like me to do that.
Quoted 21 lines · Andrew Morton
Quoted 11 lines · On Wed, 25 Mar 2026 22:13:21 +0800 Qi Zheng <[email protected]> wrote:
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
If the parameter types are not large enough (like 'int') or are handled incorrectly, it can lead to severe truncation, potential overflow issues, and unexpected type conversion bugs.
This series aims to address these issues by correcting the parameter types in the relevant functions, and fixing an implicit conversion bug in
memcg_state_val_in_pages().Thanks. I'll add this to mm.git's mm-new branch.
AI review
(https://sashiko.dev/#/patchset/cover.1774447069.git.zhengqi.arch%40bytedance.com)
still points at the problem in [2/4], now describing it as a bisection hole.
Quoted 4 lines
In
https://lkml.kernel.org/r/[email protected]
you said you were going to address this by usingabs(), and I see that being done in [4/4] so yup, runtime bisection hole.
Right.
Quoted 7 lines
I'm inclined to mark this "don't care". But if we decide to backport [2/4] ("to prevent potential overflow issues") then we might have a problem.
Also, if some downstream person decides to backport [2/4] into their kernel without [4/4] then they'll have a bad day.
As Harry Yoo pointed out:
``` Let's look at an example (assuming unit is 1). val = val * unit = -16384 (-16 KiB) val * unit / PAGE_SIZE = 0xFFFFFFFFFFFFC000 / PAGE_SIZE = 0x3FFFFFFFFFFFFF max(0x3FFFFFFFFFFFFF, 1UL) = 0x3FFFFFFFFFF Yeah, that's a massive positive number. Hmm but how did it work when it was int? val = val * unit = -16384 (-16KiB) val * unit / PAGE_SIZE = 0xFFFFFFFFFFFFC000 / PAGE_SIZE = 0x3FFFFFFFFFFFFF max(val * unit / PAGE_SIZE, 1UL) = 0x3FFFFFFFFFFFFF (int)0x3FFFFFFFFFFFFF = 0xFFFFFFFF = (-1) That's incorrect. It should have been -4? ```
this can already be an issue even without [2/4], and [2/4] just amplify it.
Before LRU folio reparenting was introduced, we wouldn’t pass in such a large value, so this wasn’t a problem. Since LRU folio reparenting is still in mm-unstable, so I didn't add a Fixes tag in [4/4].
Quoted 2 lines
So perhaps this issue should be addressed within [2/4]?
Because they fix different existing problems, I previously split them into two patches.
However, since the issues are all quite minor, I feel that combining them into a single patch is fine.
Quoted 1 line
Quoted 1 line
Actually I can trivially do this locally if you like - just turn [4/4]
i>> nto a -fix patch against [2/4], squash them together later on.
Quoted 2 lines
Quoted 2 lines
Please lmk if you'd like me to do that.
That's fine, thank you!
Andrew Morton
maintainerQuoted 6 lines · On Thu, 26 Mar 2026 10:30:29 +0800 Qi Zheng <[email protected]> wrote:
Quoted 1 line
Quoted 1 line
Actually I can trivially do this locally if you like - just turn [4/4]
i>> nto a -fix patch against [2/4], squash them together later on.
Quoted 2 lines
Quoted 2 lines
Please lmk if you'd like me to do that.
That's fine, thank you!
Done, thanks.
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
Hi all,
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
Is this a real problem without reparenting? In other words is this a follow up for that work or something that we need on its own?
Signature · 2 lines
Michal Hocko SUSE Labs
Quoted 11 lines · Michal Hocko
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
Hi all,
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
Is this a real problem without reparenting? In other words is this a follow up for that work or something that we need on its own?
This is not a problem (as discussed in [1]) without reparenting.
[1] https://lore.kernel.org/linux-mm/[email protected]
Signature · 2 lines
Cheers, Harry / Hyeonggon
Quoted 10 lines · Harry Yoo (Oracle)
Quoted 9 lines · Michal Hocko
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
Hi all,
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
Quoted 5 lines
Quoted 2 lines
Is this a real problem without reparenting? In other words is this a follow up for that work or something that we need on its own?
This is not a problem (as discussed in [1]) without reparenting.
[1] https://lore.kernel.org/linux-mm/[email protected]
Please make it explicit in the changelogs and/or make them part of the series.
Thanks!
Signature · 2 lines
Michal Hocko SUSE Labs
Quoted 11 lines · Michal Hocko
Quoted 10 lines · Harry Yoo (Oracle)
Quoted 9 lines · Michal Hocko
Quoted 7 lines · Qi Zheng
From: Qi Zheng <[email protected]>
Hi all,
As Harry Yoo pointed out [1], in scenarios where massive state updates occur (e.g., during the reparenting of LRU folios), the values passed to memcg stat update functions can accumulate and exceed the upper limit of a 32-bit integer.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Is this a real problem without reparenting? In other words is this a follow up for that work or something that we need on its own?
This is not a problem (as discussed in [1]) without reparenting.
[1] https://lore.kernel.org/linux-mm/[email protected]Please make it explicit in the changelogs and/or make them part of the series.
Agreed!
Quoted 5 lines
Thanks!
--
Michal Hocko
SUSE Labs
Cheers, Lorenzo