perf/x86/amd/uncore: turn amd_uncore_ctx events into a flexible array
Subsystems (2)
PATCH 1/1 perf/x86/amd/uncore: turn amd_uncore_ctx events into a flexible array
The events pointer was kzalloc_node()'d immediately after the parent struct allocation, with the count (pmu->num_counters) trivially available beforehand. Move events to the struct tail as a flexible array member and fold the two allocations into a single kzalloc_node() with struct_size(), dropping the separate kfree(ctx->events) in the free path.
--- arch/x86/events/amd/uncore.c | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-)
arch/x86/events/amd/uncore.c +3 -12
@@ -39,11 +39,11 @@ static int pmu_version;struct amd_uncore_ctx {int refcnt;int cpu; -struct perf_event **events;unsigned long active_mask[BITS_TO_LONGS (NUM_COUNTERS_MAX )];int nr_active;struct hrtimer hrtimer;u64 hrtimer_duration; +struct perf_event *events[]; };struct amd_uncore_pmu {
@@ -488,10 +488,8 @@ static void amd_uncore_ctx_free(struct amd_uncore *uncore, unsigned int cpu)if (cpu == ctx->cpu)cpumask_clear_cpu (cpu, &pmu->active_mask); -if (!--ctx->refcnt) { -kfree (ctx->events); +if (!--ctx->refcnt)kfree (ctx); - } *per_cpu_ptr (pmu->ctx, cpu) =NULL ; }
@@ -536,18 +534,11 @@ static int amd_uncore_ctx_init(struct amd_uncore *uncore, unsigned int cpu)/* Allocate context if sibling does not exist */ if (!curr) { node =cpu_to_node (cpu); - curr =kzalloc_node (sizeof (*curr),GFP_KERNEL , node); + curr =kzalloc_node (struct_size (curr, events, pmu->num_counters),GFP_KERNEL , node);if (!curr) goto fail; curr->cpu = cpu; - curr->events =kzalloc_node (sizeof (*curr->events) * - pmu->num_counters, -GFP_KERNEL , node); -if (!curr->events) { -kfree (curr); - goto fail; - }amd_uncore_init_hrtimer (curr); curr->hrtimer_duration = (u64 )update_interval *NSEC_PER_MSEC ;
Signature · 1 line
2.54.0
Ian Rogers
reviewerQuoted 19 lines · Rosen Penev
The events pointer was kzalloc_node()'d immediately after the parent struct allocation, with the count (
pmu->num_counters) trivially available beforehand. Move events to the struct tail as a flexible array member and fold the two allocations into a singlekzalloc_node()withstruct_size(), dropping the separatekfree(ctx->events)in the free path.Assisted-by Claude:Opus-4.7Signed-off-by Rosen Penev <[email protected]>--- arch/x86/events/amd/uncore.c | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-)arch/x86/events/amd/uncore.c +0
-0
arch/x86/events/amd/uncore.c static int pmu_version; · line 46
Quoted 9 lines
arch/x86/events/amd/uncore.c +1
-1@@ -39,11 +39,11 @@ static int pmu_version;struct amd_uncore_ctx {int refcnt;int cpu; -struct perf_event **events;unsigned long active_mask[BITS_TO_LONGS (NUM_COUNTERS_MAX )];int nr_active;struct hrtimer hrtimer;u64 hrtimer_duration; +struct perf_event *events[];
As the number of counters is fixed and events doesn't grow, this makes sense to me. Dapeng, what do you think?
Thanks,
Ian
Quoted 38 lines
arch/x86/events/amd/uncore.c +2
-12};struct amd_uncore_pmu {@@ -488,10 +488,8 @@ static void amd_uncore_ctx_free(struct amd_uncore *uncore, unsigned int cpu)if (cpu == ctx->cpu)cpumask_clear_cpu (cpu, &pmu->active_mask); -if (!--ctx->refcnt) { -kfree (ctx->events); +if (!--ctx->refcnt)kfree (ctx); - } *per_cpu_ptr (pmu->ctx, cpu) =NULL ; }@@ -536,18 +534,11 @@ static int amd_uncore_ctx_init(struct amd_uncore *uncore, unsigned int cpu)/* Allocate context if sibling does not exist */ if (!curr) { node =cpu_to_node (cpu); - curr =kzalloc_node (sizeof (*curr),GFP_KERNEL , node); + curr =kzalloc_node (struct_size (curr, events, pmu->num_counters),GFP_KERNEL , node);if (!curr) goto fail; curr->cpu = cpu; - curr->events =kzalloc_node (sizeof (*curr->events) * - pmu->num_counters, -GFP_KERNEL , node); -if (!curr->events) { -kfree (curr); - goto fail; - }amd_uncore_init_hrtimer (curr); curr->hrtimer_duration = (u64 )update_interval *NSEC_PER_MSEC ; --2.54.0
Quoted 14 lines · Ian Rogers
Quoted 13 lines · Rosen Penev
The events pointer was kzalloc_node()'d immediately after the parent struct allocation, with the count (
pmu->num_counters) trivially available beforehand. Move events to the struct tail as a flexible array member and fold the two allocations into a singlekzalloc_node()withstruct_size(), dropping the separatekfree(ctx->events)in the free path.Assisted-by Claude:Opus-4.7Signed-off-by Rosen Penev <[email protected]>--- arch/x86/events/amd/uncore.c | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-)
Quoted 16 lines
Quoted 5 lines
arch/x86/events/amd/uncore.c +0
-0arch/x86/events/amd/uncore.c static int pmu_version; · line 46
Quoted 9 lines
arch/x86/events/amd/uncore.c +1
-1@@ -39,11 +39,11 @@ static int pmu_version;struct amd_uncore_ctx {int refcnt;int cpu; -struct perf_event **events;unsigned long active_mask[BITS_TO_LONGS (NUM_COUNTERS_MAX )];int nr_active;struct hrtimer hrtimer;u64 hrtimer_duration; +struct perf_event *events[];As the number of counters is fixed and events doesn't grow, this makes sense to me. Dapeng, what do you think?
It looks good to me. But it seems there is no Sashikio review, so better resend to trigger Sashiko review for double check. Thanks.
Quoted 42 lines
Thanks,
IanQuoted 38 lines
arch/x86/events/amd/uncore.c +2
-12};struct amd_uncore_pmu {@@ -488,10 +488,8 @@ static void amd_uncore_ctx_free(struct amd_uncore *uncore, unsigned int cpu)if (cpu == ctx->cpu)cpumask_clear_cpu (cpu, &pmu->active_mask); -if (!--ctx->refcnt) { -kfree (ctx->events); +if (!--ctx->refcnt)kfree (ctx); - } *per_cpu_ptr (pmu->ctx, cpu) =NULL ; }@@ -536,18 +534,11 @@ static int amd_uncore_ctx_init(struct amd_uncore *uncore, unsigned int cpu)/* Allocate context if sibling does not exist */ if (!curr) { node =cpu_to_node (cpu); - curr =kzalloc_node (sizeof (*curr),GFP_KERNEL , node); + curr =kzalloc_node (struct_size (curr, events, pmu->num_counters),GFP_KERNEL , node);if (!curr) goto fail; curr->cpu = cpu; - curr->events =kzalloc_node (sizeof (*curr->events) * - pmu->num_counters, -GFP_KERNEL , node); -if (!curr->events) { -kfree (curr); - goto fail; - }amd_uncore_init_hrtimer (curr); curr->hrtimer_duration = (u64 )update_interval *NSEC_PER_MSEC ; --2.54.0