Skip to contentSkip to conversation

perf lock contention: Support pre-5.14 kernels

PATCH In review Ian Rogers 3 participants 5 messages Apr 8, 2023 – Apr 10, 2023

Subsystems (2)
lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

PATCH 1/2 perf lock contention: Support pre-5.14 kernels

struct rq's variable __lock was renamed from lock in 5.14.

---
 tools/perf/util/bpf_skel/lock_contention.bpf.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)
tools/perf/util/bpf_skel/lock_contention.bpf.c +14 -1
index 23f6e63544ed..8911e2a077d8 100644
@@ -418,6 +418,14 @@ int contention_end(u64 *ctx)
 
 extern struct rq runqueues __ksym;
 
+struct rq__old {
+	raw_spinlock_t lock;
+} __attribute__((preserve_access_index));
+
+struct rq__new {
+	raw_spinlock_t __lock;
+} __attribute__((preserve_access_index));
+
 SEC("raw_tp/bpf_test_finish")
 int BPF_PROG(collect_lock_syms)
 {
@@ -426,11 +434,16 @@ int BPF_PROG(collect_lock_syms)
 
 	for (int i = 0; i < MAX_CPUS; i++) {
 		struct rq *rq = bpf_per_cpu_ptr(&runqueues, i);
+		struct rq__new *rq_new = (void *)rq;
+		struct rq__old *rq_old = (void *)rq;
 
 		if (rq == NULL)
 			break;
 
-		lock_addr = (__u64)&rq->__lock;
+		if (bpf_core_field_exists(rq_new->__lock))
+			lock_addr = (__u64)&rq_new->__lock;
+		else
+			lock_addr = (__u64)&rq_old->lock;
 		lock_flag = LOCK_CLASS_RQLOCK;
 		bpf_map_update_elem(&lock_syms, &lock_addr, &lock_flag, BPF_ANY);
 	}
Signature · 1 line
2.40.0.577.gac1e443424-goog
author reviewer
#1 lore raw

PATCH 2/2 perf bpf filter: Support pre-5.16 kernels

The mem_hops bits were added in 5.16 with no prior equivalent.

---
 tools/perf/util/bpf_skel/sample_filter.bpf.c | 28 ++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)
tools/perf/util/bpf_skel/sample_filter.bpf.c +26 -2
index 57e3c67d6d37..cffe493af1ed 100644
@@ -24,6 +24,24 @@ struct perf_sample_data___new {
 	__u64 sample_flags;
 } __attribute__((preserve_access_index));
 
+/* new kernel perf_mem_data_src definition */
+union perf_mem_data_src__new {
+	__u64 val;
+	struct {
+		__u64   mem_op:5,	/* type of opcode */
+			mem_lvl:14,	/* memory hierarchy level */
+			mem_snoop:5,	/* snoop mode */
+			mem_lock:2,	/* lock instr */
+			mem_dtlb:7,	/* tlb access */
+			mem_lvl_num:4,	/* memory hierarchy level number */
+			mem_remote:1,   /* remote */
+			mem_snoopx:2,	/* snoop mode, ext */
+			mem_blk:3,	/* access blocked */
+			mem_hops:3,	/* hop level */
+			mem_rsvd:18;
+	};
+};
+
 /* helper function to return the given perf sample data */
 static inline __u64 perf_get_sample(struct bpf_perf_event_data_kern *kctx,
 				    struct perf_bpf_filter_entry *entry)
@@ -89,8 +107,14 @@ static inline __u64 perf_get_sample(struct bpf_perf_event_data_kern *kctx,
 			return kctx->data->data_src.mem_dtlb;
 		if (entry->part == 7)
 			return kctx->data->data_src.mem_blk;
-		if (entry->part == 8)
-			return kctx->data->data_src.mem_hops;
+		if (entry->part == 8) {
+			union perf_mem_data_src__new *data = (void *)&kctx->data->data_src;
+
+			if (bpf_core_field_exists(data->mem_hops))
+				return data->mem_hops;
+
+			return 0;
+		}
 		/* return the whole word */
 		return kctx->data->data_src.val;
 	default:
Signature · 1 line
2.40.0.577.gac1e443424-goog
maintainer Acked-by on v1 2/2
#2 lore raw

Re: [PATCH v1 2/2] perf bpf filter: Support pre-5.16 kernels

Quoted 4 lines · Ian Rogers

The mem_hops bits were added in 5.16 with no prior equivalent.

Thanks,
Namhyung

Quoted 53 lines
---
 tools/perf/util/bpf_skel/sample_filter.bpf.c | 28 ++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)
tools/perf/util/bpf_skel/sample_filter.bpf.c +26 -3
index 57e3c67d6d37..cffe493af1ed 100644
@@ -24,6 +24,24 @@ struct perf_sample_data___new {
        __u64 sample_flags;
 } __attribute__((preserve_access_index));
 
+/* new kernel perf_mem_data_src definition */
+union perf_mem_data_src__new {
+       __u64 val;
+       struct {
+               __u64   mem_op:5,       /* type of opcode */
+                       mem_lvl:14,     /* memory hierarchy level */
+                       mem_snoop:5,    /* snoop mode */
+                       mem_lock:2,     /* lock instr */
+                       mem_dtlb:7,     /* tlb access */
+                       mem_lvl_num:4,  /* memory hierarchy level number */
+                       mem_remote:1,   /* remote */
+                       mem_snoopx:2,   /* snoop mode, ext */
+                       mem_blk:3,      /* access blocked */
+                       mem_hops:3,     /* hop level */
+                       mem_rsvd:18;
+       };
+};
+
 /* helper function to return the given perf sample data */
 static inline __u64 perf_get_sample(struct bpf_perf_event_data_kern *kctx,
                                    struct perf_bpf_filter_entry *entry)
@@ -89,8 +107,14 @@ static inline __u64 perf_get_sample(struct bpf_perf_event_data_kern *kctx,
                        return kctx->data->data_src.mem_dtlb;
                if (entry->part == 7)
                        return kctx->data->data_src.mem_blk;
-               if (entry->part == 8)
-                       return kctx->data->data_src.mem_hops;
+               if (entry->part == 8) {
+                       union perf_mem_data_src__new *data = (void *)&kctx->data->data_src;
+
+                       if (bpf_core_field_exists(data->mem_hops))
+                               return data->mem_hops;
+
+                       return 0;
+               }
                /* return the whole word */
                return kctx->data->data_src.val;
        default:
--

2.40.0.577.gac1e443424-goog

#3 lore raw

Re: [PATCH v1 2/2] perf bpf filter: Support pre-5.16 kernels

Em Mon, Apr 10, 2023 at 08:34:34AM -0700, Namhyung Kim escreveu:

Quoted 7 lines · Ian Rogers
Quoted 4 lines

The mem_hops bits were added in 5.16 with no prior equivalent.

Both applied, thanks,

- Arnaldo

Quoted 57 lines

Thanks,
Namhyung

Quoted 53 lines
---
 tools/perf/util/bpf_skel/sample_filter.bpf.c | 28 ++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)
tools/perf/util/bpf_skel/sample_filter.bpf.c +26 -3
index 57e3c67d6d37..cffe493af1ed 100644
@@ -24,6 +24,24 @@ struct perf_sample_data___new {
        __u64 sample_flags;
 } __attribute__((preserve_access_index));
 
+/* new kernel perf_mem_data_src definition */
+union perf_mem_data_src__new {
+       __u64 val;
+       struct {
+               __u64   mem_op:5,       /* type of opcode */
+                       mem_lvl:14,     /* memory hierarchy level */
+                       mem_snoop:5,    /* snoop mode */
+                       mem_lock:2,     /* lock instr */
+                       mem_dtlb:7,     /* tlb access */
+                       mem_lvl_num:4,  /* memory hierarchy level number */
+                       mem_remote:1,   /* remote */
+                       mem_snoopx:2,   /* snoop mode, ext */
+                       mem_blk:3,      /* access blocked */
+                       mem_hops:3,     /* hop level */
+                       mem_rsvd:18;
+       };
+};
+
 /* helper function to return the given perf sample data */
 static inline __u64 perf_get_sample(struct bpf_perf_event_data_kern *kctx,
                                    struct perf_bpf_filter_entry *entry)
@@ -89,8 +107,14 @@ static inline __u64 perf_get_sample(struct bpf_perf_event_data_kern *kctx,
                        return kctx->data->data_src.mem_dtlb;
                if (entry->part == 7)
                        return kctx->data->data_src.mem_blk;
-               if (entry->part == 8)
-                       return kctx->data->data_src.mem_hops;
+               if (entry->part == 8) {
+                       union perf_mem_data_src__new *data = (void *)&kctx->data->data_src;
+
+                       if (bpf_core_field_exists(data->mem_hops))
+                               return data->mem_hops;
+
+                       return 0;
+               }
                /* return the whole word */
                return kctx->data->data_src.val;
        default:
--

2.40.0.577.gac1e443424-goog

Signature · 1 line
- Arnaldo
maintainer Acked-by on v1 1/2
#4 lore raw
In reply to Ian Rogers

Hi Ian,

Quoted 4 lines · Ian Rogers

struct rq's variable __lock was renamed from lock in 5.14.

Thanks for fixing this.

Thanks,
Namhyung

Quoted 44 lines
---
 tools/perf/util/bpf_skel/lock_contention.bpf.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)
tools/perf/util/bpf_skel/lock_contention.bpf.c +14 -2
index 23f6e63544ed..8911e2a077d8 100644
@@ -418,6 +418,14 @@ int contention_end(u64 *ctx)
 
 extern struct rq runqueues __ksym;
 
+struct rq__old {
+       raw_spinlock_t lock;
+} __attribute__((preserve_access_index));
+
+struct rq__new {
+       raw_spinlock_t __lock;
+} __attribute__((preserve_access_index));
+
 SEC("raw_tp/bpf_test_finish")
 int BPF_PROG(collect_lock_syms)
 {
@@ -426,11 +434,16 @@ int BPF_PROG(collect_lock_syms)
 
        for (int i = 0; i < MAX_CPUS; i++) {
                struct rq *rq = bpf_per_cpu_ptr(&runqueues, i);
+               struct rq__new *rq_new = (void *)rq;
+               struct rq__old *rq_old = (void *)rq;
 
                if (rq == NULL)
                        break;
 
-               lock_addr = (__u64)&rq->__lock;
+               if (bpf_core_field_exists(rq_new->__lock))
+                       lock_addr = (__u64)&rq_new->__lock;
+               else
+                       lock_addr = (__u64)&rq_old->lock;
                lock_flag = LOCK_CLASS_RQLOCK;
                bpf_map_update_elem(&lock_syms, &lock_addr, &lock_flag, BPF_ANY);
        }
--

2.40.0.577.gac1e443424-goog