[PATCH bpf] bpf: Add LBR data to BPF_PROG_TYPE_PERF_EVENT prog context

[Daniel Xu]

Last-branch-record is an intel CPU feature that can be configured to
record certain branches that are taken during code execution. This data
is particularly interesting for profile guided optimizations. perf has
had LBR support for a while but the data collection can be a bit coarse
grained.

We (Facebook) have recently run a lot of experiments with feeding
filtered LBR data to various PGO pipelines. We've seen really good
results (+2.5% throughput with lower cpu util and lower latency) by
feeding high request latency LBR branches to the compiler on a
request-oriented service. We used bpf to read a special request context
ID (which is how we associate branches with latency) from a fixed
userspace address. Reading from the fixed address is why bpf support is
useful.

Aside from this particular use case, having LBR data available to bpf
progs can be useful to get stack traces out of userspace applications
that omit frame pointers.

This patch adds support for LBR data to bpf perf progs.

Some notes:
* We use `__u64 entries[BPF_MAX_LBR_ENTRIES * 3]` instead of
  `struct perf_branch_entry[BPF_MAX_LBR_ENTRIES]` because checkpatch.pl
  warns about including a uapi header from another uapi header

* We define BPF_MAX_LBR_ENTRIES as 32 (instead of using the value from
  arch/x86/events/perf_events.h) because including arch specific headers
  seems wrong and could introduce circular header includes.

Signed-off-by: Daniel Xu <dxu@xxxxxxxxx>
---
 include/uapi/linux/bpf_perf_event.h |  5 ++++
 kernel/trace/bpf_trace.c            | 39 +++++++++++++++++++++++++++++
 2 files changed, 44 insertions(+)

diff --git a/include/uapi/linux/bpf_perf_event.h b/include/uapi/linux/bpf_perf_event.h
index eb1b9d21250c..dc87e3d50390 100644
--- a/include/uapi/linux/bpf_perf_event.h
+++ b/include/uapi/linux/bpf_perf_event.h
@@ -10,10 +10,15 @@
 
 #include <asm/bpf_perf_event.h>
 
+#define BPF_MAX_LBR_ENTRIES 32
+
 struct bpf_perf_event_data {
 	bpf_user_pt_regs_t regs;
 	__u64 sample_period;
 	__u64 addr;
+	__u64 nr_lbr;
+	/* Cast to struct perf_branch_entry* before using */
+	__u64 entries[BPF_MAX_LBR_ENTRIES * 3];
 };
 
 #endif /* _UAPI__LINUX_BPF_PERF_EVENT_H__ */
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index ffc91d4935ac..96ba7995b3d7 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -1259,6 +1259,14 @@ static bool pe_prog_is_valid_access(int off, int size, enum bpf_access_type type
 		if (!bpf_ctx_narrow_access_ok(off, size, size_u64))
 			return false;
 		break;
+	case bpf_ctx_range(struct bpf_perf_event_data, nr_lbr):
+		bpf_ctx_record_field_size(info, size_u64);
+		if (!bpf_ctx_narrow_access_ok(off, size, size_u64))
+			return false;
+		break;
+	case bpf_ctx_range(struct bpf_perf_event_data, entries):
+		/* No narrow loads */
+		break;
 	default:
 		if (size != sizeof(long))
 			return false;
@@ -1273,6 +1281,7 @@ static u32 pe_prog_convert_ctx_access(enum bpf_access_type type,
 				      struct bpf_prog *prog, u32 *target_size)
 {
 	struct bpf_insn *insn = insn_buf;
+	int off;
 
 	switch (si->off) {
 	case offsetof(struct bpf_perf_event_data, sample_period):
@@ -1291,6 +1300,36 @@ static u32 pe_prog_convert_ctx_access(enum bpf_access_type type,
 				      bpf_target_off(struct perf_sample_data, addr, 8,
 						     target_size));
 		break;
+	case offsetof(struct bpf_perf_event_data, nr_lbr):
+		/* Load struct perf_sample_data* */
+		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct bpf_perf_event_data_kern,
+						       data), si->dst_reg, si->src_reg,
+				      offsetof(struct bpf_perf_event_data_kern, data));
+		/* Load struct perf_branch_stack* */
+		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct perf_sample_data, br_stack),
+				      si->dst_reg, si->dst_reg,
+				      offsetof(struct perf_sample_data, br_stack));
+		/* Load nr */
+		*insn++ = BPF_LDX_MEM(BPF_DW, si->dst_reg, si->dst_reg,
+				      bpf_target_off(struct perf_branch_stack, nr, 8,
+						     target_size));
+		break;
+	case bpf_ctx_range(struct bpf_perf_event_data, entries):
+		off = si->off;
+		off -= offsetof(struct bpf_perf_event_data, entries);
+
+		/* Load struct perf_sample_data* */
+		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct bpf_perf_event_data_kern,
+						       data), si->dst_reg, si->src_reg,
+				      offsetof(struct bpf_perf_event_data_kern, data));
+		/* Load struct perf_branch_stack* */
+		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct perf_sample_data, br_stack),
+				      si->dst_reg, si->dst_reg,
+				      offsetof(struct perf_sample_data, br_stack));
+		/* Load requested memory */
+		*insn++ = BPF_LDX_MEM(BPF_SIZE(si->code), si->dst_reg, si->dst_reg,
+				      offsetof(struct perf_branch_stack, entries) + off);
+		break;
 	default:
 		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct bpf_perf_event_data_kern,
 						       regs), si->dst_reg, si->src_reg,
-- 
2.24.0