From: Puranjay Mohan <puranjay@kernel.org>
To: bpf@vger.kernel.org
Cc: Puranjay Mohan <puranjay@kernel.org>,
Puranjay Mohan <puranjay12@gmail.com>,
Alexei Starovoitov <ast@kernel.org>,
Andrii Nakryiko <andrii@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Martin KaFai Lau <martin.lau@kernel.org>,
Eduard Zingerman <eddyz87@gmail.com>,
Kumar Kartikeya Dwivedi <memxor@gmail.com>,
Will Deacon <will@kernel.org>,
Mark Rutland <mark.rutland@arm.com>,
Catalin Marinas <catalin.marinas@arm.com>,
Rob Herring <robh@kernel.org>, Breno Leitao <leitao@debian.org>,
linux-arm-kernel@lists.infradead.org,
linux-perf-users@vger.kernel.org, kernel-team@meta.com
Subject: [RFC PATCH] perf/arm64: Add BRBE support for bpf_get_branch_snapshot()
Date: Fri, 2 Jan 2026 13:40:41 -0800 [thread overview]
Message-ID: <20260102214043.1410242-1-puranjay@kernel.org> (raw)
Enable the bpf_get_branch_snapshot() BPF helper on ARM64 by implementing
the perf_snapshot_branch_stack static call for ARM's Branch Record
Buffer Extension (BRBE).
The BPF helper bpf_get_branch_snapshot() allows BPF programs to capture
hardware branch records on-demand. This was previously only available on
x86 (Intel LBR, AMD BRS) but not on ARM64 despite BRBE being available
since ARMv9.
This implementation:
- Follows the x86 snapshot pattern (intel_pmu_snapshot_branch_stack)
- Performs atomic snapshot by pausing BRBE, reading records, and
restoring previous state without disrupting ongoing perf events
- Reads branch records directly from BRBE registers without
event-specific filtering to minimize branch pollution
- Handles all BRBE record types (complete, source-only, target-only)
- Complies with ARM ARM synchronization requirements (ISB barriers per
rule PPBZP)
- Reuses existing BRBE infrastructure (select_brbe_bank,
__read_brbe_regset, brbe_set_perf_entry_type, etc.)
Signed-off-by: Puranjay Mohan <puranjay@kernel.org>
---
This patch is only compile tested as I don't have access to hardware with BRBE.
---
drivers/perf/arm_brbe.c | 95 ++++++++++++++++++++++++++++++++++++++++
drivers/perf/arm_brbe.h | 9 ++++
drivers/perf/arm_pmuv3.c | 5 ++-
3 files changed, 108 insertions(+), 1 deletion(-)
diff --git a/drivers/perf/arm_brbe.c b/drivers/perf/arm_brbe.c
index ba554e0c846c..cda7bf522c06 100644
--- a/drivers/perf/arm_brbe.c
+++ b/drivers/perf/arm_brbe.c
@@ -803,3 +803,98 @@ void brbe_read_filtered_entries(struct perf_branch_stack *branch_stack,
done:
branch_stack->nr = nr_filtered;
}
+
+/*
+ * ARM-specific callback invoked through perf_snapshot_branch_stack static
+ * call, defined in include/linux/perf_event.h. See its definition for API
+ * details. It's up to caller to provide enough space in *entries* to fit all
+ * branch records, otherwise returned result will be truncated to *cnt* entries.
+ *
+ * This is similar to brbe_read_filtered_entries but optimized for snapshot mode:
+ * - No filtering based on event attributes (captures everything)
+ * - Minimal branches to avoid polluting the branch buffer
+ * - Direct register reads without event-specific processing
+ */
+int arm_brbe_snapshot_branch_stack(struct perf_branch_entry *entries, unsigned int cnt)
+{
+ unsigned long flags;
+ int nr_hw, nr_banks, nr_copied = 0;
+ u64 brbidr, brbfcr, brbcr;
+
+ /*
+ * The sequence of steps to freeze BRBE should be completely inlined
+ * and contain no branches to minimize contamination of branch snapshot.
+ */
+ local_irq_save(flags);
+
+ /* Save current BRBE configuration */
+ brbfcr = read_sysreg_s(SYS_BRBFCR_EL1);
+ brbcr = read_sysreg_s(SYS_BRBCR_EL1);
+
+ /* Pause BRBE to freeze the buffer */
+ write_sysreg_s(brbfcr | BRBFCR_EL1_PAUSED, SYS_BRBFCR_EL1);
+ isb();
+
+ /* Read BRBIDR to determine number of records */
+ brbidr = read_sysreg_s(SYS_BRBIDR0_EL1);
+ if (!valid_brbidr(brbidr))
+ goto out_restore;
+
+ nr_hw = FIELD_GET(BRBIDR0_EL1_NUMREC_MASK, brbidr);
+ nr_banks = DIV_ROUND_UP(nr_hw, BRBE_BANK_MAX_ENTRIES);
+
+ /* Read branch records from BRBE banks */
+ for (int bank = 0; bank < nr_banks; bank++) {
+ int nr_remaining = nr_hw - (bank * BRBE_BANK_MAX_ENTRIES);
+ int nr_this_bank = min(nr_remaining, BRBE_BANK_MAX_ENTRIES);
+
+ select_brbe_bank(bank);
+
+ for (int i = 0; i < nr_this_bank; i++) {
+ struct brbe_regset bregs;
+ struct perf_branch_entry *entry;
+
+ if (nr_copied >= cnt)
+ goto out_restore;
+
+ if (!__read_brbe_regset(&bregs, i))
+ goto out_restore;
+
+ entry = &entries[nr_copied];
+ perf_clear_branch_entry_bitfields(entry);
+
+ /* Simple conversion without filtering */
+ if (brbe_record_is_complete(bregs.brbinf)) {
+ entry->from = bregs.brbsrc;
+ entry->to = bregs.brbtgt;
+ } else if (brbe_record_is_source_only(bregs.brbinf)) {
+ entry->from = bregs.brbsrc;
+ entry->to = 0;
+ } else if (brbe_record_is_target_only(bregs.brbinf)) {
+ entry->from = 0;
+ entry->to = bregs.brbtgt;
+ }
+
+ brbe_set_perf_entry_type(entry, bregs.brbinf);
+ entry->cycles = brbinf_get_cycles(bregs.brbinf);
+
+ if (!brbe_record_is_target_only(bregs.brbinf)) {
+ entry->mispred = brbinf_get_mispredict(bregs.brbinf);
+ entry->predicted = !entry->mispred;
+ }
+
+ if (!brbe_record_is_source_only(bregs.brbinf))
+ entry->priv = brbinf_get_perf_priv(bregs.brbinf);
+
+ nr_copied++;
+ }
+ }
+
+out_restore:
+ /* Restore BRBE to its previous state */
+ write_sysreg_s(brbcr, SYS_BRBCR_EL1);
+ isb();
+ write_sysreg_s(brbfcr, SYS_BRBFCR_EL1);
+ local_irq_restore(flags);
+ return nr_copied;
+}
diff --git a/drivers/perf/arm_brbe.h b/drivers/perf/arm_brbe.h
index b7c7d8796c86..c2a1824437fb 100644
--- a/drivers/perf/arm_brbe.h
+++ b/drivers/perf/arm_brbe.h
@@ -10,6 +10,7 @@
struct arm_pmu;
struct perf_branch_stack;
struct perf_event;
+struct perf_branch_entry;
#ifdef CONFIG_ARM64_BRBE
void brbe_probe(struct arm_pmu *arm_pmu);
@@ -22,6 +23,8 @@ void brbe_disable(void);
bool brbe_branch_attr_valid(struct perf_event *event);
void brbe_read_filtered_entries(struct perf_branch_stack *branch_stack,
const struct perf_event *event);
+int arm_brbe_snapshot_branch_stack(struct perf_branch_entry *entries,
+ unsigned int cnt);
#else
static inline void brbe_probe(struct arm_pmu *arm_pmu) { }
static inline unsigned int brbe_num_branch_records(const struct arm_pmu *armpmu)
@@ -44,4 +47,10 @@ static void brbe_read_filtered_entries(struct perf_branch_stack *branch_stack,
const struct perf_event *event)
{
}
+
+static inline int arm_brbe_snapshot_branch_stack(struct perf_branch_entry *entries,
+ unsigned int cnt)
+{
+ return 0;
+}
#endif
diff --git a/drivers/perf/arm_pmuv3.c b/drivers/perf/arm_pmuv3.c
index 8014ff766cff..1a9f129a0f94 100644
--- a/drivers/perf/arm_pmuv3.c
+++ b/drivers/perf/arm_pmuv3.c
@@ -1449,8 +1449,11 @@ static int armv8_pmu_init(struct arm_pmu *cpu_pmu, char *name,
cpu_pmu->set_event_filter = armv8pmu_set_event_filter;
cpu_pmu->pmu.event_idx = armv8pmu_user_event_idx;
- if (brbe_num_branch_records(cpu_pmu))
+ if (brbe_num_branch_records(cpu_pmu)) {
cpu_pmu->pmu.sched_task = armv8pmu_sched_task;
+ static_call_update(perf_snapshot_branch_stack,
+ arm_brbe_snapshot_branch_stack);
+ }
cpu_pmu->name = name;
cpu_pmu->map_event = map_event;
base-commit: c286e7e9d1f1f3d90ad11c37e896f582b02d19c4
--
2.47.3
next reply other threads:[~2026-01-02 21:41 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-01-02 21:40 Puranjay Mohan [this message]
2026-01-02 21:51 ` [RFC PATCH] perf/arm64: Add BRBE support for bpf_get_branch_snapshot() Puranjay Mohan
2026-01-05 21:22 ` Leo Yan
2026-01-05 21:46 ` Puranjay Mohan
2026-01-06 10:54 ` Leo Yan
2026-01-05 17:49 ` Mark Rutland
2026-01-05 19:31 ` Puranjay Mohan
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260102214043.1410242-1-puranjay@kernel.org \
--to=puranjay@kernel.org \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=catalin.marinas@arm.com \
--cc=daniel@iogearbox.net \
--cc=eddyz87@gmail.com \
--cc=kernel-team@meta.com \
--cc=leitao@debian.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=martin.lau@kernel.org \
--cc=memxor@gmail.com \
--cc=puranjay12@gmail.com \
--cc=robh@kernel.org \
--cc=will@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox