perf sample: Make user_regs and intr_regs optional

The struct dump_regs contains 512 bytes of cache_regs, meaning the two
values in perf_sample contribute 1088 bytes of its total 1384 bytes
size. Initializing this much memory has a cost reported by Tavian
Barnes <tavianator@tavianator.com> as about 2.5% when running `perf
script --itrace=i0`:
https://lore.kernel.org/lkml/d841b97b3ad2ca8bcab07e4293375fb7c32dfce7.1736618095.git.tavianator@tavianator.com/

Adrian Hunter <adrian.hunter@intel.com> replied that the zero
initialization was necessary and couldn't simply be removed.

This patch aims to strike a middle ground of still zeroing the
perf_sample, but removing 79% of its size by make user_regs and
intr_regs optional pointers to zalloc-ed memory. To support the
allocation accessors are created for user_regs and intr_regs. To
support correct cleanup perf_sample__init and perf_sample__exit
functions are created and added throughout the code base.

Signed-off-by: Ian Rogers <irogers@google.com>
Link: https://lore.kernel.org/r/20250113194345.1537821-1-irogers@google.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>

1 files changed, 36 insertions, 26 deletions

diff --git a/tools/perf/tests/sample-parsing.c b/tools/perf/tests/sample-parsing.c
index 25a3f6cece50..72411580f869 100644
--- a/tools/perf/tests/sample-parsing.c
+++ b/tools/perf/tests/sample-parsing.c
@@ -40,8 +40,8 @@
 #define BS_EXPECTED_LE	0x1aa00000000
 #define FLAG(s)	s->branch_stack->entries[i].flags
 
-static bool samples_same(const struct perf_sample *s1,
-			 const struct perf_sample *s2,
+static bool samples_same(struct perf_sample *s1,
+			 struct perf_sample *s2,
 			 u64 type, u64 read_format, bool needs_swap)
 {
 	size_t i;
@@ -126,13 +126,15 @@ static bool samples_same(const struct perf_sample *s1,
 	}
 
 	if (type & PERF_SAMPLE_REGS_USER) {
-		size_t sz = hweight_long(s1->user_regs.mask) * sizeof(u64);
-
-		COMP(user_regs.mask);
-		COMP(user_regs.abi);
-		if (s1->user_regs.abi &&
-		    (!s1->user_regs.regs || !s2->user_regs.regs ||
-		     memcmp(s1->user_regs.regs, s2->user_regs.regs, sz))) {
+		struct regs_dump *s1_regs = perf_sample__user_regs(s1);
+		struct regs_dump *s2_regs = perf_sample__user_regs(s2);
+		size_t sz = hweight_long(s1_regs->mask) * sizeof(u64);
+
+		COMP(user_regs->mask);
+		COMP(user_regs->abi);
+		if (s1_regs->abi &&
+		    (!s1_regs->regs || !s2_regs->regs ||
+		     memcmp(s1_regs->regs, s2_regs->regs, sz))) {
 			pr_debug("Samples differ at 'user_regs'\n");
 			return false;
 		}
@@ -157,13 +159,15 @@ static bool samples_same(const struct perf_sample *s1,
 		COMP(transaction);
 
 	if (type & PERF_SAMPLE_REGS_INTR) {
-		size_t sz = hweight_long(s1->intr_regs.mask) * sizeof(u64);
-
-		COMP(intr_regs.mask);
-		COMP(intr_regs.abi);
-		if (s1->intr_regs.abi &&
-		    (!s1->intr_regs.regs || !s2->intr_regs.regs ||
-		     memcmp(s1->intr_regs.regs, s2->intr_regs.regs, sz))) {
+		struct regs_dump *s1_regs = perf_sample__intr_regs(s1);
+		struct regs_dump *s2_regs = perf_sample__intr_regs(s2);
+		size_t sz = hweight_long(s1_regs->mask) * sizeof(u64);
+
+		COMP(intr_regs->mask);
+		COMP(intr_regs->abi);
+		if (s1_regs->abi &&
+		    (!s1_regs->regs || !s2_regs->regs ||
+		     memcmp(s1_regs->regs, s2_regs->regs, sz))) {
 			pr_debug("Samples differ at 'intr_regs'\n");
 			return false;
 		}
@@ -223,6 +227,16 @@ static int do_test(u64 sample_type, u64 sample_regs, u64 read_format)
 	const u32 raw_data[] = {0x12345678, 0x0a0b0c0d, 0x11020304, 0x05060708, 0 };
 	const u64 data[] = {0x2211443366558877ULL, 0, 0xaabbccddeeff4321ULL};
 	const u64 aux_data[] = {0xa55a, 0, 0xeeddee, 0x0282028202820282};
+	struct regs_dump user_regs = {
+		.abi	= PERF_SAMPLE_REGS_ABI_64,
+		.mask	= sample_regs,
+		.regs	= regs,
+	};
+	struct regs_dump intr_regs = {
+		.abi	= PERF_SAMPLE_REGS_ABI_64,
+		.mask	= sample_regs,
+		.regs	= regs,
+	};
 	struct perf_sample sample = {
 		.ip		= 101,
 		.pid		= 102,
@@ -241,11 +255,7 @@ static int do_test(u64 sample_type, u64 sample_regs, u64 read_format)
 		.callchain	= &callchain.callchain,
 		.no_hw_idx      = false,
 		.branch_stack	= &branch_stack.branch_stack,
-		.user_regs	= {
-			.abi	= PERF_SAMPLE_REGS_ABI_64,
-			.mask	= sample_regs,
-			.regs	= regs,
-		},
+		.user_regs	= &user_regs,
 		.user_stack	= {
 			.size	= sizeof(data),
 			.data	= (void *)data,
@@ -254,11 +264,7 @@ static int do_test(u64 sample_type, u64 sample_regs, u64 read_format)
 			.time_enabled = 0x030a59d664fca7deULL,
 			.time_running = 0x011b6ae553eb98edULL,
 		},
-		.intr_regs	= {
-			.abi	= PERF_SAMPLE_REGS_ABI_64,
-			.mask	= sample_regs,
-			.regs	= regs,
-		},
+		.intr_regs	= &intr_regs,
 		.phys_addr	= 113,
 		.cgroup		= 114,
 		.data_page_size = 115,
@@ -273,6 +279,8 @@ static int do_test(u64 sample_type, u64 sample_regs, u64 read_format)
 	size_t i, sz, bufsz;
 	int err, ret = -1;
 
+	perf_sample__init(&sample_out, /*all=*/false);
+	perf_sample__init(&sample_out_endian, /*all=*/false);
 	if (sample_type & PERF_SAMPLE_REGS_USER)
 		evsel.core.attr.sample_regs_user = sample_regs;
 
@@ -361,6 +369,8 @@ static int do_test(u64 sample_type, u64 sample_regs, u64 read_format)
 	ret = 0;
 out_free:
 	free(event);
+	perf_sample__exit(&sample_out_endian);
+	perf_sample__exit(&sample_out);
 	if (ret && read_format)
 		pr_debug("read_format %#"PRIx64"\n", read_format);
 	return ret;