[PATCH V5 02/25] perf evlist: Add initial support for mmapping an Instruction Trace buffer

From: Adrian Hunter
Date: Sat Mar 07 2015 - 16:15:26 EST


This patch anticipates the addition to the kernel
of an "aux" buffer that can be mmapped separately
from the perf-events buffer.

The expectation is that this buffer can be configured
to contain hardware-produced trace information for
Instruction Tracing, hence the name "itrace".
The first implementation will support Intel BTS and
Intel PT.

One itrace buffer is mmapped per perf-events buffer.
If the requested itrace buffer size is zero, which
it will be until further support is added, then
no itrace mmapping is attempted.

Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
---
tools/perf/util/Build | 1 +
tools/perf/util/evlist.c | 60 ++++++++++++++++++++++++++--
tools/perf/util/evlist.h | 5 +++
tools/perf/util/itrace.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++
tools/perf/util/itrace.h | 96 ++++++++++++++++++++++++++++++++++++++++++++
5 files changed, 260 insertions(+), 3 deletions(-)
create mode 100644 tools/perf/util/itrace.c
create mode 100644 tools/perf/util/itrace.h

diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 972a6e0..b1b7be8 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -74,6 +74,7 @@ libperf-y += data.o
libperf-$(CONFIG_X86) += tsc.o
libperf-y += cloexec.o
libperf-y += thread-stack.o
+libperf-y += itrace.o

libperf-$(CONFIG_LIBELF) += symbol-elf.o
libperf-$(CONFIG_LIBELF) += probe-event.o
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 8d0b623..62be7ef 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -725,6 +725,34 @@ void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
perf_evlist__mmap_put(evlist, idx);
}

+int __weak itrace_mmap__mmap(struct itrace_mmap *mm __maybe_unused,
+ struct itrace_mmap_params *mp __maybe_unused,
+ void *userpg __maybe_unused,
+ int fd __maybe_unused)
+{
+ return 0;
+}
+
+void __weak itrace_mmap__munmap(struct itrace_mmap *mm __maybe_unused)
+{
+}
+
+void __weak itrace_mmap_params__init(
+ struct itrace_mmap_params *mp __maybe_unused,
+ off_t itrace_offset __maybe_unused,
+ unsigned int itrace_pages __maybe_unused,
+ bool itrace_overwrite __maybe_unused)
+{
+}
+
+void __weak itrace_mmap_params__set_idx(
+ struct itrace_mmap_params *mp __maybe_unused,
+ struct perf_evlist *evlist __maybe_unused,
+ int idx __maybe_unused,
+ bool per_cpu __maybe_unused)
+{
+}
+
static void __perf_evlist__munmap(struct perf_evlist *evlist, int idx)
{
if (evlist->mmap[idx].base != NULL) {
@@ -732,6 +760,7 @@ static void __perf_evlist__munmap(struct perf_evlist *evlist, int idx)
evlist->mmap[idx].base = NULL;
evlist->mmap[idx].refcnt = 0;
}
+ itrace_mmap__munmap(&evlist->mmap[idx].itrace_mmap);
}

void perf_evlist__munmap(struct perf_evlist *evlist)
@@ -759,6 +788,7 @@ static int perf_evlist__alloc_mmap(struct perf_evlist *evlist)
struct mmap_params {
int prot;
int mask;
+ struct itrace_mmap_params itrace_mp;
};

static int __perf_evlist__mmap(struct perf_evlist *evlist, int idx,
@@ -789,6 +819,10 @@ static int __perf_evlist__mmap(struct perf_evlist *evlist, int idx,
return -1;
}

+ if (itrace_mmap__mmap(&evlist->mmap[idx].itrace_mmap,
+ &mp->itrace_mp, evlist->mmap[idx].base, fd))
+ return -1;
+
return 0;
}

@@ -853,6 +887,8 @@ static int perf_evlist__mmap_per_cpu(struct perf_evlist *evlist,
for (cpu = 0; cpu < nr_cpus; cpu++) {
int output = -1;

+ itrace_mmap_params__set_idx(&mp->itrace_mp, evlist, cpu, true);
+
for (thread = 0; thread < nr_threads; thread++) {
if (perf_evlist__mmap_per_evsel(evlist, cpu, mp, cpu,
thread, &output))
@@ -878,6 +914,9 @@ static int perf_evlist__mmap_per_thread(struct perf_evlist *evlist,
for (thread = 0; thread < nr_threads; thread++) {
int output = -1;

+ itrace_mmap_params__set_idx(&mp->itrace_mp, evlist, thread,
+ false);
+
if (perf_evlist__mmap_per_evsel(evlist, thread, mp, 0, thread,
&output))
goto out_unmap;
@@ -981,19 +1020,25 @@ int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str,
}

/**
- * perf_evlist__mmap - Create mmaps to receive events.
+ * perf_evlist__mmap_ex - Create mmaps to receive events.
* @evlist: list of events
* @pages: map length in pages
* @overwrite: overwrite older events?
+ * @itrace_pages - itrace map length in pages
+ * @itrace_overwrite - overwrite older itrace data?
*
* If @overwrite is %false the user needs to signal event consumption using
* perf_mmap__write_tail(). Using perf_evlist__mmap_read() does this
* automatically.
*
+ * Similarly, if @itrace_overwrite is %false the user needs to signal data
+ * consumption using itrace_mmap__write_tail().
+ *
* Return: %0 on success, negative error code otherwise.
*/
-int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
- bool overwrite)
+int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
+ bool overwrite, unsigned int itrace_pages,
+ bool itrace_overwrite)
{
struct perf_evsel *evsel;
const struct cpu_map *cpus = evlist->cpus;
@@ -1013,6 +1058,9 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
pr_debug("mmap size %zuB\n", evlist->mmap_len);
mp.mask = evlist->mmap_len - page_size - 1;

+ itrace_mmap_params__init(&mp.itrace_mp, evlist->mmap_len, itrace_pages,
+ itrace_overwrite);
+
evlist__for_each(evlist, evsel) {
if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
evsel->sample_id == NULL &&
@@ -1026,6 +1074,12 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
return perf_evlist__mmap_per_cpu(evlist, &mp);
}

+int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
+ bool overwrite)
+{
+ return perf_evlist__mmap_ex(evlist, pages, overwrite, 0, false);
+}
+
int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
{
evlist->threads = thread_map__new_str(target->pid, target->tid,
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index f07c984..293f6e8 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -8,6 +8,7 @@
#include "event.h"
#include "evsel.h"
#include "util.h"
+#include "itrace.h"
#include <unistd.h>

struct pollfd;
@@ -28,6 +29,7 @@ struct perf_mmap {
int mask;
int refcnt;
unsigned int prev;
+ struct itrace_mmap itrace_mmap;
char event_copy[PERF_SAMPLE_MAX_SIZE] __attribute__((aligned(8)));
};

@@ -126,6 +128,9 @@ int perf_evlist__parse_mmap_pages(const struct option *opt,
const char *str,
int unset);

+int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
+ bool overwrite, unsigned int itrace_pages,
+ bool itrace_overwrite);
int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
bool overwrite);
void perf_evlist__munmap(struct perf_evlist *evlist);
diff --git a/tools/perf/util/itrace.c b/tools/perf/util/itrace.c
new file mode 100644
index 0000000..021f49f
--- /dev/null
+++ b/tools/perf/util/itrace.c
@@ -0,0 +1,101 @@
+/*
+ * itrace.c: Instruction Tracing support
+ * Copyright (c) 2013-2014, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <sys/types.h>
+#include <sys/mman.h>
+#include <stdbool.h>
+
+#include <linux/kernel.h>
+#include <linux/perf_event.h>
+#include <linux/types.h>
+#include <linux/bitops.h>
+#include <linux/log2.h>
+
+#include "../perf.h"
+#include "util.h"
+#include "evlist.h"
+#include "cpumap.h"
+#include "thread_map.h"
+#include "itrace.h"
+
+int itrace_mmap__mmap(struct itrace_mmap *mm, struct itrace_mmap_params *mp,
+ void *userpg, int fd)
+{
+#if BITS_PER_LONG != 64 && !defined(HAVE_SYNC_COMPARE_AND_SWAP_SUPPORT)
+ pr_err("Cannot use Instruction Tracing mmaps\n");
+ return -1;
+#endif
+
+ mm->userpg = userpg;
+ mm->mask = mp->mask;
+ mm->len = mp->len;
+ mm->prev = 0;
+ mm->idx = mp->idx;
+ mm->tid = mp->tid;
+ mm->cpu = mp->cpu;
+
+ if (!mp->len) {
+ mm->base = NULL;
+ return 0;
+ }
+
+ mm->base = mmap(NULL, mp->len, mp->prot, MAP_SHARED, fd, mp->offset);
+ if (mm->base == MAP_FAILED) {
+ pr_debug2("failed to mmap itrace ring buffer\n");
+ mm->base = NULL;
+ return -1;
+ }
+
+ return 0;
+}
+
+void itrace_mmap__munmap(struct itrace_mmap *mm)
+{
+ if (mm->base)
+ munmap(mm->base, mm->len);
+}
+
+void itrace_mmap_params__init(struct itrace_mmap_params *mp,
+ off_t itrace_offset,
+ unsigned int itrace_pages, bool itrace_overwrite)
+{
+ if (itrace_pages) {
+ mp->offset = itrace_offset;
+ mp->len = itrace_pages * (size_t)page_size;
+ mp->mask = is_power_of_2(mp->len) ? mp->len - 1 : 0;
+ mp->prot = PROT_READ | (itrace_overwrite ? 0 : PROT_WRITE);
+ pr_debug2("itrace mmap length %zu\n", mp->len);
+ } else {
+ mp->len = 0;
+ }
+}
+
+void itrace_mmap_params__set_idx(struct itrace_mmap_params *mp,
+ struct perf_evlist *evlist, int idx,
+ bool per_cpu)
+{
+ mp->idx = idx;
+
+ if (per_cpu) {
+ mp->cpu = evlist->cpus->map[idx];
+ if (evlist->threads)
+ mp->tid = evlist->threads->map[0];
+ else
+ mp->tid = -1;
+ } else {
+ mp->cpu = -1;
+ mp->tid = evlist->threads->map[idx];
+ }
+}
diff --git a/tools/perf/util/itrace.h b/tools/perf/util/itrace.h
new file mode 100644
index 0000000..00ba409
--- /dev/null
+++ b/tools/perf/util/itrace.h
@@ -0,0 +1,96 @@
+/*
+ * itrace.h: Instruction Tracing support
+ * Copyright (c) 2013-2014, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#ifndef __PERF_ITRACE_H
+#define __PERF_ITRACE_H
+
+#include <sys/types.h>
+#include <stdbool.h>
+
+#include <linux/perf_event.h>
+#include <linux/types.h>
+
+#include "../perf.h"
+
+struct perf_evlist;
+
+/**
+ * struct itrace_mmap - records an mmap of the itrace buffer.
+ * @base: address of mapped area
+ * @userpg: pointer to buffer's perf_event_mmap_page
+ * @mask: %0 if @len is not a power of two, otherwise (@len - %1)
+ * @len: size of mapped area
+ * @prev: previous aux_head
+ * @idx: index of this mmap
+ * @tid: tid for a per-thread mmap (also set if there is only 1 tid on a per-cpu
+ * mmap) otherwise %0
+ * @cpu: cpu number for a per-cpu mmap otherwise %-1
+ */
+struct itrace_mmap {
+ void *base;
+ void *userpg;
+ size_t mask;
+ size_t len;
+ u64 prev;
+ int idx;
+ pid_t tid;
+ int cpu;
+};
+
+/**
+ * struct itrace_mmap_params - parameters to set up struct itrace_mmap.
+ * @mask: %0 if @len is not a power of two, otherwise (@len - %1)
+ * @offset: file offset of mapped area
+ * @len: size of mapped area
+ * @prot: mmap memory protection
+ * @idx: index of this mmap
+ * @tid: tid for a per-thread mmap (also set if there is only 1 tid on a per-cpu
+ * mmap) otherwise %0
+ * @cpu: cpu number for a per-cpu mmap otherwise %-1
+ */
+struct itrace_mmap_params {
+ size_t mask;
+ off_t offset;
+ size_t len;
+ int prot;
+ int idx;
+ pid_t tid;
+ int cpu;
+};
+
+static inline u64 itrace_mmap__read_head(struct itrace_mmap *mm __maybe_unused)
+{
+ /* Not yet implemented */
+ return 0;
+}
+
+static inline void itrace_mmap__write_tail(struct itrace_mmap *mm __maybe_unused,
+ u64 tail __maybe_unused)
+{
+ /* Not yet implemented */
+}
+
+int itrace_mmap__mmap(struct itrace_mmap *mm,
+ struct itrace_mmap_params *mp,
+ void *userpg, int fd);
+void itrace_mmap__munmap(struct itrace_mmap *mm);
+void itrace_mmap_params__init(struct itrace_mmap_params *mp,
+ off_t itrace_offset,
+ unsigned int itrace_pages, bool itrace_overwrite);
+void itrace_mmap_params__set_idx(struct itrace_mmap_params *mp,
+ struct perf_evlist *evlist, int idx,
+ bool per_cpu);
+
+#endif
--
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/