* [PATCH bpf-next v11 1/2] selftests/bpf: trace_helpers.c: optimize kallsyms cache
[not found] <cover.1693922135.git.rongtao@cestc.cn>
@ 2023-09-05 14:04 ` Rong Tao
2023-09-06 20:54 ` Jiri Olsa
2023-09-05 14:04 ` [PATCH bpf-next v11 2/2] selftests/bpf: trace_helpers.c: Add a global ksyms initialization mutex Rong Tao
1 sibling, 1 reply; 3+ messages in thread
From: Rong Tao @ 2023-09-05 14:04 UTC (permalink / raw)
To: olsajiri, andrii, daniel, sdf
Cc: Rong Tao, Alexei Starovoitov, Martin KaFai Lau, Song Liu,
Yonghong Song, John Fastabend, KP Singh, Hao Luo, Jiri Olsa,
Mykola Lysenko, Shuah Khan, Maxime Coquelin, Alexandre Torgue,
Yafang Shao,
open list:BPF [GENERAL] (Safe Dynamic Programs and Tools),
open list, open list:KERNEL SELFTEST FRAMEWORK,
moderated list:ARM/STM32 ARCHITECTURE,
moderated list:ARM/STM32 ARCHITECTURE
From: Rong Tao <rongtao@cestc.cn>
Static ksyms often have problems because the number of symbols exceeds the
MAX_SYMS limit. Like changing the MAX_SYMS from 300000 to 400000 in
commit e76a014334a6("selftests/bpf: Bump and validate MAX_SYMS") solves
the problem somewhat, but it's not the perfect way.
This commit uses dynamic memory allocation, which completely solves the
problem caused by the limitation of the number of kallsyms. At the same
time, add APIs:
load_kallsyms_local()
ksym_search_local()
ksym_get_addr_local()
free_kallsyms_local()
There are used to solve the problem of selftests/bpf updating kallsyms
after attach new symbols during testmod testing.
Acked-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Rong Tao <rongtao@cestc.cn>
---
v11: Remove useless load_kallsyms_refresh() and modify code some format
v10: https://lore.kernel.org/lkml/tencent_0A73B402B1D440480838ABF7124CE5EA5505@qq.com/
Keep the original load_kallsyms().
v9: https://lore.kernel.org/lkml/tencent_254B7015EED7A5D112C45E033DA1822CF107@qq.com/
Add load_kallsyms_local,ksym_search_local,ksym_get_addr_local functions.
v8: https://lore.kernel.org/lkml/tencent_6D23FE187408D965E95DFAA858BC7E8C760A@qq.com/
Resolves inter-thread contention for ksyms global variables.
v7: https://lore.kernel.org/lkml/tencent_BD6E19C00BF565CD5C36A9A0BD828CFA210A@qq.com/
Fix __must_check macro.
v6: https://lore.kernel.org/lkml/tencent_4A09A36F883A06EA428A593497642AF8AF08@qq.com/
Apply libbpf_ensure_mem()
v5: https://lore.kernel.org/lkml/tencent_0E9E1A1C0981678D5E7EA9E4BDBA8EE2200A@qq.com/
Release the allocated memory once the load_kallsyms_refresh() upon error
given it's dynamically allocated.
v4: https://lore.kernel.org/lkml/tencent_59C74613113F0C728524B2A82FE5540A5E09@qq.com/
Make sure most cases we don't need the realloc() path to begin with,
and check strdup() return value.
v3: https://lore.kernel.org/lkml/tencent_50B4B2622FE7546A5FF9464310650C008509@qq.com/
Do not use structs and judge ksyms__add_symbol function return value.
v2: https://lore.kernel.org/lkml/tencent_B655EE5E5D463110D70CD2846AB3262EED09@qq.com/
Do the usual len/capacity scheme here to amortize the cost of realloc, and
don't free symbols.
v1: https://lore.kernel.org/lkml/tencent_AB461510B10CD484E0B2F62E3754165F2909@qq.com/
---
samples/bpf/Makefile | 4 +
.../selftests/bpf/prog_tests/fill_link_info.c | 2 +-
.../prog_tests/kprobe_multi_testmod_test.c | 20 ++-
tools/testing/selftests/bpf/trace_helpers.c | 132 +++++++++++++-----
tools/testing/selftests/bpf/trace_helpers.h | 9 +-
5 files changed, 122 insertions(+), 45 deletions(-)
diff --git a/samples/bpf/Makefile b/samples/bpf/Makefile
index 4ccf4236031c..6c707ebcebb9 100644
--- a/samples/bpf/Makefile
+++ b/samples/bpf/Makefile
@@ -175,6 +175,7 @@ TPROGS_CFLAGS += -I$(srctree)/tools/testing/selftests/bpf/
TPROGS_CFLAGS += -I$(LIBBPF_INCLUDE)
TPROGS_CFLAGS += -I$(srctree)/tools/include
TPROGS_CFLAGS += -I$(srctree)/tools/perf
+TPROGS_CFLAGS += -I$(srctree)/tools/lib
TPROGS_CFLAGS += -DHAVE_ATTR_TEST=0
ifdef SYSROOT
@@ -314,6 +315,9 @@ XDP_SAMPLE_CFLAGS += -Wall -O2 \
$(obj)/$(XDP_SAMPLE): TPROGS_CFLAGS = $(XDP_SAMPLE_CFLAGS)
$(obj)/$(XDP_SAMPLE): $(src)/xdp_sample_user.h $(src)/xdp_sample_shared.h
+# Override includes for trace_helpers.o because __must_check won't be defined
+# in our include path.
+$(obj)/$(TRACE_HELPERS): TPROGS_CFLAGS := $(TPROGS_CFLAGS) -D__must_check=
-include $(BPF_SAMPLES_PATH)/Makefile.target
diff --git a/tools/testing/selftests/bpf/prog_tests/fill_link_info.c b/tools/testing/selftests/bpf/prog_tests/fill_link_info.c
index 9d768e083714..97142a4db374 100644
--- a/tools/testing/selftests/bpf/prog_tests/fill_link_info.c
+++ b/tools/testing/selftests/bpf/prog_tests/fill_link_info.c
@@ -308,7 +308,7 @@ void test_fill_link_info(void)
return;
/* load kallsyms to compare the addr */
- if (!ASSERT_OK(load_kallsyms_refresh(), "load_kallsyms_refresh"))
+ if (!ASSERT_OK(load_kallsyms(), "load_kallsyms"))
goto cleanup;
kprobe_addr = ksym_get_addr(KPROBE_FUNC);
diff --git a/tools/testing/selftests/bpf/prog_tests/kprobe_multi_testmod_test.c b/tools/testing/selftests/bpf/prog_tests/kprobe_multi_testmod_test.c
index 1fbe7e4ac00a..8a8ad0613861 100644
--- a/tools/testing/selftests/bpf/prog_tests/kprobe_multi_testmod_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/kprobe_multi_testmod_test.c
@@ -4,6 +4,8 @@
#include "trace_helpers.h"
#include "bpf/libbpf_internal.h"
+static struct ksyms *ksyms;
+
static void kprobe_multi_testmod_check(struct kprobe_multi *skel)
{
ASSERT_EQ(skel->bss->kprobe_testmod_test1_result, 1, "kprobe_test1_result");
@@ -50,12 +52,12 @@ static void test_testmod_attach_api_addrs(void)
LIBBPF_OPTS(bpf_kprobe_multi_opts, opts);
unsigned long long addrs[3];
- addrs[0] = ksym_get_addr("bpf_testmod_fentry_test1");
- ASSERT_NEQ(addrs[0], 0, "ksym_get_addr");
- addrs[1] = ksym_get_addr("bpf_testmod_fentry_test2");
- ASSERT_NEQ(addrs[1], 0, "ksym_get_addr");
- addrs[2] = ksym_get_addr("bpf_testmod_fentry_test3");
- ASSERT_NEQ(addrs[2], 0, "ksym_get_addr");
+ addrs[0] = ksym_get_addr_local(ksyms, "bpf_testmod_fentry_test1");
+ ASSERT_NEQ(addrs[0], 0, "ksym_get_addr_local");
+ addrs[1] = ksym_get_addr_local(ksyms, "bpf_testmod_fentry_test2");
+ ASSERT_NEQ(addrs[1], 0, "ksym_get_addr_local");
+ addrs[2] = ksym_get_addr_local(ksyms, "bpf_testmod_fentry_test3");
+ ASSERT_NEQ(addrs[2], 0, "ksym_get_addr_local");
opts.addrs = (const unsigned long *) addrs;
opts.cnt = ARRAY_SIZE(addrs);
@@ -79,11 +81,15 @@ static void test_testmod_attach_api_syms(void)
void serial_test_kprobe_multi_testmod_test(void)
{
- if (!ASSERT_OK(load_kallsyms_refresh(), "load_kallsyms_refresh"))
+ ksyms = load_kallsyms_local(NULL);
+ if (!ASSERT_OK_PTR(ksyms, "load_kallsyms_local"))
return;
if (test__start_subtest("testmod_attach_api_syms"))
test_testmod_attach_api_syms();
+
if (test__start_subtest("testmod_attach_api_addrs"))
test_testmod_attach_api_addrs();
+
+ free_kallsyms_local(ksyms);
}
diff --git a/tools/testing/selftests/bpf/trace_helpers.c b/tools/testing/selftests/bpf/trace_helpers.c
index f83d9f65c65b..7d026c128252 100644
--- a/tools/testing/selftests/bpf/trace_helpers.c
+++ b/tools/testing/selftests/bpf/trace_helpers.c
@@ -14,104 +14,166 @@
#include <linux/limits.h>
#include <libelf.h>
#include <gelf.h>
+#include "bpf/libbpf_internal.h"
#define TRACEFS_PIPE "/sys/kernel/tracing/trace_pipe"
#define DEBUGFS_PIPE "/sys/kernel/debug/tracing/trace_pipe"
-#define MAX_SYMS 400000
-static struct ksym syms[MAX_SYMS];
-static int sym_cnt;
+struct ksyms {
+ struct ksym *syms;
+ size_t sym_cap;
+ size_t sym_cnt;
+};
+
+static struct ksyms *ksyms;
+
+static int ksyms__add_symbol(struct ksyms *ksyms, const char *name,
+ unsigned long addr)
+{
+ void *tmp;
+
+ tmp = strdup(name);
+ if (!tmp)
+ return -ENOMEM;
+ ksyms->syms[ksyms->sym_cnt].addr = addr;
+ ksyms->syms[ksyms->sym_cnt].name = tmp;
+
+ ksyms->sym_cnt++;
+
+ return 0;
+}
+
+void free_kallsyms_local(struct ksyms *ksyms)
+{
+ unsigned int i;
+
+ if (!ksyms)
+ return;
+
+ if (!ksyms->syms) {
+ free(ksyms);
+ return;
+ }
+
+ for (i = 0; i < ksyms->sym_cnt; i++)
+ free(ksyms->syms[i].name);
+ free(ksyms->syms);
+ free(ksyms);
+}
static int ksym_cmp(const void *p1, const void *p2)
{
return ((struct ksym *)p1)->addr - ((struct ksym *)p2)->addr;
}
-int load_kallsyms_refresh(void)
+struct ksyms *load_kallsyms_local(struct ksyms *ksyms)
{
FILE *f;
char func[256], buf[256];
char symbol;
void *addr;
- int i = 0;
+ int ret;
- sym_cnt = 0;
+ /* flush kallsyms, free the previously allocated dynamic memory */
+ free_kallsyms_local(ksyms);
f = fopen("/proc/kallsyms", "r");
if (!f)
- return -ENOENT;
+ return NULL;
+
+ ksyms = calloc(1, sizeof(struct ksyms));
+ if (!ksyms)
+ return NULL;
while (fgets(buf, sizeof(buf), f)) {
if (sscanf(buf, "%p %c %s", &addr, &symbol, func) != 3)
break;
if (!addr)
continue;
- if (i >= MAX_SYMS)
- return -EFBIG;
- syms[i].addr = (long) addr;
- syms[i].name = strdup(func);
- i++;
+ ret = libbpf_ensure_mem((void **) &ksyms->syms, &ksyms->sym_cap,
+ sizeof(struct ksym), ksyms->sym_cnt + 1);
+ if (ret)
+ goto error;
+ ret = ksyms__add_symbol(ksyms, func, (unsigned long)addr);
+ if (ret)
+ goto error;
}
fclose(f);
- sym_cnt = i;
- qsort(syms, sym_cnt, sizeof(struct ksym), ksym_cmp);
- return 0;
+ qsort(ksyms->syms, ksyms->sym_cnt, sizeof(struct ksym), ksym_cmp);
+ return ksyms;
+
+error:
+ free_kallsyms_local(ksyms);
+ return NULL;
}
int load_kallsyms(void)
{
- /*
- * This is called/used from multiplace places,
- * load symbols just once.
- */
- if (sym_cnt)
- return 0;
- return load_kallsyms_refresh();
+ if (!ksyms)
+ ksyms = load_kallsyms_local(NULL);
+ return ksyms ? 0 : 1;
}
-struct ksym *ksym_search(long key)
+struct ksym *ksym_search_local(struct ksyms *ksyms, long key)
{
- int start = 0, end = sym_cnt;
+ int start = 0, end = ksyms->sym_cnt;
int result;
+ if (!ksyms)
+ return NULL;
+
/* kallsyms not loaded. return NULL */
- if (sym_cnt <= 0)
+ if (ksyms->sym_cnt <= 0)
return NULL;
while (start < end) {
size_t mid = start + (end - start) / 2;
- result = key - syms[mid].addr;
+ result = key - ksyms->syms[mid].addr;
if (result < 0)
end = mid;
else if (result > 0)
start = mid + 1;
else
- return &syms[mid];
+ return &ksyms->syms[mid];
}
- if (start >= 1 && syms[start - 1].addr < key &&
- key < syms[start].addr)
+ if (start >= 1 && ksyms->syms[start - 1].addr < key &&
+ key < ksyms->syms[start].addr)
/* valid ksym */
- return &syms[start - 1];
+ return &ksyms->syms[start - 1];
/* out of range. return _stext */
- return &syms[0];
+ return &ksyms->syms[0];
}
-long ksym_get_addr(const char *name)
+struct ksym *ksym_search(long key)
+{
+ if (!ksyms)
+ return NULL;
+ return ksym_search_local(ksyms, key);
+}
+
+long ksym_get_addr_local(struct ksyms *ksyms, const char *name)
{
int i;
- for (i = 0; i < sym_cnt; i++) {
- if (strcmp(syms[i].name, name) == 0)
- return syms[i].addr;
+ for (i = 0; i < ksyms->sym_cnt; i++) {
+ if (strcmp(ksyms->syms[i].name, name) == 0)
+ return ksyms->syms[i].addr;
}
return 0;
}
+long ksym_get_addr(const char *name)
+{
+ if (!ksyms)
+ return 0;
+ return ksym_get_addr_local(ksyms, name);
+}
+
/* open kallsyms and read symbol addresses on the fly. Without caching all symbols,
* this is faster than load + find.
*/
diff --git a/tools/testing/selftests/bpf/trace_helpers.h b/tools/testing/selftests/bpf/trace_helpers.h
index 876f3e711df6..d6eeec85a5e4 100644
--- a/tools/testing/selftests/bpf/trace_helpers.h
+++ b/tools/testing/selftests/bpf/trace_helpers.h
@@ -11,13 +11,18 @@ struct ksym {
long addr;
char *name;
};
+struct ksyms;
int load_kallsyms(void);
-int load_kallsyms_refresh(void);
-
struct ksym *ksym_search(long key);
long ksym_get_addr(const char *name);
+struct ksyms *load_kallsyms_local(struct ksyms *ksyms);
+struct ksym *ksym_search_local(struct ksyms *ksyms, long key);
+long ksym_get_addr_local(struct ksyms *ksyms, const char *name);
+
+void free_kallsyms_local(struct ksyms *ksyms);
+
/* open kallsyms and find addresses on the fly, faster than load + search. */
int kallsyms_find(const char *sym, unsigned long long *addr);
--
2.41.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* [PATCH bpf-next v11 2/2] selftests/bpf: trace_helpers.c: Add a global ksyms initialization mutex
[not found] <cover.1693922135.git.rongtao@cestc.cn>
2023-09-05 14:04 ` [PATCH bpf-next v11 1/2] selftests/bpf: trace_helpers.c: optimize kallsyms cache Rong Tao
@ 2023-09-05 14:04 ` Rong Tao
1 sibling, 0 replies; 3+ messages in thread
From: Rong Tao @ 2023-09-05 14:04 UTC (permalink / raw)
To: olsajiri, andrii, daniel, sdf
Cc: Rong Tao, Alexei Starovoitov, Martin KaFai Lau, Song Liu,
Yonghong Song, John Fastabend, KP Singh, Hao Luo, Jiri Olsa,
Mykola Lysenko, Shuah Khan, Maxime Coquelin, Alexandre Torgue,
Yafang Shao,
open list:BPF [GENERAL] (Safe Dynamic Programs and Tools),
open list, open list:KERNEL SELFTEST FRAMEWORK,
moderated list:ARM/STM32 ARCHITECTURE,
moderated list:ARM/STM32 ARCHITECTURE
From: Rong Tao <rongtao@cestc.cn>
As Jirka said [0], we just need to make sure that global ksyms
initialization won't race.
[0] https://lore.kernel.org/lkml/ZPCbAs3ItjRd8XVh@krava/
Signed-off-by: Rong Tao <rongtao@cestc.cn>
---
tools/testing/selftests/bpf/trace_helpers.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/tools/testing/selftests/bpf/trace_helpers.c b/tools/testing/selftests/bpf/trace_helpers.c
index 7d026c128252..411f87d5aac7 100644
--- a/tools/testing/selftests/bpf/trace_helpers.c
+++ b/tools/testing/selftests/bpf/trace_helpers.c
@@ -7,6 +7,7 @@
#include <errno.h>
#include <fcntl.h>
#include <poll.h>
+#include <pthread.h>
#include <unistd.h>
#include <linux/perf_event.h>
#include <sys/mman.h>
@@ -26,6 +27,7 @@ struct ksyms {
};
static struct ksyms *ksyms;
+static pthread_mutex_t ksyms_mutex = PTHREAD_MUTEX_INITIALIZER;
static int ksyms__add_symbol(struct ksyms *ksyms, const char *name,
unsigned long addr)
@@ -110,8 +112,10 @@ struct ksyms *load_kallsyms_local(struct ksyms *ksyms)
int load_kallsyms(void)
{
+ pthread_mutex_lock(&ksyms_mutex);
if (!ksyms)
ksyms = load_kallsyms_local(NULL);
+ pthread_mutex_unlock(&ksyms_mutex);
return ksyms ? 0 : 1;
}
--
2.41.0
^ permalink raw reply related [flat|nested] 3+ messages in thread