libbpf: Add unique_match option for multi kprobe
Subsystems (4)
Yonghong Song
reviewerPATCH 1/2 libbpf: Add unique_match option for multi kprobe
Jordan reported an issue in Meta production environment where func try_to_wake_up() is renamed to try_to_wake_up.llvm.<hash>() by clang compiler at lto mode. The original 'kprobe/try_to_wake_up' does not work any more since try_to_wake_up() does not match the actual func name in /proc/kallsyms.
There are a couple of ways to resolve this issue. For example, in attach_kprobe(), we could do lookup in /proc/kallsyms so try_to_wake_up() can be replaced by try_to_wake_up.llvm.<hach>(). Or we can force users to use bpf_program__attach_kprobe() where they need to lookup /proc/kallsyms to find out try_to_wake_up.llvm.<hach>(). But these two approaches requires extra work by either libbpf or user.
Luckily, suggested by Andrii, multi kprobe already supports wildcard ('*') for symbol matching. In the above example, 'try_to_wake_up*' can match to try_to_wake_up() or try_to_wake_up.llvm.<hash>() and this allows bpf prog works for different kernels as some kernels may have try_to_wake_up() and some others may have try_to_wake_up.llvm.<hash>().
The original intention is to kprobe try_to_wake_up() only, so an optional field unique_match is added to struct bpf_kprobe_multi_opts. If the field is set to true, the number of matched functions must be one. Otherwise, the attachment will fail. In the above case, multi kprobe with 'try_to_wake_up*' and unique_match preserves user functionality.
--- tools/lib/bpf/libbpf.c | 10 +++++++++- tools/lib/bpf/libbpf.h | 4 +++- 2 files changed, 12 insertions(+), 2 deletions(-)
tools/lib/bpf/libbpf.c +9 -1
@@ -11522,7 +11522,7 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,struct bpf_link *link =NULL ;const unsigned long *addrs;int err, link_fd, prog_fd; -bool retprobe, session; +bool retprobe, session, unique_match;const __u64 *cookies;const char **syms;size_t cnt;
@@ -11558,6 +11558,14 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, err =libbpf_available_kallsyms_parse (&res);if (err) goto error; + + unique_match =OPTS_GET (opts, unique_match, false); +if (unique_match && res.cnt !=1 ) { +pr_warn ("prog '%s': failed to find unique match: cnt %lu\n" , + prog->name, res.cnt); +return libbpf_err_ptr (-EINVAL ); + } + addrs = res.addrs; cnt = res.cnt; }
tools/lib/bpf/libbpf.h +3 -1
@@ -552,10 +552,12 @@ struct bpf_kprobe_multi_opts {bool retprobe;/* create session kprobes */ bool session; +/* enforce unique match */ +bool unique_match; size_t :0; }; -#define bpf_kprobe_multi_opts__last_field session +#define bpf_kprobe_multi_opts__last_field unique_match LIBBPF_API struct bpf_link * bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,
Signature · 1 line
2.43.5
Yonghong Song
reviewerPATCH 2/2 selftests/bpf: Add a test for kprobe multi with unique_match
Add a kprobe multi subtest to test kprobe multi unique_match option.
--- .../bpf/prog_tests/kprobe_multi_test.c | 27 +++++++++++++++++++ 1 file changed, 27 insertions(+)
tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c +27 -0
@@ -397,6 +397,31 @@ static void test_session_cookie_skel_api(void)kprobe_multi_session_cookie__destroy (skel); } +static void test_unique_match (void ) +{ +LIBBPF_OPTS (bpf_kprobe_multi_opts, opts); +struct kprobe_multi *skel =NULL ; +struct bpf_link *link =NULL ; + + skel =kprobe_multi__open_and_load (); +if (!ASSERT_OK_PTR (skel,"kprobe_multi__open_and_load" )) +return ; + + opts.unique_match = true; + skel->bss->pid =getpid (); + link =bpf_program__attach_kprobe_multi_opts (skel->progs.test_kprobe_manual, +"bpf_fentry_test*" , &opts); +if (!ASSERT_ERR_PTR (link,"bpf_program__attach_kprobe_multi_opts" )) +bpf_link__destroy (link); + + link =bpf_program__attach_kprobe_multi_opts (skel->progs.test_kprobe_manual, +"bpf_fentry_test8*" , &opts); +if (ASSERT_OK_PTR (link,"bpf_program__attach_kprobe_multi_opts" )) +bpf_link__destroy (link); + +kprobe_multi__destroy (skel); +} +static size_t symbol_hash (long key,void *ctx__maybe_unused ) {return str_hash ((const char *) key);
@@ -765,5 +790,7 @@ void test_kprobe_multi_test(void)test_session_skel_api ();if (test__start_subtest ("session_cookie" ))test_session_cookie_skel_api (); +if (test__start_subtest ("unique_match" )) +test_unique_match ();RUN_TESTS (kprobe_multi_verifier); }
Signature · 1 line
2.43.5
Quoted 48 lines · Yonghong Song
Jordan reported an issue in Meta production environment where func
try_to_wake_up()is renamed to try_to_wake_up.llvm.<hash>() by clang compiler at lto mode. The original 'kprobe/try_to_wake_up' does not work any more sincetry_to_wake_up()does not match the actual func name in/proc/kallsyms.There are a couple of ways to resolve this issue. For example, in
attach_kprobe(), we could do lookup in/proc/kallsymssotry_to_wake_up()can be replaced by try_to_wake_up.llvm.<hach>(). Or we can force users to usebpf_program__attach_kprobe()where they need to lookup/proc/kallsymsto find out try_to_wake_up.llvm.<hach>(). But these two approaches requires extra work by either libbpf or user.Luckily, suggested by Andrii, multi kprobe already supports wildcard ('*') for symbol matching. In the above example, 'try_to_wake_up*' can match to
try_to_wake_up()or try_to_wake_up.llvm.<hash>() and this allows bpf prog works for different kernels as some kernels may havetry_to_wake_up()and some others may have try_to_wake_up.llvm.<hash>().The original intention is to kprobe
try_to_wake_up()only, so an optional fieldunique_matchis added to structbpf_kprobe_multi_opts. If the field is set to true, the number of matched functions must be one. Otherwise, the attachment will fail. In the above case, multi kprobe with 'try_to_wake_up*' andunique_matchpreserves user functionality.Reported-by Jordan Rome <[email protected]>Suggested-by Andrii Nakryiko <[email protected]>Signed-off-by Yonghong Song <[email protected]>--- tools/lib/bpf/libbpf.c | 10 +++++++++- tools/lib/bpf/libbpf.h | 4 +++- 2 files changed, 12 insertions(+), 2 deletions(-)tools/lib/bpf/libbpf.c +1
-1@@ -11522,7 +11522,7 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,struct bpf_link *link =NULL ;const unsigned long *addrs;int err, link_fd, prog_fd; -bool retprobe, session; +bool retprobe, session, unique_match;const __u64 *cookies;const char **syms;size_t cnt;
tools/lib/bpf/libbpf.c bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, · line 11565
Quoted 8 lines
tools/lib/bpf/libbpf.c +5
-0@@ -11558,6 +11558,14 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, err =libbpf_available_kallsyms_parse (&res);if (err) goto error; + + unique_match =OPTS_GET (opts, unique_match, false); +if (unique_match && res.cnt !=1 ) { +pr_warn ("prog '%s': failed to find unique match: cnt %lu\n" , + prog->name, res.cnt);
nit: "failed to find a unique match. Num matches: %lu\n" or "Failed to find a unique match for %s'. Num matches: %lu\n"
Quoted 25 lines
tools/lib/bpf/libbpf.c +3
-0+return libbpf_err_ptr (-EINVAL ); + } + addrs = res.addrs; cnt = res.cnt; }tools/lib/bpf/libbpf.h +3
-2@@ -552,10 +552,12 @@ struct bpf_kprobe_multi_opts {bool retprobe;/* create session kprobes */ bool session; +/* enforce unique match */ +bool unique_match; size_t :0; }; -#define bpf_kprobe_multi_opts__last_field session +#define bpf_kprobe_multi_opts__last_field unique_match LIBBPF_API struct bpf_link * bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, --
Quoted 2 lines
2.43.5
Ack. Thanks for the quick work, Yonghong!
Quoted 48 lines · Yonghong Song
Jordan reported an issue in Meta production environment where func
try_to_wake_up()is renamed to try_to_wake_up.llvm.<hash>() by clang compiler at lto mode. The original 'kprobe/try_to_wake_up' does not work any more sincetry_to_wake_up()does not match the actual func name in/proc/kallsyms.There are a couple of ways to resolve this issue. For example, in
attach_kprobe(), we could do lookup in/proc/kallsymssotry_to_wake_up()can be replaced by try_to_wake_up.llvm.<hach>(). Or we can force users to usebpf_program__attach_kprobe()where they need to lookup/proc/kallsymsto find out try_to_wake_up.llvm.<hach>(). But these two approaches requires extra work by either libbpf or user.Luckily, suggested by Andrii, multi kprobe already supports wildcard ('*') for symbol matching. In the above example, 'try_to_wake_up*' can match to
try_to_wake_up()or try_to_wake_up.llvm.<hash>() and this allows bpf prog works for different kernels as some kernels may havetry_to_wake_up()and some others may have try_to_wake_up.llvm.<hash>().The original intention is to kprobe
try_to_wake_up()only, so an optional fieldunique_matchis added to structbpf_kprobe_multi_opts. If the field is set to true, the number of matched functions must be one. Otherwise, the attachment will fail. In the above case, multi kprobe with 'try_to_wake_up*' andunique_matchpreserves user functionality.Reported-by Jordan Rome <[email protected]>Suggested-by Andrii Nakryiko <[email protected]>Signed-off-by Yonghong Song <[email protected]>--- tools/lib/bpf/libbpf.c | 10 +++++++++- tools/lib/bpf/libbpf.h | 4 +++- 2 files changed, 12 insertions(+), 2 deletions(-)tools/lib/bpf/libbpf.c +1
-1@@ -11522,7 +11522,7 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,struct bpf_link *link =NULL ;const unsigned long *addrs;int err, link_fd, prog_fd; -bool retprobe, session; +bool retprobe, session, unique_match;const __u64 *cookies;const char **syms;size_t cnt;
tools/lib/bpf/libbpf.c bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, · line 11566
Quoted 9 lines
tools/lib/bpf/libbpf.c +6
-0@@ -11558,6 +11558,14 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, err =libbpf_available_kallsyms_parse (&res);if (err) goto error; + + unique_match =OPTS_GET (opts, unique_match, false); +if (unique_match && res.cnt !=1 ) { +pr_warn ("prog '%s': failed to find unique match: cnt %lu\n" , + prog->name, res.cnt); +return libbpf_err_ptr (-EINVAL );
goto error, leaking resources here
we should also think about interaction of unique_match interaction for !pattern case, and either reject it (if it makes no sense), or enforce it (if it does, I haven't really thought about which case do we have)
pw-bot: cr
Quoted 26 lines
tools/lib/bpf/libbpf.c +2
-0+ } + addrs = res.addrs; cnt = res.cnt; }tools/lib/bpf/libbpf.h +3
-2@@ -552,10 +552,12 @@ struct bpf_kprobe_multi_opts {bool retprobe;/* create session kprobes */ bool session; +/* enforce unique match */ +bool unique_match; size_t :0; }; -#define bpf_kprobe_multi_opts__last_field session +#define bpf_kprobe_multi_opts__last_field unique_match LIBBPF_API struct bpf_link * bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, --2.43.5
Yonghong Song
reviewerQuoted 50 lines · Andrii Nakryiko
Quoted 49 lines · Yonghong Song
Jordan reported an issue in Meta production environment where func
try_to_wake_up()is renamed to try_to_wake_up.llvm.<hash>() by clang compiler at lto mode. The original 'kprobe/try_to_wake_up' does not work any more sincetry_to_wake_up()does not match the actual func name in/proc/kallsyms.There are a couple of ways to resolve this issue. For example, in
attach_kprobe(), we could do lookup in/proc/kallsymssotry_to_wake_up()can be replaced by try_to_wake_up.llvm.<hach>(). Or we can force users to usebpf_program__attach_kprobe()where they need to lookup/proc/kallsymsto find out try_to_wake_up.llvm.<hach>(). But these two approaches requires extra work by either libbpf or user.Luckily, suggested by Andrii, multi kprobe already supports wildcard ('*') for symbol matching. In the above example, 'try_to_wake_up*' can match to
try_to_wake_up()or try_to_wake_up.llvm.<hash>() and this allows bpf prog works for different kernels as some kernels may havetry_to_wake_up()and some others may have try_to_wake_up.llvm.<hash>().The original intention is to kprobe
try_to_wake_up()only, so an optional fieldunique_matchis added to structbpf_kprobe_multi_opts. If the field is set to true, the number of matched functions must be one. Otherwise, the attachment will fail. In the above case, multi kprobe with 'try_to_wake_up*' andunique_matchpreserves user functionality.Reported-by Jordan Rome <[email protected]>Suggested-by Andrii Nakryiko <[email protected]>Signed-off-by Yonghong Song <[email protected]>--- tools/lib/bpf/libbpf.c | 10 +++++++++- tools/lib/bpf/libbpf.h | 4 +++- 2 files changed, 12 insertions(+), 2 deletions(-)tools/lib/bpf/libbpf.c +1
-1@@ -11522,7 +11522,7 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,struct bpf_link *link =NULL ;const unsigned long *addrs;int err, link_fd, prog_fd; -bool retprobe, session; +bool retprobe, session, unique_match;const __u64 *cookies;const char **syms;size_t cnt;@@ -11558,6 +11558,14 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, err =libbpf_available_kallsyms_parse (&res);if (err)
Quoted 8 lines
tools/lib/bpf/libbpf.c bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, · line 11566
Quoted 7 lines
tools/lib/bpf/libbpf.c +6
-0goto error; + + unique_match =OPTS_GET (opts, unique_match, false); +if (unique_match && res.cnt !=1 ) { +pr_warn ("prog '%s': failed to find unique match: cnt %lu\n" , + prog->name, res.cnt); +return libbpf_err_ptr (-EINVAL );goto error, leaking resources here
Ack. Will fix.
Quoted 5 lines
we should also think about interaction of
unique_matchinteraction for !pattern case, and either reject it (if it makes no sense), or enforce it (if it does, I haven't really thought about which case do we have)
The unique_match only makes sense for pattern case. So I suggest to reject the case unique_match && !pattern. WDYT?
Quoted 29 lines
pw-bot: cr
Quoted 26 lines
tools/lib/bpf/libbpf.c +2
-0+ } + addrs = res.addrs; cnt = res.cnt; }tools/lib/bpf/libbpf.h +3
-2@@ -552,10 +552,12 @@ struct bpf_kprobe_multi_opts {bool retprobe;/* create session kprobes */ bool session; +/* enforce unique match */ +bool unique_match; size_t :0; }; -#define bpf_kprobe_multi_opts__last_field session +#define bpf_kprobe_multi_opts__last_field unique_match LIBBPF_API struct bpf_link * bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, --2.43.5
Quoted 68 lines · Yonghong Song
Quoted 50 lines · Andrii Nakryiko
Quoted 49 lines · Yonghong Song
Jordan reported an issue in Meta production environment where func
try_to_wake_up()is renamed to try_to_wake_up.llvm.<hash>() by clang compiler at lto mode. The original 'kprobe/try_to_wake_up' does not work any more sincetry_to_wake_up()does not match the actual func name in/proc/kallsyms.There are a couple of ways to resolve this issue. For example, in
attach_kprobe(), we could do lookup in/proc/kallsymssotry_to_wake_up()can be replaced by try_to_wake_up.llvm.<hach>(). Or we can force users to usebpf_program__attach_kprobe()where they need to lookup/proc/kallsymsto find out try_to_wake_up.llvm.<hach>(). But these two approaches requires extra work by either libbpf or user.Luckily, suggested by Andrii, multi kprobe already supports wildcard ('*') for symbol matching. In the above example, 'try_to_wake_up*' can match to
try_to_wake_up()or try_to_wake_up.llvm.<hash>() and this allows bpf prog works for different kernels as some kernels may havetry_to_wake_up()and some others may have try_to_wake_up.llvm.<hash>().The original intention is to kprobe
try_to_wake_up()only, so an optional fieldunique_matchis added to structbpf_kprobe_multi_opts. If the field is set to true, the number of matched functions must be one. Otherwise, the attachment will fail. In the above case, multi kprobe with 'try_to_wake_up*' andunique_matchpreserves user functionality.Reported-by Jordan Rome <[email protected]>Suggested-by Andrii Nakryiko <[email protected]>Signed-off-by Yonghong Song <[email protected]>--- tools/lib/bpf/libbpf.c | 10 +++++++++- tools/lib/bpf/libbpf.h | 4 +++- 2 files changed, 12 insertions(+), 2 deletions(-)tools/lib/bpf/libbpf.c +1
-1@@ -11522,7 +11522,7 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,struct bpf_link *link =NULL ;const unsigned long *addrs;int err, link_fd, prog_fd; -bool retprobe, session; +bool retprobe, session, unique_match;const __u64 *cookies;const char **syms;size_t cnt;@@ -11558,6 +11558,14 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, err =libbpf_available_kallsyms_parse (&res);if (err)Quoted 8 lines
tools/lib/bpf/libbpf.c bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, · line 11566
Quoted 7 lines
tools/lib/bpf/libbpf.c +6
-0goto error; + + unique_match =OPTS_GET (opts, unique_match, false); +if (unique_match && res.cnt !=1 ) { +pr_warn ("prog '%s': failed to find unique match: cnt %lu\n" , + prog->name, res.cnt); +return libbpf_err_ptr (-EINVAL );goto error, leaking resources here
Ack. Will fix.
Quoted 2 lines
Quoted 7 lines
Quoted 3 lines
we should also think about interaction of
unique_matchinteraction for !pattern case, and either reject it (if it makes no sense), or enforce it (if it does, I haven't really thought about which case do we have)The
unique_matchonly makes sense for pattern case. So I suggest to reject the caseunique_match&& !pattern. WDYT?
Yep, let's reject (we could make it behave well, just making sure that cnt == 1 if unique_match == true, but why bother, it's not intended to be used together).
Quoted 30 lines
Quoted 29 lines
pw-bot: cr
Quoted 26 lines
tools/lib/bpf/libbpf.c +2
-0+ } + addrs = res.addrs; cnt = res.cnt; }tools/lib/bpf/libbpf.h +3
-2@@ -552,10 +552,12 @@ struct bpf_kprobe_multi_opts {bool retprobe;/* create session kprobes */ bool session; +/* enforce unique match */ +bool unique_match; size_t :0; }; -#define bpf_kprobe_multi_opts__last_field session +#define bpf_kprobe_multi_opts__last_field unique_match LIBBPF_API struct bpf_link * bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog, --2.43.5