perf tools: Fix prologue generation
Subsystems (5)
Jiri Olsa
reviewerv2 cover perf tools: Fix prologue generation
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.
Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
struct bpf_prog_prep_result
v2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- realloc prog->insns array in bpf_program__set_insns [Andrii]
- squash patch 1 from previous version with bpf_program__set_insns change [Daniel]
- patch 3 already merged [Arnaldo]
- added more comments
meanwhile.. perf/core and bpf-next diverged, so:
- libbpf bpf_program__set_insns change is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 with bpf_program__set_insns change, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is released
thanks,
jirka
[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Add bpf_program__set_insns function
tools/lib/bpf/libbpf.c | 22 ++++++++++++++++++++++ tools/lib/bpf/libbpf.h | 18 ++++++++++++++++++ tools/lib/bpf/libbpf.map | 1 + 3 files changed, 41 insertions(+)
Jiri Olsa (2):
perf tools: Register fallback libbpf section handler perf tools: Rework prologue generation code
tools/perf/util/bpf-loader.c | 175 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 157 insertions(+), 18 deletions(-)
Jiri Olsa
reviewerPATCH v2 1/3 libbpf: Add bpf_program__set_insns function
Adding bpf_program__set_insns that allows to set new instructions for program.
This is a very advanced libbpf API and users need to know what they are doing. This should be used from prog_prepare_load_fn callback only.
We can have changed instructions after calling prog_prepare_load_fn callback, reloading them.
--- tools/lib/bpf/libbpf.c | 22 ++++++++++++++++++++++ tools/lib/bpf/libbpf.h | 18 ++++++++++++++++++ tools/lib/bpf/libbpf.map | 1 + 3 files changed, 41 insertions(+)
tools/lib/bpf/libbpf.c +22 -0
@@ -6862,6 +6862,8 @@ static int bpf_object_load_prog_instance(struct bpf_object *obj, struct bpf_prog prog->name, err);return err; } + insns = prog->insns; + insns_cnt = prog->insns_cnt; }if (obj->gen_loader) {
@@ -8790,6 +8792,26 @@ size_t bpf_program__insn_cnt(const struct bpf_program *prog)return prog->insns_cnt; } +int bpf_program__set_insns (struct bpf_program *prog, +struct bpf_insn *new_insns,size_t new_insn_cnt) +{ +struct bpf_insn *insns; + +if (prog->obj->loaded) +return -EBUSY ; + + insns =libbpf_reallocarray (prog->insns, new_insn_cnt,sizeof (*insns)); +if (!insns) { +pr_warn ("prog '%s': failed to realloc prog code\n" , prog->name); +return -ENOMEM ; + } +memcpy (insns, new_insns, new_insn_cnt *sizeof (*insns)); + + prog->insns = insns; + prog->insns_cnt = new_insn_cnt; +return 0 ; +} +int bpf_program__set_prep (struct bpf_program *prog,int nr_instances,bpf_program_prep_t prep) {
tools/lib/bpf/libbpf.h +18 -0
@@ -323,6 +323,24 @@ struct bpf_insn;* different. */ LIBBPF_API const struct bpf_insn *bpf_program__insns(const struct bpf_program *prog); + +/** +* @brief **bpf_program__set_insns()** can set BPF program's underlying +* BPF instructions. +* +* WARNING: This is a very advanced libbpf API and users need to know +* what they are doing. This should be used from prog_prepare_load_fn +* callback only. +* +* @param prog BPF program for which to return instructions +* @param new_insns a pointer to an array of BPF instructions +* @param new_insn_cnt number of `struct bpf_insn`'s that form +* specified BPF program +* @return 0, on success; negative error code, otherwise +*/ +LIBBPF_API int bpf_program__set_insns(struct bpf_program *prog, +struct bpf_insn *new_insns,size_t new_insn_cnt); +/** * @brief **bpf_program__insn_cnt()** returns number of `struct bpf_insn`'s * that form specified BPF program.
tools/lib/bpf/libbpf.map +1 -0
@@ -448,6 +448,7 @@ LIBBPF_0.8.0 { bpf_object__open_subskeleton; bpf_program__attach_kprobe_multi_opts; bpf_program__attach_usdt; + bpf_program__set_insns; libbpf_register_prog_handler; libbpf_unregister_prog_handler; } LIBBPF_0.7.0;
Signature · 1 line
2.35.3
Jiri Olsa
reviewerPATCH v2 2/3 perf tools: Register fallback libbpf section handler
In reply to Jiri OlsaPerf is using section name to declare special kprobe arguments, which no longer works with current libbpf, that either requires certain form of the section name or allows to register custom handler.
Adding perf support to register 'fallback' section handler to take care of perf kprobe programs. The fallback means that it handles any section definition besides the ones that libbpf handles.
The handler serves two purposes:
- allows perf programs to have special arguments in section name
- allows perf to use pre-load callback where we can attach init code (zeroing all argument registers) to each perf program
The second is essential part of new prologue generation code, that's coming in following patch.
--- tools/perf/util/bpf-loader.c | 47 ++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+)
tools/perf/util/bpf-loader.c +47 -0
@@ -86,6 +86,7 @@ bpf_perf_object__next(struct bpf_perf_object *prev) (perf_obj) = (tmp), (tmp) = bpf_perf_object__next(tmp))static bool libbpf_initialized; +static int libbpf_sec_handler;static int bpf_perf_object__add (struct bpf_object *obj) {
@@ -99,12 +100,58 @@ static int bpf_perf_object__add(struct bpf_object *obj)return perf_obj ?0 : -ENOMEM ; } +static struct bpf_insn prologue_init_insn[] = { +BPF_MOV64_IMM (BPF_REG_0 ,0 ), +BPF_MOV64_IMM (BPF_REG_1 ,0 ), +BPF_MOV64_IMM (BPF_REG_2 ,0 ), +BPF_MOV64_IMM (BPF_REG_3 ,0 ), +BPF_MOV64_IMM (BPF_REG_4 ,0 ), +BPF_MOV64_IMM (BPF_REG_5 ,0 ), +}; + +static int libbpf_prog_prepare_load_fn (struct bpf_program *prog, +struct bpf_prog_load_opts *opts__maybe_unused , +long cookie__maybe_unused ) +{ +size_t init_size_cnt =ARRAY_SIZE (prologue_init_insn); +size_t orig_insn_cnt, insn_cnt, init_size, orig_size; +const struct bpf_insn *orig_insn; +struct bpf_insn *insn; + +/* prepend initialization code to program instructions */ + orig_insn =bpf_program__insns (prog); + orig_insn_cnt =bpf_program__insn_cnt (prog); + init_size = init_size_cnt *sizeof (*insn); + orig_size = orig_insn_cnt *sizeof (*insn); + + insn_cnt = orig_insn_cnt + init_size_cnt; + insn =malloc (insn_cnt *sizeof (*insn)); +if (!insn) +return -ENOMEM ; + +memcpy (insn, prologue_init_insn, init_size); +memcpy ((char *) insn + init_size, orig_insn, orig_size); +bpf_program__set_insns (prog, insn, insn_cnt); +return 0 ; +} +static int libbpf_init (void ) { +LIBBPF_OPTS (libbpf_prog_handler_opts, handler_opts, + .prog_prepare_load_fn = libbpf_prog_prepare_load_fn, + ); +if (libbpf_initialized)return 0 ;libbpf_set_print (libbpf_perf_print); + libbpf_sec_handler =libbpf_register_prog_handler (NULL ,BPF_PROG_TYPE_KPROBE , +0 , &handler_opts); +if (libbpf_sec_handler <0 ) { +pr_debug ("bpf: failed to register libbpf section handler: %d\n" , + libbpf_sec_handler); +return -BPF_LOADER_ERRNO__INTERNAL ; + } libbpf_initialized = true;return 0 ; }
Signature · 1 line
2.35.3
Jiri Olsa
reviewerPATCH v2 3/3 perf tools: Rework prologue generation code
In reply to Jiri OlsaSome functions we use for bpf prologue generation are going to be deprecated. This change reworks current code not to use them.
We need to replace following functions/struct:
bpf_program__set_prep
bpf_program__nth_fd
struct bpf_prog_prep_result
Currently we use bpf_program__set_prep to hook perf callback before program is loaded and provide new instructions with the prologue.
We replace this function/ality by taking instructions for specific program, attaching prologue to them and load such new ebpf programs with prologue using separate bpf_prog_load calls (outside libbpf load machinery).
Before we can take and use program instructions, we need libbpf to actually load it. This way we get the final shape of its instructions with all relocations and verifier adjustments).
There's one glitch though.. perf kprobe program already assumes generated prologue code with proper values in argument registers, so loading such program directly will fail in the verifier.
That's where the fallback pre-load handler fits in and prepends the initialization code to the program. Once such program is loaded we take its instructions, cut off the initialization code and prepend the prologue.
I know.. sorry ;-)
--- tools/perf/util/bpf-loader.c | 128 ++++++++++++++++++++++++++++++----- 1 file changed, 110 insertions(+), 18 deletions(-)
tools/perf/util/bpf-loader.c +110 -18
@@ -9,6 +9,7 @@#include <linux/bpf.h> #include <bpf/libbpf.h> #include <bpf/bpf.h> +#include <linux/filter.h> #include <linux/err.h> #include <linux/kernel.h> #include <linux/string.h>
@@ -49,6 +50,7 @@ struct bpf_prog_priv {struct bpf_insn *insns_buf;int nr_types;int *type_mapping; +int *proglogue_fds; };struct bpf_perf_object {
@@ -56,6 +58,11 @@ struct bpf_perf_object {struct bpf_object *obj; }; +struct bpf_preproc_result { +struct bpf_insn *new_insn_ptr; +int new_insn_cnt; +}; +static LIST_HEAD (bpf_objects_list);static struct hashmap *bpf_program_hash;static struct hashmap *bpf_map_hash;
@@ -235,14 +242,31 @@ struct bpf_object *bpf__prepare_load(const char *filename, bool source)return obj; } +static void close_prologue_programs (struct bpf_prog_priv *priv) +{ +struct perf_probe_event *pev; +int i, fd; + +if (!priv->need_prologue) +return ; + pev = &priv->pev; +for (i =0 ; i < pev->ntevs; i++) { + fd = priv->proglogue_fds[i]; +if (fd !=-1 ) +close (fd); + } +} +static void clear_prog_priv (const struct bpf_program *prog__maybe_unused ,void *_priv) {struct bpf_prog_priv *priv = _priv; +close_prologue_programs (priv);cleanup_perf_probe_events (&priv->pev,1 );zfree (&priv->insns_buf); +zfree (&priv->proglogue_fds);zfree (&priv->type_mapping);zfree (&priv->sys_name);zfree (&priv->evt_name);
@@ -605,8 +629,8 @@ static int bpf__prepare_probe(void)static int preproc_gen_prologue (struct bpf_program *prog,int n, -struct bpf_insn *orig_insns,int orig_insns_cnt, -struct bpf_prog_prep_result *res) +const struct bpf_insn *orig_insns,int orig_insns_cnt, +struct bpf_preproc_result *res) {struct bpf_prog_priv *priv =program_priv (prog);struct probe_trace_event *tev;
@@ -654,7 +678,6 @@ preproc_gen_prologue(struct bpf_program *prog, int n, res->new_insn_ptr = buf; res->new_insn_cnt = prologue_cnt + orig_insns_cnt; - res->pfd =NULL ;return 0 ; errout:
@@ -762,7 +785,7 @@ static int hook_load_preprocessor(struct bpf_program *prog)struct bpf_prog_priv *priv =program_priv (prog);struct perf_probe_event *pev;bool need_prologue = false; -int err, i; +int i;if (IS_ERR_OR_NULL (priv)) {pr_debug ("Internal error when hook preprocessor\n" );
@@ -800,6 +823,13 @@ static int hook_load_preprocessor(struct bpf_program *prog)return -ENOMEM ; } + priv->proglogue_fds =malloc (sizeof (int ) * pev->ntevs); +if (!priv->proglogue_fds) { +pr_debug ("Not enough memory: alloc prologue fds failed\n" ); +return -ENOMEM ; + } +memset (priv->proglogue_fds,-1 ,sizeof (int ) * pev->ntevs); + priv->type_mapping =malloc (sizeof (int ) * pev->ntevs);if (!priv->type_mapping) {pr_debug ("Not enough memory: alloc type_mapping failed\n" );
@@ -808,13 +838,7 @@ static int hook_load_preprocessor(struct bpf_program *prog)memset (priv->type_mapping,-1 ,sizeof (int ) * pev->ntevs); - err =map_prologue (pev, priv->type_mapping, &priv->nr_types); -if (err) -return err; - - err =bpf_program__set_prep (prog, priv->nr_types, - preproc_gen_prologue); -return err; +return map_prologue (pev, priv->type_mapping, &priv->nr_types); }int bpf__probe (struct bpf_object *obj)
@@ -921,6 +945,77 @@ int bpf__unprobe(struct bpf_object *obj)return ret; } +static int bpf_object__load_prologue (struct bpf_object *obj) +{ +int init_cnt =ARRAY_SIZE (prologue_init_insn); +const struct bpf_insn *orig_insns; +struct bpf_preproc_result res; +struct perf_probe_event *pev; +struct bpf_program *prog; +int orig_insns_cnt; + +bpf_object__for_each_program (prog, obj) { +struct bpf_prog_priv *priv =program_priv (prog); +int err, i, fd; + +if (IS_ERR_OR_NULL (priv)) { +pr_debug ("bpf: failed to get private field\n" ); +return -BPF_LOADER_ERRNO__INTERNAL ; + } + +if (!priv->need_prologue) +continue ; + +/* +* For each program that needs prologue we do following: +* +* - take its current instructions and use them +* to generate the new code with prologue +* - load new instructions with bpf_prog_load +* and keep the fd in proglogue_fds +* - new fd will be used in bpf__foreach_event +* to connect this program with perf evsel +*/ + orig_insns =bpf_program__insns (prog); + orig_insns_cnt =bpf_program__insn_cnt (prog); + + pev = &priv->pev; +for (i =0 ; i < pev->ntevs; i++) { +/* +* Skipping artificall prologue_init_insn instructions +* (init_cnt), so the prologue can be generated instead +* of them. +*/ + err =preproc_gen_prologue (prog, i, + orig_insns + init_cnt, + orig_insns_cnt - init_cnt, + &res); +if (err) +return err; + + fd =bpf_prog_load (bpf_program__get_type (prog), +bpf_program__name (prog),"GPL" , + res.new_insn_ptr, + res.new_insn_cnt,NULL ); +if (fd <0 ) { +char bf[128 ]; + +libbpf_strerror (-errno, bf,sizeof (bf)); +pr_debug ("bpf: load objects with prologue failed: err=%d: (%s)\n" , + -errno, bf); +return -errno; + } + priv->proglogue_fds[i] = fd; + } +/* +* We no longer need the original program, +* we can unload it. +*/ +bpf_program__unload (prog); + } +return 0 ; +} +int bpf__load (struct bpf_object *obj) {int err;
@@ -932,7 +1027,7 @@ int bpf__load(struct bpf_object *obj)pr_debug ("bpf: load objects failed: err=%d: (%s)\n" , err, bf);return err; } -return 0 ; +return bpf_object__load_prologue (obj); } int bpf__foreach_event(struct bpf_object *obj,
@@ -967,13 +1062,10 @@ int bpf__foreach_event(struct bpf_object *obj,for (i =0 ; i < pev->ntevs; i++) { tev = &pev->tevs[i]; -if (priv->need_prologue) { -int type = priv->type_mapping[i]; - - fd =bpf_program__nth_fd (prog, type); - }else { +if (priv->need_prologue) + fd = priv->proglogue_fds[i]; +else fd =bpf_program__fd (prog); - }if (fd <0 ) {pr_debug ("bpf: failed to get file descriptor\n" );
Signature · 1 line
2.35.3
Re: [PATCHv2 perf/core 2/3] perf tools: Register fallback libbpf section handler
In reply to Jiri OlsaQuoted 36 lines · Jiri Olsa
Perf is using section name to declare special kprobe arguments, which no longer works with current libbpf, that either requires certain form of the section name or allows to register custom handler.
Adding perf support to register 'fallback' section handler to take care of perf kprobe programs. The fallback means that it handles any section definition besides the ones that libbpf handles.
The handler serves two purposes:
- allows perf programs to have special arguments in section name
- allows perf to use pre-load callback where we can attach init code (zeroing all argument registers) to each perf programThe second is essential part of new prologue generation code, that's coming in following patch.
Signed-off-by Jiri Olsa <[email protected]>--- tools/perf/util/bpf-loader.c | 47 ++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+)tools/perf/util/bpf-loader.c +1
-0@@ -86,6 +86,7 @@ bpf_perf_object__next(struct bpf_perf_object *prev) (perf_obj) = (tmp), (tmp) = bpf_perf_object__next(tmp))static bool libbpf_initialized; +static int libbpf_sec_handler;static int bpf_perf_object__add (struct bpf_object *obj) {
tools/perf/util/bpf-loader.c static int bpf_perf_object__add(struct bpf_object *obj) · line 105
Quoted 6 lines
tools/perf/util/bpf-loader.c +3
-0@@ -99,12 +100,58 @@ static int bpf_perf_object__add(struct bpf_object *obj)return perf_obj ?0 : -ENOMEM ; } +static struct bpf_insn prologue_init_insn[] = { +BPF_MOV64_IMM (BPF_REG_0 ,0 ), +BPF_MOV64_IMM (BPF_REG_1 ,0 ),
R0 should be initialized before exit anyway. R1 contains context, so doesn't need initialization, so I think you only need R2-R5?
Quoted 55 lines
tools/perf/util/bpf-loader.c +43
-1+BPF_MOV64_IMM (BPF_REG_2 ,0 ), +BPF_MOV64_IMM (BPF_REG_3 ,0 ), +BPF_MOV64_IMM (BPF_REG_4 ,0 ), + BPF_MOV64_IMM(BPF_REG_5, 0), +}; + +static int libbpf_prog_prepare_load_fn (struct bpf_program *prog, +struct bpf_prog_load_opts *opts__maybe_unused , +long cookie__maybe_unused ) +{ +size_t init_size_cnt =ARRAY_SIZE (prologue_init_insn); +size_t orig_insn_cnt, insn_cnt, init_size, orig_size; +const struct bpf_insn *orig_insn; +struct bpf_insn *insn; + +/* prepend initialization code to program instructions */ + orig_insn =bpf_program__insns (prog); + orig_insn_cnt =bpf_program__insn_cnt (prog); + init_size = init_size_cnt *sizeof (*insn); + orig_size = orig_insn_cnt *sizeof (*insn); + + insn_cnt = orig_insn_cnt + init_size_cnt; + insn =malloc (insn_cnt *sizeof (*insn)); +if (!insn) +return -ENOMEM ; + +memcpy (insn, prologue_init_insn, init_size); +memcpy ((char *) insn + init_size, orig_insn, orig_size); +bpf_program__set_insns (prog, insn, insn_cnt); +return 0 ; +} +static int libbpf_init (void ) { +LIBBPF_OPTS (libbpf_prog_handler_opts, handler_opts, + .prog_prepare_load_fn = libbpf_prog_prepare_load_fn, + ); +if (libbpf_initialized)return 0 ;libbpf_set_print (libbpf_perf_print); + libbpf_sec_handler =libbpf_register_prog_handler (NULL ,BPF_PROG_TYPE_KPROBE , +0 , &handler_opts); +if (libbpf_sec_handler <0 ) { +pr_debug ("bpf: failed to register libbpf section handler: %d\n" , + libbpf_sec_handler); +return -BPF_LOADER_ERRNO__INTERNAL ; + } libbpf_initialized = true;return 0 ; } --2.35.3
Quoted 31 lines · Jiri Olsa
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is released
Quoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunction
The first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the bpf_program__set_insns() patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?
Quoted 11 lines
tools/lib/bpf/libbpf.c | 22 ++++++++++++++++++++++ tools/lib/bpf/libbpf.h | 18 ++++++++++++++++++ tools/lib/bpf/libbpf.map | 1 + 3 files changed, 41 insertions(+)Jiri Olsa (2):
perf tools: Register fallback libbpf section handler perf tools: Rework prologue generation codetools/perf/util/bpf-loader.c | 175 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 157 insertions(+), 18 deletions(-)
Quoted 45 lines · Andrii Nakryiko
Quoted 31 lines · Jiri Olsa
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?
we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
at the moment please just take libbpf change and I'll resend the perf change later if needed
thanks,
jirka
Re: [PATCHv2 perf/core 2/3] perf tools: Register fallback libbpf section handler
In reply to Andrii NakryikoQuoted 40 lines · Andrii Nakryiko
Quoted 39 lines · Jiri Olsa
Perf is using section name to declare special kprobe arguments, which no longer works with current libbpf, that either requires certain form of the section name or allows to register custom handler.
Adding perf support to register 'fallback' section handler to take care of perf kprobe programs. The fallback means that it handles any section definition besides the ones that libbpf handles.
The handler serves two purposes:
- allows perf programs to have special arguments in section name
- allows perf to use pre-load callback where we can attach init code (zeroing all argument registers) to each perf programThe second is essential part of new prologue generation code, that's coming in following patch.
Signed-off-by Jiri Olsa <[email protected]>--- tools/perf/util/bpf-loader.c | 47 ++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+)tools/perf/util/bpf-loader.c +1
-0@@ -86,6 +86,7 @@ bpf_perf_object__next(struct bpf_perf_object *prev) (perf_obj) = (tmp), (tmp) = bpf_perf_object__next(tmp))static bool libbpf_initialized; +static int libbpf_sec_handler;static int bpf_perf_object__add (struct bpf_object *obj) {@@ -99,12 +100,58 @@ static int bpf_perf_object__add(struct bpf_object *obj)return perf_obj ?0 : -ENOMEM ; }
Quoted 6 lines
tools/perf/util/bpf-loader.c static int bpf_perf_object__add(struct bpf_object *obj) · line 105
Quoted 3 lines
tools/perf/util/bpf-loader.c +3
-0+static struct bpf_insn prologue_init_insn[] = { +BPF_MOV64_IMM (BPF_REG_0 ,0 ), +BPF_MOV64_IMM (BPF_REG_1 ,0 ),R0 should be initialized before exit anyway. R1 contains context, so doesn't need initialization, so I think you only need R2-R5?
ah right, I'll remove that
thanks,
jirka
Quoted 56 lines
Quoted 56 lines
tools/perf/util/bpf-loader.c +43
-1+BPF_MOV64_IMM (BPF_REG_2 ,0 ), +BPF_MOV64_IMM (BPF_REG_3 ,0 ), +BPF_MOV64_IMM (BPF_REG_4 ,0 ), + BPF_MOV64_IMM(BPF_REG_5, 0), +}; + +static int libbpf_prog_prepare_load_fn (struct bpf_program *prog, +struct bpf_prog_load_opts *opts__maybe_unused , +long cookie__maybe_unused ) +{ +size_t init_size_cnt =ARRAY_SIZE (prologue_init_insn); +size_t orig_insn_cnt, insn_cnt, init_size, orig_size; +const struct bpf_insn *orig_insn; +struct bpf_insn *insn; + +/* prepend initialization code to program instructions */ + orig_insn =bpf_program__insns (prog); + orig_insn_cnt =bpf_program__insn_cnt (prog); + init_size = init_size_cnt *sizeof (*insn); + orig_size = orig_insn_cnt *sizeof (*insn); + + insn_cnt = orig_insn_cnt + init_size_cnt; + insn =malloc (insn_cnt *sizeof (*insn)); +if (!insn) +return -ENOMEM ; + +memcpy (insn, prologue_init_insn, init_size); +memcpy ((char *) insn + init_size, orig_insn, orig_size); +bpf_program__set_insns (prog, insn, insn_cnt); +return 0 ; +} +static int libbpf_init (void ) { +LIBBPF_OPTS (libbpf_prog_handler_opts, handler_opts, + .prog_prepare_load_fn = libbpf_prog_prepare_load_fn, + ); +if (libbpf_initialized)return 0 ;libbpf_set_print (libbpf_perf_print); + libbpf_sec_handler =libbpf_register_prog_handler (NULL ,BPF_PROG_TYPE_KPROBE , +0 , &handler_opts); +if (libbpf_sec_handler <0 ) { +pr_debug ("bpf: failed to register libbpf section handler: %d\n" , + libbpf_sec_handler); +return -BPF_LOADER_ERRNO__INTERNAL ; + } libbpf_initialized = true;return 0 ; } --2.35.3
This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:
Hello:
This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:
Quoted 8 lines · On Tue, 10 May 2022 09:46:56 +0200 you wrote:
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
[...]
Here is the summary with links:
- [PATCHv2,bpf-next,1/3] libbpf: Add bpf_program__set_insns function
https://git.kernel.org/bpf/bpf-next/c/b63b3c490eee
- [PATCHv2,perf/core,2/3] perf tools: Register fallback libbpf section handler (no matching commit)
- [PATCHv2,perf/core,3/3] perf tools: Rework prologue generation code (no matching commit)
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Arnaldo Carvalho de Melo
maintainerEm Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 54 lines · Andrii Nakryiko
Quoted 45 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?
Quoted 6 lines
we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
at the moment please just take libbpf change and I'll resend the perf change later if needed
Ok.
- Arnaldo
Quoted 56 lines · Arnaldo Carvalho de Melo
Em Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 54 lines · Andrii Nakryiko
Quoted 52 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?
Quoted 9 lines
Quoted 6 lines
we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
at the moment please just take libbpf change and I'll resend the perf change later if needed
Ok.
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
Quoted 1 line
- Arnaldo
Ian Rogers
reviewerQuoted 63 lines · Andrii Nakryiko
Quoted 60 lines · Arnaldo Carvalho de Melo
Em Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 58 lines · Andrii Nakryiko
Quoted 45 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
at the moment please just take libbpf change and I'll resend the perf change later if needed
Ok.
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
We still haven't done as Andrii suggested in:
https://lore.kernel.org/lkml/CAEf4BzbbOHQZUAe6iWaehKCPQAf3VC=hq657buqe2_yRKxaK-A@mail.gmail.com/
so for LIBBPF_DYNAMIC I believe we're compiling against the header files in tools/lib rather than the installed libbpf's header files. This may have some unexpected consequences.
Thanks,
Ian
Quoted 2 lines
Quoted 2 lines
- Arnaldo
Quoted 62 lines · Andrii Nakryiko
Quoted 60 lines · Arnaldo Carvalho de Melo
Em Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 58 lines · Andrii Nakryiko
Quoted 45 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
at the moment please just take libbpf change and I'll resend the perf change later if needed
Ok.
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
jirka
Quoted 71 lines · Ian Rogers
Quoted 69 lines · Andrii Nakryiko
Quoted 65 lines · Arnaldo Carvalho de Melo
Em Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 54 lines · Andrii Nakryiko
Quoted 45 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?Quoted 6 lines
we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
at the moment please just take libbpf change and I'll resend the perf change later if needed
Ok.
Quoted 7 lines
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
We still haven't done as Andrii suggested in:
https://lore.kernel.org/lkml/CAEf4BzbbOHQZUAe6iWaehKCPQAf3VC=hq657buqe2_yRKxaK-A@mail.gmail.com/
so forLIBBPF_DYNAMICI believe we're compiling against the header files intools/librather than the installed libbpf's header files. This may have some unexpected consequences.
like your program is not doing what you expected.. been there ;-) sounds good, will check
jirka
Quoted 66 lines · Jiri Olsa
Quoted 65 lines · Andrii Nakryiko
Quoted 63 lines · Arnaldo Carvalho de Melo
Em Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 60 lines · Andrii Nakryiko
Quoted 45 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
at the moment please just take libbpf change and I'll resend the perf change later if needed
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
Ok.
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the bpf_program__set_insns change.. Arnaldo, any idea when this could happen?
thanks,
jirka
Quoted 71 lines · Jiri Olsa
Quoted 69 lines · Jiri Olsa
Quoted 68 lines · Andrii Nakryiko
Quoted 65 lines · Arnaldo Carvalho de Melo
Em Wed, May 11, 2022 at 09:35:34AM +0200, Jiri Olsa escreveu:
Quoted 54 lines · Andrii Nakryiko
Quoted 45 lines · Jiri Olsa
Quoted 31 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultv2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedQuoted 9 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctionThe first patch looks good to me. The rest I can't really review and test properly, so I'll leave it up to Arnaldo.
Quoted 7 lines
Arnaldo, how do we coordinate these patches? Should they go through bpf-next (after you Ack them) or you want them in your tree?
I'd like to get the
bpf_program__set_insns()patch into bpf-next so that I can do libbpf v0.8 release, having it in a separate tree is extremely inconvenient. Please let me know how you think we should proceed?Quoted 6 lines
we need to wait with perf changes after the libbpf is merged and libbpf 0.8.0 is released.. so we don't break dynamic linking for perf
at the moment please just take libbpf change and I'll resend the perf change later if needed
Ok.
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?
Jiri, Arnaldo,
Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
- bpf_map__next;
- bpf_program__next;
- bpf_load_program;
- btf__get_from_id;
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Quoted 2 lines
thanks,
jirka
Arnaldo Carvalho de Melo
maintainerEm Fri, May 20, 2022 at 02:46:49PM -0700, Andrii Nakryiko escreveu:
Quoted 19 lines · Jiri Olsa
Quoted 9 lines · Jiri Olsa
Quoted 4 lines · Andrii Nakryiko
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
-bpf_map__next;
-bpf_program__next;
-bpf_load_program;
-btf__get_from_id;
Quoted 4 lines
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Yeah, that should be ok, the only consideration is that I'm submitting this today to Linus:
To address this:
https://lore.kernel.org/linux-perf-users/[email protected]/
- Arnaldo
Quoted 19 lines · Arnaldo Carvalho de Melo
Em Fri, May 20, 2022 at 02:46:49PM -0700, Andrii Nakryiko escreveu:
Quoted 18 lines · Jiri Olsa
Quoted 9 lines · Jiri Olsa
Quoted 4 lines · Andrii Nakryiko
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
-bpf_map__next;
-bpf_program__next;
-bpf_load_program;
-btf__get_from_id;
these were added just to bypass the time window when they were not available in the package, so can be removed now (in the patch below)
Quoted 6 lines
Quoted 5 lines
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Quoted 8 lines
Yeah, that should be ok, the only consideration is that I'm submitting this today to Linus:
To address this:
ok, we can do that via bpf-next, but of course there's a problem ;-)
perf/core already has dependency commit [1]
so either we wait for perf/core and bpf-next/master to sync or:
- perf/core reverts [1] and
- bpf-next/master takes [1] and the rest
I have the changes ready if you guys are ok with that
thanks,
jirka
[1] https://lore.kernel.org/bpf/[email protected]/
---
Subject: [PATCH bpf-next] perf tools: Remove the weak libbpf functions
We added weak functions for some new libbpf functions because they were not packaged at that time [1].
These functions are now available in package, so we can remove their weak perf variants.
[1] https://lore.kernel.org/linux-perf-users/[email protected]/
--- tools/perf/util/bpf-event.c | 51 ------------------------------------- 1 file changed, 51 deletions(-)
tools/perf/util/bpf-event.c +0 -51
@@ -22,57 +22,6 @@#include "record.h" #include "util/synthetic-events.h" -struct btf *__weak btf__load_from_kernel_by_id (__u32 id) -{ -struct btf *btf; -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -int err =btf__get_from_id (id, &btf); -#pragma GCC diagnostic pop - -return err ?ERR_PTR (err) : btf; -} - -int __weak bpf_prog_load (enum bpf_prog_type prog_type, -const char *prog_name__maybe_unused , -const char *license, -const struct bpf_insn *insns,size_t insn_cnt, -const struct bpf_prog_load_opts *opts) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return bpf_load_program (prog_type, insns, insn_cnt, license, - opts->kern_version, opts->log_buf, opts->log_size); -#pragma GCC diagnostic pop -} - -struct bpf_program *__weak -bpf_object__next_program (const struct bpf_object *obj,struct bpf_program *prev) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return bpf_program__next (prev, obj); -#pragma GCC diagnostic pop -} - -struct bpf_map *__weak -bpf_object__next_map (const struct bpf_object *obj,const struct bpf_map *prev) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return bpf_map__next (prev, obj); -#pragma GCC diagnostic pop -} - -const void *__weak -btf__raw_data (const struct btf *btf_ro,__u32 *size) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return btf__get_raw_data (btf_ro, size); -#pragma GCC diagnostic pop -} -static int snprintf_hex (char *buf,size_t size,unsigned char *data,size_t len) {int ret =0 ;
Signature · 1 line
2.35.3
Quoted 42 lines · Jiri Olsa
Quoted 19 lines · Arnaldo Carvalho de Melo
Em Fri, May 20, 2022 at 02:46:49PM -0700, Andrii Nakryiko escreveu:
Quoted 18 lines · Jiri Olsa
Quoted 9 lines · Jiri Olsa
Quoted 4 lines · Andrii Nakryiko
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
-bpf_map__next;
-bpf_program__next;
-bpf_load_program;
-btf__get_from_id;these were added just to bypass the time window when they were not available in the package, so can be removed now (in the patch below)
Quoted 6 lines
Quoted 5 lines
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Quoted 8 lines
Yeah, that should be ok, the only consideration is that I'm submitting this today to Linus:
To address this:
ok, we can do that via bpf-next, but of course there's a problem ;-)
Quoted 8 lines
perf/core already has dependency commit [1]
so either we wait for perf/core and bpf-next/master to sync or:
- perf/core reverts [1] and
- bpf-next/master takes [1] and the restI have the changes ready if you guys are ok with that
So, if I understand correctly, with merge window open bpf-next/master will get code from perf/core soon when we merge tip back in. So we can wait for that to happen and not revert anything.
So please add the below patch to your series and resend once tip is merged into bpf-next? Thanks!
Quoted 88 lines
thanks,
jirka[1] https://lore.kernel.org/bpf/[email protected]/
---
Subject: [PATCH bpf-next] perf tools: Remove the weak libbpf functionsWe added weak functions for some new libbpf functions because they were not packaged at that time [1].
These functions are now available in package, so we can remove their weak perf variants.
[1] https://lore.kernel.org/linux-perf-users/[email protected]/
Signed-off-by Jiri Olsa <[email protected]>--- tools/perf/util/bpf-event.c | 51 ------------------------------------- 1 file changed, 51 deletions(-)tools/perf/util/bpf-event.c +0
-52@@ -22,57 +22,6 @@#include "record.h" #include "util/synthetic-events.h" -struct btf *__weak btf__load_from_kernel_by_id (__u32 id) -{ -struct btf *btf; -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -int err =btf__get_from_id (id, &btf); -#pragma GCC diagnostic pop - -return err ?ERR_PTR (err) : btf; -} - -int __weak bpf_prog_load (enum bpf_prog_type prog_type, -const char *prog_name__maybe_unused , -const char *license, -const struct bpf_insn *insns,size_t insn_cnt, -const struct bpf_prog_load_opts *opts) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return bpf_load_program (prog_type, insns, insn_cnt, license, - opts->kern_version, opts->log_buf, opts->log_size); -#pragma GCC diagnostic pop -} - -struct bpf_program *__weak -bpf_object__next_program (const struct bpf_object *obj,struct bpf_program *prev) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return bpf_program__next (prev, obj); -#pragma GCC diagnostic pop -} - -struct bpf_map *__weak -bpf_object__next_map (const struct bpf_object *obj,const struct bpf_map *prev) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return bpf_map__next (prev, obj); -#pragma GCC diagnostic pop -} - -const void *__weak -btf__raw_data (const struct btf *btf_ro,__u32 *size) -{ -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wdeprecated-declarations" -return btf__get_raw_data (btf_ro, size); -#pragma GCC diagnostic pop -} -static int snprintf_hex (char *buf,size_t size,unsigned char *data,size_t len) {int ret =0 ; --2.35.3
Quoted 50 lines · Andrii Nakryiko
Quoted 49 lines · Jiri Olsa
Quoted 19 lines · Arnaldo Carvalho de Melo
Em Fri, May 20, 2022 at 02:46:49PM -0700, Andrii Nakryiko escreveu:
Quoted 18 lines · Jiri Olsa
Quoted 9 lines · Jiri Olsa
Quoted 4 lines · Andrii Nakryiko
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
-bpf_map__next;
-bpf_program__next;
-bpf_load_program;
-btf__get_from_id;these were added just to bypass the time window when they were not available in the package, so can be removed now (in the patch below)
Quoted 6 lines
Quoted 5 lines
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Quoted 8 lines
Yeah, that should be ok, the only consideration is that I'm submitting this today to Linus:
To address this:
ok, we can do that via bpf-next, but of course there's a problem ;-)
perf/core already has dependency commit [1]
so either we wait for perf/core and bpf-next/master to sync or:
- perf/core reverts [1] and
- bpf-next/master takes [1] and the rest
Quoted 8 lines
Quoted 1 line
I have the changes ready if you guys are ok with that
So, if I understand correctly, with merge window open bpf-next/master will get code from perf/core soon when we merge tip back in. So we can wait for that to happen and not revert anything.
So please add the below patch to your series and resend once tip is merged into bpf-next? Thanks!
ok
jirka
Arnaldo Carvalho de Melo
maintainerEm Tue, May 10, 2022 at 09:46:56AM +0200, Jiri Olsa escreveu:
Quoted 11 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_result
So, the first patch is already in torvalds/master, I tried applying the other two patches to my local perf/core, that already is merged with torvalds/master and:
[root@quaco ~]# perf test 42
42: BPF filter : 42.1: Basic BPF filtering : FAILED! 42.2: BPF pinning : FAILED! 42.3: BPF prologue generation : FAILED!
[root@quaco ~]#
I'll push my local perf/core to tmp.perf/core and continue tomorrow.
Its failing around here:
Open Debuginfo file: /root/.cache/debuginfod_client/e1c3de4b4c5db158f2098e80f2bf9140e8cfbdb6/debuginfo Try to find probe point from debuginfo.
Matched function: do_epoll_wait [3806bb5]
Probe point found: do_epoll_wait+0
Found 1 probe_trace_events.
Looking at the vmlinux_path (8 entries long)symsrc__init: build id mismatch for vmlinux.symsrc__init: cannot get elf header.
Using /proc/kcore for kernel data
Using /proc/kallsyms for symbolsdo_epoll_wait is out of .text, skip it.
Post processing failed or all events are skipped. (1)
Probe point 'do_epoll_wait' not found.bpf_probe: failed to convert perf probe events
Failed to add events selected by BPF
test child finished with -1
---- end ----
BPF filter subtest 1: FAILED
But:
[root@quaco ~]# grep do_epoll_wait /proc/kallsymsffffffff973c2a30 t do_epoll_wait
[root@quaco ~]#
- Arnaldo
Quoted 38 lines
v2 changes:
- use fallback section prog handler, so we don't need to use section prefix [Andrii]
- reallocprog->insnsarray inbpf_program__set_insns[Andrii]
- squash patch 1 from previous version withbpf_program__set_insnschange [Daniel]
- patch 3 already merged [Arnaldo]
- added more commentsmeanwhile.. perf/core and bpf-next diverged, so:
- libbpfbpf_program__set_insnschange is based on bpf-next/master
- perf changes do not apply on bpf-next/master so they are based on perf/core ... however they can be merged only after we release libbpf 0.8.0 withbpf_program__set_insnschange, so we don't break the dynamic linking
I'm sending perf changes now just for review, I'll resend them once libbpf 0.8.0 is releasedthanks,
jirka[1] https://lore.kernel.org/bpf/CAEf4BzaiBO3_617kkXZdYJ8hS8YF--ZLgapNbgeeEJ-pY0H88g@mail.gmail.com/
---
Jiri Olsa (1):
libbpf: Addbpf_program__set_insnsfunctiontools/lib/bpf/libbpf.c | 22 ++++++++++++++++++++++ tools/lib/bpf/libbpf.h | 18 ++++++++++++++++++ tools/lib/bpf/libbpf.map | 1 + 3 files changed, 41 insertions(+)Jiri Olsa (2):
perf tools: Register fallback libbpf section handler perf tools: Rework prologue generation codetools/perf/util/bpf-loader.c | 175 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 157 insertions(+), 18 deletions(-)
Signature · 1 line
- Arnaldo
Quoted 54 lines · Jiri Olsa
Quoted 52 lines · Andrii Nakryiko
Quoted 50 lines · Jiri Olsa
Quoted 19 lines · Arnaldo Carvalho de Melo
Em Fri, May 20, 2022 at 02:46:49PM -0700, Andrii Nakryiko escreveu:
Quoted 18 lines · Jiri Olsa
Quoted 9 lines · Jiri Olsa
Quoted 4 lines · Andrii Nakryiko
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
-bpf_map__next;
-bpf_program__next;
-bpf_load_program;
-btf__get_from_id;these were added just to bypass the time window when they were not available in the package, so can be removed now (in the patch below)
Quoted 6 lines
Quoted 5 lines
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Quoted 8 lines
Yeah, that should be ok, the only consideration is that I'm submitting this today to Linus:
To address this:
ok, we can do that via bpf-next, but of course there's a problem ;-)
perf/core already has dependency commit [1]
so either we wait for perf/core and bpf-next/master to sync or:
- perf/core reverts [1] and
- bpf-next/master takes [1] and the restI have the changes ready if you guys are ok with that
Quoted 8 lines
Quoted 6 lines
So, if I understand correctly, with merge window open bpf-next/master will get code from perf/core soon when we merge tip back in. So we can wait for that to happen and not revert anything.
So please add the below patch to your series and resend once tip is merged into bpf-next? Thanks!
ok
Hm.. Ok, so I don't see your patches in tip yet. I see them in perf/core only. Which means things won't happen naturally soon. How should we proceed? I'm sitting on a pile of patches removing a lot of code from libbpf and I'd rather get it out soon, but I can't because of them breaking perf in bpf-next without Jiri's changes.
Arnaldo, what's your suggestion? Can we land remaining Jiri's patches into perf/core and then you can create a tag for us to merge into bpf-next, so that we avoid any conflicts later? Would that work? I think we did something like that with other trees (e.g., RCU), when we had dependencies like this before.
Thoughts?
Quoted 2 lines
jirka
Quoted 62 lines · Andrii Nakryiko
Quoted 61 lines · Jiri Olsa
Quoted 58 lines · Andrii Nakryiko
Quoted 42 lines · Jiri Olsa
Quoted 19 lines · Arnaldo Carvalho de Melo
Em Fri, May 20, 2022 at 02:46:49PM -0700, Andrii Nakryiko escreveu:
Quoted 18 lines · Jiri Olsa
Quoted 9 lines · Jiri Olsa
Quoted 4 lines · Andrii Nakryiko
Quoted 1 line
Jiri, libbpf v0.8 is out, can you please re-send your perf patches?
yep, just made new fedora package.. will resend the perf changes soon
fedora package is on the way, but I'll need perf/core to merge the
bpf_program__set_insnschange.. Arnaldo, any idea when this could happen?Can we land these patches through bpf-next to avoid such complicated cross-tree dependencies? As I started removing libbpf APIs I also noticed that perf is still using few other deprecated APIs:
-bpf_map__next;
-bpf_program__next;
-bpf_load_program;
-btf__get_from_id;these were added just to bypass the time window when they were not available in the package, so can be removed now (in the patch below)
Quoted 6 lines
Quoted 5 lines
It's trivial to fix up, but doing it across few trees will delay libbpf work as well.
So let's land this through bpf-next, if Arnaldo doesn't mind?
Quoted 8 lines
Yeah, that should be ok, the only consideration is that I'm submitting this today to Linus:
To address this:
ok, we can do that via bpf-next, but of course there's a problem ;-)
Quoted 8 lines
perf/core already has dependency commit [1]
so either we wait for perf/core and bpf-next/master to sync or:
- perf/core reverts [1] and
- bpf-next/master takes [1] and the restI have the changes ready if you guys are ok with that
So, if I understand correctly, with merge window open bpf-next/master will get code from perf/core soon when we merge tip back in. So we can wait for that to happen and not revert anything.
So please add the below patch to your series and resend once tip is merged into bpf-next? Thanks!
Quoted 7 lines
Quoted 1 line
ok
Hm.. Ok, so I don't see your patches in tip yet. I see them in perf/core only. Which means things won't happen naturally soon. How should we proceed? I'm sitting on a pile of patches removing a lot of code from libbpf and I'd rather get it out soon, but I can't because of them breaking perf in bpf-next without Jiri's changes.
sorry it's merged in linus master, Arnaldo no longer goes through tip tree
Quoted 6 lines
Arnaldo, what's your suggestion? Can we land remaining Jiri's patches into perf/core and then you can create a tag for us to merge into bpf-next, so that we avoid any conflicts later? Would that work? I think we did something like that with other trees (e.g., RCU), when we had dependencies like this before.
either way is fine for me.. I just rebased those changes on top of perf/core
jirka
Quoted 17 lines · Arnaldo Carvalho de Melo
Em Tue, May 10, 2022 at 09:46:56AM +0200, Jiri Olsa escreveu:
Quoted 11 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultSo, the first patch is already in torvalds/master, I tried applying the other two patches to my local perf/core, that already is merged with torvalds/master and:
Quoted 8 lines
[root@quaco ~]# perf test 42
42: BPF filter : 42.1: Basic BPF filtering : FAILED! 42.2: BPF pinning : FAILED! 42.3: BPF prologue generation : FAILED![root@quaco ~]#
I'll push my local perf/core to tmp.perf/core and continue tomorrow.
hi,
I just rebased my changes on top of your perf/core and it seems to work:
[root@krava perf]# ./perf test bpf
40: LLVM search and compile :
40.1: Basic BPF llvm compile : Ok
40.3: Compile source for BPF prologue generation : Ok
40.4: Compile source for BPF relocation : Ok
42: BPF filter :
42.1: Basic BPF filtering : Ok
42.2: BPF pinning : Ok
42.3: BPF prologue generation : Ok
is it still a problem?
jirka
Quoted 41 lines · Jiri Olsa
Quoted 17 lines · Arnaldo Carvalho de Melo
Em Tue, May 10, 2022 at 09:46:56AM +0200, Jiri Olsa escreveu:
Quoted 11 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultSo, the first patch is already in torvalds/master, I tried applying the other two patches to my local perf/core, that already is merged with torvalds/master and:
Quoted 8 lines
[root@quaco ~]# perf test 42
42: BPF filter : 42.1: Basic BPF filtering : FAILED! 42.2: BPF pinning : FAILED! 42.3: BPF prologue generation : FAILED![root@quaco ~]#
I'll push my local perf/core to tmp.perf/core and continue tomorrow.
hi,
I just rebased my changes on top of your perf/core and it seems to work:[root@krava perf]# ./perf test bpf
40: LLVM search and compile :
40.1: Basic BPF llvm compile : Ok
40.3: Compile source for BPF prologue generation : Ok
40.4: Compile source for BPF relocation : Ok
42: BPF filter :
42.1: Basic BPF filtering : Ok
42.2: BPF pinning : Ok
42.3: BPF prologue generation : Ok
Quoted 2 lines
is it still a problem?
Ok, so I checked with Jakub, net-next will be forwarded to linus/master tomorrow or so, so after that bpf-next will get forwarded as well and we'll have all those patches of yours. So let's go back to plan A: send your perf changes based on bpf-next. Thanks and sorry for the extra noise with all the back and forth.
Quoted 1 line
jirka
Quoted 42 lines · Andrii Nakryiko
Quoted 41 lines · Jiri Olsa
Quoted 17 lines · Arnaldo Carvalho de Melo
Em Tue, May 10, 2022 at 09:46:56AM +0200, Jiri Olsa escreveu:
Quoted 11 lines
hi,
sending change we discussed some time ago [1] to get rid of some deprecated functions we use in perf prologue code.Despite the gloomy discussion I think the final code does not look that bad ;-)
This patchset removes following libbpf functions from perf:
bpf_program__set_prep
bpf_program__nth_fd
structbpf_prog_prep_resultSo, the first patch is already in torvalds/master, I tried applying the other two patches to my local perf/core, that already is merged with torvalds/master and:
Quoted 8 lines
[root@quaco ~]# perf test 42
42: BPF filter : 42.1: Basic BPF filtering : FAILED! 42.2: BPF pinning : FAILED! 42.3: BPF prologue generation : FAILED![root@quaco ~]#
I'll push my local perf/core to tmp.perf/core and continue tomorrow.
hi,
I just rebased my changes on top of your perf/core and it seems to work:[root@krava perf]# ./perf test bpf
40: LLVM search and compile :
40.1: Basic BPF llvm compile : Ok
40.3: Compile source for BPF prologue generation : Ok
40.4: Compile source for BPF relocation : Ok
42: BPF filter :
42.1: Basic BPF filtering : Ok
42.2: BPF pinning : Ok
42.3: BPF prologue generation : Ok
Quoted 8 lines
Quoted 2 lines
is it still a problem?
Ok, so I checked with Jakub, net-next will be forwarded to linus/master tomorrow or so, so after that bpf-next will get forwarded as well and we'll have all those patches of yours. So let's go back to plan A: send your perf changes based on bpf-next. Thanks and sorry for the extra noise with all the back and forth.
ok, np
jirka