include/linux/btf.h | 4 +++ kernel/bpf/btf.c | 56 +++++++++++-------------------------- kernel/trace/trace_btf.c | 15 +++++----- kernel/trace/trace_output.c | 8 ++---- 4 files changed, 31 insertions(+), 52 deletions(-)
From: Steven Rostedt <rostedt@goodmis.org>
Several functions need to call btf_put() on the btf pointer before it
returns leading to using "goto" branches to jump to the end to call
btf_put(btf). This can be simplified by introducing DEFINE_FREE() to allow
functions to define the btf descriptor with:
struct btf *btf __free(btf_put) = NULL;
Then the btf descriptor will always have btf_put() called on it if it
isn't NULL or ERR before exiting the function.
Where needed, no_free_ptr(btf) is used to assign the btf descriptor to a
pointer that will be used outside the function.
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
---
include/linux/btf.h | 4 +++
kernel/bpf/btf.c | 56 +++++++++++--------------------------
kernel/trace/trace_btf.c | 15 +++++-----
kernel/trace/trace_output.c | 8 ++----
4 files changed, 31 insertions(+), 52 deletions(-)
diff --git a/include/linux/btf.h b/include/linux/btf.h
index b2983706292f..e118c69c4996 100644
--- a/include/linux/btf.h
+++ b/include/linux/btf.h
@@ -8,6 +8,7 @@
#include <linux/bpfptr.h>
#include <linux/bsearch.h>
#include <linux/btf_ids.h>
+#include <linux/cleanup.h>
#include <uapi/linux/btf.h>
#include <uapi/linux/bpf.h>
@@ -150,6 +151,9 @@ struct btf *btf_get_by_fd(int fd);
int btf_get_info_by_fd(const struct btf *btf,
const union bpf_attr *attr,
union bpf_attr __user *uattr);
+
+DEFINE_FREE(btf_put, struct btf *, if (!IS_ERR_OR_NULL(_T)) btf_put(_T))
+
/* Figure out the size of a type_id. If type_id is a modifier
* (e.g. const), it will be resolved to find out the type with size.
*
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 1d2cf898e21e..480657912c96 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -3788,7 +3788,7 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
/* If a matching btf type is found in kernel or module BTFs, kptr_ref
* is that BTF, otherwise it's program BTF
*/
- struct btf *kptr_btf;
+ struct btf *kptr_btf __free(btf_put) = NULL;
int ret;
s32 id;
@@ -3827,23 +3827,17 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
* the same time.
*/
dtor_btf_id = btf_find_dtor_kfunc(kptr_btf, id);
- if (dtor_btf_id < 0) {
- ret = dtor_btf_id;
- goto end_btf;
- }
+ if (dtor_btf_id < 0)
+ return dtor_btf_id;
dtor_func = btf_type_by_id(kptr_btf, dtor_btf_id);
- if (!dtor_func) {
- ret = -ENOENT;
- goto end_btf;
- }
+ if (!dtor_func)
+ return -ENOENT;
if (btf_is_module(kptr_btf)) {
mod = btf_try_get_module(kptr_btf);
- if (!mod) {
- ret = -ENXIO;
- goto end_btf;
- }
+ if (!mod)
+ return -ENXIO;
}
/* We already verified dtor_func to be btf_type_is_func
@@ -3860,13 +3854,11 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
found_dtor:
field->kptr.btf_id = id;
- field->kptr.btf = kptr_btf;
+ field->kptr.btf = no_free_ptr(kptr_btf);
field->kptr.module = mod;
return 0;
end_mod:
module_put(mod);
-end_btf:
- btf_put(kptr_btf);
return ret;
}
@@ -8699,7 +8691,7 @@ u32 *btf_kfunc_is_modify_return(const struct btf *btf, u32 kfunc_btf_id,
static int __register_btf_kfunc_id_set(enum btf_kfunc_hook hook,
const struct btf_kfunc_id_set *kset)
{
- struct btf *btf;
+ struct btf *btf __free(btf_put) = NULL;
int ret, i;
btf = btf_get_module_btf(kset->owner);
@@ -8712,14 +8704,10 @@ static int __register_btf_kfunc_id_set(enum btf_kfunc_hook hook,
ret = btf_check_kfunc_protos(btf, btf_relocate_id(btf, kset->set->pairs[i].id),
kset->set->pairs[i].flags);
if (ret)
- goto err_out;
+ return ret;
}
- ret = btf_populate_kfunc_set(btf, hook, kset);
-
-err_out:
- btf_put(btf);
- return ret;
+ return btf_populate_kfunc_set(btf, hook, kset);
}
/* This function must be invoked only from initcalls/module init functions */
@@ -8807,7 +8795,7 @@ int register_btf_id_dtor_kfuncs(const struct btf_id_dtor_kfunc *dtors, u32 add_c
struct module *owner)
{
struct btf_id_dtor_kfunc_tab *tab;
- struct btf *btf;
+ struct btf *btf __free(btf_put) = NULL;
u32 tab_cnt, i;
int ret;
@@ -8873,7 +8861,6 @@ int register_btf_id_dtor_kfuncs(const struct btf_id_dtor_kfunc *dtors, u32 add_c
end:
if (ret)
btf_free_dtor_kfunc_tab(btf);
- btf_put(btf);
return ret;
}
EXPORT_SYMBOL_GPL(register_btf_id_dtor_kfuncs);
@@ -9484,9 +9471,8 @@ bpf_struct_ops_find(struct btf *btf, u32 type_id)
int __register_bpf_struct_ops(struct bpf_struct_ops *st_ops)
{
- struct bpf_verifier_log *log;
- struct btf *btf;
- int err = 0;
+ struct bpf_verifier_log *log __free(kfree) = NULL;
+ struct btf *btf __free(btf_put) = NULL;
btf = btf_get_module_btf(st_ops->owner);
if (!btf)
@@ -9495,20 +9481,12 @@ int __register_bpf_struct_ops(struct bpf_struct_ops *st_ops)
return PTR_ERR(btf);
log = kzalloc(sizeof(*log), GFP_KERNEL | __GFP_NOWARN);
- if (!log) {
- err = -ENOMEM;
- goto errout;
- }
+ if (!log)
+ return -ENOMEM;
log->level = BPF_LOG_KERNEL;
- err = btf_add_struct_ops(btf, st_ops, log);
-
-errout:
- kfree(log);
- btf_put(btf);
-
- return err;
+ return btf_add_struct_ops(btf, st_ops, log);
}
EXPORT_SYMBOL_GPL(__register_bpf_struct_ops);
#endif
diff --git a/kernel/trace/trace_btf.c b/kernel/trace/trace_btf.c
index 5bbdbcbbde3c..1c9111ab538b 100644
--- a/kernel/trace/trace_btf.c
+++ b/kernel/trace/trace_btf.c
@@ -13,26 +13,25 @@
const struct btf_type *btf_find_func_proto(const char *func_name, struct btf **btf_p)
{
const struct btf_type *t;
+ struct btf *btf __free(btf_put) = NULL;
s32 id;
- id = bpf_find_btf_id(func_name, BTF_KIND_FUNC, btf_p);
+ id = bpf_find_btf_id(func_name, BTF_KIND_FUNC, &btf);
if (id < 0)
return NULL;
/* Get BTF_KIND_FUNC type */
- t = btf_type_by_id(*btf_p, id);
+ t = btf_type_by_id(btf, id);
if (!t || !btf_type_is_func(t))
- goto err;
+ return NULL;
/* The type of BTF_KIND_FUNC is BTF_KIND_FUNC_PROTO */
- t = btf_type_by_id(*btf_p, t->type);
+ t = btf_type_by_id(btf, t->type);
if (!t || !btf_type_is_func_proto(t))
- goto err;
+ return NULL;
+ *btf_p = no_free_ptr(btf);
return t;
-err:
- btf_put(*btf_p);
- return NULL;
}
/*
diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index 0b3db02030a7..c11adfa83d5c 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -698,7 +698,7 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
const char *param_name;
char name[KSYM_NAME_LEN];
unsigned long arg;
- struct btf *btf;
+ struct btf *btf __free(btf_put) = NULL;
s32 tid, nr = 0;
int a, p, x;
@@ -716,7 +716,7 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
param = btf_get_func_param(t, &nr);
if (!param)
- goto out_put;
+ goto out;
for (a = 0, p = 0; p < nr; a++, p++) {
if (p)
@@ -756,7 +756,7 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
trace_seq_putc(s, ':');
if (++a == FTRACE_REGS_MAX_ARGS) {
trace_seq_puts(s, "...]");
- goto out_put;
+ goto out;
}
trace_seq_printf(s, "0x%lx", args[a]);
}
@@ -764,8 +764,6 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
break;
}
}
-out_put:
- btf_put(btf);
out:
trace_seq_printf(s, ")");
}
--
2.47.2On Fri, Aug 1, 2025 at 4:16 AM Steven Rostedt <rostedt@goodmis.org> wrote: > > From: Steven Rostedt <rostedt@goodmis.org> > > Several functions need to call btf_put() on the btf pointer before it > returns leading to using "goto" branches to jump to the end to call > btf_put(btf). This can be simplified by introducing DEFINE_FREE() to allow > functions to define the btf descriptor with: > > struct btf *btf __free(btf_put) = NULL; > > Then the btf descriptor will always have btf_put() called on it if it > isn't NULL or ERR before exiting the function. > > Where needed, no_free_ptr(btf) is used to assign the btf descriptor to a > pointer that will be used outside the function. > > Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> > --- > include/linux/btf.h | 4 +++ > kernel/bpf/btf.c | 56 +++++++++++-------------------------- > kernel/trace/trace_btf.c | 15 +++++----- > kernel/trace/trace_output.c | 8 ++---- > 4 files changed, 31 insertions(+), 52 deletions(-) > > diff --git a/include/linux/btf.h b/include/linux/btf.h > index b2983706292f..e118c69c4996 100644 > --- a/include/linux/btf.h > +++ b/include/linux/btf.h > @@ -8,6 +8,7 @@ > #include <linux/bpfptr.h> > #include <linux/bsearch.h> > #include <linux/btf_ids.h> > +#include <linux/cleanup.h> > #include <uapi/linux/btf.h> > #include <uapi/linux/bpf.h> > > @@ -150,6 +151,9 @@ struct btf *btf_get_by_fd(int fd); > int btf_get_info_by_fd(const struct btf *btf, > const union bpf_attr *attr, > union bpf_attr __user *uattr); > + > +DEFINE_FREE(btf_put, struct btf *, if (!IS_ERR_OR_NULL(_T)) btf_put(_T)) > + > /* Figure out the size of a type_id. If type_id is a modifier > * (e.g. const), it will be resolved to find out the type with size. > * > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c > index 1d2cf898e21e..480657912c96 100644 > --- a/kernel/bpf/btf.c > +++ b/kernel/bpf/btf.c > @@ -3788,7 +3788,7 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field, > /* If a matching btf type is found in kernel or module BTFs, kptr_ref > * is that BTF, otherwise it's program BTF > */ > - struct btf *kptr_btf; > + struct btf *kptr_btf __free(btf_put) = NULL; Sorry I hate this __free() style. It's not a simplification, but an obfuscation of code and logic. -- pw-bot: cr
On Fri, 1 Aug 2025 08:02:24 -0700 Alexei Starovoitov <alexei.starovoitov@gmail.com> wrote: > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c > > index 1d2cf898e21e..480657912c96 100644 > > --- a/kernel/bpf/btf.c > > +++ b/kernel/bpf/btf.c > > @@ -3788,7 +3788,7 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field, > > /* If a matching btf type is found in kernel or module BTFs, kptr_ref > > * is that BTF, otherwise it's program BTF > > */ > > - struct btf *kptr_btf; > > + struct btf *kptr_btf __free(btf_put) = NULL; > > Sorry I hate this __free() style. > It's not a simplification, but an obfuscation of code and logic. Well, it's becoming more common. But you are in control of this, so it's your decision. I wanted this just to get rid of the gotos in my code. -- Steve
On Fri, Aug 1, 2025 at 8:06 AM Steven Rostedt <rostedt@goodmis.org> wrote: > > On Fri, 1 Aug 2025 08:02:24 -0700 > Alexei Starovoitov <alexei.starovoitov@gmail.com> wrote: > > > > diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c > > > index 1d2cf898e21e..480657912c96 100644 > > > --- a/kernel/bpf/btf.c > > > +++ b/kernel/bpf/btf.c > > > @@ -3788,7 +3788,7 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field, > > > /* If a matching btf type is found in kernel or module BTFs, kptr_ref > > > * is that BTF, otherwise it's program BTF > > > */ > > > - struct btf *kptr_btf; > > > + struct btf *kptr_btf __free(btf_put) = NULL; > > > > Sorry I hate this __free() style. > > It's not a simplification, but an obfuscation of code and logic. > > Well, it's becoming more common. But you are in control of this, so it's > your decision. I wanted this just to get rid of the gotos in my code. You can use it in kernel/trace/trace_output.c, of course, but I really think it's a step back in maintainability. All this cleanup.h is not a silver bullet. It needs to be used sparingly.
On Fri, 1 Aug 2025 08:12:08 -0700 Alexei Starovoitov <alexei.starovoitov@gmail.com> wrote: > > You can use it in kernel/trace/trace_output.c, of course, I guess that means I should just use the DEFINE_FREE() in that file directly and not in the btf.h header file? > but I really think it's a step back in maintainability. > All this cleanup.h is not a silver bullet. It needs to be used sparingly. I have my reservations about the cleanup.h code too. But the more I use it, the more I like it. My biggest worry is guard() leak. That is, having a lock or interrupt/preemption disabled for longer than they need to be, because code was added at the end of the function after the protection is needed. -- Steve
On Fri, Aug 1, 2025 at 8:19 AM Steven Rostedt <rostedt@goodmis.org> wrote: > > On Fri, 1 Aug 2025 08:12:08 -0700 > Alexei Starovoitov <alexei.starovoitov@gmail.com> wrote: > > > > > You can use it in kernel/trace/trace_output.c, of course, > > I guess that means I should just use the DEFINE_FREE() in that file > directly and not in the btf.h header file? yes. I don't want this to proliferate and people spam us with this kind of "cleanups". > > but I really think it's a step back in maintainability. > > All this cleanup.h is not a silver bullet. It needs to be used sparingly. > > I have my reservations about the cleanup.h code too. But the more I use it, > the more I like it. My biggest worry is guard() leak. That is, having a > lock or interrupt/preemption disabled for longer than they need to be, > because code was added at the end of the function after the protection is > needed. guard() is ok. We use it too, but __free() is imo garbage. It's essence of what's wrong with C++
pá 1. 8. 2025 v 17:29 odesílatel Alexei Starovoitov <alexei.starovoitov@gmail.com> napsal: > > but __free() is imo garbage. It's essence of what's wrong with C++ > Here, you at least can read the beginning of the function though, and see that a free will be done at the end, like Go's defer, right?
On Fri, 1 Aug 2025 17:34:51 +0200 Tomas Glozar <tglozar@redhat.com> wrote: > pá 1. 8. 2025 v 17:29 odesílatel Alexei Starovoitov > <alexei.starovoitov@gmail.com> napsal: > > > > but __free() is imo garbage. It's essence of what's wrong with C++ > > > > Here, you at least can read the beginning of the function though, and > see that a free will be done at the end, like Go's defer, right? I don't know Go, but I'll assume "yes". I prefer the __free() over goto end; where you need to free up a temporary variable on error paths. But regardless. This will just be a difference of opinion, and I respect that Alexei doesn't want to use it in his code. -- Steve
pá 1. 8. 2025 v 17:56 odesílatel Steven Rostedt <rostedt@goodmis.org> napsal: > > But regardless. This will just be a difference of opinion, and I respect > that Alexei doesn't want to use it in his code. > Of course. I just wanted to point out this is (in my opinion not nearly) as bad as what C++ is doing. Tomas
On Fri, 1 Aug 2025 07:16:22 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:
> From: Steven Rostedt <rostedt@goodmis.org>
>
> Several functions need to call btf_put() on the btf pointer before it
> returns leading to using "goto" branches to jump to the end to call
> btf_put(btf). This can be simplified by introducing DEFINE_FREE() to allow
> functions to define the btf descriptor with:
>
> struct btf *btf __free(btf_put) = NULL;
>
> Then the btf descriptor will always have btf_put() called on it if it
> isn't NULL or ERR before exiting the function.
>
> Where needed, no_free_ptr(btf) is used to assign the btf descriptor to a
> pointer that will be used outside the function.
>
Yeah, this looks good to me.
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Thank you,
> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
> ---
> include/linux/btf.h | 4 +++
> kernel/bpf/btf.c | 56 +++++++++++--------------------------
> kernel/trace/trace_btf.c | 15 +++++-----
> kernel/trace/trace_output.c | 8 ++----
> 4 files changed, 31 insertions(+), 52 deletions(-)
>
> diff --git a/include/linux/btf.h b/include/linux/btf.h
> index b2983706292f..e118c69c4996 100644
> --- a/include/linux/btf.h
> +++ b/include/linux/btf.h
> @@ -8,6 +8,7 @@
> #include <linux/bpfptr.h>
> #include <linux/bsearch.h>
> #include <linux/btf_ids.h>
> +#include <linux/cleanup.h>
> #include <uapi/linux/btf.h>
> #include <uapi/linux/bpf.h>
>
> @@ -150,6 +151,9 @@ struct btf *btf_get_by_fd(int fd);
> int btf_get_info_by_fd(const struct btf *btf,
> const union bpf_attr *attr,
> union bpf_attr __user *uattr);
> +
> +DEFINE_FREE(btf_put, struct btf *, if (!IS_ERR_OR_NULL(_T)) btf_put(_T))
> +
> /* Figure out the size of a type_id. If type_id is a modifier
> * (e.g. const), it will be resolved to find out the type with size.
> *
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index 1d2cf898e21e..480657912c96 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -3788,7 +3788,7 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
> /* If a matching btf type is found in kernel or module BTFs, kptr_ref
> * is that BTF, otherwise it's program BTF
> */
> - struct btf *kptr_btf;
> + struct btf *kptr_btf __free(btf_put) = NULL;
> int ret;
> s32 id;
>
> @@ -3827,23 +3827,17 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
> * the same time.
> */
> dtor_btf_id = btf_find_dtor_kfunc(kptr_btf, id);
> - if (dtor_btf_id < 0) {
> - ret = dtor_btf_id;
> - goto end_btf;
> - }
> + if (dtor_btf_id < 0)
> + return dtor_btf_id;
>
> dtor_func = btf_type_by_id(kptr_btf, dtor_btf_id);
> - if (!dtor_func) {
> - ret = -ENOENT;
> - goto end_btf;
> - }
> + if (!dtor_func)
> + return -ENOENT;
>
> if (btf_is_module(kptr_btf)) {
> mod = btf_try_get_module(kptr_btf);
> - if (!mod) {
> - ret = -ENXIO;
> - goto end_btf;
> - }
> + if (!mod)
> + return -ENXIO;
> }
>
> /* We already verified dtor_func to be btf_type_is_func
> @@ -3860,13 +3854,11 @@ static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
>
> found_dtor:
> field->kptr.btf_id = id;
> - field->kptr.btf = kptr_btf;
> + field->kptr.btf = no_free_ptr(kptr_btf);
> field->kptr.module = mod;
> return 0;
> end_mod:
> module_put(mod);
> -end_btf:
> - btf_put(kptr_btf);
> return ret;
> }
>
> @@ -8699,7 +8691,7 @@ u32 *btf_kfunc_is_modify_return(const struct btf *btf, u32 kfunc_btf_id,
> static int __register_btf_kfunc_id_set(enum btf_kfunc_hook hook,
> const struct btf_kfunc_id_set *kset)
> {
> - struct btf *btf;
> + struct btf *btf __free(btf_put) = NULL;
> int ret, i;
>
> btf = btf_get_module_btf(kset->owner);
> @@ -8712,14 +8704,10 @@ static int __register_btf_kfunc_id_set(enum btf_kfunc_hook hook,
> ret = btf_check_kfunc_protos(btf, btf_relocate_id(btf, kset->set->pairs[i].id),
> kset->set->pairs[i].flags);
> if (ret)
> - goto err_out;
> + return ret;
> }
>
> - ret = btf_populate_kfunc_set(btf, hook, kset);
> -
> -err_out:
> - btf_put(btf);
> - return ret;
> + return btf_populate_kfunc_set(btf, hook, kset);
> }
>
> /* This function must be invoked only from initcalls/module init functions */
> @@ -8807,7 +8795,7 @@ int register_btf_id_dtor_kfuncs(const struct btf_id_dtor_kfunc *dtors, u32 add_c
> struct module *owner)
> {
> struct btf_id_dtor_kfunc_tab *tab;
> - struct btf *btf;
> + struct btf *btf __free(btf_put) = NULL;
> u32 tab_cnt, i;
> int ret;
>
> @@ -8873,7 +8861,6 @@ int register_btf_id_dtor_kfuncs(const struct btf_id_dtor_kfunc *dtors, u32 add_c
> end:
> if (ret)
> btf_free_dtor_kfunc_tab(btf);
> - btf_put(btf);
> return ret;
> }
> EXPORT_SYMBOL_GPL(register_btf_id_dtor_kfuncs);
> @@ -9484,9 +9471,8 @@ bpf_struct_ops_find(struct btf *btf, u32 type_id)
>
> int __register_bpf_struct_ops(struct bpf_struct_ops *st_ops)
> {
> - struct bpf_verifier_log *log;
> - struct btf *btf;
> - int err = 0;
> + struct bpf_verifier_log *log __free(kfree) = NULL;
> + struct btf *btf __free(btf_put) = NULL;
>
> btf = btf_get_module_btf(st_ops->owner);
> if (!btf)
> @@ -9495,20 +9481,12 @@ int __register_bpf_struct_ops(struct bpf_struct_ops *st_ops)
> return PTR_ERR(btf);
>
> log = kzalloc(sizeof(*log), GFP_KERNEL | __GFP_NOWARN);
> - if (!log) {
> - err = -ENOMEM;
> - goto errout;
> - }
> + if (!log)
> + return -ENOMEM;
>
> log->level = BPF_LOG_KERNEL;
>
> - err = btf_add_struct_ops(btf, st_ops, log);
> -
> -errout:
> - kfree(log);
> - btf_put(btf);
> -
> - return err;
> + return btf_add_struct_ops(btf, st_ops, log);
> }
> EXPORT_SYMBOL_GPL(__register_bpf_struct_ops);
> #endif
> diff --git a/kernel/trace/trace_btf.c b/kernel/trace/trace_btf.c
> index 5bbdbcbbde3c..1c9111ab538b 100644
> --- a/kernel/trace/trace_btf.c
> +++ b/kernel/trace/trace_btf.c
> @@ -13,26 +13,25 @@
> const struct btf_type *btf_find_func_proto(const char *func_name, struct btf **btf_p)
> {
> const struct btf_type *t;
> + struct btf *btf __free(btf_put) = NULL;
> s32 id;
>
> - id = bpf_find_btf_id(func_name, BTF_KIND_FUNC, btf_p);
> + id = bpf_find_btf_id(func_name, BTF_KIND_FUNC, &btf);
> if (id < 0)
> return NULL;
>
> /* Get BTF_KIND_FUNC type */
> - t = btf_type_by_id(*btf_p, id);
> + t = btf_type_by_id(btf, id);
> if (!t || !btf_type_is_func(t))
> - goto err;
> + return NULL;
>
> /* The type of BTF_KIND_FUNC is BTF_KIND_FUNC_PROTO */
> - t = btf_type_by_id(*btf_p, t->type);
> + t = btf_type_by_id(btf, t->type);
> if (!t || !btf_type_is_func_proto(t))
> - goto err;
> + return NULL;
>
> + *btf_p = no_free_ptr(btf);
> return t;
> -err:
> - btf_put(*btf_p);
> - return NULL;
> }
>
> /*
> diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
> index 0b3db02030a7..c11adfa83d5c 100644
> --- a/kernel/trace/trace_output.c
> +++ b/kernel/trace/trace_output.c
> @@ -698,7 +698,7 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
> const char *param_name;
> char name[KSYM_NAME_LEN];
> unsigned long arg;
> - struct btf *btf;
> + struct btf *btf __free(btf_put) = NULL;
> s32 tid, nr = 0;
> int a, p, x;
>
> @@ -716,7 +716,7 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
>
> param = btf_get_func_param(t, &nr);
> if (!param)
> - goto out_put;
> + goto out;
>
> for (a = 0, p = 0; p < nr; a++, p++) {
> if (p)
> @@ -756,7 +756,7 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
> trace_seq_putc(s, ':');
> if (++a == FTRACE_REGS_MAX_ARGS) {
> trace_seq_puts(s, "...]");
> - goto out_put;
> + goto out;
> }
> trace_seq_printf(s, "0x%lx", args[a]);
> }
> @@ -764,8 +764,6 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
> break;
> }
> }
> -out_put:
> - btf_put(btf);
> out:
> trace_seq_printf(s, ")");
> }
> --
> 2.47.2
>
--
Masami Hiramatsu (Google) <mhiramat@kernel.org>
On Mon, 4 Aug 2025 22:26:45 +0900 Masami Hiramatsu (Google) <mhiramat@kernel.org> wrote: > On Fri, 1 Aug 2025 07:16:22 -0400 > Steven Rostedt <rostedt@goodmis.org> wrote: > > > From: Steven Rostedt <rostedt@goodmis.org> > > > > Several functions need to call btf_put() on the btf pointer before it > > returns leading to using "goto" branches to jump to the end to call > > btf_put(btf). This can be simplified by introducing DEFINE_FREE() to allow > > functions to define the btf descriptor with: > > > > struct btf *btf __free(btf_put) = NULL; > > > > Then the btf descriptor will always have btf_put() called on it if it > > isn't NULL or ERR before exiting the function. > > > > Where needed, no_free_ptr(btf) is used to assign the btf descriptor to a > > pointer that will be used outside the function. > > > > Yeah, this looks good to me. > > Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org> Thanks but Alexei already declined it. I'm gonna respin where it is only used in the tracing code. Thanks! -- Steve
© 2016 - 2025 Red Hat, Inc.