bpf: Add load_time in bpf_prog fdinfo

[PATCH bpf-next] bpf: Add load_time in bpf_prog fdinfo

Posted by Tao Chen 3 months, 2 weeks ago

The field run_time_ns can tell us the run time of the bpf_prog,
and load_time_s can tell us how long the bpf_prog loaded on the
machine.

Signed-off-by: Tao Chen <chen.dylane@linux.dev>
---
 kernel/bpf/syscall.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 51ba1a7aa43..407841ea296 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -2438,6 +2438,7 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
 	const struct bpf_prog *prog = filp->private_data;
 	char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
 	struct bpf_prog_kstats stats;
+	u64 now = ktime_get_boottime_ns();
 
 	bpf_prog_get_stats(prog, &stats);
 	bin2hex(prog_tag, prog->tag, sizeof(prog->tag));
@@ -2450,7 +2451,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
 		   "run_time_ns:\t%llu\n"
 		   "run_cnt:\t%llu\n"
 		   "recursion_misses:\t%llu\n"
-		   "verified_insns:\t%u\n",
+		   "verified_insns:\t%u\n"
+		   "load_time_s:\t%llu\n",
 		   prog->type,
 		   prog->jited,
 		   prog_tag,
@@ -2459,7 +2461,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
 		   stats.nsecs,
 		   stats.cnt,
 		   stats.misses,
-		   prog->aux->verified_insns);
+		   prog->aux->verified_insns,
+		   (now - prog->aux->load_time) / NSEC_PER_SEC);
 }
 #endif
 
-- 
2.48.1

Re: [PATCH bpf-next] bpf: Add load_time in bpf_prog fdinfo

Posted by kernel test robot 3 months, 2 weeks ago

Hi Tao,

kernel test robot noticed the following build errors:

[auto build test ERROR on bpf-next/master]

url:    https://github.com/intel-lab-lkp/linux/commits/Tao-Chen/bpf-Add-load_time-in-bpf_prog-fdinfo/20250620-131249
base:   https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git master
patch link:    https://lore.kernel.org/r/20250620051017.111559-1-chen.dylane%40linux.dev
patch subject: [PATCH bpf-next] bpf: Add load_time in bpf_prog fdinfo
config: m68k-randconfig-r123-20250622 (https://download.01.org/0day-ci/archive/20250622/202506221641.xWzKiW3U-lkp@intel.com/config)
compiler: m68k-linux-gcc (GCC) 8.5.0
reproduce: (https://download.01.org/0day-ci/archive/20250622/202506221641.xWzKiW3U-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202506221641.xWzKiW3U-lkp@intel.com/

All errors (new ones prefixed by >>):

   m68k-linux-ld: kernel/bpf/syscall.o: in function `bpf_prog_show_fdinfo':
>> syscall.c:(.text+0x1088): undefined reference to `__udivdi3'
   m68k-linux-ld: drivers/clocksource/timer-tegra186.o: in function `tegra186_wdt_get_timeleft':
   timer-tegra186.c:(.text+0x130): undefined reference to `__udivdi3'

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH bpf-next] bpf: Add load_time in bpf_prog fdinfo

Posted by Alexei Starovoitov 3 months, 2 weeks ago

On Thu, Jun 19, 2025 at 10:10 PM Tao Chen <chen.dylane@linux.dev> wrote:
>
> The field run_time_ns can tell us the run time of the bpf_prog,
> and load_time_s can tell us how long the bpf_prog loaded on the
> machine.
>
> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
> ---
>  kernel/bpf/syscall.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> index 51ba1a7aa43..407841ea296 100644
> --- a/kernel/bpf/syscall.c
> +++ b/kernel/bpf/syscall.c
> @@ -2438,6 +2438,7 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
>         const struct bpf_prog *prog = filp->private_data;
>         char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
>         struct bpf_prog_kstats stats;
> +       u64 now = ktime_get_boottime_ns();
>
>         bpf_prog_get_stats(prog, &stats);
>         bin2hex(prog_tag, prog->tag, sizeof(prog->tag));
> @@ -2450,7 +2451,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
>                    "run_time_ns:\t%llu\n"
>                    "run_cnt:\t%llu\n"
>                    "recursion_misses:\t%llu\n"
> -                  "verified_insns:\t%u\n",
> +                  "verified_insns:\t%u\n"
> +                  "load_time_s:\t%llu\n",
>                    prog->type,
>                    prog->jited,
>                    prog_tag,
> @@ -2459,7 +2461,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
>                    stats.nsecs,
>                    stats.cnt,
>                    stats.misses,
> -                  prog->aux->verified_insns);
> +                  prog->aux->verified_insns,
> +                  (now - prog->aux->load_time) / NSEC_PER_SEC);

I don't like where it's going.
Soon fdinfo will be printing the xlated insns of the prog too,
since why not?
Let's stop here. We have syscall query api-s for that and
bpftool to print such things.
No more fdinfo "improvements".

Re: [PATCH bpf-next] bpf: Add load_time in bpf_prog fdinfo

Posted by Tao Chen 3 months, 2 weeks ago

在 2025/6/21 02:27, Alexei Starovoitov 写道:
> On Thu, Jun 19, 2025 at 10:10 PM Tao Chen <chen.dylane@linux.dev> wrote:
>>
>> The field run_time_ns can tell us the run time of the bpf_prog,
>> and load_time_s can tell us how long the bpf_prog loaded on the
>> machine.
>>
>> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
>> ---
>>   kernel/bpf/syscall.c | 7 +++++--
>>   1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> index 51ba1a7aa43..407841ea296 100644
>> --- a/kernel/bpf/syscall.c
>> +++ b/kernel/bpf/syscall.c
>> @@ -2438,6 +2438,7 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
>>          const struct bpf_prog *prog = filp->private_data;
>>          char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
>>          struct bpf_prog_kstats stats;
>> +       u64 now = ktime_get_boottime_ns();
>>
>>          bpf_prog_get_stats(prog, &stats);
>>          bin2hex(prog_tag, prog->tag, sizeof(prog->tag));
>> @@ -2450,7 +2451,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
>>                     "run_time_ns:\t%llu\n"
>>                     "run_cnt:\t%llu\n"
>>                     "recursion_misses:\t%llu\n"
>> -                  "verified_insns:\t%u\n",
>> +                  "verified_insns:\t%u\n"
>> +                  "load_time_s:\t%llu\n",
>>                     prog->type,
>>                     prog->jited,
>>                     prog_tag,
>> @@ -2459,7 +2461,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
>>                     stats.nsecs,
>>                     stats.cnt,
>>                     stats.misses,
>> -                  prog->aux->verified_insns);
>> +                  prog->aux->verified_insns,
>> +                  (now - prog->aux->load_time) / NSEC_PER_SEC);
> 
> I don't like where it's going.
> Soon fdinfo will be printing the xlated insns of the prog too,
> since why not?
> Let's stop here. We have syscall query api-s for that and
> bpftool to print such things.
> No more fdinfo "improvements".

Got it, Alexei, thanks for your patient explanation.

-- 
Best Regards
Tao Chen