Re: [RFC v2 PATCH bpf-next 1/4] bpf: add percpu stats for bpf_map elements insertions/deletions

[Anton Protopopov <aspsk@isovalent.com>]

On Thu, Jun 22, 2023 at 01:11:58PM -0700, Alexei Starovoitov wrote:
> On Thu, Jun 22, 2023 at 09:53:27AM +0000, Anton Protopopov wrote:
> > Add a generic percpu stats for bpf_map elements insertions/deletions in order
> > to keep track of both, the current (approximate) number of elements in a map
> > and per-cpu statistics on update/delete operations.
> > 
> > To expose these stats a particular map implementation should initialize the
> > counter and adjust it as needed using the 'bpf_map_*_elements_counter' helpers
> > provided by this commit. The counter can be read by an iterator program.
> > 
> > A bpf_map_sum_elements_counter kfunc was added to simplify getting the sum of
> > the per-cpu values. If a map doesn't implement the counter, then it will always
> > return 0.
> > 
> > Signed-off-by: Anton Protopopov <aspsk@isovalent.com>
> > ---
> >  include/linux/bpf.h   | 30 +++++++++++++++++++++++++++
> >  kernel/bpf/map_iter.c | 48 ++++++++++++++++++++++++++++++++++++++++++-
> >  2 files changed, 77 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index f58895830ada..20292a096188 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -275,6 +275,7 @@ struct bpf_map {
> >  	} owner;
> >  	bool bypass_spec_v1;
> >  	bool frozen; /* write-once; write-protected by freeze_mutex */
> > +	s64 __percpu *elements_count;
> >  };
> >  
> >  static inline const char *btf_field_type_name(enum btf_field_type type)
> > @@ -2040,6 +2041,35 @@ bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
> >  }
> >  #endif
> >  
> > +static inline int
> > +bpf_map_init_elements_counter(struct bpf_map *map)
> > +{
> > +	size_t size = sizeof(*map->elements_count), align = size;
> > +	gfp_t flags = GFP_USER | __GFP_NOWARN;
> > +
> > +	map->elements_count = bpf_map_alloc_percpu(map, size, align, flags);
> > +	if (!map->elements_count)
> > +		return -ENOMEM;
> > +
> > +	return 0;
> > +}
> > +
> > +static inline void
> > +bpf_map_free_elements_counter(struct bpf_map *map)
> > +{
> > +	free_percpu(map->elements_count);
> > +}
> > +
> > +static inline void bpf_map_inc_elements_counter(struct bpf_map *map)
> 
> bpf_map_inc_elem_count() to match existing inc_elem_count() ?
> 
> > +{
> > +	this_cpu_inc(*map->elements_count);
> > +}
> > +
> > +static inline void bpf_map_dec_elements_counter(struct bpf_map *map)
> > +{
> > +	this_cpu_dec(*map->elements_count);
> > +}
> > +
> >  extern int sysctl_unprivileged_bpf_disabled;
> >  
> >  static inline bool bpf_allow_ptr_leaks(void)
> > diff --git a/kernel/bpf/map_iter.c b/kernel/bpf/map_iter.c
> > index b0fa190b0979..26ca00dde962 100644
> > --- a/kernel/bpf/map_iter.c
> > +++ b/kernel/bpf/map_iter.c
> > @@ -93,7 +93,7 @@ static struct bpf_iter_reg bpf_map_reg_info = {
> >  	.ctx_arg_info_size	= 1,
> >  	.ctx_arg_info		= {
> >  		{ offsetof(struct bpf_iter__bpf_map, map),
> > -		  PTR_TO_BTF_ID_OR_NULL },
> > +		  PTR_TO_BTF_ID_OR_NULL | PTR_TRUSTED },
> 
> this and below should be in separate patch.
> 
> >  	},
> >  	.seq_info		= &bpf_map_seq_info,
> >  };
> > @@ -193,3 +193,49 @@ static int __init bpf_map_iter_init(void)
> >  }
> >  
> >  late_initcall(bpf_map_iter_init);
> > +
> > +__diag_push();
> > +__diag_ignore_all("-Wmissing-prototypes",
> > +		  "Global functions as their definitions will be in vmlinux BTF");
> > +
> > +__bpf_kfunc s64 bpf_map_sum_elements_counter(struct bpf_map *map)
> > +{
> > +	s64 *pcount;
> > +	s64 ret = 0;
> > +	int cpu;
> > +
> > +	if (!map || !map->elements_count)
> > +		return 0;
> > +
> > +	for_each_possible_cpu(cpu) {
> > +		pcount = per_cpu_ptr(map->elements_count, cpu);
> > +		ret += READ_ONCE(*pcount);
> > +	}
> > +	return ret;
> > +}
> > +
> > +__diag_pop();
> > +
> > +BTF_SET8_START(bpf_map_iter_kfunc_ids)
> > +BTF_ID_FLAGS(func, bpf_map_sum_elements_counter, KF_TRUSTED_ARGS)
> > +BTF_SET8_END(bpf_map_iter_kfunc_ids)
> > +
> > +static int tracing_iter_filter(const struct bpf_prog *prog, u32 kfunc_id)
> > +{
> > +	if (btf_id_set8_contains(&bpf_map_iter_kfunc_ids, kfunc_id) &&
> > +	    prog->expected_attach_type != BPF_TRACE_ITER)
> 
> why restrict to trace_iter?

Thanks, I will remove it.

All your other comments in this series make sense as well, will address them.

> > +		return -EACCES;
> > +	return 0;
> > +}
> > +
> > +static const struct btf_kfunc_id_set bpf_map_iter_kfunc_set = {
> > +	.owner = THIS_MODULE,
> > +	.set   = &bpf_map_iter_kfunc_ids,
> > +	.filter = tracing_iter_filter,
> > +};
> > +
> > +static int init_subsystem(void)
> > +{
> > +	return register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &bpf_map_iter_kfunc_set);
> > +}
> > +late_initcall(init_subsystem);
> > -- 
> > 2.34.1
> >