On Tue, Aug 09, 2016 at 05:23:50PM -0700, Alexei Starovoitov wrote:
> On Tue, Aug 09, 2016 at 05:00:12PM -0700, Sargun Dhillon wrote:
> > This adds a bpf helper that's similar to the skb_in_cgroup helper to check
> > whether the probe is currently executing in the context of a specific
> > subset of the cgroupsv2 hierarchy. It does this based on membership test
> > for a cgroup arraymap. It is invalid to call this in an interrupt, and
> > it'll return an error. The helper is primarily to be used in debugging
> > activities for containers, where you may have multiple programs running in
> > a given top-level "container".
> >
> > This patch also genericizes some of the arraymap fetching logic between the
> > skb_in_cgroup helper and this new helper.
> >
> > Signed-off-by: Sargun Dhillon <sar...@sargun.me>
> > Cc: Alexei Starovoitov <a...@kernel.org>
> > Cc: Daniel Borkmann <dan...@iogearbox.net>
> > ---
> > include/linux/bpf.h | 24 ++++++++++++++++++++++++
> > include/uapi/linux/bpf.h | 11 +++++++++++
> > kernel/bpf/arraymap.c | 2 +-
> > kernel/bpf/verifier.c | 4 +++-
> > kernel/trace/bpf_trace.c | 34 ++++++++++++++++++++++++++++++++++
> > net/core/filter.c | 11 ++++-------
> > 6 files changed, 77 insertions(+), 9 deletions(-)
> >
> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index 1113423..9adf712 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -319,4 +319,28 @@ extern const struct bpf_func_proto
> > bpf_get_stackid_proto;
> > void bpf_user_rnd_init_once(void);
> > u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
> >
> > +#ifdef CONFIG_CGROUPS
> > +/* Helper to fetch a cgroup pointer based on index.
> > + * @map: a cgroup arraymap
> > + * @idx: index of the item you want to fetch
> > + *
> > + * Returns pointer on success,
> > + * Error code if item not found, or out-of-bounds access
> > + */
> > +static inline struct cgroup *fetch_arraymap_ptr(struct bpf_map *map, int
> > idx)
> > +{
> > + struct cgroup *cgrp;
> > + struct bpf_array *array = container_of(map, struct bpf_array, map);
> > +
> > + if (unlikely(idx >= array->map.max_entries))
> > + return ERR_PTR(-E2BIG);
> > +
> > + cgrp = READ_ONCE(array->ptrs[idx]);
> > + if (unlikely(!cgrp))
> > + return ERR_PTR(-EAGAIN);
> > +
> > + return cgrp;
> > +}
> > +#endif /* CONFIG_CGROUPS */
> > +
> > #endif /* _LINUX_BPF_H */
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index da218fe..64b1a07 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -375,6 +375,17 @@ enum bpf_func_id {
> > */
> > BPF_FUNC_probe_write_user,
> >
> > + /**
> > + * bpf_current_task_in_cgroup(map, index) - Check cgroup2 membership of
> > current task
> > + * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
> > + * @index: index of the cgroup in the bpf_map
> > + * Return:
> > + * == 0 current failed the cgroup2 descendant test
> > + * == 1 current succeeded the cgroup2 descendant test
> > + * < 0 error
> > + */
> > + BPF_FUNC_current_task_in_cgroup,
> > +
> > __BPF_FUNC_MAX_ID,
> > };
> >
> > diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c
> > index 633a650..a2ac051 100644
> > --- a/kernel/bpf/arraymap.c
> > +++ b/kernel/bpf/arraymap.c
> > @@ -538,7 +538,7 @@ static int __init register_perf_event_array_map(void)
> > }
> > late_initcall(register_perf_event_array_map);
> >
> > -#ifdef CONFIG_SOCK_CGROUP_DATA
> > +#ifdef CONFIG_CGROUPS
> > static void *cgroup_fd_array_get_ptr(struct bpf_map *map,
> > struct file *map_file /* not used */,
> > int fd)
> > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > index 7094c69..80efab8 100644
> > --- a/kernel/bpf/verifier.c
> > +++ b/kernel/bpf/verifier.c
> > @@ -1053,7 +1053,8 @@ static int check_map_func_compatibility(struct
> > bpf_map *map, int func_id)
> > goto error;
> > break;
> > case BPF_MAP_TYPE_CGROUP_ARRAY:
> > - if (func_id != BPF_FUNC_skb_in_cgroup)
> > + if (func_id != BPF_FUNC_skb_in_cgroup &&
> > + func_id != BPF_FUNC_current_task_in_cgroup)
> > goto error;
> > break;
> > default:
> > @@ -1075,6 +1076,7 @@ static int check_map_func_compatibility(struct
> > bpf_map *map, int func_id)
> > if (map->map_type != BPF_MAP_TYPE_STACK_TRACE)
> > goto error;
> > break;
> > + case BPF_FUNC_current_task_in_cgroup:
> > case BPF_FUNC_skb_in_cgroup:
> > if (map->map_type != BPF_MAP_TYPE_CGROUP_ARRAY)
> > goto error;
> > diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> > index b20438f..39f0290 100644
> > --- a/kernel/trace/bpf_trace.c
> > +++ b/kernel/trace/bpf_trace.c
> > @@ -376,6 +376,36 @@ static const struct bpf_func_proto
> > bpf_get_current_task_proto = {
> > .ret_type = RET_INTEGER,
> > };
> >
> > +#ifdef CONFIG_CGROUPS
> > +static u64 bpf_current_task_in_cgroup(u64 r1, u64 r2, u64 r3, u64 r4, u64
> > r5)
>
> please don't introduce #ifdef into .c code.
> In this case add #else in .h to fetch_arraymap_ptr() that returns -EOPNOTSUPP.
> Also why guard it with CONFIG_CGROUPS in .h at all?
> I think it should compile fine even when cgroups are not defined.
> The helper won't be functional anyway, since no cgroup_fd can be added
> to cgroup map.
>
Pardon my ignorance, but if I don't have an ifdef wont I be referencing a bunch
of non-existent code (like cgroup_is_descendant)? Unless your suggestion is to
move this to the .h file as well.
> > +{
> > + struct bpf_map *map = (struct bpf_map *)(long)r1;
> > + struct css_set *cset;
> > + struct cgroup *cgrp;
> > + u32 idx = (u32)r2;
> > +
> > + if (unlikely(in_interrupt()))
> > + return -EINVAL;
> > +
> > + cgrp = fetch_arraymap_ptr(map, idx);
> > +
> > + if (unlikely(IS_ERR(cgrp)))
> > + return PTR_ERR(cgrp);
> > +
> > + cset = task_css_set(current);
> > +
> > + return cgroup_is_descendant(cset->dfl_cgrp, cgrp);
> > +}
> > +
> > +static const struct bpf_func_proto bpf_current_task_in_cgroup_proto = {
> > + .func = bpf_current_task_in_cgroup,
> > + .gpl_only = false,
> > + .ret_type = RET_INTEGER,
> > + .arg1_type = ARG_CONST_MAP_PTR,
> > + .arg2_type = ARG_ANYTHING,
> > +};
> > +#endif /* CONFIG_CGROUPS */
> > +
> > static const struct bpf_func_proto *tracing_func_proto(enum bpf_func_id
> > func_id)
> > {
> > switch (func_id) {
> > @@ -407,6 +437,10 @@ static const struct bpf_func_proto
> > *tracing_func_proto(enum bpf_func_id func_id)
> > return &bpf_perf_event_read_proto;
> > case BPF_FUNC_probe_write_user:
> > return bpf_get_probe_write_proto();
> > +#ifdef CONFIG_CGROUPS
> > + case BPF_FUNC_current_task_in_cgroup:
> > + return &bpf_current_task_in_cgroup_proto;
> > +#endif
>
> same here. looks unnecessary and #ifdef in .c are frowned upon.
>
