Re: [PATCH bpf-next v5 4/8] bpf: Introduce cgroup iter

From: Hao Luo
Date: Tue Aug 02 2022 - 18:28:17 EST

Hi Andrii,

On Mon, Aug 1, 2022 at 8:43 PM Andrii Nakryiko
<andrii.nakryiko@xxxxxxxxx> wrote:
>
> On Fri, Jul 22, 2022 at 10:48 AM Yosry Ahmed <yosryahmed@xxxxxxxxxx> wrote:
> >
> > From: Hao Luo <haoluo@xxxxxxxxxx>
> >
> > Cgroup_iter is a type of bpf_iter. It walks over cgroups in three modes:
> >
> > - walking a cgroup's descendants in pre-order.
> > - walking a cgroup's descendants in post-order.
> > - walking a cgroup's ancestors.
> >
> > When attaching cgroup_iter, one can set a cgroup to the iter_link
> > created from attaching. This cgroup is passed as a file descriptor and
> > serves as the starting point of the walk. If no cgroup is specified,
> > the starting point will be the root cgroup.
> >
> > For walking descendants, one can specify the order: either pre-order or
> > post-order. For walking ancestors, the walk starts at the specified
> > cgroup and ends at the root.
> >
> > One can also terminate the walk early by returning 1 from the iter
> > program.
> >
> > Note that because walking cgroup hierarchy holds cgroup_mutex, the iter
> > program is called with cgroup_mutex held.
> >
> > Currently only one session is supported, which means, depending on the
> > volume of data bpf program intends to send to user space, the number
> > of cgroups that can be walked is limited. For example, given the current
> > buffer size is 8 * PAGE_SIZE, if the program sends 64B data for each
> > cgroup, the total number of cgroups that can be walked is 512. This is
> > a limitation of cgroup_iter. If the output data is larger than the
> > buffer size, the second read() will signal EOPNOTSUPP. In order to work
> > around, the user may have to update their program to reduce the volume
> > of data sent to output. For example, skip some uninteresting cgroups.
> > In future, we may extend bpf_iter flags to allow customizing buffer
> > size.
> >
> > Signed-off-by: Hao Luo <haoluo@xxxxxxxxxx>
> > Signed-off-by: Yosry Ahmed <yosryahmed@xxxxxxxxxx>
> > Acked-by: Yonghong Song <yhs@xxxxxx>
> > ---
> > include/linux/bpf.h | 8 +
> > include/uapi/linux/bpf.h | 30 +++
> > kernel/bpf/Makefile | 3 +
> > kernel/bpf/cgroup_iter.c | 252 ++++++++++++++++++
> > tools/include/uapi/linux/bpf.h | 30 +++
> > .../selftests/bpf/prog_tests/btf_dump.c | 4 +-
> > 6 files changed, 325 insertions(+), 2 deletions(-)
> > create mode 100644 kernel/bpf/cgroup_iter.c
> >
> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index a97751d845c9..9061618fe929 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -47,6 +47,7 @@ struct kobject;
> > struct mem_cgroup;
> > struct module;
> > struct bpf_func_state;
> > +struct cgroup;
> >
> > extern struct idr btf_idr;
> > extern spinlock_t btf_idr_lock;
> > @@ -1717,7 +1718,14 @@ int bpf_obj_get_user(const char __user *pathname, int flags);
> > int __init bpf_iter_ ## target(args) { return 0; }
> >
> > struct bpf_iter_aux_info {
> > + /* for map_elem iter */
> > struct bpf_map *map;
> > +
> > + /* for cgroup iter */
> > + struct {
> > + struct cgroup *start; /* starting cgroup */
> > + int order;
> > + } cgroup;
> > };
> >
> > typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index ffcbf79a556b..fe50c2489350 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -87,10 +87,30 @@ struct bpf_cgroup_storage_key {
> > __u32 attach_type; /* program attach type (enum bpf_attach_type) */
> > };
> >
> > +enum bpf_iter_cgroup_traversal_order {
> > + BPF_ITER_CGROUP_PRE = 0, /* pre-order traversal */
> > + BPF_ITER_CGROUP_POST, /* post-order traversal */
> > + BPF_ITER_CGROUP_PARENT_UP, /* traversal of ancestors up to the root */
>
> I've just put up my arguments why it's a good idea to also support a
> "trivial" mode of only traversing specified cgroup and no descendants
> or parents. Please see [0].

cc Kui-Feng in this thread.

Yeah, I think it's a good idea. It's useful when we only want to show
a single object, which can be common. Going further, I think we may
want to restructure bpf_iter to optimize for this case.

> I think the same applies here, especially
> considering that it seems like a good idea to support
> task/task_vma/task_files iteration within a cgroup.

I have reservations on these use cases. I don't see immediate use of
iterating vma or files within a cgroup. Tasks within a cgroup? Maybe.
:)

> So depending on
> how successful I am in arguing for supporting task iterator with
> target cgroup, I think we should reuse *exactly* this
> bpf_iter_cgroup_traversal_order and how we specify cgroup (FD or ID,
> see some more below) *as is* in task iterators as well. In the latter
> case, having an ability to say "iterate task for only given cgroup" is
> very useful, and for such mode all the PRE/POST/PARENT_UP is just an
> unnecessary nuisance.
>
> So please consider also adding and supporting BPF_ITER_CGROUP_SELF (or
> whatever naming makes most sense).
>

PRE/POST/UP can be reused for iter of tree-structured containers, like
rbtree [1]. SELF can be reused for any iters like iter/task,
iter/cgroup, etc. Promoting all of them out of cgroup-specific struct
seems valuable.

[1] https://lwn.net/Articles/902405/

>
> Some more naming nits. I find BPF_ITER_CGROUP_PRE and
> BPF_ITER_CGROUP_POST a bit confusing. Even internally in kernel we
> have css_next_descendant_pre/css_next_descendant_post, so why not
> reflect the fact that we are going to iterate descendants:
> BPF_ITER_CGROUP_DESCENDANTS_{PRE,POST}. And now that we use
> "descendants" terminology, PARENT_UP should be ANCESTORS. ANCESTORS_UP
> probably is fine, but seems a bit redundant (unless we consider a
> somewhat weird ANCESTORS_DOWN, where we find the furthest parent and
> then descend through preceding parents until we reach specified
> cgroup; seems a bit exotic).
>

BPF_ITER_CGROUP_DESCENDANTS_PRE is too verbose. If there is a
possibility of merging rbtree and supporting walk order of rbtree
iter, maybe the name here could be general, like
BPF_ITER_DESCENDANTS_PRE, which seems better.

> [0] https://lore.kernel.org/bpf/f92e20e9961963e20766e290ee6668edd4bacf06.camel@xxxxxx/T/#m5ce50632aa550dd87a99241efb168cbcde1ee98f
>
> > +};
> > +
> > union bpf_iter_link_info {
> > struct {
> > __u32 map_fd;
> > } map;
> > +
> > + /* cgroup_iter walks either the live descendants of a cgroup subtree, or the
> > + * ancestors of a given cgroup.
> > + */
> > + struct {
> > + /* Cgroup file descriptor. This is root of the subtree if walking
> > + * descendants; it's the starting cgroup if walking the ancestors.
> > + * If it is left 0, the traversal starts from the default cgroup v2
> > + * root. For walking v1 hierarchy, one should always explicitly
> > + * specify the cgroup_fd.
> > + */
> > + __u32 cgroup_fd;
>
> Now, similar to what I argued in regard of pidfd vs pid, I think the
> same applied to cgroup_fd vs cgroup_id. Why can't we support both?
> cgroup_fd has some benefits, but cgroup_id is nice due to simplicity
> and not having to open/close/keep extra FDs (which can add up if we
> want to periodically query something about a large set of cgroups).
> Please see my arguments from [0] above.
>
> Thoughts?
>

We can support both, it's a good idea IMO. But what exactly is the
interface going to look like? Can you be more specific about that?
Below is something I tried based on your description.

@@ -91,6 +91,18 @@ union bpf_iter_link_info {
struct {
__u32 map_fd;
} map;
+ struct {
+ /* PRE/POST/UP/SELF */
+ __u32 order;
+ struct {
+ __u32 cgroup_fd;
+ __u64 cgroup_id;
+ } cgroup;
+ struct {
+ __u32 pid_fd;
+ __u64 pid;
+ } task;
+ };
};

> > + __u32 traversal_order;
> > + } cgroup;
> > };
> >
> > /* BPF syscall commands, see bpf(2) man-page for more details. */
>
> [...]