[PATCH bpf-next v2 1/4] bpf: Generalize bpf_sk_storage

[Martin KaFai Lau]

On Mon, Jun 29, 2020 at 06:01:00PM +0200, KP Singh wrote:
> Thanks for your feedback! Apologies it took some time for me
> to incorporate this into another revision.
> 
> On 18-Jun 23:43, Martin KaFai Lau wrote:
> > On Wed, Jun 17, 2020 at 10:29:38PM +0200, KP Singh wrote:
> > > From: KP Singh <kpsingh at google.com>
> > > 
> > > Refactor the functionality in bpf_sk_storage.c so that concept of
> > > storage linked to kernel objects can be extended to other objects like
> > > inode, task_struct etc.
> > > 
> > > bpf_sk_storage is updated to be bpf_local_storage with a union that
> > > contains a pointer to the owner object. The type of the
> > > bpf_local_storage can be determined using the newly added
> > > bpf_local_storage_type enum.
> > > 
> > > Each new local storage will still be a separate map and provide its own
> > > set of helpers. This allows for future object specific extensions and
> > > still share a lot of the underlying implementation.
> > Thanks for taking up this effort to refactor sk_local_storage.
> > 
> > I took a quick look.  I have some comments and would like to explore
> > some thoughts.
> > 
> > > --- a/net/core/bpf_sk_storage.c
> > > +++ b/kernel/bpf/bpf_local_storage.c
> > > @@ -1,19 +1,22 @@
> > >  // SPDX-License-Identifier: GPL-2.0
> > >  /* Copyright (c) 2019 Facebook  */
> > > +#include "linux/bpf.h"
> > > +#include "asm-generic/bug.h"
> > > +#include "linux/err.h"
> > "<" ">"
> > 
> > >  #include <linux/rculist.h>
> > >  #include <linux/list.h>
> > >  #include <linux/hash.h>
> > >  #include <linux/types.h>
> > >  #include <linux/spinlock.h>
> > >  #include <linux/bpf.h>
> > > -#include <net/bpf_sk_storage.h>
> > > +#include <linux/bpf_local_storage.h>
> > >  #include <net/sock.h>
> > >  #include <uapi/linux/sock_diag.h>
> > >  #include <uapi/linux/btf.h>
> > >  
> > >  static atomic_t cache_idx;
> > inode local storage and sk local storage probably need a separate
> > cache_idx.  An improvement on picking cache_idx has just been
> > landed also.
> 
> I see, thanks! I rebased and I now see that cache_idx is now a:
> 
>   static u64 cache_idx_usage_counts[BPF_STORAGE_CACHE_SIZE];
> 
> which tracks the free cache slots rather than using a single atomic
> cache_idx. I guess all types of local storage can share this now
> right?
I believe they have to be separated.  A sk-storage will not be cached/stored
in inode.  Caching a sk-storage at idx=0 of a sk should not stop
an inode-storage to be cached at the same idx of a inode.

> 
> > 
> > [ ... ]
> > 
> > > +struct bpf_local_storage {
> > > +	struct bpf_local_storage_data __rcu *cache[BPF_STORAGE_CACHE_SIZE];
> > >  	return NULL;
> 
> [...]
> 
> > >  }
> > >  
> > > -/* sk_storage->lock must be held and selem->sk_storage == sk_storage.
> > > +static void __unlink_local_storage(struct bpf_local_storage *local_storage,
> > > +				   bool uncharge_omem)
> > Nit. indent is off.  There are a few more cases like this.
> 
> Thanks, will fix this. (note to self: don't trust the editor's
> clang-format blindly).
> 
> > 
> > > +{
> > > +	struct sock *sk;
> > > +
> > > +	switch (local_storage->stype) {
> > Does it need a new bpf_local_storage_type?  Is map_type as good?
> > 
> > Instead of adding any new member (e.g. stype) to
> > "struct bpf_local_storage",  can the smap pointer be directly used
> > here instead?
> > 
> > For example in __unlink_local_storage() here, it should
> > have a hold to the selem which then has a hold to smap.
> 
> Good point, Updated to using the map->map_type.
> 
> > 
> > > +	case BPF_LOCAL_STORAGE_SK:
> > > +		sk = local_storage->sk;
> > > +		if (uncharge_omem)
> > > +			atomic_sub(sizeof(struct bpf_local_storage),
> > > +				   &sk->sk_omem_alloc);
> > > +
> > > +		/* After this RCU_INIT, sk may be freed and cannot be used */
> > > +		RCU_INIT_POINTER(sk->sk_bpf_storage, NULL);
> > > +		local_storage->sk = NULL;
> > > +		break;
> > > +	}
> > Another thought on the stype switch cases.
> > 
> > Instead of having multiple switches on stype in bpf_local_storage.c which may
> > not be scalable soon if we are planning to support a few more kernel objects,
> > have you considered putting them into its own "ops".  May be a few new
> > ops can be added to bpf_map_ops to do local storage unlink/update/alloc...etc.
> 
> Good idea, I was able to refactor this with the following ops:
> 
>         /* Functions called by bpf_local_storage maps */
> 	void (*map_local_storage_unlink)(struct bpf_local_storage *local_storage,
>                                          bool uncharge_omem);
> 	struct bpf_local_storage_elem *(*map_selem_alloc)(
> 		struct bpf_local_storage_map *smap, void *owner, void *value,
> 		bool charge_omem);
> 	struct bpf_local_storage_data *(*map_local_storage_update)(
> 		void  *owner, struct bpf_map *map, void *value, u64 flags);
> 	int (*map_local_storage_alloc)(void *owner,
> 				       struct bpf_local_storage_map *smap,
> 				       struct bpf_local_storage_elem *elem);
> 
> Let me know if you have any particular thoughts/suggestions about
> this.
Make sense to me.

It seems the fast-path's lookup can all be done within __local_storage_lookup().
No indirect call is needed, so should be good.

> 
> > 
> > > +}
> > > +
> > > +/* local_storage->lock must be held and selem->local_storage == local_storage.
> > >   * The caller must ensure selem->smap is still valid to be
> > >   * dereferenced for its smap->elem_size and smap->cache_idx.
> > > + *
> > > + * uncharge_omem is only relevant when:
> 
> [...]
> 
> > > +	/* bpf_local_storage_map is currently limited to CAP_SYS_ADMIN as
> > >  	 * the map_alloc_check() side also does.
> > >  	 */
> > >  	if (!bpf_capable())
> > > @@ -1025,10 +1127,10 @@ bpf_sk_storage_diag_alloc(const struct nlattr *nla_stgs)
> > >  }
> > >  EXPORT_SYMBOL_GPL(bpf_sk_storage_diag_alloc);
> > Would it be cleaner to leave bpf_sk specific function, map_ops, and func_proto
> > in net/core/bpf_sk_storage.c?
> 
> Sure, I can also keep the sk_clone code their as well for now.
Just came to my mind.  For easier review purpose, may be
first do the refactoring/renaming within bpf_sk_storage.c first and
then create another patch to move the common parts to a new
file bpf_local_storage.c.

Not sure if it will be indeed easier to read the diff in practice.
I probably should be able to follow it either way.

> 
> > 
> > There is a test in map_tests/sk_storage_map.c, in case you may not notice.
> 
> I will try to make it generic as a part of this series. If it takes
> too much time, I will send a separate patch for testing
> inode_storage_map and till then we have some assurance with
> test_local_storage in test_progs.
Sure. no problem.  It is mostly for you to test sk_storage to ensure things ;)
Also give some ideas on what racing conditions have
been considered in the sk_storage test and may be the inode storage
test want to stress similar code path.