BPF Archive on lore.kernel.org
 help / color / Atom feed
From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Martin KaFai Lau <kafai@fb.com>
Cc: bpf <bpf@vger.kernel.org>, Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Kernel Team <kernel-team@fb.com>,
	Networking <netdev@vger.kernel.org>
Subject: Re: [PATCH bpf-next 0/3] bpf: Allow inner map with different max_entries
Date: Thu, 21 May 2020 16:10:36 -0700
Message-ID: <CAEf4BzZyhT6D6F2A+cN6TKvLFoH5GU5QVEVW7ZkG+KQRgJC-1w@mail.gmail.com> (raw)
In-Reply-To: <20200521225939.7nmw7l5dk3wf557r@kafai-mbp.dhcp.thefacebook.com>

On Thu, May 21, 2020 at 3:59 PM Martin KaFai Lau <kafai@fb.com> wrote:
>
> On Thu, May 21, 2020 at 03:39:10PM -0700, Andrii Nakryiko wrote:
> > On Thu, May 21, 2020 at 12:18 PM Martin KaFai Lau <kafai@fb.com> wrote:
> > >
> > > This series allows the outer map to be updated with inner map in different
> > > size as long as it is safe (meaning the max_entries is not used in the
> > > verification time during prog load).
> > >
> > > Please see individual patch for details.
> > >
> >
> > Few thoughts:
> >
> > 1. You describe WHAT, but not necessarily WHY. Can you please
> > elaborate in descriptions what motivates these changes?
> There are cases where people want to update a bigger size
> inner map.  I will update the cover letter.
>
> > 2. IMO, "capabilities" is word that way too strongly correlates with
> > Linux capabilities framework, it's just confusing. It's also more of a
> > property of a map type, than what map is capable of, but it's more
> > philosophical distinction, of course :)
> Sure. I can rename it to "property"
>
> > 3. I'm honestly not convinced that patch #1 qualifies as a clean up. I
> > think one specific check for types of maps that are not compatible
> > with map-in-map is just fine. Instead you are spreading this bit flags
> > into a long list of maps, most of which ARE compatible.
> but in one place and at the same time a new map type is added to
> bpf_types.h
>
> > It's just hard
> > to even see which ones are not compatible. I like current way better.
> There are multiple cases that people forgot to exclude a new map
> type from map-in-map in the first attempt and fix it up later.
>
> During the map-in-map implementation, this same concern was raised also
> about how to better exclude future map type from map-in-map since
> not all people has used map-in-map and it is easy to forget during
> review.  Having it in one place in bpf_types.h will make this
> more obvious in my opinion.  Patch 1 is an attempt to address
> this earlier concern in the map-in-map implementation.

Ok, just invert the condition and list only types that **are** allowed
inside map-in-map. If someone forgot to add it to map-in-map check, it
can be done later when someone needs it. The point is that we have a
check and a list in one place, close to where it matters, instead of
tracing where the value of ->capabilities comes from. Finding that in
bpf_types.h is not easy and not obvious, unfortunately, and is very
distant from where it's actually checked.


>
> > 4. Then for size check change, again, it's really much simpler and
> > cleaner just to have a special case in check in bpf_map_meta_equal for
> > cases where map size matters.
> It may be simpler but not necessary less fragile for future map type.
>
> I am OK for removing patch 1 and just check for a specific
> type in patch 2 but I think it is fragile for future map
> type IMO.

Well, if we think that the good default needs to be to check size,
then similar to above, explicitly list stuff that *does not* follow
the default, i.e., maps that don't want max_elements verification. My
point still stands.

>
> > 5. I also wonder if for those inner maps for which size doesn't
> > matter, maybe we should set max_elements to zero when setting
> > inner_meta to show that size doesn't matter? This is minor, though.
> >
> >
> > > Martin KaFai Lau (3):
> > >   bpf: Clean up inner map type check
> > >   bpf: Relax the max_entries check for inner map
> > >   bpf: selftests: Add test for different inner map size
> > >
> > >  include/linux/bpf.h                           | 18 +++++-
> > >  include/linux/bpf_types.h                     | 64 +++++++++++--------
> > >  kernel/bpf/btf.c                              |  2 +-
> > >  kernel/bpf/map_in_map.c                       | 12 ++--
> > >  kernel/bpf/syscall.c                          | 19 +++++-
> > >  kernel/bpf/verifier.c                         |  2 +-
> > >  .../selftests/bpf/prog_tests/btf_map_in_map.c | 12 ++++
> > >  .../selftests/bpf/progs/test_btf_map_in_map.c | 31 +++++++++
> > >  8 files changed, 119 insertions(+), 41 deletions(-)
> > >
> > > --
> > > 2.24.1
> > >

  reply index

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-21 19:17 Martin KaFai Lau
2020-05-21 19:17 ` [PATCH bpf-next 1/3] bpf: Clean up inner map type check Martin KaFai Lau
2020-05-21 19:18 ` [PATCH bpf-next 2/3] bpf: Relax the max_entries check for inner map Martin KaFai Lau
2020-05-21 19:18 ` [PATCH bpf-next 3/3] bpf: selftests: Add test for different inner map size Martin KaFai Lau
2020-05-21 22:39 ` [PATCH bpf-next 0/3] bpf: Allow inner map with different max_entries Andrii Nakryiko
2020-05-21 22:59   ` Martin KaFai Lau
2020-05-21 23:10     ` Andrii Nakryiko [this message]
2020-05-21 23:16       ` Alexei Starovoitov
2020-05-22  0:06         ` Martin KaFai Lau
2020-05-26 17:47         ` Andrii Nakryiko

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAEf4BzZyhT6D6F2A+cN6TKvLFoH5GU5QVEVW7ZkG+KQRgJC-1w@mail.gmail.com \
    --to=andrii.nakryiko@gmail.com \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=kafai@fb.com \
    --cc=kernel-team@fb.com \
    --cc=netdev@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

BPF Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/bpf/0 bpf/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 bpf bpf/ https://lore.kernel.org/bpf \
		bpf@vger.kernel.org
	public-inbox-index bpf

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.bpf


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git