[v9] bpf: fix wrong copied_seq calculation and add tests

[PATCH bpf v9 0/5] bpf: fix wrong copied_seq calculation and add tests

Posted by Jiayuan Chen 2 weeks ago

A previous commit described in this topic
http://lore.kernel.org/bpf/20230523025618.113937-9-john.fastabend@gmail.com
directly updated 'sk->copied_seq' in the tcp_eat_skb() function when the
action of a BPF program was SK_REDIRECT. For other actions, like SK_PASS,
the update logic for 'sk->copied_seq' was moved to
tcp_bpf_recvmsg_parser() to ensure the accuracy of the 'fionread' feature.

That commit works for a single stream_verdict scenario, as it also
modified 'sk_data_ready->sk_psock_verdict_data_ready->tcp_read_skb'
to remove updating 'sk->copied_seq'.

However, for programs where both stream_parser and stream_verdict are
active (strparser purpose), tcp_read_sock() was used instead of
tcp_read_skb() (sk_data_ready->strp_data_ready->tcp_read_sock).
tcp_read_sock() now still updates 'sk->copied_seq', leading to duplicated
updates.

In summary, for strparser + SK_PASS, copied_seq is redundantly calculated
in both tcp_read_sock() and tcp_bpf_recvmsg_parser().

The issue causes incorrect copied_seq calculations, which prevent
correct data reads from the recv() interface in user-land.

Also we added test cases for bpf + strparser and separated them from
sockmap_basic, as strparser has more encapsulation and parsing
capabilities compared to sockmap.

---
V8 -> v9
https://lore.kernel.org/bpf/20250121050707.55523-1-mrpre@163.com/
Fixed some issues suggested by Jakub Sitnicki.

V7 -> V8
https://lore.kernel.org/bpf/20250116140531.108636-1-mrpre@163.com/
Avoid using add read_sock to psock. (Jakub Sitnicki)
Avoid using warpper function to check whether strparser is supported.

V3 -> V7:
https://lore.kernel.org/bpf/20250109094402.50838-1-mrpre@163.com/
https://lore.kernel.org/bpf/20241218053408.437295-1-mrpre@163.com/
Avoid introducing new proto_ops. (Jakub Sitnicki).
Add more edge test cases for strparser + bpf.
Fix patchwork fail of test cases code.
Fix psock fetch without rcu lock.
Move code of modifying to tcp_bpf.c.

V1 -> V3:
https://lore.kernel.org/bpf/20241209152740.281125-1-mrpre@163.com/
Fix patchwork fail by adding Fixes tag.
Save skb data offset for ENOMEM. (John Fastabend)
---

Jiayuan Chen (5):
  strparser: add read_sock callback
  bpf: fix wrong copied_seq calculation
  bpf: disable non stream socket for strparser
  selftests/bpf: fix invalid flag of recv()
  selftests/bpf: add strparser test for bpf

 Documentation/networking/strparser.rst        |   9 +-
 include/linux/skmsg.h                         |   2 +
 include/net/strparser.h                       |   2 +
 include/net/tcp.h                             |   8 +
 net/core/skmsg.c                              |   7 +
 net/core/sock_map.c                           |   5 +-
 net/ipv4/tcp.c                                |  29 +-
 net/ipv4/tcp_bpf.c                            |  36 ++
 net/strparser/strparser.c                     |  11 +-
 .../selftests/bpf/prog_tests/sockmap_basic.c  |  59 +--
 .../selftests/bpf/prog_tests/sockmap_strp.c   | 454 ++++++++++++++++++
 .../selftests/bpf/progs/test_sockmap_strp.c   |  53 ++
 12 files changed, 610 insertions(+), 65 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/sockmap_strp.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_sockmap_strp.c

-- 
2.43.5

Re: [PATCH bpf v9 0/5] bpf: fix wrong copied_seq calculation and add tests

Posted by Jakub Sitnicki 1 week, 3 days ago

On Wed, Jan 22, 2025 at 06:09 PM +08, Jiayuan Chen wrote:
> A previous commit described in this topic
> http://lore.kernel.org/bpf/20230523025618.113937-9-john.fastabend@gmail.com
> directly updated 'sk->copied_seq' in the tcp_eat_skb() function when the
> action of a BPF program was SK_REDIRECT. For other actions, like SK_PASS,
> the update logic for 'sk->copied_seq' was moved to
> tcp_bpf_recvmsg_parser() to ensure the accuracy of the 'fionread' feature.
>
> That commit works for a single stream_verdict scenario, as it also
> modified 'sk_data_ready->sk_psock_verdict_data_ready->tcp_read_skb'
> to remove updating 'sk->copied_seq'.
>
> However, for programs where both stream_parser and stream_verdict are
> active (strparser purpose), tcp_read_sock() was used instead of
> tcp_read_skb() (sk_data_ready->strp_data_ready->tcp_read_sock).
> tcp_read_sock() now still updates 'sk->copied_seq', leading to duplicated
> updates.
>
> In summary, for strparser + SK_PASS, copied_seq is redundantly calculated
> in both tcp_read_sock() and tcp_bpf_recvmsg_parser().
>
> The issue causes incorrect copied_seq calculations, which prevent
> correct data reads from the recv() interface in user-land.
>
> Also we added test cases for bpf + strparser and separated them from
> sockmap_basic, as strparser has more encapsulation and parsing
> capabilities compared to sockmap.
>
> ---
> V8 -> v9
> https://lore.kernel.org/bpf/20250121050707.55523-1-mrpre@163.com/
> Fixed some issues suggested by Jakub Sitnicki.
>
> V7 -> V8
> https://lore.kernel.org/bpf/20250116140531.108636-1-mrpre@163.com/
> Avoid using add read_sock to psock. (Jakub Sitnicki)
> Avoid using warpper function to check whether strparser is supported.
>
> V3 -> V7:
> https://lore.kernel.org/bpf/20250109094402.50838-1-mrpre@163.com/
> https://lore.kernel.org/bpf/20241218053408.437295-1-mrpre@163.com/
> Avoid introducing new proto_ops. (Jakub Sitnicki).
> Add more edge test cases for strparser + bpf.
> Fix patchwork fail of test cases code.
> Fix psock fetch without rcu lock.
> Move code of modifying to tcp_bpf.c.
>
> V1 -> V3:
> https://lore.kernel.org/bpf/20241209152740.281125-1-mrpre@163.com/
> Fix patchwork fail by adding Fixes tag.
> Save skb data offset for ENOMEM. (John Fastabend)
> ---

Thanks for addressing all feedback, Jiayuan. Series LGTM.
Feel free to carry my tags if there is another iteration.

-jkbs

Re: [PATCH bpf v9 0/5] bpf: fix wrong copied_seq calculation and add tests

Posted by John Fastabend 1 week, 2 days ago

On 2025-01-26 15:16:47, Jakub Sitnicki wrote:
> On Wed, Jan 22, 2025 at 06:09 PM +08, Jiayuan Chen wrote:
> > A previous commit described in this topic
> > http://lore.kernel.org/bpf/20230523025618.113937-9-john.fastabend@gmail.com
> > directly updated 'sk->copied_seq' in the tcp_eat_skb() function when the
> > action of a BPF program was SK_REDIRECT. For other actions, like SK_PASS,
> > the update logic for 'sk->copied_seq' was moved to
> > tcp_bpf_recvmsg_parser() to ensure the accuracy of the 'fionread' feature.
> >
> > That commit works for a single stream_verdict scenario, as it also
> > modified 'sk_data_ready->sk_psock_verdict_data_ready->tcp_read_skb'
> > to remove updating 'sk->copied_seq'.
> >
> > However, for programs where both stream_parser and stream_verdict are
> > active (strparser purpose), tcp_read_sock() was used instead of
> > tcp_read_skb() (sk_data_ready->strp_data_ready->tcp_read_sock).
> > tcp_read_sock() now still updates 'sk->copied_seq', leading to duplicated
> > updates.
> >
> > In summary, for strparser + SK_PASS, copied_seq is redundantly calculated
> > in both tcp_read_sock() and tcp_bpf_recvmsg_parser().
> >
> > The issue causes incorrect copied_seq calculations, which prevent
> > correct data reads from the recv() interface in user-land.
> >
> > Also we added test cases for bpf + strparser and separated them from
> > sockmap_basic, as strparser has more encapsulation and parsing
> > capabilities compared to sockmap.
> >
> > ---
> > V8 -> v9
> > https://lore.kernel.org/bpf/20250121050707.55523-1-mrpre@163.com/
> > Fixed some issues suggested by Jakub Sitnicki.
> >
> > V7 -> V8
> > https://lore.kernel.org/bpf/20250116140531.108636-1-mrpre@163.com/
> > Avoid using add read_sock to psock. (Jakub Sitnicki)
> > Avoid using warpper function to check whether strparser is supported.
> >
> > V3 -> V7:
> > https://lore.kernel.org/bpf/20250109094402.50838-1-mrpre@163.com/
> > https://lore.kernel.org/bpf/20241218053408.437295-1-mrpre@163.com/
> > Avoid introducing new proto_ops. (Jakub Sitnicki).
> > Add more edge test cases for strparser + bpf.
> > Fix patchwork fail of test cases code.
> > Fix psock fetch without rcu lock.
> > Move code of modifying to tcp_bpf.c.
> >
> > V1 -> V3:
> > https://lore.kernel.org/bpf/20241209152740.281125-1-mrpre@163.com/
> > Fix patchwork fail by adding Fixes tag.
> > Save skb data offset for ENOMEM. (John Fastabend)
> > ---
> 
> Thanks for addressing all feedback, Jiayuan. Series LGTM.
> Feel free to carry my tags if there is another iteration.

+1 Thanks Jiayuan for sticking with this.

I've reviewed this a couple times. I had one nit on the if/else branch
for a read call, but I haven't come up with anything better on my end
and this fixes a real bug. So lets take it.

For the series.

Acked-by: John Fastabend <john.fastabend@gmail.com>

> 
> -jkbs