From: Jakub Sitnicki <jakub@cloudflare.com>
To: Alexei Starovoitov <alexei.starovoitov@gmail.com>,
John Fastabend <john.fastabend@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>, bpf <bpf@vger.kernel.org>,
Network Development <netdev@vger.kernel.org>,
kernel-team <kernel-team@cloudflare.com>
Subject: Re: [PATCH bpf 3/3] selftests/bpf: Test freeing sockmap/sockhash with a socket in it
Date: Mon, 10 Feb 2020 11:52:36 +0000 [thread overview]
Message-ID: <87blq6accb.fsf@cloudflare.com> (raw)
In-Reply-To: <5e40d43715474_2a9a2abf5f7f85c025@john-XPS-13-9370.notmuch>
On Mon, Feb 10, 2020 at 03:55 AM GMT, John Fastabend wrote:
> Jakub Sitnicki wrote:
>> On Sun, Feb 09, 2020 at 03:41 AM CET, Alexei Starovoitov wrote:
>> > On Thu, Feb 6, 2020 at 3:28 AM Jakub Sitnicki <jakub@cloudflare.com> wrote:
>> >>
>> >> Commit 7e81a3530206 ("bpf: Sockmap, ensure sock lock held during tear
>> >> down") introduced sleeping issues inside RCU critical sections and while
>> >> holding a spinlock on sockmap/sockhash tear-down. There has to be at least
>> >> one socket in the map for the problem to surface.
>> >>
>> >> This adds a test that triggers the warnings for broken locking rules. Not a
>> >> fix per se, but rather tooling to verify the accompanying fixes. Run on a
>> >> VM with 1 vCPU to reproduce the warnings.
>> >>
>> >> Fixes: 7e81a3530206 ("bpf: Sockmap, ensure sock lock held during tear down")
>> >> Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
>> >
>> > selftests/bpf no longer builds for me.
>> > make
>> > BINARY test_maps
>> > TEST-OBJ [test_progs] sockmap_basic.test.o
>> > /data/users/ast/net/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c:
>> > In function ‘connected_socket_v4’:
>> > /data/users/ast/net/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c:20:11:
>> > error: ‘TCP_REPAIR_ON’ undeclared (first use in this function); did
>> > you mean ‘TCP_REPAIR’?
>> > 20 | repair = TCP_REPAIR_ON;
>> > | ^~~~~~~~~~~~~
>> > | TCP_REPAIR
>> > /data/users/ast/net/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c:20:11:
>> > note: each undeclared identifier is reported only once for each
>> > function it appears in
>> > /data/users/ast/net/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c:29:11:
>> > error: ‘TCP_REPAIR_OFF_NO_WP’ undeclared (first use in this function);
>> > did you mean ‘TCP_REPAIR_OPTIONS’?
>> > 29 | repair = TCP_REPAIR_OFF_NO_WP;
>> > | ^~~~~~~~~~~~~~~~~~~~
>> > | TCP_REPAIR_OPTIONS
>> >
>> > Clearly /usr/include/linux/tcp.h is too old.
>> > Suggestions?
>>
>> Sorry for the inconvenience. I see that tcp.h header is missing under
>> linux/tools/include/uapi/.
>
> How about we just add the couple defines needed to sockmap_basic.c I don't
> see a need to pull in all of tcp.h just for a couple defines that wont
> change anyways.
Looking back at how this happened. test_progs.h pulls in netinet/tcp.h:
# 19 "/home/jkbs/src/linux/tools/testing/selftests/bpf/test_progs.h" 2
# 1 "/usr/include/netinet/tcp.h" 1 3 4
# 92 "/usr/include/netinet/tcp.h" 3 4
A glibc header, which gained TCP_REPAIR_* constants in 2.29 [0]:
$ git describe --contains 5cd7dbdea13eb302620491ef44837b17e9d39c5a
glibc-2.29~510
Pulling in linux/tcp.h would conflict with struct definitions in
netinet/tcp.h. So redefining the possibly missing constants, like John
suggests, is the right way out.
I'm not sure, though, how to protect against such mistakes in the
future. Any ideas?
[0] https://sourceware.org/git/?p=glibc.git;a=commit;h=5cd7dbdea13eb302620491ef44837b17e9d39c5a
>
>>
>> I have been building against my distro kernel headers, completely
>> unaware of this. This is an oversight on my side.
>>
>> Can I ask for a revert? I'm traveling today with limited ability to
>> post patches.
>
> I don't think we need a full revert.
>
>>
>> I can resubmit the test with the missing header for bpf-next once it
>> reopens.
>
> If you are traveling I'll post a patch with the defines.
Thanks, again.
next prev parent reply other threads:[~2020-02-10 11:52 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-06 11:16 [PATCH bpf 0/3] Fix locking order and synchronization on sockmap/sockhash tear-down Jakub Sitnicki
2020-02-06 11:16 ` [PATCH bpf 1/3] bpf, sockmap: Don't sleep while holding RCU lock on tear-down Jakub Sitnicki
2020-02-06 18:59 ` John Fastabend
2020-02-06 11:16 ` [PATCH bpf 2/3] bpf, sockhash: synchronize_rcu before free'ing map Jakub Sitnicki
2020-02-06 19:01 ` John Fastabend
2020-02-06 11:16 ` [PATCH bpf 3/3] selftests/bpf: Test freeing sockmap/sockhash with a socket in it Jakub Sitnicki
2020-02-06 19:03 ` John Fastabend
2020-02-09 2:41 ` Alexei Starovoitov
2020-02-09 4:12 ` Yonghong Song
2020-02-09 15:29 ` Jakub Sitnicki
2020-02-10 3:55 ` John Fastabend
2020-02-10 11:52 ` Jakub Sitnicki [this message]
2020-02-10 23:38 ` Daniel Borkmann
2020-02-06 19:43 ` [PATCH bpf 0/3] Fix locking order and synchronization on sockmap/sockhash tear-down John Fastabend
2020-02-07 10:45 ` Jakub Sitnicki
2020-03-10 16:44 ` John Fastabend
2020-03-11 11:51 ` Jakub Sitnicki
2020-03-10 11:30 ` Jakub Sitnicki
2020-02-07 21:56 ` Daniel Borkmann
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=87blq6accb.fsf@cloudflare.com \
--to=jakub@cloudflare.com \
--cc=alexei.starovoitov@gmail.com \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=john.fastabend@gmail.com \
--cc=kernel-team@cloudflare.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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).