[PATCH bpf 2/6] selftest/bpf: Socket pair creation, cleanups

Following create_pair() changes, remove unused function argument in create_socket_pairs() and adapt its callers, i.e. drop the open-coded loopback socket creation. Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_basic.c | 9

[PATCH bpf 5/6] selftest/bpf: Exercise SOCK_STREAM unix_inet_redir_to_connected()

Constants got switched reducing the test's coverage. Replace SOCK_DGRAM with SOCK_STREAM in one of unix_inet_skb_redir_to_connected() tests. Fixes: 51354f700d40 ("bpf, sockmap: Add af_unix test with both sockets in map") Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luc

[PATCH bpf 3/6] selftest/bpf: Simplify inet_socketpair() and vsock_unix_redir_connectible()

Replace implementation with a call to a generic function. Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_listen.c | 83 +- 1 file changed, 2 insertions(+), 81 deletions(-) diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_listen.c b

[PATCH bpf 6/6] selftest/bpf: Introduce __attribute__((cleanup)) in create_pair()

Rewrite function to have (unneeded) socket descriptors automatically close()d when leaving the scope. Make sure the "ownership" of fds is correctly passed via take_fd(); i.e. descriptor returned to caller will remain valid. Suggested-by: Jakub Sitnicki Signed-off-by: Mic

[PATCH bpf 4/6] selftest/bpf: Respect the sotype of af_unix redir tests

Do actually test the sotype as specified by the caller. This picks up after commit 75e0e27db6cf ("selftest/bpf: Change udp to inet in some function names"). Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 6 +++-

[PATCH bpf 1/6] selftest/bpf: Support more socket types in create_pair()

Extend the function to allow creating socket pairs of SOCK_STREAM, SOCK_DGRAM and SOCK_SEQPACKET. Adapt direct callers and leave further cleanups for the following patch. Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_basic.c | 19

[PATCH bpf 0/6] selftest/bpf: Various sockmap-related fixes

...@rbox.co/ - No declarations in function body (Jakub) - Don't touch output arguments until function succeeds (Jakub) Signed-off-by: Michal Luczaj --- Michal Luczaj (6): selftest/bpf: Support more socket types in create_pair() selftest/bpf: Socket pair creation, cleanups sel

Re: [PATCH bpf 3/6] selftest/bpf: Simplify inet_socketpair() and vsock_unix_redir_connectible()

On 7/24/24 13:32, Michal Luczaj wrote: > Replace implementation with a call to a generic function. > > Signed-off-by: Michal Luczaj Patch subject is incorrect, it should be: "... and vsock_socketpair_connectible()". Sorry for confusion. Michal

Re: [PATCH bpf 1/6] selftest/bpf: Support more socket types in create_pair()

On 7/26/24 19:23, Jakub Sitnicki wrote: > I was going to suggest that a single return path for success is better > than two (diff below), but I see that this is what you ended up with > after patch 6. > > So I think we can leave it as is. > [...] And speaking of which, would you rather have patch

Re: [PATCH bpf 6/6] selftest/bpf: Introduce __attribute__((cleanup)) in create_pair()

On 7/26/24 19:27, Jakub Sitnicki wrote: > On Wed, Jul 24, 2024 at 01:32 PM +02, Michal Luczaj wrote: >> +#define take_fd(fd) >>\ >> +({ \ >&g

Re: [PATCH bpf 0/6] selftest/bpf: Various sockmap-related fixes

On 7/26/24 19:36, Jakub Sitnicki wrote: > On Wed, Jul 24, 2024 at 01:32 PM +02, Michal Luczaj wrote: >> Series takes care of few bugs and missing features with the aim to improve >> the test coverage of sockmap/sockhash. >> >> Last patch is a create_pair() rewrite mak

[PATCH bpf-next v2 5/6] selftests/bpf: Exercise SOCK_STREAM unix_inet_redir_to_connected()

Constants got switched reducing the test's coverage. Replace SOCK_DGRAM with SOCK_STREAM in one of unix_inet_skb_redir_to_connected() tests. Fixes: 51354f700d40 ("bpf, sockmap: Add af_unix test with both sockets in map") Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luc

[PATCH bpf-next v2 0/6] selftests/bpf: Various sockmap-related fixes

Series takes care of few bugs and missing features with the aim to improve the test coverage of sockmap/sockhash. Last patch is a create_pair() rewrite making use of __attribute__((cleanup)) to handle socket fd lifetime. Signed-off-by: Michal Luczaj --- Changes in v2: - Rebase on bpf-next

[PATCH bpf-next v2 2/6] selftests/bpf: Socket pair creation, cleanups

Following create_pair() changes, remove unused function argument in create_socket_pairs() and adapt its callers, i.e. drop the open-coded loopback socket creation. Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_basic.c | 9

[PATCH bpf-next v2 4/6] selftests/bpf: Honour the sotype of af_unix redir tests

Do actually test the sotype as specified by the caller. This picks up after commit 75e0e27db6cf ("selftest/bpf: Change udp to inet in some function names"). Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 6 +++-

[PATCH bpf-next v2 3/6] selftests/bpf: Simplify inet_socketpair() and vsock_socketpair_connectible()

Replace implementation with a call to a generic function. Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_listen.c | 83 +- 1 file changed, 2 insertions(+), 81 deletions(-) diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_listen.c b

[PATCH bpf-next v2 6/6] selftests/bpf: Introduce __attribute__((cleanup)) in create_pair()

Rewrite function to have (unneeded) socket descriptors automatically close()d when leaving the scope. Make sure the "ownership" of fds is correctly passed via take_fd(); i.e. descriptor returned to caller will remain valid. Suggested-by: Jakub Sitnicki Signed-off-by: Mic

[PATCH bpf-next v2 1/6] selftests/bpf: Support more socket types in create_pair()

Extend the function to allow creating socket pairs of SOCK_STREAM, SOCK_DGRAM and SOCK_SEQPACKET. Adapt direct callers and leave further cleanups for the following patch. Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_basic.c | 19

Re: [PATCH bpf 1/6] selftest/bpf: Support more socket types in create_pair()

On 7/30/24 19:13, Jakub Sitnicki wrote: > On Fri, Jul 26, 2024 at 10:29 PM +02, Michal Luczaj wrote: >> On 7/26/24 19:23, Jakub Sitnicki wrote: >>> I was going to suggest that a single return path for success is better >>> than two (diff below), but I see that thi

Re: [PATCH bpf-next v2 0/6] selftests/bpf: Various sockmap-related fixes

On 8/5/24 17:22, Jakub Sitnicki wrote: > On Wed, Jul 31, 2024 at 12:01 PM +02, Michal Luczaj wrote: >> Series takes care of few bugs and missing features with the aim to improve >> the test coverage of sockmap/sockhash. >> >> Last patch is a create_pair() rewrite mak

Re: [PATCH bpf-next v2 0/6] selftests/bpf: Various sockmap-related fixes

On 8/6/24 14:01, Jakub Sitnicki wrote: > On Wed, Jul 31, 2024 at 12:01 PM +02, Michal Luczaj wrote: >> Series takes care of few bugs and missing features with the aim to improve >> the test coverage of sockmap/sockhash. >> >> Last patch is a create_pair() rewrite mak

[PATCH 1/3] selftests/bpf: Support AF_UNIX SOCK_DGRAM socket pair creation

Handle AF_UNIX in init_addr_loopback(). For pair creation, bind() the peer socket to make SOCK_DGRAM connect() happy. Signed-off-by: Michal Luczaj --- .../bpf/prog_tests/sockmap_helpers.h | 29 +++ 1 file changed, 24 insertions(+), 5 deletions(-) diff --git a/tools

[PATCH 2/3] selftests/bpf: Allow setting BPF_F_INGRESS in prog_msg_verdict()

Let a selftest set BPF_F_INGRESS for map/hash redirect. In run_tests(), explicitly reset skel->bss->test_ingress to false. Earlier tests might have left it flipped. Signed-off-by: Michal Luczaj --- tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 2 ++ tools/testing/selftes

[PATCH 3/3] selftests/bpf: Add selftest for sockmap/hashmap redirection

, SOCK_STREAM AF_VSOCK, SOCK_SEQPACKET Suggested-by: Jakub Sitnicki Signed-off-by: Michal Luczaj --- .../bpf/prog_tests/sockmap_helpers.h | 58 .../selftests/bpf/prog_tests/sockmap_redir.c | 315 ++ 2 files changed, 373 insertions(+) create mode 100644 tools/testing

Re: [PATCH bpf-next v2 0/6] selftests/bpf: Various sockmap-related fixes

On 8/6/24 19:45, Jakub Sitnicki wrote: > On Tue, Aug 06, 2024 at 07:18 PM +02, Michal Luczaj wrote: >> Great, thanks for the review. With this completed, I guess we can unwind >> the (mail) stack to [1]. Is that ingress-to-local et al. something you >> wanted to take care of y

[PATCH bpf 1/4] bpf, vsock: Fix poll() missing a queue

When a verdict program simply passes a packet without redirection, sk_msg is enqueued on sk_psock::ingress_msg. Add a missing check to poll(). Fixes: 634f1a7110b4 ("vsock: support sockmap") Signed-off-by: Michal Luczaj --- net/vmw_vsock/af_vsock.c | 3 +++ 1 file changed, 3 insertion

[PATCH bpf 3/4] bpf, vsock: Invoke proto::close on close()

: changes in __vsock_release() look messy, but it's only due to indent level reduction and variables xmas tree reorder. Fixes: 634f1a7110b4 ("vsock: support sockmap") Signed-off-by: Michal Luczaj --- net/vmw_vsock/af_vsock.c | 67 +--- 1 f

[PATCH bpf 0/4] bpf, vsock: Fix poll() and close()

Two small fixes for vsock: poll() missing a queue check, and close() not invoking sockmap cleanup. Signed-off-by: Michal Luczaj --- Michal Luczaj (4): bpf, vsock: Fix poll() missing a queue selftest/bpf: Add test for af_vsock poll() bpf, vsock: Invoke proto::close on close

[PATCH bpf 4/4] selftest/bpf: Add test for vsock removal from sockmap on close()

Make sure the proto::close callback gets invoked on vsock release. Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_basic.c | 31 ++ 1 file changed, 31 insertions(+) diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c b/tools

[PATCH bpf 2/4] selftest/bpf: Add test for af_vsock poll()

Verify that vsock's poll() notices when sk_psock::ingress_msg isn't empty. Signed-off-by: Michal Luczaj --- .../selftests/bpf/prog_tests/sockmap_basic.c | 46 ++ 1 file changed, 46 insertions(+) diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_

[PATCH bpf 1/3] bpf, sockmap: Fix update element with same

gger the cleanup, i.e. a closed socket will not be automatically removed from the sockmap. Stop tearing the links when a matching link_raw is found. Signed-off-by: Michal Luczaj --- net/core/sock_map.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/core/sock_map.c b/net/core/sock_map

[PATCH bpf 2/3] selftest/bpf: Extend test for sockmap update with same

Verify that the sockmap link was not severed, and socket's entry is indeed removed from the map when the corresponding descriptor gets closed. Signed-off-by: Michal Luczaj --- tools/testing/selftests/bpf/prog_tests/sockmap_basic.c | 8 +--- 1 file changed, 5 insertions(+), 3 dele

[PATCH bpf 0/3] bpf, sockmap: Fix the element replace

detail of the current __sock_map_delete() implementation. I'd be grateful for comments, thanks. Signed-off-by: Michal Luczaj --- Michal Luczaj (3): bpf, sockmap: Fix update element with same selftest/bpf: Extend test for sockmap update with same bpf, sockmap: Fix race

[PATCH bpf 3/3] bpf, sockmap: Fix race between element replace and close()

: [] asm_sysvec_apic_timer_interrupt+0x16/0x20 hardirqs last disabled at (10740): [] handle_softirqs+0x60d/0x770 softirqs last enabled at (10506): [] __irq_exit_rcu+0x109/0x210 softirqs last disabled at (10301): [] __irq_exit_rcu+0x109/0x210 Fixes: 604326b41a6f ("bpf, sockmap: convert to gen

Re: [PATCH bpf 1/3] bpf, sockmap: Fix update element with same

On 12/9/24 06:47, John Fastabend wrote: > Michal Luczaj wrote: >> Consider a sockmap entry being updated with the same socket: >> >> osk = stab->sks[idx]; >> sock_map_add_link(psock, link, map, &stab->sks[idx]); >> stab->sks[idx] = sk; &

Re: [PATCH bpf 3/4] bpf, vsock: Invoke proto::close on close()

On 11/21/24 10:22, Stefano Garzarella wrote: > On Mon, Nov 18, 2024 at 10:03:43PM +0100, Michal Luczaj wrote: >> vsock defines a BPF callback to be invoked when close() is called. However, >> this callback is never actually executed. As a result, a closed vsock >> socket