| Message ID | 20211129153929.3457-1-Jason@zx2c4.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | wireguard/siphash patches for 5.16-rc6 | expand |
On Mon, 29 Nov 2021 10:39:19 -0500 Jason A. Donenfeld wrote: > Here's quite a largeish set of stable patches I've had queued up and > testing for a number of months now: > > - Patch (1) squelches a sparse warning by fixing an annotation. > - Patches (2), (3), and (5) are minor improvements and fixes to the > test suite. > - Patch (4) is part of a tree-wide cleanup to have module-specific > init and exit functions. > - Patch (6) fixes a an issue with dangling dst references, by having a > function to release references immediately rather than deferring, > and adds an associated test case to prevent this from regressing. > - Patches (7) and (8) help mitigate somewhat a potential DoS on the > ingress path due to the use of skb_list's locking hitting contention > on multiple cores by switching to using a ring buffer and dropping > packets on contention rather than locking up another core spinning. > - Patch (9) switches kvzalloc to kvcalloc for better form. > - Patch (10) fixes alignment traps in siphash with clang-13 (and maybe > other compilers) on armv6, by switching to using the unaligned > functions by default instead of the aligned functions by default. Typo in the subject, right? No particular connection to -rc6 here? Just checking.
On Mon, Nov 29, 2021 at 1:17 PM Jakub Kicinski <kuba@kernel.org> wrote: > > On Mon, 29 Nov 2021 10:39:19 -0500 Jason A. Donenfeld wrote: > > Here's quite a largeish set of stable patches I've had queued up and > > testing for a number of months now: > > > > - Patch (1) squelches a sparse warning by fixing an annotation. > > - Patches (2), (3), and (5) are minor improvements and fixes to the > > test suite. > > - Patch (4) is part of a tree-wide cleanup to have module-specific > > init and exit functions. > > - Patch (6) fixes a an issue with dangling dst references, by having a > > function to release references immediately rather than deferring, > > and adds an associated test case to prevent this from regressing. > > - Patches (7) and (8) help mitigate somewhat a potential DoS on the > > ingress path due to the use of skb_list's locking hitting contention > > on multiple cores by switching to using a ring buffer and dropping > > packets on contention rather than locking up another core spinning. > > - Patch (9) switches kvzalloc to kvcalloc for better form. > > - Patch (10) fixes alignment traps in siphash with clang-13 (and maybe > > other compilers) on armv6, by switching to using the unaligned > > functions by default instead of the aligned functions by default. > > Typo in the subject, right? No particular connection to -rc6 here? > Just checking. I definitely meant to type 5.16-rc4, yes, sorry. Noticed this after it was too late.
Hi Dave/Jakub, Here's quite a largeish set of stable patches I've had queued up and testing for a number of months now: - Patch (1) squelches a sparse warning by fixing an annotation. - Patches (2), (3), and (5) are minor improvements and fixes to the test suite. - Patch (4) is part of a tree-wide cleanup to have module-specific init and exit functions. - Patch (6) fixes a an issue with dangling dst references, by having a function to release references immediately rather than deferring, and adds an associated test case to prevent this from regressing. - Patches (7) and (8) help mitigate somewhat a potential DoS on the ingress path due to the use of skb_list's locking hitting contention on multiple cores by switching to using a ring buffer and dropping packets on contention rather than locking up another core spinning. - Patch (9) switches kvzalloc to kvcalloc for better form. - Patch (10) fixes alignment traps in siphash with clang-13 (and maybe other compilers) on armv6, by switching to using the unaligned functions by default instead of the aligned functions by default. Thanks, Jason Arnd Bergmann (1): siphash: use _unaligned version by default Gustavo A. R. Silva (1): wireguard: ratelimiter: use kvcalloc() instead of kvzalloc() Jason A. Donenfeld (6): wireguard: allowedips: add missing __rcu annotation to satisfy sparse wireguard: selftests: increase default dmesg log size wireguard: selftests: actually test for routing loops wireguard: device: reset peer src endpoint when netns exits wireguard: receive: use ring buffer for incoming handshakes wireguard: receive: drop handshakes if queue lock is contended Li Zhijian (1): wireguard: selftests: rename DEBUG_PI_LIST to DEBUG_PLIST Randy Dunlap (1): wireguard: main: rename 'mod_init' & 'mod_exit' functions to be module-specific drivers/net/wireguard/allowedips.c | 2 +- drivers/net/wireguard/device.c | 39 ++++++++++--------- drivers/net/wireguard/device.h | 9 ++--- drivers/net/wireguard/main.c | 8 ++-- drivers/net/wireguard/queueing.c | 6 +-- drivers/net/wireguard/queueing.h | 2 +- drivers/net/wireguard/ratelimiter.c | 4 +- drivers/net/wireguard/receive.c | 39 +++++++++++-------- drivers/net/wireguard/socket.c | 2 +- include/linux/siphash.h | 14 ++----- include/net/dst_cache.h | 11 ++++++ lib/siphash.c | 12 +++--- net/core/dst_cache.c | 19 +++++++++ tools/testing/selftests/wireguard/netns.sh | 30 +++++++++++++- .../selftests/wireguard/qemu/debug.config | 2 +- .../selftests/wireguard/qemu/kernel.config | 1 + 16 files changed, 129 insertions(+), 71 deletions(-)