wireguard-openbsd/sys/kern/sys_pipe.c, branch jd/queueboosts WireGuard implementation for the OpenBSD kernel https://git.zx2c4.com/wireguard-openbsd/atom/sys/kern/sys_pipe.c?h=jd%2Fqueueboosts 2020-06-19T02:08:48Z backout pipe change, it crashes some arch 2020-06-19T02:08:48Z deraadt deraadt@openbsd.org 2020-06-19T02:08:48Z urn:sha1:8631d7d972792c6c7fa7099b8e25765d2df2a208 Instead of performing three distinct allocations per created pipe, 2020-06-17T18:29:28Z anton anton@openbsd.org 2020-06-17T18:29:28Z urn:sha1:15ad1813789d0ac9ab1cb6bcd5eabf14996ca0d9 reduce it to a single one. Not only should this be more performant, it also solves a kqueue related issue found by visa@ who also requested this change: if you attach an EVFILT_WRITE filter to a pipe fd, the knote gets added to the peer's klist. This is a problem for kqueue because if you close the peer's fd, the knote is left in the list whose head is about to be freed. knote_fdclose() is not able to clear the knote because it is not registered with the peer's fd. FreeBSD also takes a similar approach to pipe allocations. ok mpi@ visa@ Set __EV_HUP when the conditions matching poll(2)'s POLLUP are found. 2020-06-15T15:29:40Z mpi mpi@openbsd.org 2020-06-15T15:29:40Z urn:sha1:2b88cdafd374c425bef518b6071e7ccde4677902 This is only done in poll-compatibility mode, when __EV_POLL is set. ok visa@, millert@ Abstract the head of knote lists. This allows extending the lists, 2020-04-07T13:27:50Z visa visa@openbsd.org 2020-04-07T13:27:50Z urn:sha1:9c969c9ab4dbfc69101f439c13329bb852e9fe18 for example, with locking assertions. OK mpi@, anton@ Replace field f_isfd with field f_flags in struct filterops to allow 2020-02-20T16:56:51Z visa visa@openbsd.org 2020-02-20T16:56:51Z urn:sha1:b821368957f060d7b1e9eeaa4fbc563e892edb32 adding more filter properties without cluttering the struct. OK mpi@, anton@ Unconditionally acquiring a write lock in pipe_ioctl() is quite 2020-02-16T07:59:08Z anton anton@openbsd.org 2020-02-16T07:59:08Z urn:sha1:42e31c228413cc665b52f53adf7450715f9bcdcd excessive as only one command actually modifies the pipe. The sigio subsystem is already internally protected using its own lock. This is similar to what soo_ioctl() already does. ok mpi@ visa@ Push the KERNEL_LOCK() insidge pgsigio() and selwakeup(). 2020-02-14T14:32:44Z mpi mpi@openbsd.org 2020-02-14T14:32:44Z urn:sha1:112e0b7cd8c0810f5e6ec8c11d147cff217a606d The 3 subsystems: signal, poll/select and kqueue can now be addressed separatly. Note that bpf(4) and audio(4) currently delay the wakeups to a separate context in order to respect the KERNEL_LOCK() requirement. Sockets (UDP, TCP) and pipes spin to grab the lock for the sames reasons. ok anton@, visa@ Back out previous. Nothing wrong with the diff per se but I should have 2020-02-01T15:52:34Z anton anton@openbsd.org 2020-02-01T15:52:34Z urn:sha1:52cbe52912ac7d75938196a4b47625f1d6d61bb3 asked for more oks; my bad! Grab the kernel lock in pgsigio() as it's strictly needed while 2020-02-01T15:06:21Z anton anton@openbsd.org 2020-02-01T15:06:21Z urn:sha1:fec696e8b19f2ebeba1c5bca5b14987979b47d40 operating on the process structure and issuing signals. This is similar to what sigio_setown() already does. With this in place, the pipe subsystem is no longer required to grab the kernel lock before calling pgsigio(). ok visa@ condense comment 2020-01-12T09:18:39Z anton anton@openbsd.org 2020-01-12T09:18:39Z urn:sha1:ea998dad1e2381e4e26aeb8c67ce1425a6411c43