On Mon, Nov 22, 2021 at 03:13:44PM +0300, Vitaliy Makkoveev wrote: > Since the rev1.267 of kern/uipc_socket solock() is used as klist lock > for sockets to make socket event filters MP-safe. This means KNOTE(9) > within doaccept() doesn't require kernel lock to be held and the > accept(2) and accept4(2) syscalls could be unlocked. This makes sense > because all our sockets use rwlock(9)s for protection. This also makes > sense for the next step of unix(4) sockets unlocking. > > The sys_accept() and sys_accept4() are just wrappers around doaccept() > which does all the work and there is no differences within so they could > be unlocked together. > > I also tested this diff with all incarnation of "Rework UNIX sockets > locking to be fine grained" diff [1]. > > ok?
regress passes on i386 and amd64. OK bluhm@ > > 1.https://marc.info/?l=openbsd-tech&m=163758154427455&w=2 > > Index: sys/kern/syscalls.master > =================================================================== > RCS file: /cvs/src/sys/kern/syscalls.master,v > retrieving revision 1.219 > diff -u -p -r1.219 syscalls.master > --- sys/kern/syscalls.master 27 Oct 2021 03:24:44 -0000 1.219 > +++ sys/kern/syscalls.master 22 Nov 2021 11:51:33 -0000 > @@ -95,7 +95,7 @@ > 29 STD NOLOCK { ssize_t sys_recvfrom(int s, void *buf, size_t len, \ > int flags, struct sockaddr *from, \ > socklen_t *fromlenaddr); } > -30 STD { int sys_accept(int s, struct sockaddr *name, \ > +30 STD NOLOCK { int sys_accept(int s, struct sockaddr *name, \ > socklen_t *anamelen); } > 31 STD { int sys_getpeername(int fdes, struct sockaddr *asa, \ > socklen_t *alen); } > @@ -204,7 +204,7 @@ > 91 STD NOLOCK { int sys_nanosleep(const struct timespec *rqtp, \ > struct timespec *rmtp); } > 92 STD NOLOCK { int sys_fcntl(int fd, int cmd, ... void *arg); } > -93 STD { int sys_accept4(int s, struct sockaddr *name, \ > +93 STD NOLOCK { int sys_accept4(int s, struct sockaddr *name, \ > socklen_t *anamelen, int flags); } > 94 STD NOLOCK { int sys___thrsleep(const volatile void *ident, \ > clockid_t clock_id, const struct timespec *tp, \
