Repply this change with a couple of bugs fixed:

- Do away with separate pool_cache for some kernel objects that have no special
requirements and use the general purpose allocator instead. On one of my
test systems this makes for a small (~1%) but repeatable reduction in system
time during builds presumably because it decreases the kernel's cache /
memory bandwidth footprint a little.
- vfs_lockf: cache a pointer to the uidinfo and put mutex in the data segment.

Back out recent change to replace pool_cache with then general allocator.
Will return to this when I have time again.

- Do away with separate pool_cache for some kernel objects that have no special
requirements and use the general purpose allocator instead. On one of my
test systems this makes for a small (~1%) but repeatable reduction in system
time during builds presumably because it decreases the kernel's cache /
memory bandwidth footprint a little.
- vfs_lockf: cache a pointer to the uidinfo and put mutex in the data segment.

kern/vfs_lockf.c: Parenthesize to make arithmetic match check.

We verified start + (fl->fl_len - 1) would not overflow, but then
computed (start + fl->fl_len) - 1 instead, and it is possible for
start + fl->fl_len to overflow before we subtract 1.

Reported-by: syzbot+762480b00cb14085f63a@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=d69313b2460a12715315f9f5f74fbe44b8bc38ba

kern/vfs_lockf.c: Fix overflow in overflow detection.

Reported-by: syzbot+cda9440741a1516747c8@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=030eb71324790093d467799263cd0789e5097229

vfs(9): Avoid arithmetic overflow in lf_advlock.

syzbot+897abbbe59467cbf6e98@syzkaller.appspotmail.com

fix various typos in comments and log messages.

Tie the maximum file lock per unprivilegied uid to kern.maxfiles

This makes the limit simple to raise at run time. While there, document
that fcntl(2) and flock(2) may return ENOMEM when this limit is reached.

branches: 1.73.60; 1.73.72;
Fix build with LOCKF_DEBUG on LP64 platforms; from Alexander Nasonov
in PR 44490.

branches: 1.72.4; 1.72.6; 1.72.8;
lockf() passes its arguments through to fcntl() but is supposed to
support -ve lengths (lock area before current offset).
Nothing in libc or the kernel allowed for this, so some random part
of the file would get locked (no idea which bits).
Although this could probably be fixed in libc, the stubs for posix file
locks for emulations could easily get into the kernel with -ve lengths.
So fixing in the kernel avoids those problems.
This also fixes PR/41620 (attempting to lock negative offsets) - which
is what I was looking into!

don't make F_GETLK or the common case of F_UNLCK fail for per-user limit.

lf_split: cv_destroy a condvar before clobbering it.

branches: 1.69.4; 1.69.8; 1.69.10;
Move uidinfo to its own module in kern_uidinfo.c and include in rump.
No functional change to uidinfo.

lf_setlock: g/c dead code.

memcpy'ing a sleepq (TAILQ) doesn't work.

Re-init the condvar after memcpy'ing so that the TAILQ is valid.

regress/sys/kern/lockf now completes rather than failing in cv_wait_sig
with KASSERT(cv_has_waiters()).

Improve the debug output ever so slightly.

branches: 1.65.4;
PR kern/38355 lockf deadlock detection is broken after vmlocking

- Fix it; tested with Sun's libMicro.
- Use pool_cache.
- Use a global lock, so the deadlock detection code is safer.

branches: 1.64.2; 1.64.4;
Merge proc::p_mutex and proc::p_smutex into a single adaptive mutex, since
we no longer need to guard against access from hardware interrupt handlers.

Additionally, if cloning a process with CLONE_SIGHAND, arrange to have the
child process share the parent's lock so that signal state may be kept in
sync. Partially addresses PR kern/37437.

branches: 1.63.2;
need <sys/atomic.h> now.

- Replace uihashtbl_lock and struct uidinfo::ui_lock with atomic operations.
This make uid_find(), chgproccnt(), chgsbsize() and lf_alloc(), lf_free()
functions lock-less.
- Increase the size of uihashtbl in case of MP system, as suggested by <ad>.
- Add HASH_SLIST type for hashinit().

Reviewed by <ad>.

branches: 1.61.2; 1.61.6;
Merge vmlocking2 to head.

branches: 1.60.8; 1.60.14; 1.60.16; 1.60.20;
Merge some of the less invasive changes from the vmlocking branch:

- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements

branches: 1.59.2;
Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.

branches: 1.58.2;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.57.2;
Merge newlock2 to head.

branches: 1.56.2;
Fix all the -D*DEBUG* code that it was rotting away and did not even compile.
Mostly from Arnaud Lacombe, many thanks!

Use the LWP cached credentials where sane.

remove NOLOCKF and use normal NULL instead.

remove an debug printf slipped into the previous.

fix F_SETLKW deadlock detection, which has been broken since lwpify.
although this doesn't work for processes with multiple lwps, it's better
than not working at all.

make lockfpool static.

move lockf implementation details from sys/lockf.h to kern/vfs_lockf.c.

branches: 1.49.2;
integrate kauth.

branches: 1.48.4; 1.48.6; 1.48.8; 1.48.10; 1.48.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

merge ktrace-lwp.

use ap->a_flags instead of fl->f_type; pointed out by: Devon H. O'Dell

branches: 1.45.2; 1.45.4;
Re-arrange slighty, eliminate prototypes that are unnecessary when using
ANSI function decls with "static".

- add const.
- remove unnecessary casts.
- add __UNCONST casts and mark them with XXXUNCONST as necessary.

No 0x in front of %p...

don't unlock twice.

lock all uses of uidhash. provide macros to lock and unlock. based on more
discussions with yamt.

PR/30154: YAMAMOTO Takashi: tcp_close locking botch
chgsbsize() as mentioned in the PR can be called from an interrupt context
via tcp_close(). Avoid calling uid_find() in chgsbsize().
- Instead of storing so_uid in struct socketvar, store *so_uidinfo
- Add a simple lock to struct uidinfo.

Don't free the wrong lock. Pointed out by Konrad Schroder

Keep track and limit the number of locks per uid. From OpenBSD

nuke trailing whitespace

branches: 1.36.4; 1.36.6;
Fix a typo, dependancy -> dependency

ok wiz@

Initialise (most) pools from a link set instead of explicit calls
to pool_init. Untouched pools are ones that either in arch-specific
code, or aren't initialiased during initial system startup.

Convert struct session, ucred and lockf to pools.

fix uninitialized variable

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

branches: 1.32.2;
revert rev.1.28. (free -> FREE changes)
pointed by David Laight on source-changes about similar changes.

Make gcc3 happy ("label at end of compound statement").

get pointer to v_interlock directly
as we are not interested in the rest of the vnode here.

- KNF.
- remove unneeded caddr_t casts.

use FREE instead of free for MALLOC'ed memory.

MP locking.

remove a part of comment that isn't true anymore.

ANSIfy.

don't export lockf internal functions.

DANGER WILL ROBINSON!

We cannot store LWP pointers permanently in lock structures, for two reasons:
1) They are somewhat ephemeral. Dangling pointers are bad.
2) A different LWP may issue the unlock, and in this case, we were not actually
doing the unlock at all. This was causing processes to exit without undoing
fcntl(2) locks. Furthermore, the locks are process-specific to begin with,
so the test was just plain wrong.

Instead, we go back to storing a proc pointer for POSIX locks. In addition, we
add an extra pointer to the LWP, which is used in deadlock detection. After
the lock is granted, this pointer is 0ed and there is no reference to the LWP.

Now evolution can inc my mail again.

Add extensible malloc types, adapted from FreeBSD. This turns
malloc types into a structure, a pointer to which is passed around,
instead of an int constant. Allow the limit to be adjusted when the
malloc type is defined, or with a function call, as suggested by
Jonathan Stone.

Merge the nathanw_sa branch.

"ovcase" is set but not used in lf_getblock(); remove it.

Use the queue macros from <sys/queue.h> instead of referring to the queue
members directly. Use *_FOREACH whenever possible.

add RCSIDs

branches: 1.17.2; 1.17.4; 1.17.8;
change the lf_advlock() arguments from

int lf_advlock __P((struct lockf **,
off_t, caddr_t, int, struct flock *, int));
to

int lf_advlock __P((struct vop_advlock_args *, struct lockf **, off_t));

This matches common usage and is also compatible with similar change
in FreeBSD (though they use u_quad_t as last arg).

branches: 1.16.2;
Fix several problems with lockf/fcntl byte range locks:
- document a data structure invariant in lockf.h
- add KASSERT() to check the invariant.
- be more consistent about dequeuing ourselves from the blocked list
after a tsleep().
- Fix two places where the invariant is violated.
- correct a few comments here and there
- If we're still following a lock dependancy chain after maxlockdepth
processes and haven't gotten back to the start, assume that we're in a
cycle anyway and return EDEADLK.

Fix is a superset of an existing fix in FreeBSD, but independantly
derived.

Fixes kern/3860.

branches: 1.15.2;
Get rid of register declarations.

branches: 1.14.6; 1.14.12;
Abolition of bcopy, ovbcopy, bcmp, and bzero, phase one.
bcopy(x, y, z) -> memcpy(y, x, z)
ovbcopy(x, y, z) -> memmove(y, x, z)
bcmp(x, y, z) -> memcmp(x, y, z)
bzero(x, y) -> memset(x, 0, y)

fix sizeofs so they comply with the KNF style guide. yes, it is pedantic.

branches: 1.12.2;
Merge with Lite2 + local changes

make it compile with -DLOCKF_DEBUG

Moved the `unlocking a lock-free inode' shortcut behind
the l_whence and starting offset sanity checks.

backout previous kprintf change

printf -> kprintf, sprintf -> ksprintf

First pass at prototyping

Use %p.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite, plus some local changes.

stupidity for prototypes...

some prototype cleanup, eliminate/replace bogus types (e.g. quad and
u_quad) -> use better types (e.g. quad_t & u_quad_t in inodes),
some cleanup.

Make FFS optional

Back out recent change to replace pool_cache with then general allocator.
Will return to this when I have time again.

- Do away with separate pool_cache for some kernel objects that have no special
requirements and use the general purpose allocator instead. On one of my
test systems this makes for a small (~1%) but repeatable reduction in system
time during builds presumably because it decreases the kernel's cache /
memory bandwidth footprint a little.
- vfs_lockf: cache a pointer to the uidinfo and put mutex in the data segment.

kern/vfs_lockf.c: Parenthesize to make arithmetic match check.

We verified start + (fl->fl_len - 1) would not overflow, but then
computed (start + fl->fl_len) - 1 instead, and it is possible for
start + fl->fl_len to overflow before we subtract 1.

Reported-by: syzbot+762480b00cb14085f63a@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=d69313b2460a12715315f9f5f74fbe44b8bc38ba

kern/vfs_lockf.c: Fix overflow in overflow detection.

Reported-by: syzbot+cda9440741a1516747c8@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=030eb71324790093d467799263cd0789e5097229

vfs(9): Avoid arithmetic overflow in lf_advlock.

syzbot+897abbbe59467cbf6e98@syzkaller.appspotmail.com

fix various typos in comments and log messages.

Tie the maximum file lock per unprivilegied uid to kern.maxfiles

This makes the limit simple to raise at run time. While there, document
that fcntl(2) and flock(2) may return ENOMEM when this limit is reached.

branches: 1.73.60; 1.73.72;
Fix build with LOCKF_DEBUG on LP64 platforms; from Alexander Nasonov
in PR 44490.

branches: 1.72.4; 1.72.6; 1.72.8;
lockf() passes its arguments through to fcntl() but is supposed to
support -ve lengths (lock area before current offset).
Nothing in libc or the kernel allowed for this, so some random part
of the file would get locked (no idea which bits).
Although this could probably be fixed in libc, the stubs for posix file
locks for emulations could easily get into the kernel with -ve lengths.
So fixing in the kernel avoids those problems.
This also fixes PR/41620 (attempting to lock negative offsets) - which
is what I was looking into!

don't make F_GETLK or the common case of F_UNLCK fail for per-user limit.

lf_split: cv_destroy a condvar before clobbering it.

branches: 1.69.4; 1.69.8; 1.69.10;
Move uidinfo to its own module in kern_uidinfo.c and include in rump.
No functional change to uidinfo.

lf_setlock: g/c dead code.

memcpy'ing a sleepq (TAILQ) doesn't work.

Re-init the condvar after memcpy'ing so that the TAILQ is valid.

regress/sys/kern/lockf now completes rather than failing in cv_wait_sig
with KASSERT(cv_has_waiters()).

Improve the debug output ever so slightly.

branches: 1.65.4;
PR kern/38355 lockf deadlock detection is broken after vmlocking

- Fix it; tested with Sun's libMicro.
- Use pool_cache.
- Use a global lock, so the deadlock detection code is safer.

branches: 1.64.2; 1.64.4;
Merge proc::p_mutex and proc::p_smutex into a single adaptive mutex, since
we no longer need to guard against access from hardware interrupt handlers.

Additionally, if cloning a process with CLONE_SIGHAND, arrange to have the
child process share the parent's lock so that signal state may be kept in
sync. Partially addresses PR kern/37437.

branches: 1.63.2;
need <sys/atomic.h> now.

- Replace uihashtbl_lock and struct uidinfo::ui_lock with atomic operations.
This make uid_find(), chgproccnt(), chgsbsize() and lf_alloc(), lf_free()
functions lock-less.
- Increase the size of uihashtbl in case of MP system, as suggested by <ad>.
- Add HASH_SLIST type for hashinit().

Reviewed by <ad>.

branches: 1.61.2; 1.61.6;
Merge vmlocking2 to head.

branches: 1.60.8; 1.60.14; 1.60.16; 1.60.20;
Merge some of the less invasive changes from the vmlocking branch:

- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements

branches: 1.59.2;
Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.

branches: 1.58.2;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.57.2;
Merge newlock2 to head.

branches: 1.56.2;
Fix all the -D*DEBUG* code that it was rotting away and did not even compile.
Mostly from Arnaud Lacombe, many thanks!

Use the LWP cached credentials where sane.

remove NOLOCKF and use normal NULL instead.

remove an debug printf slipped into the previous.

fix F_SETLKW deadlock detection, which has been broken since lwpify.
although this doesn't work for processes with multiple lwps, it's better
than not working at all.

make lockfpool static.

move lockf implementation details from sys/lockf.h to kern/vfs_lockf.c.

branches: 1.49.2;
integrate kauth.

branches: 1.48.4; 1.48.6; 1.48.8; 1.48.10; 1.48.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

merge ktrace-lwp.

use ap->a_flags instead of fl->f_type; pointed out by: Devon H. O'Dell

branches: 1.45.2; 1.45.4;
Re-arrange slighty, eliminate prototypes that are unnecessary when using
ANSI function decls with "static".

- add const.
- remove unnecessary casts.
- add __UNCONST casts and mark them with XXXUNCONST as necessary.

No 0x in front of %p...

don't unlock twice.

lock all uses of uidhash. provide macros to lock and unlock. based on more
discussions with yamt.

PR/30154: YAMAMOTO Takashi: tcp_close locking botch
chgsbsize() as mentioned in the PR can be called from an interrupt context
via tcp_close(). Avoid calling uid_find() in chgsbsize().
- Instead of storing so_uid in struct socketvar, store *so_uidinfo
- Add a simple lock to struct uidinfo.

Don't free the wrong lock. Pointed out by Konrad Schroder

Keep track and limit the number of locks per uid. From OpenBSD

nuke trailing whitespace

branches: 1.36.4; 1.36.6;
Fix a typo, dependancy -> dependency

ok wiz@

Initialise (most) pools from a link set instead of explicit calls
to pool_init. Untouched pools are ones that either in arch-specific
code, or aren't initialiased during initial system startup.

Convert struct session, ucred and lockf to pools.

fix uninitialized variable

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

branches: 1.32.2;
revert rev.1.28. (free -> FREE changes)
pointed by David Laight on source-changes about similar changes.

Make gcc3 happy ("label at end of compound statement").

get pointer to v_interlock directly
as we are not interested in the rest of the vnode here.

- KNF.
- remove unneeded caddr_t casts.

use FREE instead of free for MALLOC'ed memory.

MP locking.

remove a part of comment that isn't true anymore.

ANSIfy.

don't export lockf internal functions.

DANGER WILL ROBINSON!

We cannot store LWP pointers permanently in lock structures, for two reasons:
1) They are somewhat ephemeral. Dangling pointers are bad.
2) A different LWP may issue the unlock, and in this case, we were not actually
doing the unlock at all. This was causing processes to exit without undoing
fcntl(2) locks. Furthermore, the locks are process-specific to begin with,
so the test was just plain wrong.

Instead, we go back to storing a proc pointer for POSIX locks. In addition, we
add an extra pointer to the LWP, which is used in deadlock detection. After
the lock is granted, this pointer is 0ed and there is no reference to the LWP.

Now evolution can inc my mail again.

Add extensible malloc types, adapted from FreeBSD. This turns
malloc types into a structure, a pointer to which is passed around,
instead of an int constant. Allow the limit to be adjusted when the
malloc type is defined, or with a function call, as suggested by
Jonathan Stone.

Merge the nathanw_sa branch.

"ovcase" is set but not used in lf_getblock(); remove it.

Use the queue macros from <sys/queue.h> instead of referring to the queue
members directly. Use *_FOREACH whenever possible.

add RCSIDs

branches: 1.17.2; 1.17.4; 1.17.8;
change the lf_advlock() arguments from

int lf_advlock __P((struct lockf **,
off_t, caddr_t, int, struct flock *, int));
to

int lf_advlock __P((struct vop_advlock_args *, struct lockf **, off_t));

This matches common usage and is also compatible with similar change
in FreeBSD (though they use u_quad_t as last arg).

branches: 1.16.2;
Fix several problems with lockf/fcntl byte range locks:
- document a data structure invariant in lockf.h
- add KASSERT() to check the invariant.
- be more consistent about dequeuing ourselves from the blocked list
after a tsleep().
- Fix two places where the invariant is violated.
- correct a few comments here and there
- If we're still following a lock dependancy chain after maxlockdepth
processes and haven't gotten back to the start, assume that we're in a
cycle anyway and return EDEADLK.

Fix is a superset of an existing fix in FreeBSD, but independantly
derived.

Fixes kern/3860.

branches: 1.15.2;
Get rid of register declarations.

branches: 1.14.6; 1.14.12;
Abolition of bcopy, ovbcopy, bcmp, and bzero, phase one.
bcopy(x, y, z) -> memcpy(y, x, z)
ovbcopy(x, y, z) -> memmove(y, x, z)
bcmp(x, y, z) -> memcmp(x, y, z)
bzero(x, y) -> memset(x, 0, y)

fix sizeofs so they comply with the KNF style guide. yes, it is pedantic.

branches: 1.12.2;
Merge with Lite2 + local changes

make it compile with -DLOCKF_DEBUG

Moved the `unlocking a lock-free inode' shortcut behind
the l_whence and starting offset sanity checks.

backout previous kprintf change

printf -> kprintf, sprintf -> ksprintf

First pass at prototyping

Use %p.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite, plus some local changes.

stupidity for prototypes...

some prototype cleanup, eliminate/replace bogus types (e.g. quad and
u_quad) -> use better types (e.g. quad_t & u_quad_t in inodes),
some cleanup.

Make FFS optional

Back out recent change to replace pool_cache with then general allocator.
Will return to this when I have time again.

- Do away with separate pool_cache for some kernel objects that have no special
requirements and use the general purpose allocator instead. On one of my
test systems this makes for a small (~1%) but repeatable reduction in system
time during builds presumably because it decreases the kernel's cache /
memory bandwidth footprint a little.
- vfs_lockf: cache a pointer to the uidinfo and put mutex in the data segment.

kern/vfs_lockf.c: Parenthesize to make arithmetic match check.

We verified start + (fl->fl_len - 1) would not overflow, but then
computed (start + fl->fl_len) - 1 instead, and it is possible for
start + fl->fl_len to overflow before we subtract 1.

Reported-by: syzbot+762480b00cb14085f63a@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=d69313b2460a12715315f9f5f74fbe44b8bc38ba

kern/vfs_lockf.c: Fix overflow in overflow detection.

Reported-by: syzbot+cda9440741a1516747c8@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=030eb71324790093d467799263cd0789e5097229

vfs(9): Avoid arithmetic overflow in lf_advlock.

syzbot+897abbbe59467cbf6e98@syzkaller.appspotmail.com

fix various typos in comments and log messages.

Tie the maximum file lock per unprivilegied uid to kern.maxfiles

This makes the limit simple to raise at run time. While there, document
that fcntl(2) and flock(2) may return ENOMEM when this limit is reached.

branches: 1.73.60; 1.73.72;
Fix build with LOCKF_DEBUG on LP64 platforms; from Alexander Nasonov
in PR 44490.

branches: 1.72.4; 1.72.6; 1.72.8;
lockf() passes its arguments through to fcntl() but is supposed to
support -ve lengths (lock area before current offset).
Nothing in libc or the kernel allowed for this, so some random part
of the file would get locked (no idea which bits).
Although this could probably be fixed in libc, the stubs for posix file
locks for emulations could easily get into the kernel with -ve lengths.
So fixing in the kernel avoids those problems.
This also fixes PR/41620 (attempting to lock negative offsets) - which
is what I was looking into!

don't make F_GETLK or the common case of F_UNLCK fail for per-user limit.

lf_split: cv_destroy a condvar before clobbering it.

branches: 1.69.4; 1.69.8; 1.69.10;
Move uidinfo to its own module in kern_uidinfo.c and include in rump.
No functional change to uidinfo.

lf_setlock: g/c dead code.

memcpy'ing a sleepq (TAILQ) doesn't work.

Re-init the condvar after memcpy'ing so that the TAILQ is valid.

regress/sys/kern/lockf now completes rather than failing in cv_wait_sig
with KASSERT(cv_has_waiters()).

Improve the debug output ever so slightly.

branches: 1.65.4;
PR kern/38355 lockf deadlock detection is broken after vmlocking

- Fix it; tested with Sun's libMicro.
- Use pool_cache.
- Use a global lock, so the deadlock detection code is safer.

branches: 1.64.2; 1.64.4;
Merge proc::p_mutex and proc::p_smutex into a single adaptive mutex, since
we no longer need to guard against access from hardware interrupt handlers.

Additionally, if cloning a process with CLONE_SIGHAND, arrange to have the
child process share the parent's lock so that signal state may be kept in
sync. Partially addresses PR kern/37437.

branches: 1.63.2;
need <sys/atomic.h> now.

- Replace uihashtbl_lock and struct uidinfo::ui_lock with atomic operations.
This make uid_find(), chgproccnt(), chgsbsize() and lf_alloc(), lf_free()
functions lock-less.
- Increase the size of uihashtbl in case of MP system, as suggested by <ad>.
- Add HASH_SLIST type for hashinit().

Reviewed by <ad>.

branches: 1.61.2; 1.61.6;
Merge vmlocking2 to head.

branches: 1.60.8; 1.60.14; 1.60.16; 1.60.20;
Merge some of the less invasive changes from the vmlocking branch:

- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements

branches: 1.59.2;
Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.

branches: 1.58.2;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.57.2;
Merge newlock2 to head.

branches: 1.56.2;
Fix all the -D*DEBUG* code that it was rotting away and did not even compile.
Mostly from Arnaud Lacombe, many thanks!

Use the LWP cached credentials where sane.

remove NOLOCKF and use normal NULL instead.

remove an debug printf slipped into the previous.

fix F_SETLKW deadlock detection, which has been broken since lwpify.
although this doesn't work for processes with multiple lwps, it's better
than not working at all.

make lockfpool static.

move lockf implementation details from sys/lockf.h to kern/vfs_lockf.c.

branches: 1.49.2;
integrate kauth.

branches: 1.48.4; 1.48.6; 1.48.8; 1.48.10; 1.48.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

merge ktrace-lwp.

use ap->a_flags instead of fl->f_type; pointed out by: Devon H. O'Dell

branches: 1.45.2; 1.45.4;
Re-arrange slighty, eliminate prototypes that are unnecessary when using
ANSI function decls with "static".

- add const.
- remove unnecessary casts.
- add __UNCONST casts and mark them with XXXUNCONST as necessary.

No 0x in front of %p...

don't unlock twice.

lock all uses of uidhash. provide macros to lock and unlock. based on more
discussions with yamt.

PR/30154: YAMAMOTO Takashi: tcp_close locking botch
chgsbsize() as mentioned in the PR can be called from an interrupt context
via tcp_close(). Avoid calling uid_find() in chgsbsize().
- Instead of storing so_uid in struct socketvar, store *so_uidinfo
- Add a simple lock to struct uidinfo.

Don't free the wrong lock. Pointed out by Konrad Schroder

Keep track and limit the number of locks per uid. From OpenBSD

nuke trailing whitespace

branches: 1.36.4; 1.36.6;
Fix a typo, dependancy -> dependency

ok wiz@

Initialise (most) pools from a link set instead of explicit calls
to pool_init. Untouched pools are ones that either in arch-specific
code, or aren't initialiased during initial system startup.

Convert struct session, ucred and lockf to pools.

fix uninitialized variable

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

branches: 1.32.2;
revert rev.1.28. (free -> FREE changes)
pointed by David Laight on source-changes about similar changes.

Make gcc3 happy ("label at end of compound statement").

get pointer to v_interlock directly
as we are not interested in the rest of the vnode here.

- KNF.
- remove unneeded caddr_t casts.

use FREE instead of free for MALLOC'ed memory.

MP locking.

remove a part of comment that isn't true anymore.

ANSIfy.

don't export lockf internal functions.

DANGER WILL ROBINSON!

We cannot store LWP pointers permanently in lock structures, for two reasons:
1) They are somewhat ephemeral. Dangling pointers are bad.
2) A different LWP may issue the unlock, and in this case, we were not actually
doing the unlock at all. This was causing processes to exit without undoing
fcntl(2) locks. Furthermore, the locks are process-specific to begin with,
so the test was just plain wrong.

Instead, we go back to storing a proc pointer for POSIX locks. In addition, we
add an extra pointer to the LWP, which is used in deadlock detection. After
the lock is granted, this pointer is 0ed and there is no reference to the LWP.

Now evolution can inc my mail again.

Add extensible malloc types, adapted from FreeBSD. This turns
malloc types into a structure, a pointer to which is passed around,
instead of an int constant. Allow the limit to be adjusted when the
malloc type is defined, or with a function call, as suggested by
Jonathan Stone.

Merge the nathanw_sa branch.

"ovcase" is set but not used in lf_getblock(); remove it.

Use the queue macros from <sys/queue.h> instead of referring to the queue
members directly. Use *_FOREACH whenever possible.

add RCSIDs

branches: 1.17.2; 1.17.4; 1.17.8;
change the lf_advlock() arguments from

int lf_advlock __P((struct lockf **,
off_t, caddr_t, int, struct flock *, int));
to

int lf_advlock __P((struct vop_advlock_args *, struct lockf **, off_t));

This matches common usage and is also compatible with similar change
in FreeBSD (though they use u_quad_t as last arg).

branches: 1.16.2;
Fix several problems with lockf/fcntl byte range locks:
- document a data structure invariant in lockf.h
- add KASSERT() to check the invariant.
- be more consistent about dequeuing ourselves from the blocked list
after a tsleep().
- Fix two places where the invariant is violated.
- correct a few comments here and there
- If we're still following a lock dependancy chain after maxlockdepth
processes and haven't gotten back to the start, assume that we're in a
cycle anyway and return EDEADLK.

Fix is a superset of an existing fix in FreeBSD, but independantly
derived.

Fixes kern/3860.

branches: 1.15.2;
Get rid of register declarations.

branches: 1.14.6; 1.14.12;
Abolition of bcopy, ovbcopy, bcmp, and bzero, phase one.
bcopy(x, y, z) -> memcpy(y, x, z)
ovbcopy(x, y, z) -> memmove(y, x, z)
bcmp(x, y, z) -> memcmp(x, y, z)
bzero(x, y) -> memset(x, 0, y)

fix sizeofs so they comply with the KNF style guide. yes, it is pedantic.

branches: 1.12.2;
Merge with Lite2 + local changes

make it compile with -DLOCKF_DEBUG

Moved the `unlocking a lock-free inode' shortcut behind
the l_whence and starting offset sanity checks.

backout previous kprintf change

printf -> kprintf, sprintf -> ksprintf

First pass at prototyping

Use %p.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite, plus some local changes.

stupidity for prototypes...

some prototype cleanup, eliminate/replace bogus types (e.g. quad and
u_quad) -> use better types (e.g. quad_t & u_quad_t in inodes),
some cleanup.

Make FFS optional

kern/vfs_lockf.c: Parenthesize to make arithmetic match check.

We verified start + (fl->fl_len - 1) would not overflow, but then
computed (start + fl->fl_len) - 1 instead, and it is possible for
start + fl->fl_len to overflow before we subtract 1.

Reported-by: syzbot+762480b00cb14085f63a@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=d69313b2460a12715315f9f5f74fbe44b8bc38ba

kern/vfs_lockf.c: Fix overflow in overflow detection.

Reported-by: syzbot+cda9440741a1516747c8@syzkaller.appspotmail.com
https://syzkaller.appspot.com/bug?id=030eb71324790093d467799263cd0789e5097229

vfs(9): Avoid arithmetic overflow in lf_advlock.

syzbot+897abbbe59467cbf6e98@syzkaller.appspotmail.com

fix various typos in comments and log messages.

Tie the maximum file lock per unprivilegied uid to kern.maxfiles

This makes the limit simple to raise at run time. While there, document
that fcntl(2) and flock(2) may return ENOMEM when this limit is reached.

branches: 1.73.60; 1.73.72;
Fix build with LOCKF_DEBUG on LP64 platforms; from Alexander Nasonov
in PR 44490.

branches: 1.72.4; 1.72.6; 1.72.8;
lockf() passes its arguments through to fcntl() but is supposed to
support -ve lengths (lock area before current offset).
Nothing in libc or the kernel allowed for this, so some random part
of the file would get locked (no idea which bits).
Although this could probably be fixed in libc, the stubs for posix file
locks for emulations could easily get into the kernel with -ve lengths.
So fixing in the kernel avoids those problems.
This also fixes PR/41620 (attempting to lock negative offsets) - which
is what I was looking into!

don't make F_GETLK or the common case of F_UNLCK fail for per-user limit.

lf_split: cv_destroy a condvar before clobbering it.

branches: 1.69.4; 1.69.8; 1.69.10;
Move uidinfo to its own module in kern_uidinfo.c and include in rump.
No functional change to uidinfo.

lf_setlock: g/c dead code.

memcpy'ing a sleepq (TAILQ) doesn't work.

Re-init the condvar after memcpy'ing so that the TAILQ is valid.

regress/sys/kern/lockf now completes rather than failing in cv_wait_sig
with KASSERT(cv_has_waiters()).

Improve the debug output ever so slightly.

branches: 1.65.4;
PR kern/38355 lockf deadlock detection is broken after vmlocking

- Fix it; tested with Sun's libMicro.
- Use pool_cache.
- Use a global lock, so the deadlock detection code is safer.

branches: 1.64.2; 1.64.4;
Merge proc::p_mutex and proc::p_smutex into a single adaptive mutex, since
we no longer need to guard against access from hardware interrupt handlers.

Additionally, if cloning a process with CLONE_SIGHAND, arrange to have the
child process share the parent's lock so that signal state may be kept in
sync. Partially addresses PR kern/37437.

branches: 1.63.2;
need <sys/atomic.h> now.

- Replace uihashtbl_lock and struct uidinfo::ui_lock with atomic operations.
This make uid_find(), chgproccnt(), chgsbsize() and lf_alloc(), lf_free()
functions lock-less.
- Increase the size of uihashtbl in case of MP system, as suggested by <ad>.
- Add HASH_SLIST type for hashinit().

Reviewed by <ad>.

branches: 1.61.2; 1.61.6;
Merge vmlocking2 to head.

branches: 1.60.8; 1.60.14; 1.60.16; 1.60.20;
Merge some of the less invasive changes from the vmlocking branch:

- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements

branches: 1.59.2;
Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.

branches: 1.58.2;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.57.2;
Merge newlock2 to head.

branches: 1.56.2;
Fix all the -D*DEBUG* code that it was rotting away and did not even compile.
Mostly from Arnaud Lacombe, many thanks!

Use the LWP cached credentials where sane.

remove NOLOCKF and use normal NULL instead.

remove an debug printf slipped into the previous.

fix F_SETLKW deadlock detection, which has been broken since lwpify.
although this doesn't work for processes with multiple lwps, it's better
than not working at all.

make lockfpool static.

move lockf implementation details from sys/lockf.h to kern/vfs_lockf.c.

branches: 1.49.2;
integrate kauth.

branches: 1.48.4; 1.48.6; 1.48.8; 1.48.10; 1.48.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

merge ktrace-lwp.

use ap->a_flags instead of fl->f_type; pointed out by: Devon H. O'Dell

branches: 1.45.2; 1.45.4;
Re-arrange slighty, eliminate prototypes that are unnecessary when using
ANSI function decls with "static".

- add const.
- remove unnecessary casts.
- add __UNCONST casts and mark them with XXXUNCONST as necessary.

No 0x in front of %p...

don't unlock twice.

lock all uses of uidhash. provide macros to lock and unlock. based on more
discussions with yamt.

PR/30154: YAMAMOTO Takashi: tcp_close locking botch
chgsbsize() as mentioned in the PR can be called from an interrupt context
via tcp_close(). Avoid calling uid_find() in chgsbsize().
- Instead of storing so_uid in struct socketvar, store *so_uidinfo
- Add a simple lock to struct uidinfo.

Don't free the wrong lock. Pointed out by Konrad Schroder

Keep track and limit the number of locks per uid. From OpenBSD

nuke trailing whitespace

branches: 1.36.4; 1.36.6;
Fix a typo, dependancy -> dependency

ok wiz@

Initialise (most) pools from a link set instead of explicit calls
to pool_init. Untouched pools are ones that either in arch-specific
code, or aren't initialiased during initial system startup.

Convert struct session, ucred and lockf to pools.

fix uninitialized variable

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

branches: 1.32.2;
revert rev.1.28. (free -> FREE changes)
pointed by David Laight on source-changes about similar changes.

Make gcc3 happy ("label at end of compound statement").

get pointer to v_interlock directly
as we are not interested in the rest of the vnode here.

- KNF.
- remove unneeded caddr_t casts.

use FREE instead of free for MALLOC'ed memory.

MP locking.

remove a part of comment that isn't true anymore.

ANSIfy.

don't export lockf internal functions.

DANGER WILL ROBINSON!

We cannot store LWP pointers permanently in lock structures, for two reasons:
1) They are somewhat ephemeral. Dangling pointers are bad.
2) A different LWP may issue the unlock, and in this case, we were not actually
doing the unlock at all. This was causing processes to exit without undoing
fcntl(2) locks. Furthermore, the locks are process-specific to begin with,
so the test was just plain wrong.

Instead, we go back to storing a proc pointer for POSIX locks. In addition, we
add an extra pointer to the LWP, which is used in deadlock detection. After
the lock is granted, this pointer is 0ed and there is no reference to the LWP.

Now evolution can inc my mail again.

Add extensible malloc types, adapted from FreeBSD. This turns
malloc types into a structure, a pointer to which is passed around,
instead of an int constant. Allow the limit to be adjusted when the
malloc type is defined, or with a function call, as suggested by
Jonathan Stone.

Merge the nathanw_sa branch.

"ovcase" is set but not used in lf_getblock(); remove it.

Use the queue macros from <sys/queue.h> instead of referring to the queue
members directly. Use *_FOREACH whenever possible.

add RCSIDs

branches: 1.17.2; 1.17.4; 1.17.8;
change the lf_advlock() arguments from

int lf_advlock __P((struct lockf **,
off_t, caddr_t, int, struct flock *, int));
to

int lf_advlock __P((struct vop_advlock_args *, struct lockf **, off_t));

This matches common usage and is also compatible with similar change
in FreeBSD (though they use u_quad_t as last arg).

branches: 1.16.2;
Fix several problems with lockf/fcntl byte range locks:
- document a data structure invariant in lockf.h
- add KASSERT() to check the invariant.
- be more consistent about dequeuing ourselves from the blocked list
after a tsleep().
- Fix two places where the invariant is violated.
- correct a few comments here and there
- If we're still following a lock dependancy chain after maxlockdepth
processes and haven't gotten back to the start, assume that we're in a
cycle anyway and return EDEADLK.

Fix is a superset of an existing fix in FreeBSD, but independantly
derived.

Fixes kern/3860.

branches: 1.15.2;
Get rid of register declarations.

branches: 1.14.6; 1.14.12;
Abolition of bcopy, ovbcopy, bcmp, and bzero, phase one.
bcopy(x, y, z) -> memcpy(y, x, z)
ovbcopy(x, y, z) -> memmove(y, x, z)
bcmp(x, y, z) -> memcmp(x, y, z)
bzero(x, y) -> memset(x, 0, y)

fix sizeofs so they comply with the KNF style guide. yes, it is pedantic.

branches: 1.12.2;
Merge with Lite2 + local changes

make it compile with -DLOCKF_DEBUG

Moved the `unlocking a lock-free inode' shortcut behind
the l_whence and starting offset sanity checks.

backout previous kprintf change

printf -> kprintf, sprintf -> ksprintf

First pass at prototyping

Use %p.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite, plus some local changes.

stupidity for prototypes...

some prototype cleanup, eliminate/replace bogus types (e.g. quad and
u_quad) -> use better types (e.g. quad_t & u_quad_t in inodes),
some cleanup.

Make FFS optional