MFC @248461

When pidptr was passed as NULL to pidfile_open(3), we were returning
EAGAIN/EWOULDBLOCK when another daemon was running and had the pidfile open.
We should return EEXIST in that case, fix it.

Report

When pidptr was passed as NULL to pidfile_open(3), we were returning
EAGAIN/EWOULDBLOCK when another daemon was running and had the pidfile open.
We should return EEXIST in that case, fix it.

Reported by: Dirk Engling <erdgeist@erdgeist.org>
Reviewed by: jhb, Dirk Engling <erdgeist@erdgeist.org>
MFC after: 1 week

show more ...

IFC @ r234692

sys/amd64/include/cpufunc.h
sys/amd64/include/fpu.h
sys/amd64/amd64/fpu.c
sys/amd64/vmm/vmm.c

- Add API to allow vmm FPU state init/save/restore.

FP stuff discussed with: kib

Set the O_CLOEXEC flag when opening the pidfile to avoid leaking the
file descriptor via exec(3).

Now that daemon(8) has been fixed to resolve the issue noted by trociny,
the consensus is that this

Set the O_CLOEXEC flag when opening the pidfile to avoid leaking the
file descriptor via exec(3).

Now that daemon(8) has been fixed to resolve the issue noted by trociny,
the consensus is that this change should be OK.

show more ...

IFC @231845

Sponsored by: Cisco Systems, Inc.

Using the O_CLOEXEC flag on open(2) caused the pidfile lock to be lost
when the child process execs daemon's target program thanks to flock(2)
semantics. So, we apparently have to leak the open pidfi

Using the O_CLOEXEC flag on open(2) caused the pidfile lock to be lost
when the child process execs daemon's target program thanks to flock(2)
semantics. So, we apparently have to leak the open pidfile's file
descriptor to keep the lock for the pidfile(3) functions to work properly.

Test case demonstrated by trociny:

ref8-amd64:/home/trociny% uname -r
8.2-STABLE
ref8-amd64:/home/trociny% daemon -p /tmp/sleep.pid sleep 10
ref8-amd64:/home/trociny% daemon -p /tmp/sleep.pid sleep 10
daemon: process already running, pid: 19799

kopusha:~% uname -r
10.0-CURRENT
kopusha:~% daemon -p /tmp/sleep.pid sleep 10
kopusha:~% daemon -p /tmp/sleep.pid sleep 10
kopusha:~%

show more ...

IFC @230489 (pending review).

Move struct pidfh definition into pidfile.c, and leave a forward declaration
for pidfh in libutil.h in its place.
This allows us to hide the contents of the pidfh structure, and also
allowed removal

Move struct pidfh definition into pidfile.c, and leave a forward declaration
for pidfh in libutil.h in its place.
This allows us to hide the contents of the pidfh structure, and also
allowed removal of the "#ifdef _SYS_PARAM_H" guard from around the
pidfile_* function prototypes.

Suggested by pjd.

show more ...

jilles pointed out that O_CLOEXEC could be used in the open(2) flags
rather than using fcntl(2) later, and in addition to saving a system
call, removes a possible race with fork/exec from threads or

jilles pointed out that O_CLOEXEC could be used in the open(2) flags
rather than using fcntl(2) later, and in addition to saving a system
call, removes a possible race with fork/exec from threads or signal
handlers.

show more ...

Constify arguments.

Style fixes courtesy of pjd.

Add pidfile_fileno() to obtain the file descriptor for an open
pidfile.

Set the FD_CLOEXEC flag on the open pidfile file descriptor.

Discussed with: pjd, des

IFC @ r226824

In pidfile_open(), if the pidfile is locked, but empty (PID is not stored yet)
and the caller requested other process' PID by passing non-NULL pidptr
argument, we will wait at most 100ms for the PID

In pidfile_open(), if the pidfile is locked, but empty (PID is not stored yet)
and the caller requested other process' PID by passing non-NULL pidptr
argument, we will wait at most 100ms for the PID to show up in the file and if
it won't, we will store -1 in *pidptr.

From now on, pidfile_open() function never sets errno to EAGAIN on failure.

In collaboration with: des
MFC after: 1 week

show more ...

There is no point in releasing a lock on a file which we've unlinked and
are about to close, so don't. As a bonus, pidfile_remove(3) will now
work with an fcntl(2)-based flopen(3).

When pidfile is already locked and has zero length, do not return
success and zero pid from pidfile_read(). Return EAGAIN instead. Sleep
up to three times for 5 ms while waiting for pidfile to be wri

When pidfile is already locked and has zero length, do not return
success and zero pid from pidfile_read(). Return EAGAIN instead. Sleep
up to three times for 5 ms while waiting for pidfile to be written.

mount(8) does the kill(mountpid, SIGHUP). If mountd pidfile is truncated,
that would result in the SIGHUP delivered to the mount' process group
instead of the mountd.

Found and analyzed by: Peter Holm
Tested by: Peter Holm, kris
Reviewed by: pjd
MFC after: 1 week

show more ...

Back out previous commit until I figure out why my regression test fails.

Approved by: re (kensmith)

Use fcntl(2)-style locks instead of less-portable flock(2)-style locks.

Approved by: re (kensmith)

strlcpy() may be faster than snprintf(), but it is less portable, and this
is not performance critical code anyway. Also, avoid using strlen() to
obtain information which we already have.

MFC after

strlcpy() may be faster than snprintf(), but it is less portable, and this
is not performance critical code anyway. Also, avoid using strlen() to
obtain information which we already have.

MFC after: 3 weeks

show more ...

Use flopen(3).

MFC after: 3 weeks

Remove some unused variables

use pwrite to always write at the begining of the file.. If multiple calls
to pidfile_write happen, the pidfile will have nul characters prepended
due to the cached file descriptor offset...

Review

use pwrite to always write at the begining of the file.. If multiple calls
to pidfile_write happen, the pidfile will have nul characters prepended
due to the cached file descriptor offset...

Reviewed by: scottl
MFC after: 3 days

show more ...

Add a family of functions for reliable pidfiles handling.

Idea from: jmg
Discussed on: arch@

There is no point in releasing a lock on a file which we've unlinked and
are about to close, so don't. As a bonus, pidfile_remove(3) will now
work with an fcntl(2)-based flopen(3).