to vclean(). Prevents a deadlock with two null mounts on the same
physical mount where one thread tries to vclean() a layer node and
another thread tries to vget() a layer node pointing to the same
physical node.
Fixes PR kern/50375 layerfs (nullfs) locking problem leading to livelock
sysctl data belongs with the module code. Move it from kern/init_sysctl.c
to kern/uipc_sem.c
While here, add a new sysctl variable kern.posix.semcnt (current count of
semaphores) to complement the existing kern.posix.semmax (maximum number
of semaphores).
module code from module_init() rather than waiting until after calling
exec_init(). Use a RUN_ONCE routine at entry to each sys_sem* syscall
to establish the exithook, and no longer KASSERT that the hook has
been set before removing it. (A manually loaded module can be unloaded
before any syscalls have been invoked.)
Remove the conditional calls to the various xxx_init() routines from
init_main.c - we now rely on module_init() to handle initialization.
Let each sub-component's xxx_init() routine handle its own sysctl
sub-tree initialization; this removes another set of #ifdef ugliness.
Tested both built-in and loadable versions and verified that atf
test kernel/t_sysv passes.
we can correctly clean up on process exit or fork.
Without this, firefox attaches to a shared memory segment but doesn't
detach before exit. Thus once firefox causes an autoload for sysv_ipc
it cannot be unloaded since the segment still retains references.
a kernel history, settable via the SCDEBUG_KERNHIST flag.
this requires a fairly significantly different set of messages than the
normal debug as histories are restricted:
- each message can take one literal format string and upto 4
arguments
- the arguments can not be strings if you want vmstat -u to
work (this could be fixed, and i might, as it would be nice
if we could print syscall names as well as numbers.)
introduce SCDEBUG_DEFAULT that is settable in the kernel config.
fix a problem in kernhist_dump_histories() where it would crash when a
history with no allocated entries was found.
extend kernhist_dumpmask() to handle the usbhist and scdebughist.
drive instead of its associated vfsops. Makes it more friendly, and allows
compat binaries to autoload VFS modules if needed.
sent on tech-kern@, ok christos@
p_stat is SACTIVE yet p_sflag is PS_STOPPING (while waiting for other
lwp's to stop). In that case, we don't want to adjust the parent's
p_nstopchild count.
Found by Robert Elz.
XXX Pullups to: NetBSD-7, -6{,-0,-1}, and -5{,-0,-1,-2}
without updating its p_waited value or its parent's p_nstopchild
counter. Later, we restore the original state, again without any
adjustment of the related values. This leaves a relatively short
window when the values are inconsistent and could interfere with the
proper operation of sys_wait() for the parent (if it manages to be
scheduled; it's not totally clear what, if anything, prevents
scheduling/execution of the parent).
If during this window, any of the checks being made result in an
error, we call exit1() which will eventually migrate the process's
state to SDEAD (with an intermediate transition to SDYING). At
this point the other variables get updated, and we finally restore
a consistent state.
This change updates the p_waited and parent's p_nstopchild at each
step to eliminate any windows during which the values could lead to
incorrect decisions.
Fixes PR kern/50330
Pullups will be requested for NetBSD-7, -6, -6-0, and -6-1
value, and update its parent's p_nstopchild value when marking the
process's p_stat to SSTOP. The process needed to be SACTIVE to get
here, so this transition represents an additional process for which
the parent needs to wait.
Fixes PR kern/50308
Pullups will be requested for:
NetBSD-7, -6, -6-0, -6-1, -5, -5-0, -5-1, and -5-2
of reaping the process (nor any other children), the process wil get
reparented to init. Since the state of the exiting process at this point
is SDEAD, proc_reparent() will not update either the old or new parent's
p_nstopchild counters.
This change causes both old and new parents to be properly updated.
Fixes PR kern/50300
Pullups will be requested for:
NetBSD-7, -6, -6-0, -6-1, -5, -5-0, -5-1, and -5-2
all processes to the SSTOP state. Make sure we update each process's
p_waited and the parents' p_nstopchild counters to maintain consistent
values. Should not make any real difference this late in the shutdown
process, but we should still be consistent just in case.
Fixes PR kern/50318
Pullups will be requested for:
NetBSD-7, -6, -6-0, -6-1, -5, -5-0, -5-1, and -5-2
moved to SSTOP state, not for its parent. (It is correct to update
the parent's p_nstopchild count.) If the value is not already zero,
it could prevent its parent from waiting for the process.
Fixes PR kern/50298
Pullups will be requested for:
NetBSD-7, -6, -6-0, -6-1, -5, -5-0, -5-1, and -5-2
months ago, but no one reviewed it - probably because it's not a trivial
change.
This change fixes the following bug: when loading a PaX'ed binary, the
kernel updates the PaX flag of the calling process before it makes sure
the new process is actually launched. If the kernel fails to launch the
new process, it does not restore the PaX flag of the calling process,
leaving it in an inconsistent state.
Actually, simply restoring it would be horrible as well, since in the
meantime another thread may have used the flag.
The solution is therefore: modify all the functions used by PaX so that
they take as argument the exec package instead of the lwp, and set the PaX
flag in the process *right before* launching the new process - it cannot
fail in the meantime.
libc api uses. For example syscallnames[293] = "__sigprocmask14",
altsyscallnames[293] = "sigprocmask". This is so that things like dtrace
can use the system call names everyone uses. The array is sparse; if the
names were the same (or for compat names) they are not copied and the array
has NULL for them.
