used in an ad hoc way by a couple of eval board ports, so might as well
tidy it up a little and add some formality. (And, yes, I need to use it
in another eval board port.)
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
whether there is anything to do - almost as if it were a predicate
test outside of a condition wait. This prevents returning to userland
when tsleep() has woken up spuriously, such as from a signal that was
caught and then removed by a tracing process.
Kills off some double-stops in GDB due to signals as well as a couple
of pthread__idle assertions when detaching from a process.
XXX stopping inside tsleep, via CURSIG(), is evil.
is set to 0, the function will always return 0 (no packets/events
are permitted)." Before this patch, ppsratecheck returned 1 once
a second when maxpps was 0.
1. sa_len was not properly checked.
2. sa_family was not properly checked [even used as an array index!]
3. we only know about inet4 and inet6, so make sure that the corresponding
data is valid before using it.
4. keep reference counts of addresses used (is that necessary?)
out the writing of an lwp's registers to a separate function. XXX Although
not really the correct way to do this, make the thread that caused the
coredump has it's register set written first so GDB is happy. (this is a
bridge until TRT is done).
- Under chroot it displays only the visible filesystems with appropriate paths.
- The statfs f_mntonname gets adjusted to contain the real path from root.
- While was there, fixed a bug in ext2fs, locking problems with vfs_getfsstat(),
and factored out some of the vfsop statfs() code to copy_statfs_info(). This
fixes the problem where some filesystems forgot to set fsid.
- Made coda look more like a normal fs.
* m_apply(), which applies a function to each mbuf in chain
starting at a specified offset, for a specified length.
* m_getptr(), which returns a pointer to the mbuf, as well as
the offset into that mbuf, corresponding to an offset from
the beginning of an mbuf chain.
From OpenBSD, cleaned up slightly by me.
in the mbuf header.
* Use the new cached paddr feature of the pool_cache API to record
the physical address of mbuf clusters. (We cannot use a ctor for
clusters, since clusters have no constructed form; they are merely
buffers).
Bus_dma back-ends may use the cached physical addresses to save having to
extract the physical address from virtual.
* Provide space in m_ext recording the vm_page *'s for an SOSEND_LOAN_CHUNK-
sized non-cluster external buffer. Use this in the sosend_loan code to
save having to extract the physical address from virtual and then look
up the vm_page *'s.
* Provide an indication that an external buffer is mapped read-only at
the MMU. Set this flag for the external buffer in the sosend_loan
case, since loaned pages are always mapped read-only. Bus_dma back-ends
may use this information to save cache flushing, since a cache flush of
a read-only mapping is redundant on some architectures (the cache would
have already been flushed when making the mapping read-only).
Part 2 in a series of simple patches contributed by Wasabi Systems
to improve network performance.
objects. Clients of the pool_cache API must consistently use
the "paddr" variants or not, otherwise behavior is undefined.
Enable this on Alpha, ARM, MIPS, and x86. Other platforms must
define POOL_VTOPHYS() in the appropriate manner in order to enable
the feature.
Part 1 of a series of simple patches contributed by Wasabi Systems
to improve network performance.
checking against nprocs is wrong in any case btw - we do allow
maxproc higher than number of current processes, it would just mean
no new process could be started until number of processes would
be lower than the new limit
Avoids a lot of casting and removes the need for some line breaks.
Removed a load of (caddr_t) casts from calls to copyin/copyout as well.
(approved by christos - he has a plan to remove caddr_t...)
- #ifdef out a DIAGNOSTIC printf() that was too annoying (rule of thumb,
don't make DIAGNOSTIC printfs() that print *very* frequently...)
- fix DIAGNOSTIC test that would always get triggered on a new session.
lookup and allocation, and any dependency on NPROC or MAXUSERS.
NO_PID changed to -1 (and renamed NO_PGID) to remove artificial limit
on PID_MAX.
As discussed on tech-kern.
having some #ifdef UNION code in vfs_vnops.c, introduce variable
'vn_union_readdir_hook' which is set to address of appropriate
vn_readdir() hook by union filesystem when it's loaded & mounted
Treat +ve numbers as process group ids and -ve as pids (see tcsetpgrp() in part of current session.
Treat +ve numbers as process group ids and -ve as pids - see tcsetpgrp(3).
(approved by christos)
syscall sys_timer_settime()) to take an absolute value for realtime
timers. This avoids a pair of gratiuitous conversions with the
possibility that the timer's intermediate value would be 0.0, which
would signal timer_settime() to cancel the timer.
Adjust callers of timer_settime() to compensate; catch the case where
sys_timer_settime() with an absolute time value of now and a virtual
timer would also be subtracted down to a timer-cancelling 0.0.
This should fix the bug seen in libpthread's nanosleep() where certain
applications, such as xmms, would wedge with unexpired userlevel
alarms.
people would read that list in a more timely fashion!), change the new
64-bit memory reporting sysctl nodes to report bytes. This should not
be a problem, since it's only a week old, and no applications use the
new nodes yet.
be returning because the code path that calls is will very likely call
setrunnable() again on the returned thread, leading to a panic because
the thread returned is already at LSRUN. This fixes a problem where netbsd
would panic when using gdb (5.3) on a process with multiple lwp's like this:
% gdb program
(gdb) run
^C
(gdb) quit
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.
and in libkvm. Then teach ps how to show them to you.
Also, teach ps how to show the names for all the uids, the rest of the
group numbers, and the "group access list".
this alignment would have been backward into the dataspace covered by
MAXDSIZ. Now the alignment is done forward. XXX It is expected that
in the TOPDOWN case, VM_DEFAULT_ADDRESS will make sure any address it
returns has the proper alignment for that architecure.
psection alignment. XXX If the psection alignment is greater than the page
alignment, extra pages may be mapped that will never be needed. This is
inefficient and wasteful of swap space and needs to be fixed.
VM_DEFAULT_ADDRESS from elf*_makecmds to elf*_load_file. In load_file,
actually determine ahead of time how much space will be needed and pass
that to VM_DEFAULT_ADDRESS. Now we have a relatistic starting address
so we can do the loading of psections normally with no extra topdown
code in load_psection. Also, if there is a gap in betweeen psections
zero map an inaccessible region between (just like ld.elf_so does) to
avoid inadvertant mmaps in the gap.
Do a little mbuf rework while here. Change all uses of MGET*(*, M_WAIT, *)
to m_get*(M_WAIT, *). These are not performance critical and making them
call m_get saves considerable space. Add m_clget analogue of MCLGET and
make corresponding change for M_WAIT uses.
Modify netinet, gem, fxp, tulip, nfs to support MBUFTRACE.
Begin to change netstat to use sysctl.
means that the dynamic linker gets mapped in at the top of available
user virtual memory (typically just below the stack), shared libraries
get mapped downwards from that point, and calls to mmap() that don't
specify a preferred address will get mapped in below those.
This means that the heap and the mmap()ed allocations will grow
towards each other, allowing one or the other to grow larger than
before. Previously, the heap was limited to MAXDSIZ by the placement
of the dynamic linker (and the process's rlimits) and the space
available to mmap was hobbled by this reservation.
This is currently only enabled via an *option* for the i386 platform
(though other platforms are expected to follow). Add "options
USE_TOPDOWN_VM" to your kernel config file, rerun config, and rebuild
your kernel to take advantage of this.
Note that the pmap_prefer() interface has not yet been modified to
play nicely with this, so those platforms require a bit more work
(most notably the sparc) before they can use this new memory
arrangement.
This change also introduces a VM_DEFAULT_ADDRESS() macro that picks
the appropriate default address based on the size of the allocation or
the size of the process's text segment accordingly. Several drivers
and the SYSV SHM address assignment were changed to use this instead
of each one picking their own "default".
sa_switch() invocations while exiting. Test P_SA instead of L_SA, out
of paranoia. Avoids a possible remrunqueue panic reported by Havard
Eidnes.
Release the kernel lock before calling the userret function to exit in
sigexit(). Problem noted by Paul Kranenburg.
timeout
the semantics of 'timeout' parameter differ to POSIX for the syscall
(not const, may be modified by kernel if interrupted from the wait) -
libc will provide appropriate wrapper
since sigwaitinfo(2) will be implemented as wrapper around sigtimedwait()
too, remove it's reserved slot and move sigqueue slot 'up', freeing
slot #246
(XXX sa_yieldcall() and sa_switchcall() should be combined and take
arg as the function to call, but I'm somewhat nervous about void *
vs. void (*)()).
conditions at points where it's necessary to access both the up-stream
and down-stream parts of the bi-directional pipe data structure. These
are marked `XXXSMP' in the code.
Also, since the changes are pretty invasive, there little point in keeping
all the "#ifdef FreeBSD" code around; so all of that has been stripped out.
kernel config option) that controls whether the kernel dumps to the
dump device on panic. Dumps can still be forced via the ``sync''
command from ddb. Defaults to ``on''.
* Change the semantics of proc_unstop() slightly, so that it is
responsible for making all stopped LWPs runnable, instead of
all-but-one. Return value is a LWP that can be interrupted if doing
so is necessary to take a signal. Adjust callers of proc_stop() to
the new, simpler semantics.
* When a non-continue signal is delivered to a stopped process and
there is a LWP sleeping interruptably, call setrunnable() (by way
of the 'out:' target in psignal1) instead of calling unsleep() so
that it becomes LSSTOP in issignal() and continuable by
proc_unstop(). Addresses PR kern/19990 by Martin Husemann, with
suggestions from enami tsugutomo.
This is derived from alpha/microtime.c and i386/tsc_microtime.c,
and will share with both ports.
This should fix PR port-sparc64/18452.
(approved by martin)
use 'l' for sa_yield() tsleep call; sa->sa_idle == l there
add DIAGNOSTIC printf on one more place in sa_upcall_userret() where
we kill process with SIGILL
fire close together, the second (and every other) timer would be
added to mask incorrectly - timerid value would be shifted twice,
and sa_upcall() would later kill process with SIGILL
arbitrary number of timers, both oneshot and periodic.
from FreeBSD, only adapted to NetBSD kernel API - mstohz() instead
of tvtohz(), and takes advantage of callout_schedule() in filt_timerexpire()
done by Artur Grabowski and Thomas Nordin for OpenBSD, which is more
efficient in several ways than the callwheel implementation that it is
replacing. It has been adapted to our pre-existing callout API, and
also provides the slightly more efficient (and much more intuitive)
API (adapted to the callout_*() naming scheme) that the OpenBSD version
provides.
Among other things, this shaves a bunch of cycles off rescheduling-in-
the-future a callout which is already scheduled, which the common case
for TCP timers (notably REXMT and KEEP).
The API has been simplified a bit, as well. The (very confusing to
a good many people) "ACTIVE" state for callouts has gone away. There
is now only "PENDING" (scheduled to fire in the future) and "EXPIRED"
(has fired, and the function called).
Kernel version bump not done; we'll ride the 1.6N bump that happened
with the malloc(9) change.
mechanism by keeping a list (bitset) of which timers have fired and using
that list in the upcall (Does this sound familiar? SEND HELP NEED SIGINFO).
Provoke the idle LWP into running again with setrunnable(sa->sa_idle)
instead of a wakeup() call, since we know what it is.
being woken up by the the reaper when a child process is cleaned up
(SIGCHLD will still cause this to run, and threads actually waiting
for the child will still see the wakeup, of course).
Should fix various spurious wakeups that manifest as assertion
failures in pthread__idle().
in the device is either nul or a digit. this avoids "raid0" being
matched as the "ra" device (and thus failing to find anything at all
causing my raid0 root to fail) on my vax.
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.
(1) ELFNAME(load_file)() now takes a pointer to the entry point
offset, instead of taking a pointer to the entry point itself. This
allows proper adjustment of the ultimate entry point at a higher level
if the object containing the entry point is moved before the exec is
finished.
(2) Introduce VMCMD_FIXED, which means the address at which a given
vmcmd describes a mapping is fixed (ie, should not be moved). Don't
set this for entries pertaining to ld.so.
Also some minor comment/whitespace tweaks.
as ltsleep() may call callout_reset() with the scheduler lock held.
So, prevent interrupts that may take the scheduler lock while holding
the callwheel lock.
recorded by p_nlwps) *or* if the process was a SA process. Since
cached SA LWPs aren't counted in p_nlwps, it was possible for
them to not be cleaned up and remain on the alllwp list, pointing to a
dead proc.
