and then allocate it on demand. Rename some common symbols (__bss_start,
_edata, _end, __start_link_set_*, __stop_link_set_*) so that ".<module>"
is appended to them. This shrinks an amd64 kernel by 20KB of BSS.
of using on-stack memory, so that this wouldn't eventually cause kernel
panic if the process get swapped out and another process runs kqueue_scan()
problem pointed out in kern/24220 by Stephan Uphoff
called with every buffer written through spec_strategy().
Used by fss(4). Future file-system-internal snapshots will need them too.
Welcome to 1.6ZK
Approved by: Jason R. Thorpe <thorpej@netbsd.org>
so the last change has us comparing pages to bytes instead of pages
to buffers! The consequence was to try to free radically less memory
than UVM wanted us to -- though always at least one buffer, which is
probably why the results weren't dire.
This does suggest that buf_canrelease() could be a *lot* more
conservative about how much to release than "2 * page deficit". In
fact, serious trouble seems to ensue if it's not -- when anything
else on the system demands enough pages, we slam down to the low
water mark nd stay there. I've adjusted it to use min(page defecit,
buffer memory / 16), which still isn't quite right but seems better.
Another change: consider the case of an infinite loop that does
"tar xzf pkgsrc.tar.gz ; rm -rf pkgsrc". Each time the rm runs,
all the dead metadata will go on the AGE list -- and, until we hit
the high-water mark, stay there, at which point it may be slowly
recycled. Two adjustments seem to solve this: 1) whack buf_lotsfree()
to return 0 if there's anything on the AGE list; 2) whack buf_canrelease()
to count the memory used by the AGE list and always return at least
that much.
This basically turns the AGE list into a "delayed free" list, since we
can't entirely eliminate it as we can't free pool items from interrupt
context (e.g. from biodone()).
To consider: with the bookkeeping corrected, should buf_drain() move
back to the _end_ of the pagedaemon, and should the calculation then
try to give back at least the current defecit?
the latter is not a appropriate place to do so and it broke vfork.
- deactivate pmap before calling cpu_exit() to keep a balance of
pmap_activate/deactivate.
and uncontrolled growth.
The key fix is from Dan Carasone, who noticed that buf_canfree() was
counting in _bytes_ but freeing in _buffers_, which caused the instant
drop to lowater observed by some users.
We now control the rate of growth; the probability of getting a new
allocation is inversely proportional to the current size of the
cache. This idea is from a long-ago conversation with Kirk McKusick
and, if memory serves, was used for the file-system cache in some
other BSD variant at some point in history.
With growth and shrinkage more or less dealt with, we return the
default maximum cache size to 15%. The default _minimum_ cache size
is raised from 1/16 of the maximum cache size to 1/8, since 1/16 was
chosen when the maximum size was 30% of memory.
Finally, after observing the behaviour of the pagedaemon and the
buffer cache drainer under pathological workloads (e.g. a benchmark
that steps through 75% of available memory backwards) I have moved
the call to buf_drain() to the beginning of the pagedaemon from the
end; if the pagedaemon bogs down, it still won't get run as often
as it should, but at least this way it will see the state of the
free count and free target _before_ the scan step does its thing.
PR#23470, with minor updates by me. This is only the syscall support
from that PR, for now.
Changes: port over fix from FreeBSD for multicast address generation.
Changed bcopy to memcpy. For now, #ifdef notyet the portions of
kern_uuid.c that are meant to be used by (currently nonexistent) other
things in the kernel. Added syscall to COMPAT_FREEBSD as well, though
that's currently not useful, as any program new enough to use this call
also uses other syscalls we don't (yet) emulate.
tripping over this getting too large, and suffering other performance
problems due to the lack of good backpressure shrinking the bufcache
when other memory is required. Again, this tunable should be
revisited when the backpressure mechanism has been improved.
sysctl vm.bufcache can be used to manually tune those rare machines
that might need more than this.
See comments in rev 1.106 for more detail.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
From PR kern/13702 from Charles Carvalho. Tested on alpha and
i386 with a Laipac TF10 PPS-capable GPS. The com.c change was
copied wholesale from Charles' z8530tty.c patch.
"rv".
In sysctl_destroyv(), deal with deleting alias nodes, and pass a token
size_t to sysctl_destroy().
In sysctl_free(), check that "node" has not reached "rnode", not that
"pnode" has.
In sysctl_realloc(), don't bother setting sysctl_clen...the value is
unchanged.
- delete ktrsyscall32()
- add a check #ifdef _LP64 to do the conversion if P_32 is set to the
standard ktrsyscall()
- add a couple of similar _LP64/P_32 checks to the systrace code.
this should get systrace working for 32 bit apps as well as complete
ktrace support for "trace_enter/trace_exit" using platforms such as amd64.
XXX: systrace isn't supported on sparc64 currently... (it doesn't use
trace_enter/trace_exit, or have it's own calls to systrace_xxx()...)
suspending.
Move vfs_write_suspend() and vfs_write_resume() from kern/vfs_vnops.c
to kern/vfs_subr.c.
Change vnode write gating in ufs/ffs/ffs_softdep.c (from FreeBSD).
When vnodes are throttled in softdep_trackbufs() check for
file system suspension every 10 msecs to avoid a deadlock.
idle pool pages to be returned to the system immediately upon becoming
de-fragmented.
Also, in pool_do_put(), don't free back an idle page unless we are over
our minimum page claim.
cache) from 30% to 20%. This seems to significantly smooth the oscillation
between "almost no memory available" and "UVM free target available" caused
by the current sudden, heavy backpressure on the metadata cache. We should
revisit this again once the backpressure mechanism is better tuned; ideally,
the hard limit should almost never come into play, because the metadata
cache should gradually give back pages as buffers hit the AGE list and as
the page cache demands them, rather than giving back a big slug of pages
all at once when UVM decides it's in a hurry and fires off the page daemon.
Just how well this adjustment works is likely to vary significantly from
machine to machine depending on I/O mix, filesystem frag size, and total
memory. However, 20% seems to be quite a bit better than 30% on several
systems I've tested and is, coincidentally, more than enough to cache
the entire metadata working set of the AnonCVS server with 100 clients,
which is a useful worst-case stake in the ground...
0.5%, based on some quick measurements on a number of workstations and
small fileservers (including my home fileserver running simultaneous
builds of the NetBSD source tree and several NetBSD kernels). This
brings the hit rate on my machines from below 70% to above 90%. We
should be able to tune this as we run, by tracking the hit rate and
increasing the size of the cache if memory permits.
Some systems will still require significantly larger cache sizes. Some
ports -- notably the 64-bit ones -- probably should use more than 1% of
physmem as the default due to the larger size of struct vnode.
buf_mrelease(). Without this, though the pages are returned to the
relevant *pool*, they are never available for any other use in the
system.
Now the backpressure on the physical size of the buffer cache through
the buf_drain() call in the pagedaemon works correctly. If anything,
it may be a bit more aggressive than intended. On my 256MB system,
with vm.bufcache set to the default 30% of physmem, a kernel with this
fix can do 5 simultaneous config/makedep/builds of different NetBSD
kernels in 1313 seconds; with the "traditional" buffer cache code it
requires 1320 seconds. Running "find / -type d -exec ls -l {}" while
the build is going demonstrates that the backpressure is working
correctly: free memory oscillates slowly between close to none and
the UVM target free, and vmstat -m shows a large number of releases
for the buffer pools.
For future work: how is "bufpl" memory returned to the system? This
is not obvious to me (I must be looking in the wrong place). Also,
buf_mrelease() is also called from brelse() in some cases. Would it
be better to add a pool flag causing automatic release of full pages
as they become available (not fragmented)? Jason Thorpe proposed this
and it seems more elegant than cleaning the _entire_ pool only upon
memory pressure.
Greg Oster did a lot of the work of figuring this out. Jason proposed
the use of pool_reclaim as a way to fix it.
Split the sysctl setup routine into two routines, one for each
"subtree". Perhaps it's a little pedantic, but it's cleaner. Also,
assert that the "kern" and "vm" nodes exist.
process context ('reaper').
From within the exiting process context:
* deactivate pmap and free vmspace while we can still block
* introduce MD cpu_lwp_free() - this cleans all MD-specific context (such
as FPU state), and is the last potentially blocking operation;
all of cpu_wait(), and most of cpu_exit(), is now folded into cpu_lwp_free()
* process is now immediatelly marked as zombie and made available for pickup
by parent; the remaining last lwp continues the exit as fully detached
* MI (rather than MD) code bumps uvmexp.swtch, cpu_exit() is now same
for both 'process' and 'lwp' exit
uvm_lwp_exit() is modified to never block; the u-area memory is now
always just linked to the list of available u-areas. Introduce (blocking)
uvm_uarea_drain(), which is called to release the excessive u-area memory;
this is called by parent within wait4(), or by pagedaemon on memory shortage.
uvm_uarea_free() is now private function within uvm_glue.c.
MD process/lwp exit code now always calls lwp_exit2() immediatelly after
switching away from the exiting lwp.
g/c now unneeded routines and variables, including the reaper kernel thread
an offset between ss_sp and struct sa_stackinfo_t (located in struct
__pthread_st) when calling sa_register. The kernel increments the
sast_gen counter in struct sastack when an upcall stack is used.
libpthread increments the sasi_stackgen counter in struct
sa_stackinfo_t when an upcall stack is freed. The kernel compares the
two counters to decide if a stack is free or in use.
- add struct sa_stackinfo_t with sasi_stackgen to count stack use in
userland
- add sast_gen to struct sastack to count stack use in kernel
- add SA_FLAG_STACKINFO to enable the stackinfo_offset argument in the
sa_register syscall
- add sa_stackinfo_offset to struct sadata for offset between ss_sp
and struct sa_stackinfo_t
- add ssize_t stackinfo_offset argument to sa_register, initialize
struct sadata's sa_stackinfo_offset from it if SA_FLAG_STACKINFO is
set
- add sa_getstack, sa_getstack0, sa_stackused and sa_setstackfree
functions to find/use/free upcall stacks and use these where
appropriate
- don't record stack for upcall in sa_upcall0
- pass sau to sa_switchcall instead of l2 (l2 = curlwp in sa_switchcall)
- add sa_vp_blocker to struct sadata to pass recently blocked lwp to
sa_switchcall
- delay finding a stack for blocked upcalls to sa_switchcall
- add sa_stacknext to struct sadata pointing to next most likely free
upcall stack; also g/c sa_stackslist in struct sadata and sast_list
in struct sastack
- add L_SA_WOKEN flag: LWP is on sa_woken queue
- add L_SA_RECYCLE flag: LWP should be recycled in sa_setwoken
- replace l_upcallstack with L_SA_WOKEN/L_SA_RECYCLE/L_SA_BLOCKING
flags
- g/c now unused sast_blocker in struct sastack
- make sa_switchcall, sa_upcall0 and sa_upcall_getstate static in
kern_sa.c
- call sa_upcall_userret only once in userret
- split sa_makeupcalls out of sa_upcall_userret and use to process
the sa_upcalls queue
- on process exit: mark LWPs sleeping in saunblock interruptible; also
there are no LWPs sleeping on l->l_upcallstack anymore; also clear
sa_wokenq_head to prevent unblocked upcalls
additional changes:
- cleanup timerupcall sa_vp == curlwp check
- add check in sa_yield if we didn't block on our way here and we
wouldn't any longer be the LWP on the VP
- invalidate sa_vp_ofaultaddr after resolving pagefault
virtual memory reservation and a private pool of memory pages -- by a scheme
based on memory pools.
This allows better utilization of memory because buffers can now be allocated
with a granularity finer than the system's native page size (useful for
filesystems with e.g. 1k or 2k fragment sizes). It also avoids fragmentation
of virtual to physical memory mappings (due to the former fixed virtual
address reservation) resulting in better utilization of MMU resources on some
platforms. Finally, the scheme is more flexible by allowing run-time decisions
on the amount of memory to be used for buffers.
On the other hand, the effectiveness of the LRU queue for buffer recycling
may be somewhat reduced compared to the traditional method since, due to the
nature of the pool based memory allocation, the actual least recently used
buffer may release its memory to a pool different from the one needed by a
newly allocated buffer. However, this effect will kick in only if the
system is under memory pressure.
need the data in the mbuf later and m_clget() changes some fields
overlaid to regular mbuf data. Instead, rearange code a bit, create
data into a new allocated buffer and and use MEXTADD to attach it to
the mbuf, if the mbuf internal space is not sufficient.
This fixes a crash on sparc64 (and probably all other archs where
sizeof(int) != sizeof(struct file *)) when running
regress/sys/kern/unfdpass.
Idea for solution from Matt Thomas, with additional input from YAMAMOTO
Takashi.
being requested so that (1) the uniprocessor case and the
multiprocessor case are more similar and (2) so that we return ENOENT
when a non-existent processor is requested (which is both more
sensible and follows the general order of things anyway).
domainname. Note that there's no need to copy rnode since we're not
changing any of it, nor protecting anything from change.
Thanks to martin for initial work.
fit what it does.
The softsignal feature is used in Darwin to trace processes. When the
traced process gets a signal, this raises an exception. The debugger will
receive the exception message, use ptrace with PT_THUPDATE to pass the
signal to the child or discard it, and then it will send a reply to the
exception message, to resume the child.
With the hook at the beginnng of kpsignal2, we are in the context of the
signal sender, which can be the kill(1) command, for instance. We cannot
afford to sleep until the debugger tells us if the signal should be
delivered or not.
Therefore, the hook to generate the Mach exception must be in the traced
process context. That was we can sleep awaiting for the debugger opinion
about the signal, this is not a problem. The hook is hence located into
issignal, at the place where normally SIGCHILD is sent to the debugger,
whereas the traced process is stopped. If the hook returns 0, we bypass
thoses operations, the Mach exception mecanism will take care of notifying
the debugger (through a Mach exception), and stop the faulting thread.
exec case, as the emulation already has the ability to intercept that
with the e_proc_exec hook. It is the responsability of the emulation to
take appropriaye action about lwp_emuldata in e_proc_exec.
Patch reviewed by Christos.
to a 2-clause licence (retaining UCB clauses (1) and (2)), per PR
22409 from Joel Baker, approved by Theo de Raadt, and ratified by
myself - the only discrepancy being the handling of the original
clause 3 in src/usr.sbin/yppoll/yppoll.c.
Uses a hook in spec_strategy() to save data written from a mounted
file system to its block device and a hook in dounmount().
Not enabled by default in any kernel config.
Approved by: Frank van der Linden <fvdl@netbsd.org>
correctly, free original instead of incremented pointer, copy results for
n = -2 case too, so top shows correct stats.
Additionaly, rearange code for better readability (from Andrew).
Gone are the old kern_sysctl(), cpu_sysctl(), hw_sysctl(),
vfs_sysctl(), etc, routines, along with sysctl_int() et al. Now all
nodes are registered with the tree, and nodes can be added (or
removed) easily, and I/O to and from the tree is handled generically.
Since the nodes are registered with the tree, the mapping from name to
number (and back again) can now be discovered, instead of having to be
hard coded. Adding new nodes to the tree is likewise much simpler --
the new infrastructure handles almost all the work for simple types,
and just about anything else can be done with a small helper function.
All existing nodes are where they were before (numerically speaking),
so all existing consumers of sysctl information should notice no
difference.
PS - I'm sorry, but there's a distinct lack of documentation at the
moment. I'm working on sysctl(3/8/9) right now, and I promise to
watch out for buses.
so that a specific emulation has the oportunity to filter out some signals.
if sigfilter returns 0, then no signal is sent by kpsignal2().
There is another place where signals can be generated: trapsignal. Since this
function is already an emulation hook, no call to the sigfilter hook was
introduced in trapsignal.
This is needed to emulate the softsignal feature in COMPAT_DARWIN (signals
sent as Mach exception messages)
accepted. However, this time this behavor is not the default. Instead
it must enabled by using the LOCAL_CONNWAIT socket option on either the
connecting or accepting socket.
- always wait for unblocked upcall if we have to continue a blocked
thread.
=> removes wakeup from sys_sa_stacks when a stack is returned.
=> avoids extra sa_unblockyield syscall when unblocked upcall is
delivered before blocked upcall is processed.
=> avoids double pagefault if we continued a thread before the
pagefault was resolved.
=> avoids losing unblocked state if we continued a thread after
skipping the unblocked upcall.
- use splay tree for the pagefault check if the thread was running on
an upcall stack.
=> removes the limitation that all upcall stacks need to be
adjoining and that all upcall stacks have to be loaded with the
1st sys_sa_stacks call.
=> enables keeping information associated with a stack in the kernel
which makes it simpler to find out which LWP is using a stack.
=> allows increasing the SA_MAXNUMSTACKS without having to
allocate an array of that size.
symbols, and made it impossible for the kernel to use that value, and
correctly find symbols from LKMs.
o Allow LKM users to use DDB to debug the entry function of a LKM by
loading the symbol table with the temporary name /lkmtemp/ before calling
it, and then renaming it once we know the module name.
Approved by ragge@.
copyin() or copyout().
uvm_useracc() tells us whether the mapping permissions allow access to
the desired part of an address space, and many callers assume that
this is the same as knowing whether an attempt to access that part of
the address space will succeed. however, access to user space can
fail for reasons other than insufficient permission, most notably that
paging in any non-resident data can fail due to i/o errors. most of
the callers of uvm_useracc() make the above incorrect assumption. the
rest are all misguided optimizations, which optimize for the case
where an operation will fail. we'd rather optimize for operations
succeeding, in which case we should just attempt the access and handle
failures due to insufficient permissions the same way we handle i/o
errors. since there appear to be no good uses of uvm_useracc(), we'll
just remove it.
(1) split the single list of pages allocated to a pool into three lists:
completely full, partially full, and completely empty.
there is no longer any need to traverse any list looking for a
certain type of page.
(2) replace the 8-element hash table for out-of-page page headers
with a splay tree.
these two changes (together with the recent enhancements to the wait code)
give us linear scaling for a fork+exit microbenchmark.
mbuf chains which are recycled (e.g., ICMP reflection, loopback
interface). A consensus was reached that such recycled packets should
behave (more-or-less) the same way if a new chain had been allocated
and the contents copied to that chain.
Some packet tags may in future be marked as "persistent" (e.g., for
mandatory access controls) and should persist across such deletion.
NetBSD as yet hos no persistent tags, so m_tag_delete_nonpersistent()
just deletes all tags. This should not be relied upon.
This should fix PR 23418 which was also reported by Thomas Klausner and
Ian Fry (who also provided core dumps for analysis - thanks!).
Also g/c sa_yieldcall since it's now safe to put LWPs back into the cache.
Also return stacks in failure case.
of the sibling list so that find_stopped_child can be optimised to avoid
traversing the entire sibling list - helps when a process has a lot of
children.
- Modify locking in pfind() and pgfind() to that the caller can rely on the
result being valid, allow caller to request that zombies be findable.
- Rename pfind() to p_find() to ensure we break binary compatibility.
- Remove svr4_pfind since p_find willnow do the job.
- Modify some of the SMP locking of the proc lists - signals are still stuffed.
Welcome to 1.6ZF
specifically, don't keep a stale pointer in fd_ofiles.
it isn't needed anymore as fd allocation is now done using bitmaps.
- clean up dupfdopen() a little.
- don't call fd_used() unnecessarily.
the not-entered symbols will be found anyway but via a linear-search.
This only happens if something is wrong when linking the kernel.
Fixes problems reported on port-hp700.
Remove p_raslock and rename p_lwplock p_lock (one lock is enough).
Simplify window test when adding a ras and correct test on VM_MAXUSER_ADDRESS.
Avoid unpredictable branch in i386 locore.S
(pad fields left in struct proc to avoid kernel bump)
combined. Also prepare for adding VP repossession later.
- kern_sa.c: sa_yield/sa_switch: detect if there are pending unblocked
upcalls.
- kern_sa.c: sa_unblock_userret/sa_setwoken: queue LWPs about to invoke
an unblocked upcall on the sa_wokenq. put queued LWPs in a state where
they can be put in the cache. notify LWP on the VP about pending
upcalls.
- kern_sa.c: sa_upcall_userret: check sa_wokenq for pending upcalls,
generate unblocked upcalls with multiple event sas
- kern_sa.c: sa_vp_repossess/sa_vp_donate: g/c, restore original
sa_vp_repossess
General idea: only consider the LWP on the VP for signal delivery, all
other LWPs are either asleep or running from waking up until repossessing
the VP.
- in kern_sig.c:kpsignal2: handle all states the LWP on the VP can be in
- in kern_sig.c:proc_stop: only try to stop the LWP on the VP. All other
LWPs will suspend in sa_vp_repossess() until the VP-LWP donates the VP.
Restore original behaviour (before SA-specific hacks were added) for
non-SA processes.
- in kern_sig.c:proc_unstop: only return the LWP on the VP
- handle sa_yield as case 0 in sa_switch instead of clearing L_SA, add an
L_SA_YIELD flag
- replace sa_idle by L_SA_IDLE flag since it was either NULL or == sa_vp
Also don't output itimerfire overrun warning if the process is already
exiting.
Also g/c sa_woken because it's not used.
Also g/c some #if 0 code.
we pass via sigctx, so that it guaranteed that the memory wouldn't be
paged out at the time the signal arrives
potential problem pointed out by YAMAMOTO Takashi
generate unblocked upcalls in sa_unblock_userret(), before signal
delivery/p_userret handling in userret().
Also defer getting state for preempted upcalls because on some ports
preemption can happen between sa_unblock_userret() and sa_upcall_userret().
its state is saved:
- don't sa_putcachelwp() in sa_vp_repossess/sa_vp_donate
- only defer saving the event LWP's state
- sa_putcachelwp() after the interrupted LWP's state is saved
-obey ELF_LINK_ADDR in ELF_load_file()
-set ELF_LINK_ADDR in the probe() function if needed
-make ELF_NULL_ADDR the default, so that probe() functions dont need
to set it explicitely
-allocate buffer for interpreter name only if needed
initialized. Update the txp(4) to compensate.
- Statically initialize the TCP timer callout handles in the tcpcb
template. We still use callout_setfunc(), but that call is now much
less expensive. Add a comment that the compiler is likely to unroll
the loop (so don't sweat that it's there).
separately from the bufpages, so that it would be possible to eventually
make their limits changeable in runtime
make static all local variables which do not need to be exported to other
kernel parts
set using a pointer, to save couple bytes in struct sigctx
also fix fallout from recent lwp_wakeup() change, where we failed to properly
detect if tsleep() returned as result of lwp_wakeup() or signal outside
our wait set; could have caused problems for threaded apps using sigwait(2)
et.al.
matt