uidinfo is an SLIST.

For vmstat -mW, add a total KB consumed column and widen a couple of
other fields.

Remove extraneous "d" in a printf format string in UVM history dump
(leftover int format specifier from conversion to FMTd32?).

When operating on core files or /dev/mem when using the -M option,
use 64-bit math to calculate pool sizes. Fixes overflow errors for
pools larger than 4GB and gives the correct output with "vmstat -m"
for in use, total allocation and utilisation numbers.

Move the call to getnlist() to after we check if kvm_openfiles(3) succeeded.
Avoids a coredump when called with "vmstat -M /dev/mem".

Bump some column widths for "vmstat -mW".

vmstat: unexport file-scope variable numdisks

There is no need to make this variable externally visible. There are
several more variables in this file that could be unexported, leave
these for someone who knows whether vmstat.c is used by other parts of
the tree as well.

No functional change, OK mrg.

Fix typo in usage.

allow the number of disks displayed in the default output
to be controlled.

when picking which 2 disks to show in the 'vmstat' default output
and they haven't been specified on the command line, pick the two
devices that have had the largest read+write IO bytes count.

usr.bin: remove unnecessary CONSTCOND, lint no longer needs it

Since 2021-01-31, lint no longer requires a CONSTCOND comment in a
do-while-0 statement since this is a common code pattern, especially in
statement-like macros.

usr.bin: enable lint checks for most programs in /usr/bin

Previously, lint was only activated for the libraries, but not for the
kernel or userland programs. Activate lint for the programs in /usr/bin
for now, more will follow later.

This only affects builds that set MKLINT=yes.

Drop setgid kmem, simplify nlist setup.

Use kernel sysctl hashstat collection instead of kmem grovelling
directly. Also GC a few old hash nlist entries that no longer exist.

Now that ports that use legacy intrcnt interrupt accounting have event
counters to show that data, for the live kernel case just show INTR
events for "vmstat -i".

For vmstat -h/-H, calculate the hash element size correctly instead of
assuming that everything that isn't a list is a tailq. Fixes random
reads from kmem that either fail or return incorrect data for the vcache
hash table.

Remove PG_ZERO. It worked brilliantly on x86 machines from the mid-90s but
having spent an age experimenting with it over the last 6 months on various
machines and with different use cases it's always either break-even or a
slight net loss for me.

pool_cache:

- make all counters per-CPU and make cache layer do its work with atomic ops.
- conserve memory by caching empty groups globally.

Counter tweaks:

- Don't need to count anonpages+filepages any more; clean+unknown+dirty for
each kind of page can be summed to get the totals.

- Track the number of free pages with a counter so that it's one less thing
for the allocator to do, which opens up further options there.

- Remove cpu_count_sync_one(). It has no users and doesn't save a whole lot.
For the cheap option, give cpu_count_sync() a boolean parameter indicating
that a cached value is okay, and rate limit the updates for cached values
to hz.

ncs_collisions is gone.

Process concurrent page faults on individual uvm_objects / vm_amaps in
parallel, where the relevant pages are already in-core. Proposed on
tech-kern.

Temporarily disabled on MP architectures with __HAVE_UNLOCKED_PMAP until
adjustments are made to their pmaps.

- nchash is gone.
- Report new namecache stats.

Fix alignment of Flags column for vmstat -mW.

vmstat -s: report new stats:

0 per-cpu stats one synced
7246 per-cpu stats all synced
4092 anon pages possibly dirty
8881 anon pages dirty
0 anon pages clean
68 file pages possibly dirty
0 file pages dirty
2367889 file pages clean

Show reverse misses too.

- Report namecache reverse hits with vmstat -s.
- ncvhashtbl is no more.

move the time nlist fetches into their own namelist and only
fetch them when necessary. allow for fallback uses of older
time sources if others are not present.

this stops vmstat from exiting if it can't get the addresses
of these time values it often doesn't need (eg, running kernels
use the sysctl method), which has cropped up recently wit the
removal of boottime variable.


a slighly modified version of this patch (modified to handle
the old boottime variable over the new one) works against a
netbsd-9 vmstat in -current too.

XXX: pullup

boottime in the kernel is no more. Instead, read timebasebin and convert
from bintime to timespec.

s/sucess/success/ in comment.

Have vmstat -H report on vcache_hashtab.

Expose struct namecache.

branches: 1.227.2;
when dumping a kernhist history replace "%s" with "%p" so that the
kernhist in kernel can be a little more usable without making the
userland dumping code crash.

Use PRWORD for printing -m/-mW totals so fields don't run in to each other.

While here, update field widths for 'vmstat -mW' for modern machines.

typo

branches: 1.224.2; 1.224.4;
deref the right address to get the pool_allocator.

remove now wrong comment. minor KNF.

fix the kvm version of counting pool pages for vmstat -s.

add two new members to uvmexp_sysctl{}: bootpages and poolpages.
bootpages is set to the pages allocated via uvm_pageboot_alloc().
poolpages is calculated from the list of pools nr_pages members.

this brings us closer to having a valid total of pages known by
the system, vs actual pages originally managed.

XXX: poolpages needs some handling for PR_RECURSIVE pools still.

Remove the ABI version-and-length check that was recently introduced;
sysctl(9) ABIs should be stable across versions.

XXX Pull-up to -8

Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...

(As proposed on tech-kern@ with additional changes and enhancements.)

Details of changes:

* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)

* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.

* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.

* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."

* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.

* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).

* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).

* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.

* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.

[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3) format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".

[2] I've tried very hard to find "all [the] existing users of kernhist(9)"
but it is possible that I've missed some of them. I would be glad to
update any stragglers that anyone identifies.

Don't print "nan" when there is no disk activity.

Use I/O timestamps to compute disk statistics for better precision.

Disk statistics are collected in a fixed size array, that got corrupted
when a disk was detached. Adapt by skipping entries of detached disks
and detect reused disknames at the array end.

branches: 1.216.6;
"vmstat -ie[v]" auto fit to minimum columns

Reviewed by msaitoh@

Adapt to use of bintime(9) for kernel history timestamps (7.99.55).

Fix error handling confusion between error and errno

Adapt to new version of the sysctl export structure

Use appropriate PRI* format specifiers for fixed-bit-size numbers

Should fix i386 build.

For kvm-based kernhist code, print size of history table and "next free"
pointer, same as we do for sysctl-based code.

Restore the kvm-grovelling code for use on crash files (where sysctl(3)
is not available). Otherwise, this would always report on the current
running kernel regardless of the use of -M option.

For kernel history display, replace the kvm grovelling with the newly
available sysctl access to the history data.

XXX vmstat still uses kvm for several other displays

No point in using float here; makes evbmips64 need __truncdfsf2, and that
causes static linking issues because of softfloat.

Drop main() prototype.

branches: 1.206.2;
Update stats-keeping in sys/kern/vfs_cache.c to remove (most)
races while allowing consistent lockless sampling of the per-cpu
statistics without atomic operations. Update comment describing
the locking protocol to include this.

These files were fumble-fingered out of the last commit.

Trailing whitespace.

Print the cpu in the kernhist record.

branches: 1.203.2;
Use u_long to avoid negative value outputs.

Init summary counters.

Fix format string for ILP32.

Use kern.pool for the live system.

GC left-over kread.

Use CPU statistics from uvmexp when possible.

Use kern.clockrate if possible.

Convert the remaining uvmexp users to the sysctl as default source.

Use vm.uvmexp data as much as possible for vmstat -s.
Temporarily disable shadowing warning until the other users are cleaned
up.

Provide sysctl for namecache statistics.

Make missing kern.evcnt an error, don't try to fallback to kmem use.

Add a global vnode cache:

- vcache_get() retrieves a referenced and initialised vnode / fs node pair.
- vcache_remove() removes a vnode / fs node pair from the cache.

On cache miss vcache_get() calls new vfs operation vfs_loadvnode() to
initialise a vnode / fs node pair. This call is guaranteed exclusive,
no other thread will try to load this vnode / fs node pair.

Convert ufs/ext2fs, ufs/ffs and ufs/mfs to use this interface.

Remove now unused ufs/ufs_ihash

Discussed on tech-kern.

Welcome to 6.99.41

branches: 1.191.2;
Remove some pointless inclusions os sys/user.h

Remove cpu_queue (and thus eleminate another use of CIRCLEQ) by replacing
its uses with cpu_infos array. Extra testing by christos@.

add a row with totals for some pool statistics.

branches: 1.188.2;
remove malloc(9) output function from vmstat
it is not required and working any more

Use C89 function definitions

branches: 1.186.2;
use warn instead of perror.

unbreak -f -H -h, etc, part 1
- get nlist once in the beginning and fix the flags. Not each time! It is
not like the kernel will change it symbols in the middle of the run.

Use __arraycount(), as done above.

Mark usage as static and __dead.

dopoolcache: don't print unused caches unless verbose

move and rename the uvm history code out of uvm_stat to "kernhist".

rename "UVMHIST" option to enable the uvm histories.

TODO:
- make UVMHIST properly depend upon KERNHIST
- enable dynamic registration of histories. this is mostly just
allocating something in a bitmap, and is only for viewing multiple
histories in a merged form.


tested on amd64 and sparc64.

Fix header, format and total count of system interrupt counters.

dosum also needs getnlist

deal with detecting the first and last cpu differently.

Make sure to getnlist in dovmstat

Defer nlist processing until we know we need to it. If everything can be
obtained via sysctl, we can skip it entirely. This means we can run even
if not setgid.

getuptime will now use sysctl/clock_gettime if memf is NULL.
doevcnt now sues sysctl(kern.evcnt) is memf is NULL. It falls back to
groveling if sysctl returns an error of ENOENT.
dointr will call doevcnt to evcnt based intr stats.

branches: 1.175.2;
can't play games with structure names since cpu_data refers to cpu_info, so
provide our own copy for everything.

use a local definition of cpu_info if we have __HAVE_CPU_DATA_FIRST

Don't crash on old kernels that don't have the new per cpu counters.

We still need to update ouvmexp to output the differene from
previous iteration.

make this work with the per-cpu counters.

simplify and modernize:
- make sysctl mibs static
- use warn/warnx for warnings
makes binary 4K smaller.

Make "i" unsigned to mirror change to struct uvm_history

Follow christos' suggestions, and make ks_active a u_short, and
also only use 16 u_shorts instead of 32 ints. Also add panic()
calls for under- and overflow of the ks_active members under
DIAGNOSTIC. The MAXBUCKET constant ended up in sys/mallocvar.h
and not sys/param.h, as the latter caused build problems.

Ride the kernel revision bump of my previous change.

Extend struct malloc_type to count the number of active allocations
per size, and make vmstat report this information under the "Memory
statistics by type" display, which is only printed when the kernel
has been compiled with KMEMSTATS defined, like this:

Memory statistics by type Type Kern
Type InUse MemUse HighUse Limit Requests Limit Limit Size(s)
wapbl 15 4192K 4192K 78644K 376426 0 0 32:0,256:3,512:6,131072:1,262144:2,524288:3

Since struct malloc_type is user-visible and is changed, bump kernel
revision to 5.99.26.

While it is true that malloc(9) is in general on the path of slowly
being replaced by kmem(9) (kmem_alloc/kmem_free), there remains a
lot of points of usage of malloc/free, and this could aid in finding
any leaks. (It helped finding the leak fixed in PR#42661.)

This was discussed with and somewhat hestitantly OKed by rmind@

Remove uarea swap-out functionality:

- Addresses the issue described in PR/38828.
- Some simplification in threading and sleepq subsystems.
- Eliminates pmap_collect() and, as a side note, allows pmap optimisations.
- Eliminates XS_CTL_DATA_ONSTACK in scsipi code.
- Avoids few scans on LWP list and thus potentially long holds of proc_lock.
- Cuts ~1.5k lines of code. Reduces amd64 kernel size by ~4k.
- Removes __SWAP_BROKEN cases.

Tested on x86, mips, acorn32 (thanks <mpumford>) and partly tested on
acorn26 (thanks to <bjh21>).

Discussed on <tech-kern>, reviewed by <ad>.

fix -Wsign-compare issues

Add missing fflush() in one of three places that print stuff repeatedly
every so often. PR 32836 from Takuro KUBOTA.

branches: 1.163.6;
NFS hash tables no longer exist.

Restore the "pages actige" and "pages inactive" value printouts
in "vmstat -s" output when run on the active kernel.

The reason we can't easily provide these values out of a crash
dump, is that these fields are no longer proper members of
struct uvmexp, but rather are estimated by and dependent on
the currently active page replacement policy in the kernel.

Remove the \n and tabs from the __COPYRIGHT() strings.
Tweak to use a consistent format.

Fix alignment of fields for "vmstat -m" when KMEMSTATS are available:
o Make "prop dictionary" fit in the "Type" coloumn
o Ensure that fields don't run into each other (adds a space between fields)
o Ensure that a Limit of >100000K fits in the field width

vmstat -s: report cpu pagealloc hits/misses.

branches: 1.158.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.157.4;
Merge more changes from vmlocking2, mainly:

- Locking improvements.
- Use pool_cache for more items.

- Support two different sizes of pool_cache group. The default has 14 or 15
items, and the new large groups (for busy caches) have 62 or 63 items.
- Add PR_LARGECACHE flag as a hint that a pool_cache should use large groups.
This should be eventually be tuned at runtime.
- Report group size for vmstat -C.

add -C to usage.

Add -C option to display pool_cache info, eg:

Pool cache statistics.
Name Spin Full Empty PoolLayer CacheLayer Hit% CpuLayer Hit%
pipepl 0 1 3 385 42862 99.1 985780 95.7
fdescpl 0 0 2 310 31133 99.0 1029620 97.0
cwdipl 0 0 2 310 31227 99.0 1029714 97.0
filepl 0 1 4 680 141415 99.5 24011135 99.4

branches: 1.153.8;
warns=4, pass lint.

use howmany(), instead of open-coding the calculation.

increase a chance to fit within 80 char width consoles.
- use PRWORD for dovmstat as well.
- reduce number of drives shown by default from 3 to 2.

sync with kernel after yamt-pdpolicy merge.

PR 34205: workaround struct timeval tv_sec not being SUS compliant issue (reported by Takeshi Nakayama)

branches: 1.148.2;
Dump some column widths.

accept either "time_second" or "time" as kernel timevariable in
order to support common userland in enviroments where
timecounter kernels and non timecounter kernel co-exist
(e.g. xen & i386)

include evcnt.h rather than device.h.

support timecounters (struct time gone, use time_second then)

branches: 1.144.2;
Coverity CID 884: Avoid NULL deref.

Coverity CID 885: Avoid NULL deref.

Coverity CID 886: Avoid NULL deref.

Unwrap a not too long line.

Update statistics reporting to use new kernel framework, rename dkstats*
to a more generic name since it no longer just deals with disk.

Split dkreadstats() into dkreadstats(), tkreadstats() and cpureadstats().
Similarly for dkswap().
Allows code to only call one of these.
Rename struct _disk to _vminfo to be nearer its contents.
Change 'systat vm' so that it always shows the 'current' cpu times, even
after ':run' or ':boot' commands. The code in vmstat.c doesn't support :run.

branches: 1.138.2;
Don't forget original allocated memory address.

optimize pool_caches similarly to how I optimized pools before:
split the single list of pool cache groups into three lists:
completely full, partially full, and completely empty.
use LIST instead of TAILQ where appropriate.

Add support for tape statistics.

appease gcc -Wuninitialized

rename "kbucket" to "kmembuckets", for greater clarity.
don't forget to update vmstat this time.

do not rely on the pre-evcnt interrupt counter kernel symbols being present.
if they do not exist, then interrupt stats are all in evcnts.

remove anon related statistics which are no longer used.

Add -t to usage.

Add a '-t' option to print the contents of the vmtotal structure,
possibly repeatedly.

WARNS=3

Remove a blank line.

Sync with manpage (add -W, sort).

Add a "wide" mode output (selected with -W) to show more information
about memory pools, including pool page size, flags and utilisation.

Correct typo in an error message: allocatior -> allocator

Add (unsigned char) cast to ctype functions

Make vmstat -s prints:
- uvmexp.pdre{anon,file,exec}.
- real number of namecache stats.

pidhash and pgrphash should have been removed from here as well when the
new pid allocator was committed 3/2003.

Uppercase CPU, plural is CPUs.

fixes to keep default output under 80 columns
reduce default number of disks to 3 from 4
adjust widths of cpu stats if one of them is 100%
always put spaces between # of processes runnable (from pr 12691)

Extra columns appeared in one line of the "Memory statistics by type"
header in '94 and '97. Grow the other line a little so that it looks
pretty again.

realloc pedant

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

Patches provided by Joel Baker in PR 22365, verified by myself.

snprintf

No need to cast the third argument of ioctl() to "char *".

no need to #include <sys/dkstat.h>

Make vmstat -m skip memory pools that have never been used - unless -v given.
Update man page to match.

Update for cached physical addresses in the pool caches.

Alternative pid/proc allocater, removes all searches associated with pid
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.

The last NULL entry should not be printed on missing symbols.

Update for extensible malloc types.

The -u option takes an argument, so we need the corresponding : in the
getopt() string. Does no one use uvm histories any more, or does
everyone just dump all the histories?

Remove pc532-only hacks to access wildly different interrupt accounting
now that interrupts are tracked with evcnt(9) on this port.

Include <machine/intr.h> to get "struct iv" (in pc532-specific code).

In getuptime(), use kread() to get the current time as well as boottime,
instead of calling time(3).

This makes rate calculations of crash dumps sensible; in contrast,
subtracting the boot time of the crash dump from the current time of the
running system produces a not-very-meaningful number.

iostat(8) and systat(8) support for separate read/write disk statistics.
"iostat -x" now shows these (ala linux/solaris), but this is only splitting
the read/write bytes/transactions, not adding any new metrics. "systat
iostat" now has two new commands to switch between combined/separate mode
for both it's numbers & bar modes.

sleep() -> nanosleep()

Print newline before pool stats.

Since the KERN_CP_TIME sysctl reports total clock ticks across all
cpus and hz is measured in ticks per cpu, divide tick count by ncpu to
determine elapsed time since last sample.

Fixes I/O rate deflation observed on multiprocessors.

branches: 1.100.2;
Don't provide a local declaration of 'struct pool_cache_group' now that
it's in <sys/pool.h>.

Fix broken indentation.

fix pool_allocator fallout. bad boy jason!

Don't forget to fetch pool cache group.

When printing pool stats, print pool cache stats also if -v is given.

Use the new disk and tty stats sysctls when operating on a live kernel,
dk*() don't need to worry about setgid() games anymore.

replace "vnode" and "vtext" with "file" and "exec" in uvmexp field names.

print average chain length as %8.2f instead of %8d

KERNBASE is a constant that kernland has no business using; it's
existence and value are machine-dependent, and using KERNBASE
either breaks binary compatibility across the same MACHINE_ARCH
or prevents this program from compiling at all.

Cast the printf'ing of the 2 size_t vars in dohashstat to unsigned long long as
some ports have size_t's which are larger than int's.

- Separate hashstat namelist into separate hashnl[], and don't barf if
some of the symbols can't be found
- Only kvm_nlist() hashnl[] and histnl[] once
- Add a description to struct kernel_hash, and print with -L
- Sort entries in khashes[]
- Don't exit on unknown hashes; just display a warning and move on

add stats for nchash & ncvhash

- Use -H (instead of -h) to list all hash stats.
- Add '-h hashname' to only list the stats for hashname.
- Add -L to list valid hashes
- Fix up some whitespace

- replace -h uvmhist with -u uvmhist, and -H with -U, since uvm history stuff
isn't generally enabled by default, and i wanted -h for something else.
- implement -h, which displays statistics for various kernel hash tables
- add deref_kptr(), which does the hard work of kvm_read() et al
- various minor cleanups

Check array boundary correctly.

Print evcnt/intrcnt as unsigned.

When summarizing pool memory usage, account for PR_RECURSIVE pools
specially so as to avoid counting the same memory twice.

Adjust vmstat -i / -e for wider event names

when printing "avm fre" fields, use " %5ld %5ld " instead of "%6ld%6ld ",
because it's better to slightly right-shift the display than run numbers
together when you've got > 99999KB avm or fre.

Move the number of page colors to the top of the display.

Add the number of page colors to uvmexp.

Print print color bucket hit/miss stats.

Make -c and -w work properly when printing non-default statistics.

Show number of pages used by executables now that they're are counted
by UVM.

Prevent integer overflow.

Cosmetic changes.

Don't "extern" variables and function in dkstats.c, declare them in
dkstats.h instead.

ANSIfy.

Use a better description for the vnode page cache pages.

Show anonpages and vnodepages in "vmstat -s".

Move cpu header right 3 spaces to match grouping in man page

Make vmstat -m to put at least one space between each numbers.

Display number of zero page aborts.

remove include of <vm/vm.h>

branches: 1.66.2;
Implement the more flexiable `evcnt' interface as discussed (briefly) on
tech-kern and now documented in evcnt(9).

cp_time is no longer available in the kernel.
remove it from namelist[] to make vmstat work.

include sys/sched.h to get definition of CPUSTATES

branches: 1.63.2;
Print the zero page statistics.

Don't declare 'extern opt*' getopt variables.

Cast timeval members to types we know the printf conversions of.

More fallout from CL* purge.

Make pool output fit in 80 columns on most machines.

This program has to use KVM alrady, so use that to read uvmexp, rather
than sysctl. This can now read VM stats from crash dumps.

branches: 1.57.4;
Define __POOL_EXPOSE to get at the guts of the pool descriptor.

completely remove Mach VM support. all that is left is the all the
header files as UVM still uses (most of) these.

<sys/fcntl.h> -> <fcntl.h>

Nuke all mentions of vmstat -t. PR #6433, Geoff Wing.

bzero->memset, bcopy->memcpy, bcmp->memcmp

Display the number of idle pages per pool.

Print memory pool statistics.

const poisoning.

Fix handling of privileges - the gid was given up before dkinit()
was called.
Pass the gid to use for KVM access explicitely to dkinit(), the caller
decides whether to call setgid(getgid()) or not.

- change setgid kmem programs (that lend themselves to this) so setegid(getgid())
and the top, and then set the effective gid back to kmem around the call to
kvm_openfiles(). this reduces the time group kmem is available.
- for those above that also allow this, setgid(getgid()) after the call to
kvm_openfiles() to fully revoke priviledges.
- some KNF
- use err(3) over fprintf(3) in some places

KNF.

use the uvmexp.pageins member for "pi".

Add support for listing and dumping UVM history buffers.

print almost all of uvmexp with -s. do pagein/pageout counts correctly.
also include freed pages count (that was missing from mach vm, but present
in the "output" but hard coded to 0 in vmstat.c.

print out the contents of uvmexp for "vmstat -s"

flesh out standard vmstat output with UVM.

GC old old old "NEWVM". make run under UVM, if not report everything yet..

Revive the -f option: report fork statistics.

getopt returns -1 not EOF

WARNSify; deprecate register; fix some broken fprintf() statements found from WARNS.

merge lite-2.

Repair the formatting of the by-memory-type table of "vmstat -m"
("MSDOSFS mount" is too big for %12s).
Note that this breaks naive parser scripts, but they would not parse
the earlier "MSDOSFS mount" line correctly, anyway.

"undefined" -> "undef/%d" for memory type names. Similar, but not identical
to Ken Raeburns proposal from PR 1012, herewith closed.

If kmem stats are not being gathered by the kernel, print a useful
message to that effect. Fixes PR #3187, submitted by
Guenther Grau <s_grau@ira.uka.de> (fix by me).

Compute `halfuptime' correctly when `uptime' == 1.
From Lennart Augustsson <augustss@cs.chalmers.se>, PR #2944.

print all evcnt structures, not just those which have the name "intr".
Printing only those with the name "intr" is too restrictive, because it
means that devices can only have one interrupt, which makes little
sense for many devices and absolutely no sense for several common busses.

Remove straggling references to dk_ndrive and dk_xfer.

the kernel 'boottime' variable is a timeval. copy it out of kernel
memory as such (not as a time_t).

branches: 1.29.4;
New userland code to read the statistics kept by the NetBSD generic
disk framework, from John M. Vinopal <banshee@gabriella.resort.com>

update for the fact that 'allevents' is now a TAILQ.

print interrupt names in 14 cols. really should be dynamic.

don't forget to accumulate interrupt counts found via allevents.

fix spacing problem in interrupt/total/rate output line of vmstat -i

For the pc532, get struct iv from the protected part of <machine/psl.h>.

pc532 specific interrupt code. (From Matthias Pfaller.)

sys/signal.h -> signal.h

fix up RCS ids

cast pointers to longs, not ints.

VAX machine dependent defs removed.

Make use of the new kernel event counters; from Theo.

size is a size_t!

Fix RCS id format.

re-enable new bucket foo, also widen type name field by one for miscfs mount

rcsids

clean up import. also, stub read_names for the i386, and comment out
some refs to kmem stats fields that we don't have yet. This will need
hacking, (1) for the various ports, (2) for the kmem stuff, and won't
work completely right until those happen, and until the new vm system
is integrated.

clean, exported variable (and contents) changes

fix to work with new config. SCSI device name support is totally hosed.

get rid of __386BSD__ and replace as appropriate with __NetBSD__ and i386

some nice new things

widen "Type" field of vmstat -m by 2 characters

make various vm counters part of the "standard" nlist (for mw)

Add RCS identifiers.

clean up interrupt count printing a bit...

fix dependencies on M_LAST's old off-by-one error...
(arguably, the old error wasn't an error, but, given some
of the old usage of M_LAST and its name, i say it was)

add rcs ids, and update for "arch"

iostat and vmstat being taught about drives..

branches: 1.1.1;
added, from net/2 (patch 124).

For vmstat -mW, add a total KB consumed column and widen a couple of
other fields.

Remove extraneous "d" in a printf format string in UVM history dump
(leftover int format specifier from conversion to FMTd32?).

When operating on core files or /dev/mem when using the -M option,
use 64-bit math to calculate pool sizes. Fixes overflow errors for
pools larger than 4GB and gives the correct output with "vmstat -m"
for in use, total allocation and utilisation numbers.

Move the call to getnlist() to after we check if kvm_openfiles(3) succeeded.
Avoids a coredump when called with "vmstat -M /dev/mem".

Bump some column widths for "vmstat -mW".

vmstat: unexport file-scope variable numdisks

There is no need to make this variable externally visible. There are
several more variables in this file that could be unexported, leave
these for someone who knows whether vmstat.c is used by other parts of
the tree as well.

No functional change, OK mrg.

Fix typo in usage.

allow the number of disks displayed in the default output
to be controlled.

when picking which 2 disks to show in the 'vmstat' default output
and they haven't been specified on the command line, pick the two
devices that have had the largest read+write IO bytes count.

usr.bin: remove unnecessary CONSTCOND, lint no longer needs it

Since 2021-01-31, lint no longer requires a CONSTCOND comment in a
do-while-0 statement since this is a common code pattern, especially in
statement-like macros.

usr.bin: enable lint checks for most programs in /usr/bin

Previously, lint was only activated for the libraries, but not for the
kernel or userland programs. Activate lint for the programs in /usr/bin
for now, more will follow later.

This only affects builds that set MKLINT=yes.

Drop setgid kmem, simplify nlist setup.

Use kernel sysctl hashstat collection instead of kmem grovelling
directly. Also GC a few old hash nlist entries that no longer exist.

Now that ports that use legacy intrcnt interrupt accounting have event
counters to show that data, for the live kernel case just show INTR
events for "vmstat -i".

For vmstat -h/-H, calculate the hash element size correctly instead of
assuming that everything that isn't a list is a tailq. Fixes random
reads from kmem that either fail or return incorrect data for the vcache
hash table.

Remove PG_ZERO. It worked brilliantly on x86 machines from the mid-90s but
having spent an age experimenting with it over the last 6 months on various
machines and with different use cases it's always either break-even or a
slight net loss for me.

pool_cache:

- make all counters per-CPU and make cache layer do its work with atomic ops.
- conserve memory by caching empty groups globally.

Counter tweaks:

- Don't need to count anonpages+filepages any more; clean+unknown+dirty for
each kind of page can be summed to get the totals.

- Track the number of free pages with a counter so that it's one less thing
for the allocator to do, which opens up further options there.

- Remove cpu_count_sync_one(). It has no users and doesn't save a whole lot.
For the cheap option, give cpu_count_sync() a boolean parameter indicating
that a cached value is okay, and rate limit the updates for cached values
to hz.

ncs_collisions is gone.

Process concurrent page faults on individual uvm_objects / vm_amaps in
parallel, where the relevant pages are already in-core. Proposed on
tech-kern.

Temporarily disabled on MP architectures with __HAVE_UNLOCKED_PMAP until
adjustments are made to their pmaps.

- nchash is gone.
- Report new namecache stats.

Fix alignment of Flags column for vmstat -mW.

vmstat -s: report new stats:

0 per-cpu stats one synced
7246 per-cpu stats all synced
4092 anon pages possibly dirty
8881 anon pages dirty
0 anon pages clean
68 file pages possibly dirty
0 file pages dirty
2367889 file pages clean

Show reverse misses too.

- Report namecache reverse hits with vmstat -s.
- ncvhashtbl is no more.

move the time nlist fetches into their own namelist and only
fetch them when necessary. allow for fallback uses of older
time sources if others are not present.

this stops vmstat from exiting if it can't get the addresses
of these time values it often doesn't need (eg, running kernels
use the sysctl method), which has cropped up recently wit the
removal of boottime variable.


a slighly modified version of this patch (modified to handle
the old boottime variable over the new one) works against a
netbsd-9 vmstat in -current too.

XXX: pullup

boottime in the kernel is no more. Instead, read timebasebin and convert
from bintime to timespec.

s/sucess/success/ in comment.

Have vmstat -H report on vcache_hashtab.

Expose struct namecache.

branches: 1.227.2;
when dumping a kernhist history replace "%s" with "%p" so that the
kernhist in kernel can be a little more usable without making the
userland dumping code crash.

Use PRWORD for printing -m/-mW totals so fields don't run in to each other.

While here, update field widths for 'vmstat -mW' for modern machines.

typo

branches: 1.224.2; 1.224.4;
deref the right address to get the pool_allocator.

remove now wrong comment. minor KNF.

fix the kvm version of counting pool pages for vmstat -s.

add two new members to uvmexp_sysctl{}: bootpages and poolpages.
bootpages is set to the pages allocated via uvm_pageboot_alloc().
poolpages is calculated from the list of pools nr_pages members.

this brings us closer to having a valid total of pages known by
the system, vs actual pages originally managed.

XXX: poolpages needs some handling for PR_RECURSIVE pools still.

Remove the ABI version-and-length check that was recently introduced;
sysctl(9) ABIs should be stable across versions.

XXX Pull-up to -8

Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...

(As proposed on tech-kern@ with additional changes and enhancements.)

Details of changes:

* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)

* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.

* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.

* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."

* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.

* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).

* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).

* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.

* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.

[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3) format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".

[2] I've tried very hard to find "all [the] existing users of kernhist(9)"
but it is possible that I've missed some of them. I would be glad to
update any stragglers that anyone identifies.

Don't print "nan" when there is no disk activity.

Use I/O timestamps to compute disk statistics for better precision.

Disk statistics are collected in a fixed size array, that got corrupted
when a disk was detached. Adapt by skipping entries of detached disks
and detect reused disknames at the array end.

branches: 1.216.6;
"vmstat -ie[v]" auto fit to minimum columns

Reviewed by msaitoh@

Adapt to use of bintime(9) for kernel history timestamps (7.99.55).

Fix error handling confusion between error and errno

Adapt to new version of the sysctl export structure

Use appropriate PRI* format specifiers for fixed-bit-size numbers

Should fix i386 build.

For kvm-based kernhist code, print size of history table and "next free"
pointer, same as we do for sysctl-based code.

Restore the kvm-grovelling code for use on crash files (where sysctl(3)
is not available). Otherwise, this would always report on the current
running kernel regardless of the use of -M option.

For kernel history display, replace the kvm grovelling with the newly
available sysctl access to the history data.

XXX vmstat still uses kvm for several other displays

No point in using float here; makes evbmips64 need __truncdfsf2, and that
causes static linking issues because of softfloat.

Drop main() prototype.

branches: 1.206.2;
Update stats-keeping in sys/kern/vfs_cache.c to remove (most)
races while allowing consistent lockless sampling of the per-cpu
statistics without atomic operations. Update comment describing
the locking protocol to include this.

These files were fumble-fingered out of the last commit.

Trailing whitespace.

Print the cpu in the kernhist record.

branches: 1.203.2;
Use u_long to avoid negative value outputs.

Init summary counters.

Fix format string for ILP32.

Use kern.pool for the live system.

GC left-over kread.

Use CPU statistics from uvmexp when possible.

Use kern.clockrate if possible.

Convert the remaining uvmexp users to the sysctl as default source.

Use vm.uvmexp data as much as possible for vmstat -s.
Temporarily disable shadowing warning until the other users are cleaned
up.

Provide sysctl for namecache statistics.

Make missing kern.evcnt an error, don't try to fallback to kmem use.

Add a global vnode cache:

- vcache_get() retrieves a referenced and initialised vnode / fs node pair.
- vcache_remove() removes a vnode / fs node pair from the cache.

On cache miss vcache_get() calls new vfs operation vfs_loadvnode() to
initialise a vnode / fs node pair. This call is guaranteed exclusive,
no other thread will try to load this vnode / fs node pair.

Convert ufs/ext2fs, ufs/ffs and ufs/mfs to use this interface.

Remove now unused ufs/ufs_ihash

Discussed on tech-kern.

Welcome to 6.99.41

branches: 1.191.2;
Remove some pointless inclusions os sys/user.h

Remove cpu_queue (and thus eleminate another use of CIRCLEQ) by replacing
its uses with cpu_infos array. Extra testing by christos@.

add a row with totals for some pool statistics.

branches: 1.188.2;
remove malloc(9) output function from vmstat
it is not required and working any more

Use C89 function definitions

branches: 1.186.2;
use warn instead of perror.

unbreak -f -H -h, etc, part 1
- get nlist once in the beginning and fix the flags. Not each time! It is
not like the kernel will change it symbols in the middle of the run.

Use __arraycount(), as done above.

Mark usage as static and __dead.

dopoolcache: don't print unused caches unless verbose

move and rename the uvm history code out of uvm_stat to "kernhist".

rename "UVMHIST" option to enable the uvm histories.

TODO:
- make UVMHIST properly depend upon KERNHIST
- enable dynamic registration of histories. this is mostly just
allocating something in a bitmap, and is only for viewing multiple
histories in a merged form.


tested on amd64 and sparc64.

Fix header, format and total count of system interrupt counters.

dosum also needs getnlist

deal with detecting the first and last cpu differently.

Make sure to getnlist in dovmstat

Defer nlist processing until we know we need to it. If everything can be
obtained via sysctl, we can skip it entirely. This means we can run even
if not setgid.

getuptime will now use sysctl/clock_gettime if memf is NULL.
doevcnt now sues sysctl(kern.evcnt) is memf is NULL. It falls back to
groveling if sysctl returns an error of ENOENT.
dointr will call doevcnt to evcnt based intr stats.

branches: 1.175.2;
can't play games with structure names since cpu_data refers to cpu_info, so
provide our own copy for everything.

use a local definition of cpu_info if we have __HAVE_CPU_DATA_FIRST

Don't crash on old kernels that don't have the new per cpu counters.

We still need to update ouvmexp to output the differene from
previous iteration.

make this work with the per-cpu counters.

simplify and modernize:
- make sysctl mibs static
- use warn/warnx for warnings
makes binary 4K smaller.

Make "i" unsigned to mirror change to struct uvm_history

Follow christos' suggestions, and make ks_active a u_short, and
also only use 16 u_shorts instead of 32 ints. Also add panic()
calls for under- and overflow of the ks_active members under
DIAGNOSTIC. The MAXBUCKET constant ended up in sys/mallocvar.h
and not sys/param.h, as the latter caused build problems.

Ride the kernel revision bump of my previous change.

Extend struct malloc_type to count the number of active allocations
per size, and make vmstat report this information under the "Memory
statistics by type" display, which is only printed when the kernel
has been compiled with KMEMSTATS defined, like this:

Memory statistics by type Type Kern
Type InUse MemUse HighUse Limit Requests Limit Limit Size(s)
wapbl 15 4192K 4192K 78644K 376426 0 0 32:0,256:3,512:6,131072:1,262144:2,524288:3

Since struct malloc_type is user-visible and is changed, bump kernel
revision to 5.99.26.

While it is true that malloc(9) is in general on the path of slowly
being replaced by kmem(9) (kmem_alloc/kmem_free), there remains a
lot of points of usage of malloc/free, and this could aid in finding
any leaks. (It helped finding the leak fixed in PR#42661.)

This was discussed with and somewhat hestitantly OKed by rmind@

Remove uarea swap-out functionality:

- Addresses the issue described in PR/38828.
- Some simplification in threading and sleepq subsystems.
- Eliminates pmap_collect() and, as a side note, allows pmap optimisations.
- Eliminates XS_CTL_DATA_ONSTACK in scsipi code.
- Avoids few scans on LWP list and thus potentially long holds of proc_lock.
- Cuts ~1.5k lines of code. Reduces amd64 kernel size by ~4k.
- Removes __SWAP_BROKEN cases.

Tested on x86, mips, acorn32 (thanks <mpumford>) and partly tested on
acorn26 (thanks to <bjh21>).

Discussed on <tech-kern>, reviewed by <ad>.

fix -Wsign-compare issues

Add missing fflush() in one of three places that print stuff repeatedly
every so often. PR 32836 from Takuro KUBOTA.

branches: 1.163.6;
NFS hash tables no longer exist.

Restore the "pages actige" and "pages inactive" value printouts
in "vmstat -s" output when run on the active kernel.

The reason we can't easily provide these values out of a crash
dump, is that these fields are no longer proper members of
struct uvmexp, but rather are estimated by and dependent on
the currently active page replacement policy in the kernel.

Remove the \n and tabs from the __COPYRIGHT() strings.
Tweak to use a consistent format.

Fix alignment of fields for "vmstat -m" when KMEMSTATS are available:
o Make "prop dictionary" fit in the "Type" coloumn
o Ensure that fields don't run into each other (adds a space between fields)
o Ensure that a Limit of >100000K fits in the field width

vmstat -s: report cpu pagealloc hits/misses.

branches: 1.158.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.157.4;
Merge more changes from vmlocking2, mainly:

- Locking improvements.
- Use pool_cache for more items.

- Support two different sizes of pool_cache group. The default has 14 or 15
items, and the new large groups (for busy caches) have 62 or 63 items.
- Add PR_LARGECACHE flag as a hint that a pool_cache should use large groups.
This should be eventually be tuned at runtime.
- Report group size for vmstat -C.

add -C to usage.

Add -C option to display pool_cache info, eg:

Pool cache statistics.
Name Spin Full Empty PoolLayer CacheLayer Hit% CpuLayer Hit%
pipepl 0 1 3 385 42862 99.1 985780 95.7
fdescpl 0 0 2 310 31133 99.0 1029620 97.0
cwdipl 0 0 2 310 31227 99.0 1029714 97.0
filepl 0 1 4 680 141415 99.5 24011135 99.4

branches: 1.153.8;
warns=4, pass lint.

use howmany(), instead of open-coding the calculation.

increase a chance to fit within 80 char width consoles.
- use PRWORD for dovmstat as well.
- reduce number of drives shown by default from 3 to 2.

sync with kernel after yamt-pdpolicy merge.

PR 34205: workaround struct timeval tv_sec not being SUS compliant issue (reported by Takeshi Nakayama)

branches: 1.148.2;
Dump some column widths.

accept either "time_second" or "time" as kernel timevariable in
order to support common userland in enviroments where
timecounter kernels and non timecounter kernel co-exist
(e.g. xen & i386)

include evcnt.h rather than device.h.

support timecounters (struct time gone, use time_second then)

branches: 1.144.2;
Coverity CID 884: Avoid NULL deref.

Coverity CID 885: Avoid NULL deref.

Coverity CID 886: Avoid NULL deref.

Unwrap a not too long line.

Update statistics reporting to use new kernel framework, rename dkstats*
to a more generic name since it no longer just deals with disk.

Split dkreadstats() into dkreadstats(), tkreadstats() and cpureadstats().
Similarly for dkswap().
Allows code to only call one of these.
Rename struct _disk to _vminfo to be nearer its contents.
Change 'systat vm' so that it always shows the 'current' cpu times, even
after ':run' or ':boot' commands. The code in vmstat.c doesn't support :run.

branches: 1.138.2;
Don't forget original allocated memory address.

optimize pool_caches similarly to how I optimized pools before:
split the single list of pool cache groups into three lists:
completely full, partially full, and completely empty.
use LIST instead of TAILQ where appropriate.

Add support for tape statistics.

appease gcc -Wuninitialized

rename "kbucket" to "kmembuckets", for greater clarity.
don't forget to update vmstat this time.

do not rely on the pre-evcnt interrupt counter kernel symbols being present.
if they do not exist, then interrupt stats are all in evcnts.

remove anon related statistics which are no longer used.

Add -t to usage.

Add a '-t' option to print the contents of the vmtotal structure,
possibly repeatedly.

WARNS=3

Remove a blank line.

Sync with manpage (add -W, sort).

Add a "wide" mode output (selected with -W) to show more information
about memory pools, including pool page size, flags and utilisation.

Correct typo in an error message: allocatior -> allocator

Add (unsigned char) cast to ctype functions

Make vmstat -s prints:
- uvmexp.pdre{anon,file,exec}.
- real number of namecache stats.

pidhash and pgrphash should have been removed from here as well when the
new pid allocator was committed 3/2003.

Uppercase CPU, plural is CPUs.

fixes to keep default output under 80 columns
reduce default number of disks to 3 from 4
adjust widths of cpu stats if one of them is 100%
always put spaces between # of processes runnable (from pr 12691)

Extra columns appeared in one line of the "Memory statistics by type"
header in '94 and '97. Grow the other line a little so that it looks
pretty again.

realloc pedant

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

Patches provided by Joel Baker in PR 22365, verified by myself.

snprintf

No need to cast the third argument of ioctl() to "char *".

no need to #include <sys/dkstat.h>

Make vmstat -m skip memory pools that have never been used - unless -v given.
Update man page to match.

Update for cached physical addresses in the pool caches.

Alternative pid/proc allocater, removes all searches associated with pid
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.

The last NULL entry should not be printed on missing symbols.

Update for extensible malloc types.

The -u option takes an argument, so we need the corresponding : in the
getopt() string. Does no one use uvm histories any more, or does
everyone just dump all the histories?

Remove pc532-only hacks to access wildly different interrupt accounting
now that interrupts are tracked with evcnt(9) on this port.

Include <machine/intr.h> to get "struct iv" (in pc532-specific code).

In getuptime(), use kread() to get the current time as well as boottime,
instead of calling time(3).

This makes rate calculations of crash dumps sensible; in contrast,
subtracting the boot time of the crash dump from the current time of the
running system produces a not-very-meaningful number.

iostat(8) and systat(8) support for separate read/write disk statistics.
"iostat -x" now shows these (ala linux/solaris), but this is only splitting
the read/write bytes/transactions, not adding any new metrics. "systat
iostat" now has two new commands to switch between combined/separate mode
for both it's numbers & bar modes.

sleep() -> nanosleep()

Print newline before pool stats.

Since the KERN_CP_TIME sysctl reports total clock ticks across all
cpus and hz is measured in ticks per cpu, divide tick count by ncpu to
determine elapsed time since last sample.

Fixes I/O rate deflation observed on multiprocessors.

branches: 1.100.2;
Don't provide a local declaration of 'struct pool_cache_group' now that
it's in <sys/pool.h>.

Fix broken indentation.

fix pool_allocator fallout. bad boy jason!

Don't forget to fetch pool cache group.

When printing pool stats, print pool cache stats also if -v is given.

Use the new disk and tty stats sysctls when operating on a live kernel,
dk*() don't need to worry about setgid() games anymore.

replace "vnode" and "vtext" with "file" and "exec" in uvmexp field names.

print average chain length as %8.2f instead of %8d

KERNBASE is a constant that kernland has no business using; it's
existence and value are machine-dependent, and using KERNBASE
either breaks binary compatibility across the same MACHINE_ARCH
or prevents this program from compiling at all.

Cast the printf'ing of the 2 size_t vars in dohashstat to unsigned long long as
some ports have size_t's which are larger than int's.

- Separate hashstat namelist into separate hashnl[], and don't barf if
some of the symbols can't be found
- Only kvm_nlist() hashnl[] and histnl[] once
- Add a description to struct kernel_hash, and print with -L
- Sort entries in khashes[]
- Don't exit on unknown hashes; just display a warning and move on

add stats for nchash & ncvhash

- Use -H (instead of -h) to list all hash stats.
- Add '-h hashname' to only list the stats for hashname.
- Add -L to list valid hashes
- Fix up some whitespace

- replace -h uvmhist with -u uvmhist, and -H with -U, since uvm history stuff
isn't generally enabled by default, and i wanted -h for something else.
- implement -h, which displays statistics for various kernel hash tables
- add deref_kptr(), which does the hard work of kvm_read() et al
- various minor cleanups

Check array boundary correctly.

Print evcnt/intrcnt as unsigned.

When summarizing pool memory usage, account for PR_RECURSIVE pools
specially so as to avoid counting the same memory twice.

Adjust vmstat -i / -e for wider event names

when printing "avm fre" fields, use " %5ld %5ld " instead of "%6ld%6ld ",
because it's better to slightly right-shift the display than run numbers
together when you've got > 99999KB avm or fre.

Move the number of page colors to the top of the display.

Add the number of page colors to uvmexp.

Print print color bucket hit/miss stats.

Make -c and -w work properly when printing non-default statistics.

Show number of pages used by executables now that they're are counted
by UVM.

Prevent integer overflow.

Cosmetic changes.

Don't "extern" variables and function in dkstats.c, declare them in
dkstats.h instead.

ANSIfy.

Use a better description for the vnode page cache pages.

Show anonpages and vnodepages in "vmstat -s".

Move cpu header right 3 spaces to match grouping in man page

Make vmstat -m to put at least one space between each numbers.

Display number of zero page aborts.

remove include of <vm/vm.h>

branches: 1.66.2;
Implement the more flexiable `evcnt' interface as discussed (briefly) on
tech-kern and now documented in evcnt(9).

cp_time is no longer available in the kernel.
remove it from namelist[] to make vmstat work.

include sys/sched.h to get definition of CPUSTATES

branches: 1.63.2;
Print the zero page statistics.

Don't declare 'extern opt*' getopt variables.