appear to be never called:
(1) If a function is never called according to its call count but it
must have been called because its child time is nonzero, then print
it in the flat profile. Previously, if its call count was zero
then we only printed it in the flat profile if its self time was
nonzero.
(2) If a function has a zero call count but has a nonzero self or child
time, then print its total self time in the self time per call
column as a percentage of the total (self + child) time. It is
not possible to print the times per call in this case because the
call count is zero. Previously, this was handled by leaving both
per-call columns blank. The self time is printed in another column
but there was no way to recover the total time.
(1) partially fixes the case of the "never called" function main() and
prepares for (2) to apply to main() and other functions. Profiling
of main() was lost in the conversion from a.out to ELF, so main()'s
call count has always been zero for many years; then in the common
case where main() is a tiny function, it gets no profiling ticks, so
main() was completely lost in the flat profile.
(2) improves mainly cases like kernel threads. Most kernel threads
appear to be never called because they are always started before
userland can run to turn on profiling. As for main(), the fact that
they are called is not very interesting and their callers are
uninteresting, but their relative self time is interesting since they
are long-running.
Almost always printing percentages in the per-call columns would be
more useful than almost always printing 0.0ms. 0.1ms is now a long
time, so only very large functions take that long per call. The accuracy
per call can approach 1-10 nsec provided programs are run for about
100000 times as long as is necessary to get this accuracy with high
resolution kernel profiling.
you want to see, e.g., sendmail arguments mail(1) will use.
-H is not an independent flag, it's a modifier. Also explicitly
say that -H will cause mail(1) to exit as soon as it prints the headers.
MFC after: 5 days
Add a flags argument to wait_for_lock so that O_NONBLOCK can be
passed to open if a user doesn't want the open to sleep until the
lock becomes available.
Submitted by: Amir Shalem (partially modified)
for mutual exclusion:
A brief description of the problem:
1) Proc A picks up non-blocking lock on file X
2) Proc B attempts to pickup lock, fails then waits
3) Proc C attempts to pickup lock, fails then waits
4) Proc A releases lock
5) Proc B acquires lock, release it to pickup a non-blocking version
6) Proc C acquires lock, release it to pickup a non-blocking version
7) Both process B and C race each other to pickup lock again
This occurs mainly because the processes do not keep the lock after they have
been waiting on it. They drop it, attempt to re-acquire it. (They use the wait
to notify when the lock has become available then race to pick it up). This
results in additional CPU utilization during the race, and can also result
in processes picking locks up out of order.
This change attempts to correct this problem by eliminating the test/acquire
race and having the operating system handle it.
Reported by: kris
Tested by: kris
MFC after: 1 week
Split commands into two groups: one with optional count and one with
required argument. Changed synopsis line accordingly.
Added some hopefully-helpful comments based on experiments, knowing
that not all hardware works the same.
PR: docs/84101
Approved by: keramida
MFC after: 3 days
replacement and has additional features which make it superior.
Discussed on: -arch
Reviewed by: thompsa
X-MFC-after: never (RELENG_6 as transition period)
-- Made the synopses more precise.
-- Added argument to flag in option description.
-- Moved -b default and limits to option description (to un-hide).
-- Noted several behaviors that were not mentioned.
-- A few more trivial changes.
PR: docs/46787
Approved by: keramida
MFC after: 3 days
a -B option which causes bpf peers to be printed. This option can be
used in conjunction with -I if information about specific interfaces
is desired. This is similar to what NetBSD added to their version of
netstat.
$ netstat -B
Pid Netif Flags Recv Drop Match Sblen Hblen Command
1137 lo0 p--s-- 0 0 0 0 0 tcpdump
205 sis0 -ifs-l 37331 0 1 0 0 dhclient
$
$ netstat -I lo0 -B
Pid Netif Flags Recv Drop Match Sblen Hblen Command
1174 lo0 p--s-- 0 0 0 0 0 tcpdump
$
-Add bpf.c which stores all the code for retrieving and parsing bpf
related statistics.
-Modify main.c to add support for the -B option and hook it into the
program logic.
-Add bpf.c to the build.
-Document this new functionality in the man page and bump the revision
date.
-Add prototype for bpf_stats function.
if none was specified on the command line. This is not permitted by
POSIX, and no longer needed now that we have the -a option.
PR: 85099
Submitted by: Toby Peterson (Apple Computer)
integer to an unsigned long. This lifts variables like the maximum
number of pages available for shared memory from 2^31 to 2^32 on 32
bit architectures, and from 2^31 to 2^64 on 64 bit architectures.
It should be noted that this changes breaks ABI on 64 bit architectures
because the size of the shmmax, shmmin, shmmni, shmseg and shmall members
of the shminfo structure has changed.
Silence on: current@
constructing and applying binary patches; in particular, they perform
well (in the sense of constructing small patches) for executable code.
Both portsnap (coming to the base system Real Soon Now) and FreeBSD
Update (coming to the base system a bit later) use bspatch.
This is the same code as the bsdiff-4.2 which has been in the ports
tree (misc/bsdiff) for the past year, with the following exceptions:
1. The license is now the traditional 2-clause BSD;
2. Instead of forking and execing bzip2, the code now uses libbz2; and
3. Some minor changes have been made to fit this code into the base
system (adding $FreeBSD$ tags, putting bsdiff and bspatch into separate
directories, etc.)
This code is rather ugly and has lots of style bugs (mostly because I
wrote it before I had ever heard of style(9)). Some day I'll come
back and clean it up.
Discussed on: freebsd-arch
MFC before: 5.5-RELEASE
Tested by: Several million users (earlier version).
set up before it is called, so move the progname initialization before
the first possible call to bsdtar_warnc().
Thanks to: Stanislav Sedov
PR: bin/83366
MFC after: 7 days
cdev structure, returns the device name associated with it through
the __si_namebuf member. This un-breaks the processing of devices.
This is a RELENG_6 candidate.
Reviewed by: phk
- Remove some extra blank lines.
- Remove comments that don't contribute to understanding.
- Remove additional blank lines in output added to maximize compatibility
with older vmstat output, but that is actually somewhat gratuitous.
Submitted by: bde
MFC with: other vmstat libmemstat(3) changes
statistics from -z are now a bit different due to changes in the
way statistics are now measured. Reproduce with some amount of
accuracy the slightly obscure layouts adopted by the two kernel
sysctls. In the future, we might want to normalize them.
GC dosysctl(), which is now no longer used.
MFC after: 1 week
avg/median/stddev bars onto separate lines for readability if the
ranges overlapped. In 2005, ministat was extended to support more than
2 datasets, but the -s code was not updated. It will coredump if run
with -s and >2 sets.
PR: 82909
Submitted by: Dan Nelson <dnelson@allantgroup.com>
commands for this target are appended to the .END target instead
of beeing executed now. They are executed when the graph is finished.
There was a bug with executing the .END target which came in when
doing conversion to LST_FOREACH() which caused make to dump core.
PR: bin/83698
Submitted by: Max Okumoto <okumoto@ucsd.edu>
MFC after: 3 days
with a number of positive benefits:
- Start using UMA(9) statistics for mbufs and clusters, which avoids
using the mbuf allocator statistics which suffer from races under
load on SMP. This should eliminate "negative" mbuf counts in
netstat -mb.
- We are now able to track cached (free) mbufs and clusters and count
it towards memory allocated by the network stack.
- We are now also able to track memory allocated to mbuf tags since
libmemstat(3) can also query malloc(9). We don't print this except
as part of the total (for now - #if 0).
- We are now able to track mbuf/cluster/packet allocation failures,
although they are not currently printed (#if 0).
- Don't print out sfbuf statistics when running on a kernel core, as
currently that code is able only to query sysctl for statistics.
MFC after: 1 week
1) An unquoted space is always a separator, even when not "in_arg".
2) When a new destination buffer must be allocated during variable
substitution, only copy data from the active buffer to the new
one when we *are* "in_arg".
These were noticed when testing variable-substitution of variables
which have null values, and are not inside quoted strings...
MFC plans: after a few days, and re@ approval...
was a separator character immediately before it. This wasn't likely to
happen in #-lines, but we might as well get it right. Also fix it so
that "" and "" will create a zero-length argument.
Approved by: re (blanket `env')
start with a '/', they are more supported (by POSIX and SUSv3) than
filenames which have an '=' in them.
Noticed by: tjr
Approved by: re (blanket `env')
apart a string, and supports some text substitutions. This can be
used to provide all the flexibility (and more!) that was lost by recent
changes to how the kernel parses #!-lines in shell scripts.
The '-P' option provides a way to specify an alternate set of directories
to use when searching for the 'utility' program to run. This way you can
be sure what directories are used for that search, without changing the
value of PATH that the user has set. Note that on FreeBSD 6.0, this
option is worthless unless the '-S' option is also used.
Approved by: re (blanket `env')
without checking it for an equals-sign. If it starts with a slash, then
it cannot be a request to set the value of a valid environment variable.
Approved by: re (blanket `env')
it does not happen until all single-letter options are processed. This will
be important for the -S option, which will be coming soon.
Approved by: re (blanket `env')
malloc(9) statistics from kernel memory or a kernel coredump, to catch
up with recent changes to adopt per-CPU malloc(9) statistics. The new
routines walk the per-CPU statistics pools and coalesce them for
presentation to the user.
the string. Until now this caused no harm, because the buffer code used
to tack two NULs onto buffers. With the new, soon to come, parsing code
this isn't the case anymore in all cases, so fix this.
rename the function to be consistent with the naming scheme in the rest
of make. No functional changes.
Obtained from: DragonFlyBSD (idea and most of shell.h)
before executing the shell. Until now this was done when the default
shell was the ksh. This failed if the default shell was sh or csh and
the user switched to ksh.
set the current shell to DEFSHELL. Put all these specifications into
a list. Add user specified new shells to this list. If the user
just selects one of the already know shells just pick the right one
off the list. This let's one do something like:
# Full specification of the user's shell. This also selects the shell.
.SHELL: name=myshell path=/somewhere/foo echo=loud ...
FOO != bar # use myshell here
.SHELL: name=sh
BAR != baz # use /bin/sh here
.SHELL: name=myshell # no need for full spec here.
# continue to use the user's special shell.
the list of shell builtins. Both of these are needed for the compat
mode where make directly executes commands if the command line contains
neither a shell meta character nor a shell builtin. The list of builtins
is not changed, but csh has '@' added as a meta-character.
Initialize the default shell by parsing a string as one would specify
to the .SHELL target. So we get rid of the CShell clone of struct Shell which
just contained const char * where struct Shell had char *.
Add a debugging function for dumping a parsed shell description to
stdout.
introduce a struct that holds all the information about an argument
vector and pass that around.
Author: Max Okumoto <okumoto@ucsd.edu>
Obtained from: DragonFlyBSD
command that toggles between the two and update the ORDER_PCTCPU()
macro to sort correctly by the visible "cpu" value.
This saves 6 more columns in 80-column terminals, making things a lot
better for the COMMAND column.
Tested on: i386, sparc64 (panther), amd64 (sledge)
Approved by: davidxu (in principle)
similar to the zmore script that comes with gzip (and in fact, in most
Linux distros, zless is a symlink to that very same zmore script) but has
the advantage that you get the correct file name on the less status line,
and can use :n and :p to navigate back and forth between multiple files.
MFC after: 1 week
there are users on the system (even if not running a single process)
with a login > 8 chars.
I'm not all that happy limiting the username width like this, but it
restores sanity to top(1) output.
Discussed with: keramida
after that many values have been printed. The line length is not
considered anymore.
o Add option -x. The -x option will cause the byte values to be
printed in hexadecimal instead of decimal.
o Bump WARNS to 6.
o Update the manpage accordingly.
Make sure we don't end up with shellPath beeing non-zero, but shellName
beeing zero in the error case - back out cleanly from the error.
When executing a command for macro assignment in Cmd_Exec() stuff the
path of the shell into argv[0], not the name. This makes no difference
from the functionality point of view, but allows the regression tests to
determine whether make executes the correct shell.
match text; it also doesn't warn() for files which can't be opened. Remove
global variables. Use bool. fopen(3) the files instead of freopen(3)ing stdin.
used so there is no need to stuff the value of .MAKE into it,
which btw isn't set for quite a while already.
Submitted by: Max Okumoto <okumoto@ucsd.edu> (7.239)
into job.c. Move retrieving of environment nearer to the place where it
is actually used and invert the preprocessor conditionals to use
positive logic.
Submitted by: Max Okumoto <okumoto@ucsd.edu> (7.236)
been two maxJobs variables: one static in job.c and one global used in
main.c and parse.c. Makeing one global out of these was the wrong way
to fix the problem. Instead rename the global one to jobLimit and keep
maxJobs static in job.c.
Suggested by: rwatson
PR: bin/72510
takes place in the child process in a function ProcExec(). Make sure,
that the child does not call malloc() or other potential dangerous
functions (there are still calls to Punt() in the error case that
should go away). Allocate the argv string via malloc to overcome
the non-constness bug of the execvp prototype. Change the handling of
shell meta-characters and move the builtin list near the list of shell
builtins. Both of these lists should actuall be configurable by the .SHELL
target since they depend on the shell used.
Patch: 7.21[2-9], 7.22[0-46]
Submitted by: Max Okumoto <okumoto@ucsd.edu>
kldstat -m finds geom_uzip module even if it is compiled in statically.
- create output file with x bit set.
- build mkuzip on all architectures (verified with "make universe").
- fix typo in info message.
where they actually belong to. Move the definitions of the strings
for special macros like "$*" from make.h to parse.h - they're used
only in the parser.
Submitted by: Max Okumoto <okumoto@ucsd.edu> (7.211)
context (and only in one place to substitute the .for variable). Therefor
there is no need to pass the context as a parameter.
Patch: 7.197
Submitted by: Max Okumoto <okumoto@ucsd.edu>
take everything after -- as either a macro assignment or a target.
Note that make still reorders arguments before --: anything starting
with a dash is considered an option, anything which contains an equal
sign is considered a macro assignment and everything else a target.
This still is not POSIX with regard to the options, but it will probably
not change because it has been make's behaviour for ages.
Add a new function Var_Match() that correctly skips a macro call by just
doing the same as Var_Subst() but without producing output. This will help
making the parser more robust.
Patches: 7.190,7.191
Submitted by: Max Okumoto <okumoto@ucsd.edu>
only one variable and Var_Subst() which substitutes all. Split out the
test whether a variable should not be expanded into match_var().
Make access to the input string consistently using str[]. Remove two
unused functions: Var_GetTail() and Var_GetHead().
Patches: 7.184-7.189
Submitted by: Max Okumoto <okumoto@ucsd.edu>
make macro into the environment of programs executed by make. This
has approximately the same function as gmake's export directive.
The form of a pseudo target was deliberately choosen to minimize work
for POSIX compatibility (Makefiles are not allowed to use any targets
starting with a dot and consisting only of uppercase letters except those
specified in the standard when they want POSIX compatible behaviour, so
such a Makefile can never contain .EXPORTVAR.)
Change the handling of macros coming from the environment: instead
of asking the environment for each variable we could not find otherwise
put all the environment variables in a special variable environment just
at start up.
This has been tested on the ports cluster by kris.
Submitted by: Max Okumoto <okumoto@ucsd.edu>
given is looked up in a table and no longer stored literally in the
header.
Submitted by: Divacky Roman <xdivac02@stud.fit.vutbr.cz
PR: 80499
MFC After: 1 month
that there are more than one hash table in them. There is no
history to preserve here, so go without a repo-copy.
Asked for by: Max Okumoto <okumoto@ucsd.edu>
to the id_print() function.
Use getgrouplist(3) for the case when an user was specified,
and getgroups(2) when no user was given.
That reverts to the expected behaviour and makes it easy to
implement an option later to force using getgrouplist(3).
NGROUPS groups. getgrouplist(3) may put a duplicate group
id into the passed array (it sets [0] and [1] to the value
of the gid argument), but id_print() sorts them out.
Showing the ids of both an user given by an argument to `id',
and the current user, is now handled in a single function.
Displaying the current user's ids was inaccurate because
getgroups(2) had been used. getgroups(2) returns the current
kernel state of a user's groups, which may not always be
correct if /etc/group was recently changed.
- Fix a few style bugs.
PR: bin/78085
Negative values would produce undefined behaviour including
a possible segmentation fault.
- Explicitly initialize the global row and column variables
to zero.
PR: bin/80348
a Makefile target to re-created this file. Note, that there is no
explicite dependency to automatically re-create the file, because this
is needed only when the directive table changes and it requires the
(yet to come) devel/mph port.
Submitted by: Max Okumoto <okumoto@ucsd.edu> (first version)
(combine with existing seconds-based), treat '-' as punctuation rather
than a negative number indicator (eliminates several special cases),
use a single list of special words instead of several separate lists,
use table-driven abbreviation logic (eliminate duplicate word entries
and special-case abbreviation and plural handling). The result is
shorter, simpler (judging from comments, earlier maintainers didn't
understand the special handling for "negative years"), handles more
cases (e.g., "tu" is now a recognized abbreviation for "tuesday",
"3rd" is now equivalent to "third") and it has 2 fewer shift/reduce
conflicts.
particular, acl support is no longer enabled on FreeBSD 4, acl support
should be correctly enabled on Linux, dirent.d_namlen should be
correctly detected on platforms that support it.
Thanks to: Greg Lewis, Juergen Lock, and Jaakko Heinonen
extraction and creation. While I'm here, fix a bug reported by Garrett
Wollman: when stripping the leading '/' from the path "/", don't produce
an entry with an empty name; produce "." instead.
of lines in SMP machines (which are wider), until we have a better way
of handling window sizes & columns in top.
Caught by: ache, Andre Guibert de Bruet <andy@siliconlandmark.com>
Point hat: keramida
threads a process has. The THR column is disabled and disappears
when 'H' is hit, because then every thread gets its own output line.
- Allow sorting processes by "threads".
Approved by: davidxu
Inspired by: Jiawei Ye <leafy7382@gmail.com>
