Beginning with linux-4.5, the following new fields are
being added under that /proc/<pid>/status pseudo file:
. RssAnon - size of resident anonymous memory
. RssFile - size of resident file mappings
. RssShmem - size of resident shared memory
This patch just represents the initial library and top
support, sharing a commit message with 2 more patches.
p.s. locked resident memory support was also added but
isn't directly related to the kernel 4.5 enhancements.
Reference(s):
commit 1f8e41d019
Signed-off-by: Jim Warner <james.warner@comcast.net>
If the mapping descriptor is longer than 128 chars, the last parsed
character won't be a newline even if the current buffer contains it a
bit further than that. The current code always interprets it as a short
fgets() read instead, and thus keeps calling fgets() until it gets a
newline, dropping valid lines and failing with the following error:
pmap: ERROR: inconsistent detail field in smaps file, line:
Rss: 212 kB
[ patch has been adapted expressly for newlib branch ]
In response to Issue #21, the commit referred to below
provided some much needed improvements and corrections
to topic `3a. DESCRIPTIONS of Fields' in the man page.
However, it assumed a reader possessed much background
knowledge that may not, in truth, actually be present.
So without, hopefully, insulting anyone's intelligence
this patch offers an expanded discussion of some terms
and concepts within a separate section under OVERVIEW.
[ plus it affords an opportunity to incorporate that ]
[ extremely useful table from Florent Bruneau's post ]
Reference(s):
commit 6a36bd7afdhttps://techtalk.intersec.com/2013/07/memory-part-1-memory-types/
Signed-off-by: Jim Warner <james.warner@comcast.net>
Due to quirks in kernel memory management plus limited
information available from /proc/<pid>/status & statm,
some of the top resident memory fields were capable of
exceeding available physical memory. So this commit is
a bit of a band-aid until the kernel has been changed.
Such a change appears to be on the horizon in the form
of three new fields to be added to /proc/<pid>/status.
While not preventing 'resident' memory from apparently
exceeding physical memory, the new fields will help to
clarify any such contingency, if/when we exploit them.
Reference(s):
. original post by Samuel Thibault
https://gitlab.com/procps-ng/procps/issues/21
. informative memory analysis
https://techtalk.intersec.com/2013/07/memory-part-1-memory-types/https://techtalk.intersec.com/2013/07/memory-part-2-understanding-process-memory/
. kernel changes to /proc/<pid>/status
commit 8cee852ec53fb530f10ccabf1596734209ae336b
commit eca56ff906bdd0239485e8b47154a6e73dd9a2f3
Signed-off-by: Jim Warner <james.warner@comcast.net>
The commit referenced below claims to disable vertical
scrolling when idle tasks weren't being shown. However
it really addresses only a point in time when that 'i'
toggle is keyed. Left untouched were the up/down keys.
So this commit will simply finish the job of disabling
vertical scrolling whenever tasks which have used some
CPU are the only ones which are currently being shown.
Reference(s):
commit c07f6c5e6d
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit is just a cleanup of some extraneous cruft
left after the newlib migrations and summarized below:
. whack the 2nd line of what was an eliminated comment
. eliminated reference to NUM_NS and thus that need to
include the otherwise defunct 'readproc.h' header file
. reference to Frame_maxtask made properly generalized
. all former sort support #define's are now eliminated
Signed-off-by: Jim Warner <james.warner@comcast.net>
Summarized below, miscellany addressed in this commit:
. deleted extraneous newline(s) for consistent spacing
Signed-off-by: Jim Warner <james.warner@comcast.net>
Historically LFS mattered only to open/read large files. A few programs
here use open/seek, but not generally on files that are large. However,
LFS also applies to stat which procps does in a bunch of places -- some
filesystems have 64bit inodes and attempts to do a 32bit stat will throw
an error.
Enable transparent LFS everywhere to avoid this.
There is a need in some utilities to have a way of accepting both
types of decimal points "." and ",". The only way seems to be to
rebuild strtod().
This new function will accept "123.456" and "123,456" as 123.456
and considers them the same number. It means we lose thousands
separator, but this is rarely used.
test scripts are added to check the function returns the proper
values. There was simpler predecessor that got stuck on negative
0 or -0.123 which these tests flushed out.
References:
Commit 99fa7f removed watch crashing from having the null ANSI
sequence. However it just ignored that sequence instead of fixing it.
This change correctly checks buf (not undefined endptr) and sends
the reset attribute if found.
Closes: #12
It is documented behavior that when certain other keys
are active, sorts column highlighting will temporarily
be disabled. Among those keys is the 'L' (locate/find)
provision. The equals ('=') key can be used to restore
column highlighting by resetting other keys, except 1.
When a locate/find is active, the '=' key will have no
effect on 'x' column highlighting, which still remains
disabled. Further, when 'L' is active an 'x' keystroke
is processed changing the state of column highlighting
but without any visual clue (since it's yet disabled).
So this commit just extends the '=' key to embrace 'L'
processing resets, just like other highlight disabling
keys while avoiding 'x' state changes if approproiate.
Signed-off-by: Jim Warner <james.warner@comcast.net>
We'll following the newlib <pids> approach to hashing:
. a 'PIDs at max depth:' portion of that UNREF_RPTHASH
enabled #define is now published only when the maximum
depth of hash table entry chains exceed depths of one.
Signed-off-by: Jim Warner <james.warner@comcast.net>
Craig's recent commit under that newlib branch dealing
with namespace support has prompted me to review top's
handling of those fields. Currently, when such a field
is zero, top displays a dash ('-'). This will mean the
justification toggles ('j/J') will behave incorrectly.
This patch simply allows the potential zero to display
or be suppressed with the already existing '0' toggle.
Signed-off-by: Jim Warner <james.warner@comcast.net>
A patch containing the following miscellaneous tweaks:
. remove a function that handled former library errors
[ that function should have gone bye-bye with 3.3.11 ]
[ when those 'wchan' provisions were much simplified ]
Signed-off-by: Jim Warner <james.warner@comcast.net>
When reference counts were added to some string fields
the 3 true string vector fields were not duplicated as
were those other fields. Instead they were supposed to
disallow a duplicate stack reference beyond the first.
However, the actual implementation gave NULL for every
true vector field whenever such items were duplicated.
More importantly, such true string vector fields never
considered references to the shared proc_t source root
which would have forced the conversion of such vectors
into a single string form via the '_CVT' library flag.
So this commit restores the intended outcome with true
string vectors. There's only 1 valid reference allowed
and duplicates and converted fields will yield a NULL.
Signed-off-by: Jim Warner <james.warner@comcast.net>
While not changing generated code this commit corrects
one free reference from 'str' to a more proper 'strv'.
Signed-off-by: Jim Warner <james.warner@comcast.net>
ps previously followed the Unix98 standard when it comes to
user-defined output, sometimes. This meant you could have
user output format with a header that included commas and
equals signs. It was dependent on if ps thought you wanted
sysv or bsd format and THAT was dependent on things in previous
options.
It was very confusing to a user because
ps p $$ -o pid=,comm=
gave you a two-column output but
ps -p $$ -o pid=,comm=
would give you a one column output with the header ",comm="
The -p versus p means (to ps) you want sysv or bsd parsing.
Unix98 standard or not, this is plainly just silly.
The commit removes any of the quirks Unix98 has with user defined
output. If you really wanted a ps header with commas in the output,
today isn't your day.
Signed-off-by: Craig Small <csmall@enc.com.au>
configure.ac: do not check for C++ compiler
We do not have any C++ code, so there is no point in checking for C++
tools in the configure script.
Signed-off-by: John Keeping <john@keeping.me.uk>
See merge request !8
Ok, here is that rather major internal redesign hinted
at in the three previous commits. Its need was quickly
revealed after adapting top then attempting to display
newly added 'CGNAME' fields and an existing 'CGROUPS'.
That very quickly generated a SEGV. And the reason was
just as quickly recognized. Both fields relied on that
proc_t.cgroup member yet whichever result structure is
first in a stack is the one which assumes ownership of
of the vectored sting by resetting its cgroup to NULL.
So this commit introduces reference counting for a few
of the fields in the proc_t. Specifically there are 17
entries in the Item_table dealing with strings/vectors
where ownership is transferred to newlib. Now whenever
such fields are represented more than once in a stack,
the strings will be duplicated instead of transferred.
In this way we can generally remain optimized avoiding
string copies, yet still accommodate them when needed.
There's an exception to this scheme: those true string
vectors (CGROUP_V, CMDLINE_V and ENVIRON_V). When such
fields are duplicated in a stack the result structures
beyond the first will be set to NULL, which the caller
will (should) already be equipped to deal with anyway.
Signed-off-by: Jim Warner <james.warner@comcast.net>
[ but stay tuned! there is a commit coming soon that ]
[ represents a rather major internal redesign, which ]
[ was prompted by the ps and top adaptation testing. ]
Signed-off-by: Jim Warner <james.warner@comcast.net>
[ but stay tuned! there is a commit coming soon that ]
[ represents a rather major internal redesign, which ]
[ was prompted by the ps and top adaptation testing. ]
Signed-off-by: Jim Warner <james.warner@comcast.net>
The ps program was modified to print the control group
names, based on the library provided list of all those
control groups to which a process belongs. But this is
probably something the newlib should be doing for all.
So this commit borrows the ps approach to cg names and
thus will make that available to all future consumers.
[ but stay tuned! there is a commit coming soon that ]
[ represents a rather major internal redesign, which ]
[ was prompted by the ps and top adaptation testing. ]
Signed-off-by: Jim Warner <james.warner@comcast.net>
The above function was the sole public function in the
<pids> interface to use the word 'stacks' in its name.
All of the others dealt exclusively with their duties,
So this commit normalizes that outlier by renaming it.
Signed-off-by: Jim Warner <james.warner@comcast.net>
The above function had been disabled via '#if 0' so as
to prevent a compiler warning. But it really should be
called by that 'procps_pids_read_shut' function rather
than it duplicating/reinventing the same logic itself.
Signed-off-by: Jim Warner <james.warner@comcast.net>
It is documented behavior that when certain other keys
are active, sorts column highlighting will temporarily
be disabled. Among those keys is the 'L' (locate/find)
provision. The equals ('=') key can be used to restore
column highlighting by resetting other keys, except 1.
When a locate/find is active, the '=' key will have no
effect on 'x' column highlighting, which still remains
disabled. Further, when 'L' is active an 'x' keystroke
is processed changing the state of column highlighting
but without any visual clue (since it's yet disabled).
So this commit just extends the '=' key to embrace 'L'
processing resets, just like other highlight disabling
keys while avoiding 'x' state changes if approproiate.
Signed-off-by: Jim Warner <james.warner@comcast.net>
The cgroup field while shown as a vector is a concatenated
string, so alot of the complexity of sorting and displaying
has gone.
This change simplifies the cgroup sorting and adds display
and sorting for the name attribute of the cgroup, if found.
Signed-off-by: Craig Small <csmall@enc.com.au>
Ported-by: Jim Warner <james.warner@comcast.net>
From original:
commit 0ee090ae16
w would error out if the window size was smaller than 71 or some
other fields through environment grew too big. The code was a little
convoluted as well. The minimum length for command was 3, which is
pretty useless.
This change does the following:
w doesn't care by default the window size
w will adjust the command length up and down, to a minimum of 7
characters.
if the fields don't fit, w will line-wrap each line.
The idea being its better the line-wrap than it is to error out.
References: https://bugs.debian.org/183394
Signed-off-by: Craig Small <csmall@enc.com.au>
Ported-by: Jim Warner <james.warner@comcast.net>
From original:
commit 151c05b497
Previous commit fixed pkill for trailing garbage on pkill
signal when it was an integer. This check now ensures that
pkill complains about it.
References:
commit a3975a9c60
This commit prevents pkill from accepting something like `-1garbage` as
a SIGHUP. The previous code was using atoi() which does not check for
trailing garbage and would parse the above as 1.
Handling numeric signals in signal_option() is not really necessary,
since signal_name_to_number() will recognize numeric signals and parse
them properly using strtol() and checking for trailing garbage. It also
checks that the numeric signals are in the proper range. So all we need
to do is remove the buggy numeric signal handling here.
Tested with `pkill -1garbage sleep`, after this patch it will complain
that "1" is not a valid option, which is the expected.
Signed-off-by: Filipe Brandenburger <filbranden@google.com>
Ported-by: Jim Warner <james.warner@comcast.net>
From original:
commit 9646f7cba4
Ever since top was adapted to the new <pids> interface
there has been a bug that would cause an abnormal exit
when the '-p' argument contained *no* valid pids. This
was never revealed until now since the QA folks tested
only with valid, existing pids. (bunch of morons, eh?)
And even though the program author is blameless he has
taken it upon himself to clean up after the QA jokers.
Signed-off-by: Jim Warner <james.warner@comcast.net>
Gosh, just because nobody uses some newlib provision I
guess, since it is being offered, it ought to actually
be tested at some point. Well, that point just arrived
and guess what? A surprise: some bugs were discovered.
The procps_pids_select function established a for loop
wherein readproc is called until the passed 'maxthese'
limit. Unfortunately this was incorrect for 2 reasons:
1. For PROCPS_FILL_PID results are limited by what the
oldlib finds, having established the pid list at open.
Total found already cannot exceed a passed 'maxthese';
2. With PROCPS_FILL_UID, returned results could exceed
a 'maxthese' thus making the for loop incorrect again.
[ plus yours truly neglected to forward the required ]
[ UIDs total to our old library, another oops biggie ]
In summary: the loop should have been forever, exiting
only when all those identified procs had been located.
So, while addressing those bugs, I've consolidated all
the retrieval code (initialize, iterate, summarize) in
a single helper function which will now serve both the
procps_pids_reap and select functions. And as a result
those guys were reduced to quite trivial housekeeping.
This patch, hopefully, completes the normalization for
reap/select (fill), which began with references shown.
Reference(s):
commit 0c953eccc5
commit 747dfc5987
Signed-off-by: Jim Warner <james.warner@comcast.net>
With the conversion to the new <pids> interface, a few
comments (only) are being adjusted, as detailed below.
. Escapes '\' crept into some comments containing '|'.
. For consistency, add '.' dot qualifier to a comment.
Signed-off-by: Jim Warner <james.warner@comcast.net>
