The current OpenBSD-specific CPU usage code is broken. The `cpu`
parameter of `Platform_setCPUValues` is an integer in the interval
[0, cpuCount], not [0, cpuCount-1]: Actual CPUs are numbered from
1, the “zero” CPU is a “virtual” one which represents the average
of actual CPUs (I guess it’s inherited from Linux’s `/proc/stats`).
This off-by-one error leads to random crashes.
Moreover, the displayed CPU usage is more detailed with system,
user and nice times.
I made the OpenBSD CPU code more similar to the Linux CPU code,
removing a few old bits from OpenBSD’s top(1). I think it will be
easier to understand, maintain and evolve.
I’d love some feedback from experienced OpenBSD people.
The source code correctly states that the maximum PID number in
the OpenBSD kernel is fixed in sys/sys/proc.h, however this was
updated in revision 1.215 (two years ago!) from 32766 to 99999.
Introduction of CP_SPIN sched state broke hard-coded state indexes
resulting in the meters incorrectly reporting bogus intr data instead of
CPU usage. Change hardcoded values to sched.h macros.
Specifically, Platform_signals[] and Platform_numberOfSignals. Both are
not supposed to be mutable. Marking them 'const' puts them into rodata
sections in binary. And for Platform_numberOfSignals, this aids
optimization (aids only Link Time Optimization for now). :)
Signed-off-by: Kang-Che Sung <explorer09@gmail.com>
Namely:
o use malloc where an xCalloc slipped in
o safeguard against an empty arg list - I don't think it's possible,
but it would be potentially exploitable
o we need to initialize the arg string to an empty string because we no
longer use strlcpy(3)
o annotate a tricky use of strlcpy(3)'s truncation
Including:
o set *basenameEnd even in error cases (FreeBSD probably needs this)
o use kvm_openfiles(3) rather than kvm_open(3) so that we can report
errors (as with FreeBSD)
o sanify the process argument list creation by using strlcat(3)
o drop the pageSizeKb variable and use the PAGE_SIZE_KB macro directly,
as the page size can't change anyway
o clean up a few macros, add MINIMUM() and MAXIMUM() (should be
mirrored to FreeBSD)
o fix some syntax
o add some useful comments
With the CLAMP macro replacing the combination of MIN and MAX, we will
have at least two advantages:
1. It's more obvious semantically.
2. There are no more mixes of confusing uses like MIN(MAX(a,b),c) and
MAX(MIN(a,b),c) and MIN(a,MAX(b,c)) appearing everywhere. We unify
the 'clamping' with a single macro.
Note that the behavior of this CLAMP macro is different from
the combination `MAX(low,MIN(x,high))`.
* This CLAMP macro expands to two comparisons instead of three from
MAX and MIN combination. In theory, this makes the code slightly
smaller, in case that (low) or (high) or both are computed at
runtime, so that compilers cannot optimize them. (The third
comparison will matter if (low)>(high); see below.)
* CLAMP has a side effect, that if (low)>(high) it will produce weird
results. Unlike MIN & MAX which will force either (low) or (high) to
win. No assertion of ((low)<=(high)) is done in this macro, for now.
This CLAMP macro is implemented like described in glib
<http://developer.gnome.org/glib/stable/glib-Standard-Macros.html>
and does not handle weird uses like CLAMP(a++, low++, high--) .
This is what OpenBSD's top(1) does when the libkvm call fails, and it's
a good idea.
This commit also fixes process name construction. The space was being
written one character too far.
I forgot how awful the process name logic was. It was an initial hack to
get it running, and I forgot to clean it up.
I also had to change a few includes and error function uses.