#!/usr/perl5/bin/perl # # CDDL HEADER START # # The contents of this file are subject to the terms of the # Common Development and Distribution License, Version 1.0 only # (the "License"). You may not use this file except in compliance # with the License. # # You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE # or http://www.opensolaris.org/os/licensing. # See the License for the specific language governing permissions # and limitations under the License. # # When distributing Covered Code, include this CDDL HEADER in each # file and include the License file at usr/src/OPENSOLARIS.LICENSE. # If applicable, add the following below this CDDL HEADER, with the # fields enclosed by brackets "[]" replaced with your own identifying # information: Portions Copyright [yyyy] [name of copyright owner] # # CDDL HEADER END # # # Copyright 2005 Sun Microsystems, Inc. All rights reserved. # Use is subject to license terms. # #ident "%Z%%M% %I% %E% SMI" # require 5.6.1; use strict; use warnings; use POSIX; use File::Basename("basename"); my $cmdname = basename($0); my $using_scengen = 0; # 1 if using scenario simulator my $debug = 0; my $min_sleeptime = 1; my $max_sleeptime = 15; my $onecpu_sleeptime = (60 * 15); # used if only 1 CPU on system my $sleeptime = $min_sleeptime; # time to sleep between kstat updates # For timerange_foo variables, see comments at tail of &getstat() my $timerange_toohi = .01; my $timerange_hithresh = .0003; my $timerange_lothresh = $timerange_hithresh / 2; my $unsafe_timerange = .02; my $statslen = 60; # time period (in secs) to keep in @deltas # Parse arguments. intrd does not accept any public arguments; the two # arguments below are meant for testing purposes. -D generates a significant # amount of syslog output. -S loads the filename as a perl # script. That file is expected to implement a kstat "simulator" which # can be used to feed information to intrd and verify intrd's responses. while ($_ = shift @ARGV) { if ($_ eq "-S" && $#ARGV != -1) { $using_scengen = 1; do $ARGV[0]; # load simulator shift @ARGV; } elsif ($_ eq "-D") { $debug = 1; } } if ($using_scengen == 0) { require Sun::Solaris::Kstat; require Sun::Solaris::Intrs; import Sun::Solaris::Intrs(qw(intrmove)); require Sys::Syslog; import Sys::Syslog; openlog($cmdname, 'pid', 'daemon'); setlogmask(Sys::Syslog::LOG_UPTO($debug > 0 ? &Sys::Syslog::LOG_DEBUG : &Sys::Syslog::LOG_INFO)); } my $asserted = 0; my $assert_level = 'debug'; # syslog level for assertion failures sub VERIFY($@) { my $bad = (shift() == 0); # $_[0] == 0 means assert failed if ($bad) { my $msg = shift(); syslog($assert_level, "VERIFY: $msg", @_); $asserted++; } return ($bad); } sub getstat($); sub generate_delta($$); sub compress_deltas($); sub dumpdelta($); sub goodness($); sub imbalanced($$); sub do_reconfig($); sub goodness_cpu($$); # private function sub move_intr($$$$); # private function sub ivecs_to_string(@); # private function sub do_find_goal($$$$); # private function sub find_goal($$); # private function sub do_reconfig_cpu2cpu($$$$); # private function sub do_reconfig_cpu($$$); # private function # # What follow are the basic data structures routines of intrd. # # getstat() is responsible for reading the kstats and generating a "stat" hash. # # generate_delta() is responsible for taking two "stat" hashes and creating # a new "delta" hash that represents what has changed over time. # # compress_deltas() is responsible for taking a list of deltas and generating # a single delta hash that encompasses all the time periods described by the # deltas. # # getstat() is handed a reference to a kstat and generates a hash, returned # by reference, containing all the fields from the kstats which we need. # If it returns the scalar 0, it failed to gather the kstats, and the caller # should react accordingly. # # getstat() is also responsible for maintaining a reasonable $sleeptime. # # {"snaptime"} kstat's snaptime # {} one hash reference per online cpu # ->{"tot"} == cpu::sys:cpu_nsec_{user + kernel + idle} # ->{"crtime"} == cpu::sys:crtime # ->{"ivecs"} # ->{} iterates over pci_intrs::config:cookie # ->{"time"} == pci_intrs::config:time (in nsec) # ->{"pil"} == pci_intrs::config:pil # ->{"crtime"} == pci_intrs::config:crtime # ->{"ino"} == pci_intrs::config:ino # ->{"buspath"} == pci_intrs::config:buspath # ->{"name"} == pci_intrs::config:name # ->{"ihs"} == pci_intrs::config:ihs # sub getstat($) { my ($ks) = @_; my $cpucnt = 0; my %stat = (); my ($minsnap, $maxsnap); # kstats are not generated atomically. Each kstat hierarchy will # have been generated within the kernel at a different time. On a # thrashing system, we may not run quickly enough in order to get # coherent kstat timing information across all the kstats. To # determine if this is occurring, $minsnap/$maxsnap are used to # find the breadth between the first and last snaptime of all the # kstats we access. $maxsnap - $minsnap roughly represents the # total time taken up in getstat(). If this time approaches the # time between snapshots, our results may not be useful. $minsnap = -1; # snaptime is always a positive number $maxsnap = $minsnap; # Iterate over the cpus in cpu:::. Check # cpu_info::cpu_info:state to make sure the # processor is "on-line". If not, it isn't accepting interrupts # and doesn't concern us. # # Record cpu::sys:snaptime, and check $minsnap/$maxsnap. while (my ($cpu, $cpst) = each %{$ks->{cpu}}) { next if !exists($ks->{cpu_info}{$cpu}{"cpu_info$cpu"}{state}); my $state = $ks->{cpu_info}{$cpu}{"cpu_info$cpu"}{state}; next if ($state !~ /^on-line\0/); my $cpu_sys = $cpst->{sys}; $stat{$cpu}{tot} = ($cpu_sys->{cpu_nsec_idle} + $cpu_sys->{cpu_nsec_user} + $cpu_sys->{cpu_nsec_kernel}); $stat{$cpu}{crtime} = $cpu_sys->{crtime}; $stat{$cpu}{ivecs} = {}; if ($minsnap == -1 || $cpu_sys->{snaptime} < $minsnap) { $minsnap = $cpu_sys->{snaptime}; } if ($cpu_sys->{snaptime} > $maxsnap) { $maxsnap = $cpu_sys->{snaptime}; } $cpucnt++; } if ($cpucnt <= 1) { $sleeptime = $onecpu_sleeptime; return (0); # nothing to do with 1 CPU } # Iterate over the ivecs. If the cpu is not on-line, ignore the # ivecs mapped to it, if any. # # Record pci_intrs:{inum}:config:time, snaptime, crtime, pil, # ino, name, and buspath. Check $minsnap/$maxsnap. foreach my $inst (values(%{$ks->{pci_intrs}})) { my $intrcfg = $inst->{config}; my $cpu = $intrcfg->{cpu}; next unless exists $stat{$cpu}; if ($intrcfg->{snaptime} < $minsnap) { $minsnap = $intrcfg->{snaptime}; } elsif ($intrcfg->{snaptime} > $maxsnap) { $maxsnap = $intrcfg->{snaptime}; } my $cookie = "$intrcfg->{buspath} $intrcfg->{ino}"; if (exists $stat{$cpu}{ivecs}{$cookie}) { my $cookiestats = $stat{$cpu}{ivecs}{$cookie}; $cookiestats->{time} += $intrcfg->{time}; $cookiestats->{name} .= "/$intrcfg->{name}"; # If this new interrupt sharing $cookie represents a # change from an earlier getstat, make sure that # generate_delta will see the change by setting # crtime to the most recent crtime of its components. if ($intrcfg->{crtime} > $cookiestats->{crtime}) { $cookiestats->{crtime} = $intrcfg->{crtime}; } $cookiestats->{ihs}++; next; } $stat{$cpu}{ivecs}{$cookie}{time} = $intrcfg->{time}; $stat{$cpu}{ivecs}{$cookie}{crtime} = $intrcfg->{crtime}; $stat{$cpu}{ivecs}{$cookie}{pil} = $intrcfg->{pil}; $stat{$cpu}{ivecs}{$cookie}{ino} = $intrcfg->{ino}; $stat{$cpu}{ivecs}{$cookie}{buspath} = $intrcfg->{buspath}; $stat{$cpu}{ivecs}{$cookie}{name} = $intrcfg->{name}; $stat{$cpu}{ivecs}{$cookie}{ihs} = 1; } # We define the timerange as the amount of time spent gathering the # various kstats, divided by our sleeptime. If we take a lot of time # to access the kstats, and then we create a delta comparing these # kstats with a prior set of kstats, that delta will cover # substaintially different amount of time depending upon which # interrupt or CPU is being examined. # # By checking the timerange here, we guarantee that any deltas # created from these kstats will contain self-consistent data, # in that all CPUs and interrupts cover a similar span of time. # # We attempt to keep this timerange between $timerange_lothresh and # $timerange_hithresh. If the timerange gets too large, not only are # there the accuracy concerns above, but it means that intrd is using # a lot of CPU time. If the timerange gets too small, that means our # sleep time is large, and we could fail to react quickly enough to a # sudden change. # # Finally, $timerange_toohi is the upper bound. Any timerange above # this is thrown out as garbage. If the stat is safely within this # bound, we treat the stat as representing an instant in time, rather # than the time range it actually spans. We arbitrarily choose minsnap # as the snaptime of the stat. $stat{snaptime} = $minsnap; my $timerange = ($maxsnap - $minsnap) / $sleeptime; if ($sleeptime == $onecpu_sleeptime) { $sleeptime = $min_sleeptime; # time to come out of idling } elsif ($timerange > $timerange_hithresh && $sleeptime < $max_sleeptime) { $sleeptime++; } elsif ($timerange < $timerange_lothresh && $sleeptime > $min_sleeptime) { $sleeptime--; } return (0) if ($timerange > $timerange_toohi); # i.e. failure return (\%stat); } # # dumpdelta takes a reference to our "delta" structure: # {"missing"} "1" if the delta's component stats had inconsistencies # {"minsnap"} time of the first kstat snaptime used in this delta # {"maxsnap"} time of the last kstat snaptime used in this delta # {"goodness"} cost function applied to this delta # {"avgintrload"} avg of interrupt load across cpus, as a percentage # {"avgintrnsec"} avg number of nsec spent in interrupts, per cpu # {} iterates over on-line cpus # ->{"intrs"} cpu's movable intr time (sum of "time" for each ivec) # ->{"tot"} CPU load from all sources # ->{"bigintr"} largest value of {ivecs}{}{time} from below # ->{"intrload"} intrs / tot # ->{"ivecs"} # ->{} iterates over ivecs for this cpu # ->{"time"} time used by this interrupt (in nsec) # ->{"pil"} pil level of this interrupt # ->{"ino"} interrupt number # ->{"buspath"} filename of the directory of the device's bus # ->{"name"} device name # ->{"ihs"} number of different handlers sharing this ino # # It prints out the delta structure in a nice, human readable display. # sub dumpdelta($) { my ($delta) = @_; # print global info syslog('debug', "dumpdelta:"); syslog('debug', " RECONFIGURATION IN DELTA") if $delta->{missing} > 0; syslog('debug', " avgintrload: %5.2f%% avgintrnsec: %d", $delta->{avgintrload} * 100, $delta->{avgintrnsec}); syslog('debug', " goodness: %5.2f%%", $delta->{goodness} * 100) if exists($delta->{goodness}); # iterate over cpus while (my ($cpu, $cpst) = each %$delta) { next if !ref($cpst); # skip non-cpuid entries my $tot = $cpst->{tot}; syslog('debug', " cpu %3d intr %7.3f%% (bigintr %7.3f%%)", $cpu, $cpst->{intrload}*100, $cpst->{bigintr}*100/$tot); syslog('debug', " intrs %d, bigintr %d", $cpst->{intrs}, $cpst->{bigintr}); # iterate over ivecs on this cpu while (my ($ivec, $ivst) = each %{$cpst->{ivecs}}) { syslog('debug', " %15s:\"%s\": %7.3f%% %d", ($ivst->{ihs} > 1 ? "$ivst->{name}($ivst->{ihs})" : $ivst->{name}), $ivec, $ivst->{time}*100 / $tot, $ivst->{time}); } } } # # generate_delta($stat, $newstat) takes two stat references, returned from # getstat(), and creates a %delta. %delta (not surprisingly) contains the # same basic info as stat and newstat, but with the timestamps as deltas # instead of absolute times. We return a reference to the delta. # sub generate_delta($$) { my ($stat, $newstat) = @_; my %delta = (); my $intrload; my $intrnsec; my $cpus; # Take the worstcase timerange $delta{minsnap} = $stat->{snaptime}; $delta{maxsnap} = $newstat->{snaptime}; if (VERIFY($delta{maxsnap} > $delta{minsnap}, "generate_delta: stats aren't ascending")) { $delta{missing} = 1; return (\%delta); } # if there are a different number of cpus in the stats, set missing $delta{missing} = (keys(%$stat) != keys(%$newstat)); if (VERIFY($delta{missing} == 0, "generate_delta: number of CPUs changed")) { return (\%delta); } # scan through every cpu in %newstat and compare against %stat while (my ($cpu, $newcpst) = each %$newstat) { next if !ref($newcpst); # skip non-cpuid fields # If %stat is missing a cpu from %newstat, then it was just # onlined. Mark missing. if (VERIFY(exists $stat->{$cpu} && $stat->{$cpu}{crtime} == $newcpst->{crtime}, "generate_delta: cpu $cpu changed")) { $delta{missing} = 1; return (\%delta); } my $cpst = $stat->{$cpu}; $delta{$cpu}{tot} = $newcpst->{tot} - $cpst->{tot}; if (VERIFY($delta{$cpu}{tot} >= 0, "generate_delta: deltas are not ascending?")) { $delta{missing} = 1; delete($delta{$cpu}); return (\%delta); } # Avoid remote chance of division by zero $delta{$cpu}{tot} = 1 if $delta{$cpu}{tot} == 0; $delta{$cpu}{intrs} = 0; $delta{$cpu}{bigintr} = 0; my %ivecs = (); $delta{$cpu}{ivecs} = \%ivecs; # if the number of ivecs differs, set missing if (VERIFY(keys(%{$cpst->{ivecs}}) == keys(%{$newcpst->{ivecs}}), "generate_delta: cpu $cpu has more/less". " interrupts")) { $delta{missing} = 1; return (\%delta); } while (my ($inum, $newivec) = each %{$newcpst->{ivecs}}) { # If this ivec doesn't exist in $stat, or if $stat # shows a different crtime, set missing. if (VERIFY(exists $cpst->{ivecs}{$inum} && $cpst->{ivecs}{$inum}{crtime} == $newivec->{crtime}, "generate_delta: cpu $cpu inum $inum". " has changed")) { $delta{missing} = 1; return (\%delta); } my $ivec = $cpst->{ivecs}{$inum}; # Create $delta{$cpu}{ivecs}{$inum}. my %dltivec = (); $delta{$cpu}{ivecs}{$inum} = \%dltivec; # calculate time used by this interrupt my $time = $newivec->{time} - $ivec->{time}; if (VERIFY($time >= 0, "generate_delta: ivec went backwards?")) { $delta{missing} = 1; delete($delta{$cpu}{ivecs}{$inum}); return (\%delta); } $delta{$cpu}{intrs} += $time; $dltivec{time} = $time; if ($time > $delta{$cpu}{bigintr}) { $delta{$cpu}{bigintr} = $time; } # Transfer over basic info about the kstat. We # don't have to worry about discrepancies between # ivec and newivec because we verified that both # have the same crtime. $dltivec{pil} = $newivec->{pil}; $dltivec{ino} = $newivec->{ino}; $dltivec{buspath} = $newivec->{buspath}; $dltivec{name} = $newivec->{name}; $dltivec{ihs} = $newivec->{ihs}; } if ($delta{$cpu}{tot} < $delta{$cpu}{intrs}) { # Ewww! Hopefully just a rounding error. # Make something up. $delta{$cpu}{tot} = $delta{$cpu}{intrs}; } $delta{$cpu}{intrload} = $delta{$cpu}{intrs} / $delta{$cpu}{tot}; $intrload += $delta{$cpu}{intrload}; $intrnsec += $delta{$cpu}{intrs}; $cpus++; } if ($cpus > 0) { $delta{avgintrload} = $intrload / $cpus; $delta{avgintrnsec} = $intrnsec / $cpus; } else { $delta{avgintrload} = 0; $delta{avgintrnsec} = 0; } return (\%delta); } # compress_delta takes a list of deltas, and returns a single new delta # which represents the combined information from all the deltas. The deltas # provided are assumed to be sequential in time. The resulting compressed # delta looks just like any other delta. This new delta is also more accurate # since its statistics are averaged over a longer period than any of the # original deltas. sub compress_deltas ($) { my ($deltas) = @_; my %newdelta = (); my ($intrs, $tot); my $cpus = 0; if (VERIFY($#$deltas != -1, "compress_deltas: list of delta is empty?")) { return (0); } $newdelta{minsnap} = $deltas->[0]{minsnap}; $newdelta{maxsnap} = $deltas->[$#$deltas]{maxsnap}; $newdelta{missing} = 0; foreach my $delta (@$deltas) { if (VERIFY($delta->{missing} == 0, "compressing bad deltas?")) { return (0); } while (my ($cpuid, $cpu) = each %$delta) { next if !ref($cpu); $intrs += $cpu->{intrs}; $tot += $cpu->{tot}; $newdelta{$cpuid}{intrs} += $cpu->{intrs}; $newdelta{$cpuid}{tot} += $cpu->{tot}; if (!exists $newdelta{$cpuid}{ivecs}) { my %ivecs = (); $newdelta{$cpuid}{ivecs} = \%ivecs; } while (my ($inum, $ivec) = each %{$cpu->{ivecs}}) { my $newivecs = $newdelta{$cpuid}{ivecs}; $newivecs->{$inum}{time} += $ivec->{time}; $newivecs->{$inum}{pil} = $ivec->{pil}; $newivecs->{$inum}{ino} = $ivec->{ino}; $newivecs->{$inum}{buspath} = $ivec->{buspath}; $newivecs->{$inum}{name} = $ivec->{name}; $newivecs->{$inum}{ihs} = $ivec->{ihs}; } } } foreach my $cpu (values(%newdelta)) { next if !ref($cpu); # ignore non-cpu fields $cpus++; my $bigintr = 0; foreach my $ivec (values(%{$cpu->{ivecs}})) { if ($ivec->{time} > $bigintr) { $bigintr = $ivec->{time}; } } $cpu->{bigintr} = $bigintr; $cpu->{intrload} = $cpu->{intrs} / $cpu->{tot}; $cpu->{tot} = 1 if $cpu->{tot} <= 0; } if ($cpus == 0) { $newdelta{avgintrnsec} = 0; $newdelta{avgintrload} = 0; } else { $newdelta{avgintrnsec} = $intrs / $cpus; $newdelta{avgintrload} = $intrs / $tot; } return (\%newdelta); } # What follow are the core functions responsible for examining the deltas # generated above and deciding what to do about them. # # goodness() and its helper goodness_cpu() return a heuristic which describe # how good (or bad) the current interrupt balance is. The value returned will # be between 0 and 1, with 0 representing maximum goodness, and 1 representing # maximum badness. # # imbalanced() compares a current and historical value of goodness, and # determines if there has been enough change to warrant evaluating a # reconfiguration of the interrupts # # do_reconfig(), and its helpers, do_reconfig_cpu(), do_reconfig_cpu2cpu(), # find_goal(), do_find_goal(), and move_intr(), are responsible for examining # a delta and determining the best possible assignment of interrupts to CPUs. # # It is important that do_reconfig() be in alignment with goodness(). If # do_reconfig were to generate a new interrupt distribution that worsened # goodness, we could get into a pathological loop with intrd fighting itself, # constantly deciding that things are imbalanced, and then changing things # only to make them worse. # any goodness over $goodness_unsafe_load is considered really bad # goodness must drop by at least $goodness_mindelta for a reconfig my $goodness_unsafe_load = .9; my $goodness_mindelta = .1; # goodness(%delta) examines a delta and return its "goodness". goodness will # be between 0 (best) and 1 (major bad). goodness is determined by evaluating # the goodness of each individual cpu, and returning the worst case. This # helps on systems with many CPUs, where otherwise a single pathological CPU # might otherwise be ignored because the average was OK. # # To calculate the goodness of an individual CPU, we start by looking at its # load due to interrupts. If the load is above a certain high threshold and # there is more than one interrupt assigned to this CPU, we set goodness # to worst-case. If the load is below the average interrupt load of all CPUs, # then we return best-case, since what's to complain about? # # Otherwise we look at how much the load is above the average, and return # that as the goodness, with one caveat: we never return more than the CPU's # interrupt load ignoring its largest single interrupt source. This is # because a CPU with one high-load interrupt, and no other interrupts, is # perfectly balanced. Nothing can be done to improve the situation, and thus # it is perfectly balanced even if the interrupt's load is 100%. sub goodness($) { my ($delta) = @_; return (1) if $delta->{missing} > 0; my $high_goodness = 0; my $goodness; foreach my $cpu (values(%$delta)) { next if !ref($cpu); # skip non-cpuid fields $goodness = goodness_cpu($cpu, $delta->{avgintrload}); if (VERIFY($goodness >= 0 && $goodness <= 1, "goodness: cpu goodness out of range?")) { dumpdelta($delta); return (1); } if ($goodness == 1) { return (1); # worst case, no need to continue } if ($goodness > $high_goodness) { $high_goodness = $goodness; } } return ($high_goodness); } sub goodness_cpu($$) # private function { my ($cpu, $avgintrload) = @_; my $goodness; my $load = $cpu->{intrs} / $cpu->{tot}; return (0) if ($load < $avgintrload); # low loads are perfectly good # Calculate $load_no_bigintr, which represents the load # due to interrupts, excluding the one biggest interrupt. # This is the most gain we can get on this CPU from # offloading interrupts. my $load_no_bigintr = ($cpu->{intrs} - $cpu->{bigintr}) / $cpu->{tot}; # A major imbalance is indicated if a CPU is saturated # with interrupt handling, and it has more than one # source of interrupts. Those other interrupts could be # starved if of a lower pil. Return a goodness of 1, # which is the worst possible return value, # which will effectively contaminate this entire delta. my $cnt = keys(%{$cpu->{ivecs}}); if ($load > $goodness_unsafe_load && $cnt > 1) { return (1); } $goodness = $load - $avgintrload; if ($goodness > $load_no_bigintr) { $goodness = $load_no_bigintr; } return ($goodness); } # imbalanced() is used by the main routine to determine if the goodness # has shifted far enough from our last baseline to warrant a reassignment # of interrupts. A very high goodness indicates that a CPU is way out of # whack. If the goodness has varied too much since the baseline, then # perhaps a reconfiguration is worth considering. sub imbalanced ($$) { my ($goodness, $baseline) = @_; # Return 1 if we are pathological, or creeping away from the baseline return (1) if $goodness > .50; return (1) if abs($goodness - $baseline) > $goodness_mindelta; return (0); } # do_reconfig(), do_reconfig_cpu(), and do_reconfig_cpu2cpu(), are the # decision-making functions responsible for generating a new interrupt # distribution. They are designed with the definition of goodness() in # mind, i.e. they use the same definition of "good distribution" as does # goodness(). # # do_reconfig() is responsible for deciding whether a redistribution is # actually warranted. If the goodness is already pretty good, it doesn't # waste the CPU time to generate a new distribution. If it # calculates a new distribution and finds that it is not sufficiently # improved from the prior distirbution, it will not do the redistribution, # mainly to avoid the disruption to system performance caused by # rejuggling interrupts. # # Its main loop works by going through a list of cpus sorted from # highest to lowest interrupt load. It removes the highest-load cpus # one at a time and hands them off to do_reconfig_cpu(). This function # then re-sorts the remaining CPUs from lowest to highest interrupt load, # and one at a time attempts to rejuggle interrupts between the original # high-load CPU and the low-load CPU. Rejuggling on a high-load CPU is # considered finished as soon as its interrupt load is within # $goodness_mindelta of the average interrupt load. Such a CPU will have # a goodness of below the $goodness_mindelta threshold. # # move_intr(\%delta, $inum, $oldcpu, $newcpu) # used by reconfiguration code to move an interrupt between cpus within # a delta. This manipulates data structures, and does not actually move # the interrupt on the running system. # sub move_intr($$$$) # private function { my ($delta, $inum, $oldcpuid, $newcpuid) = @_; my $ivec = $delta->{$oldcpuid}{ivecs}{$inum}; # Remove ivec from old cpu my $oldcpu = $delta->{$oldcpuid}; $oldcpu->{intrs} -= $ivec->{time}; $oldcpu->{intrload} = $oldcpu->{intrs} / $oldcpu->{tot}; delete($oldcpu->{ivecs}{$inum}); VERIFY($oldcpu->{intrs} >= 0, "move_intr: intr's time > total time?"); VERIFY($ivec->{time} <= $oldcpu->{bigintr}, "move_intr: intr's time > bigintr?"); if ($ivec->{time} >= $oldcpu->{bigintr}) { my $bigtime = 0; foreach my $ivec (values(%{$oldcpu->{ivecs}})) { $bigtime = $ivec->{time} if $ivec->{time} > $bigtime; } $oldcpu->{bigintr} = $bigtime; } # Add ivec onto new cpu my $newcpu = $delta->{$newcpuid}; $ivec->{nowcpu} = $newcpuid; $newcpu->{intrs} += $ivec->{time}; $newcpu->{intrload} = $newcpu->{intrs} / $newcpu->{tot}; $newcpu->{ivecs}{$inum} = $ivec; $newcpu->{bigintr} = $ivec->{time} if $ivec->{time} > $newcpu->{bigintr}; } sub move_intr_check($$$) # private function { my ($delta, $oldcpuid, $newcpuid) = @_; VERIFY($delta->{$oldcpuid}{tot} >= $delta->{$oldcpuid}{intrs}, "Moved interrupts left 100+%% load on src cpu"); VERIFY($delta->{$newcpuid}{tot} >= $delta->{$newcpuid}{intrs}, "Moved interrupts left 100+%% load on tgt cpu"); } sub ivecs_to_string(@) # private function { my $str = ""; foreach my $ivec (@_) { $str = "$str $ivec->{inum}"; } return ($str); } sub do_reconfig($) { my ($delta) = @_; my $goodness = $delta->{goodness}; # We can't improve goodness to better than 0. We should stop here # if, even if we achieve a goodness of 0, the improvement is still # too small to merit the action. if ($goodness - 0 < $goodness_mindelta) { syslog('debug', "goodness good enough, don't reconfig"); return (0); } syslog('notice', "Optimizing interrupt assignments"); if (VERIFY ($delta->{missing} == 0, "RECONFIG Aborted: should not ". "have a delta with missing")) { return (-1); } # Make a list of all cpuids, and also add some extra information # to the ivec structures. my @cpusortlist = (); while (my ($cpuid, $cpu) = each %$delta) { next if !ref($cpu); # skip non-cpu entries push(@cpusortlist, $cpuid); while (my ($inum, $ivec) = each %{$cpu->{ivecs}}) { $ivec->{origcpu} = $cpuid; $ivec->{nowcpu} = $cpuid; $ivec->{inum} = $inum; } } # Sort the list of CPUs from highest to lowest interrupt load. # Remove the top CPU from that list and attempt to redistribute # its interrupts. If the CPU has a goodness below a threshold, # just ignore the CPU and move to the next one. If the CPU's # load falls below the average load plus that same threshold, # then there are no CPUs left worth reconfiguring, and we're done. while (@cpusortlist) { # Re-sort cpusortlist each time, since do_reconfig_cpu can # move interrupts around. @cpusortlist = sort({$delta->{$b}{intrload} <=> $delta->{$a}{intrload}} @cpusortlist); my $cpu = shift(@cpusortlist); if (($delta->{$cpu}{intrload} <= $goodness_unsafe_load) && ($delta->{$cpu}{intrload} <= $delta->{avgintrload} + $goodness_mindelta)) { syslog('debug', "finished reconfig: cpu $cpu load ". "$delta->{$cpu}{intrload} avgload ". "$delta->{avgintrload}"); last; } if (goodness_cpu($delta->{$cpu}, $delta->{avgintrload}) < $goodness_mindelta) { next; } do_reconfig_cpu($delta, \@cpusortlist, $cpu); } # How good a job did we do? If the improvement was minimal, and # our goodness wasn't pathological (and thus needing any help it # can get), then don't bother moving the interrupts. my $newgoodness = goodness($delta); VERIFY($newgoodness <= $goodness, "reconfig: result has worse goodness?"); if (($goodness != 1 || $newgoodness == 1) && $goodness - $newgoodness < $goodness_mindelta) { syslog('debug', "goodness already near optimum, ". "don't reconfig"); return (0); } syslog('debug', "goodness %5.2f%% --> %5.2f%%", $goodness*100, $newgoodness*100); # Time to move those interrupts! my $ret = 1; my $warned = 0; while (my ($cpuid, $cpu) = each %$delta) { next if $cpuid =~ /\D/; while (my ($inum, $ivec) = each %{$cpu->{ivecs}}) { next if ($ivec->{origcpu} == $cpuid); if (!intrmove($ivec->{buspath}, $ivec->{ino}, $cpuid)) { syslog('warning', "Unable to move interrupts") if $warned++ == 0; syslog('debug', "Unable to move buspath ". "$ivec->{buspath} ino $ivec->{ino} to ". "cpu $cpuid"); $ret = -1; } } } syslog('notice', "Interrupt assignments optimized"); return ($ret); } sub do_reconfig_cpu($$$) # private function { my ($delta, $cpusortlist, $oldcpuid) = @_; # We have been asked to rejuggle interrupts between $oldcpuid and # other CPUs found on $cpusortlist so as to improve the load on # $oldcpuid. We reverse $cpusortlist to get our own copy of the # list, sorted from lowest to highest interrupt load. One at a # time, shift a CPU off of this list of CPUs, and attempt to # rejuggle interrupts between the two CPUs. Don't do this if the # other CPU has a higher load than oldcpuid. We're done rejuggling # once $oldcpuid's goodness falls below a threshold. syslog('debug', "reconfiguring $oldcpuid"); my $cpu = $delta->{$oldcpuid}; my $avgintrload = $delta->{avgintrload}; my @cputargetlist = reverse(@$cpusortlist); # make a copy of the list while ($#cputargetlist != -1) { last if goodness_cpu($cpu, $avgintrload) < $goodness_mindelta; my $tgtcpuid = shift(@cputargetlist); my $tgt = $delta->{$tgtcpuid}; my $load = $cpu->{intrload}; my $tgtload = $tgt->{intrload}; last if $tgtload > $load; do_reconfig_cpu2cpu($delta, $oldcpuid, $tgtcpuid, $load); } } sub do_reconfig_cpu2cpu($$$$) # private function { my ($delta, $srccpuid, $tgtcpuid, $srcload) = @_; # We've been asked to consider interrupt juggling between srccpuid # (with a high interrupt load) and tgtcpuid (with a lower interrupt # load). First, make a single list with all of the ivecs from both # CPUs, and sort the list from highest to lowest load. syslog('debug', "exchanging intrs between $srccpuid and $tgtcpuid"); # Gather together all the ivecs and sort by load my @ivecs = (values(%{$delta->{$srccpuid}{ivecs}}), values(%{$delta->{$tgtcpuid}{ivecs}})); return if $#ivecs == -1; @ivecs = sort({$b->{time} <=> $a->{time}} @ivecs); # Our "goal" load for srccpuid is the average load across all CPUs. # find_goal() will find determine the optimum selection of the # available interrupts which comes closest to this goal without # falling below the goal. my $goal = $delta->{avgintrnsec}; # We know that the interrupt load on tgtcpuid is less than that on # srccpuid, but its load could still be above avgintrnsec. Don't # choose a goal which would bring srccpuid below the load on tgtcpuid. my $avgnsec = ($delta->{$srccpuid}{intrs} + $delta->{$tgtcpuid}{intrs}) / 2; if ($goal < $avgnsec) { $goal = $avgnsec; } # If the largest of the interrupts is on srccpuid, leave it there. # This can help minimize the disruption caused by moving interrupts. if ($ivecs[0]->{origcpu} == $srccpuid) { syslog('debug', "Keeping $ivecs[0]->{inum} on $srccpuid"); $goal -= $ivecs[0]->{time}; shift(@ivecs); } syslog('debug', "GOAL: inums should total $goal"); find_goal(\@ivecs, $goal); # find_goal() returned its results to us by setting $ivec->{goal} if # the ivec should be on srccpuid, or clearing it for tgtcpuid. # Call move_intr() to update our $delta with the new results. foreach my $ivec (@ivecs) { syslog('debug', "ivec $ivec->{inum} goal $ivec->{goal}"); VERIFY($ivec->{nowcpu} == $srccpuid || $ivec->{nowcpu} == $tgtcpuid, "cpu2cpu found an ". "interrupt not currently on src or tgt cpu"); if ($ivec->{goal} && $ivec->{nowcpu} != $srccpuid) { move_intr($delta, $ivec->{inum}, $ivec->{nowcpu}, $srccpuid); } elsif ($ivec->{goal} == 0 && $ivec->{nowcpu} != $tgtcpuid) { move_intr($delta, $ivec->{inum}, $ivec->{nowcpu}, $tgtcpuid); } } move_intr_check($delta, $srccpuid, $tgtcpuid); # asserts my $newload = $delta->{$srccpuid}{intrs} / $delta->{$srccpuid}{tot}; VERIFY($newload <= $srcload && $newload > $delta->{avgintrload}, "cpu2cpu: new load didn't end up in expected range"); } # find_goal() and its helper do_find_goal() are used to find the best # combination of interrupts in order to generate a load that is as close # as possible to a goal load without falling below that goal. Before returning # to its caller, find_goal() sets a new value in the hash of each interrupt, # {goal}, which if set signifies that this interrupt is one of the interrupts # identified as part of the set of interrupts which best meet the goal. # # The arguments to find_goal are a list of ivecs (hash references), sorted # by descending {time}, and the goal load. The goal is relative to {time}. # The best fit is determined by performing a depth-first search. do_find_goal # is the recursive subroutine which carries out the search. # # It is passed an index as an argument, originally 0. On a given invocation, # it is only to consider interrupts in the ivecs array starting at that index. # It then considers two possibilities: # 1) What is the best goal-fit if I include ivecs[index]? # 2) What is the best goal-fit if I exclude ivecs[index]? # To determine case 1, it subtracts the load of ivecs[index] from the goal, # and calls itself recursively with that new goal and index++. # To determine case 2, it calls itself recursively with the same goal and # index++. # # It then compares the two results, decide which one best meets the goals, # and returns the result. The return value is the best-fit's interrupt load, # followed by a list of all the interrupts which make up that best-fit. # # As an optimization, a second array loads[] is created which mirrors ivecs[]. # loads[i] will equal the total loads of all ivecs[i..$#ivecs]. This is used # by do_find_goal to avoid recursing all the way to the end of the ivecs # array if including all remaining interrupts will still leave the best-fit # at below goal load. If so, it then includes all remaining interrupts on # the goal list and returns. # sub find_goal($$) # private function { my ($ivecs, $goal) = @_; my @goals; my $load; my $ivec; if ($goal <= 0) { @goals = (); # the empty set will best meet the goal } else { syslog('debug', "finding goal from intrs %s", ivecs_to_string(@$ivecs)); # Generate @loads array my $tot = 0; foreach $ivec (@$ivecs) { $tot += $ivec->{time}; } my @loads = (); foreach $ivec (@$ivecs) { push(@loads, $tot); $tot -= $ivec->{time}; } ($load, @goals) = do_find_goal($ivecs, \@loads, $goal, 0); VERIFY($load >= $goal, "find_goal didn't meet goals"); } syslog('debug', "goals found: %s", ivecs_to_string(@goals)); # Set or clear $ivec->{goal} for each ivec, based on returned @goals foreach $ivec (@$ivecs) { if ($#goals > -1 && $ivec == $goals[0]) { syslog('debug', "inum $ivec->{inum} on source cpu"); $ivec->{goal} = 1; shift(@goals); } else { syslog('debug', "inum $ivec->{inum} on target cpu"); $ivec->{goal} = 0; } } } sub do_find_goal($$$$) # private function { my ($ivecs, $loads, $goal, $idx) = @_; if ($idx > $#{$ivecs}) { return (0); } syslog('debug', "$idx: finding goal $goal inum $ivecs->[$idx]{inum}"); my $load = $ivecs->[$idx]{time}; my @goals_with = (); my @goals_without = (); my ($with, $without); # If we include all remaining items and we're still below goal, # stop here. We can just return a result that includes $idx and all # subsequent ivecs. Since this will still be below goal, there's # nothing better to be done. if ($loads->[$idx] <= $goal) { syslog('debug', "$idx: including all remaining intrs %s with load %d", ivecs_to_string(@$ivecs[$idx .. $#{$ivecs}]), $loads->[$idx]); return ($loads->[$idx], @$ivecs[$idx .. $#{$ivecs}]); } # Evaluate the "with" option, i.e. the best matching goal which # includes $ivecs->[$idx]. If idx's load is more than our goal load, # stop here. Once we're above the goal, there is no need to consider # further interrupts since they'll only take us further from the goal. if ($goal <= $load) { $with = $load; # stop here } else { ($with, @goals_with) = do_find_goal($ivecs, $loads, $goal - $load, $idx + 1); $with += $load; } syslog('debug', "$idx: with-load $with intrs %s", ivecs_to_string($ivecs->[$idx], @goals_with)); # Evaluate the "without" option, i.e. the best matching goal which # excludes $ivecs->[$idx]. ($without, @goals_without) = &do_find_goal($ivecs, $loads, $goal, $idx + 1); syslog('debug', "$idx: without-load $without intrs %s", ivecs_to_string(@goals_without)); # We now have our "with" and "without" options, and we choose which # best fits the goal. If one is greater than goal and the other is # below goal, we choose the one that is greater. If they are both # below goal, then we choose the one that is greater. If they are # both above goal, then we choose the smaller. my $which; # 0 == with, 1 == without if ($with >= $goal && $without < $goal) { $which = 0; } elsif ($with < $goal && $without >= $goal) { $which = 1; } elsif ($with >= $goal && $without >= $goal) { $which = ($without < $with); } else { $which = ($without > $with); } # Return the load of our best case scenario, followed by all the ivecs # which compose that goal. if ($which == 1) { # without syslog('debug', "$idx: going without"); return ($without, @goals_without); } else { syslog('debug', "$idx: going with"); return ($with, $ivecs->[$idx], @goals_with); } # Not reached } syslog('debug', "intrd is starting".($debug ? " (debug)" : "")); my @deltas = (); my $deltas_tottime = 0; # sum of maxsnap-minsnap across @deltas my $avggoodness; my $baseline_goodness = 0; my $compdelta; my $do_reconfig; # temp variables my $goodness; my $deltatime; my $olddelta; my $olddeltatime; my $delta; my $newstat; my $below_statslen; my $newtime; my $ret; my $gotsig = 0; $SIG{INT} = sub { $gotsig = 1; }; # don't die in the middle of retargeting $SIG{HUP} = $SIG{INT}; $SIG{TERM} = $SIG{INT}; my $ks; if ($using_scengen == 0) { $ks = Sun::Solaris::Kstat->new(); } else { $ks = myks_update(); # supplied by the simulator } # If no pci_intrs kstats were found, we need to exit, but we can't because # SMF will restart us and/or report an error to the administrator. But # there's nothing an administrator can do. So print out a message for SMF # logs and silently pause forever. if (!exists($ks->{pci_intrs})) { print STDERR "$cmdname: no interrupts were found; ". "your PCI bus may not yet be supported\n"; pause() while $gotsig == 0; exit 0; } my $stat = getstat($ks); for (;;) { sub clear_deltas { @deltas = (); $deltas_tottime = 0; $stat = 0; # prevent next gen_delta() from setting {missing} } # 1. Sleep, update the kstats, and save the new stats in $newstat. exit 0 if $gotsig; # if we got ^C / SIGTERM, exit if ($using_scengen == 0) { sleep($sleeptime); exit 0 if $gotsig; # if we got ^C / SIGTERM, exit $ks->update(); } else { $ks = myks_update(); } $newstat = getstat($ks); # $stat or $newstat could be zero if they're uninitialized, or if # getstat() failed. If $stat is zero, move $newstat to $stat, sleep # and try again. If $newstat is zero, then we also sleep and try # again, hoping the problem will clear up. next if (!ref $newstat); if (!ref $stat) { $stat = $newstat; next; } # 2. Compare $newstat with the prior set of values, result in %$delta. $delta = generate_delta($stat, $newstat); dumpdelta($delta) if $debug; # Dump most recent stats to stdout. $stat = $newstat; # The new stats now become the old stats. # 3. If $delta->{missing}, then there has been a reconfiguration of # either cpus or interrupts (probably both). We need to toss out our # old set of statistics and start from scratch. # # Also, if the delta covers a very long range of time, then we've # been experiencing a system overload that has resulted in intrd # not being allowed to run effectively for a while now. As above, # toss our old statistics and start from scratch. $deltatime = $delta->{maxsnap} - $delta->{minsnap}; if ($delta->{missing} > 0 || $deltatime > $statslen) { clear_deltas(); syslog('debug', "evaluating interrupt assignments"); next; } # 4. Incorporate new delta into the list of deltas, and associated # statistics. If we've just now received $statslen deltas, then it's # time to evaluate a reconfiguration. $below_statslen = ($deltas_tottime < $statslen); $deltas_tottime += $deltatime; $do_reconfig = ($below_statslen && $deltas_tottime >= $statslen); push(@deltas, $delta); # 5. Remove old deltas if total time is more than $statslen. We use # @deltas as a moving average of the last $statslen seconds. Shift # off the olders deltas, but only if that doesn't cause us to fall # below $statslen seconds. while (@deltas > 1) { $olddelta = $deltas[0]; $olddeltatime = $olddelta->{maxsnap} - $olddelta->{minsnap}; $newtime = $deltas_tottime - $olddeltatime; last if ($newtime < $statslen); shift(@deltas); $deltas_tottime = $newtime; } # 6. The brains of the operation are here. First, check if we're # imbalanced, and if so set $do_reconfig. If $do_reconfig is set, # either because of imbalance or above in step 4, we evaluate a # new configuration. # # First, take @deltas and generate a single "compressed" delta # which summarizes them all. Pass that to do_reconfig and see # what it does with it: # # $ret == -1 : failure # $ret == 0 : current config is optimal (or close enough) # $ret == 1 : reconfiguration has occurred # # If $ret is -1 or 1, dump all our deltas and start from scratch. # Step 4 above will set do_reconfig soon thereafter. # # If $ret is 0, then nothing has happened because we're already # good enough. Set baseline_goodness to current goodness. $compdelta = compress_deltas(\@deltas); if (VERIFY(ref($compdelta) eq "HASH", "couldn't compress deltas")) { clear_deltas(); next; } $compdelta->{goodness} = goodness($compdelta); dumpdelta($compdelta) if $debug; $goodness = $compdelta->{goodness}; syslog('debug', "GOODNESS: %5.2f%%", $goodness * 100); if ($deltas_tottime >= $statslen && imbalanced($goodness, $baseline_goodness)) { $do_reconfig = 1; } if ($do_reconfig) { $ret = do_reconfig($compdelta); if ($ret != 0) { clear_deltas(); syslog('debug', "do_reconfig FAILED!") if $ret == -1; } else { syslog('debug', "setting new baseline of $goodness"); $baseline_goodness = $goodness; } } syslog('debug', "---------------------------------------"); }