原文链接 https://liutaihua.github.io/2013/04/25/process-smaps-analysis.html
注：以下为加速网络访问所做的原文缓存，经过重新格式化，可能存在格式方面的问题，或偶有遗漏信息，请以原文为准。
2.6.16后的内核, 对于查看进程内存使用分布, 更方便了. 在/proc/{pid} 路径下有一个smaps文件, 记录了进程内存使用情况, 在老的内核系统上, 这个文件是maps或memap , 而且老的内核下maps或memap文件记录的数据真不是人读的.
现在有了高内核, 当然可以用起来了.
smaps文件内容格式是: 7f4913d8f000-7f4913ddd000 r-xp 00000000 fd:00 791940 /usr/local/boost149/lib/libboost_python.so.1.49.0 Size: 312 kB Rss: 20 kB Pss: 2 kB Shared_Clean: 20 kB Shared_Dirty: 0 kB Private_Clean: 0 kB Private_Dirty: 0 kB Referenced: 20 kB Anonymous: 0 kB AnonHugePages: 0 kB Swap: 0 kB KernelPageSize: 4 kB MMUPageSize: 4 kB size： 是进程使用内存空间，并不一定实际分配了物理内存；
Rss： "Resident Set Size"，实际驻留"在内存中"的内存数. 不包括已经交换出去的页面。RSS还包括了与其它进程共享的内存区域，通常用于共享库；
Pss： Private Rss， Rss中私有的内存页面；
Shared_Clean： Rss中和其他进程共享的未改写页面；
Shared_Dirty： Rss和其他进程共享的已改写页面；
Private_Clean： Rss中改写的私有页面页面；
Private_Dirty： Rss中已改写的私有页面页面；
(其中Dirty页面如果没有交换机制的情况下，应该是不能回收的)
网上有仁兄使用的过滤分析此文件内容的perl脚本, 借来用:
!/usr/bin/perl
Copyright Ben Maurer
you can distribute this under the MIT/X11 License
use Linux::Smaps;
my $pid=shift @ARGV; unless ($pid) { print "./smem.pl /n"; exit 1; } my $map=Linux::Smaps->new($pid); my @VMAs = $map->vmas;
format STDOUT = VMSIZE: @######## kb $map->size RSS: @######## kb total $map->rss @######## kb shared $map->shared_clean + $map->shared_dirty @######## kb private clean $map->private_clean @######## kb private dirty $map->private_dirty .
write;
printPrivateMappings (); printSharedMappings ();
sub sharedMappings () { return grep { ($->shared_clean + $->shared_dirty) > 0 } @VMAs; }
sub privateMappings () { return grep { ($->private_clean + $->private_dirty) > 0 } @VMAs; }
sub printPrivateMappings () { $TYPE = "PRIVATE MAPPINGS"; $^ = 'SECTION_HEADER'; $~ = 'SECTION_ITEM'; $- = 0; $= = 100000000; foreach $vma (sort {-($a->private_dirty <=> $b->private_dirty)} privateMappings ()) { $size = $vma->size; $dirty = $vma->private_dirty; $clean = $vma->private_clean; $file = $vma->file_name; write; } }
sub printSharedMappings () { $TYPE = "SHARED MAPPINGS"; $^ = 'SECTION_HEADER'; $~ = 'SECTION_ITEM'; $- = 0; $= = 100000000;
foreach $vma (sort {-(($a->shared_clean + $a->shared_dirty)
<=>
($b->shared_clean + $b->shared_dirty))}
sharedMappings ()) {
$size = $vma->size; $dirty = $vma->shared_dirty; $clean = $vma->shared_clean; $file = $vma->file_name; write;
}
}
format SECTION_HEADER = @<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< $TYPE @>>>>>>>>>> @>>>>>>>>>> @>>>>>>>>> @<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< "vmsize" "rss clean" "rss dirty" "file" .
format SECTION_ITEM = @####### kb @####### kb @####### kb @<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< $size $clean $dirty $file .
使用之前需要先安装Linux::Smaps模块: perl -MCPAN -e 'install Linux::Smaps'
使用之: [root@211 tmp]# perl p.pl 6490 VMSIZE: 835412 kb RSS: 274928 kb total 2880 kb shared 8112 kb private clean 263936 kb private dirty PRIVATE MAPPINGS vmsize rss clean rss dirty file 406380 kb 7432 kb 258816 kb [heap] 6272 kb 0 kb 3612 kb 4048 kb 28 kb 364 kb /terminus/crown/bin/ares 520 kb 56 kb 216 kb 604 kb 112 kb 176 kb 260 kb 20 kb 160 kb 240 kb 32 kb 128 kb /usr/lib64/libpython2.6.so.1.0 260 kb 44 kb 92 kb 772 kb 236 kb 44 kb 224 kb 0 kb 44 kb [stack] 56 kb 8 kb 40 kb 80 kb 20 kb 36 kb 72 kb 12 kb 24 kb /terminus/crown/bin/ares 928 kb 0 kb 24 kb /usr/lib64/libstdc++.so.6.0.13 10240 kb 0 kb 20 kb 132 kb 0 kb 16 kb 20 kb 0 kb 16 kb 1476 kb 0 kb 16 kb /usr/lib64/libpython2.6.so.1.0 10240 kb 0 kb 12 kb 10240 kb 0 kb 8 kb 16 kb 0 kb 8 kb /usr/lib64/python2.6/lib-dynload/datetime.so 8 kb 0 kb 8 kb /usr/lib64/libstdc++.so.6.0.13 4 kb 0 kb 4 kb /usr/lib64/python2.6/lib-dynload/syslog.so 16 kb 12 kb 4 kb /lib64/libc-2.12.so 4 kb 0 kb 4 kb /lib64/libc-2.12.so 4 kb 0 kb 4 kb /lib64/libm-2.12.so 4 kb 0 kb 4 kb /lib64/libm-2.12.so 28 kb 20 kb 4 kb /usr/lib64/libstdc++.so.6.0.13 84 kb 8 kb 4 kb 92 kb 0 kb 4 kb /lib64/libpthread-2.12.so 4 kb 0 kb 4 kb /lib64/libpthread-2.12.so 4 kb 0 kb 4 kb /lib64/libpthread-2.12.so 16 kb 0 kb 4 kb 12 kb 4 kb 4 kb /usr/lib64/libcurl.so.4.1.1 128 kb 0 kb 4 kb /lib64/ld-2.12.so 4 kb 0 kb 4 kb /lib64/ld-2.12.so 4 kb 4 kb 0 kb /usr/lib64/python2.6/lib-dynload/_randommodule.so 8 kb 8 kb 0 kb /usr/lib64/python2.6/lib-dynload/timemodule.so 4 kb 4 kb 0 kb /usr/lib64/python2.6/lib-dynload/_functoolsmodule.so 4 kb 4 kb 0 kb /usr/lib64/python2.6/lib-dynload/_json.so 4 kb 4 kb 0 kb /lib64/librt-2.12.so 16 kb 8 kb 0 kb /usr/local/boost149/lib/libboost_python.so.1.49.0 4 kb 4 kb 0 kb /usr/local/boost149/lib/libboost_system.so.1.49.0 8 kb 4 kb 0 kb /usr/local/boost149/lib/libboost_thread.so.1.49.0 16 kb 16 kb 0 kb /usr/local/lib/libzmq.so.1.0.0 12 kb 4 kb 0 kb 4 kb 4 kb 0 kb /lib64/ld-2.12.so 4 kb 4 kb 0 kb
SHARED MAPPINGS vmsize rss clean rss dirty file 4048 kb 948 kb 0 kb /terminus/crown/bin/ares 1476 kb 784 kb 0 kb /usr/lib64/libpython2.6.so.1.0 1576 kb 448 kb 0 kb /lib64/libc-2.12.so 324 kb 172 kb 0 kb /usr/lib64/libcurl.so.4.1.1 928 kb 160 kb 0 kb /usr/lib64/libstdc++.so.6.0.13 524 kb 104 kb 0 kb /lib64/libm-2.12.so 192 kb 76 kb 0 kb /usr/local/lib/libzmq.so.1.0.0 64 kb 32 kb 0 kb /usr/lib64/python2.6/lib-dynload/datetime.so 72 kb 28 kb 0 kb /terminus/crown/bin/ares 92 kb 28 kb 0 kb /lib64/libpthread-2.12.so 312 kb 20 kb 0 kb /usr/local/boost149/lib/libboost_python.so.1.49.0 12 kb 12 kb 0 kb /usr/lib64/python2.6/lib-dynload/_json.so 96 kb 12 kb 0 kb /usr/local/boost149/lib/libboost_thread.so.1.49.0 128 kb 12 kb 0 kb /lib64/ld-2.12.so 8 kb 8 kb 0 kb /usr/lib64/python2.6/lib-dynload/syslog.so 12 kb 8 kb 0 kb /usr/lib64/python2.6/lib-dynload/_randommodule.so 12 kb 8 kb 0 kb /usr/lib64/python2.6/lib-dynload/timemodule.so 8 kb 8 kb 0 kb /usr/lib64/python2.6/lib-dynload/_functoolsmodule.so 28 kb 4 kb 0 kb /lib64/librt-2.12.so 8 kb 4 kb 0 kb /usr/local/boost149/lib/libboost_system.so.1.49.0 4 kb 4 kb 0 kb [vdso]
从上面看到rss大小被分成了两个部分: private(私有)和shared(共享). private rss就是我们最关心的进程实际占用的内存数.