dovecot - Jun 2014 - dovecot/lmtp munmap()-ing a lot

I observed several long running dovecot/lmtp processes hogging the
CPU. I then strace'd them (strace -c -p 6375) and found them
munmap()-ing a lot:

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
 97.18   19.592537        1878     10430           munmap
  2.28    0.458984          36     12696           epoll_ctl
  0.26    0.052926          10      5288           fdatasync
  0.21    0.042472           3     13679           epoll_wait
... snip ...  

Why would that happen? (dovecot 2.1.17)
    
-- 
[*] sys4 AG

http://sys4.de, +49 (89) 30 90 46 64
Franziskanerstra?e 15, 81669 M?nchen

Sitz der Gesellschaft: M?nchen, Amtsgericht M?nchen: HRB 199263
Vorstand: Patrick Ben Koetter, Marc Schiffbauer
Aufsichtsratsvorsitzender: Florian Kirstein

On 8.6.2014, at 11.59, Ralf Hildebrandt <r at sys4.de> wrote:
> I observed several long running dovecot/lmtp processes hogging the
> CPU. I then strace'd them (strace -c -p 6375) and found them
> munmap()-ing a lot:
> 
> % time     seconds  usecs/call     calls    errors syscall
> ------ ----------- ----------- --------- --------- ----------------
> 97.18   19.592537        1878     10430           munmap
>  2.28    0.458984          36     12696           epoll_ctl
>  0.26    0.052926          10      5288           fdatasync
>  0.21    0.042472           3     13679           epoll_wait
> ... snip ...  
> 
> Why would that happen? (dovecot 2.1.17)
Difficult to say. It could be munmap()ing memory allocations or it could be
munmap()ing Dovecot index files. Weren't there equivalent number of mmap()
calls?..

BTW. In v2.2 the index file handling is faster for LDA/LMTP because it
doesn't even try to mmap() the full indexes into memory.