Threaded 6.4 code compiled under 9.0 uses a lot more memory?..

[Konstantin Belousov]

On Wed, Oct 31, 2012 at 09:49:21AM +0000, Karl Pielorz wrote:
> 
> --On 30 October 2012 19:51 +0200 Konstantin Belousov <kostikbel at gmail.com> 
> wrote:
> 
> > I suggest to take a look at where the actual memory goes.
> >
> > Start with procstat -v.
> 
> Ok, running that for the milter PID I get seem to be able to see smallish 
> chunks used for things like 'libmilter.so', and 'libthr.so' / 'libc.so' - 
> e.g.
Since you neglected to provide the verbatim output of procstat, nothing
conclusive can be said. Obviously, you can make an investigation on your
own.

> 
> 2010           0x400000           0x413000 r-x   19    0   1   0 CN-- vn 
> /usr/local/sbin/milter
> 2010           0x613000           0x800000 rw-   15    0   1   0 ---- df
> 2010        0x800613000        0x80062b000 r-x   24    0  97   0 CN-- vn 
> /libexec/ld-elf.so.1
> 2010        0x80062b000        0x800634000 rw-    9    0   1   0 ---- df
> 2010        0x800634000        0x80065f000 rw-   33    0   1   0 ---- df
> 2010        0x80082b000        0x80082d000 rw-    2    0   1   0 ---- df
> 2010        0x80082d000        0x800839000 r-x   12   12   2   1 CN-- vn 
> /usr/lib/libmilter.so.5
> 2010        0x800839000        0x800a38000 ---    0    0   1   0 ---- df
> 2010        0x800a38000        0x800a39000 rw-    1    0   1   0 C--- vn 
> /usr/lib/libmilter.so.5
> 2010        0x800a39000        0x800a3c000 rw-    0    0   0   0 ---- --
> 
> Then there's a bunch of 'large' blocks e.g..
> 
>  PID              START                END PRT  RES PRES REF SHD  FL TP PATH
> 2010        0x801c00000        0x802800000 rw- 2869    0   4   0 ---- df
> 2010        0x802800000        0x803400000 rw- 1880    0   1   0 ---- df
> 2010        0x803400000        0x803800000 rw-  920    0   1   0 ---- df
> 2010        0x803800000        0x804000000 rw-  865    0   1   0 ---- df
> 2010        0x804000000        0x804400000 rw- 1024    0   4   0 ---- df
> 2010        0x804400000        0x804c00000 rw-  627    0   1   0 ---- df
> 2010        0x804c00000        0x805000000 rw- 1021    0   4   0 ---- df
Most likely, these are malloc arenas.

> 
> Then lots of 'little' blocks,
> 
> 2010     0x7ffff0161000     0x7ffff0181000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0362000     0x7ffff0382000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0563000     0x7ffff0583000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0764000     0x7ffff0784000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0965000     0x7ffff0985000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0b66000     0x7ffff0b86000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0d67000     0x7ffff0d87000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff0f68000     0x7ffff0f88000 rw-   16    0   1   0 ---D df
> 2010     0x7ffff1169000     0x7ffff1189000 rw-   16    0   1   0 ---D df
And those are thread stacks.

> 
> 
> The memory usage figures seem to have 'stabilized' now - SIZE/RES seem to 
> track around 9 times the size of the 6.4 system, for a comparative load.
> 
> We can probably live with this (as they have come back down overnight as 
> load fell off) - i.e. they don't appear to be continuously growing (the 
> binaries were heavily profiled under 6.x and found to have no internal 
> leaks).
> 
> -Karl
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-hackers/attachments/20121031/9cca67ec/attachment.sig>