HIgh Memory Footprint of Coovachilli

[drkhosla]

I compile latest version of Coovachilli (29th July, 2015) in openwrt (r46569) successfully. I am able run chilli in device having RAM 64 Mb but unable to run in device with 32 Mb RAM. I used to run two instances in a device having only 32 Mb RAM older version.

When chilli is not running, free command show

root@test:~# free
             total         used         free       shared      buffers
Mem:         61300        21676        39624          172            0
-/+ buffers:              21676        39624
Swap:            0            0            0
root@test:~#

After running first instance of chilli

root@test:~# chilli
coova-chilli[7553]: (Re)processing options [/var/run/chilli.7553.cfg.bin]
coova-chilli[7554]: running chilli_opt on /var/run/chilli.7553.cfg.bin
coova-chilli[7553]: PID 7553 rereading binary file /var/run/chilli.7553.cfg.bin
coova-chilli[7553]: PID 7553 reloaded binary options file
root@test:~# free
             total         used         free       shared      buffers
Mem:         61300        39596        21704          184            0
-/+ buffers:              39596        21704
Swap:            0            0            0
root@test:~# 

After running second instance of chilli what I get

root@test:~# chilli -c /etc/chilli/main2.conf
coova-chilli[9144]: (Re)processing options [/var/run/chilli.9144.cfg.bin]
coova-chilli[9145]: running chilli_opt on /var/run/chilli.9144.cfg.bin
coova-chilli[9144]: PID 9144 rereading binary file /var/run/chilli.9144.cfg.bin
coova-chilli[9144]: PID 9144 reloaded binary options file
root@test:~# free
             total         used         free       shared      buffers
Mem:         61300        56920         4380          200            0
-/+ buffers:              56920         4380
Swap:            0            0            0
root@test:~# 

Every instance eat almost 17 Mb of RAM which is very high memory footprint for any package in openwrt. It is almost impossible to run chilli in device with 32 or 16 Mb RAM. Is it a memory leak

[wlanmac]

What were your compile time options?

[gareth41]

I can confirm I'm having the same memory problem as well when compiling from the latest commit, however compiling from a commit going back about a month there's no problems, I'm also compiling with chilliproxy using the makefile from openwrt packages git repo and compiling for ar17xx.

On 14/08/2015, at 08:31, David Bird [email protected] wrote:

What were your compile time options?

—
Reply to this email directly or view it on GitHub.

[drkhosla]

chilli --help show
....
Compiled with ENABLE_CHILLIQUERY ENABLE_COA ENABLE_RADPROXY ENABLE_DHCPRADIUS ENABLE_IEEE8021Q ENABLE_LEAKYBUCKET ENABLE_UAMANYIP ENABLE_UAMUIPORT

My target is ar71xx.

[sevan]

Can you update from git, try again & report back if issue still persists

[drkhosla]

I update current version eb97f84 and try again but the issue still persists.

[dusterio]

@drkhosla @gareth41 did you find a way to optimise memory use? I turned off a lot of stuff during the compilation, but it has no effect on memory footprint - still 17 Mb per instance

[dusterio]

I installed old OpenWrt and coova-chilli (1.3.0) there only uses 700kb of RAM, same config.

coova-chilli 1.3.1 (latest version that's shipped with OpenWrt or LEDE) and 1.4.0 (git) uses 17mb from the get-go

[gareth41]

I've moved away from coovachilli and went with wifidog for my project, it's memory footprint is far less and allows high throughout as it uses layer3. I ended up integrating a DHCP server into it along with Mac auth - the same as coovachilli, in all I've changed a lot of the original code and fixed a major memory leak which happened on every web server request when new pthreads are created. There's a few companies here in NZ using wifidog and if you send hundreds of http requests to their hotspots they stop responding after some time.

…

On 3/11/2017, at 13:38, Denis Mysenko ***@***.***> wrote: @drkhosla @gareth41 did you find a way to optimise memory use? I turned off a lot of stuff during the compilation, but it has no effect on memory footprint - still 17 Mb per instance — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.