On 2012-05-15 1:55 PM, Daniel Golle wrote: > On 14/05/12 02:35, Daniel Golle wrote: >> On Sat, May 05, 2012 at 08:23:06PM +0200, Felix Fietkau wrote: >>> Any package that adds a protocol handler script now needs to have netifd >>> support (netifd will be enabled by default soon, once the remaining >>> scripts have been ported). >>> Today I finished porting over the existing pptp script. The netifd >>> version is quite simple, you can use it as template. >> I used the netifd'd pptp.sh script and integrated it into ppp.sh. >> To remain compatible, I also added support for ppp-mod-pptp to files.old. >> Besides this, everything as it was. >> I'll test it extensively on ar71xx and kirkwood in the next days. > With some small corrections in options.pptp (which should probably be retired > in > favour of having stuff configurable through UCI anyway) I got stable and > well-performing results when using the pptp plugin and kernel-mode pptp > instead > of user-space pptpd. > I even overdid it and chained several L2TPv2 and PPTP connections within each > other until the resulting MTU of the inner-most tunnel ended up ridiculously > low > -- it's all working great for more than 18 hours now on a small Atheros > AR7240. > > Considerung the footprint, a single quite busy PPtP session looks like this on > that box: > PID PPID USER STAT VSZ %VSZ %CPU COMMAND > 12267 1 root S 1644 6% 0% /usr/sbin/pppd plugin pptp.so > pptp_se > 12270 1 root S 1644 6% 0% /usr/sbin/pppd plugin pptp.so > pptp_se > > uptime: 11:53:50 up 18:29, load average: 0.21, 0.24, 0.19 > loadavg: 0.29 0.26 0.20 1/48 12601 > > So obviously, obviously the call-manager is still fork'ed into a seperate > process which is created and being controllled by the plugin. > > Besides that I tested it on rt305x(mipsel), kirkwood(armel) and x86. > > After integrating #31724 this also starts up correctly and one-by-one after a > reboot, not needing manual ifup/ifdown calls any more if initially > unresolvable > hostnames instead of plain IP addresses are used. > > Please advise me what is still missing or needs to be changed. > How is everybody's schedule regarding the integration and changing of > procotols > to netifd. The question is basically, if I should work towards getting this > integrated asap or if we want to get stuff entirely polished and working in > netifd with the existing pptp implementation and then eventually switch over > to > kernel-mode pptp. Based on your test results, it sounds like these changes (including switching to kernel-mode pptp) are ready for merging now. If you want to work on getting rid of options.pptp, that's fine with me too. Feel free to send me your updated patches and I'll review/merge them.
- Felix _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel
