Hi Daniel, On Wed, Oct 13, 2010 at 10:00:04AM +0200, Daniel Wagner wrote: > Good Morning Samuel, > > On Wed, Oct 13, 2010 at 01:20:52AM +0200, Samuel Ortiz wrote: > > Hi Daniel, > > > > On Tue, Oct 12, 2010 at 03:36:21PM +0200, Daniel Wagner wrote: > > > Hi, > > > > > > with these patches the number of open files for stats is reduced > > > by half. Hopefully with these patches we don't get 'Too many open' > > > anymore. > > Thanks, both patches applied. > > Cool. BTW, I'm seeing another possible improvement to reduce the > number of open files. I have a connection over ethernet and a > connection through a MBM and seeing this: > > $ sudo lsof +p 7061 | grep stats > connmand 7061 root mem REG 8,18 4096 927984 > /var/lib/connman/stats/ethernet_028037ec0200_cable.data > connmand 7061 root mem REG 8,18 4096 927988 > /var/lib/connman/stats/cellular_262029903630655_context1.data > connmand 7061 root mem REG 8,18 4096 916554 > /var/lib/connman/stats/ethernet_002481c57c73_cable.data > connmand 7061 root 11u REG 8,18 4096 927988 > /var/lib/connman/stats/cellular_262029903630655_context1.data > connmand 7061 root 12u REG 8,18 4096 916554 > /var/lib/connman/stats/ethernet_002481c57c73_cable.data > connmand 7061 root 18u REG 8,18 4096 927984 > /var/lib/connman/stats/ethernet_028037ec0200_cable.data > Which connman version are you using ? With connman trunk, I'm no longer seeing the fake MBM ethernet device, only the cellular one.
> and test-counter shows only traffic for the build in ethernet device > but none for the MBM ethernet device. > > /profile/default/ethernet_002481c57c73_cable > Home > RX.Bytes = 187616800 (178.9 MiB) > RX.Packets = 294298 > Time = 10200 > /profile/default/cellular_262029903630655_context1 > Home > RX.Bytes = 24766 (24.2 KiB) > RX.Packets = 298 > TX.Bytes = 32776 (32.0 KiB) > TX.Packets = 321 > Time = 383 > > Maybe we have to decided which type of Service should register itself > at stats.c. Currently we do statistic for all. It might even not > possible to distinguish between the build in ethernet device and the > one from MBM without udev, right? We can, and that's what we currently do with the latest detect.c code. We're reading the device's uevent and that let us know the device's real type. Cheers, Samuel. > > cheers, > daniel -- Intel Open Source Technology Centre http://oss.intel.com/ _______________________________________________ connman mailing list [email protected] http://lists.connman.net/listinfo/connman
