problem with bytes counter in ifconfig

phoenix_me

The best way to see my problem is to look at below example.
Especially look at RX bytes on eth0.

mjrosenb · Posted: Mon Mar 05, 2007 11:24 am Post subject:

it's a problem with x86. All of the older x86 hardware uses 32 bit integers for everything, which gives a maximum value of 2^32-1 = 4,294,967,295 = 4GB
If I'm not mistaken, it's something that is defined within the kernel, so the two best options are to re-write the proper kernel modules, or write a script that will monitor eth0 and find out when it's rolled over.
_________________
I'll be mjrosenb on #${COMPUTER_RELATED_SUBJECT}
1x i386 laptop w/ Gentoo
4x i386 desktop w/ Gentoo
2x dual proc i386 w/dragonfly bsd
1x x86-64 desktop w/ Gentoo
1x i386 desktop w/ FreeBSD
1x alpha workstation w/ Gentoo
looking for more

phoenix_me · Posted: Mon Mar 05, 2007 1:47 pm Post subject:

mjrosenb · Posted: Wed Mar 07, 2007 10:48 am Post subject:

Gcc supports long long, but the support is entirely in software. I would call the long long a poor hack, and something that's not exactly suitable to be in the kernel (not that the kernel is by any means good). Back to the other point that I was attempting to make. The limitation is in the kernel, not in ifconfig
at least in the e100 driver (/usr/src/linux/drivers/net/e100.c) they have the struct that's exported to /sys/class/net/eth1/statistics.

struct stats {
u32 tx_good_frames, tx_max_collisions, tx_late_collisions,
tx_underruns, tx_lost_crs, tx_deferred, tx_single_collisions,
tx_multiple_collisions, tx_total_collisions;
u32 rx_good_frames, rx_crc_errors, rx_alignment_errors,
rx_resource_errors, rx_overrun_errors, rx_cdt_errors,
rx_short_frame_errors;
u32 fc_xmt_pause, fc_rcv_pause, fc_rcv_unsupported;
u16 xmt_tco_frames, rcv_tco_frames;
u32 complete;
};
You can try changing the corresponding line in your driver, but since all of the accesor functions deal with ints, you'll most likely run into the same problem.
_________________
I'll be mjrosenb on #${COMPUTER_RELATED_SUBJECT}
1x i386 laptop w/ Gentoo
4x i386 desktop w/ Gentoo
2x dual proc i386 w/dragonfly bsd
1x x86-64 desktop w/ Gentoo
1x i386 desktop w/ FreeBSD
1x alpha workstation w/ Gentoo
looking for more

dj_farid · l33t Joined: 14 Jun 2004 Posts: 613

I found this thread while wondering the same thing. ( knew that I had downloaded more than 4 gigs, but the counter showed something else).

Is there a way to get better statistics that shows the real world numbers?
There is probably a script that gets the numbers from /proc/net/dev and keeps track of the resets every 4 GB...