On Tuesday 31 January 2006 01:12, Andi Kleen wrote: > > > 1) Get rid of the debugging printk()'s > > I would suggest to replace them with statistic counters for netstat -s
I just tried adding a few linux-mib entries for this, and netstat started dying. There is a hard-coded limit of 1024 in netstat statistics.c:process_fd() for the length of the lines in the proc files. /proc/net/netstat is currently at 968; not much room left. Argh. What do we want to do about this? > From reading the patch it seems to have quite a lot of duplicated > code with the output path in tcp_mtu_probe. Can't you consolidate > this a bit more? Anything in particular? It doesn't seem to duplicate too much to me. The coalescing code is fairly particular to this procedure, and the actual output is only a couple lines. > Also if()s with 5 line conditions and magic numbers are always somewhat > suspicious. I tried to make this as clear as possible with a comment. The constant is straight out of the internet-draft, which has justification. It could be a #define, but I don't know if there's much point in that. > Are you sure the checksum handling is correct? AFAIK. It works for me w/ and w/o checksum offloading enabled. > Also we're trying to get away from the giant inlines in tcp.h, so > better move the new functions you add somewhere else out of line. Done. Will send out an updated patch as soon as I know what to do about the mib entries. Thanks, -John - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
