Date: Tue, 30 Sep 2014 22:48:03 +0400 From: "Alexander V. Chernikov" <melifaro@FreeBSD.org> To: Rumen Telbizov <telbizov@gmail.com> Cc: "freebsd-stable@freebsd.org" <freebsd-stable@freebsd.org> Subject: Re: 10.1-BETA2 possible kernel memory leak in routing table Message-ID: <542AFAE3.9030705@FreeBSD.org> In-Reply-To: <CAENR%2B_XX4jnD6SBi8S1dGfWM68tmcm0aE2iMVA3LDR3R8ygQYw@mail.gmail.com> References: <CAENR%2B_UVLDDrsef2W4CXCFX65EYaxeKN4MNWbgoyaZ5qDGe1Pg@mail.gmail.com> <542AAA3C.1080803@ipfw.ru> <CAENR%2B_X5KTdeb00f9NShN1YK%2BT2aY1vG5YcTCgu4aXZO=%2Bpa=g@mail.gmail.com> <542AE376.6000003@FreeBSD.org> <CAENR%2B_XX4jnD6SBi8S1dGfWM68tmcm0aE2iMVA3LDR3R8ygQYw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 30.09.2014 21:56, Rumen Telbizov wrote:
> Hi Alexander,
>
> Sorry. That meant to be "route -Arn".
> It would be even better if you show something like
> "route -Arn ; sleep 10; route -Arn"
>
>
> OK I would assume you really meant netstat ;) I apologize for not
> pasting the entire routing table (security considerations). If you're
> only after the size of it - like I said it's pretty static and doesn't
> change very often. Is there anything specific that you're after?:
>
> netstat -rna | wc -l ; sleep 10; netstat -rna | wc -l
> 145
> 145
Well, I was going to run vimdiff on old / new output but it looks like
it is not necessary.
Can you do some dtrace profiling?
e.g. do
"
kldload dtraceall
cat > xx.d <<EOT
#!/usr/sbin/dtrace -s
fbt:kernel:rt_msg2:entry
{
stack();
}
fbt:kernel:rn_addroute:entry
{
stack();
}
EOT
dtrace -s xx.d
"
According to your stats, you're getting 52 requests per second, so It
won't take too long.
(and don't run utilities like route / netstat / ifconfig - they will
pollute the output).
Btw, you're not running ipfw with tables?
>
> If you do need something specific from it I'll spend some time to
> anonymize it later.
>
> Ok. Are there any (relevant) messages floating inside route socket?
> e.g does "route -n monitor" shows a lot of output?
>
>
> time route -n monitor
>
> real 13m15.960s
> user 0m0.000s
> sys 0m0.002s
>
> No output. No changes whatsoever! dmesg doesn't report anything either.
>
> Very strange. Do you have "normal" ipv4/ipv6 traffic? (e.g. not
> tunneled, not ipsec, not all-fragmented, etc?)
> ixgbe marks all packets by received queue id number, and queue 0
> is always selected for "corner case" traffic.
> I see more or less equal traffic distribution in you original
> post, so this looks very strange for me.
>
>
>
> Most of it should be pretty standard. Very few fragments and some
> ipsec (not terminated on the firewall). Again without this option -
> nothing goes over ix1 for some reason.
>
> Can you show "sysctl dev.ix" output? (and any ixgbe tunables set
> if any)?
>
>
> I do have 3 settings that I changed via loader.conf as reported in the
> original email. Please find the full output of loader.conf,
> sysctl.conf as well assysctl hw.ix and sysctl dev.ix at
> http://pastebin.com/M5Stnfvt
>
>
> Thank you for your help.
>
> Regards,
> --
> Rumen Telbizov
> Unix Systems Administrator <http://telbizov.com>;
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?542AFAE3.9030705>