Oh the s10 zones were upstream gate, I though that got removed there ages ago with the sparc paltform... guess not

hrm... sata.c is approaching sd.c levels of ridiculous size... I wonder how objectionable it'd be to split the scsa bits and the sxlt (translation) bits to keep the sizes from getting too ridiculous

(as part of support for more translation of stuff.. we don't implement the entire scsi/sata translation spec)

github.com/illumos/illumos-gate/blo…ommon/io/sata/impl/sata.c#L852-L857 -- is it me, or wouldn't it just be simpler to snprintf(taskq_name, sizeof (taskq_name), "%s-%d", ...) ?

21k lines is really too many lol

jbk: yes that taskq_name business is obscene

i mean i think sd.c probably has the record

(32k)

Hmmm... lemme look...

i got pulled away on a different issue, but I was going to add some additional log page support to our satl is why I was asking

Rolling back to the first commit of illumos-gate, ip.c was 25k lines. It's now 15k post-datapath-refactoring.

wc -l on the on10 usr/src/uts/common/inet/ip/ip.c gives 31387

that's after all the update releases though, so may have some backports that happened after the nevada gate was split off

Thank you alanc; I thought it was even bigger before.

if you do a ping w/ a non-default (larger) data size, is there a limit on the size of pings we'll reply to?

just to test some stuff, i'm doing a ping IP 8000 (mtu is set to 9000)... snoop shows it received, but dtrace suggests the driver is not getting a reply packet to send

if i lower it to 4000, then i do

jbk: ---> If MTU == 9000 and you send an 8000 byte ping, you should get back a 8000 byte ping unless you hit a router that has MTU == 1500. pathMTU should kick in with subsequent pings.

Is this an on-link peer?

If so I wonder if on-link peer has MTU < 5000 ?

(Or some other mismatch?)

ANd there should be no limit.

Yeah, I just sent 8k pings over a 9kMTU local link.

(and got them back).

For 9000 MTU, use "8972" as the argument for illumos ping(1) for size:

`ping -svn DEST 8972 <count>`

ahh.. i think I've found the problem (kinda)..

You should set -D if you do this.

To ensure there's no fragmenting.