В Чтв, 25/02/2010 в 20:40 +0200, Покотиленко Костик пишет:
> В Чтв, 25/02/2010 в 20:37 +0200, Покотиленко Костик пишет:
> > В Чтв, 25/02/2010 в 16:19 +0200, Покотиленко Костик пишет:
> > > Hi,
> > > 
> > > We've switched back from 82576 to 82574L+82578DM. 82576 is still there
> > > for a while before I put it into another server for testing.
> > 
> > Well, I don't believe we have problems with 82576 as well as 82574L and
> > 82578DM. Configuration with 82574L+82578DM has rebooted with 5 hour
> > uptime.
> > 
> > Also there was a mention in bugreport that "pcie_aspm=off" helped some
> > people with 82574L issues, but not for me.
> > 
> > As to me this is defenitelly motherboard/irq/dma/pci-e/driver problem.
> > I'm lost on how to debug further... 
> > 
> > Info on the case:
> 

Also, don't remember if I posted my setup. Just in case the problem is
setup related:

               br0
                |
   +--------+---+--+-------+
   |        |      |       |
eth1.11 eth1.12 eth1.13 eth1.19   eth0.23 eth0.27 eth0.31
   |        |      |       |         |       |       |
   +--------+---+--+-------+         +----+--+-------+
                |                         |
              eth1                      eth0

- everything is connected to one switch: D-link DES-1226G with vlans
(port number = vlan-10)
- port 25 and 26 are gigabit, tagged, rest 1-24 are 100 Mbit/s untagged
- eth1 is connected port 25
- eth0 is connected port 26
- br0 is client side
- eth0.23 is server's DMZ
- eth0.31 is Internet
- most of the traffic is between eth1 and eth0, more preciselly
eth0->eth1

- there is pfifo_fast on every eth+
- there is PREROUTING imq hook to imq0 with HTB and ESFQ at leafs with
  fixed number of "-j MARK" and "tc filter fw" filters and
  many dynamic "tc filter u32 police rate" filters
- there is POSTROUTING imq hook to imq1 with HTB and ESFQ at leafs with
  fixed number of "-j CLASSIFY"

Here is more visual scheme:

http://meteor.dp.ua/Lan-r_VLANs-24.12.2009-03_stripped_indexed.png

Also, I'm already have problems with this setup: it is impossible to
ifdown vlan interfaces because they are in use, usage countdown can last
many munites. This also affects reboots.

I didn't have such problems in same setup but without vlans, just bridge
+imq. Vlans seem to add some sort of conflict.

I've found "right" down sequence and made a script "reboot.sh":

/etc/init.d/imq stop
rmmod imq
ifconfig br0 down
brctl delbr br0
ifconfig eth1.11 down
ifconfig eth1.12 down
ifconfig eth1.13 down
ifconfig eth1.19 down
ifconfig eth0.23 down
ifconfig eth0.27 down
ifconfig eth0.31 down
vconfig rem eth1.11
vconfig rem eth1.12
vconfig rem eth1.13
vconfig rem eth1.19
vconfig rem eth0.23
vconfig rem eth0.27
vconfig rem eth0.31
rmmod 8021q
ifconfig eth0 down
ifconfig eth1 down
rmmod igb
rmmod e1000e
reboot

-- 
Покотиленко Костик <[email protected]>


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel&#174; Ethernet, visit 
http://communities.intel.com/community/wired

Reply via email to