Re: NFS on 10G interface terribly slow
I wrote: Gerrit Kuhn wrote: On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet Yea, I don't know how to decode this either. I took a look at the driver and, if I read it correctly, most chips (including all the X540 ones) use IXGBE_82599_SCATTER. As such, you will be doing lots of m_defrag() calls, but since disabling TSO didn't help, that doesn't seem to be the bottleneck. rick I was actually interested in what chip Scott was using and getting wire speed. As noted in the other reply, since disabling TSO didn't help, you probably aren't affected by this issue. rick RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... RM Btw, it appears that the driver in head/current now sets RM if_hw_tsomaxsegcount, but the driver in stable/10 does not. This means RM that the 82599 chip will end up doing the m_defrag() calls for 10.x. So the next step could even be updating to -current... OTOH, I get the same (bad) resulsts, no matter if TSO is enabled or disabled on the interface. cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
Hi Rick On 06/29/2015 02:20 PM, Rick Macklem wrote: If the Solaris server is using ZFS, setting sync=disabled might help w.r.t. write performance. It is, however, somewhat dangerous w.r.t. loss of recently written data when the server crashes. (Server has told client data is safely on stable storage so client will not re-write the block(s) although data wasn't on stable storage and is lost.) (I'm not a ZFS guy, so I can't suggest more w.r.t. ZFS.) The system on the other side uses SAM/QFS, i.e. there is no such option for the file system per se (only the file system metadata is in a zvol thus not a full featured zfs). In parallel we are working also with Oracle to see where there may be a matching knob to turn as we see about the same performance issues from a Linux host (NFS client, Debian Jessie) with a Mellanox Technologies MT27500 Family [ConnectX-3] controller. Cheers Carsten -- Dr. Carsten Aulbert, Atlas cluster administration Max Planck Institute for Gravitational Physics (Albert Einstein Institute) Callinstraße 38, 30167 Hannover, Germany Tel: +49 511 762 17185, Fax: +49 511 762 17193 ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
On 6/29/2015 8:20 AM, Rick Macklem wrote: If the Solaris server is using ZFS, setting sync=disabled might help w.r.t. On my FreeBSD zfs server, this is a must for decent and consistent write throughput. Using FreeBSD as an iSCSI target and a Linux initiator, I can saturate a 1G nic no problem with sync disabled. Its barely usable with the default sync standard as its so bursty ---Mike -- --- Mike Tancsa, tel +1 519 651 3400 Sentex Communications, m...@sentex.net Providing Internet services since 1994 www.sentex.net Cambridge, Ontario Canada http://www.tancsa.com/ ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
On Mon, Jun 29, 2015 at 9:19 AM, Gerrit Kühn gerrit.ku...@aei.mpg.de wrote: On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... extract first 4 numbers of chip, then try a grep: grep 1528 /usr/src/sys/dev/ixgbe/* /usr/src/sys/dev/ixgbe/ixgbe_type.h:#define IXGBE_DEV_ID_X540T 0x1528 = Then your chipset is X540 ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
Gerrit Kuhn wrote: On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... RM Btw, it appears that the driver in head/current now sets RM if_hw_tsomaxsegcount, but the driver in stable/10 does not. This means RM that the 82599 chip will end up doing the m_defrag() calls for 10.x. So the next step could even be updating to -current... OTOH, I get the same (bad) resulsts, no matter if TSO is enabled or disabled on the interface. Since disabling TSO had no effect, I don't think updating would matter. If you can test against a different NFS server, that might indicate whether or not the Solaris server is the bottleneck. If the Solaris server is using ZFS, setting sync=disabled might help w.r.t. write performance. It is, however, somewhat dangerous w.r.t. loss of recently written data when the server crashes. (Server has told client data is safely on stable storage so client will not re-write the block(s) although data wasn't on stable storage and is lost.) (I'm not a ZFS guy, so I can't suggest more w.r.t. ZFS.) rick cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
[Bug 200323] BPF userland misuse can crash the system
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=200323 Ermal Luçi e...@pfsense.org changed: What|Removed |Added Resolution|--- |FIXED Status|Open|Closed -- You are receiving this mail because: You are the assignee for the bug. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
Gerrit Kuhn wrote: On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet Yea, I don't know how to decode this either. I was actually interested in what chip Scott was using and getting wire speed. As noted in the other reply, since disabling TSO didn't help, you probably aren't affected by this issue. rick RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... RM Btw, it appears that the driver in head/current now sets RM if_hw_tsomaxsegcount, but the driver in stable/10 does not. This means RM that the 82599 chip will end up doing the m_defrag() calls for 10.x. So the next step could even be updating to -current... OTOH, I get the same (bad) resulsts, no matter if TSO is enabled or disabled on the interface. cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
netmap custom RSS and custom packet info
Working with netmap and modern hardware I am lacking some features: a) some spare space before packet (64/128/192/256 bytes) for application data. For example: application do some pre-analysig packet, filled structure in this space and routed packet (via NETMAP pipe) to other thread. Received thread got packet and linked inforamtion about this packet for processing w/o additional overhead. b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, PPPoE encapsulated packets placed in queue 0, different tunneling don't recognised and etc. May be NETMAP can be used custom RSS hashing from loadable kernel module, provideng by user? Function frm this module can be packet analysing, tunnel removing, custom RSS hashnig with direction-independly maner, filled some structure prepended to buffer (see above) and pass this information to application. This is possible? This is useful not only to me? ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 5:17 PM, Slawa Olhovchenkov s...@zxy.spb.ru wrote: Working with netmap and modern hardware I am lacking some features: a) some spare space before packet (64/128/192/256 bytes) for application data. For example: application do some pre-analysig packet, filled structure in this space and routed packet (via NETMAP pipe) to other thread. Received thread got packet and linked inforamtion about this packet for processing w/o additional overhead. spare space in front of the packet is something we have been considering for a different purpose, namely better support for encapsulation/decapsulation and things like vhost-net header. Note though that the annotation is transferred for free only in the case of pipes or ports sharing the same memory region; vale ports would have to explicitly copy the extra bytes which is (moderately) expensive. A quick and dirty way to support what you want is the following: - in the kernel code, modify NMB(), PNMB() and the offset between the netmap_ring and the first buffer to add the extra space you want in front of the packet. You can possibly make this offset a sysctl-controlled value - in netmap_vale.c, make a small change to the code that copies buffers so that it includes also the space before the actual packet. That should be all. b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, PPPoE encapsulated packets placed in queue 0, different tunneling don't recognised and etc. May be NETMAP can be used custom RSS hashing from loadable kernel module, provideng by user? Function frm this module can be packet analysing, tunnel removing, custom RSS hashnig with direction-independly maner, filled some structure prepended to buffer (see above) and pass this information to application. RSS is completely orthogonal to netmap and I strongly suggest to keep it this way, using either use the NIC-specific tools to control RSS or some generic mechanism (on linux there is ethtool, and we should implement something similar also on freebsd). cheers luigi This is possible? This is useful not only to me? ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org -- -+--- Prof. Luigi RIZZO, ri...@iet.unipi.it . Dip. di Ing. dell'Informazione http://www.iet.unipi.it/~luigi/. Universita` di Pisa TEL +39-050-2217533 . via Diotisalvi 2 Mobile +39-338-6809875 . 56122 PISA (Italy) -+--- ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
82599 in our case. One problem I do have is the stack likes to blow up on occasion with the right combo of high load and high throughput while TSO is enabled, possibly relating to the 10.x driver issue you've pointed out. But when it comes to the throughput they'll blast 10G with no problem. *[image: userimage]Scott Larson[image: los angeles] https://www.google.com/maps/place/4216+Glencoe+Ave,+Marina+Del+Rey,+CA+90292/@33.9892151,-118.4421334,17z/data=!3m1!4b1!4m2!3m1!1s0x80c2ba88ffae914d:0x14e1d00084d4d09cLead Systems Administrator[image: wdlogo] https://www.wiredrive.com/ [image: linkedin] https://www.linkedin.com/company/wiredrive [image: facebook] https://www.twitter.com/wiredrive [image: twitter] https://www.facebook.com/wiredrive [image: instagram] https://www.instagram.com/wiredriveT 310 823 8238 x1106 310%20823%208238%20x1106 | M 310 904 8818 310%20904%208818* On Mon, Jun 29, 2015 at 5:22 AM, Rick Macklem rmack...@uoguelph.ca wrote: Gerrit Kuhn wrote: On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet Yea, I don't know how to decode this either. I was actually interested in what chip Scott was using and getting wire speed. As noted in the other reply, since disabling TSO didn't help, you probably aren't affected by this issue. rick RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... RM Btw, it appears that the driver in head/current now sets RM if_hw_tsomaxsegcount, but the driver in stable/10 does not. This means RM that the 82599 chip will end up doing the m_defrag() calls for 10.x. So the next step could even be updating to -current... OTOH, I get the same (bad) resulsts, no matter if TSO is enabled or disabled on the interface. cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 06:05:41PM +0200, Luigi Rizzo wrote: On Mon, Jun 29, 2015 at 5:17 PM, Slawa Olhovchenkov s...@zxy.spb.ru wrote: Working with netmap and modern hardware I am lacking some features: a) some spare space before packet (64/128/192/256 bytes) for application data. For example: application do some pre-analysig packet, filled structure in this space and routed packet (via NETMAP pipe) to other thread. Received thread got packet and linked inforamtion about this packet for processing w/o additional overhead. spare space in front of the packet is something we have been considering for a different purpose, namely better support for encapsulation/decapsulation and things like vhost-net header. Adding more space (sysctl or ioctl controled may be satisfy both: 4-8-20 bytes for encapsulation and rest for application). Note though that the annotation is transferred for free only in the case of pipes or ports sharing the same memory region; vale ports would have to explicitly copy the extra bytes which is (moderately) expensive. I think this bytes don't be transfered throw VALE. This is only packet-processing information, like tags, opposite to VALE that is like packet transfered by wire. A quick and dirty way to support what you want is the following: - in the kernel code, modify NMB(), PNMB() and the offset between the netmap_ring and the first buffer to add the extra space you want in front of the packet. You can possibly make this offset a sysctl-controlled value - in netmap_vale.c, make a small change to the code that copies buffers so that it includes also the space before the actual packet. That should be all. Do you plan to do this? I am don't like have permanenty private branch/patchs. b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, PPPoE encapsulated packets placed in queue 0, different tunneling don't recognised and etc. May be NETMAP can be used custom RSS hashing from loadable kernel module, provideng by user? Function frm this module can be packet analysing, tunnel removing, custom RSS hashnig with direction-independly maner, filled some structure prepended to buffer (see above) and pass this information to application. RSS is completely orthogonal to netmap and I strongly suggest to keep it this way, using either use the NIC-specific tools to control RSS or some generic mechanism (on linux there is ethtool, and we should implement something similar also on freebsd). This is not true RSS. This is only trick for reassigning RX packets to different netmap rings. All hardware avalable RSS mechanism is fully inacceptable for this: - don't support different encapsulation (PPPoE, GRE, GTP and etc) - give different rings for packet 1.2.3.4-5.6.7.8 and 5.6.7.8-1.2.3.4 Producing unversal hashing/distributing mechanism is too complex. But using user-providing kernel module (syncing to application) may be acceptable? This is like ephemeral permanent NETMAP pipe between real hardware RX rings/driver and application visible rings. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On 06/29/2015 08:17, Slawa Olhovchenkov wrote: ... b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, ... This is default behavior because the default hash (Toeplitz) is not symmetrical. There are modern NICs that do support other, symmetrical hashes. Regards, Navdeep ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
Hi, Turns out there are a class of symmetric RSS Toeplitz keys. Use google to find the paper. :) -a On 29 June 2015 at 10:19, Slawa Olhovchenkov s...@zxy.spb.ru wrote: On Mon, Jun 29, 2015 at 10:07:15AM -0700, Navdeep Parhar wrote: On 06/29/2015 08:17, Slawa Olhovchenkov wrote: ... b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, ... This is default behavior because the default hash (Toeplitz) is not symmetrical. There are modern NICs that do support other, symmetrical hashes. Anyway this is still hardware-depended. I am don't see symmetrical hashes for 1G/10G Intel cards. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 10:29:14AM -0700, Adrian Chadd wrote: Hi, Turns out there are a class of symmetric RSS Toeplitz keys. Use google to find the paper. :) How this interopperate with PPPoE encapsulation? With GRE/GTP/MPLS encapsulation? On 29 June 2015 at 10:19, Slawa Olhovchenkov s...@zxy.spb.ru wrote: On Mon, Jun 29, 2015 at 10:07:15AM -0700, Navdeep Parhar wrote: On 06/29/2015 08:17, Slawa Olhovchenkov wrote: ... b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, ... This is default behavior because the default hash (Toeplitz) is not symmetrical. There are modern NICs that do support other, symmetrical hashes. Anyway this is still hardware-depended. I am don't see symmetrical hashes for 1G/10G Intel cards. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 10:07:15AM -0700, Navdeep Parhar wrote: On 06/29/2015 08:17, Slawa Olhovchenkov wrote: ... b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, ... This is default behavior because the default hash (Toeplitz) is not symmetrical. There are modern NICs that do support other, symmetrical hashes. Anyway this is still hardware-depended. I am don't see symmetrical hashes for 1G/10G Intel cards. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 6:22 PM, Slawa Olhovchenkov s...@zxy.spb.ru wrote: On Mon, Jun 29, 2015 at 06:05:41PM +0200, Luigi Rizzo wrote: On Mon, Jun 29, 2015 at 5:17 PM, Slawa Olhovchenkov s...@zxy.spb.ru wrote: Working with netmap and modern hardware I am lacking some features: a) some spare space before packet (64/128/192/256 bytes) for application data. For example: application do some pre-analysig packet, filled structure in this space and routed packet (via NETMAP pipe) to other thread. Received thread got packet and linked inforamtion about this packet for processing w/o additional overhead. spare space in front of the packet is something we have been considering for a different purpose, namely better support for encapsulation/decapsulation and things like vhost-net header. Adding more space (sysctl or ioctl controled may be satisfy both: 4-8-20 bytes for encapsulation and rest for application). Note though that the annotation is transferred for free only in the case of pipes or ports sharing the same memory region; vale ports would have to explicitly copy the extra bytes which is (moderately) expensive. I think this bytes don't be transfered throw VALE. This is only packet-processing information, like tags, opposite to VALE that is like packet transfered by wire. A quick and dirty way to support what you want is the following: - in the kernel code, modify NMB(), PNMB() and the offset between the netmap_ring and the first buffer to add the extra space you want in front of the packet. You can possibly make this offset a sysctl-controlled value - in netmap_vale.c, make a small change to the code that copies buffers so that it includes also the space before the actual packet. That should be all. Do you plan to do this? I am don't like have permanenty private branch/patchs. possibly in the long term yes, but before doing it i want to design it properly so that it does not look like a custom hack. b) custom RSS. Modern NIC have RSS poorly interoperable with packet analysing: packets from same flow, but different direction placed in different queue, PPPoE encapsulated packets placed in queue 0, different tunneling don't recognised and etc. May be NETMAP can be used custom RSS hashing from loadable kernel module, provideng by user? Function frm this module can be packet analysing, tunnel removing, custom RSS hashnig with direction-independly maner, filled some structure prepended to buffer (see above) and pass this information to application. RSS is completely orthogonal to netmap and I strongly suggest to keep it this way, using either use the NIC-specific tools to control RSS or some generic mechanism (on linux there is ethtool, and we should implement something similar also on freebsd). This is not true RSS. This is only trick for reassigning RX packets to different netmap rings. All hardware avalable RSS mechanism is fully inacceptable for this: - don't support different encapsulation (PPPoE, GRE, GTP and etc) - give different rings for packet 1.2.3.4-5.6.7.8 and 5.6.7.8-1.2.3.4 Producing unversal hashing/distributing mechanism is too complex. But using user-providing kernel module (syncing to application) may be acceptable? This is like ephemeral permanent NETMAP pipe between real hardware RX rings/driver and application visible rings. this particular function would also need to deal with notifications between the physical NIC and the exported netmap rings, and i would probably leave it to userspace. You should be able to do what you have in mind using the programmable forwarding function that already exists for VALE ports (at the cost of a memory copy, which could be avoided when/if we decide to support VALE ports that share the same memory region hence using zero copy. Don't hold your breath though. cheers luigi -- -+--- Prof. Luigi RIZZO, ri...@iet.unipi.it . Dip. di Ing. dell'Informazione http://www.iet.unipi.it/~luigi/. Universita` di Pisa TEL +39-050-2217533 . via Diotisalvi 2 Mobile +39-338-6809875 . 56122 PISA (Italy) -+--- ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
Hi, PPPoE will not be hashed according to RSS on the 1g/10g (igb, ixgbe) intel hardware. you're going to have to figure out some other method for traffic redistribution. If it's inside GRE, then it's IPv4/IPv6 and thus yes, you can do symmetric hashing. But if it's raw pppoe coming in, you're SoL. -a ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 10:41:39AM -0700, Adrian Chadd wrote: Hi, PPPoE will not be hashed according to RSS on the 1g/10g (igb, ixgbe) intel hardware. you're going to have to figure out some other method for traffic redistribution. I propose ephemeral but permanent NETMAP RX pipe with redistribution function from user-loadable kernel module. Hardware RxRing(i) = loadable hash = user-visible RxRing(j). As I perceive overhead is packet-parsing and hash computation + swaping of slot indexes (zero-copying between hardware ring and user-visible ring). In other (Tx) direction user-visible direct mapped to hardware ring. If it's inside GRE, then it's IPv4/IPv6 and thus yes, you can do symmetric hashing. But if it's raw pppoe coming in, you're SoL. This is will be only hashed tunnel IPs, yes? Not sessions inside tunnel. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: FreeBSD 10.1-REL - network unaccessible after high traffic
hi, I asked for the output of vmstat -z and vmstat -m in a loop. :) -a On 29 June 2015 at 02:02, Csaba Banhalmi bim...@field.hu wrote: Hi All, vmstat 5 output when system freezes: procs memory pagedisks faults cpu r b w avmfre flt re pi pofr sr ad0 ad1 in sy cs us sy id 0 0 0 8752M 126M 5663 0 0 0 4042 445 66 0 1219 7148 4870 3 2 95 0 0 0 8650M 145M 2167 0 0 0 3501 447 79 0 974 4042 3578 1 1 98 0 0 0 8374M 201M 3113 0 0 0 6790 441 5 0 1130 6670 3729 3 1 96 0 0 0 8252M 220M 2632 0 0 0 4014 435 4 0 726 11653 2401 2 1 97 0 0 0 8188M 224M 1625 0 0 0 2189 434 5 0 713 6714 2376 1 1 98 0 0 0 7992M 233M 1504 0 0 0 2254 433 2 0 867 2890 2868 1 1 98 4 0 0 8032M 216M 2145 0 0 0 1995 435 18 0 526 3769 2048 1 1 98 0 0 0 8180M 195M 1949 0 0 0 1741 435 50 0 593 3441 2363 1 1 98 0 0 0 8186M 178M 2859 0 0 0 2525 436 6 0 499 3313 1733 2 1 97 1 0 0 8410M 146M 2521 0 0 0 1764 440 11 0 736 67271 2121 4 2 94 0 0 0 8182M 205M 2910 0 0 0 6378 927 8 0 495 16043 1775 1 1 98 1 1 0 7944M 210M 3009 0 0 0 3696 438 8 0 522 4247 1963 2 1 97 0 0 0 8091M 169M 7529 0 0 0 3601 436 105 0 1359 75290 4400 9 3 88 0 0 0 8121M 141M 4607 0 0 0 3288 444 62 0 949 12169 3268 5 1 94 0 0 0 8044M 201M 1782 0 0 0 4954 1795 9 0 446 3025 1927 1 1 99 0 0 0 7916M 222M 1296 0 0 0 2671 438 5 0 525 2984 1920 1 1 98 1 0 0 7870M 230M 888 0 0 0 1677 432 8 0 473 6424 2126 1 1 99 0 0 0 7968M 228M 3375 0 0 0 2625 433 51 0 768 4100 2852 3 1 96 0 0 0 8238M 194M 7586 0 0 0 4758 436 88 0 1026 9631 3908 4 2 94 0 0 0 8293M 185M 3253 0 0 0 2362 437 52 0 747 4475 3105 2 1 97 I increased the vm.v_free_min, but did not help. It was a different froze, the system was unreacheable even through IPMI, needed a hard reset. Regards, Csaba 2015.06.12. 20:17 keltezéssel, Adrian Chadd írta: On 12 June 2015 at 10:57, Christopher Forgeron csforge...@gmail.com wrote: I agree it shouldn't run out of memory. Here's what mine does under network load, or rsync load: 2 0 9 1822M 1834M 0 0 0 014 8 0 0 22750 724 136119 0 23 77 0 0 9 1822M 1823M 0 0 0 0 0 8 0 0 44317 347 138151 0 16 84 0 0 9 1822M 1761M 0 0 0 017 8 0 0 23818 820 92198 0 12 88 0 0 9 1822M 1727M 0 0 0 014 8 0 0 40768 634 126688 0 17 83 0 0 9 1822M 8192B 0 8 0 015 3 3 0 9236 305 57149 0 33 67 That's with a 5 second vmstat output. After the 8KiB, the system is nearly completely brain-dead and needs a hard power-off. I've seen it go from 6 GiB free to 8KiB in 5 sec as well. Currently my large machines are set to 12 GiB free to keep them from crashing, from what I presume is just network load due to lots of iSCSI / NFS traffic on my 10GiB network. I haven't had time to type this up for the list yet, but I'm putting it here just to make sure people know it's real. Hi, Then something is leaking or holding onto memory when it shouldn't be. Try doing vmstat -z and vmstat -m in a one second loop, post the data just before it falls over. -adrian ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: Same NIC name to MAC mapping on FreeBSD
On my production systems, I've never seen it deviate without hardware changes. Are you seeing otherwise? On 6/29/2015 午後 04:23, Wei Hu wrote: Hi, On a FreeBSD system with multiple NICs, ie, multiple MAC addresses, is there a way to keep the same network interface name to MAC address mapping across reboot? It seems on Linux udev rule can help achieve this. Anything similar on FreeBSD? Thanks, Wei ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On Mon, Jun 29, 2015 at 10:29:14AM -0700, Adrian Chadd wrote: Hi, Turns out there are a class of symmetric RSS Toeplitz keys. Use google to find the paper. :) Do someone work on using different RSS keys and hash fields (selecting L2/L3/L4 or just L3 hash for example) in FreeBSD? ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
[Bug 159294] [em] em watchdog timeouts
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=159294 --- Comment #3 from Kurt Jaeger p...@freebsd.org --- I still have a box with those interfaces, but I have avoided using em4 and em5 on this box. It now says: dev.em.4.%desc: Intel(R) PRO/1000 Legacy Network Connection 1.0.6 Hmm, if I have to reproduce the problem, this will take some time. -- You are receiving this mail because: You are the assignee for the bug. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
[Bug 159294] [em] em watchdog timeouts
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=159294 --- Comment #4 from Sean Bruno sbr...@freebsd.org --- (In reply to Kurt Jaeger from comment #3) I'll try to do the same here with my ATOM test box. -- You are receiving this mail because: You are the assignee for the bug. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
[Bug 175734] no ethernet detected on system with EG20T PCH chipset ATOM E6xx series
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=175734 Sean Bruno sbr...@freebsd.org changed: What|Removed |Added Keywords|IntelNetworking | -- You are receiving this mail because: You are the assignee for the bug. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
[Bug 159294] [em] em watchdog timeouts
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=159294 --- Comment #2 from Sean Bruno sbr...@freebsd.org --- (In reply to pi from comment #0) There has been a lot of changes to em(4) handling in FreeBSD. I have not, however, touched lem(4) which is what is controlling the 82541EI. Does this still happen for you? I can take a look at the watchdog handling and see if it can be fixed with the same code that em(4) has used. -- You are receiving this mail because: You are the assignee for the bug. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
[Bug 193620] Problem with igb multiqueue together with pf
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=193620 Sean Bruno sbr...@freebsd.org changed: What|Removed |Added Status|New |In Progress --- Comment #1 from Sean Bruno sbr...@freebsd.org --- hw.igb.enable_msix=0 disables MSIX features which *does* disable multiqueue. You can however, set hw.igb.num_queues=1 to use MSIX and only 1 queue. Can you test this with 10.1 release and 10.2 release beta when available? -- You are receiving this mail because: You are the assignee for the bug. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
Scott Larson wrote: 82599 in our case. One problem I do have is the stack likes to blow up on occasion with the right combo of high load and high throughput while TSO is enabled, possibly relating to the 10.x driver issue you've pointed out. But when it comes to the throughput they'll blast 10G with no problem. Thanks for the info. So long as your mbuf cluster pool is large enough, I think the m_defrag() calls will just result in increased CPU overheads and probably don't introduce much delay. I have no idea why the stack would blow up sometimes. If you can catch the backtrace for one of these and post it, it might become obvious. (Or you could just try increasing KSTACK_PAGES in sys/amd64/include/param.h and see if the stack still blows up. Alternately, I think you can set KSTACK_PAGES in your kernel config file.) rick *[image: userimage]Scott Larson[image: los angeles] https://www.google.com/maps/place/4216+Glencoe+Ave,+Marina+Del+Rey,+CA+90292/@33.9892151,-118.4421334,17z/data=!3m1!4b1!4m2!3m1!1s0x80c2ba88ffae914d:0x14e1d00084d4d09cLead Systems Administrator[image: wdlogo] https://www.wiredrive.com/ [image: linkedin] https://www.linkedin.com/company/wiredrive [image: facebook] https://www.twitter.com/wiredrive [image: twitter] https://www.facebook.com/wiredrive [image: instagram] https://www.instagram.com/wiredriveT 310 823 8238 x1106 310%20823%208238%20x1106 | M 310 904 8818 310%20904%208818* On Mon, Jun 29, 2015 at 5:22 AM, Rick Macklem rmack...@uoguelph.ca wrote: Gerrit Kuhn wrote: On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet Yea, I don't know how to decode this either. I was actually interested in what chip Scott was using and getting wire speed. As noted in the other reply, since disabling TSO didn't help, you probably aren't affected by this issue. rick RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... RM Btw, it appears that the driver in head/current now sets RM if_hw_tsomaxsegcount, but the driver in stable/10 does not. This means RM that the 82599 chip will end up doing the m_defrag() calls for 10.x. So the next step could even be updating to -current... OTOH, I get the same (bad) resulsts, no matter if TSO is enabled or disabled on the interface. cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
On Fri, 26 Jun 2015 19:58:42 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM The default (auto tuned) value is reported by nfsstat -m. RM It can be set with a mount option (should be something in man RM mount_nfs). If you are doing a test with 1 megabyte writes, I'd set RM it to at least 1 megabyte. (Basically, writing will be slower for write RM (2) syscalls that are larger than wcommitsize. After mav@'s patch, the RM difference isn't nearly as noticable. His other commit makes the auto RM tuned value more reasonable). RM RM If you set it large enough with the wcommitsize=N mount option, you RM don't need the updates stable/10. Ok, I set it way over 1MB now: hellpool:/samqfs/K1/Gerrit on /net/hellpool nfsv3,tcp,resvport,hard,cto,lockd,rdirplus,sec=sys,acdirmin=3,acdirmax=60,acregmin=5,acregmax=60,nametimeo=60,negnametimeo=60,rsize=65536,wsize=65536,readdirsize=8192,readahead=1,wcommitsize=2048576,timeout=120,retrans=2 However, this still gives me the same bad write performance: root@crest: # dd if=/dev/zero of=/net/hellpool/Z bs=1024k count=1000 1000+0 records in 1000+0 records out 1048576000 bytes transferred in 22.939049 secs (45711398 bytes/sec) So I guess I can postpone the update for now, and look for some other reason for this instead. cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Same NIC name to MAC mapping on FreeBSD
Hi, On a FreeBSD system with multiple NICs, ie, multiple MAC addresses, is there a way to keep the same network interface name to MAC address mapping across reboot? It seems on Linux udev rule can help achieve this. Anything similar on FreeBSD? Thanks, Wei ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: FreeBSD 10.1-REL - network unaccessible after high traffic
Hi All, vmstat 5 output when system freezes: procs memory pagedisks faults cpu r b w avmfre flt re pi pofr sr ad0 ad1 in sy cs us sy id 0 0 0 8752M 126M 5663 0 0 0 4042 445 66 0 1219 7148 4870 3 2 95 0 0 0 8650M 145M 2167 0 0 0 3501 447 79 0 974 4042 3578 1 1 98 0 0 0 8374M 201M 3113 0 0 0 6790 441 5 0 1130 6670 3729 3 1 96 0 0 0 8252M 220M 2632 0 0 0 4014 435 4 0 726 11653 2401 2 1 97 0 0 0 8188M 224M 1625 0 0 0 2189 434 5 0 713 6714 2376 1 1 98 0 0 0 7992M 233M 1504 0 0 0 2254 433 2 0 867 2890 2868 1 1 98 4 0 0 8032M 216M 2145 0 0 0 1995 435 18 0 526 3769 2048 1 1 98 0 0 0 8180M 195M 1949 0 0 0 1741 435 50 0 593 3441 2363 1 1 98 0 0 0 8186M 178M 2859 0 0 0 2525 436 6 0 499 3313 1733 2 1 97 1 0 0 8410M 146M 2521 0 0 0 1764 440 11 0 736 67271 2121 4 2 94 0 0 0 8182M 205M 2910 0 0 0 6378 927 8 0 495 16043 1775 1 1 98 1 1 0 7944M 210M 3009 0 0 0 3696 438 8 0 522 4247 1963 2 1 97 0 0 0 8091M 169M 7529 0 0 0 3601 436 105 0 1359 75290 4400 9 3 88 0 0 0 8121M 141M 4607 0 0 0 3288 444 62 0 949 12169 3268 5 1 94 0 0 0 8044M 201M 1782 0 0 0 4954 1795 9 0 446 3025 1927 1 1 99 0 0 0 7916M 222M 1296 0 0 0 2671 438 5 0 525 2984 1920 1 1 98 1 0 0 7870M 230M 888 0 0 0 1677 432 8 0 473 6424 2126 1 1 99 0 0 0 7968M 228M 3375 0 0 0 2625 433 51 0 768 4100 2852 3 1 96 0 0 0 8238M 194M 7586 0 0 0 4758 436 88 0 1026 9631 3908 4 2 94 0 0 0 8293M 185M 3253 0 0 0 2362 437 52 0 747 4475 3105 2 1 97 I increased the vm.v_free_min, but did not help. It was a different froze, the system was unreacheable even through IPMI, needed a hard reset. Regards, Csaba 2015.06.12. 20:17 keltezéssel, Adrian Chadd írta: On 12 June 2015 at 10:57, Christopher Forgeron csforge...@gmail.com wrote: I agree it shouldn't run out of memory. Here's what mine does under network load, or rsync load: 2 0 9 1822M 1834M 0 0 0 014 8 0 0 22750 724 136119 0 23 77 0 0 9 1822M 1823M 0 0 0 0 0 8 0 0 44317 347 138151 0 16 84 0 0 9 1822M 1761M 0 0 0 017 8 0 0 23818 820 92198 0 12 88 0 0 9 1822M 1727M 0 0 0 014 8 0 0 40768 634 126688 0 17 83 0 0 9 1822M 8192B 0 8 0 015 3 3 0 9236 305 57149 0 33 67 That's with a 5 second vmstat output. After the 8KiB, the system is nearly completely brain-dead and needs a hard power-off. I've seen it go from 6 GiB free to 8KiB in 5 sec as well. Currently my large machines are set to 12 GiB free to keep them from crashing, from what I presume is just network load due to lots of iSCSI / NFS traffic on my 10GiB network. I haven't had time to type this up for the list yet, but I'm putting it here just to make sure people know it's real. Hi, Then something is leaking or holding onto memory when it shouldn't be. Try doing vmstat -z and vmstat -m in a one second loop, post the data just before it falls over. -adrian ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: NFS on 10G interface terribly slow
On Fri, 26 Jun 2015 20:42:08 -0400 (EDT) Rick Macklem rmack...@uoguelph.ca wrote about Re: NFS on 10G interface terribly slow: RM Btw, can you tell us what Intel chip(s) you're using? I have ix0@pci0:5:0:0: class=0x02 card=0x00028086 chip=0x15288086 rev=0x01 hdr=0x00 vendor = 'Intel Corporation' device = 'Ethernet Controller 10-Gigabit X540-AT2' class = network subclass = ethernet RM For example, from the ix driver: RM #define IXGBE_82598_SCATTER 100 RM #define IXGBE_82599_SCATTER 32 Hm, I cannot find out into which chipset number this translates for my device... RM Btw, it appears that the driver in head/current now sets RM if_hw_tsomaxsegcount, but the driver in stable/10 does not. This means RM that the 82599 chip will end up doing the m_defrag() calls for 10.x. So the next step could even be updating to -current... OTOH, I get the same (bad) resulsts, no matter if TSO is enabled or disabled on the interface. cu Gerrit ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: patm device on FreeBSD 9.2
Yes, FreeBSD detects my card. When I ping the FreeBSD side from Linux side, netstat -s -p ip shows that arrived packets are incorrect version number. When both sides are FreeBSD packets are not received by the other side, I mean even netstat does not show the received packets! ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
Re: netmap custom RSS and custom packet info
On 29 June 2015 at 12:11, Slawa Olhovchenkov s...@zxy.spb.ru wrote: On Mon, Jun 29, 2015 at 10:29:14AM -0700, Adrian Chadd wrote: Hi, Turns out there are a class of symmetric RSS Toeplitz keys. Use google to find the paper. :) Do someone work on using different RSS keys and hash fields (selecting L2/L3/L4 or just L3 hash for example) in FreeBSD? The -HEAD RSS stuff has a global set of (compiled in) config options that say whether L2/L3 RSS hashing is enabled. It's more complicated than that, as the alignment of RSS NIC config, expected RSS hash info and the tcp/udp pcb table hashing has to align, or weird stuff happens. It still needs a bunch more work. Unfortunately it's not in my would be useful for work right at the moment, so other things are taking priority. -adrian ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org
RE: Same NIC name to MAC mapping on FreeBSD
-Original Message- From: owner-freebsd-...@freebsd.org [mailto:owner-freebsd- n...@freebsd.org] On Behalf Of Paul S. Sent: Monday, June 29, 2015 7:53 PM To: freebsd-net@freebsd.org Subject: Re: Same NIC name to MAC mapping on FreeBSD On my production systems, I've never seen it deviate without hardware changes. Are you seeing otherwise? In Hyper-V, if say three NICs were assigned to the VM, I got following mapping Initially: Hn0 - MAC 0 Hn1 - MAC 1 Hn2 - MAC2 Then if I remove the NIC with MAC 1 and reboot, I want the other two interfaces to keep the same Names instead of reassigning hn1 to MAC2. This is a requirement from virtual appliance Vendor to retain such mappings. I am wondering if there is any way to do this without Asking customer or manually editing any config files. Thanks, Wei ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to freebsd-net-unsubscr...@freebsd.org