0.95.4 Embedded - slow speeds problem



  • Hi guys,
    I didn't know where to post this and since I have some reasons to suspect it is hardware related…
    This is not the 1st time that I've noticed this and this has happened to me with m0n0wall as well (with 5.3 betas) but not with 1.2final.
    I'm running 0.95.4 on a Soekris net4521 which is actually very similar to a net4501 but only has 2 NICs and instead it has 2 PCMCIA.
    It runs @133Mhz and it has 64MB RAM and I'm booting from a 256MB CF Card.
    Since I've upgraded from m0n0 1.2 to pfS 0.95.4 my Internet connection speed degraded from steady 48-52KB/s to a poor 5-12KB/s.
    Since my cable connection is a 512/128Kb/s I should be able to go as high as 64KB/s but I was ok with 48-52KB/s.
    Now, 5-12KB/s is unacceptable :(.
    My config is a very light one with simple Inet NATing + Firewall.
    No traffic shapping or VPN stuff.
    Looking at the system log I found something funny when there's high activity (downloads):

    Dec 6 18:56:02 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 6 19:00:08 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 6 19:01:07 	last message repeated 21 times
    Dec 6 19:06:26 	last message repeated 120 times
    Dec 6 19:06:28 	kernel: sis1: Applying short cable fix (
    Dec 6 19:06:28 	kernel: stray irq7
    Dec 6 19:06:28 	kernel: reg=e8)
    Dec 6 19:06:30 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 6 19:06:53 	last message repeated 11 times
    Dec 6 19:08:58 	last message repeated 40 times
    Dec 6 19:09:00 	kernel: sis1: Applying short cable
    Dec 6 19:09:00 	kernel: stray irq7
    Dec 6 19:09:00 	kernel: fix (reg=e8)
    Dec 6 19:09:02 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 6 19:09:33 	last message repeated 18 times
    Dec 6 19:09:49 	last message repeated 9 times
    Dec 6 19:24:31 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 6 19:24:35 	kernel: sis1: Applying short cable fix (reg=e8)
    

    Why would it be applying so many times the same "short cable" fix?
    In fact the UTP cable that connects my cablemodem to my net4521 is a very short cable.
    Perhaps less than 0.5 meter but I've had other standard 2 meter cables connecting both devices in the past and still I would experience this "slow downloading" issue.
    Anyone else experiencing this on net4501s or net4521s?
    Any ideas to fix this?
    TIA
    Cheers



  • Can you try an even longer cable? Try 5 meter or even 10 meters and see if this makes a difference. Also check duplexsettings reported at status>interfaces and what the switch/hub reports (if it has an indicator for this). Maybe there is a mismatch somewhere.



  • I could try a longer cable of 5-10 meters but I won't be able to do that ASAP.
    Anyway the status interfaces reports "100baseTX <full-duplex>" for both WAN and LAN.
    Now that is odd because my cable modem is a Dlink DCM-200 with a 10Mbit LAN port on it's back which I also believe to be half-duplex…
    Could this be the catch?
    What can I do to change this settings?
    TIA
    Cheers</full-duplex>



  • Try the hidden options to force the desired mode manually. Sounds like this is an autonegotiation issue. Maybe it solves itself with a longer cable though.

    http://faq.pfsense.com/index.php?sid=12261&lang=en&action=artikel&cat=10&id=38&artlang=en&highlight=hidden



  • Hi there,
    I ran ifconfig-m and I got this:

    supported media:
    media autoselect
    media 100baseTX mediaopt full-duplex
    media 100baseTX
    media 10baseT/UTP mediaopt full-duplex
    media 10baseT/UTP
    

    Now, my config.xml looks like this:

     <wan><if>sis1</if>
    	 <mtu><ipaddr>dhcp</ipaddr>
    	 <subnet><gateway><blockpriv><disableftpproxy><dhcphostname><media><mediaopt><bandwidth>100</bandwidth>
    	<bandwidthtype>Mb</bandwidthtype></mediaopt></media></dhcphostname></disableftpproxy></blockpriv></gateway></subnet></mtu></wan> 
    

    Should I edit it to end up like this?

    ...
    	 <dhcphostname><media>10baseT/UTP</media>
    	 <mediaopt>...</mediaopt></dhcphostname>
    

    What about the bandwidth and bandwidthtype?
    Should I leave them like this?

    ...
    	<bandwidth>100</bandwidth>
    	<bandwidthtype>Mb</bandwidthtype>
    ...
    
    ```Or should I edit them to something like this?
    

    ...
    <bandwidth><bandwidthtype>...</bandwidthtype></bandwidth>

    Sorry for all the trouble but I really don't know how to edit the config.xml and this performance issues are really a pain in the a**.
    TIA
    Cheers :)


  • No, it would look something like this:

    <media>media 100baseTX</media>
    <mediaopt>full-duplex</mediaopt>



  • Gee, thanks Scott.
    What about those bandwidth and bandwidthtype fields?
    Do I leave them as they are?
    EDIT:
    Oops, almost did this blindly.
    You're telling me to use "media 100baseTX" but my cable modem comes with a 10Mbit LAN port.
    What do I do?
    Cheers



  • Leave bandwidth alone.

    Change the value to 10 if you need to.



  • Ok ;).



  • @rds_correia:

    Ok ;).

    Also, leave out "media" inside the xml string.



  • I did exactly as you told me to and my box still won't go faster than 18KB/s…
    Here's my log:

    Dec 7 01:19:27 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 7 01:19:58 	last message repeated 22 times
    Dec 7 01:20:21 	last message repeated 12 times
    Dec 7 01:20:24 	kernel: sis1: Applying
    Dec 7 01:20:24 	kernel: too many stray irq 7's: not logging anymore
    Dec 7 01:20:24 	kernel: short cable fix (reg=e8)
    Dec 7 01:20:26 	kernel: sis1: Applying short cable fix (reg=e8)
    

    Oh but my Status > Interfaces > WAN still show "Media 100baseTX <full-duplex>".
    Maybe I'm doing something wrong here.
    BTW I checked some screenshots I took from my last m0n0 config on this box and it also shows "Media 100baseTX <full-duplex>".
    Maybe this media/mediaopt ain't related to my performance issues…
    I don't know.
    Cheers</full-duplex></full-duplex>



  • I'm at a loss then.  The repeated short cable message makes me wonder.  It should not be repeating that much.



  • Yep, me too.
    This cable has been replaced more than a dozen times specifically because of logs like this one when I was using m0n0's 1.2betas based upon FreeBSD 5.3 (or was it 5.4?).
    But ever since Manuel went back to FreeBSD 4.11 on the latest betas and on the 1.2final the speed issues went away and I made a new CAT-5 cable with ~ 0.5 meter because the modem and the Soekris box are actually stacked on my table.
    Maybe the speed issues don't even have anything to do with these "short cable fixes".
    I promise that as soon as I feel better I'll go out and buy another cable just to make sure but I think this ain't related with long or small cables.
    Cheers



  • Yes, it's me again ;D.
    Take a look at this:
    But this goes on:

    Dec 7 13:08:42 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 7 13:08:44 	php: : phpDynDNS: (Unknown Response)
    Dec 7 13:08:48 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 7 13:09:17 	last message repeated 10 times
    Dec 7 13:09:19 	kernel: s
    Dec 7 13:09:19 	kernel: stray irq7
    Dec 7 13:09:19 	kernel: is1: Applying short cable fix (reg=e8)
    Dec 7 13:17:49 	dnsmasq[780]: reading /etc/resolv.conf
    Dec 7 13:17:49 	dnsmasq[780]: using nameserver 213.228.128.6#53
    Dec 7 13:17:49 	dnsmasq[780]: using nameserver 213.228.128.99#53
    Dec 7 13:17:55 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 7 13:18:18 	last message repeated 4 times
    Dec 7 13:20:26 	last message repeated 8 times
    Dec 7 13:25:01 	kernel: sis1: Applying short cable fix (reg=e8)
    Dec 7 13:29:18 	kernel: sis1: Applying short cab
    Dec 7 13:29:18 	kernel: too many stray irq 7's: not logging anymore
    Dec 7 13:29:18 	kernel: le fix (reg=e8)
    

    So, what's a stray IRQ7 and why do I have too many of those.
    Could this be the reason for so many short cable fixes and ultimately for slow netwk performance?

    EDITED:
    Just went back to m0n0wall 1.2final with the same config (but not the same config.xml file) and networking speed is back.
    Here's a m0n0's log:

    Dec 7 14:33:31 	/kernel: Copyright (c) 1992-2005 The FreeBSD Project.
    Dec 7 14:33:31 	/kernel: Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989, 1991, 1992, 1993, 1994
    Dec 7 14:33:31 	/kernel: The Regents of the University of California. All rights reserved.
    Dec 7 14:33:31 	/kernel: FreeBSD 4.11-RELEASE-p11 #0: Wed Sep 7 13:51:02 CEST 2005
    Dec 7 14:33:31 	/kernel: root@fb411.neon1.net:/usr/src/sys/compile/M0N0WALL_NET45XX
    Dec 7 14:33:31 	/kernel: Timecounter "i8254" frequency 1189158 Hz
    Dec 7 14:33:31 	/kernel: CPU: AMD Enhanced Am486DX4/Am5x86 Write-Back (486-class CPU)
    Dec 7 14:33:31 	/kernel: Origin = "AuthenticAMD" Id = 0x494 Stepping = 4
    Dec 7 14:33:31 	/kernel: Features=0x1 <fpu>Dec 7 14:33:31 	/kernel: real memory = 67108864 (65536K bytes)
    Dec 7 14:33:31 	/kernel: avail memory = 50819072 (49628K bytes)
    Dec 7 14:33:31 	/kernel: Preloaded elf kernel "kernel" at 0xc0e0f000.
    Dec 7 14:33:31 	/kernel: Preloaded mfs_root "/mfsroot" at 0xc0e0f0a8.
    Dec 7 14:33:31 	/kernel: md0: Preloaded image  11534336 bytes at 0xc030d3bc
    Dec 7 14:33:31 	/kernel: md1: Malloc disk
    Dec 7 14:33:31 	/kernel: Timecounter "ELAN" frequency 8333333 Hz
    Dec 7 14:33:31 	/kernel: npx0: <math processor=""> on motherboard</math>
    Dec 7 14:33:31 	/kernel: npx0: INT 16 interface
    Dec 7 14:33:31 	/kernel: pcib0: <amd elan="" sc520="" host="" to="" pci="" bridge=""> on motherboard
    Dec 7 14:33:31 	/kernel: pci0: <pci bus=""> on pcib0
    Dec 7 14:33:31 	/kernel: pcic0: <ti pci-1420="" pci-cardbus="" bridge=""> mem 0xa0000000-0xa0000fff irq 10 at device 17.0 on pci0
    Dec 7 14:33:31 	/kernel: pcic0: TI12XX PCI Config Reg: [pwr save][pci only]
    Dec 7 14:33:31 	/kernel: pccard0: <pc card="" 16-bit="" bus="" (classic)=""> on pcic0
    Dec 7 14:33:31 	/kernel: pcic1: <ti pci-1420="" pci-cardbus="" bridge=""> mem 0xa0001000-0xa0001fff irq 10 at device 17.1 on pci0
    Dec 7 14:33:31 	/kernel: pcic1: TI12XX PCI Config Reg: [pwr save][pci only]
    Dec 7 14:33:31 	/kernel: pccard1: <pc card="" 16-bit="" bus="" (classic)=""> on pcic1
    Dec 7 14:33:31 	/kernel: sis0: <natsemi 10="" dp83815="" 100basetx=""> port 0xe100-0xe1ff mem 0xa0002000-0xa0002fff irq 11 at device 18.0 on pci0
    Dec 7 14:33:31 	/kernel: sis0: Ethernet address: 00:00:24:c0:e7:e0
    Dec 7 14:33:31 	/kernel: miibus0: <mii bus=""> on sis0
    Dec 7 14:33:31 	/kernel: ukphy0: <generic ieee="" 802.3u="" media="" interface=""> on miibus0
    Dec 7 14:33:31 	/kernel: ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
    Dec 7 14:33:31 	/kernel: sis1: <natsemi 10="" dp83815="" 100basetx=""> port 0xe200-0xe2ff mem 0xa0003000-0xa0003fff irq 5 at device 19.0 on pci0
    Dec 7 14:33:31 	/kernel: sis1: Ethernet address: 00:00:24:c0:e7:e1
    Dec 7 14:33:31 	/kernel: miibus1: <mii bus=""> on sis1
    Dec 7 14:33:31 	/kernel: ukphy1: <generic ieee="" 802.3u="" media="" interface=""> on miibus1
    Dec 7 14:33:31 	/kernel: ukphy1: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
    Dec 7 14:33:31 	/kernel: isa0: <isa bus=""> on motherboard
    Dec 7 14:33:31 	/kernel: orm0: <option rom=""> at iomem 0xc8000-0xd0fff on isa0
    Dec 7 14:33:31 	/kernel: pmtimer0 on isa0
    Dec 7 14:33:31 	/kernel: ata0 at port 0x1f0-0x1f7,0x3f6 irq 14 on isa0
    Dec 7 14:33:31 	/kernel: ata1 at port 0x170-0x177,0x376 irq 15 on isa0
    Dec 7 14:33:31 	/kernel: sio0 at port 0x3f8-0x3ff irq 4 flags 0x30 on isa0
    Dec 7 14:33:31 	/kernel: sio0: type 16550A, console
    Dec 7 14:33:31 	/kernel: sio1 at port 0x2f8-0x2ff irq 3 on isa0
    Dec 7 14:33:31 	/kernel: sio1: type 16550A
    Dec 7 14:33:31 	/kernel: Elan-mmcr driver: MMCR at 0xc597e000
    Dec 7 14:33:31 	/kernel: BRIDGE 020214 loaded
    Dec 7 14:33:31 	/kernel: IPsec: Initialized Security Association Processing.
    Dec 7 14:33:31 	/kernel: IP Filter: v3.4.35 initialized. Default = block all, Logging = enabled
    Dec 7 14:33:31 	/kernel: ad0: 15MB <sandisk sdcfb-16=""> [490/2/32] at ata0-master BIOSPIO
    Dec 7 14:33:31 	/kernel: Mounting root from ufs:/dev/md0c
    Dec 7 14:33:31 	dnsmasq[95]: started, version 2.22 cachesize 150
    Dec 7 14:33:31 	dnsmasq[95]: setting --bind-interfaces option because of OS limitations
    Dec 7 14:33:31 	dnsmasq[95]: setting --bind-interfaces option because of OS limitations
    Dec 7 14:33:31 	dnsmasq[95]: read /etc/hosts - 4 addresses
    Dec 7 14:33:31 	dnsmasq[95]: failed to access /var/db/dhcpd.leases: No such file or directory
    Dec 7 14:33:31 	dnsmasq[95]: failed to access /var/db/dhcpd.leases: No such file or directory
    Dec 7 14:33:31 	dhclient: DHCPOFFER from 10.39.0.1
    Dec 7 14:33:31 	dnsmasq[95]: reading /etc/resolv.conf
    Dec 7 14:33:31 	dhclient: DHCPREQUEST on sis1 to 255.255.255.255 port 67
    Dec 7 14:33:31 	/usr/local/bin/ez-ipupdate[98]: ez-ipupdate Version 3.0.11b8, Copyright (C) 1998-2001 Angus Mackay.
    Dec 7 14:33:31 	/usr/local/bin/ez-ipupdate[98]: /usr/local/bin/ez-ipupdate started for interface sis1 host WWWWWWW.TTTTTT.net using server members.dyndns.org and service dyndns
    Dec 7 14:33:31 	/usr/local/bin/ez-ipupdate[98]: got last update xx.xx.xx.xx on 2005/12/06 16:39 from cache file
    Dec 7 14:33:31 	/usr/local/bin/ez-ipupdate[98]: failure to update sis1->0.0.0.0 (WWWWWWW.TTTTTT.net)
    Dec 7 14:33:31 	dhclient: DHCPACK from 10.39.0.1
    Dec 7 14:33:31 	dhclient: New Network Number: xx.xx.xx.0
    Dec 7 14:33:32 	dhclient: New Broadcast Address: xx.xx.xx.255
    Dec 7 14:33:32 	dhclient: New IP Address (sis1): xx.xx.xx.58
    Dec 7 14:33:32 	dhclient: New Subnet Mask (sis1): 255.255.248.0
    Dec 7 14:33:32 	dhclient: New Broadcast Address (sis1): xx.xx.xx.255
    Dec 7 14:33:32 	dhclient: New Routers: xx.xx.xx.1
    Dec 7 14:33:32 	dhcpd: Internet Systems Consortium DHCP Server V3.0.3
    Dec 7 14:33:33 	dhcpd: Copyright 2004-2005 Internet Systems Consortium.
    Dec 7 14:33:33 	dhcpd: All rights reserved.
    Dec 7 14:33:33 	dhcpd: For info, please visit http://www.isc.org/sw/dhcp/
    Dec 7 14:33:33 	dnsmasq[95]: reading /var/db/dhcpd.leases
    Dec 7 14:33:37 	/usr/local/bin/ez-ipupdate[98]: SIGHUP recieved, re-reading config file
    Dec 7 14:33:37 	/usr/local/bin/ez-ipupdate[98]: failure to update sis1->xx.xx.xx.58 (WWWWWWW.TTTTTT.net)
    Dec 7 14:33:41 	dhclient: bound to 84.90.74.58 -- renewal in 2799 seconds.
    Dec 7 14:37:03 	dnsmasq[95]: reading /etc/resolv.conf
    Dec 7 14:37:03 	dnsmasq[95]: using nameserver 213.228.128.6#53
    Dec 7 14:37:03 	dnsmasq[95]: using nameserver 213.228.128.99#53</sandisk></option></isa></generic></mii></natsemi></generic></mii></natsemi></pc></ti></pc></ti></pci></amd></fpu>
    

    I'm lost here.
    Why will m0n0 work fine and pfS have these issues?
    Same hardware and software configs…
    TIA
    Cheers



  • I have no idea what is causing this.  Perhaps this is one of those cases where m0n0wall is a better fit.

    Sorry!



  • But aren't these embedded images supposed to work on net4501/net4801 and WRAP?
    Because my net4521 is really a net4501 with just 2 NICs.
    Anyway, no problem pal.
    I know you fried your mind with this yesterday.
    Thanks for the help.
    Cheers



  • @rds_correia:

    But aren't these embedded images supposed to work on net4501/net4801 and WRAP?
    Because my net4521 is really a net4501 with just 2 NICs.
    Anyway, no problem pal.
    I know you fried your mind with this yesterday.
    Thanks for the help.
    Cheers

    Yes they work fine.  I have tons of 4501's, 4511's and 4801's.  In fact, I have 50+.

    At any rate, I cannot duplicate this here.  I'm pushing 15+ megabit on a 4511.



  • @sullrich:

    At any rate, I cannot duplicate this here.  I'm pushing 15+ megabit on a 4511.

    Hey! Don't say that!
    That will only make me feel more frustrated ;D.
    No problem, Scott.
    I'll catch you guys when we're more closer to 1.0.
    Any time frame for 1.0?
    Cheers



  • Ooops, just noticed the change log for 0.95.8 - currently under download ;-).
    Would it be reasonable to say that previous releases had issues with WoL/SNMP/DynDNS and that these have been solved?
    Because you see, these were some of my other complaints, that is, besides the slow networking performance.
    Cheers



  • @rds_correia:

    Ooops, just noticed the change log for 0.95.8 - currently under download ;-).
    Would it be reasonable to say that previous releases had issues with WoL/SNMP/DynDNS and that these have been solved?
    Because you see, these were some of my other complaints, that is, besides the slow networking performance.
    Cheers

    Yep.  Don't know about the speed problem, but its possible.



  • Alright!
    Cause now I also have a 2 meter CAT5 cable.
    I'll test 1st thing in the morning ;).
    Cheers



  • UPDATE! UPDATE!
    Currently testing 0.96.
    Problem with SNMP and WoL is fixed.
    EDITED: WoL is not working. Service still stops whenever I creat a clients list for WoL. Sorry for misleading you.
    Still hasn't tested the DynDNS problem but I will do so later on today.
    But the real UPDATE is that with a 2 meter cable I still have speeds of up to 100Kb/s.
    But then I changed to a (most probably) 5 meter CAT5 cable and I am now surfing even faster than with m0n0.
    Speed is now stable between the 400Kb/s and the 456Kb/s.
    It's simply the fastest firewall I have seen here at home.
    BTW I still get "kernel: sis1: Applying short cable fix (reg=e8)" messages but now they are only a few compared to 0.95.4.
    I was wrong about the cable lenght because with m0n0wall I have never seen these issues.
    But remember, it took me a 5 meter cable to do the trick.
    Thanks a bunch for all the help guys.



  • What do you mean by creating a client list for WOL?  Can't say that I am familiar with that.



  • Maybe I didn't express myself clear in the previous post.
    Sorry but English is not my native language.
    I meant the MAC table that is under Services > Wake on LAN.
    When you open that part of the config, on the top you'll see an Interface drop down and MAC address field so that you can Wake up a host on your LAN/WAN/OPT.
    Below you see a a button that says "You also can wake all clients at once" if you press it.
    And below that you have a table of MACs and nicknames of your most critical hosts.
    That table makes it easier to wake up host when you're not there and don't have a list of all the MACs with you.
    That's the table that I'm talking about.
    If I add a host for my home PC (MAC and nickname) and then I look at the Status > Services I will find out that the WoL service has stopped.
    Then if I remove my PC from the WoL hosts table and I check the Status > Services again I can see that the service is not even mentioned there.
    I am positive that this is what's happening.
    Cheers



  • Okay that's strange, the two shouldn't even really be related.

    I'll check it out.



  • Thanks ;).
    I still haven't tested DynDNS because I'm affraid they ban me for abuse of their service.
    But I think I'll give it a try later this evening.
    Cheers



  • @rds_correia:

    Thanks ;).
    I still haven't tested DynDNS because I'm affraid they ban me for abuse of their service.
    But I think I'll give it a try later this evening.
    Cheers

    It works.  See this thread:

    http://www.mail-archive.com/support@pfsense.com/msg03300.html

    Another reason why the mailing lists give more up to date information.



  • I had already read it before you posted the link.
    But since the DynDNS issue has appeared more than once and since I have moved on to 0.96 and the mail stated 0.95x…
    BTW, if all the people in the mailing list used the forum the forum would have all the info updated.
    That's why at 1st I thought that you should close the mailing lists.
    But then I thought "I like forums better than mailing lists but there may be people don't think in the opposite way".
    And that's why I wrote in the Feedback sub-forum that maybe the mailing list should keep working.
    What suites me the best might not be what suites everybody else the best.
    But then we come to a fact that information is too much spread siuch as in this particular case...
    But maybe this is pure BS.
    Anyway thanks for the info ;).
    Cheers



  • Reemplaze the card of net of your pc and try again, it me worked to my  :)



  • This thread is OLD.  Please let it DIE!

    And please find a new translator… Whatever you are using is incomprehensible.


Locked