Bug 688454

are you using a wired network interface ?

(In reply to comment #1)
> are you using a wired network interface ?


Yes. I should add that I'm currently running 7 suse machines on this (university) network. The SLES 11, the 11.2 and the 11.3 machines  all download all of their updates perfectly happily.

Being a university network, please double check that there is or there is not proxy settings, on the affected machine and the others, for comparison. It is in /etc/sysconfig/proxy, IIRC.

There's no proxy on the machine, and there's no proxy on the network. Historically there was but it was turned off last year.

I have had the same problem (with the same files as listed in the examples here) since upgrading to 11.4.
Using zypper upgrade, or Yast doesn't seem to make any difference, the download stalls. It would appear from the messages that a partial file is downloaded in some examples, but the directory entry stored on the system shows 0 bytes.

The error message using zypper update shows only a partial file specification, for example:
./suse/i586/qt3-3.3.8b-111.1.i586.rpm

full file description example:
http://download.opensuse.org/distrib...111.1.i586.rpm

It is not possible to download the file on the target 11.4 system using Firefox. The download stalls. However, it is possible to use this same full file specification listed in the zypper update to download the file from Firefox on another computer, in my case, a SuSE 11.2 machine.

It is then possible to copy the file from the other system into a directory tree that reflects the shortened path as shown above, and from top of the directory tree run "zypper refresh", followed by "zypper update" to install the file. If "zypper update" is not run from the directory at the top of the directory tree, zypper attempts to download the file, which fails. I have noticed in a couple of instances that zypper downloads the file, but is able to do so and complete the installation without stalling if the file is already on the system as described.

By this time, a number of sub-directories have been created to hold update files:
./Essentials/i586/
./rpm/i586
./suse/i586 ./suse/i686/ ./suse/noarch/

Searching the internet using a search engine shows similar problems with earlier SuSE versions (and other varieties of Linux). However, I have been unable to find any resolution. Search using the terms 'SuSE 11.4 Download (curl) error for: Can't provide'. My apologies if that leads to a goose chase, but it lead me to comments regarding this issue from other users.

Upgrade attempts from runlevel 3 produce the same result.

HTH

Bill

Created attachment 426053 [details]
Wireshark output while attempting to download konqueror-plugins-lang

Zypper got to 89% downloaded then the rate went down to 0. I left it like that for a few seconds then aborted.

The last received data packets are correctly acknowledged. No more packets are received until the connection is locally closed, at in that moment the ACK for the FIN packet shows more packets were lost.
So... network card driver problem? Would be good to know which network cards affected users have.

My network card is a 3C905CX-TX/TM Fast ethernet link, used with the 3c59x driver.

Ok, let's ensure it is not an already fixed possible kernel bug, can you install latest kernel from Kernel:HEAD repository ?

(In reply to comment #7)

> So... network card driver problem? Would be good to know which network cards
> affected users have.

Extracting info from forum posts:

Trollicus: Dual core 2.88 MHz intel, Nvidia 8800 GTX 784 MB, 4 GB 800MHz RAM, 2x
500GB WD Serial ATA HD, 3c905x network card. HDMI Audio.

hbco2: not known.

akwatton: (home, no proxy; runlevel 3 works)
          "3Com 3C905c-TX/TXM ethernet card in PC."


So, at least three people have the same card.

Just to say that I'm gone for Easter now but I'll try with the new kernel when I'm back towards the end of next week.

I too use a 3com 3C905c-TX/TXM ethernet card and have had problems downloading with YaST software manager or wget. Many files will download OK, but I'm getting repeated problems with acroread (which always hangs in the same spot at 99% of the download), VLC (hangs downloading vlc-Nox) or the restricted formats (hangs downloading Ladspa). I also get the same problem when I try and download or update using failsafe mode, which other users on the forums say works for them. However, I have tried wget and YaST software manager with acroread on runlevel 3 as "su -" and it downloaded successfully. Tried wget again later in failsafe or normal logon and it sticks at the same point as before. Also still hangs using yast online update etc.

When I tried to use wget on one of the update files that was hanging the output I got was

 wget http://download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm
asking libproxy about url 'http://download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm'
libproxy suggest to use 'direct://'
--2011-04-14 15:13:24--  http://download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm
Resolving download.opensuse.org... 195.135.221.134
Connecting to download.opensuse.org|195.135.221.134|:80... connected.
HTTP request sent, awaiting response... 302 Found
Location: http://www.mirrorservice.org/sites/download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm [following]
asking libproxy about url 'http://www.mirrorservice.org/sites/download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm'
libproxy suggest to use 'direct://'
--2011-04-14 15:13:24--  http://www.mirrorservice.org/sites/download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm
Resolving www.mirrorservice.org... 212.219.56.134, 212.219.56.135, 212.219.56.138, ...
Connecting to www.mirrorservice.org|212.219.56.134|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 93811 (92K) [application/x-redhat-package-manager]
Saving to: `sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm'

40% [=======================================================>                                                                                   ] 37,899      --.-K/s   in 15m 0s  

2011-04-14 15:28:25 (42.1 B/s) - Read error at byte 37899/93811 (Connection timed out). Retrying.

--2011-04-14 15:28:26--  (try: 2)  http://www.mirrorservice.org/sites/download.opensuse.org/update/11.4/rpm/i586/sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm
Connecting to www.mirrorservice.org|212.219.56.134|:80... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 93811 (92K), 55912 (55K) remaining [application/x-redhat-package-manager]
Saving to: `sane-backends-1.0.22-1.3.3_1.5.1.i586.delta.rpm'

41% [++++++++++++++++++++++++++++++++++++++++++++++++++++++++>       




...and this repeats ad infinitum.

I can confirm that the download worked when using a level 3 boot, but no other mode will work for me, not even failsafe.

(In reply to comment #10)
> (In reply to comment #7)
> 
> > So... network card driver problem? Would be good to know which network cards
> > affected users have.
> 
> Extracting info from forum posts:
> 
> Trollicus: Dual core 2.88 MHz intel, Nvidia 8800 GTX 784 MB, 4 GB 800MHz RAM,
> 2x
> 500GB WD Serial ATA HD, 3c905x network card. HDMI Audio.
> 
> hbco2: not known.
> 
> akwatton: (home, no proxy; runlevel 3 works)
>           "3Com 3C905c-TX/TXM ethernet card in PC."
> 
> 
> So, at least three people have the same card.

hbco2:

  Card model 3Com 3C902C-TXM
  Model: "Ethernet network interface"
  Driver: "3c59x"

  Great analysis!  Should another card be used?

Ok, clearly is a kernel bug, cannot be just coincidence that affected users are using the same kernel driver...

(In reply to comment #14)
> Ok, clearly is a kernel bug, cannot be just coincidence that affected users are
> using the same kernel driver...

Could it be a certain sequence of bytes triggering this? Is there a way to find out?

Well, this is a strange one...

Can someone experiencing the problem try adding "noapic" (without quotes) to the kernel boot commandline and try like that?

I've just added noapic to the kernel boot command line and tried to install acroread using Yast software manager. Still hangs in exactly the same place and gives the same error message as before.

(In reply to comment #18)
> I've just added noapic to the kernel boot command line and tried to install
> acroread using Yast software manager. Still hangs in exactly the same place and
> gives the same error message as before.

Can you try a newer kernel from Kernel:HEAD repository ?

(In reply to comment #19)
> 
> Can you try a newer kernel from Kernel:HEAD repository ?

I have installed the latest Kernel from Kernel:HEAD repo the output from cat
/proc/version is:

Linux version 2.6.39-rc4-10-default (geeko@buildhost) (gcc version 4.5.1
20101208 [gcc-4_5-branch revision 167585] (SUSE Linux) ) #1 SMP 2011-04-22
00:02:11 +0200


I rebooted the PC, logged in with the new kernel, ran software manager in yast
and selected acroread for install. Unfortunately, this still hung at 99% of
download.

(In reply to comment #20)

> I rebooted the PC, logged in with the new kernel, ran software manager in yast
> and selected acroread for install. Unfortunately, this still hung at 99% of
> download.

Thanks, will be cool if you keep running Kernel:HEAD packages in case kernel devs need more information on your setup.

Also, please upload a txt file with the output of "hwinfo" as root.

also, try

rmmod 3c59x
modprobe 3c59x debug=6

and then attempt to download stuff, after it hangs, post the the output of "dmesg"

Created attachment 426333 [details]
Output of hwinfo as root

(In reply to comment #22)
> also, try
> 
> rmmod 3c59x
> modprobe 3c59x debug=6
> 
> and then attempt to download stuff, after it hangs, post the the output of
> "dmesg"

This is very strange. I tried the above last night, then tried to download acroread, and believe it or not it downloaded. This is the first time it hasn't hung on download in the four weeks I've had the system installed. I couldn't see the point in posting the output of "dmesg" as the download had been successful. I then tried to download all the other problem files I'd had and they downloaded successfully too. That was with the original kernel 2.6.37.1

This morning I thought I try and install acroread again using the later kernel from the Kernel:HEAD repo. When I loaded software manager there were 3 files waiting for update, so I accepted them and they hung when trying to download. I typed "rmmod 3c59x" and "modprobe 3c59x debug=6" again and the same as last night, the files downloaded OK. I've since tried to download acroread again and that also downloads fine.

This is really puzzling...

Since my last post, comment 24, I have deleted acroread. powered the PC down, rebooted and tried to reinstall acroread. As before, it hung, So again I
typed "rmmod 3c59x" and "modprobe 3c59x debug=6", tried to the download again and it worked. There's obviously something happening after typing "rmmod 3c59x" and "modprobe 3c59x debug=6" that's making the files download successfully. I will post the output of dmesg as requested in case there is something in there that may point to something.

Created attachment 426360 [details]
output of "dmesg" after downloading

(In reply to comment #25)
> Since my last post, comment 24, I have deleted acroread. powered the PC down,
> rebooted and tried to reinstall acroread. As before, it hung, So again I
> typed "rmmod 3c59x" and "modprobe 3c59x debug=6", tried to the download again
> and it worked. There's obviously something happening after typing "rmmod 3c59x"
> and "modprobe 3c59x debug=6" that's making the files download successfully. I
> will post the output of dmesg as requested in case there is something in there
> that may point to something.

try making "debug" permanent ;)

echo "options 3c59x debug=6" >> /etc/modprobe.d/99-local.conf
reboot

if you get an stall, post the dmesg output, thanks.

(In reply to comment #27)

> try making "debug" permanent ;)
> 
> 
> reboot
> 
> if you get an stall, post the dmesg output, thanks.

I've just typed echo "options 3c59x debug=6" >> /etc/modprobe.d/99-local.conf

I rebooted then tried to install acroread via yast software manager and it stalled. I aborted the install after a while. I have posted the output of dmesg. Hope I've done this correctly and it has what you need in there.

Created attachment 426415 [details]
ouput of dmesg after trying to download acroread and it stalls.

  0.000000] Using APIC driver default
[    0.000000] ACPI: PM-Timer IO Port: 0x808
[    0.000000] SMP: Allowing 1 CPUs, 0 hotplug CPUs
[    0.000000] Local APIC disabled by BIOS -- you can enable it with "lapic"
[    0.016583] ACPI: Core revision 20110316
[    0.019211] ACPI: setting ELCR to 0200 (from 0820)
[    0.020157] weird, boot CPU (#0) not listed by the BIOS.
[    0.020160] SMP motherboard not detected.
[    0.020163] Local APIC not detected. Using dummy APIC emulation.

can you try to boot the system with kernel boot option "lapic" (without quotes)
does it actually boots and/or works correctly ?

(In reply to comment #30)
>   0.000000] Using APIC driver default
> [    0.000000] ACPI: PM-Timer IO Port: 0x808
> [    0.000000] SMP: Allowing 1 CPUs, 0 hotplug CPUs
> [    0.000000] Local APIC disabled by BIOS -- you can enable it with "lapic"
> [    0.016583] ACPI: Core revision 20110316
> [    0.019211] ACPI: setting ELCR to 0200 (from 0820)
> [    0.020157] weird, boot CPU (#0) not listed by the BIOS.
> [    0.020160] SMP motherboard not detected.
> [    0.020163] Local APIC not detected. Using dummy APIC emulation.
> 
> can you try to boot the system with kernel boot option "lapic" (without quotes)
> does it actually boots and/or works correctly ?

OK, I've booted with kernel boot option "lapic" and kernel 2.6.39-rc4-10-default. Again tried to install acroread with TaST software manager and it stalled. i waited for the error message to show, and aborted at that point. I have posted output of dmesg again.

When I've booted with this version of the kernel, it boots, but the graphics are all broken up on the bootsplash/login screen, but system seems to work ok after. When using the kernel boot option "lapic" the bootsplash/login screen is ok. 

Don't have problems with login screen at all if use kernel 2.6.37.1

Only other problems I've come across are dowmloading.......

Hope this file is OK for you this time.

Created attachment 426456 [details]
output of dmesg after dowload stalls with kernel boot option "lapic"

(In reply to comment #31)

> When I've booted with this version of the kernel, it boots, but the graphics
> are all broken up on the bootsplash/login screen, but system seems to work ok
> after. When using the kernel boot option "lapic" the bootsplash/login screen is
> ok. 

Thta's a different issue, the problems star in your system BIOS, if you can please do this:

Enter your BIOS setup program and select "load optimized defaults(or something similar)"

Then 

a) in the power managment section, ensure ACPI is enabled.
b) try to find an option like this one http://www.techarp.com/showfreebog.aspx?lang=0&bogno=269 and ensure the IOAPIC is enabled or LAPIC..whatever.
b) This system is quite old it seems, ensure you have the latest BIOS from the motherboard manufacturer..

About the system messages,they seem normal

Message:  234.896368] eth0: vortex_error(), status=0x8081

comes from http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=drivers/net/3c59x.c line 1972 it always prints the device status when debug > 2... not something to worry about...

It is either a regression somewhere else in kernel since 11.3 or is just that the new kernel doesnt handle bugs in your BIOS...huh. I'm done some kernel developer has to check this one, my knownledge stops aroud here ;-(

*** Bug 688585 has been marked as a duplicate of this bug. ***

(In reply to comment #33)
> (In reply to comment #31)
> 
> > When I've booted with this version of the kernel, it boots, but the graphics
> > are all broken up on the bootsplash/login screen, but system seems to work ok
> > after. When using the kernel boot option "lapic" the bootsplash/login screen is
> > ok. 
> 
> Thta's a different issue, the problems star in your system BIOS, if you can
> please do this:
> 
> Enter your BIOS setup program and select "load optimized defaults(or something
> similar)"
> 
> Then 
> 
> a) in the power managment section, ensure ACPI is enabled.
> b) try to find an option like this one
> http://www.techarp.com/showfreebog.aspx?lang=0&bogno=269 and ensure the IOAPIC
> is enabled or LAPIC..whatever.
> b) This system is quite old it seems, ensure you have the latest BIOS from the
> motherboard manufacturer..
> 
> About the system messages,they seem normal
> 
> Message:  234.896368] eth0: vortex_error(), status=0x8081
> 
> comes from
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=drivers/net/3c59x.c
> line 1972 it always prints the device status when debug > 2... not something to
> worry about...
> 
> It is either a regression somewhere else in kernel since 11.3 or is just that
> the new kernel doesnt handle bugs in your BIOS...huh. I'm done some kernel
> developer has to check this one, my knownledge stops aroud here ;-(

Because I got exactly the same problem installing updates/patches and downloading acroread, Inkscape or Scribus as Tony Watson described earlier I would like to remind you that this bug is in the first place a problem within the 2.6.37.1-kernel for 32-bit-machines and not a possible failure of the motherboard BIOS. Until openSUSE 11.3 everything worked as it should and updating my 64-machine to 11.4 was successful also.

(In reply to comment #35)
> 
> Because I got exactly the same problem installing updates/patches and
> downloading acroread, Inkscape or Scribus as Tony Watson described earlier I
> would like to remind you that this bug is in the first place a problem within
> the 2.6.37.1-kernel for 32-bit-machines and not a possible failure of the
> motherboard BIOS. Until openSUSE 11.3 everything worked as it should and
> updating my 64-machine to 11.4 was successful also.


I agree with you Juergen, 11.3 worked fine and downloaded OK on this PC, so does Ubuntu, Windows XP and Linux Mint. Problems started for me the minute I installed opensuse 11.4 with 2.6.37.1 kernel. As you can see I've updated to the latest kernel from the Kernel:Head repo but the problems remain. 11.4 is the only OS I'm getting this problem with.

I am affected by the same problem. I have an older machine (Dell Dimension 4550) with 3 NICs running 11.4 (i586) in runlevel 3 as router/firewall:

eth0 - 3c59x - LAN
eth1 - 3c59x - Internet 1 (ADSL - websurfing/updates etc.) - default gateway
eth2 - e100  - Internet 2 (SDSL - corporate VPN)

Updates stall for several files, always at the same percent level. Same for downloading this file with wget or firefox/chrome/arora.

If I change the default gatway to eth2 (e100 NIC), file transfer works fine.

I can confirm that if I "rmmod 3c59x; modprobe 3c59x" after booting file transfers work fine via eth1 (3c59x) too.

Other client machines with different NICs running 11.4 (X64) are unaffected.

Okay, rather unhelpfully I have come back from my holiday and am now unable to reproduce this bug, including with packages which I know were failing, and despite not having installed any relevant updates over the holiday period. (Literally - the switches in our building were replaced so we were offline for most of it.)

(In reply to comment #37)

> I can confirm that if I "rmmod 3c59x; modprobe 3c59x" after booting file
> transfers work fine via eth1 (3c59x) too.

Ok, that's interesting, plz do the following.
install package "netdiag"
reboot
when the card stalls, run

# vortex-diag -aaee > vortex-diag.bad
# mii-diag -v > mii-diag.bad

then do rmmod 3c59x; modprobe 3c59x
and then

# vortex-diag -aaee > vortex-diag.good
# mii-diag -v > mii-diag.good

after that , create a tarball, with those files

tar -jcvf 3comdiag.tar.bz2 *-diag.*

upload the resulting tarball here.

(In reply to comment #39)
> (In reply to comment #37)
> 
> > I can confirm that if I "rmmod 3c59x; modprobe 3c59x" after booting file
> > transfers work fine via eth1 (3c59x) too.
> 
> Ok, that's interesting, plz do the following.
> install package "netdiag"
> reboot
> when the card stalls, run
> 
> # vortex-diag -aaee > vortex-diag.bad
> # mii-diag -v > mii-diag.bad
> 
> then do rmmod 3c59x; modprobe 3c59x
> and then
> 
> # vortex-diag -aaee > vortex-diag.good
> # mii-diag -v > mii-diag.good
> 
> after that , create a tarball, with those files
> 
> tar -jcvf 3comdiag.tar.bz2 *-diag.*
> 
> upload the resulting tarball here.

OK, I've tried the above and uploaded the tarball as requested. I've also updated kernel to 11.4 - 2.6.39-rc5-16 but problem still remains. I also updated bios as you suggested to latest from manufacturers site. PC did not work next day. Took two days to get working again with original bios.

Created attachment 427456 [details]
tarball of netbios tests

(In reply to comment #40)
> I also
> updated bios as you suggested to latest from manufacturers site. PC did not
> work next day. 

huh, sometimes you can mess stuff up.. ;)

>Took two days to get working again with original bios.

ok, Saw your files, they dont provide any useful hint :| we are stuck here.

(In reply to comment #42)
> (In reply to comment #40)
> > I also
> > updated bios as you suggested to latest from manufacturers site. PC did not
> > work next day. 
> 
> huh, sometimes you can mess stuff up.. ;)
> 
> >Took two days to get working again with original bios.
> 
> ok, Saw your files, they dont provide any useful hint :| we are stuck here.


You guys at Novell should check the kernels of openSUSE 11.3 and openSUSE 11.4 of the 32-bit-version  and surge for differencies. And you should also check the kernels of openSUSE 11.4 (32-bit) and openSUSE 11.4 (64-bit).

Both my PCs have nearly the same hardware-equipment, the 64-bit-machine works fine with openSUSE 11.4, the 32-bit-machine causes the o.m. troubles.

To download the latest kernel-update I had to go down to runlevel 3, but this is not the solution I wanted!

Again: The problem are not Tony Wattons or mine "old" PCs but the bug lies inside the software. I'm very interested to know if there is any smooth running 32-bit-machine with openSUSE 11.4 and if "YES" which hardware is installed.

So I have installed openSUSE 11.4 on a machine with:

0a:01.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 74)
# uname -a
Linux linux-21c0 2.6.37.1-1.2-desktop #1 SMP PREEMPT 2011-02-21 10:34:10 +0100 i686 i686 i386 GNU/Linux

and I am not seeing any issues thus far.

curl http://anorien.csc.warwick.ac.uk/mirrors/download.opensuse.org/update/11.4/rpm/noarch/timezone-java-2011b_2011d-1.2.1_0.2.1.noarch.delta.rpm > /dev/null
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 48677  100 48677    0     0  31783      0  0:00:01  0:00:01 --:--:-- 51238

(In reply to comment #39)
> Ok, that's interesting, plz do the following.
> ...

After the mentioned rmmod/modprobe action the problem did not reapear - even not after rebooting. I noticed that the kernel-firmware package was not installed by default, so I installed it manually. (Don't know, if it's relevant.)

Unfortunatley I can't test very much with this computer as this is a production machine 24/7.

Created attachment 429599 [details]
Tarball with 3com diags

Now the problem has happened again. I have provided the requested logs in the attached tarball. This computer has two 3com nics, so I provide logs for both of them. The affected nic however was the external nic (eth1).

This problem cropped up again this morning. Various updates had failed over the weekend, specifically kdelibs4, perl, python-base, and python-xml. Interestingly closely related packages such as kdelibs4-core, libkde4, libkdecore4, libksuseinstall1, libpython2_7-1_0, perl-base and python were fine.

As suggested, I did rmmod 3c59x then modprobe 3c59x debug=6 and was able to download kdelibs4. I then rebooted and, without having added these lines to rc.local, had no problem downloading the rest of the packages.

I am seeing this problem (same symptoms, same modprobe workaround) on my x86_64 system with a 64-bit 11.4 installation from DVD. The card (from lspci)

00:09.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 74)

(adding this comment because a post above suggested this is a 32-bit only issue, but I am seeing it on the 64 bit kernel).

Hmmm, this is pretty bizarre... here are some higher-level observations regarding this problem, but first some system configuration information:

* openSUSE 11.4
* 2.6.37.1-1.2-desktop #1 SMP PREEMPT 2011-02-21 10:34:10 +0100 i686 athlon i386 GNU/Linux
* 00:0d.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 74)
* Thomson ST536 v6 ADSL modem
* ISP: NZ Telecom / Xtra broadband

Symptoms: stalling RPM downloads on certain RPM files (but not all RPMs!) when using either Yast SW Management, zypper CLI, HTTP or FTP downloads via Firefox, or wget / wget -c. The stalling does not always happen, but when it does happen it is persistent for certain RPM files, irrespectively of the repository server being used. Otherwise networking appears to be fine at the same time, with DNS POP3, SMTP, web browsing, working fine and speedy without issues.

Now here comes the strange (and possibly interesting) thing: when I was in this stuck state, I tried going to runlevel 3 and then repeating the downloads with wget - again the same problem. Then I rebooted (into runlevel 5), but did not log into KDE. Instead, I only logged into tty1 and ran the same wget again, and it went perfectly smoothly on two files straight away.

So is there something that gets loaded or can become stuck when logging into KDE that may cause the trouble? I've got no idea...

Let me know if this is useful and if there is more specific information I could add.

Just to confirm - I see exactly the same symptoms on an old (2003-ish??) 64bit system with that network card. So it is not just the 32bit.

The workaround

> rmmod 3c59x
> modprobe 3c59x debug=6

gets things downloading again for me.

Michal, can you assign this bug to someone knowledgeable on the networking code, clearly a bug in the driver..

Until finding this bug report, my first thought was about faulty NIC hardware (as 3C905C-TX is quiet old card) or broken cable / connector, because problematic RPMs were those of bigger sizes, while the smaller packages were downloaded without error. But then I read this thread and eventually found that some bigger updates are also received just fine, as well as other files downloaded by HTTP, while in contrary at least one tiny delta-package (less than 40K) fails.

So my "test suite" was:

> $ wget -v -d -S http://download.opensuse.org/update/11.4/rpm/i586/ntp-4.2.6p3-6.9.2_6.11.1.i586.delta.rpm

running on 2.6.37.6-0.7-default i686 athlon. I must also add that I have a soft-bridge (created by system for Xen hypervisor), and that booting to either Xen Dom0 or simple bare-metal produces the same result.


My observations are:

1. In runlevel 3 everything works fine.

2. Proceeding to runlevel 5 without logging into KDE4 leaves things working.

3. Logging into KDE4 desktop introduces the discussed problem.

4. Restarting the NIC driver (rmmod/modprobe) does not help; moreover, it puts the network subsystem in an unusable state, so "service network restart" is required.

5. Simply running "service network restart" (without preliminary rmmod/modprobe) solves the problem.

6. Logging out of KDE4 does not affect things (neither improves nor worsens).

7. Subsequent logging into KDE4 does not affect things, too.


When I start the NIC driver with "debug=6", I actually do not see any additional information in dmesg, so I made a couple of tcpdump logs (attached in the next message).

Created attachment 447209 [details]
tcpdump IP/TCP headers log when downloading RPM with wget

logs of "tcpdump -n -vvv tcp port 80"

1. good (in runlevel 5, before logging into KDE4)
2. bad (after logging into KDE4)
3. bad (another try)
4. good (after running "service network restart")

Can you confirm that this is not just a broken mirror that you are being redirected to? If you open http://download.opensuse.org/update/11.4/rpm in a browser, there is a "Details" link for each file that lists all available mirrors for that file. Try downloading from several of them directly and share the results, thanks.

> Can you confirm that this is not just a broken mirror [...]

Clearly it is not a case of a broken mirror, because the workaround mentioned in earlier comments hides the problem! Also, I tried specifying different mirrors by hand, without that helping any.

(In reply to comment #55)
> Can you confirm that this is not just a broken mirror that you are being
[...]

Yes, I had tried various mirrors already, and it does not help. I have also seen the problem occur for a range of repositories hosted on different servers.

My usual work-around is logging into a text console straight after booting (and before anyone logs in via the graphical login) and running the downloads / installs from there using the old text-based yast, and I have never seen the problem when using that approach.

(In reply to comment #12)
> I too use a 3com 3C905c-TX/TXM ethernet card and have had problems downloading
> with YaST software manager or wget. Many files will download OK, but I'm
> getting repeated problems with acroread (which always hangs in the same spot 

I also have a 3com 3C905c-TX/TXM card and have been having problems with downloads. Just installed opensuse 12.1. My solution was to use my wireless card instead of wired. Problem went away.

K Brewer

(In reply to comment #58)

> I also have a 3com 3C905c-TX/TXM card and have been having problems with
> downloads. Just installed opensuse 12.1. My solution was to use my wireless
> card instead of wired. Problem went away.
> 
> K Brewer


There is a small improvement with openSUSE 12.1:

Still I have to go down to runlevel 3 for installation of some patches. But now with the new version 12.1 I'm also able to install (zypper install ....) AdobeReader (acroread), Scribus etc. which was not possible in openSUSE 11.4

I don't believe that the only source of our problem is the 3com ethernet card because most of all patches are downloaded and installed properly.

Juergen Moeller

IMHO the problem has nothing to do with a certain runlevel - with changing runlevels  - and because of this reinitializing network - maybe?

In our case only 32bit machines with 3C905 cards are affected. All operate in runlevel 3 only (firewalls - no gui).

These machines use ifup (not networkmanger) and static ip addresses.

My observations:

* "rmmod 3c59x ; modprobe 3c59x" always worked as a workaround.

* The problem seems to go away if I disable "assign hostname to loopback ip" option in yast - at least it didn't show up for about 2 weeks.

*** Bug 689094 has been marked as a duplicate of this bug. ***

> I don't believe that the only source of our problem is the 3com ethernet card
> because most of all patches are downloaded and installed properly.

My theory is that some rare bit pattern in the data triggers this bug, which is approaching its anniversary :-(. I know the idea sounds bizarre (why should
the driver care about the data), but I think I have seen the bug bite in other 
net usage as well: sometimes (rarely, and I cannot point to an example) large
images or Flashes hang, unless I have done the modprobe hack described in prior
comments. So I nowadays do it always by hand. For whatever reason I cannot
automate it in RCs, it has to be done after the system has "settled" (possibly
I have been doing some total loser's mistake in my attempts at automation).

With the coming release of openSUSE 12.2, openSUSE kernel developers are focusing their efforts there. Reports against openSUSE 11.4 and prior will not get the attention needed to resolve them before openSUSE 12.2 is release and openSUSE 11.4 becomes unmaintained.

Please re-test with openSUSE 12.1 or openSUSE RC2+ and re-open with an updated Product if you still encounter your issue.

We apologize for this issue not getting the attention it deserves but we are focusing our resources in the area where they will have the most impact for our users.  We're working hard to make openSUSE 12.2 the best openSUSE release yet!

> Please re-test with openSUSE 12.1

I'm using openSUSE 12.1 i586 on a Pentium III,
with the ethernet adapter 3C905CX-TX/TX-M Fast Etherlink for PC
running with driver 3c59x,
and I still experience the same problems,
downloading software packages, or updating.

Albeit with some difficulties
(see http://forums.opensuse.org/english/get-technical-help-here/install-boot-login/477084-online-update-will-shoot-running-os-12-1-i586.html)
I have installed all online updates for 12.1.

Reopening and moving to openSUSE 12.1 according to comment #64.

This report is against openSUSE 12.1 which is no longer under maintenance. If
you are able to reproduce it with openSUSE 13.1 or openSUSE Factory, please
re-open and reset the the "Product" field to the appropriate release.