Since I get any response I try again -- hoping that somebody have ANY idea! A network with a RedHat 6.0 linux server and samba 2.0.6 and Windows 95 clients: When I for example copy large (binary?) files from local drives to one of my Samba shares I VERY often experience that the network connection to the Win95 machine crashes. The lovely Windows Blue-Screen message about an exception appears, with a message something like (translated from danish): --------------------------------------------------------------------- An exception OE has appeared at 0028:C104311B in VxD ELNK3(01)+ 00000E4B. This was called from 0028:C004374B in VxD NDIS(01)+ 00006733. You may be able to continue as normal. --------------------------------------------------------------------- I am NOT able to continue, I am not even able to restart the computer normally, have to use power off/power on method.... I have found a workaround -- I can just establish a ftp-connection to the server and copy the files via that. Apparantly that never fails. It is rather ackward, though. PLEASE someone -- how do I diagnose the problem? - Carsten
On Mon, 3 Apr 2000 17:59:01 +1000, hai scritto:>When I for example copy large (binary?) files from >local drives to one of my Samba shares I VERY often >experience that the network connection to the >Win95 machine crashes. The lovely Windows Blue-ScreenWhat happens if you copy the same file on another win9x share? -- giulioo@pobox.com
Carsten Nordstr?m Jensen wrote:> --------------------------------------------------------------------- > An exception OE has appeared at 0028:C104311B in VxD ELNK3(01)+ > 00000E4B. This was called from 0028:C004374B in VxD NDIS(01)+ > 00006733. You may be able to continue as normal. > --------------------------------------------------------------------- > I am NOT able to continue, I am not even able to restart > the computer normally, have to use power off/power on > method....Looks like a NIC drivers problem. Have you downloaded the latest drivers from your NIC manufacturer? You should also install the "newer" Win95 releases ("Win95b" or "Win95c") on your clients or apply MS' Win95 service releases. -- Thomas
Giulio Orsero wrote:> On Mon, 03 Apr 2000 11:15:47 +0200, hai scritto: > > >On the "usual share I am able to get a crash at anytime. Can't > >that be exploited for diagnosing? > Then I would set the samba log level to 4 or more (up to 10), be sure > noone else is doing anything with samba, try the copy, and see if the > samba log shows a particular error when the win9x crashes.Ok, I have experimented somewhat. If the debug level is set to 4 or less, I can create a crash anytime. But NOTHING is written in the samba log (/var/log/samba/log.smb) at the time of the crash on the Win95 client. If the debug level is set to 5 or above, apparently I cannot reproduce the crash!! Isn't that a bit weird? - Carsten
> | A network with a RedHat 6.0 linux server and samba 2.0.6 and > | Windows 95 clients: > | When I for example copy large (binary?) files from > | local drives to one of my Samba shares I VERY often > | experience that the network connection to the > | Win95 machine crashes. > > first, measure the transfer speed of an "average" > file by copying it with smbclient a few times. > > Then set raw read = no and raw write = no and > see how much it slows down (amd mail me the results, > please for my archive) > > If it's not too much, leave these settings in place > and see if this affects the crashing behavior. > > A colleague was reporting a less severe problem > which **seemed** to respond to this, but I'm > not at home and can't reproduce it anyway...That doesn't change the crashing behaviour. Till now the only thing that DOES change the crashing behaviour is to set the debug level to 5 or above -- obviously not a proper solution. Let me note that until last summer the clients were sitting on a Novell network and did NOT experience such behaviour -- which of course makes me believe it is a SAMBA problem.... - Carsten
Carsten Jensen <north@get2net.dk> wrote: | That doesn't change the crashing behaviour. Till now | the only thing that DOES change the crashing behaviour | is to set the debug level to 5 or above -- obviously not | a proper solution. Cool: it's a timing issue, I'll bet! Running high logging levels slows Samba a lot, so let's try an experiment with one of the new timing options (suggested by yet another commentator on the list, Mark Ruedy) oplock break wait time = 100 This was discussed by Jeremy quite some time ago: | Someone just reported on the mailing list that upping | "oplock break wait time" parameter from the default of 10 ms | to 100ms fixed oplock break problems with Windows 98. | If anyone is still having this problem with 2.0.6 could you | try this fix and let me know the result. I want to find out if | it improves things enough to make it the default. --dave David Collier-Brown in Boston Phone: (781) 442-0734, Room BUR03-3632