Author Topic: MultiMac ethernet drivers updated on BetaZone - Intel, NVIDIA, Realtek  (Read 2715 times)

eirik

  • Newbie
  • *
  • Posts: 14
    • View Profile
http://www.os2world.com/content/view/21584/2/
Quote
Posted by Martin Iturbide - Tuesday, 20 March 2012


MultiMac i. .. .

If you run a ThinkPad T410 do not install the new Multimac driver, Intel e1000e-0.2.1.zip, as boot fails, and it trashes your file system.

If you installed this driver and had the same unfortunate experience as I did, boot eCS from CD, go to management console, and run chkdsk /f on all your jfs drives.  From a command line facility, revert back to your past Intel e1000 MultiMac driver (in my case version 0.1.8), and reboot.

Luckily, all went well and I am up and running again :-) after some "scary" minutes.



miturbide

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 1154
    • View Profile
    • OS2World
thanks for the warning eirik.

I assumed that only file system driver were dangerous, but I never thought a network driver can corrupt a partition. (that's my empiric knowledge talking)

Did you get any reference for someone else having the same issue? or do you think also it may be the intel driver combined with something specific of your configuration?
Martín Itúrbide
OS2World.com NewsMaster
Open Source Advocate

Skype - martiniturbide
Google Talk - martiniturbide@gmail.com

DougB

  • Sr. Member
  • ****
  • Posts: 407
    • View Profile
Quote
If you run a ThinkPad T410 do not install the new Multimac driver, Intel e1000e-0.2.1.zip, as boot fails, and it trashes your file system.

The 0.2.1 driver does stop the boot at NETBIND.EXE, on my ThinkPad T510, but there was no file system corruption. I went back to the 0.1.8 version, which works perfectly.

I expect that here really is no corruption, but the file system will be left dirty when you force a reboot. If you don't have the proper /AUTOCHECK parameter on the IFS line, in CONFIG.SYS, it will appear that the file system is broken until you run CHKDSK /F.

On the bright side, the r8169 0.0.3 driver works on my main system, where the older r8110 driver did not.

eirik

  • Newbie
  • *
  • Posts: 14
    • View Profile
Doug,

the proper autocheck is in place (it checks the jfs volumes on every startup), but this time the power off was "too brutal" for that to work, necessitating a chdksk /f (which only has happened once before in conjunction with some other updates that did not work).

On the bright side,
  • version 0.1.8 of the Intel e10000e MultiMac driver works excellently, and boots fast in my case
  • the RealTek driver works in your case

As the MultiMac project matures, I am certain we will have more functional drivers, both for wired and wireless access.  Hence, to MultiMac developers: Keep up the good work!

rwklein

  • Full Member
  • ***
  • Posts: 166
    • View Profile
http://www.os2world.com/content/view/21584/2/
Quote
Posted by Martin Iturbide - Tuesday, 20 March 2012


MultiMac i. .. .

If you run a ThinkPad T410 do not install the new Multimac driver, Intel e1000e-0.2.1.zip, as boot fails, and it trashes your file system.

If you installed this driver and had the same unfortunate experience as I did, boot eCS from CD, go to management console, and run chkdsk /f on all your jfs drives.  From a command line facility, revert back to your past Intel e1000 MultiMac driver (in my case version 0.1.8), and reboot.

Luckily, all went well and I am up and running again :-) after some "scary" minutes.




Well the driver it self most likely did not trash you file system. If the NIC driver however causes a RING 0 exception it can leave the file system in a bad state. Please open a support ticket as documented in the readme.
Or goto http://svn.ecomstation.nl/multimac

Did the system trap ? (black screen with numbers and stated it trapped in module "e1000e" ?

Thanks,

Roderick Klein
Mensys


If the system is trappking in e1000e driver (listed module) then please make a copy of the trap screen.

eirik

  • Newbie
  • *
  • Posts: 14
    • View Profile
The system did not trap, but after 15 minutes waiting I chose to power down the system.  It had then come to the e1000e driver, but stopped there.

I have checked log files, but none of those give any indication of what may have gone wrong.

As a user it is of little comfort to me if it is the driver or ring0 that kills my file system (that I was able to restore using chkdsk /f).  I am a user of eCS because I like the user interface and the way the system (usually works), but I am no computer expert.  Hence I am not able to offer much assistance on the error search (and I was unable to logon to the trac or find how to get a user name/password - if I spent some more time, I may have figured that out, but some times time is just not there)

abwillis

  • Sr. Member
  • ****
  • Posts: 325
    • View Profile
As a user it is of little comfort to me if it is the driver or ring0 that kills my file system (that I was able to restore using chkdsk /f). 
It seems odd that the file system would get into that bad a state just from a hang during boot.  The problem was that the file system was not clean when it shut down (you had no choice, of course, if Ctrl-Alt-Del fails to respond)... I almost suspect there was already something going on with your file system prior to this and this just brought it to the surface.  If this is JFS, I have found that it doesn't always do a full chkdsk when it really should but rather just verifies the log and the metadata all checks out it goes on while there really is a problem it needed to run a full chkdsk on.  It appears to be a design decision to get servers back up quickly after crash or sudden power loss etc. and is more often than not just fine (JFS used to be quite unstable but has improved to quite a good file system and I haven't seen this issue in a few years now but I used to have to run a full chkdsk /f:3 any time the system crashed to make sure there were not any problems). 

DougB

  • Sr. Member
  • ****
  • Posts: 407
    • View Profile
Quote
It had then come to the e1000e driver, but stopped there.

If you had used Alt-F2 to make the system list the drivers etc. as they load, you probably would have seen that it got to NETBIND.EXE before it hung. I keep a file named AltF2ON.$$$ in the root of my boot drive, and that causes the same thing as doing Alt-F2 at boot time. I always see the list of what got loaded, and the system stops at NETBIND.EXE when I try the e1000e driver. I have done that about a dozen times, with no file system problems.

Quote
JFS used to be quite unstable but has improved to quite a good file system and I haven't seen this issue in a few years now

I find the new JFS to be far better than HPFS ever was. The startup chkdsk rarely does more than the quick check, and when it does the longer check, it rarely finds anything to "fix". I think I have seen that only once since the bootable JFS was released. HPFS used to "fix" something every time it decided to do the chkdsk at boot time (usually just log files, but...). I do, however, make a point of running CHKDSK /F to each JFS drive when I happen to boot to my bootAble maintenance partition. It never finds anything to "fix", but I do it anyway because I know that it is possible for the boot time chkdsk to miss things. I suspect that Eirik just stumbled upon a bad combination of circumstances, and the full chkdsk was required to fix something.

Blonde Guy

  • Sr. Member
  • ****
  • Posts: 283
    • View Profile
    • Blonde Guy
I tried the new Realtek driver. While it does not crash, it also does not get a MAC address. I see now the new driver supports NETADDRESS=, which is good, but it doesn't work, which is bad.

I had a bug open (http://svn.ecomstation.nl/multimac/ticket/4), but I see that it is closed, so it's better to report here.

I installed the Realtek 8169 driver, which works, but only with the NETADDRESS= parameter.
Expert Consulting for OS/2 and eComStation

DougB

  • Sr. Member
  • ****
  • Posts: 407
    • View Profile
Quote
While it does not crash, it also does not get a MAC address.

Odd. I am using the MultiMac 8169 driver, from r8169-0.0.3.zip, on my main system (where the original 8168 NIC died), and I get a MAC address. NETSTAT -N and my router, both recognize it properly. No need for the NETADDRESS= parameter. The device is a cheap AirLink+ 10/100/1000 PCI adapter, but it seems to work well, even if VSWITCH doesn't like the new driver much (it works good with the GENMAC driver).

Quote
but I see that it is closed, so it's better to report here.

Probably better to reopen your incident, and give them hell for closing it without supplying a fix (unless the fix is in r8169-0.0.3.zip, which is available with Software Subscription).

Paul Smedley

  • Sr. Member
  • ****
  • Posts: 322
    • View Profile
Quote
but I see that it is closed, so it's better to report here.

Probably better to reopen your incident, and give them hell for closing it without supplying a fix (unless the fix is in r8169-0.0.3.zip, which is available with Software Subscription).

The tickets that David closed were raised against the R8110 driver which has been replaced with the r8169 driver. If the problem still exists with the latest r8169 driver then a new ticket should be raised - no hell raising required - that is, determining unless you want to pass David off.

Blonde Guy

  • Sr. Member
  • ****
  • Posts: 283
    • View Profile
    • Blonde Guy
I open more tickets than most, because I go through a lot of computers. In this case, the computer is sold, so I guess there is no point in raising the issue. It won't be long before I get another computer with a Realtek. If the driver doesn't work in that one, I'll open an ticket. I'd like to have my tickets correspond to hardware where I can actually test a fix.

In this case, I tried all of the MultiMac Realtek betas, but none of them got an IP, even ones that worked for other people.
Expert Consulting for OS/2 and eComStation