| Title: | Alpha Workstation Conference |
| Notice: | See note 1.* for conference notices |
| Moderator: | WRKSYS::HOUSE |
| Created: | Wed Sep 07 1994 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 1996 |
| Total number of notes: | 9122 |
A customer of mine has a problem on an AlphaStation 250 4/266:
For an unknown reason (probably AFS), the system hang. Then the halt
button is pressed and the >>> appears. An init command is issued and
the machine starts to boot (auto_action BOOT) until it detects and
(load the driver) for the fddi interface and then crash with the
following:
fta0 DEC DEFPA FDDI Module, Hardware Revision 0
apecs_read_io_port: illegal handle 0x0
panic (cpu 0): apecs_read_io_port
At that time, the only way to reboot the system is to switch it OFF
then ON. It seems that the initial hang is leaving the fddi in an
active state and when rebooting a leftover interrupt is causing the
apecs_read_io_port 's error...
Is this a Firmware problem ? (I have revision 6.3)
The configuration is as follows:
Digital UNIX V3.2C (Rev. 148)
physical memory = 256.00 megabytes
AlphaStation 250 4/266 system
Firmware revision: 6.3
PALcode: OSF version 1.46
rz0 at scsi0 bus 0 target 0 lun 0 (DEC RZ26N 0616)
rz1 at scsi0 bus 0 target 1 lun 0 (DEC RZ28M 0616)
rz4 at scsi0 bus 0 target 4 lun 0 (DEC RRD45 1645)
fta0 DEC DEFPA FDDI Module, Hardware Revision 0
fta0: Firmware rev: 2.46
Any answers will be welcomed.
Thierry Aubert/DEC at CERN
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 1931.1 | Will FW v6.4 fix this ??? | PANTER::MARTIN | Be vigilant... | Tue Apr 22 1997 09:44 | 20 |
Has anybody encountered such FDDI initialisation problem on
Alphastation 250 when Alphastation crash ??
This looks to us a problem we had on Alphastation 255 FW v6.0 sometime
ago which was corrected by FW update (cannot remember which one)...
Shall we need to IPMT this ?
The problem is we cannot provide any data for engineering as no
crash dump can be taken... the only way to get rid of this is
to switch off/on the system !
Does V6.4 FW will solve that, release notes do not specify such
a FDDI fix...
Cheers,
============================
Alain MARTIN/SSG Switzerland
| |||||
| 1931.2 | OZROCK::HARTWIG | Arthur Hartwig, TaN Engineering-Australia | Tue Apr 22 1997 15:44 | 17 | |
Probably a couple of years ago I reported a problem with AlphaStations
not resetting i/o devices on boot. I suggested this had potential for
"warm" reboot problems: the device could interrupt or DMA memory while
the console is loading the OS. The console firmware people I discussed
this with didn't seem to be very interested in doing anything about it.
I don't know if this is the same problem.
Suggestion: Disconnect the workstation from the FDDI before trying to
boot (maybe this will prevent FDDI interrupts or DMA during the boot).
Caveat: I have no idea what disconnecting the workstation from the FDDI
will do to the controller (maybe it will generate a stream of interrupt
requests) or the FDDI.
Suggeestion: Have you tried the console CRASH command - its pretty
effective on VMS and just might set the controller back into a state
such that a subsequent boot won't fail.
| |||||
| 1931.3 | NETCAD::STEFANI | FDDI Adapters R Us | Tue Apr 22 1997 17:51 | 27 | |
>> Suggestion: Disconnect the workstation from the FDDI before trying to
>> boot (maybe this will prevent FDDI interrupts or DMA during the boot).
Pulling the cable will stop receive packets from coming in, so you'll
effectively eliminate DMA writes for incoming traffic. On the other hand,
unless the driver is paying attention to the link state, DMA reads for outgoing
traffic may still occur. The only guaranteed way of stopping DMA's is to bring
the adapter to the DMA_UNAVAILABLE state. The driver can do this by issuing
the proper command. Alternatively, a bus reset will have the same effect.
As for interrupts, unless interrupts are disabled at the adapter, the
driver WILL receive a link state change interrupt when you disconnect the cable.
>> Caveat: I have no idea what disconnecting the workstation from the FDDI
>> will do to the controller (maybe it will generate a stream of interrupt
>> requests) or the FDDI.
Disconnecting the workstation is not harmful to the FDDI ring, it will just
wrap around your connection. The driver will receive a state change interrupt
unless that interrupt is masked or interrupts are disabled.
My advice for ANY operating system device driver operating a bus master DMA
controller is to take the DMA engine off-line prior to shutdown, reboot, or
reset. This is the 100% safest way to eliminate DMA operations that could
occur after the system no longer wants them.
- Larry
| |||||
| 1931.4 | I will open an IPMT. | PANTER::AUBERT | Wed Apr 23 1997 04:42 | 6 | |
Thanks for your workaround but the customer need to have a fix for
this problem.
So I will open an IPMT.
Thierry Aubert, DEC at CERN
| |||||