Re: More MP problems.

[Adam K Kirchhoff <adamk@xxxxxxxxxxxx>]

Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx> wrote:

> :After some googling, I see that dumpdev can be set as a variable in the
> :bootloader (this should be added to the help for the "set" command" in
> :the loader, imho).  I rebooted, dropped to the bootloader command
> :prompt, set that variable to /dev/ad1s2b (my swap device) and rebooted.
> :
> :When I got to the debug prompt, I ran "panic" to cause a panic.  All I
> :got was the following 4 lines repeated over and over and over :-)
> :
> :panic: assertion: gd->gd_vme_base != NULL in vm_map_entry_kreserve
> :mp_lock = 00000003; cpuid = 3; lapci_id = 07000000
> :boot() called on cpu#3
> :Uptime: 16s
> :
> :The values change, as does the cpu#, but that's the error.
> :
> :Am I missing some step in getting this core file?
> :
> :Adam
>  
>     Hmm.  This is happening in early boot.  It must be just after init
>     is started.

>     Maybe we don't have enough zbootinit map entries on startup.  Try
>     increasing MAX_MAPENT in vm/vm_map.h from 256 to 512.  See if that
>     solves the problem or delays it long enough for the system to get
>     a crash dump.

Matt,

	So far so good.  It boots up, and it stayed up for about 30
minutes while I did a "make -j4 buildworld".  Unfortunately, the CPU
temperature rose above 60 degress Celcius, and the machine shut down :-)
I'll give it another go later, making sure it can stay up for at least
24 hours :-)

Adam