[Info-vax] Current expected range for x86-64 first boot date ?

[Mike Moroney]

On 12/4/2018 8:29 AM, clair.grant at vmssoftware.com wrote:
> Calling SCHED$INIT unleashes the beginnings of a much bigger environment, namely lowering IPL and changing modes. This brings into play much more of the newly-written x86-specific code, in particular page tables for the outer modes, switching modes, and creating a process. Here is recent (yesterday) output including temporary develop helpers.
> 
> %EXECINIT-I-ACTIVESET, joining ACTIVE set
> %EXECINIT-S-CIAO, transferring to the scheduler
> Lowering ipl to 3
> Acquiring SCHED lock at IPL 8
> Find next process to schedule; curktb = 0
> New ktb: 800a9718
> Find structures for new thread
> Mark cpu active
> Switch context
> SWPCTX   FFFFFFFF858B4620
> FNDHWPCB
> LOG IT
> SAVExSP
> SAVEUNQ
> SAVEASTx
> SAVERSP
> SAVEIDEP
> SWPCC
> LOADPTBR
>    NEWPCB FFFFFFFF800A8880
>       ASN 00000000000003FF
>      PTBR 0000000000EDC000
>       CR3 0000000000EDC3FF
> LOADRSP
> LOADASTx
> LOADHPCB
> LOADIDEP
> LOADUNQ
> LOADxSP
> LOG IT
> FIRSTUSE
> XRSTOR   FFFF8300103484C0
> CLRREGS
> FUNC
>      FUNC FFFFFFFF858B4620
> Never scheduled
> find structures again
> Release sched lock
> Lower ipl to ASTDEL
> newhwpcb: 800a8880
> sch$ar_swppcb(800a9718)->pcb$l_phd(800a8800)->phd$q_hwpcb(@800a8880)
> Enter swapper main loop
> %SWAPPER-I-SHUFFLE, executing SWAPPER initialization code
> %SWAPPER-I-POOL, initializing paged pool
> %SWAPPER-I-INIT, calling initialization routines
> %SWAPPER-I-RSDM, joining system resource domain
> %SWAPPER-I-P1, initializing SWAPPER P1 cells
> %SWAPPER-I-PAGED, paged system initialization
> %SWAPPER-I-HASH, creating the logical name hash tables
> %SWAPPER-I-DEFINE, defining the system logical names
>   
> Another significant factor is that we are now up far enough that a number of things can be debugged
> in parallel. A few weeks ago it was really a straight line of single stepping execution but now we
> usually have at least five paths of debugging going on without serious dependencies on one another.

I can say it goes a little further now. I got logical names mostly 
working now and now it goes as far as the $CREPRC of SYSINIT and returns 
to the main SWAPPER loop. The $CREPRC doesn't work (known issue) and the 
SWAPPER loop, well, isn't ready yet.

At least I am away from memory management for now. Fugly. How many 
levels of page tables? Variable levels? Add to it VMS-specific stuff and 
it is that much worse. And the instruction set/registers... it's obvious 
this evolved, it wasn't designed...

Also the further along we go the more code is recompiled existing code 
rather than x86 specific like the memory management.