[its-hackers] TS NAME

Eric Swenson eric at swenson.org
Fri Nov 25 19:37:55 CET 2016


Lest anyone (besides me) is losing sleep on this, I did a bunch of 
debugging and it turns out that NAME was not failing on the CORBLK 
system calls (I was tricked by ending up in the error handling code when 
trying to step over the CORBLK calls).

After setting breakpoints after all CORBLK calls (normal and error 
returns), I discovered that we were failing much later in the program -- 
when trying to find a TTYTYP entry for this machine. Turns out we were 
missing the file SYSENG; TTYTYP > in the built system.  It was the OPEN 
call on this file that was returning the "file not found" error.

On 11/24/2016 04:14 PM, Eric Swenson wrote:
> I guess I forgot the important part. The error code, returned from 
> CORBLK was, I believe 35 (No such job).  But aren't we specifying our 
> job?  Or does there need be some daemon running to help with this 
> call?  -- Eric
>
>
> On 11/24/2016 03:51 PM, Eric Swenson wrote:
>> I'm trying to build/dump SYS; TS NAME on a new system. I do the 
>> following:
>>
>> name$j
>>
>> $l sysbin; name bin
>>
>> $g
>>
>> And I get:
>>
>> Initializing...
>>
>> .VAL 0; ERROR+3>>TLNE 400000 0/ SEKTTY+25,0
>>
>> I've also tried the following (setting the DEBUG flag from -1 to 0:
>>
>> name$j
>>
>> $l sysbin; name bin
>>
>> debug/ 0
>>
>> $g
>>
>> New sys, must initialize--
>>
>> Take paws off keys and wait
>>
>> :$ Urk! AAAAAAIIIIIEEEeeeeeee... $
>>
>> *:SL SYS; TS NAME
>>
>> *:PDUMP CRASH;NAME >
>>
>> $:BUG NAME A NAME crash was just dumped to CRASH; NAME >
>>
>> We're dying when we try to do a CORBLK system call.  It is the first 
>> SYSCAL CORBLK in the following code:
>>
>> corcpy:    syscal corblk,[cimm %cbndr+%cbndw ? cimm -1 ? cimm usrpag 
>> ? cimm %jsnew]
>>      jsr error
>>     hrlz c,a
>>     lsh c,10.    ;get addr in LH
>>     hrri c,usrfil
>>     blt c,usrfil+1777    ;copy page by hand.
>>     hrrz c,a
>>     syscal corblk,[B ? cimm -1 ? C    ;replace page with copied page.
>>         cimm -1 ? cimm usrpag]
>>      jsr error
>>     syscal corblk,[cimm 0 ? cimm -1 ? cimm usrpag] ;flush scratch page.
>>      jsr error
>>     aobjn a,corcpy    ;repeat as necessary.
>>     popj p,
>> 

>>
>> BVAR
>> errcod:    0
>> error:    0        ; JSR here when fatal error happens.
>>     skipe debug
>>      .value
>>     tlne f,%svrmod
>>      jsr exit    ;if server, just die for now.  Perhaps later report it.
>>
>>     ; We're going to dump.  Save some info.
>>     movem a,asave        ; save ac (don't smash pdl)
>>     .suset [.rbchn,,bchn]
>>     hrlz a,bchn
>>     add a,[.rios,,ios]
>>     .suset a
>>     move a,asave
>>     .value [asciz : Urk! AAAAAAIIIIIEEEEeeeeeeeeee... 
>> :SL SYS;TS NAME
>> :PDUMP CRASH;NAME >
>> :BUG NAME A NAME crash was just dumped to CRASH;NAME > !!
>> : Sorry about that.  If you have any idea why this crash
>> happened or how to reproduce it, please do another :bug name
>> and tell us...
>> ]
>>
>> Any idea where to look?  I've single-stepped through the execution, 
>> but don't know how to proceed when we are getting an error return 
>> from a CORBLK system call.  I do have the saved CRASH; NAME >.
>>
>




More information about the its-hackers mailing list