Re: sun4v trap table

On Mon, 06 Oct 2025 17:34:42 +1100
matthew green <[email protected]> wrote:

> > For sparc64 in locore.s below is the fragment from the sun4v trap table for TL (Trap Level) 0:
> >
> >         SPILL64(uspill8_sun4vt0,ASI_AIUS)                       ! 0x080 spill_0_normal -- used to save user windows in user mode
> >         SPILL32(uspill4_sun4vt0,ASI_AIUS)                       ! 0x084 spill_1_normal
> >         SPILLBOTH(uspill8_sun4vt0,uspill4_sun4vt0,ASI_AIUS)     ! 0x088 spill_2_normal
> >         sun4v_trap_entry_spill_fill_fail 1                      ! 0x08c spill_3_normal
> >         SPILL64(kspill8_sun4vt0,ASI_N)                          ! 0x090 spill_4_normal  -- used to save supervisor windows
> >         SPILL32(kspill4_sun4vt0,ASI_N)                          ! 0x094 spill_5_normal
> >         SPILLBOTH(kspill8_sun4vt0,kspill4_sun4vt0,ASI_N)        ! 0x098 spill_6_normal
> >         sun4v_trap_entry_spill_fill_fail 1                      ! 0x09c spill_7_normal
> >
> > Does anyone know how the TT (Trap Type) maps to the table entries? For
> > example, when TT=0x080 why does the kernel think it needs to use SPILL64()
> > and for example, not SPILL32(). Where is this defined?
> >
> > The architecture manual does not seem to specify the details, so I assume
> > it is left up to the OS kernel to decide? Is this correct?
> 
> i thought it was more obvious but it's been almost 30 years since
> i wrote some code related to this that eeh implemented as well in
> a far more complete form, that also worked :)
> 
> the traptype is composed of 9 bits:
> 
>    8:6 are 010 for spill and 011 for fill.
>    5 is if OTHERWIN is not 0.
>    4:2 are either WSTATE.OTHER or WSTATE.NORMAL.
> 
> (1:0 is always 0 because these are 4-trap long traps.)
> 
> in the (original) v9 manual this is in 7.5.2.1
> "Trap Type for Spill/Fill Traps".
> 
> so the control is on platform setting %WSTATE -- which is 3 bits for
> "other" and "normal".  we only use two values for this so, really
> only two of the above should be used?  it's been a while, and i'm
> still a little confused by:
> 
> include/psl.h:264:#define WSTATE_USER    022
> include/psl.h:263:#define WSTATE_KERN    026
> 
> since they both seem to set OTHER bits, which means that 4:2 in the
> trap type will always be 2?  so always uses SPILLBOTH()?  i think i
> am still missing something here...
> 
> 
> (none of the above is sun4v specific but sparc v9 generic.)
> 
> 
> .mrg.


Window State (WSTATE) register:
+-----------+
|Other|Norml|
+-----------+
 5 4 3 2 1 0


Trap Type (TT) register:
+-----+-+-----+---+
| S/F |O|WType|0 0|
+-----+-+-----+---+
 8 7 6 5 4 3 2 1 0

S/F  : Spill (010) or Fill (011)
O    : Other (0 if OTHERWIN=0, 1 if OTHERWIN>0)
WType: Window type (if Other==0 then WSTATE.NORMAL else WSTATE.OTHER)


NetBSD src/sys/arch/sparc64/include/psl.h
263:#define WSTATE_KERN 026
264:#define WSTATE_USER 022

So in binary:
WSTATE_KERN is 010 110
WSTATE_USER is 010 010

TT Hex  TT Bin        Spill/Fill  Window State
----------------------------------------------------------
0x098   010 0 110 00      S       OTHERWIN=0 and WSTATE_KERN
0x0A8   010 1 010 00      S       OTHERWIN>0 and WSTATE_KERN
0x088   010 0 010 00      S       OTHERWIN=0 and WSTATE_USER
0x0A8   010 1 010 00      S       OTHERWIN>0 and WSTATE_USER

0x0D8   011 0 110 00      F       OTHERWIN=0 and WSTATE_KERN
0x0E8   011 1 010 00      F       OTHERWIN>0 and WSTATE_KERN
0x0C8   011 0 010 00      F       OTHERWIN=0 and WSTATE_USER
0x0E8   011 1 010 00      F       OTHERWIN>0 and WSTATE_USER
----------------------------------------------------------

So when OTHERWIN>0 both WSTATE_KERN and WSTATE_USER map to the same
TTs: 0x0A8 for spill TTs and 0x0E8 for fill TTs.

Is it supposed to work like this, or OTHERWIN>0 is never set/used in
NetBSD?

I thought OTHERWIN was only used for microkernel type OS, which may
implement client/server calls between isolated modules and may need to
partition window registers between different address spaces.