starkernel: item 4.2 -- Hermes native on the Stadium (complete)

Migrates Hermes's message/channel lifecycle onto the Stadium's unified
heat/capacity economy: MSG-ALLOC/FREE-NODE and CH-ALLOC/FREE-NODE now
route entirely through stadium_admit()/stadium_evict(), replacing the
old local free-list + independent heat-field mechanism. Eight
kernel-only STADIUM-* FORTH primitives (ADMIT, EVICT, RES@, RES-PULL,
RES-PUSH, HEAT@, HEAT!, WORD-HEAT), VM.stadium_vm_id threaded through
all three vm_core.c dispatch sites (replacing item 4.1's hardcoded
vm_uuid_hera()), and the stadium_owner[idx] fix so evict-credit lands
in the VM that actually admitted a patron, not whoever owned cell 0.

This session's own contribution, on top of that pre-existing
implementation: found and fixed two bugs blocking the item's own K≡1.0
conservation self-check (HERMES-K was reading 0, not 65536):

- Q.SLOT admission-heat fix (capsules/hermes/init.4th): MSG-SEND/
  CH-ACCEPT admitted with Q.1 (the entire fleet-wide "1.0" unit) per
  item, a leftover from before the Stadium migration when each
  message/channel had its own unconstrained heat field. Instantly
  drained the shared, finite reservoir.

- Reservoir floor for word-execution admission (stadium_words.c):
  stadium_word_dispatch() (item 4.1) pulls STADIUM_WORD_HEAT_QUANTUM on
  every word dispatch, not just first admission -- exhausts a VM's
  entire reservoir in ~32 dispatches, starving any application-level
  economy sharing that VM's reservoir before it gets a chance to pull
  anything. word_dispatch_pull() now clamps word-execution's own pulls
  to leave a Q48_ONE/3 floor (same fair-share figure COMMON-CH's own
  floor already uses); application-level pulls are unaffected.

- STADIUM-WORD-HEAT primitive + stadium_words_resident_heat(): the
  floor deliberately leaves word-execution residents holding real
  heat, invisible to HERMES-K's original formula (MSG+CH+reservoir,
  no term for word patrons). Adding this term closes K to exactly
  65536 on all three architectures.

Also rules on two open scope questions in FABRIC.md: MBR-ALLOC/
MBR-FREE-NODE stay off the Stadium (membership records have no heat
field, never did -- the acceptance bullet's inclusion of them was a
completeness gesture predating a check of the actual layout), and
records the effort number (12 implementation files, +759/-120 lines).

Verified: all three architectures boot clean, full self-test passes,
Stadium conservation closes exactly (resident_sum + reservoir =
Q48_ONE) at both the C/Stadium level and the FORTH-level HERMES-K
check.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
Robert Allan James
2026-08-07 01:49:23 -04:00
co-authored by Claude Sonnet 5
parent 0a7f144367
commit 5a28458b21
19 changed files with 1034 additions and 162 deletions
+204 -5
View File
@@ -44,6 +44,8 @@
#include "starkernel/capsule_loader.h"
#include "starkernel/capsule_vm_physics.h"
#include "starkernel/vm/vm_internal.h"
#include "starkernel/vm/stadium.h" /* item 4.2 -- STADIUM-* words */
#include "starkernel/vm/stadium_words.h" /* item 4.2 -- stadium_words_resident_heat() */
#include "starkernel/repl.h"
#include "starkernel/capsule_generated.h"
#include "starkernel/console.h"
@@ -1156,16 +1158,205 @@ void register_mama_forth_words(VM *vm)
vocabulary_word_definitions(vm);
}
/* ============================================================================
* Stadium Words (FABRIC.md punch list item 4.2)
*
* The entire C surface item 4.2 is permitted to add, per HERMES.md's
* language constraint: all eight operate on the CALLING VM's own identity
* (vm->stadium_vm_id) implicitly, never a FORTH-supplied vm-id. A
* stack-passed vm-id could only ever be the caller's own (redundant) or
* another VM's (stadium_admit()/stadium_evict() would refuse it via quota
* isolation, except STADIUM-RES-PUSH, which has no such guard and would be
* an outright heat-forgery primitive against another VM's reservoir).
* Conservation is the invariant this item is verified against, so implicit
* self is not an optimization -- it's the only version that can't break it.
*
* STADIUM-ADMIT does not itself pull the candidate's heat from the
* reservoir -- that's STADIUM-RES-PULL's job, composed in FORTH by the
* caller (e.g. a rewritten MSG-ALLOC): pull first, admit with the pulled
* amount, and STADIUM-RES-PUSH it back if admission refuses. Mirrors
* stadium_words.c's C-side Option B starter-grant pattern, but the
* composition itself lives in StarForth, not here, per HERMES.md.
* ============================================================================ */
/**
* @brief STADIUM-ADMIT ( identity heat behaviour -- cell | -1 )
* Admits a mass-1 patron into the calling VM's own Stadium quota.
* `behaviour` must be a valid StadiumBehaviour tag (0..3); anything else
* refuses without calling stadium_admit() at all. `contains` is always
* explicitly STADIUM_CONTAINS_NONE -- stadium_admit()'s own doc warns a
* zero-initialized `contains` reads as "contains Hera" (index 0) and
* permanently blocks eviction, so this is never left to a zero-fill.
*/
static void mama_word_stadium_admit(VM *vm)
{
cell_t behaviour_cell, heat_cell, identity_cell;
StadiumPatronHeader candidate;
uint8_t *raw = (uint8_t *)&candidate;
size_t i;
size_t idx;
if (vm->dsp < 2) { vm->error = 1; return; }
behaviour_cell = vm_pop(vm);
heat_cell = vm_pop(vm);
identity_cell = vm_pop(vm);
if (behaviour_cell < STADIUM_BEHAVIOUR_MIGRATE || behaviour_cell > STADIUM_BEHAVIOUR_COOL) {
vm_push(vm, (cell_t)-1);
return;
}
for (i = 0; i < sizeof(candidate); i++) raw[i] = 0;
candidate.identity = (uint64_t)identity_cell;
candidate.heat = (uint64_t)heat_cell;
candidate.ttl = 0;
candidate.link = 0;
candidate.contains = STADIUM_CONTAINS_NONE;
candidate.mass = 1;
candidate.flags = 0;
candidate.behaviour = (uint8_t)behaviour_cell;
idx = stadium_admit(vm->stadium_vm_id, &candidate);
vm_push(vm, (idx == STADIUM_CELL_NONE) ? (cell_t)-1 : (cell_t)idx);
}
/**
* @brief STADIUM-EVICT ( cell -- flag )
* Reaps the patron header at `cell`. flag is FORTH true (-1) on success,
* false (0) if refused (out of range, not resident, pinned, or contains-
* gated) -- stadium_evict()'s own refusal set, unchanged here.
*/
static void mama_word_stadium_evict(VM *vm)
{
cell_t cell_cell;
if (vm->dsp < 0) { vm->error = 1; return; }
cell_cell = vm_pop(vm);
if (cell_cell < 0) {
vm_push(vm, (cell_t)0);
return;
}
vm_push(vm, (stadium_evict((size_t)cell_cell) == 0) ? (cell_t)-1 : (cell_t)0);
}
/**
* @brief STADIUM-RES@ ( -- heat )
* Read-only peek at the calling VM's own reservoir balance (Q48.16).
*/
static void mama_word_stadium_res_fetch(VM *vm)
{
vm_push(vm, (cell_t)stadium_reservoir_peek(vm->stadium_vm_id));
}
/**
* @brief STADIUM-WORD-HEAT ( -- heat )
* Sum of heat held by the calling VM's own word-execution residents
* (item 4.1's cells) -- the term a VM's own application-level conservation
* check (e.g. Hermes's HERMES-K) needs to close exactly, since word patrons
* are otherwise invisible to FORTH (FABRIC.md §25.7, ruling 2026-08-06).
*/
static void mama_word_stadium_word_heat(VM *vm)
{
vm_push(vm, (cell_t)stadium_words_resident_heat(vm->stadium_vm_id));
}
/**
* @brief STADIUM-RES-PULL ( qty -- heat )
* Pulls up to `qty` (Q48.16) from the calling VM's own reservoir. Returns
* the amount actually pulled, which may be less than requested -- never
* negative, never invents heat, mirrors stadium_reservoir_pull()'s own
* clamping exactly.
*/
static void mama_word_stadium_res_pull(VM *vm)
{
cell_t qty_cell;
if (vm->dsp < 0) { vm->error = 1; return; }
qty_cell = vm_pop(vm);
if (qty_cell < 0) {
vm_push(vm, (cell_t)0);
return;
}
vm_push(vm, (cell_t)stadium_reservoir_pull(vm->stadium_vm_id, (uint64_t)qty_cell));
}
/**
* @brief STADIUM-RES-PUSH ( heat -- )
* Credits `heat` (Q48.16) back into the calling VM's own reservoir -- the
* other half of every reservoir transfer (cooling, refused-admission
* rollback, or a departing patron's remaining heat after eviction).
*/
static void mama_word_stadium_res_push(VM *vm)
{
cell_t heat_cell;
if (vm->dsp < 0) { vm->error = 1; return; }
heat_cell = vm_pop(vm);
if (heat_cell < 0) return;
stadium_reservoir_push(vm->stadium_vm_id, (uint64_t)heat_cell);
}
/**
* @brief STADIUM-HEAT@ ( cell -- heat )
* Reads a resident cell's own heat. Requires the cell to be resident and
* owned by the calling VM's own quota -- returns 0 otherwise (out of range,
* not resident, or belongs to a different VM).
*/
static void mama_word_stadium_heat_fetch(VM *vm)
{
cell_t cell_cell;
if (vm->dsp < 0) { vm->error = 1; return; }
cell_cell = vm_pop(vm);
if (cell_cell < 0) {
vm_push(vm, (cell_t)0);
return;
}
vm_push(vm, (cell_t)stadium_cell_heat_get(vm->stadium_vm_id, (size_t)cell_cell));
}
/**
* @brief STADIUM-HEAT! ( new-heat cell -- )
* Writes a resident cell's own heat, reconciling the reservoir delta
* atomically in C (pulls on an increase, refusing silently if the
* calling VM's reservoir can't cover it; pushes back on a decrease).
* Requires the cell to be resident and owned by the calling VM's own
* quota -- silently refused otherwise, same as every other write here.
*/
static void mama_word_stadium_heat_store(VM *vm)
{
cell_t cell_cell, new_heat_cell;
if (vm->dsp < 1) { vm->error = 1; return; }
cell_cell = vm_pop(vm);
new_heat_cell = vm_pop(vm);
if (cell_cell < 0 || new_heat_cell < 0) return;
(void)stadium_cell_heat_set(vm->stadium_vm_id, (size_t)cell_cell, (uint64_t)new_heat_cell);
}
/**
* register_child_vm_words - Register the minimal word set needed by child VMs.
*
* Child VMs are not bootstrapped through sk_vm_bootstrap_parity, so they
* do not get register_mama_forth_words(). They only need STOP (self-halt)
* and EXEC (load a capsule). Keeping the registrations here — in the same
* translation unit as the word functions — avoids cross-TU function-pointer
* loads that produce R_X86_64_REX_GOTPCRELX relocations; those are not
* relaxed by the PE32+ linker, causing the function code bytes to be read
* as the pointer value instead of the actual address.
* and EXEC (load a capsule) -- plus, as of item 4.2, the eight STADIUM-*
* primitives Hermes needs to migrate her message/channel lifecycle onto the
* Stadium. Keeping the registrations here — in the same translation unit
* as the word functions — avoids cross-TU function-pointer loads that
* produce R_X86_64_REX_GOTPCRELX relocations; those are not relaxed by the
* PE32+ linker, causing the function code bytes to be read as the pointer
* value instead of the actual address.
*
* Deliberately NOT added to register_mama_forth_words(): that would put
* these words in Hera's own dictionary too and move dict_hash off item
* 4.1's baseline (0x3d4e1daf289da94f) -- a deliberate baseline change to
* state this item does not make as a side effect.
*/
void register_child_vm_words(VM *vm)
{
@@ -1173,6 +1364,14 @@ void register_child_vm_words(VM *vm)
register_word(vm, "EXEC", mama_word_exec);
register_word(vm, "VM-EXEC", mama_word_vm_exec);
register_word(vm, "VM-CALL", mama_word_vm_call);
register_word(vm, "STADIUM-ADMIT", mama_word_stadium_admit);
register_word(vm, "STADIUM-EVICT", mama_word_stadium_evict);
register_word(vm, "STADIUM-RES@", mama_word_stadium_res_fetch);
register_word(vm, "STADIUM-RES-PULL", mama_word_stadium_res_pull);
register_word(vm, "STADIUM-RES-PUSH", mama_word_stadium_res_push);
register_word(vm, "STADIUM-HEAT@", mama_word_stadium_heat_fetch);
register_word(vm, "STADIUM-HEAT!", mama_word_stadium_heat_store);
register_word(vm, "STADIUM-WORD-HEAT", mama_word_stadium_word_heat);
}
#endif /* __STARKERNEL__ */