Destroy a movie decoder's system events and free their table entries when the decoder goes.

The review of the system-event change (Ev-1) found the table still filling. Entries in host memory now count as live
for good in the full-table scan, but their objects are freed: ov_movie2.c ov_dtor freed the Video with in_ev and out_ev
still bound, so each decoder left two dead entries that nothing reclaimed (AUDIT F-1-08 measured about two entries per
reload of a save, which E-4's verifier tied to the decoder pair), and their four kernel handles stayed open. A long
session would end with "kern_bind_system_event: table full" once no dead guest entry was left to take.

kernel.c gains kern_unbind_system_event(c, ev). On a guest thread it calls the SDK's nn::os::DestroySystemEvent, which
zeroes the state byte and closes both handles, and then removes the entry under g_se_lock (the last entry takes its
place). The destroy comes first, so a signal still in flight finds the event destroyed or its handle no longer naming it
and skips it, rather than finding no entry and binding again on a host thread. Off a guest thread, or when the SDK lacks
the function (logged once), only the entry goes. ov_dtor calls it for each bound event after finalize has joined the
decoder and convert threads, then frees the Video. net_offline.c frees its HostTask with a bound event in the same way;
that file is outside this change and is handed to the coordinator.

Not reproduced before the change in a game run. tests/system_event_check.c fills the table with live guest events and
the movie's two host-memory events, unbinds one: it leaves the table, its handles are closed and its state byte is 0,
the other keeps its entry, and a second unbind changes nothing; with the table full again and the last host entry
unbound, a new guest bind succeeds where it would have ended with "table full". The check passes in both modes,
movie_planes_check passes, and the compile check of every runtime file (host.h changed) passes, adding no warnings in
kernel.c or ov_movie2.c with -Wall -Wextra.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
chrg6
2026-11-13 17:56:39 -05:00
parent 601a954676
commit d8c046acc4
5 changed files with 98 additions and 16 deletions

View File

@@ -2158,8 +2158,10 @@ void kern_event_clear(uint32_t hw) {
// Each entry also keeps the serial number of the Event its handle named at bind time (h_new hands a closed number to
// the next new object at once, so the number alone does not say it is still that event), and whether the object lies
// in guest memory: host services keep some in their own allocations (ov_movie2.c Video.in_ev/out_ev, net_offline.c
// HostTask.event), which mem_readable does not know and the SDK never destroys.
// Lock order: G before g_se_lock (system_event_set); kern_bind_system_event never holds both.
// HostTask.event), which mem_readable does not know and the game never destroys. A service that frees such an object
// gives its entry back first (kern_unbind_system_event, from ov_movie2.c ov_dtor).
// Lock order: G before g_se_lock (system_event_set); kern_bind_system_event and kern_unbind_system_event never hold
// both.
#define SE_STATE 0x28
static struct { uint64_t ev, serial; uint32_t wh; uint8_t guest; } g_se[512];
static int g_nse;
@@ -2205,10 +2207,11 @@ void kern_bind_system_event(Cpu *c, uint64_t ev, int autoclear) {
int i = 0;
while (i < g_nse && g_se[i].ev != ev) i++;
if (i == g_nse && g_nse == (int)(sizeof g_se / sizeof *g_se)) {
// Full: entries are never removed, and a long session binds events at new addresses (two per reload of a
// save). An entry in guest memory whose event the guest destroyed (its state byte is 0, or its memory is gone)
// gives its slot. Entries in host memory count as live: their services never destroy them through the SDK, and
// taking a live one would make the service's next signal bind again on a host thread, which is fatal.
// Full: entries in guest memory are never removed, and a long session binds events at new addresses (two per
// reload of a save). An entry in guest memory whose event the guest destroyed (its state byte is 0, or its
// memory is gone) gives its slot. Entries in host memory count as live: the game never destroys them, their
// services remove them before freeing them (kern_unbind_system_event), and taking a live one would make the
// service's next signal bind again on a host thread, which is fatal.
i = 0;
while (i < g_nse && (!g_se[i].guest || (mem_readable(g_se[i].ev + SE_STATE, 1) && se_live(g_se[i].ev)))) i++;
if (i == g_nse) rt_fatal("kern_bind_system_event: table full (%d live system events)", g_nse);
@@ -2259,6 +2262,34 @@ void kern_clear_system_event(uint64_t ev) {
if (se_find(ev, NULL) && se_live(ev)) system_event_set(ev, 0);
}
// For a service about to free a SystemEventType it bound in its own memory (ov_movie2.c ov_dtor), with nothing left to
// signal it: on a guest thread the SDK's DestroySystemEvent zeroes the state byte and closes both kernel handles, which
// would otherwise stay open for the rest of the run; then the table entry goes, and its slot is free for the next bind.
// The destroy comes first, so a signal still in flight finds the event destroyed (kern_signal_system_event) or its
// handle no longer naming it (system_event_set) and skips it, rather than finding no entry and binding again. Off a
// guest thread, or when the SDK lacks the function, only the entry goes.
void kern_unbind_system_event(Cpu *c, uint64_t ev) {
if (!se_find(ev, NULL)) return;
if (c && tcur && se_live(ev)) {
static uint64_t destroy; // 1: looked up and not loaded
uint64_t fn = __atomic_load_n(&destroy, __ATOMIC_RELAXED);
if (!fn) {
fn = ldr_symbol("_ZN2nn2os18DestroySystemEventEPNS0_15SystemEventTypeE");
if (!fn)
rt_log("kernel: the SDK's DestroySystemEvent is not loaded: unbound system events keep their handles");
__atomic_store_n(&destroy, fn = fn ? fn : 1, __ATOMIC_RELAXED);
}
if (fn > 1) rt_call(c, fn, 1, ev);
}
plat_mutex_lock(&g_se_lock);
for (int i = 0; i < g_nse; i++)
if (g_se[i].ev == ev) {
g_se[i] = g_se[--g_nse];
break;
}
plat_mutex_unlock(&g_se_lock);
}
// A SystemEventType that still holds an event: mapped guest memory with its state byte nonzero. For services that keep
// their own lists of bound events and reuse the slots of destroyed ones (gfx_present.c gfx_vsync_bind).
int kern_system_event_live(uint64_t ev) { return mem_readable(ev + SE_STATE, 1) && se_live(ev); }