tomoaki/hakmem
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix C0/C7 class confusion: Upgrade C7 stride to 2048B and fix meta->class_idx initialization

Browse Source 
Root Cause:
1. C7 stride was 1024B, unable to serve 1024B user requests (need 1025B with header)
2. New SuperSlabs start with meta->class_idx=0 (mmap zero-init)
3. superslab_init_slab() only sets class_idx if meta->class_idx==255
4. Multiple code paths used conditional assignment (if class_idx==255), leaving C7 slabs with class_idx=0
5. This caused C7 blocks to be misidentified as C0, leading to HDR_META_MISMATCH errors

Changes:
1. Upgrade C7 stride: 1024B → 2048B (can now serve 1024B requests)
2. Update blocks_per_slab[7]: 64 → 32 (2048B stride / 64KB slab)
3. Update size-to-class LUT: entries 513-2048 now map to C7
4. Fix superslab_init_slab() fail-safe: only reinitialize if class_idx==255 (not 0)
5. Add explicit class_idx assignment in 6 initialization paths:
   - tiny_superslab_alloc.inc.h: superslab_refill() after init
   - hakmem_tiny_superslab.c: backend_shared after init (main path)
   - ss_unified_backend_box.c: unconditional assignment
   - ss_legacy_backend_box.c: explicit assignment
   - superslab_expansion_box.c: explicit assignment
   - ss_allocation_box.c: fail-safe condition fix

Fix P0 refill bug:
- Update obsolete array access after Phase 3d-B TLS SLL unification
- g_tls_sll_head[cls] → g_tls_sll[cls].head
- g_tls_sll_count[cls] → g_tls_sll[cls].count

Results:
- HDR_META_MISMATCH: eliminated (0 errors in 100K iterations)
- 1024B allocations now routed to C7 (Tiny fast path)
- NXT_MISALIGN warnings remain (legacy 1024B SuperSlabs, separate issue)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Moe Charm (CI)
2025-11-21 13:44:05 +09:00
parent 66a29783a4
commit a78224123e
11 changed files with 266 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

84
core/tiny_nextptr.h
View File

@ -16,8 +16,9 @@
// → next_off = 1
//
// Class 7:
// 大きなクラス、互換性と実装方針により next は base+0 扱い
// → next_off = 0
// [1B header][payload 2047B]
// → C7アップグレード後も header保持、next は base+1 に格納
// → next_off = 1
//
// HAKMEM_TINY_HEADER_CLASSIDX == 0 のとき:
//
@ -34,14 +35,22 @@
#include <stdint.h>
#include <string.h>
#include "hakmem_build_flags.h"
#include "tiny_region_id.h" // HEADER_MAGIC/HEADER_CLASS_MASK for header repair/logging
#include "hakmem_super_registry.h" // hak_super_lookup
#include "superslab/superslab_inline.h" // slab_index_for
#include <stdio.h>
#include <stdatomic.h>
#include <dlfcn.h>
#include <execinfo.h> // backtrace for rare misalign diagnostics
// Compute freelist next-pointer offset within a block for the given class.
static inline __attribute__((always_inline)) size_t tiny_next_off(int class_idx) {
#if HAKMEM_TINY_HEADER_CLASSIDX
// Phase E1-CORRECT finalized rule:
// Class 0,7 → offset 0
// Class 1-6 → offset 1
return (class_idx == 0 || class_idx == 7) ? 0u : 1u;
// Phase E1-CORRECT REVISED (C7 corruption fix):
// Class 0 → offset 0 (8B block、header後に8Bポインタは入らない)
// Class 1-7 → offset 1 (header保持、nextはheader直後)
// C7も header を保持して class 判別を壊さないことを優先
return (class_idx == 0) ? 0u : 1u;
#else
(void)class_idx;
return 0u;
@ -68,6 +77,69 @@ static inline __attribute__((always_inline)) void* tiny_next_load(const void* ba
static inline __attribute__((always_inline)) void tiny_next_store(void* base, int class_idx, void* next) {
size_t off = tiny_next_off(class_idx);
#if HAKMEM_TINY_HEADER_CLASSIDX
if (class_idx != 0) {
uint8_t expected = (uint8_t)(HEADER_MAGIC | (class_idx & HEADER_CLASS_MASK));
uint8_t got = *(uint8_t*)base;
if (__builtin_expect(got != expected, 0)) {
static _Atomic uint32_t g_next_hdr_diag = 0;
uint32_t n = atomic_fetch_add_explicit(&g_next_hdr_diag, 1, memory_order_relaxed);
if (n < 16) {
fprintf(stderr, "[NXT_HDR_MISMATCH] cls=%d base=%p got=0x%02x expect=0x%02x\n",
class_idx, base, got, expected);
}
}
*(uint8_t*)base = expected; // Always restore header before writing next
}
#endif
// Misalignment detector: class stride vs base offset
do {
static _Atomic uint32_t g_next_misalign_log = 0;
extern const size_t g_tiny_class_sizes[];
size_t stride = (class_idx >= 0 && class_idx < 8) ? g_tiny_class_sizes[class_idx] : 0;
if (stride > 0) {
uintptr_t delta = ((uintptr_t)base) % stride;
if (__builtin_expect(delta != 0, 0)) {
void* ra = __builtin_return_address(0);
const char* sym = "(unknown)";
#ifdef __GLIBC__
do {
Dl_info info;
if (dladdr(ra, &info) && info.dli_sname) {
sym = info.dli_sname;
}
} while (0);
#endif
uint32_t n = atomic_fetch_add_explicit(&g_next_misalign_log, 1, memory_order_relaxed);
int meta_cls = -1;
int slab_idx = -1;
struct SuperSlab* ss = NULL;
if (class_idx >= 0 && class_idx < 8) {
ss = hak_super_lookup(base);
if (ss) {
slab_idx = slab_index_for(ss, base);
if (slab_idx >= 0) {
struct TinySlabMeta* m = &ss->slabs[slab_idx];
meta_cls = m->class_idx;
}
}
}
if (n < 16) {
fprintf(stderr,
"[NXT_MISALIGN] cls=%d base=%p stride=%zu delta_mod=%zu next=%p ra=%p fn=%s meta_cls=%d slab_idx=%d ss=%p\n",
class_idx, base, stride, (size_t)delta, next, ra, sym, meta_cls, slab_idx, (void*)ss);
if (n < 4) {
void* bt[8];
int frames = backtrace(bt, 8);
backtrace_symbols_fd(bt, frames, fileno(stderr));
}
fflush(stderr);
}
}
}
} while (0);
if (off == 0) {
// Aligned access at base.
*(void**)base = next;