## Summary
Implemented Phase 12 Shared SuperSlab Pool (mimalloc-style) to address
SuperSlab allocation churn (877 SuperSlabs → 100-200 target).
## Implementation (ChatGPT + Claude)
1. **Metadata changes** (superslab_types.h):
- Added class_idx to TinySlabMeta (per-slab dynamic class)
- Removed size_class from SuperSlab (no longer per-SuperSlab)
- Changed owner_tid (16-bit) → owner_tid_low (8-bit)
2. **Shared Pool** (hakmem_shared_pool.{h,c}):
- Global pool shared by all size classes
- shared_pool_acquire_slab() - Get free slab for class_idx
- shared_pool_release_slab() - Return slab when empty
- Per-class hints for fast path optimization
3. **Integration** (23 files modified):
- Updated all ss->size_class → meta->class_idx
- Updated all meta->owner_tid → meta->owner_tid_low
- superslab_refill() now uses shared pool
- Free path releases empty slabs back to pool
4. **Build system** (Makefile):
- Added hakmem_shared_pool.o to OBJS_BASE and TINY_BENCH_OBJS_BASE
## Status: ⚠️ Build OK, Runtime CRASH
**Build**: ✅ SUCCESS
- All 23 files compile without errors
- Only warnings: superslab_allocate type mismatch (legacy code)
**Runtime**: ❌ SEGFAULT
- Crash location: sll_refill_small_from_ss()
- Exit code: 139 (SIGSEGV)
- Test case: ./bench_random_mixed_hakmem 1000 256 42
## Known Issues
1. **SEGFAULT in refill path** - Likely shared_pool_acquire_slab() issue
2. **Legacy superslab_allocate()** still exists (type mismatch warning)
3. **Remaining TODOs** from design doc:
- SuperSlab physical layout integration
- slab_handle.h cleanup
- Remove old per-class head implementation
## Next Steps
1. Debug SEGFAULT (gdb backtrace shows sll_refill_small_from_ss)
2. Fix shared_pool_acquire_slab() or superslab_init_slab()
3. Basic functionality test (1K → 100K iterations)
4. Measure SuperSlab count reduction (877 → 100-200)
5. Performance benchmark (+650-860% expected)
## Files Changed (25 files)
core/box/free_local_box.c
core/box/free_remote_box.c
core/box/front_gate_classifier.c
core/hakmem_super_registry.c
core/hakmem_tiny.c
core/hakmem_tiny_bg_spill.c
core/hakmem_tiny_free.inc
core/hakmem_tiny_lifecycle.inc
core/hakmem_tiny_magazine.c
core/hakmem_tiny_query.c
core/hakmem_tiny_refill.inc.h
core/hakmem_tiny_superslab.c
core/hakmem_tiny_superslab.h
core/hakmem_tiny_tls_ops.h
core/slab_handle.h
core/superslab/superslab_inline.h
core/superslab/superslab_types.h
core/tiny_debug.h
core/tiny_free_fast.inc.h
core/tiny_free_magazine.inc.h
core/tiny_remote.c
core/tiny_superslab_alloc.inc.h
core/tiny_superslab_free.inc.h
Makefile
## New Files (3 files)
PHASE12_SHARED_SUPERSLAB_POOL_DESIGN.md
core/hakmem_shared_pool.c
core/hakmem_shared_pool.h
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: ChatGPT <chatgpt@openai.com>
239 lines
7.5 KiB
C
239 lines
7.5 KiB
C
// front_gate_classifier.c - Box FG: Pointer Classification Implementation
|
|
|
|
// CRITICAL: Box FG requires header-based classification
|
|
// Ensure HEADER_MAGIC and HEADER_CLASS_MASK are available
|
|
#ifndef HAKMEM_TINY_HEADER_CLASSIDX
|
|
#define HAKMEM_TINY_HEADER_CLASSIDX 1
|
|
#endif
|
|
|
|
#include <stdio.h> // For fprintf in debug
|
|
#include <stdlib.h> // For abort in debug
|
|
#include "front_gate_classifier.h"
|
|
#include "../tiny_region_id.h" // Must come before hakmem_tiny_superslab.h for HEADER_MAGIC
|
|
#include "../hakmem_tiny_superslab.h"
|
|
#include "../superslab/superslab_inline.h" // For ss_slabs_capacity
|
|
#include "../hakmem_build_flags.h"
|
|
#include "../hakmem_internal.h" // AllocHeader, HAKMEM_MAGIC, HEADER_SIZE, hak_is_memory_readable
|
|
#include "../hakmem_tiny_config.h" // For TINY_NUM_CLASSES, SLAB_SIZE
|
|
#include "../hakmem_super_registry.h" // For hak_super_lookup (Box REG)
|
|
|
|
#ifdef HAKMEM_POOL_TLS_PHASE1
|
|
#include "../pool_tls_registry.h" // Safer pool pointer lookup (no header deref)
|
|
#endif
|
|
|
|
// ========== Debug Stats ==========
|
|
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
__thread uint64_t g_classify_header_hit = 0;
|
|
__thread uint64_t g_classify_headerless_hit = 0;
|
|
__thread uint64_t g_classify_pool_hit = 0;
|
|
__thread uint64_t g_classify_unknown_hit = 0;
|
|
|
|
void front_gate_print_stats(void) {
|
|
uint64_t total = g_classify_header_hit + g_classify_headerless_hit +
|
|
g_classify_pool_hit + g_classify_unknown_hit;
|
|
if (total == 0) return;
|
|
|
|
fprintf(stderr, "\n========== Front Gate Classification Stats ==========\n");
|
|
fprintf(stderr, "Header (C0-C6): %lu (%.2f%%)\n",
|
|
g_classify_header_hit, 100.0 * g_classify_header_hit / total);
|
|
fprintf(stderr, "Headerless (C7): %lu (%.2f%%)\n",
|
|
g_classify_headerless_hit, 100.0 * g_classify_headerless_hit / total);
|
|
fprintf(stderr, "Pool TLS: %lu (%.2f%%)\n",
|
|
g_classify_pool_hit, 100.0 * g_classify_pool_hit / total);
|
|
fprintf(stderr, "Unknown: %lu (%.2f%%)\n",
|
|
g_classify_unknown_hit, 100.0 * g_classify_unknown_hit / total);
|
|
fprintf(stderr, "Total: %lu\n", total);
|
|
fprintf(stderr, "======================================================\n");
|
|
}
|
|
|
|
static void __attribute__((destructor)) front_gate_stats_destructor(void) {
|
|
front_gate_print_stats();
|
|
}
|
|
#endif
|
|
|
|
// ========== Safe Header Probe ==========
|
|
|
|
// Try to read 1-byte header at ptr-1 (safe conditions only)
|
|
// Returns: class_idx (0-7) on success, -1 on failure
|
|
//
|
|
// Safety conditions:
|
|
// 1. Same page: (ptr & 0xFFF) >= 1 → header won't cross page boundary
|
|
// 2. Valid magic: (header & 0xF0) == HEADER_MAGIC (0xa0)
|
|
// 3. Valid class: class_idx in range [0, 7]
|
|
//
|
|
// Performance: 2-3 cycles (L1 cache hit)
|
|
static inline int safe_header_probe(void* ptr) {
|
|
// Reject obviously invalid/sentinel-sized pointers (defense-in-depth)
|
|
if ((uintptr_t)ptr < 4096) {
|
|
return -1;
|
|
}
|
|
// Safety check: header must be in same page as ptr
|
|
uintptr_t offset_in_page = (uintptr_t)ptr & 0xFFF;
|
|
if (offset_in_page == 0) {
|
|
// ptr is page-aligned → header would be on previous page (unsafe)
|
|
return -1;
|
|
}
|
|
|
|
// Safe to read header (same page guaranteed)
|
|
uint8_t* header_ptr = (uint8_t*)ptr - 1;
|
|
uint8_t header = *header_ptr;
|
|
|
|
// Validate magic
|
|
if ((header & 0xF0) != HEADER_MAGIC) {
|
|
return -1; // Not a Tiny header
|
|
}
|
|
|
|
// Extract class index
|
|
int class_idx = header & HEADER_CLASS_MASK;
|
|
|
|
// Phase E1-CORRECT: Validate class range (all classes 0-7 valid)
|
|
if (class_idx < 0 || class_idx >= TINY_NUM_CLASSES) {
|
|
return -1; // Invalid class
|
|
}
|
|
|
|
return class_idx;
|
|
}
|
|
|
|
// ========== Registry Lookup ==========
|
|
|
|
// Lookup pointer in SuperSlab registry (fallback when header probe fails)
|
|
// Returns: classification result with SuperSlab + class_idx + slab_idx
|
|
//
|
|
// Performance: 50-100 cycles (hash lookup + validation)
|
|
static inline ptr_classification_t registry_lookup(void* ptr) {
|
|
ptr_classification_t result = {
|
|
.kind = PTR_KIND_UNKNOWN,
|
|
.class_idx = -1,
|
|
.ss = NULL,
|
|
.slab_idx = -1
|
|
};
|
|
|
|
// Query SuperSlab registry
|
|
struct SuperSlab* ss = hak_super_lookup(ptr);
|
|
if (!ss || ss->magic != SUPERSLAB_MAGIC) {
|
|
// Not in Tiny registry
|
|
return result;
|
|
}
|
|
|
|
// Found SuperSlab - determine slab index from ptr-1 (block base)
|
|
result.ss = ss;
|
|
|
|
uintptr_t ptr_addr = (uintptr_t)ptr;
|
|
uintptr_t ss_addr = (uintptr_t)ss;
|
|
if (ptr_addr <= ss_addr) {
|
|
result.kind = PTR_KIND_UNKNOWN;
|
|
return result;
|
|
}
|
|
|
|
// Use block base for slab index to be consistent with free paths
|
|
uintptr_t base_addr = ptr_addr - 1;
|
|
size_t offset = base_addr - ss_addr;
|
|
int slab_idx = (int)(offset / SLAB_SIZE);
|
|
if (slab_idx < 0 || slab_idx >= ss_slabs_capacity(ss)) {
|
|
result.kind = PTR_KIND_UNKNOWN;
|
|
return result;
|
|
}
|
|
|
|
result.slab_idx = slab_idx;
|
|
TinySlabMeta* meta = &ss->slabs[slab_idx];
|
|
int cls = (meta->class_idx < TINY_NUM_CLASSES) ? (int)meta->class_idx : -1;
|
|
result.class_idx = cls;
|
|
|
|
if (cls == 7) {
|
|
// 1KB headerless tiny
|
|
result.kind = PTR_KIND_TINY_HEADERLESS;
|
|
} else if (cls >= 0) {
|
|
// Other tiny classes with 1-byte header
|
|
result.kind = PTR_KIND_TINY_HEADER;
|
|
} else {
|
|
result.kind = PTR_KIND_UNKNOWN;
|
|
}
|
|
|
|
return result;
|
|
}
|
|
|
|
// ========== Pool TLS Probe ==========
|
|
|
|
#ifdef HAKMEM_POOL_TLS_PHASE1
|
|
// Registry-based Pool TLS probe (no memory deref)
|
|
static inline int is_pool_tls_reg(void* ptr) {
|
|
pid_t tid = 0; int cls = -1;
|
|
return pool_reg_lookup(ptr, &tid, &cls);
|
|
}
|
|
#endif
|
|
|
|
// ========== Front Gate Entry Point ==========
|
|
|
|
ptr_classification_t classify_ptr(void* ptr) {
|
|
ptr_classification_t result = {
|
|
.kind = PTR_KIND_UNKNOWN,
|
|
.class_idx = -1,
|
|
.ss = NULL,
|
|
.slab_idx = -1
|
|
};
|
|
|
|
if (!ptr) return result;
|
|
// Early guard: reject non-canonical tiny integers to avoid ptr-1 probe crashes
|
|
if ((uintptr_t)ptr < 4096) {
|
|
result.kind = PTR_KIND_UNKNOWN;
|
|
return result;
|
|
}
|
|
|
|
// Step 1: Check Pool TLS via registry (no pointer deref)
|
|
#ifdef HAKMEM_POOL_TLS_PHASE1
|
|
if (is_pool_tls_reg(ptr)) {
|
|
result.kind = PTR_KIND_POOL_TLS;
|
|
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
g_classify_pool_hit++;
|
|
#endif
|
|
return result;
|
|
}
|
|
#endif
|
|
|
|
// Step 2: Registry lookup for Tiny (header or headerless)
|
|
result = registry_lookup(ptr);
|
|
if (result.kind == PTR_KIND_TINY_HEADERLESS) {
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
g_classify_headerless_hit++;
|
|
#endif
|
|
return result;
|
|
}
|
|
if (result.kind == PTR_KIND_TINY_HEADER) {
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
g_classify_header_hit++;
|
|
#endif
|
|
return result;
|
|
}
|
|
|
|
// Step 3: Try AllocHeader (HAKMEM header) for Mid/Large/Mmap
|
|
do {
|
|
if (!ptr) break;
|
|
// Quick page-safety check: avoid crossing page for header read
|
|
uintptr_t off = (uintptr_t)ptr & 0xFFFu;
|
|
int safe_same_page = (off >= HEADER_SIZE);
|
|
void* raw = (char*)ptr - HEADER_SIZE;
|
|
if (!safe_same_page) {
|
|
if (!hak_is_memory_readable(raw)) break;
|
|
}
|
|
AllocHeader* hdr = (AllocHeader*)raw;
|
|
if (hdr->magic == HAKMEM_MAGIC) {
|
|
result.kind = PTR_KIND_MID_LARGE; // HAKMEM-owned (non-Tiny)
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
g_classify_unknown_hit++; // reuse for stats without adding a new counter
|
|
#endif
|
|
return result;
|
|
}
|
|
} while (0);
|
|
|
|
// Step 4: Not recognized → UNKNOWN (route to libc or slow path)
|
|
result.kind = PTR_KIND_UNKNOWN;
|
|
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
g_classify_unknown_hit++;
|
|
#endif
|
|
|
|
return result;
|
|
}
|