tomoaki/hakmem
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
c86d0d0f7b6051ebda69b9a4f342b9ae05e0756c
hakmem/hakmem_batch.d

12 lines
380 B
D
Raw Normal View History

Tiny: fix header/stride mismatch and harden refill paths - Root cause: header-based class indexing (HEADER_CLASSIDX=1) wrote a 1-byte header during allocation, but linear carve/refill and initial slab capacity still used bare class block sizes. This mismatch could overrun slab usable space and corrupt freelists, causing reproducible SEGV at ~100k iters. Changes - Superslab: compute capacity with effective stride (block_size + header for classes 0..6; class7 remains headerless) in superslab_init_slab(). Add a debug-only bound check in superslab_alloc_from_slab() to fail fast if carve would exceed usable bytes. - Refill (non-P0 and P0): use header-aware stride for all linear carving and TLS window bump operations. Ensure alignment/validation in tiny_refill_opt.h also uses stride, not raw class size. - Drain: keep existing defense-in-depth for remote sentinel and sanitize nodes before splicing into freelist (already present). Notes - This unifies the memory layout across alloc/linear-carve/refill with a single stride definition and keeps class7 (1024B) headerless as designed. - Debug builds add fail-fast checks; release builds remain lean. Next - Re-run Tiny benches (256/1024B) in debug to confirm stability, then in release. If any remaining crash persists, bisect with HAKMEM_TINY_P0_BATCH_REFILL=0 to isolate P0 batch carve, and continue reducing branch-miss as planned.
2025-11-09 18:55:50 +09:00
hakmem_batch.o: core/hakmem_batch.c core/hakmem_batch.h core/hakmem_sys.h \
release: silence runtime logs and stabilize benches - Fix HAKMEM_LOG gating to use (numeric) so release builds compile out logs. - Switch remaining prints to HAKMEM_LOG or guard with : - core/box/hak_core_init.inc.h (EVO sample warning, shutdown banner) - core/hakmem_config.c (config/feature prints) - core/hakmem.c (BigCache eviction prints) - core/hakmem_tiny_superslab.c (OOM, head init/expand, C7 init diagnostics) - core/hakmem_elo.c (init/evolution) - core/hakmem_batch.c (init/flush/stats) - core/hakmem_ace.c (33KB route diagnostics) - core/hakmem_ace_controller.c (ACE logs macro → no-op in release) - core/hakmem_site_rules.c (init banner) - core/box/hak_free_api.inc.h (unknown method error → release-gated) - Rebuilt benches and verified quiet output for release: - bench_fixed_size_hakmem/system - bench_random_mixed_hakmem/system - bench_mid_large_mt_hakmem/system - bench_comprehensive_hakmem/system Note: Kept debug logs available in debug builds and when explicitly toggled via env.
2025-11-11 01:47:06 +09:00
core/hakmem_whale.h core/hakmem_internal.h core/hakmem.h \
core/hakmem_build_flags.h core/hakmem_config.h core/hakmem_features.h
Tiny: fix header/stride mismatch and harden refill paths - Root cause: header-based class indexing (HEADER_CLASSIDX=1) wrote a 1-byte header during allocation, but linear carve/refill and initial slab capacity still used bare class block sizes. This mismatch could overrun slab usable space and corrupt freelists, causing reproducible SEGV at ~100k iters. Changes - Superslab: compute capacity with effective stride (block_size + header for classes 0..6; class7 remains headerless) in superslab_init_slab(). Add a debug-only bound check in superslab_alloc_from_slab() to fail fast if carve would exceed usable bytes. - Refill (non-P0 and P0): use header-aware stride for all linear carving and TLS window bump operations. Ensure alignment/validation in tiny_refill_opt.h also uses stride, not raw class size. - Drain: keep existing defense-in-depth for remote sentinel and sanitize nodes before splicing into freelist (already present). Notes - This unifies the memory layout across alloc/linear-carve/refill with a single stride definition and keeps class7 (1024B) headerless as designed. - Debug builds add fail-fast checks; release builds remain lean. Next - Re-run Tiny benches (256/1024B) in debug to confirm stability, then in release. If any remaining crash persists, bisect with HAKMEM_TINY_P0_BATCH_REFILL=0 to isolate P0 batch carve, and continue reducing branch-miss as planned.
2025-11-09 18:55:50 +09:00
core/hakmem_batch.h:
core/hakmem_sys.h:
core/hakmem_whale.h:
release: silence runtime logs and stabilize benches - Fix HAKMEM_LOG gating to use (numeric) so release builds compile out logs. - Switch remaining prints to HAKMEM_LOG or guard with : - core/box/hak_core_init.inc.h (EVO sample warning, shutdown banner) - core/hakmem_config.c (config/feature prints) - core/hakmem.c (BigCache eviction prints) - core/hakmem_tiny_superslab.c (OOM, head init/expand, C7 init diagnostics) - core/hakmem_elo.c (init/evolution) - core/hakmem_batch.c (init/flush/stats) - core/hakmem_ace.c (33KB route diagnostics) - core/hakmem_ace_controller.c (ACE logs macro → no-op in release) - core/hakmem_site_rules.c (init banner) - core/box/hak_free_api.inc.h (unknown method error → release-gated) - Rebuilt benches and verified quiet output for release: - bench_fixed_size_hakmem/system - bench_random_mixed_hakmem/system - bench_mid_large_mt_hakmem/system - bench_comprehensive_hakmem/system Note: Kept debug logs available in debug builds and when explicitly toggled via env.
2025-11-11 01:47:06 +09:00
core/hakmem_internal.h:
core/hakmem.h:
core/hakmem_build_flags.h:
core/hakmem_config.h:
core/hakmem_features.h:
Reference in New Issue Copy Permalink