Merge branch 'dev' into dev2

contrib/docker/alpine-arm32v7/Dockerfile

@@ -1,6 +1,6 @@
 # install from an image
 # download first an appropriate tar.gz image into the current directory
-# from: <https://github.com/alpinelinux/docker-alpine/tree/edge/armv7>
+# from <https://github.com/alpinelinux/docker-alpine/tree/edge/armv7>
 FROM scratch
 
 # Substitute the image name that was downloaded

contrib/docker/alpine-x86/Dockerfile

@@ -0,0 +1,28 @@
+# install from an image
+# download first an appropriate tar.gz image into the current directory
+# from <https://github.com/alpinelinux/docker-alpine/tree/edge/x86>
+FROM scratch
+
+# Substitute the image name that was downloaded
+ADD alpine-minirootfs-20250108-x86.tar.gz /
+
+# Install tools
+RUN apk add build-base make cmake
+RUN apk add git
+RUN apk add vim
+
+RUN mkdir -p  /home/dev
+WORKDIR /home/dev
+
+# Get mimalloc
+RUN git clone https://github.com/microsoft/mimalloc -b dev2
+RUN mkdir -p mimalloc/out/release
+RUN mkdir -p mimalloc/out/debug
+
+# Build mimalloc debug
+WORKDIR /home/dev/mimalloc/out/debug
+RUN cmake ../.. -DMI_DEBUG_FULL=ON
+# RUN make -j
+# RUN make test
+
+CMD ["/bin/sh"]

ide/vs2022/mimalloc-test-stress.vcxproj

@@ -282,8 +282,8 @@
     </ClCompile>
   </ItemGroup>
   <ItemGroup>
-    <ProjectReference Include="mimalloc-lib.vcxproj">
+    <ProjectReference Include="mimalloc-override-dll.vcxproj">
-      <Project>{abb5eae7-b3e6-432e-b636-333449892ea6}</Project>
+      <Project>{abb5eae7-b3e6-432e-b636-333449892ea7}</Project>
     </ProjectReference>
   </ItemGroup>
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />

include/mimalloc/prim.h

@@ -118,9 +118,6 @@ void _mi_prim_thread_done_auto_done(void);
 void _mi_prim_thread_associate_default_heap(mi_heap_t* heap);
 
 
-
-
-
 //-------------------------------------------------------------------
 // Access to TLS (thread local storage) slots.
 // We need fast access to both a unique thread id (in `free.c:mi_free`) and
@@ -211,7 +208,7 @@ static inline void mi_prim_tls_slot_set(size_t slot, void* value) mi_attr_noexce
 // thread-local initialization checks in the fast path.
 // We allocate a user TLS slot at process initialization (see `windows/prim.c`)
 // and store the offset `_mi_win_tls_offset`.
-#define MI_HAS_TLS_SLOT      2              // 2 = we can reliably initialize the slot (saving a test on each malloc)
+#define MI_HAS_TLS_SLOT  1              // 2 = we can reliably initialize the slot (saving a test on each malloc)
 
 extern mi_decl_hidden size_t _mi_win_tls_offset;
 

include/mimalloc/types.h

@@ -571,7 +571,6 @@ struct mi_heap_s {
   size_t                guarded_size_min;                    // minimal size for guarded objects
   size_t                guarded_size_max;                    // maximal size for guarded objects
   size_t                guarded_sample_rate;                 // sample rate (set to 0 to disable guarded pages)
-  size_t                guarded_sample_seed;                 // starting sample count
   size_t                guarded_sample_count;                // current sample count (counting down to 0)
   #endif
   mi_page_t*            pages_free_direct[MI_PAGES_DIRECT];  // optimize: array where every entry points a page with possibly free blocks in the corresponding queue for that size.

src/arena.c

@@ -951,7 +951,7 @@ void mi_debug_show_arenas(void) mi_attr_noexcept {
   for (size_t i = 0; i < max_arenas; i++) {
     mi_arena_t* arena = mi_atomic_load_ptr_relaxed(mi_arena_t, &mi_arenas[i]);
     if (arena == NULL) break;
-    _mi_message("arena %zu: %zu blocks of size %zuMiB (in %zu fields) %s\n", i, arena->block_count, MI_ARENA_BLOCK_SIZE / MI_MiB, arena->field_count, (arena->memid.is_pinned ? ", pinned" : ""));
+    _mi_message("arena %zu: %zu blocks of size %zuMiB (in %zu fields) %s\n", i, arena->block_count, (size_t)(MI_ARENA_BLOCK_SIZE / MI_MiB), arena->field_count, (arena->memid.is_pinned ? ", pinned" : ""));
     if (show_inuse) {
       inuse_total += mi_debug_show_bitmap("  ", "inuse blocks", arena->block_count, arena->blocks_inuse, arena->field_count);
     }

src/init.c

@@ -123,7 +123,7 @@ mi_decl_cache_align const mi_heap_t _mi_heap_empty = {
   false,            // can reclaim
   0,                // tag
   #if MI_GUARDED
-  0, 0, 0, 0, 1,    // count is 1 so we never write to it (see `internal.h:mi_heap_malloc_use_guarded`)
+  0, 0, 0, 1,       // count is 1 so we never write to it (see `internal.h:mi_heap_malloc_use_guarded`)
   #endif
   MI_SMALL_PAGES_EMPTY,
   MI_PAGE_QUEUES_EMPTY
@@ -172,7 +172,7 @@ mi_decl_cache_align mi_heap_t _mi_heap_main = {
   false,            // can reclaim
   0,                // tag
   #if MI_GUARDED
-  0, 0, 0, 0, 0,
+  0, 0, 0, 0,
   #endif
   MI_SMALL_PAGES_EMPTY,
   MI_PAGE_QUEUES_EMPTY
@@ -184,15 +184,14 @@ mi_stats_t _mi_stats_main = { MI_STAT_VERSION, MI_STATS_NULL };
 
 #if MI_GUARDED
 mi_decl_export void mi_heap_guarded_set_sample_rate(mi_heap_t* heap, size_t sample_rate, size_t seed) {
-  heap->guarded_sample_seed = seed;
-  if (heap->guarded_sample_seed == 0) {
-    heap->guarded_sample_seed = _mi_heap_random_next(heap);
-  }
   heap->guarded_sample_rate  = sample_rate;
-  if (heap->guarded_sample_rate >= 1) {
+  heap->guarded_sample_count = sample_rate;  // count down samples
-    heap->guarded_sample_seed = heap->guarded_sample_seed % heap->guarded_sample_rate;
+  if (heap->guarded_sample_rate > 1) {
+    if (seed == 0) {
+      seed = _mi_heap_random_next(heap);
+    }
+    heap->guarded_sample_count = (seed % heap->guarded_sample_rate) + 1;  // start at random count between 1 and `sample_rate`
   }
-  heap->guarded_sample_count = heap->guarded_sample_seed;  // count down samples
 }
 
 mi_decl_export void mi_heap_guarded_set_size_bound(mi_heap_t* heap, size_t min, size_t max) {
@@ -245,7 +244,6 @@ mi_heap_t* _mi_heap_main_get(void) {
   return &_mi_heap_main;
 }
 
-
 /* -----------------------------------------------------------
   Sub process
 ----------------------------------------------------------- */

src/prim/unix/prim.c

@@ -32,7 +32,7 @@ terms of the MIT license. A copy of the license can be found in the file
 #if defined(__linux__)
   #include <features.h>
   #include <sys/prctl.h>    // THP disable, PR_SET_VMA
-  #if !defined(PR_SET_VMA)
+  #if defined(__GLIBC__) && !defined(PR_SET_VMA)
   #include <linux/prctl.h>
   #endif
   #if defined(__GLIBC__)

src/prim/windows/prim.c

@@ -628,32 +628,41 @@ bool _mi_prim_random_buf(void* buf, size_t buf_len) {
 //----------------------------------------------------------------
 
 #if MI_WIN_USE_FIXED_TLS==1
-mi_decl_cache_align size_t _mi_win_tls_offset = sizeof(void*);  // use 2nd slot by default
+mi_decl_cache_align size_t _mi_win_tls_offset = 0;  
 #endif
 
+//static void mi_debug_out(const char* s) {
+//  HANDLE h = GetStdHandle(STD_ERROR_HANDLE);
+//  WriteConsole(h, s, (DWORD)_mi_strlen(s), NULL, NULL);
+//}
+
+static void mi_win_tls_init(DWORD reason) {
+  if (reason==DLL_PROCESS_ATTACH || reason==DLL_THREAD_ATTACH) {
+    #if MI_WIN_USE_FIXED_TLS==1  // we must allocate a TLS slot dynamically
+    if (_mi_win_tls_offset == 0 && reason == DLL_PROCESS_ATTACH) {
+      const DWORD tls_slot = TlsAlloc();  // usually returns slot 1
+      if (tls_slot == TLS_OUT_OF_INDEXES) {
+        _mi_error_message(EFAULT, "unable to allocate the a TLS slot (rebuild without MI_WIN_USE_FIXED_TLS?)\n");
+      }
+      _mi_win_tls_offset = (size_t)tls_slot * sizeof(void*);
+    }
+    #endif
+    #if MI_HAS_TLS_SLOT >= 2  // we must initialize the TLS slot before any allocation
+    if (mi_prim_get_default_heap() == NULL) {
+      _mi_heap_set_default_direct((mi_heap_t*)&_mi_heap_empty);  
+      #if MI_DEBUG && MI_WIN_USE_FIXED_TLS==1
+      void* const p = TlsGetValue((DWORD)(_mi_win_tls_offset / sizeof(void*)));
+      mi_assert_internal(p == (void*)&_mi_heap_empty);
+      #endif  
+    }
+    #endif  
+  }  
+}
+
 static void NTAPI mi_win_main(PVOID module, DWORD reason, LPVOID reserved) {
   MI_UNUSED(reserved);
   MI_UNUSED(module);
-  #if MI_HAS_TLS_SLOT >= 2  // we must initialize the TLS slot before any allocation
+  mi_win_tls_init(reason);
-  #if MI_WIN_USE_FIXED_TLS==1
-  if (reason==DLL_PROCESS_ATTACH) {
-    const DWORD tls_slot = TlsAlloc();
-    if (tls_slot == TLS_OUT_OF_INDEXES) { 
-      _mi_error_message(EFAULT, "unable to allocate the a TLS slot (rebuild without MI_WIN_USE_FIXED_TLS?)\n"); 
-    }
-    _mi_win_tls_offset = (size_t)tls_slot * sizeof(void*);
-  }
-  #endif
-  if (reason==DLL_PROCESS_ATTACH || reason==DLL_THREAD_ATTACH) {
-    if (mi_prim_get_default_heap() == NULL) {
-      _mi_heap_set_default_direct((mi_heap_t*)&_mi_heap_empty);
-    }
-    #if MI_DEBUG && MI_WIN_USE_FIXED_TLS==1
-    void* const p = TlsGetValue((DWORD)(_mi_win_tls_offset / sizeof(void*)));
-    mi_assert_internal(p == (void*)&_mi_heap_empty);
-    #endif  
-  }
-  #endif
   if (reason==DLL_PROCESS_ATTACH) {
     _mi_process_load();
   }
@@ -815,11 +824,7 @@ static void NTAPI mi_win_main(PVOID module, DWORD reason, LPVOID reserved) {
   #endif
   mi_decl_export void _mi_redirect_entry(DWORD reason) {
     // called on redirection; careful as this may be called before DllMain
-    #if MI_HAS_TLS_SLOT >= 2 // we must initialize the TLS slot before any allocation
+    mi_win_tls_init(reason);
-    if ((reason==DLL_PROCESS_ATTACH || reason==DLL_THREAD_ATTACH) && mi_prim_get_default_heap() == NULL) {
-      _mi_heap_set_default_direct((mi_heap_t*)&_mi_heap_empty);
-    }
-    #endif
     if (reason == DLL_PROCESS_ATTACH) {
       mi_redirected = true;
     }

test/main-override-static.c

@@ -43,7 +43,7 @@ int main() {
   // corrupt_free();
   // block_overflow1();
   // block_overflow2();
-  // test_canary_leak();
+  test_canary_leak();
   // test_aslr();
   // invalid_free();
   // test_reserved();