Don't expose guard pages to malloc_stack API consumers (#54591)

Whether or not a guard page is in effect is an implementation detail and
consumers of the `malloc_stack` API should not have to worry about that.
In particular, if a stack of a certain size is requested, a stack of that
size should be delivered, and not be reduced on some systems because we
park a guard page in that range.

This also helps consumers of the gcext API implementing stack scanning
(i.e., GAP.jl), as it does not have to worry about running into those
guard pages anymore.

(cherry picked from commit 5dfd57d)

Author: fingolfin

src/gc-stacks.c

@@ -22,13 +22,22 @@
 // number of stacks to always keep available per pool
 #define MIN_STACK_MAPPINGS_PER_POOL 5
 
+#if defined(_OS_WINDOWS_) || (!defined(_OS_OPENBSD_) && !defined(JL_HAVE_UCONTEXT) && !defined(JL_HAVE_SIGALTSTACK))
+#define JL_USE_GUARD_PAGE   1
 const size_t jl_guard_size = (4096 * 8);
+#else
+const size_t jl_guard_size = 0;
+#endif
+
 static _Atomic(uint32_t) num_stack_mappings = 0;
 
 #ifdef _OS_WINDOWS_
 #define MAP_FAILED NULL
 static void *malloc_stack(size_t bufsz) JL_NOTSAFEPOINT
 {
+    size_t guard_size = LLT_ALIGN(jl_guard_size, jl_page_size);
+    bufsz += guard_size;
+
     void *stk = VirtualAlloc(NULL, bufsz, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
     if (stk == NULL)
         return MAP_FAILED;
@@ -37,13 +46,21 @@ static void *malloc_stack(size_t bufsz) JL_NOTSAFEPOINT
         VirtualFree(stk, 0, MEM_RELEASE);
         return MAP_FAILED;
     }
+    stk = (char *)stk + guard_size;
+
     jl_atomic_fetch_add(&num_stack_mappings, 1);
     return stk;
 }
 
 
 static void free_stack(void *stkbuf, size_t bufsz)
 {
+#ifdef JL_USE_GUARD_PAGE
+    size_t guard_size = LLT_ALIGN(jl_guard_size, jl_page_size);
+    bufsz += guard_size;
+    stkbuf = (char *)stkbuf - guard_size;
+#endif
+
     VirtualFree(stkbuf, 0, MEM_RELEASE);
     jl_atomic_fetch_add(&num_stack_mappings, -1);
 }
@@ -52,22 +69,35 @@ static void free_stack(void *stkbuf, size_t bufsz)
 
 static void *malloc_stack(size_t bufsz) JL_NOTSAFEPOINT
 {
+#ifdef JL_USE_GUARD_PAGE
+    size_t guard_size = LLT_ALIGN(jl_guard_size, jl_page_size);
+    bufsz += guard_size;
+#endif
+
     void* stk = mmap(0, bufsz, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
     if (stk == MAP_FAILED)
         return MAP_FAILED;
-#if !defined(JL_HAVE_UCONTEXT) && !defined(JL_HAVE_SIGALTSTACK)
-    // setup a guard page to detect stack overflow
+
+#ifdef JL_USE_GUARD_PAGE
+    // set up a guard page to detect stack overflow
     if (mprotect(stk, jl_guard_size, PROT_NONE) == -1) {
         munmap(stk, bufsz);
         return MAP_FAILED;
     }
+    stk = (char *)stk + guard_size;
 #endif
     jl_atomic_fetch_add(&num_stack_mappings, 1);
     return stk;
 }
 
 static void free_stack(void *stkbuf, size_t bufsz)
 {
+#ifdef JL_USE_GUARD_PAGE
+    size_t guard_size = LLT_ALIGN(jl_guard_size, jl_page_size);
+    bufsz += guard_size;
+    stkbuf = (char *)stkbuf - guard_size;
+#endif
+
     munmap(stkbuf, bufsz);
     jl_atomic_fetch_add(&num_stack_mappings, -1);
 }