Make smallbuddy handle larger requests correctly (#556)

* Fail more abruptly if the bounds are not exact.

* Move bounding from Pool into Backend.

This commit makes the rounding and the bounding occur in the same
function.

* Enable smallbuddyrange to handle larger requests

The smallbuddy can now pass the larger requests up the range chain if
it cannot satisfy it itself.

* Test larger requests for meta-data.

Author: mjp41

src/snmalloc/aal/aal_cheri.h

@@ -86,6 +86,10 @@ namespace snmalloc
       }
 
       void* pb = __builtin_cheri_bounds_set_exact(a.unsafe_ptr(), size);
+
+      SNMALLOC_ASSERT(
+        __builtin_cheri_tag_get(pb) && "capptr_bound exactness failed.");
+
       return CapPtr<T, BOut>::unsafe_from(static_cast<T*>(pb));
     }
 

src/snmalloc/backend/backend.h

@@ -39,7 +39,7 @@ namespace snmalloc
      * does not avail itself of this degree of freedom.
      */
     template<typename T>
-    static capptr::Arena<void>
+    static capptr::Alloc<void>
     alloc_meta_data(LocalState* local_state, size_t size)
     {
       capptr::Arena<void> p;
@@ -61,9 +61,13 @@ namespace snmalloc
       }
 
       if (p == nullptr)
+      {
         errno = ENOMEM;
+        return nullptr;
+      }
 
-      return p;
+      return capptr_to_user_address_control(
+        Aal::capptr_bound<void, capptr::bounds::AllocFull>(p, size));
     }
 
     /**

src/snmalloc/backend_helpers/smallbuddyrange.h

@@ -174,13 +174,22 @@ namespace snmalloc
           base,
           length,
           [this](CapPtr<void, ChunkBounds> base, size_t align, bool) {
-            CapPtr<void, ChunkBounds> overflow =
-              buddy_small
-                .add_block(
-                  base.template as_reinterpret<FreeChunk<ChunkBounds>>(), align)
-                .template as_reinterpret<void>();
-            if (overflow != nullptr)
-              parent.dealloc_range(overflow, bits::one_at_bit(MIN_CHUNK_BITS));
+            if (align < MIN_CHUNK_SIZE)
+            {
+              CapPtr<void, ChunkBounds> overflow =
+                buddy_small
+                  .add_block(
+                    base.template as_reinterpret<FreeChunk<ChunkBounds>>(),
+                    align)
+                  .template as_reinterpret<void>();
+              if (overflow != nullptr)
+                parent.dealloc_range(
+                  overflow, bits::one_at_bit(MIN_CHUNK_BITS));
+            }
+            else
+            {
+              parent.dealloc_range(base, align);
+            }
           });
       }
 
@@ -204,7 +213,8 @@ namespace snmalloc
 
       CapPtr<void, ChunkBounds> alloc_range(size_t size)
       {
-        SNMALLOC_ASSERT(size < MIN_CHUNK_SIZE);
+        if (size >= MIN_CHUNK_SIZE)
+          return parent.alloc_range(size);
 
         auto result = buddy_small.remove_block(size);
         if (result != nullptr)
@@ -218,8 +228,6 @@ namespace snmalloc
 
       CapPtr<void, ChunkBounds> alloc_range_with_leftover(size_t size)
       {
-        SNMALLOC_ASSERT(size < MIN_CHUNK_SIZE);
-
         auto rsize = bits::next_pow2(size);
 
         auto result = alloc_range(rsize);
@@ -236,8 +244,6 @@ namespace snmalloc
 
       void dealloc_range(CapPtr<void, ChunkBounds> base, size_t size)
       {
-        SNMALLOC_ASSERT(size < MIN_CHUNK_SIZE);
-
         add_range(base, size);
       }
     };

src/snmalloc/mem/backend_concept.h

@@ -122,7 +122,7 @@ namespace snmalloc
     {
       Backend::template alloc_meta_data<void*>(local_state, size)
     }
-    ->ConceptSame<capptr::Arena<void>>;
+    ->ConceptSame<capptr::Alloc<void>>;
   }
   &&requires(
     LocalState& local_state,

src/snmalloc/mem/pool.h

@@ -137,11 +137,8 @@ namespace snmalloc
         Config::Pal::error("Failed to initialise thread local allocator.");
       }
 
-      p = capptr_to_user_address_control(
-        Aal::capptr_bound<T, capptr::bounds::AllocFull>(
-          capptr::Arena<T>::unsafe_from(new (raw.unsafe_ptr())
-                                          T(std::forward<Args>(args)...)),
-          sizeof(T)));
+      p = capptr::Alloc<T>::unsafe_from(new (raw.unsafe_ptr())
+                                          T(std::forward<Args>(args)...));
 
       FlagLock f(pool.lock);
       p->list_next = pool.list;

src/test/func/pool/pool.cc

@@ -1,3 +1,5 @@
+#include <array>
+#include <iostream>
 #include <snmalloc/snmalloc.h>
 #include <test/opt.h>
 #include <test/setup.h>
@@ -24,6 +26,23 @@ struct PoolBEntry : Pooled<PoolBEntry>
 
 using PoolB = Pool<PoolBEntry, Alloc::Config>;
 
+struct PoolLargeEntry : Pooled<PoolLargeEntry>
+{
+  std::array<int, 2'000'000> payload;
+
+  PoolLargeEntry()
+  {
+    printf(".");
+    fflush(stdout);
+    payload[0] = 1;
+    printf("first %d\n", payload[0]);
+    payload[1'999'999] = 1;
+    printf("last %d\n", payload[1'999'999]);
+  };
+};
+
+using PoolLarge = Pool<PoolLargeEntry, Alloc::Config>;
+
 void test_alloc()
 {
   auto ptr = PoolA::acquire();
@@ -116,6 +135,18 @@ void test_iterator()
   PoolA::release(after_iteration_ptr);
 }
 
+void test_large()
+{
+  printf(".");
+  fflush(stdout);
+  PoolLargeEntry* p = PoolLarge::acquire();
+  printf(".");
+  fflush(stdout);
+  PoolLarge::release(p);
+  printf(".");
+  fflush(stdout);
+}
+
 int main(int argc, char** argv)
 {
   setup();
@@ -128,10 +159,18 @@ int main(int argc, char** argv)
 #endif
 
   test_alloc();
+  std::cout << "test_alloc passed" << std::endl;
   test_constructor();
+  std::cout << "test_constructor passed" << std::endl;
   test_alloc_many();
+  std::cout << "test_alloc_many passed" << std::endl;
   test_double_alloc();
+  std::cout << "test_double_alloc passed" << std::endl;
   test_different_alloc();
+  std::cout << "test_different_alloc passed" << std::endl;
   test_iterator();
+  std::cout << "test_iterator passed" << std::endl;
+  test_large();
+  std::cout << "test_large passed" << std::endl;
   return 0;
 }