[v1,1/1] malloc: fix ASan handling for unmapped memory
Checks
Commit Message
Currently, when we free previously allocated memory, we mark the area as
"freed" for ASan purposes (flag 0xfd). However, sometimes, freeing a
malloc element will cause pages to be unmapped from memory and re-backed
with anonymous memory again. This may cause ASan's "use-after-free"
error down the line, because the allocator will try to write into
memory areas recently marked as "freed".
To fix this, we need to mark the unmapped memory area as "available",
and fixup surrounding malloc element header/trailers to enable later
malloc routines to safely write into new malloc elements' headers or
trailers.
Fixes: 6cc51b1293ce ("mem: instrument allocator for ASan")
Cc: zhihongx.peng@intel.com
Cc: stable@dpdk.org
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
---
lib/eal/common/malloc_heap.c | 35 +++++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
@@ -861,6 +861,7 @@ malloc_heap_free(struct malloc_elem *elem)
struct rte_memseg_list *msl;
unsigned int i, n_segs, before_space, after_space;
int ret;
+ bool unmapped = false;
const struct internal_config *internal_conf =
eal_get_internal_configuration();
@@ -1027,6 +1028,9 @@ malloc_heap_free(struct malloc_elem *elem)
request_to_primary(&req);
}
+ /* we didn't exit early, meaning we have unmapped some pages */
+ unmapped = true;
+
RTE_LOG(DEBUG, EAL, "Heap on socket %d was shrunk by %zdMB\n",
msl->socket_id, aligned_len >> 20ULL);
@@ -1034,6 +1038,37 @@ malloc_heap_free(struct malloc_elem *elem)
free_unlock:
asan_set_freezone(asan_ptr, asan_data_len);
+ /* if we unmapped some memory, we need to do additional work for ASan */
+ if (unmapped) {
+ void *asan_end = RTE_PTR_ADD(asan_ptr, asan_data_len);
+ void *aligned_end = RTE_PTR_ADD(aligned_start, aligned_len);
+ void *aligned_trailer = RTE_PTR_SUB(aligned_start,
+ MALLOC_ELEM_TRAILER_LEN);
+
+ /*
+ * There was a memory area that was unmapped. This memory area
+ * will have to be marked as available for ASan, because we will
+ * want to use it next time it gets mapped again. The OS memory
+ * protection should trigger a fault on access to these areas
+ * anyway, so we are not giving up any protection.
+ */
+ asan_set_zone(aligned_start, aligned_len, 0x00);
+
+ /*
+ * ...however, when we unmap pages, we create new free elements
+ * which might have been marked as "freed" with an earlier
+ * `asan_set_freezone` call. So, if there is an area past the
+ * unmapped space that was marked as freezone for ASan, we need
+ * to mark the malloc header as available.
+ */
+ if (asan_end > aligned_end)
+ asan_set_zone(aligned_end, MALLOC_ELEM_HEADER_LEN, 0x00);
+
+ /* if there's space before unmapped memory, mark as available */
+ if (asan_ptr < aligned_start)
+ asan_set_zone(aligned_trailer, MALLOC_ELEM_TRAILER_LEN, 0x00);
+ }
+
rte_spinlock_unlock(&(heap->lock));
return ret;
}