[Mesa-dev] [PATCH 2/3] draw: check for an integer overflow when computing stride

[Zack Rusin]

Our buffer overflow arithmetic was susceptible to integer
overflows which was the buffer overflow logic to break.
Lets use the llvm overflow intrinsics to check for integer
overflows while computing the stride/needed buffer size.

Signed-off-by: Zack Rusin <zackr at vmware.com>
---
 src/gallium/auxiliary/draw/draw_llvm.c |   53 ++++++++++++++++++++++++++------
 1 file changed, 43 insertions(+), 10 deletions(-)

diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c
index f27776a..c1b4acd 100644
--- a/src/gallium/auxiliary/draw/draw_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_llvm.c
@@ -699,6 +699,13 @@ generate_fetch(struct gallivm_state *gallivm,
    LLVMValueRef temp_ptr =
       lp_build_alloca(gallivm,
                       lp_build_vec_type(gallivm, lp_float32_vec4_type()), "");
+   LLVMValueRef ofbit, oresult;
+   LLVMTypeRef oelems[2] = {
+      LLVMInt32TypeInContext(gallivm->context),
+      LLVMInt1TypeInContext(gallivm->context)
+   };
+   LLVMTypeRef otype = LLVMStructTypeInContext(gallivm->context,
+                                               oelems, 2, FALSE);
    struct lp_build_if_state if_ctx;
 
    if (velem->instance_divisor) {
@@ -708,23 +715,49 @@ generate_fetch(struct gallivm_state *gallivm,
                             "instance_divisor");
    }
 
-   stride = LLVMBuildMul(builder, vb_stride, index, "");
+   oresult = lp_build_intrinsic_binary(builder,
+                                       "llvm.umul.with.overflow.i32",
+                                       otype, vb_stride, index);
+   ofbit = LLVMBuildExtractValue(builder, oresult, 1, "");
+   stride = LLVMBuildExtractValue(builder, oresult, 0, "");
+
+   oresult = lp_build_intrinsic_binary(builder,
+                                       "llvm.uadd.with.overflow.i32",
+                                       otype, stride, vb_buffer_offset);
+   ofbit = LLVMBuildOr(
+      builder, ofbit,
+      LLVMBuildExtractValue(builder, oresult,  1, ""),
+      "");
+   stride = LLVMBuildExtractValue(builder, oresult, 0, "");
+
+   oresult = lp_build_intrinsic_binary(
+      builder,
+      "llvm.uadd.with.overflow.i32",
+      otype, stride,
+      lp_build_const_int32(gallivm, velem->src_offset));
+   ofbit = LLVMBuildOr(
+      builder, ofbit,
+      LLVMBuildExtractValue(builder, oresult, 1, ""),
+      "");
+   stride = LLVMBuildExtractValue(builder, oresult,  0, "");
 
-   stride = LLVMBuildAdd(builder, stride,
-                         vb_buffer_offset,
-                         "");
-   stride = LLVMBuildAdd(builder, stride,
-                         lp_build_const_int32(gallivm, velem->src_offset),
-                         "");
-   needed_buffer_size = LLVMBuildAdd(
-      builder, stride,
+
+   oresult = lp_build_intrinsic_binary(
+      builder,
+      "llvm.uadd.with.overflow.i32",
+      otype, stride,
       lp_build_const_int32(gallivm,
-                           util_format_get_blocksize(velem->src_format)),
+                           util_format_get_blocksize(velem->src_format)));
+   ofbit = LLVMBuildOr(
+      builder, ofbit,
+      LLVMBuildExtractValue(builder, oresult, 1, ""),
       "");
+   needed_buffer_size = LLVMBuildExtractValue(builder, oresult, 0, "");
 
    buffer_overflowed = LLVMBuildICmp(builder, LLVMIntUGT,
                                      needed_buffer_size, buffer_size,
                                      "buffer_overflowed");
+   buffer_overflowed = LLVMBuildOr(builder, buffer_overflowed, ofbit, "");
 #if 0
    lp_build_printf(gallivm, "vbuf index = %u, vb_stride is %u\n",
                    index, vb_stride);
-- 
1.7.10.4