bpf: avoid load from read-only sections

If users tried to have a structure decl/init code like below
   struct test_t t = { .memeber1 = 45 };
It is very likely that compiler will generate a readonly section
to hold up the init values for variable t. Later load of t members,
e.g., t.member1 will result in a read from readonly section.

BPF program cannot handle relocation. This will force users to
write:
  struct test_t t = {};
  t.member1 = 45;
This is just inconvenient and unintuitive.

This patch addresses this issue by implementing BPF PreprocessISelDAG.
For any load from a global constant structure or an global array of
constant struct, it attempts to
translate it into a constant directly. The traversal of the
constant struct and other constant data structures are similar
to where the assembler emits read-only sections.

Four different unit test cases are also added to cover
different scenarios.

Signed-off-by: Yonghong Song <yhs@fb.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@305560 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 187 insertions, 0 deletions

diff --git a/test/CodeGen/BPF/rodata_1.ll b/test/CodeGen/BPF/rodata_1.ll
new file mode 100644
index 00000000000..5566f76bb75
--- /dev/null
+++ b/test/CodeGen/BPF/rodata_1.ll
@@ -0,0 +1,52 @@
+; RUN: llc < %s -march=bpfel -verify-machineinstrs | FileCheck %s
+; RUN: llc < %s -march=bpfeb -verify-machineinstrs | FileCheck %s
+
+; Source code:
+; struct test_t1 {
+;   char a, b, c;
+; };
+; struct test_t2 {
+;   int a, b, c, d, e;
+; };
+;
+; struct test_t1 g1;
+; struct test_t2 g2;
+; int test()
+; {
+;   struct test_t1 t1 = {.c = 1};
+;   struct test_t2 t2 = {.c = 1};
+;   g1 = t1;
+;   g2 = t2;
+;   return 0;
+; }
+
+%struct.test_t1 = type { i8, i8, i8 }
+%struct.test_t2 = type { i32, i32, i32, i32, i32 }
+
+@test.t1 = private unnamed_addr constant %struct.test_t1 { i8 0, i8 0, i8 1 }, align 1
+@test.t2 = private unnamed_addr constant %struct.test_t2 { i32 0, i32 0, i32 1, i32 0, i32 0 }, align 4
+@g1 = common local_unnamed_addr global %struct.test_t1 zeroinitializer, align 1
+@g2 = common local_unnamed_addr global %struct.test_t2 zeroinitializer, align 4
+
+; Function Attrs: nounwind
+define i32 @test() local_unnamed_addr #0 {
+; CHECK-LABEL: test:
+
+entry:
+    tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* getelementptr inbounds (%struct.test_t1, %struct.test_t1* @g1, i64 0, i32 0), i8* getelementptr inbounds (%struct.test_t1, %struct.test_t1* @test.t1, i64 0, i32 0), i64 3, i32 1, i1 false)
+    tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* bitcast (%struct.test_t2* @g2 to i8*), i8* bitcast (%struct.test_t2* @test.t2 to i8*), i64 20, i32 4, i1 false)
+; CHECK:  r1 = <MCOperand Expr:(g1)>ll
+; CHECK:  r2 = 0
+; CHECK:  *(u8 *)(r1 + 1) = r2
+; CHECK:  r3 = 1
+; CHECK:  *(u8 *)(r1 + 2) = r3
+; CHECK:  r1 = <MCOperand Expr:(g2)>ll
+; CHECK:  *(u32 *)(r1 + 8) = r3
+    ret i32 0
+}
+; CHECK: .section  .rodata,"a",@progbits
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1) #1
+
+attributes #0 = { nounwind }
+attributes #1 = { argmemonly nounwind }
diff --git a/test/CodeGen/BPF/rodata_2.ll b/test/CodeGen/BPF/rodata_2.ll
new file mode 100644
index 00000000000..74b3c3640c3
--- /dev/null
+++ b/test/CodeGen/BPF/rodata_2.ll
@@ -0,0 +1,51 @@
+; RUN: llc < %s -march=bpfel -verify-machineinstrs | FileCheck %s
+; RUN: llc < %s -march=bpfeb -verify-machineinstrs | FileCheck %s
+
+; Source code:
+; struct test_t1 {
+;   char a;
+;   int  b;
+; };
+; struct test_t2 {
+;   char a, b;
+;   struct test_t1 c[2];
+;   int d[2];
+;   int e;
+; };
+; struct test_t2 g;
+; int test()
+; {
+;    struct test_t2 t2 = {.c = {{}, {.b = 1}}, .d = {2, 3}};
+;    g = t2;
+;    return 0;
+; }
+
+%struct.test_t2 = type { i8, i8, [2 x %struct.test_t1], [2 x i32], i32 }
+%struct.test_t1 = type { i8, i32 }
+
+@test.t2 = private unnamed_addr constant %struct.test_t2 { i8 0, i8 0, [2 x %struct.test_t1] [%struct.test_t1 zeroinitializer, %struct.test_t1 { i8 0, i32 1 }], [2 x i32] [i32 2, i32 3], i32 0 }, align 4
+@g = common local_unnamed_addr global %struct.test_t2 zeroinitializer, align 4
+
+; Function Attrs: nounwind
+define i32 @test() local_unnamed_addr #0 {
+; CHECK-LABEL: test:
+
+entry:
+    tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* getelementptr inbounds (%struct.test_t2, %struct.test_t2* @g, i64 0, i32 0), i8* getelementptr inbounds (%struct.test_t2, %struct.test_t2* @test.t2, i64 0, i32 0), i64 32, i32 4, i1 false)
+; CHECK:  r1 = <MCOperand Expr:(g)>ll
+; CHECK:  r2 = 0
+; CHECK:  *(u32 *)(r1 + 28) = r2
+; CHECK:  r3 = 3
+; CHECK:  *(u32 *)(r1 + 24) = r3
+; CHECK:  r3 = 2
+; CHECK:  *(u32 *)(r1 + 20) = r3
+; CHECK:  r3 = 1
+; CHECK:  *(u32 *)(r1 + 16) = r3
+      ret i32 0
+}
+; CHECK: .section  .rodata.cst32,"aM",@progbits,32
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1) #1
+
+attributes #0 = { nounwind }
+attributes #1 = { argmemonly nounwind }
diff --git a/test/CodeGen/BPF/rodata_3.ll b/test/CodeGen/BPF/rodata_3.ll
new file mode 100644
index 00000000000..814ce764546
--- /dev/null
+++ b/test/CodeGen/BPF/rodata_3.ll
@@ -0,0 +1,41 @@
+; REQUIRES: x86_64-linux
+; RUN: llc < %s -march=bpfel -verify-machineinstrs | FileCheck --check-prefix=CHECK-EL %s
+; RUN: llc < %s -march=bpfeb -verify-machineinstrs | FileCheck --check-prefix=CHECK-EB %s
+;
+; This test requires little-endian host, so we specific x86_64-linux here.
+; Source code:
+; struct test_t1 {
+;   char a;
+;   int b, c, d;
+; };
+;
+; struct test_t1 g;
+; int test()
+; {
+;   struct test_t1 t1 = {.a = 1};
+;   g = t1;
+;   return 0;
+; }
+
+%struct.test_t1 = type { i8, i32, i32, i32 }
+
+@test.t1 = private unnamed_addr constant %struct.test_t1 { i8 1, i32 0, i32 0, i32 0 }, align 4
+@g = common local_unnamed_addr global %struct.test_t1 zeroinitializer, align 4
+
+; Function Attrs: nounwind
+define i32 @test() local_unnamed_addr #0 {
+entry:
+    tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* getelementptr inbounds (%struct.test_t1, %struct.test_t1* @g, i64 0, i32 0), i8* getelementptr inbounds (%struct.test_t1, %struct.test_t1* @test.t1, i64 0, i32 0), i64 16, i32 4, i1 false)
+; CHECK-EL:  r2 = 1
+; CHECK-EL:  *(u32 *)(r1 + 0) = r2
+; CHECK-EB:  r2 = 16777216
+; CHECK-EB:  *(u32 *)(r1 + 0) = r2
+    ret i32 0
+}
+; CHECK-EL:  .section .rodata.cst16,"aM",@progbits,16
+; CHECK-EB:  .section .rodata.cst16,"aM",@progbits,16
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1) #1
+
+attributes #0 = { nounwind }
+attributes #1 = { argmemonly nounwind }
diff --git a/test/CodeGen/BPF/rodata_4.ll b/test/CodeGen/BPF/rodata_4.ll
new file mode 100644
index 00000000000..d6b9fba5be0
--- /dev/null
+++ b/test/CodeGen/BPF/rodata_4.ll
@@ -0,0 +1,43 @@
+; RUN: llc < %s -march=bpfel -verify-machineinstrs | FileCheck %s
+; RUN: llc < %s -march=bpfeb -verify-machineinstrs | FileCheck %s
+
+; Source code:
+; struct test_t1
+; {
+;   short a;
+;   short b;
+;   char c;
+; };
+;
+; struct test_t1 g;
+; int test()
+; {
+;   struct test_t1 t1[] = {{50, 500, 5}, {60, 600, 6}, {70, 700, 7}, {80, 800, 8} };
+;
+;   g = t1[1];
+;   return 0;
+; }
+
+%struct.test_t1 = type { i16, i16, i8 }
+
+@test.t1 = private unnamed_addr constant [4 x %struct.test_t1] [%struct.test_t1 { i16 50, i16 500, i8 5 }, %struct.test_t1 { i16 60, i16 600, i8 6 }, %struct.test_t1 { i16 70, i16 700, i8 7 }, %struct.test_t1 { i16 80, i16 800, i8 8 }], align 2
+@g = common local_unnamed_addr global %struct.test_t1 zeroinitializer, align 2
+
+; Function Attrs: nounwind
+define i32 @test() local_unnamed_addr #0 {
+; CHECK-LABEL: test:
+entry:
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* bitcast (%struct.test_t1* @g to i8*), i8* bitcast (%struct.test_t1* getelementptr inbounds ([4 x %struct.test_t1], [4 x %struct.test_t1]* @test.t1, i64 0, i64 1) to i8*), i64 6, i32 2, i1 false)
+; CHECK:  r2 = 600
+; CHECK:  *(u16 *)(r1 + 2) = r2
+; CHECK:  r2 = 60
+; CHECK:  *(u16 *)(r1 + 0) = r2
+  ret i32 0
+}
+; CHECK  .section  .rodata,"a",@progbits
+
+; Function Attrs: argmemonly nounwind
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1) #1
+
+attributes #0 = { nounwind }
+attributes #1 = { argmemonly nounwind }