[RISCV] Fix evaluating %pcrel_lo against global and weak symbols

Summary:
Previously, we would erroneously turn %pcrel_lo(label), where label has
a %pcrel_hi against a weak symbol, into %pcrel_lo(label + offset), as
evaluatePCRelLo would believe the target independent logic was going to
fold it. Moreover, even if that were fixed, shouldForceRelocation lacks
an MCAsmLayout and thus cannot evaluate the %pcrel_hi fixup to a value
and check the symbol, so we would then erroneously constant-fold the
%pcrel_lo whilst leaving the %pcrel_hi intact. After D72197, this same
sequence also occurs for symbols with global binding, which is triggered
in real-world code.

Instead, as discussed in D71978, we introduce a new FKF_IsTarget flag to
avoid these kinds of issues. All the resolution logic happens in one
place, with no coordination required between RISCAsmBackend and
RISCVMCExpr to ensure they implement the same logic twice. Although the
implementation of %pcrel_hi can be left as target independent, we make
it target dependent to ensure that they are handled identically to
%pcrel_lo, otherwise we risk one of them being constant folded but the
other being preserved. This also allows us to properly support fixup
pairs where the instructions are in different fragments.

Reviewers: asb, lenary, efriedma

Reviewed By: efriedma

Subscribers: arichardson, hiraditya, rbar, johnrusso, simoncook, sabuasal, niosHD, kito-cheng, shiva0217, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, PkmX, jocewei, psnobl, benna, Jim, s.egerton, pzheng, sameer.abuasal, apazos, luismarques, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D73211

(cherry picked from commit 3f5976c97dbfefb4669abcf968bd79a9a64c18e0)

5 files changed, 81 insertions, 17 deletions

diff --git a/llvm/test/MC/RISCV/pcrel-fixups.s b/llvm/test/MC/RISCV/pcrel-fixups.s
index 1025988967a..f4f9e2c57a0 100644
--- a/llvm/test/MC/RISCV/pcrel-fixups.s
+++ b/llvm/test/MC/RISCV/pcrel-fixups.s
@@ -12,11 +12,12 @@
 # RUN:    | FileCheck --check-prefix RELAX %s
 
 # Fixups for %pcrel_hi / %pcrel_lo can be evaluated within a section,
-# regardless of the fragment containing the target address.
+# regardless of the fragment containing the target address, provided symbol
+# binding allows it.
 
 function:
 .Lpcrel_label1:
-	auipc	a0, %pcrel_hi(other_function)
+	auipc	a0, %pcrel_hi(local_function)
 	addi	a1, a0, %pcrel_lo(.Lpcrel_label1)
 # NORELAX: auipc	a0, 0
 # NORELAX-NOT: R_RISCV
@@ -24,7 +25,7 @@ function:
 # NORELAX-NOT: R_RISCV
 
 # RELAX: auipc	a0, 0
-# RELAX: R_RISCV_PCREL_HI20	other_function
+# RELAX: R_RISCV_PCREL_HI20	local_function
 # RELAX: R_RISCV_RELAX	*ABS*
 # RELAX: addi	a1, a0, 0
 # RELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label1
@@ -32,7 +33,7 @@ function:
 
 	.p2align	2   # Cause a new fragment be emitted here
 .Lpcrel_label2:
-	auipc	a0, %pcrel_hi(other_function)
+	auipc	a0, %pcrel_hi(local_function)
 	addi	a1, a0, %pcrel_lo(.Lpcrel_label2)
 # NORELAX: auipc	a0, 0
 # NORELAX-NOT: R_RISCV
@@ -40,13 +41,72 @@ function:
 # NORELAX-NOT: R_RISCV
 
 # RELAX: auipc	a0, 0
-# RELAX: R_RISCV_PCREL_HI20	other_function
+# RELAX: R_RISCV_PCREL_HI20	local_function
 # RELAX: R_RISCV_RELAX	*ABS*
 # RELAX: addi	a1, a0, 0
 # RELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label2
 # RELAX: R_RISCV_RELAX	*ABS*
 
-	.type	other_function,@function
-other_function:
+	.type	local_function,@function
+local_function:
 	ret
 
+# Check we correctly evaluate when fixups are in different fragments
+
+.Lpcrel_label3:
+	auipc	a0, %pcrel_hi(local_function)
+	.p2align	2   # Cause a new fragment be emitted here
+	addi	a1, a0, %pcrel_lo(.Lpcrel_label3)
+# NORELAX: auipc	a0, 0
+# NORELAX-NOT: R_RISCV
+# NORELAX: addi	a1, a0, -4
+# NORELAX-NOT: R_RISCV
+
+# RELAX: auipc	a0, 0
+# RELAX: R_RISCV_PCREL_HI20	local_function
+# RELAX: R_RISCV_RELAX	*ABS*
+# RELAX: addi	a1, a0, 0
+# RELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label3
+# RELAX: R_RISCV_RELAX	*ABS*
+
+# Check handling of symbol binding.
+
+.Lpcrel_label4:
+	auipc	a0, %pcrel_hi(global_function)
+	addi	a1, a0, %pcrel_lo(.Lpcrel_label4)
+# NORELAX: auipc	a0, 0
+# NORELAX: R_RISCV_PCREL_HI20	global_function
+# NORELAX: addi	a1, a0, 0
+# NORELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label4
+
+# RELAX: auipc	a0, 0
+# RELAX: R_RISCV_PCREL_HI20	global_function
+# RELAX: R_RISCV_RELAX	*ABS*
+# RELAX: addi	a1, a0, 0
+# RELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label4
+# RELAX: R_RISCV_RELAX	*ABS*
+
+.Lpcrel_label5:
+	auipc	a0, %pcrel_hi(weak_function)
+	addi	a1, a0, %pcrel_lo(.Lpcrel_label5)
+# NORELAX: auipc	a0, 0
+# NORELAX: R_RISCV_PCREL_HI20	weak_function
+# NORELAX: addi	a1, a0, 0
+# NORELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label5
+
+# RELAX: auipc	a0, 0
+# RELAX: R_RISCV_PCREL_HI20	weak_function
+# RELAX: R_RISCV_RELAX	*ABS*
+# RELAX: addi	a1, a0, 0
+# RELAX: R_RISCV_PCREL_LO12_I	.Lpcrel_label5
+# RELAX: R_RISCV_RELAX	*ABS*
+
+	.global	global_function
+	.type	global_function,@function
+global_function:
+	ret
+
+	.weak	weak_function
+	.type	weak_function,@function
+weak_function:
+	ret
diff --git a/llvm/test/MC/RISCV/pcrel-lo12-invalid.s b/llvm/test/MC/RISCV/pcrel-lo12-invalid.s
index 7cf2494ad8e..74a1f2fac92 100644
--- a/llvm/test/MC/RISCV/pcrel-lo12-invalid.s
+++ b/llvm/test/MC/RISCV/pcrel-lo12-invalid.s
@@ -3,3 +3,5 @@
 
 1:
   addi a0, a0, %pcrel_lo(1b) # CHECK: :[[@LINE]]:3: error: could not find corresponding %pcrel_hi
+  addi a0, a0, %pcrel_lo(0x123456) # CHECK: :[[@LINE]]:3: error: could not find corresponding %pcrel_hi
+  addi a0, a0, %pcrel_lo(foo) # CHECK: :[[@LINE]]:3: error: could not find corresponding %pcrel_hi
diff --git a/llvm/test/MC/RISCV/rv32i-aliases-valid.s b/llvm/test/MC/RISCV/rv32i-aliases-valid.s
index d2e142efcfc..5140fb2adfe 100644
--- a/llvm/test/MC/RISCV/rv32i-aliases-valid.s
+++ b/llvm/test/MC/RISCV/rv32i-aliases-valid.s
@@ -14,6 +14,8 @@
 # CHECK-ALIAS....Match the alias (tests instr. to alias mapping)
 # CHECK-EXPAND...Match canonical instr. unconditionally (tests alias expansion)
 
+# Needed for testing valid %pcrel_lo expressions
+.Lpcrel_hi0: auipc a0, %pcrel_hi(foo)
 
 # CHECK-INST: addi a0, zero, 0
 # CHECK-ALIAS: mv a0, zero
@@ -71,16 +73,13 @@ li x12, 0xFFFFFFFF
 
 # CHECK-EXPAND: addi a0, zero, 1110
 li a0, %lo(0x123456)
-# CHECK-OBJ-NOALIAS: addi a0, zero, 0
-# CHECK-OBJ: R_RISCV_PCREL_LO12
-li a0, %pcrel_lo(0x123456)
 
 # CHECK-OBJ-NOALIAS: addi a0, zero, 0
 # CHECK-OBJ: R_RISCV_LO12
 li a0, %lo(foo)
 # CHECK-OBJ-NOALIAS: addi a0, zero, 0
 # CHECK-OBJ: R_RISCV_PCREL_LO12
-li a0, %pcrel_lo(foo)
+li a0, %pcrel_lo(.Lpcrel_hi0)
 
 .equ CONST, 0x123456
 # CHECK-EXPAND: lui a0, 291
diff --git a/llvm/test/MC/RISCV/rv32i-valid.s b/llvm/test/MC/RISCV/rv32i-valid.s
index bbcfa5feac4..580f0f548c3 100644
--- a/llvm/test/MC/RISCV/rv32i-valid.s
+++ b/llvm/test/MC/RISCV/rv32i-valid.s
@@ -11,6 +11,9 @@
 
 .equ CONST, 30
 
+# Needed for testing valid %pcrel_lo expressions
+.Lpcrel_hi0: auipc a0, %pcrel_hi(foo)
+
 # CHECK-ASM-AND-OBJ: lui a0, 2
 # CHECK-ASM: encoding: [0x37,0x25,0x00,0x00]
 lui a0, 2
@@ -161,11 +164,11 @@ lw a0, 97(a2)
 # CHECK-OBJ: lbu s5, 0(s6)
 # CHECK-OBJ: R_RISCV_LO12
 lbu s5, %lo(foo)(s6)
-# CHECK-ASM: lhu t3, %pcrel_lo(foo)(t3)
+# CHECK-ASM: lhu t3, %pcrel_lo(.Lpcrel_hi0)(t3)
 # CHECK-ASM: encoding: [0x03,0x5e,0bAAAA1110,A]
 # CHECK-OBJ: lhu t3, 0(t3)
 # CHECK-OBJ: R_RISCV_PCREL_LO12
-lhu t3, %pcrel_lo(foo)(t3)
+lhu t3, %pcrel_lo(.Lpcrel_hi0)(t3)
 # CHECK-ASM-AND-OBJ: lb t0, 30(t1)
 # CHECK-ASM: encoding: [0x83,0x02,0xe3,0x01]
 lb t0, CONST(t1)
diff --git a/llvm/test/MC/RISCV/rv64i-aliases-valid.s b/llvm/test/MC/RISCV/rv64i-aliases-valid.s
index 551e46f8530..9f0ad246cc9 100644
--- a/llvm/test/MC/RISCV/rv64i-aliases-valid.s
+++ b/llvm/test/MC/RISCV/rv64i-aliases-valid.s
@@ -17,6 +17,9 @@
 # TODO ld
 # TODO sd
 
+# Needed for testing valid %pcrel_lo expressions
+.Lpcrel_hi0: auipc a0, %pcrel_hi(foo)
+
 # CHECK-INST: addi a0, zero, 0
 # CHECK-ALIAS: mv a0, zero
 li x10, 0
@@ -107,16 +110,13 @@ li t5, 0xFFFFFFFFFFFFFFFF
 
 # CHECK-EXPAND: addi a0, zero, 1110
 li a0, %lo(0x123456)
-# CHECK-OBJ-NOALIAS: addi a0, zero, 0
-# CHECK-OBJ: R_RISCV_PCREL_LO12
-li a0, %pcrel_lo(0x123456)
 
 # CHECK-OBJ-NOALIAS: addi a0, zero, 0
 # CHECK-OBJ: R_RISCV_LO12
 li a0, %lo(foo)
 # CHECK-OBJ-NOALIAS: addi a0, zero, 0
 # CHECK-OBJ: R_RISCV_PCREL_LO12
-li a0, %pcrel_lo(foo)
+li a0, %pcrel_lo(.Lpcrel_hi0)
 
 .equ CONST, 0x123456
 # CHECK-EXPAND: lui a0, 291