diff --git a/lld/ELF/Arch/BPF.cpp b/lld/ELF/Arch/BPF.cpp
new file mode 100644
--- /dev/null
+++ b/lld/ELF/Arch/BPF.cpp
@@ -0,0 +1,113 @@
+//===---- BPF.cpp ---------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "Target.h"
+#include "llvm/Object/ELF.h"
+
+using namespace llvm;
+using namespace llvm::ELF;
+using namespace lld;
+using namespace lld::elf;
+
+namespace {
+class BPF final : public TargetInfo {
+public:
+  BPF();
+  int64_t getImplicitAddend(const uint8_t *buf, RelType type) const override;
+  int64_t getImplicitAddendAlloc(const uint8_t *buf,
+                                 RelType type) const override;
+  void relocate(uint8_t *loc, const Relocation &rel,
+                uint64_t val) const override;
+  void relocateAlloc(uint8_t *loc, const Relocation &rel,
+                     uint64_t val) const override;
+  RelType getDynRel(RelType type) const override;
+  RelExpr getRelExpr(RelType type, const Symbol &s,
+                     const uint8_t *loc) const override;
+};
+} // namespace
+
+BPF::BPF() { noneRel = R_BPF_NONE; }
+
+int64_t BPF::getImplicitAddend(const uint8_t *buf, RelType type) const {
+  switch (type) {
+  case R_BPF_NONE:
+    return 0;
+  case R_BPF_64_64:
+    return read64(buf);
+  case R_BPF_64_32:
+    return read32(buf);
+  default:
+    llvm_unreachable("unknown relocation");
+  }
+}
+
+void BPF::relocate(uint8_t *loc, const Relocation &rel, uint64_t val) const {
+  switch (rel.type) {
+  case R_BPF_NONE:
+    break;
+  case R_BPF_64_64:
+    write64(loc, val);
+    break;
+  case R_BPF_64_32:
+    write32(loc, (uint32_t)val);
+    break;
+  default:
+    llvm_unreachable("unknown relocation");
+  }
+}
+
+int64_t BPF::getImplicitAddendAlloc(const uint8_t *buf,
+                                    RelType type) const {
+  switch (type) {
+  case R_BPF_NONE:
+    return 0;
+  case R_BPF_64_64:
+    return read32(buf + 4);
+  case R_BPF_64_32:
+    // The addend is the (sec_offset/8 - 1) which could be negative.
+    return (int32_t)(uint32_t)(read32(buf + 4) * 8);
+  default:
+    llvm_unreachable("unknown relocation");
+  }
+}
+
+void BPF::relocateAlloc(uint8_t *loc, const Relocation &rel,
+                        uint64_t val) const {
+  switch (rel.type) {
+  case R_BPF_NONE:
+    break;
+  case R_BPF_64_64:
+    write32(loc + 4, (uint32_t)val);
+    break;
+  case R_BPF_64_32:
+    write32(loc + 4, (uint32_t)(val / 8));
+    break;
+  default:
+    llvm_unreachable("unknown relocation");
+  }
+}
+
+RelType BPF::getDynRel(RelType type) const { return R_BPF_NONE; }
+
+RelExpr BPF::getRelExpr(RelType type, const Symbol &s,
+                        const uint8_t *loc) const {
+  switch (type) {
+  case R_BPF_NONE:
+    return R_NONE;
+  case R_BPF_64_32:
+  case R_BPF_64_64:
+    return R_ABS;
+  default:
+    llvm_unreachable("unknown relocation");
+  }
+}
+
+TargetInfo *elf::getBPFTargetInfo() {
+  static BPF target;
+  return ⌖
+}
diff --git a/lld/ELF/CMakeLists.txt b/lld/ELF/CMakeLists.txt
--- a/lld/ELF/CMakeLists.txt
+++ b/lld/ELF/CMakeLists.txt
@@ -8,6 +8,7 @@
   Arch/AMDGPU.cpp
   Arch/ARM.cpp
   Arch/AVR.cpp
+  Arch/BPF.cpp
   Arch/Hexagon.cpp
   Arch/Mips.cpp
   Arch/MipsArchTree.cpp
diff --git a/lld/ELF/InputFiles.cpp b/lld/ELF/InputFiles.cpp
--- a/lld/ELF/InputFiles.cpp
+++ b/lld/ELF/InputFiles.cpp
@@ -1629,6 +1629,9 @@
     return EM_ARM;
   case Triple::avr:
     return EM_AVR;
+  case Triple::bpfel:
+  case Triple::bpfeb:
+    return EM_BPF;
   case Triple::mips:
   case Triple::mipsel:
   case Triple::mips64:
diff --git a/lld/ELF/InputSection.cpp b/lld/ELF/InputSection.cpp
--- a/lld/ELF/InputSection.cpp
+++ b/lld/ELF/InputSection.cpp
@@ -470,8 +470,12 @@
 
       int64_t addend = getAddend<ELFT>(rel);
       const uint8_t *bufLoc = sec->data().begin() + rel.r_offset;
-      if (!RelTy::IsRela)
-        addend = target->getImplicitAddend(bufLoc, type);
+      if (!RelTy::IsRela) {
+        if (sec->flags & SHF_ALLOC)
+          addend = target->getImplicitAddendAlloc(bufLoc, type);
+        else
+          addend = target->getImplicitAddend(bufLoc, type);
+      }
 
       if (config->emachine == EM_MIPS &&
           target->getRelExpr(type, sym, bufLoc) == R_MIPS_GOTREL) {
@@ -1054,7 +1058,7 @@
       // opportunities but is safe.
       if (ppc64noTocRelax.count({rel.sym, rel.addend}) ||
           !tryRelaxPPC64TocIndirection(rel, bufLoc))
-        target->relocate(bufLoc, rel, targetVA);
+        target->relocateAlloc(bufLoc, rel, targetVA);
       break;
     case R_RELAX_TLS_IE_TO_LE:
       target->relaxTlsIeToLe(bufLoc, rel, targetVA);
@@ -1097,10 +1101,10 @@
         }
         write32(bufLoc + 4, 0xe8410018); // ld %r2, 24(%r1)
       }
-      target->relocate(bufLoc, rel, targetVA);
+      target->relocateAlloc(bufLoc, rel, targetVA);
       break;
     default:
-      target->relocate(bufLoc, rel, targetVA);
+      target->relocateAlloc(bufLoc, rel, targetVA);
       break;
     }
   }
diff --git a/lld/ELF/LTO.cpp b/lld/ELF/LTO.cpp
--- a/lld/ELF/LTO.cpp
+++ b/lld/ELF/LTO.cpp
@@ -84,9 +84,10 @@
   c.Options.RelaxELFRelocations = true;
   c.Options.EmitAddrsig = true;
 
-  // Always emit a section per function/datum with LTO.
-  c.Options.FunctionSections = true;
-  c.Options.DataSections = true;
+  if (config->emachine != EM_BPF || !config->relocatable) {
+    c.Options.FunctionSections = true;
+    c.Options.DataSections = true;
+  }
 
   // Check if basic block sections must be used.
   // Allowed values for --lto-basic-block-sections are "all", "labels",
diff --git a/lld/ELF/Target.h b/lld/ELF/Target.h
--- a/lld/ELF/Target.h
+++ b/lld/ELF/Target.h
@@ -34,6 +34,10 @@
   virtual void writeGotPlt(uint8_t *buf, const Symbol &s) const {};
   virtual void writeIgotPlt(uint8_t *buf, const Symbol &s) const {}
   virtual int64_t getImplicitAddend(const uint8_t *buf, RelType type) const;
+  virtual int64_t getImplicitAddendAlloc(const uint8_t *buf,
+                                         RelType type) const {
+     return getImplicitAddend(buf, type);
+  }
   virtual int getTlsGdRelaxSkip(RelType type) const { return 1; }
 
   // If lazy binding is supported, the first entry of the PLT has code
@@ -87,6 +91,10 @@
   void relocateNoSym(uint8_t *loc, RelType type, uint64_t val) const {
     relocate(loc, Relocation{R_NONE, type, 0, 0, nullptr}, val);
   }
+  virtual void relocateAlloc(uint8_t *loc, const Relocation &rel,
+                             uint64_t val) const {
+    relocate(loc, rel, val);
+  }
 
   virtual void applyJumpInstrMod(uint8_t *loc, JumpModType type,
                                  JumpModType val) const {}
@@ -175,6 +183,7 @@
 TargetInfo *getAMDGPUTargetInfo();
 TargetInfo *getARMTargetInfo();
 TargetInfo *getAVRTargetInfo();
+TargetInfo *getBPFTargetInfo();
 TargetInfo *getHexagonTargetInfo();
 TargetInfo *getMSP430TargetInfo();
 TargetInfo *getPPC64TargetInfo();
diff --git a/lld/ELF/Target.cpp b/lld/ELF/Target.cpp
--- a/lld/ELF/Target.cpp
+++ b/lld/ELF/Target.cpp
@@ -60,6 +60,8 @@
     return getARMTargetInfo();
   case EM_AVR:
     return getAVRTargetInfo();
+  case EM_BPF:
+    return getBPFTargetInfo();
   case EM_HEXAGON:
     return getHexagonTargetInfo();
   case EM_MIPS:
diff --git a/lld/docs/index.rst b/lld/docs/index.rst
--- a/lld/docs/index.rst
+++ b/lld/docs/index.rst
@@ -33,8 +33,8 @@
   machine, you can expect that LLD runs more than twice as fast as the GNU
   gold linker. Your mileage may vary, though.
 
-- It supports various CPUs/ABIs including AArch64, AMDGPU, ARM, Hexagon, MIPS
-  32/64 big/little-endian, PowerPC, PowerPC64, RISC-V, SPARC V9, x86-32 and
+- It supports various CPUs/ABIs including AArch64, AMDGPU, ARM, BPF, Hexagon,
+  MIPS 32/64 big/little-endian, PowerPC, PowerPC64, RISC-V, SPARC V9, x86-32 and
   x86-64. Among these, AArch64, ARM (>= v6), PowerPC, PowerPC64, x86-32 and
   x86-64 have production quality. MIPS seems decent too.
 
diff --git a/lld/docs/ld.lld.1 b/lld/docs/ld.lld.1
--- a/lld/docs/ld.lld.1
+++ b/lld/docs/ld.lld.1
@@ -27,7 +27,7 @@
 as GNU linkers.
 .Pp
 .Nm
-currently supports i386, x86-64, ARM, AArch64, PowerPC32, PowerPC64,
+currently supports i386, x86-64, ARM, AArch64, BPF, PowerPC32, PowerPC64,
 MIPS32, MIPS64, RISC-V, AMDGPU, Hexagon and SPARC V9 targets.
 .Nm
 acts as a Microsoft link.exe-compatible linker if invoked as
diff --git a/lld/test/ELF/bpf-basic.s b/lld/test/ELF/bpf-basic.s
new file mode 100644
--- /dev/null
+++ b/lld/test/ELF/bpf-basic.s
@@ -0,0 +1,24 @@
+// REQUIRES: bpf
+// RUN: split-file %s %t
+// RUN: llc -filetype=obj -mtriple=bpfel %t/a.ll -o %t1.el.o
+// RUN: llc -filetype=obj -mtriple=bpfel %t/b.ll -o %t2.el.o
+// RUN: ld.lld -r %t1.el.o %t2.el.o -o %t3.el
+// RUN: llvm-objdump -d --no-show-raw-insn %t3.el | FileCheck %s
+// RUN: llc -filetype=obj -mtriple=bpfeb %t/a.ll -o %t1.eb.o
+// RUN: llc -filetype=obj -mtriple=bpfeb %t/b.ll -o %t2.eb.o
+// RUN: ld.lld -r %t1.eb.o %t2.eb.o -o %t3.eb
+// RUN: llvm-objdump -d --no-show-raw-insn %t5.eb | FileCheck %s
+
+// CHECK:  <foo>:
+// CHECK:  r0 = 0
+// CHECK:  exit
+
+// CHECK:  <bar>:
+// CHECK:  r0 = 1
+// CHECK:  exit
+
+//--- a.ll
+define dso_local i32 @foo() local_unnamed_addr { entry: ret i32 0 }
+
+//--- b.ll
+define dso_local i32 @bar() local_unnamed_addr { entry: ret i32 1 }
diff --git a/lld/test/ELF/lto/bpf-diff-sec.ll b/lld/test/ELF/lto/bpf-diff-sec.ll
new file mode 100644
--- /dev/null
+++ b/lld/test/ELF/lto/bpf-diff-sec.ll
@@ -0,0 +1,111 @@
+; REQUIRES: bpf
+; RUN: split-file %s %t
+; RUN: llvm-as %t/a.ll -o %t1.bc
+; RUN: llvm-as %t/b.ll -o %t2.bc
+; RUN: ld.lld -r %t1.bc %t2.bc -o %t3
+; RUN: llvm-objdump -d --no-show-raw-insn %t3 | FileCheck %s
+; RUN: llvm-readelf -r %t3 | FileCheck -check-prefix=RELOC %s
+
+; CHECK:       section .text:
+; CHECK:       r0 = 0 ll
+; CHECK:       section sec2:
+; CHECK:       call -1
+; CHECK:       call 2
+
+; RELOC:       .rel.text
+; RELOC:       R_BPF_64_64 {{[0-9a-f]+}} sec3
+; RELOC:       .relsec2
+; RELOC:       R_BPF_64_32 {{[0-9a-f]+}} sec1
+; RELOC:       R_BPF_64_32 {{[0-9a-f]+}} sec1
+
+;--- a.ll
+; Source code:
+;   static __attribute__((noinline)) __attribute__((section("sec1")))
+;   int add(int a, int b) {
+;     return a + b;
+;   }
+;   static __attribute__((noinline)) __attribute__((section("sec1")))
+;   int sub(int a, int b) {
+;     return a - b;
+;   }
+;   __attribute__((section("sec2")))
+;   int bar(int a, int b) {
+;     return add(a, b) + sub(a, b);
+;   }
+; Compilation flags:
+;   clang -target bpfel -flto -O2 -S t1.c -o t1.ll
+
+; ModuleID = 't1.c'
+source_filename = "t1.c"
+target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
+target triple = "bpfel"
+
+; Function Attrs: nofree norecurse nosync nounwind readnone willreturn mustprogress
+define dso_local i32 @bar(i32 %a, i32 %b) local_unnamed_addr #0 section "sec2" {
+entry:
+  %call = tail call fastcc i32 @add(i32 %a, i32 %b)
+  %call1 = tail call fastcc i32 @sub(i32 %a, i32 %b)
+  %add = add nsw i32 %call1, %call
+  ret i32 %add
+}
+
+; Function Attrs: nofree noinline norecurse nosync nounwind readnone willreturn mustprogress
+define internal fastcc i32 @add(i32 %a, i32 %b) unnamed_addr #1 section "sec1" {
+entry:
+  %add = add nsw i32 %b, %a
+  ret i32 %add
+}
+
+; Function Attrs: nofree noinline norecurse nosync nounwind readnone willreturn mustprogress
+define internal fastcc i32 @sub(i32 %a, i32 %b) unnamed_addr #1 section "sec1" {
+entry:
+  %sub = sub nsw i32 %a, %b
+  ret i32 %sub
+}
+
+attributes #0 = { nofree norecurse nosync nounwind readnone willreturn mustprogress "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #1 = { nofree noinline norecurse nosync nounwind readnone willreturn mustprogress "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+
+!llvm.module.flags = !{!0, !1}
+!llvm.ident = !{!2}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 7, !"frame-pointer", i32 2}
+!2 = !{!"clang version 13.0.0 (https://github.com/llvm/llvm-project.git c16909d7d1000413a32d3fcbc765dc082e8b62c0)"}
+
+;--- b.ll
+; Source code:
+;   static __attribute__((noinline)) __attribute__((section("sec3")))
+;   int test() {
+;     return 0;
+;   }
+;   long test2() { return (long)&test; }
+; Compilation flags:
+;   clang -target bpfel -flto -O2 -S t2.c -o t2.ll
+
+; ModuleID = 't2.c'
+source_filename = "t2.c"
+target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
+target triple = "bpfel"
+
+; Function Attrs: nofree norecurse nosync nounwind readnone willreturn mustprogress
+define dso_local i64 @test2() local_unnamed_addr #0 {
+entry:
+  ret i64 ptrtoint (i32 ()* @test to i64)
+}
+
+; Function Attrs: nofree noinline norecurse nosync nounwind readnone willreturn mustprogress
+define internal i32 @test() #1 section "sec3" {
+entry:
+  ret i32 0
+}
+
+attributes #0 = { nofree norecurse nosync nounwind readnone willreturn mustprogress "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #1 = { nofree noinline norecurse nosync nounwind readnone willreturn mustprogress "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+
+!llvm.module.flags = !{!0, !1}
+!llvm.ident = !{!2}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 7, !"frame-pointer", i32 2}
+!2 = !{!"clang version 13.0.0 (https://github.com/llvm/llvm-project.git c16909d7d1000413a32d3fcbc765dc082e8b62c0)"}
diff --git a/lld/test/ELF/lto/bpf-same-sec.ll b/lld/test/ELF/lto/bpf-same-sec.ll
new file mode 100644
--- /dev/null
+++ b/lld/test/ELF/lto/bpf-same-sec.ll
@@ -0,0 +1,178 @@
+; REQUIRES: bpf
+; RUN: split-file %s %t
+; RUN: llvm-as %t/a.ll -o %t1.bc
+; RUN: llvm-as %t/b.ll -o %t2.bc
+; RUN: ld.lld -r %t1.bc %t2.bc -o %t3
+; RUN: llvm-objdump -d --no-show-raw-insn %t3 | FileCheck %s
+; RUN: llvm-readelf -r %t3 | FileCheck -check-prefix=RELOC %s
+; RUN: llvm-readelf -x ".BTF" %t3 | FileCheck -check-prefix=BTF %s
+
+; CHECK:       <bar>:
+; CHECK:       call [[#]]
+; CHECK:       call -1
+; CHECK:       <add>:
+; CHECK:       <foo>:
+
+; RELOC:       R_BPF_64_32 {{[0-9a-f]+}} foo
+
+; '9feb0100' below is the first 4 bytes of the .BTF section,
+; representing BTF Magic value (0x9feb in little endian format),
+; version (0x1) and flags (0x0).
+
+; BTF:         9feb0100
+; BTF-NOT:     9feb0100
+
+;--- a.ll
+; Source code:
+;   struct s {
+;     int a;
+;     int b;
+;   };
+;   extern int foo(int);
+;   static __attribute__((noinline)) int add(int a, int b) {
+;     return a + b;
+;   }
+;   int bar(struct s *arg) {
+;     return add(arg->a, arg->b) + foo(arg->a);
+;   }
+; Compilation flags:
+;   clang -target bpfel -flto -O2 -g -S t1.c -o t1.ll
+
+; ModuleID = 't1.c'
+source_filename = "t1.c"
+target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
+target triple = "bpfel"
+
+%struct.s = type { i32, i32 }
+
+; Function Attrs: nounwind
+define dso_local i32 @bar(%struct.s* nocapture readonly %arg) local_unnamed_addr #0 !dbg !8 {
+entry:
+  call void @llvm.dbg.value(metadata %struct.s* %arg, metadata !18, metadata !DIExpression()), !dbg !19
+  %a = getelementptr inbounds %struct.s, %struct.s* %arg, i64 0, i32 0, !dbg !20
+  %0 = load i32, i32* %a, align 4, !dbg !20, !tbaa !21
+  %b = getelementptr inbounds %struct.s, %struct.s* %arg, i64 0, i32 1, !dbg !26
+  %1 = load i32, i32* %b, align 4, !dbg !26, !tbaa !27
+  %call = tail call fastcc i32 @add(i32 %0, i32 %1), !dbg !28
+  %call2 = tail call i32 @foo(i32 %0) #4, !dbg !29
+  %add = add nsw i32 %call2, %call, !dbg !30
+  ret i32 %add, !dbg !31
+}
+
+; Function Attrs: nofree noinline norecurse nosync nounwind readnone willreturn mustprogress
+define internal fastcc i32 @add(i32 %a, i32 %b) unnamed_addr #1 !dbg !32 {
+entry:
+  call void @llvm.dbg.value(metadata i32 %a, metadata !36, metadata !DIExpression()), !dbg !38
+  call void @llvm.dbg.value(metadata i32 %b, metadata !37, metadata !DIExpression()), !dbg !38
+  %add = add nsw i32 %b, %a, !dbg !39
+  ret i32 %add, !dbg !40
+}
+
+declare !dbg !41 dso_local i32 @foo(i32) local_unnamed_addr #2
+
+; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
+declare void @llvm.dbg.value(metadata, metadata, metadata) #3
+
+attributes #0 = { nounwind "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #1 = { nofree noinline norecurse nosync nounwind readnone willreturn mustprogress "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #2 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #3 = { nofree nosync nounwind readnone speculatable willreturn }
+attributes #4 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5, !6}
+!llvm.ident = !{!7}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 13.0.0 (https://github.com/llvm/llvm-project.git 433a5046a139601b86ac0f5a8ce89c79e153ff06)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "t1.c", directory: "/tmp/home/yhs/work/tests/llvm/lto")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{i32 7, !"frame-pointer", i32 2}
+!7 = !{!"clang version 13.0.0 (https://github.com/llvm/llvm-project.git 433a5046a139601b86ac0f5a8ce89c79e153ff06)"}
+!8 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 9, type: !9, scopeLine: 9, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !17)
+!9 = !DISubroutineType(types: !10)
+!10 = !{!11, !12}
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !13, size: 64)
+!13 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "s", file: !1, line: 1, size: 64, elements: !14)
+!14 = !{!15, !16}
+!15 = !DIDerivedType(tag: DW_TAG_member, name: "a", scope: !13, file: !1, line: 2, baseType: !11, size: 32)
+!16 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !13, file: !1, line: 3, baseType: !11, size: 32, offset: 32)
+!17 = !{!18}
+!18 = !DILocalVariable(name: "arg", arg: 1, scope: !8, file: !1, line: 9, type: !12)
+!19 = !DILocation(line: 0, scope: !8)
+!20 = !DILocation(line: 10, column: 19, scope: !8)
+!21 = !{!22, !23, i64 0}
+!22 = !{!"s", !23, i64 0, !23, i64 4}
+!23 = !{!"int", !24, i64 0}
+!24 = !{!"omnipotent char", !25, i64 0}
+!25 = !{!"Simple C/C++ TBAA"}
+!26 = !DILocation(line: 10, column: 27, scope: !8)
+!27 = !{!22, !23, i64 4}
+!28 = !DILocation(line: 10, column: 10, scope: !8)
+!29 = !DILocation(line: 10, column: 32, scope: !8)
+!30 = !DILocation(line: 10, column: 30, scope: !8)
+!31 = !DILocation(line: 10, column: 3, scope: !8)
+!32 = distinct !DISubprogram(name: "add", scope: !1, file: !1, line: 6, type: !33, scopeLine: 6, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !35)
+!33 = !DISubroutineType(types: !34)
+!34 = !{!11, !11, !11}
+!35 = !{!36, !37}
+!36 = !DILocalVariable(name: "a", arg: 1, scope: !32, file: !1, line: 6, type: !11)
+!37 = !DILocalVariable(name: "b", arg: 2, scope: !32, file: !1, line: 6, type: !11)
+!38 = !DILocation(line: 0, scope: !32)
+!39 = !DILocation(line: 7, column: 12, scope: !32)
+!40 = !DILocation(line: 7, column: 3, scope: !32)
+!41 = !DISubprogram(name: "foo", scope: !1, file: !1, line: 5, type: !42, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!42 = !DISubroutineType(types: !43)
+!43 = !{!11, !11}
+
+;--- b.ll
+; Source code:
+;   int __attribute__((noinline)) foo(int a) {
+;     return a * a;
+;   }
+; Compilation flags:
+;   clang -target bpfel -flto -O2 -g -S t2.c -o t2.ll
+
+; ModuleID = 't2.c'
+source_filename = "t2.c"
+target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
+target triple = "bpfel"
+
+; Function Attrs: nofree noinline norecurse nosync nounwind readnone willreturn mustprogress
+define dso_local i32 @foo(i32 %a) local_unnamed_addr #0 !dbg !8 {
+entry:
+  call void @llvm.dbg.value(metadata i32 %a, metadata !13, metadata !DIExpression()), !dbg !14
+  %mul = mul nsw i32 %a, %a, !dbg !15
+  ret i32 %mul, !dbg !16
+}
+
+; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
+declare void @llvm.dbg.value(metadata, metadata, metadata) #1
+
+attributes #0 = { nofree noinline norecurse nosync nounwind readnone willreturn mustprogress "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #1 = { nofree nosync nounwind readnone speculatable willreturn }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5, !6}
+!llvm.ident = !{!7}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 13.0.0 (https://github.com/llvm/llvm-project.git 433a5046a139601b86ac0f5a8ce89c79e153ff06)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "t2.c", directory: "/tmp/home/yhs/work/tests/llvm/lto")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{i32 7, !"frame-pointer", i32 2}
+!7 = !{!"clang version 13.0.0 (https://github.com/llvm/llvm-project.git 433a5046a139601b86ac0f5a8ce89c79e153ff06)"}
+!8 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !9, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!9 = !DISubroutineType(types: !10)
+!10 = !{!11, !11}
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13}
+!13 = !DILocalVariable(name: "a", arg: 1, scope: !8, file: !1, line: 1, type: !11)
+!14 = !DILocation(line: 0, scope: !8)
+!15 = !DILocation(line: 2, column: 12, scope: !8)
+!16 = !DILocation(line: 2, column: 3, scope: !8)
diff --git a/lld/test/lit.cfg.py b/lld/test/lit.cfg.py
--- a/lld/test/lit.cfg.py
+++ b/lld/test/lit.cfg.py
@@ -69,6 +69,7 @@
                           'AMDGPU': 'amdgpu',
                           'ARM': 'arm',
                           'AVR': 'avr',
+                          'BPF': 'bpf',
                           'Hexagon': 'hexagon',
                           'Mips': 'mips',
                           'MSP430': 'msp430',