diff --git a/clang/lib/Driver/ToolChains/HIPAMD.cpp b/clang/lib/Driver/ToolChains/HIPAMD.cpp
--- a/clang/lib/Driver/ToolChains/HIPAMD.cpp
+++ b/clang/lib/Driver/ToolChains/HIPAMD.cpp
@@ -152,6 +152,18 @@
 
   addLinkerCompressDebugSectionsOption(TC, Args, LldArgs);
 
+  // Given that host and device linking happen in separate processes, the device
+  // linker doesn't always have the visibility as to which device symbols are
+  // needed by a program, especially for the device symbol dependencies that are
+  // introduced through the host symbol resolution.
+  // For example: host_A() (A.obj) --> host_B(B.obj) --> device_kernel_B() (B.obj)
+  // In this case, the device linker doesn't know that A.obj actually depends on 
+  // the kernel functions in B.obj.  When linking to static device library, the 
+  // device linker may drop some of the device global symbols if they aren't
+  // referenced.  As a workaround, we are adding to the --whole-archive flag such
+  // that all global symbols would be linked in.
+  LldArgs.push_back("--whole-archive");
+
   for (auto *Arg : Args.filtered(options::OPT_Xoffload_linker)) {
     LldArgs.push_back(Arg->getValue(1));
     Arg->claim();
diff --git a/clang/test/Driver/hip-toolchain-rdc-separate.hip b/clang/test/Driver/hip-toolchain-rdc-separate.hip
--- a/clang/test/Driver/hip-toolchain-rdc-separate.hip
+++ b/clang/test/Driver/hip-toolchain-rdc-separate.hip
@@ -126,6 +126,7 @@
 // LINK-NOT: ".*llc"
 // LINK: {{".*lld.*"}} {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
 // LINK-SAME: "-plugin-opt=mcpu=gfx803"
+// LINK-SAME: "--whole-archive"
 // LLD-TMP-SAME: "-o" "[[IMG_DEV1:.*.out]]"
 // LLD-FIN-SAME: "-o" "[[IMG_DEV1:a.out-.*gfx803]]"
 // LINK-SAME "[[A_BC1]]" "[[B_BC1]]"
@@ -135,10 +136,12 @@
 // LINK-NOT: ".*llc"
 // LINK: {{".*lld.*"}} {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
 // LINK-SAME: "-plugin-opt=mcpu=gfx900"
+// LINK-SAME: "--whole-archive"
 // LLD-TMP-SAME: "-o" "[[IMG_DEV2:.*.out]]"
 // LLD-FIN-SAME: "-o" "[[IMG_DEV1:a.out-.*gfx900]]"
 // LINK-SAME "[[A_BC2]]" "[[B_BC2]]"
 
+
 // LINK-BUNDLE: [[BUNDLER:".*clang-offload-bundler"]] "-type=o"
 // LINK-BUNDLE-SAME: "-targets={{.*}},hipv4-amdgcn-amd-amdhsa--gfx803,hipv4-amdgcn-amd-amdhsa--gfx900"
 // LINK-BUNDLE-SAME: "-input={{.*}}" "-input=[[IMG_DEV1]]" "-input=[[IMG_DEV2]]" "-output=[[BUNDLE:.*]]"
diff --git a/clang/test/Driver/hip-toolchain-rdc-static-lib.hip b/clang/test/Driver/hip-toolchain-rdc-static-lib.hip
--- a/clang/test/Driver/hip-toolchain-rdc-static-lib.hip
+++ b/clang/test/Driver/hip-toolchain-rdc-static-lib.hip
@@ -80,6 +80,7 @@
 // CHECK-NOT: ".*llc"
 // CHECK: [[LLD]] {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
 // CHECK-SAME: "-plugin-opt=mcpu=gfx900"
+// CHECK-SAME: "--whole-archive"
 // CHECK-SAME: "-o" "[[IMG_DEV2:.*out]]" [[A_BC2]] [[B_BC2]]
 
 // combine images generated into hip fat binary object