diff --git a/llvm/lib/Target/AMDGPU/BUFInstructions.td b/llvm/lib/Target/AMDGPU/BUFInstructions.td
--- a/llvm/lib/Target/AMDGPU/BUFInstructions.td
+++ b/llvm/lib/Target/AMDGPU/BUFInstructions.td
@@ -332,6 +332,7 @@
   bits<4> elements    = 0;
   bits<1> has_sccb    = 1;
   bits<1> sccb_value  = 0;
+  bits<1> IsBufferInv = 0;
 }
 
 class MUBUF_Real <MUBUF_Pseudo ps> :
@@ -374,7 +375,8 @@
   let mayLoad = 0;
   let mayStore = 0;
 
-  // Set everything to 0.
+  let IsBufferInv = 1;
+  // Set everything else to 0.
   let offen       = 0;
   let idxen       = 0;
   let addr64      = 0;
@@ -2606,7 +2608,10 @@
 def MUBUFInfoTable : GenericTable {
   let FilterClass = "MUBUF_Pseudo";
   let CppTypeName = "MUBUFInfo";
-  let Fields = ["Opcode", "BaseOpcode", "elements", "has_vaddr", "has_srsrc", "has_soffset"];
+  let Fields = [
+    "Opcode", "BaseOpcode", "elements", "has_vaddr", "has_srsrc", "has_soffset",
+    "IsBufferInv"
+  ];
 
   let PrimaryKey = ["Opcode"];
   let PrimaryKeyName = "getMUBUFOpcodeHelper";
diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -27,6 +27,7 @@
 #include "GCNSubtarget.h"
 #include "MCTargetDesc/AMDGPUMCTargetDesc.h"
 #include "SIMachineFunctionInfo.h"
+#include "Utils/AMDGPUBaseInfo.h"
 #include "llvm/ADT/MapVector.h"
 #include "llvm/ADT/PostOrderIterator.h"
 #include "llvm/CodeGen/MachinePostDominators.h"
@@ -945,8 +946,10 @@
   AMDGPU::Waitcnt Wait;
   bool Modified = false;
 
-  // See if this instruction has a forced S_WAITCNT VM.
-  // TODO: Handle other cases of NeedsWaitcntVmBefore()
+  // FIXME: This should have already been handled by the memory legalizer.
+  // Removing this currently doesn't affect any lit tests, but we need to
+  // verify that nothing was relying on this. The number of buffer invalidates
+  // being handled here should not be expanded.
   if (MI.getOpcode() == AMDGPU::BUFFER_WBINVL1 ||
       MI.getOpcode() == AMDGPU::BUFFER_WBINVL1_SC ||
       MI.getOpcode() == AMDGPU::BUFFER_WBINVL1_VOL ||
@@ -1317,12 +1320,7 @@
     if (FlatASCount > 1)
       ScoreBrackets->setPendingFlat();
   } else if (SIInstrInfo::isVMEM(Inst) &&
-             // TODO: get a better carve out.
-             Inst.getOpcode() != AMDGPU::BUFFER_WBINVL1 &&
-             Inst.getOpcode() != AMDGPU::BUFFER_WBINVL1_SC &&
-             Inst.getOpcode() != AMDGPU::BUFFER_WBINVL1_VOL &&
-             Inst.getOpcode() != AMDGPU::BUFFER_GL0_INV &&
-             Inst.getOpcode() != AMDGPU::BUFFER_GL1_INV) {
+             !llvm::AMDGPU::getMUBUFIsBufferInv(Inst.getOpcode())) {
     if (!ST->hasVscnt())
       ScoreBrackets->updateByEvent(TII, TRI, MRI, VMEM_ACCESS, Inst);
     else if ((Inst.mayLoad() && !SIInstrInfo::isAtomicNoRet(Inst)) ||
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -394,6 +394,9 @@
 LLVM_READONLY
 bool getMUBUFHasSoffset(unsigned Opc);
 
+LLVM_READONLY
+bool getMUBUFIsBufferInv(unsigned Opc);
+
 LLVM_READONLY
 bool getSMEMIsBuffer(unsigned Opc);
 
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -162,6 +162,7 @@
   bool has_vaddr;
   bool has_srsrc;
   bool has_soffset;
+  bool IsBufferInv;
 };
 
 struct MTBUFInfo {
@@ -257,6 +258,11 @@
   return Info ? Info->has_soffset : false;
 }
 
+bool getMUBUFIsBufferInv(unsigned Opc) {
+  const MUBUFInfo *Info = getMUBUFOpcodeHelper(Opc);
+  return Info ? Info->IsBufferInv : false;
+}
+
 bool getSMEMIsBuffer(unsigned Opc) {
   const SMInfo *Info = getSMEMOpcodeHelper(Opc);
   return Info ? Info->IsBuffer : false;
diff --git a/llvm/test/CodeGen/AMDGPU/waitcnt-preexisting.mir b/llvm/test/CodeGen/AMDGPU/waitcnt-preexisting.mir
--- a/llvm/test/CodeGen/AMDGPU/waitcnt-preexisting.mir
+++ b/llvm/test/CodeGen/AMDGPU/waitcnt-preexisting.mir
@@ -192,6 +192,7 @@
     $vgpr0 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec, implicit $flat_scr
     S_WAITCNT 3952
     KILL $vgpr0
+...
 
 # Combine preexisting waitcnt with wait added to the start of a non-entry function.
 
@@ -206,3 +207,28 @@
     S_WAITCNT 0
     S_ENDPGM 0
 ...
+
+# Verify that extra waitcnt are not added after buffer invalidate instructions.
+
+---
+name:            test_waitcnt_preexisting_buffer_inv
+body:             |
+  bb.0:
+    ; GFX9-LABEL: name: test_waitcnt_preexisting_buffer_inv
+    ; GFX9: S_WAITCNT 0
+    ; GFX9: $vgpr0_vgpr1 = GLOBAL_LOAD_DWORDX2 $vgpr0_vgpr1, 0, 0, implicit $exec
+    ; GFX9: S_WAITCNT 3952
+    ; GFX9: BUFFER_INVL2 implicit $exec
+    ; GFX9: BUFFER_WBINVL1_VOL implicit $exec
+    ; GFX9: $vgpr0 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9: S_WAITCNT 112
+    ; GFX9: FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9: S_ENDPGM 0
+    $vgpr0_vgpr1 = GLOBAL_LOAD_DWORDX2 $vgpr0_vgpr1, 0, 0, implicit $exec
+    S_WAITCNT 3952
+    BUFFER_INVL2 implicit $exec
+    BUFFER_WBINVL1_VOL implicit $exec
+    $vgpr0 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec, implicit $flat_scr
+    FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr0, 0, 0, implicit $exec, implicit $flat_scr
+    S_ENDPGM 0
+...