diff --git a/llvm/include/llvm/Analysis/TargetTransformInfo.h b/llvm/include/llvm/Analysis/TargetTransformInfo.h
--- a/llvm/include/llvm/Analysis/TargetTransformInfo.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfo.h
@@ -1643,6 +1643,9 @@
   /// false, but it shouldn't matter what it returns anyway.
   bool hasArmWideBranch(bool Thumb) const;
 
+  /// \return The maximum number of function arguments the target supports.
+  unsigned getMaxNumArgs() const;
+
   /// @}
 
 private:
@@ -2003,6 +2006,7 @@
   virtual VPLegalization
   getVPLegalizationStrategy(const VPIntrinsic &PI) const = 0;
   virtual bool hasArmWideBranch(bool Thumb) const = 0;
+  virtual unsigned getMaxNumArgs() const = 0;
 };
 
 template <typename T>
@@ -2694,6 +2698,10 @@
   bool hasArmWideBranch(bool Thumb) const override {
     return Impl.hasArmWideBranch(Thumb);
   }
+
+  unsigned getMaxNumArgs() const override {
+    return Impl.getMaxNumArgs();
+  }
 };
 
 template <typename T>
diff --git a/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h b/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
--- a/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
@@ -878,6 +878,8 @@
 
   bool hasArmWideBranch(bool) const { return false; }
 
+  unsigned getMaxNumArgs() const { return UINT_MAX; }
+
 protected:
   // Obtain the minimum required size to hold the value (without the sign)
   // In case of a vector it returns the min required size for one element.
diff --git a/llvm/lib/Analysis/TargetTransformInfo.cpp b/llvm/lib/Analysis/TargetTransformInfo.cpp
--- a/llvm/lib/Analysis/TargetTransformInfo.cpp
+++ b/llvm/lib/Analysis/TargetTransformInfo.cpp
@@ -1194,6 +1194,10 @@
   return TTIImpl->hasArmWideBranch(Thumb);
 }
 
+unsigned TargetTransformInfo::getMaxNumArgs() const {
+  return TTIImpl->getMaxNumArgs();
+}
+
 bool TargetTransformInfo::shouldExpandReduction(const IntrinsicInst *II) const {
   return TTIImpl->shouldExpandReduction(II);
 }
diff --git a/llvm/lib/Target/BPF/BPFTargetTransformInfo.h b/llvm/lib/Target/BPF/BPFTargetTransformInfo.h
--- a/llvm/lib/Target/BPF/BPFTargetTransformInfo.h
+++ b/llvm/lib/Target/BPF/BPFTargetTransformInfo.h
@@ -77,6 +77,10 @@
     return Options;
   }
 
+  unsigned getMaxNumArgs() const {
+    return 5;
+  }
+
 };
 
 } // end namespace llvm
diff --git a/llvm/lib/Transforms/IPO/ArgumentPromotion.cpp b/llvm/lib/Transforms/IPO/ArgumentPromotion.cpp
--- a/llvm/lib/Transforms/IPO/ArgumentPromotion.cpp
+++ b/llvm/lib/Transforms/IPO/ArgumentPromotion.cpp
@@ -781,6 +781,7 @@
   // Check to see which arguments are promotable.  If an argument is promotable,
   // add it to ArgsToPromote.
   DenseMap<Argument *, SmallVector<OffsetAndArgPart, 4>> ArgsToPromote;
+  unsigned NumArgsAfterPromote = F->getFunctionType()->getNumParams();
   for (Argument *PtrArg : PointerArgs) {
     // Replace sret attribute with noalias. This reduces register pressure by
     // avoiding a register copy.
@@ -804,6 +805,7 @@
         Types.push_back(Pair.second.Ty);
 
       if (areTypesABICompatible(Types, *F, TTI)) {
+        NumArgsAfterPromote += ArgParts.size() - 1;
         ArgsToPromote.insert({PtrArg, std::move(ArgParts)});
       }
     }
@@ -813,6 +815,9 @@
   if (ArgsToPromote.empty())
     return nullptr;
 
+  if (NumArgsAfterPromote > TTI.getMaxNumArgs())
+    return nullptr;
+
   return doPromotion(F, FAM, ArgsToPromote);
 }
 
diff --git a/llvm/test/CodeGen/BPF/argumentpromotion.ll b/llvm/test/CodeGen/BPF/argumentpromotion.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/BPF/argumentpromotion.ll
@@ -0,0 +1,145 @@
+; RUN: opt -O2 -mtriple=bpf-pc-linux -S %s | FileCheck %s
+; Source:
+;  struct t {
+;    int a, b, c, d, e, f, g;
+;  };
+;  __attribute__((noinline)) static int foo1(struct t *p1, struct t *p2, struct t *p3) {
+;    return p1->a + p1->b + p2->c + p2->e + p3->f + p3->g;
+;  }
+;  __attribute__((noinline)) static int foo2(struct t *p1, struct t *p2, struct t *p3) {
+;    return p1->a + p1->b + p2->c + p2->e + p3->f;
+;  }
+;  void init(void *);
+;  int bar(void) {
+;    struct t v1, v2, v3;
+;    init(&v1); init(&v2); init(&v3);
+;    return foo1(&v1, &v2, &v3) + foo2(&v1, &v2, &v3);
+;  }
+; Compilation flag:
+;   clang -target bpf -O2 -S -emit-llvm t.c -Xclang -disable-llvm-passes
+
+%struct.t = type { i32, i32, i32, i32, i32, i32, i32 }
+
+; Function Attrs: nounwind
+define dso_local i32 @bar() #0 {
+entry:
+  %v1 = alloca %struct.t, align 4
+  %v2 = alloca %struct.t, align 4
+  %v3 = alloca %struct.t, align 4
+  call void @llvm.lifetime.start.p0(i64 28, ptr %v1) #4
+  call void @llvm.lifetime.start.p0(i64 28, ptr %v2) #4
+  call void @llvm.lifetime.start.p0(i64 28, ptr %v3) #4
+  call void @init(ptr noundef %v1)
+  call void @init(ptr noundef %v2)
+  call void @init(ptr noundef %v3)
+  %call = call i32 @foo1(ptr noundef %v1, ptr noundef %v2, ptr noundef %v3)
+  %call1 = call i32 @foo2(ptr noundef %v1, ptr noundef %v2, ptr noundef %v3)
+  %add = add nsw i32 %call, %call1
+  call void @llvm.lifetime.end.p0(i64 28, ptr %v3) #4
+  call void @llvm.lifetime.end.p0(i64 28, ptr %v2) #4
+  call void @llvm.lifetime.end.p0(i64 28, ptr %v1) #4
+  ret i32 %add
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+declare dso_local void @init(ptr noundef) #2
+
+; Function Attrs: noinline nounwind
+define internal i32 @foo1(ptr noundef %p1, ptr noundef %p2, ptr noundef %p3) #3 {
+entry:
+  %p1.addr = alloca ptr, align 8
+  %p2.addr = alloca ptr, align 8
+  %p3.addr = alloca ptr, align 8
+  store ptr %p1, ptr %p1.addr, align 8, !tbaa !3
+  store ptr %p2, ptr %p2.addr, align 8, !tbaa !3
+  store ptr %p3, ptr %p3.addr, align 8, !tbaa !3
+  %0 = load ptr, ptr %p1.addr, align 8, !tbaa !3
+  %a = getelementptr inbounds %struct.t, ptr %0, i32 0, i32 0
+  %1 = load i32, ptr %a, align 4, !tbaa !7
+  %2 = load ptr, ptr %p1.addr, align 8, !tbaa !3
+  %b = getelementptr inbounds %struct.t, ptr %2, i32 0, i32 1
+  %3 = load i32, ptr %b, align 4, !tbaa !10
+  %add = add nsw i32 %1, %3
+  %4 = load ptr, ptr %p2.addr, align 8, !tbaa !3
+  %c = getelementptr inbounds %struct.t, ptr %4, i32 0, i32 2
+  %5 = load i32, ptr %c, align 4, !tbaa !11
+  %add1 = add nsw i32 %add, %5
+  %6 = load ptr, ptr %p2.addr, align 8, !tbaa !3
+  %e = getelementptr inbounds %struct.t, ptr %6, i32 0, i32 4
+  %7 = load i32, ptr %e, align 4, !tbaa !12
+  %add2 = add nsw i32 %add1, %7
+  %8 = load ptr, ptr %p3.addr, align 8, !tbaa !3
+  %f = getelementptr inbounds %struct.t, ptr %8, i32 0, i32 5
+  %9 = load i32, ptr %f, align 4, !tbaa !13
+  %add3 = add nsw i32 %add2, %9
+  %10 = load ptr, ptr %p3.addr, align 8, !tbaa !3
+  %g = getelementptr inbounds %struct.t, ptr %10, i32 0, i32 6
+  %11 = load i32, ptr %g, align 4, !tbaa !14
+  %add4 = add nsw i32 %add3, %11
+  ret i32 %add4
+}
+
+; CHECK:  i32 @foo1(ptr nocapture noundef readonly %p1, ptr nocapture noundef readonly %p2, ptr nocapture noundef readonly %p3)
+
+; Function Attrs: noinline nounwind
+define internal i32 @foo2(ptr noundef %p1, ptr noundef %p2, ptr noundef %p3) #3 {
+entry:
+  %p1.addr = alloca ptr, align 8
+  %p2.addr = alloca ptr, align 8
+  %p3.addr = alloca ptr, align 8
+  store ptr %p1, ptr %p1.addr, align 8, !tbaa !3
+  store ptr %p2, ptr %p2.addr, align 8, !tbaa !3
+  store ptr %p3, ptr %p3.addr, align 8, !tbaa !3
+  %0 = load ptr, ptr %p1.addr, align 8, !tbaa !3
+  %a = getelementptr inbounds %struct.t, ptr %0, i32 0, i32 0
+  %1 = load i32, ptr %a, align 4, !tbaa !7
+  %2 = load ptr, ptr %p1.addr, align 8, !tbaa !3
+  %b = getelementptr inbounds %struct.t, ptr %2, i32 0, i32 1
+  %3 = load i32, ptr %b, align 4, !tbaa !10
+  %add = add nsw i32 %1, %3
+  %4 = load ptr, ptr %p2.addr, align 8, !tbaa !3
+  %c = getelementptr inbounds %struct.t, ptr %4, i32 0, i32 2
+  %5 = load i32, ptr %c, align 4, !tbaa !11
+  %add1 = add nsw i32 %add, %5
+  %6 = load ptr, ptr %p2.addr, align 8, !tbaa !3
+  %e = getelementptr inbounds %struct.t, ptr %6, i32 0, i32 4
+  %7 = load i32, ptr %e, align 4, !tbaa !12
+  %add2 = add nsw i32 %add1, %7
+  %8 = load ptr, ptr %p3.addr, align 8, !tbaa !3
+  %f = getelementptr inbounds %struct.t, ptr %8, i32 0, i32 5
+  %9 = load i32, ptr %f, align 4, !tbaa !13
+  %add3 = add nsw i32 %add2, %9
+  ret i32 %add3
+}
+
+; CHECK:  i32 @foo2(i32 %p1.0.val, i32 %p1.4.val, i32 %p2.8.val, i32 %p2.16.val, i32 %p3.20.val)
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+attributes #0 = { nounwind "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #3 = { noinline nounwind "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #4 = { nounwind }
+
+!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 17.0.0 (https://github.com/llvm/llvm-project.git 48ad4296f7847db15b6bee7a465fc2cbe687ba6a)"}
+!3 = !{!4, !4, i64 0}
+!4 = !{!"any pointer", !5, i64 0}
+!5 = !{!"omnipotent char", !6, i64 0}
+!6 = !{!"Simple C/C++ TBAA"}
+!7 = !{!8, !9, i64 0}
+!8 = !{!"t", !9, i64 0, !9, i64 4, !9, i64 8, !9, i64 12, !9, i64 16, !9, i64 20, !9, i64 24}
+!9 = !{!"int", !5, i64 0}
+!10 = !{!8, !9, i64 4}
+!11 = !{!8, !9, i64 8}
+!12 = !{!8, !9, i64 16}
+!13 = !{!8, !9, i64 20}
+!14 = !{!8, !9, i64 24}