2 files changed, 89 insertions, 0 deletions

diff --git a/src/cmd/internal/obj/x86/asm6.go b/src/cmd/internal/obj/x86/asm6.go
index 782b6d4aff..5e988eaf48 100644
--- a/src/cmd/internal/obj/x86/asm6.go
+++ b/src/cmd/internal/obj/x86/asm6.go
@@ -2036,6 +2036,31 @@ type nopPad struct {
 	n int32     // Size of the pad
 }
 
+// Padding bytes to add to align code as requested.
+// Alignment is restricted to powers of 2 between 8 and 2048 inclusive.
+//
+// pc: current offset in function, in bytes
+// a: requested alignment, in bytes
+// cursym: current function being assembled
+// returns number of bytes of padding needed
+func addpad(pc, a int64, ctxt *obj.Link, cursym *obj.LSym) int {
+	if !((a&(a-1) == 0) && 8 <= a && a <= 2048) {
+		ctxt.Diag("alignment value of an instruction must be a power of two and in the range [8, 2048], got %d\n", a)
+		return 0
+	}
+
+	// By default function alignment is 32 bytes for amd64
+	if cursym.Func().Align < int32(a) {
+		cursym.Func().Align = int32(a)
+	}
+
+	if pc&(a-1) != 0 {
+		return int(a - (pc & (a - 1)))
+	}
+
+	return 0
+}
+
 func span6(ctxt *obj.Link, s *obj.LSym, newprog obj.ProgAlloc) {
 	if ctxt.Retpoline && ctxt.Arch.Family == sys.I386 {
 		ctxt.Diag("-spectre=ret not supported on 386")
@@ -2119,6 +2144,19 @@ func span6(ctxt *obj.Link, s *obj.LSym, newprog obj.ProgAlloc) {
 			c0 := c
 			c = pjc.padJump(ctxt, s, p, c)
 
+			if p.As == obj.APCALIGN {
+				aln := p.From.Offset
+				v := addpad(int64(c), aln, ctxt, s)
+				if v > 0 {
+					s.Grow(int64(c) + int64(v))
+					fillnop(s.P[c:], int(v))
+				}
+
+				c += int32(v)
+				pPrev = p
+				continue
+			}
+
 			if maxLoopPad > 0 && p.Back&branchLoopHead != 0 && c&(loopAlign-1) != 0 {
 				// pad with NOPs
 				v := -c & (loopAlign - 1)
diff --git a/src/cmd/internal/obj/x86/asm_test.go b/src/cmd/internal/obj/x86/asm_test.go
index 36c8fce675..458a91258a 100644
--- a/src/cmd/internal/obj/x86/asm_test.go
+++ b/src/cmd/internal/obj/x86/asm_test.go
@@ -7,6 +7,10 @@ package x86
 import (
 	"cmd/internal/obj"
 	"cmd/internal/objabi"
+	"internal/testenv"
+	"os"
+	"path/filepath"
+	"regexp"
 	"testing"
 )
 
@@ -289,3 +293,50 @@ func TestRegIndex(t *testing.T) {
 		}
 	}
 }
+
+// TestPCALIGN verifies the correctness of the PCALIGN by checking if the
+// code can be aligned to the alignment value.
+func TestPCALIGN(t *testing.T) {
+	testenv.MustHaveGoBuild(t)
+	dir := t.TempDir()
+	tmpfile := filepath.Join(dir, "test.s")
+	tmpout := filepath.Join(dir, "test.o")
+
+	var testCases = []struct {
+		name string
+		code string
+		out  string
+	}{
+		{
+			name: "8-byte alignment",
+			code: "TEXT ·foo(SB),$0-0\nMOVQ $0, AX\nPCALIGN $8\nMOVQ $1, BX\nRET\n",
+			out:  `0x0008\s00008\s\(.*\)\tMOVQ\t\$1,\sBX`,
+		},
+		{
+			name: "16-byte alignment",
+			code: "TEXT ·foo(SB),$0-0\nMOVQ $0, AX\nPCALIGN $16\nMOVQ $2, CX\nRET\n",
+			out:  `0x0010\s00016\s\(.*\)\tMOVQ\t\$2,\sCX`,
+		},
+	}
+
+	for _, test := range testCases {
+		if err := os.WriteFile(tmpfile, []byte(test.code), 0644); err != nil {
+			t.Fatal(err)
+		}
+		cmd := testenv.Command(t, testenv.GoToolPath(t), "tool", "asm", "-S", "-o", tmpout, tmpfile)
+		cmd.Env = append(os.Environ(), "GOARCH=amd64", "GOOS=linux")
+		out, err := cmd.CombinedOutput()
+		if err != nil {
+			t.Errorf("The %s build failed: %v, output: %s", test.name, err, out)
+			continue
+		}
+
+		matched, err := regexp.MatchString(test.out, string(out))
+		if err != nil {
+			t.Fatal(err)
+		}
+		if !matched {
+			t.Errorf("The %s testing failed!\ninput: %s\noutput: %s\n", test.name, test.code, out)
+		}
+	}
+}