winch: Implement check_stack for aarch64 (#10763)

* winch: Implement check_stack for aarch64

* winch: update aarch64 snapshot tests for check_stack change

* winch: address review comments on aarch64 check_stack

* winch: update aarch64 snapshot tests: check_stack uses unsigned compare

* winch: update aarch64 snapshot tests: stack grows down

* winch: aarch64 check_stack correct direction and alignment

* winch: aarch64 enable stack_stack wast checks

Author: graydon

cranelift/codegen/src/isa/aarch64/inst/emit.rs

@@ -107,13 +107,8 @@ fn machreg_to_gpr_or_vec(m: Reg) -> u32 {
     u32::from(m.to_real_reg().unwrap().hw_enc() & 31)
 }
 
-pub(crate) fn enc_arith_rrr(
-    bits_31_21: u32,
-    bits_15_10: u32,
-    rd: Writable<Reg>,
-    rn: Reg,
-    rm: Reg,
-) -> u32 {
+/// Encode a 3-register aeithmeric instruction.
+pub fn enc_arith_rrr(bits_31_21: u32, bits_15_10: u32, rd: Writable<Reg>, rn: Reg, rm: Reg) -> u32 {
     (bits_31_21 << 21)
         | (bits_15_10 << 10)
         | machreg_to_gpr(rd.to_reg())
@@ -204,7 +199,13 @@ fn enc_test_bit_and_branch(
         | machreg_to_gpr(reg)
 }
 
-fn enc_move_wide(op: MoveWideOp, rd: Writable<Reg>, imm: MoveWideConst, size: OperandSize) -> u32 {
+/// Encode a move-wide instruction.
+pub fn enc_move_wide(
+    op: MoveWideOp,
+    rd: Writable<Reg>,
+    imm: MoveWideConst,
+    size: OperandSize,
+) -> u32 {
     assert!(imm.shift <= 0b11);
     let op = match op {
         MoveWideOp::MovN => 0b00,
@@ -218,7 +219,8 @@ fn enc_move_wide(op: MoveWideOp, rd: Writable<Reg>, imm: MoveWideConst, size: Op
         | machreg_to_gpr(rd.to_reg())
 }
 
-fn enc_movk(rd: Writable<Reg>, imm: MoveWideConst, size: OperandSize) -> u32 {
+/// Encode a move-keep immediate instruction.
+pub fn enc_movk(rd: Writable<Reg>, imm: MoveWideConst, size: OperandSize) -> u32 {
     assert!(imm.shift <= 0b11);
     0x72800000
         | size.sf_bit() << 31

crates/cranelift/src/compiler.rs

@@ -225,7 +225,7 @@ impl wasmtime_environ::Compiler for Compiler {
         // check to all functions for how much native stack is remaining. The
         // `VMContext` pointer is the first argument to all functions, and the
         // first field of this structure is `*const VMStoreContext` and the
-        // first field of that is the stack limit. Note that the stack limit in
+        // third field of that is the stack limit. Note that the stack limit in
         // this case means "if the stack pointer goes below this, trap". Each
         // function which consumes stack space or isn't a leaf function starts
         // off by loading the stack limit, checking it against the stack

crates/test-util/src/wast.rs

@@ -418,11 +418,6 @@ impl WastTest {
         if config.compiler == Compiler::Winch {
             if cfg!(target_arch = "aarch64") {
                 let unsupported = [
-                    // Missing stack checks (#8321)
-                    "spec_testsuite/call.wast",
-                    "spec_testsuite/fac.wast",
-                    "spec_testsuite/skip-stack-guard-page.wast",
-                    "misc_testsuite/stack_overflow.wast",
                     // Fails intermittently
                     "misc_testsuite/table_copy_on_imported_tables.wast",
                     // Segfault

tests/disas/winch/aarch64/br/as_br_if_cond.wat

@@ -10,7 +10,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x10
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x58
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x10
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #8]
@@ -21,3 +28,4 @@
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   58: .byte   0x1f, 0xc1, 0x00, 0x00

tests/disas/winch/aarch64/br/as_br_value.wat

@@ -10,7 +10,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x10
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x60
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x10
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #8]
@@ -23,3 +30,4 @@
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   60: .byte   0x1f, 0xc1, 0x00, 0x00

tests/disas/winch/aarch64/br/as_if_cond.wat

@@ -15,7 +15,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x10
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x60
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x10
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #8]
@@ -28,3 +35,4 @@
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   60: .byte   0x1f, 0xc1, 0x00, 0x00

tests/disas/winch/aarch64/br/as_if_else.wat

@@ -15,7 +15,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x18
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x80
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x18
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #0x10]
@@ -24,15 +31,16 @@
 ;;       stur    w3, [x28]
 ;;       ldur    w0, [x28, #4]
 ;;       tst     w0, w0
-;;       b.eq    #0x44
-;;       b       #0x3c
-;;   3c: ldur    w0, [x28]
-;;       b       #0x4c
-;;   44: mov     x16, #4
+;;       b.eq    #0x60
+;;       b       #0x58
+;;   58: ldur    w0, [x28]
+;;       b       #0x68
+;;   60: mov     x16, #4
 ;;       mov     w0, w16
 ;;       add     x28, x28, #0x18
 ;;       mov     sp, x28
 ;;       mov     sp, x28
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   80: .byte   0x1f, 0xc1, 0x00, 0x00

tests/disas/winch/aarch64/br/as_if_then.wat

@@ -15,7 +15,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x18
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x80
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x18
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #0x10]
@@ -24,15 +31,16 @@
 ;;       stur    w3, [x28]
 ;;       ldur    w0, [x28, #4]
 ;;       tst     w0, w0
-;;       b.eq    #0x48
-;;       b       #0x3c
-;;   3c: mov     x16, #3
+;;       b.eq    #0x64
+;;       b       #0x58
+;;   58: mov     x16, #3
 ;;       mov     w0, w16
-;;       b       #0x4c
-;;   48: ldur    w0, [x28]
+;;       b       #0x68
+;;   64: ldur    w0, [x28]
 ;;       add     x28, x28, #0x18
 ;;       mov     sp, x28
 ;;       mov     sp, x28
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   80: .byte   0x1f, 0xc1, 0x00, 0x00

tests/disas/winch/aarch64/br/as_loop_first.wat

@@ -11,7 +11,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x10
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x60
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x10
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #8]
@@ -24,3 +31,4 @@
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   60: .byte   0x1f, 0xc1, 0x00, 0x00

tests/disas/winch/aarch64/br/br_jump.wat

@@ -18,7 +18,14 @@
 ;;       mov     x29, sp
 ;;       str     x28, [sp, #-0x10]!
 ;;       mov     x28, sp
-;;       mov     x9, x0
+;;       ldur    x16, [x0, #8]
+;;       ldur    x16, [x16, #0x10]
+;;       mov     x17, #0
+;;       movk    x17, #0x20
+;;       add     x16, x16, x17
+;;       cmp     sp, x16
+;;       b.lo    #0x8c
+;;   2c: mov     x9, x0
 ;;       sub     x28, x28, #0x18
 ;;       mov     sp, x28
 ;;       stur    x0, [x28, #0x10]
@@ -35,10 +42,11 @@
 ;;       stur    w16, [x28]
 ;;       add     x28, x28, #4
 ;;       mov     sp, x28
-;;       b       #0x3c
-;;   58: add     x28, x28, #0x18
+;;       b       #0x58
+;;   74: add     x28, x28, #0x18
 ;;       mov     sp, x28
 ;;       mov     sp, x28
 ;;       ldr     x28, [sp], #0x10
 ;;       ldp     x29, x30, [sp], #0x10
 ;;       ret
+;;   8c: .byte   0x1f, 0xc1, 0x00, 0x00