From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 1130) id 57B123857341; Tue, 20 Jun 2023 20:48:51 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 57B123857341 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1687294131; bh=FBIrqCahDwztwwJxtX7lLb0aYmiVUIbSW0tGy5XV5CY=; h=From:To:Subject:Date:From; b=ByrACU6UT+uyFsW72/N/RV6veD5rRjQwRlL3QBFU/OoJN3gXRhdOCi0RoPj/NQvay kgn4oSvSIRd0lMe3FEQFiANxgZtgz9RLmWzW+GtO8fDWDtnfhRnMPGgG7anovmswna GypdnXzRjPnSHR2zHsQ20FZocjRxaMiuisJvKTiM= MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="utf-8" From: Richard Sandiford To: gcc-cvs@gcc.gnu.org Subject: [gcc r14-2002] aarch64: Robustify stack tie handling X-Act-Checkin: gcc X-Git-Author: Richard Sandiford X-Git-Refname: refs/heads/trunk X-Git-Oldrev: f8f68c4ca622a24c2e8cf2b5f2f9fdcd47a7b369 X-Git-Newrev: 580b74a79146e51268dd11192d3870645adb0bbb Message-Id: <20230620204851.57B123857341@sourceware.org> Date: Tue, 20 Jun 2023 20:48:51 +0000 (GMT) List-Id: https://gcc.gnu.org/g:580b74a79146e51268dd11192d3870645adb0bbb commit r14-2002-g580b74a79146e51268dd11192d3870645adb0bbb Author: Richard Sandiford Date: Tue Jun 20 21:48:38 2023 +0100 aarch64: Robustify stack tie handling The SVE handling of stack clash protection copied the stack pointer to X11 before the probe and set up X11 as the CFA for unwind purposes: /* This is done to provide unwinding information for the stack adjustments we're about to do, however to prevent the optimizers from removing the R11 move and leaving the CFA note (which would be very wrong) we tie the old and new stack pointer together. The tie will expand to nothing but the optimizers will not touch the instruction. */ rtx stack_ptr_copy = gen_rtx_REG (Pmode, STACK_CLASH_SVE_CFA_REGNUM); emit_move_insn (stack_ptr_copy, stack_pointer_rtx); emit_insn (gen_stack_tie (stack_ptr_copy, stack_pointer_rtx)); /* We want the CFA independent of the stack pointer for the duration of the loop. */ add_reg_note (insn, REG_CFA_DEF_CFA, stack_ptr_copy); RTX_FRAME_RELATED_P (insn) = 1; -fcprop-registers is now smart enough to realise that X11 = SP, replace X11 with SP in the stack tie, and delete the instruction created above. This patch tries to prevent that by making stack_tie fussy about the register numbers. It fixes failures in gcc.target/aarch64/sve/pcs/stack_clash*.c. gcc/ * config/aarch64/aarch64.md (stack_tie): Hard-code the first register operand to the stack pointer. Require the second register operand to have the number specified in a separate const_int operand. * config/aarch64/aarch64.cc (aarch64_emit_stack_tie): New function. (aarch64_allocate_and_probe_stack_space): Use it. (aarch64_expand_prologue, aarch64_expand_epilogue): Likewise. (aarch64_expand_epilogue): Likewise. Diff: --- gcc/config/aarch64/aarch64.cc | 18 ++++++++++++++---- gcc/config/aarch64/aarch64.md | 7 ++++--- 2 files changed, 18 insertions(+), 7 deletions(-) diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc index ee37ceaa255..b99f12c99e9 100644 --- a/gcc/config/aarch64/aarch64.cc +++ b/gcc/config/aarch64/aarch64.cc @@ -9664,6 +9664,16 @@ aarch64_stack_clash_protection_alloca_probe_range (void) return STACK_CLASH_CALLER_GUARD; } +/* Emit a stack tie that acts as a scheduling barrier for all previous and + subsequent memory accesses and that requires the stack pointer and REG + to have their current values. REG can be stack_pointer_rtx if no + other register's value needs to be fixed. */ + +static void +aarch64_emit_stack_tie (rtx reg) +{ + emit_insn (gen_stack_tie (reg, gen_int_mode (REGNO (reg), DImode))); +} /* Allocate POLY_SIZE bytes of stack space using TEMP1 and TEMP2 as scratch registers. If POLY_SIZE is not large enough to require a probe this function @@ -9776,7 +9786,7 @@ aarch64_allocate_and_probe_stack_space (rtx temp1, rtx temp2, the instruction. */ rtx stack_ptr_copy = gen_rtx_REG (Pmode, STACK_CLASH_SVE_CFA_REGNUM); emit_move_insn (stack_ptr_copy, stack_pointer_rtx); - emit_insn (gen_stack_tie (stack_ptr_copy, stack_pointer_rtx)); + aarch64_emit_stack_tie (stack_ptr_copy); /* We want the CFA independent of the stack pointer for the duration of the loop. */ @@ -10145,7 +10155,7 @@ aarch64_expand_prologue (void) aarch64_add_cfa_expression (insn, regno_reg_rtx[reg1], hard_frame_pointer_rtx, 0); } - emit_insn (gen_stack_tie (stack_pointer_rtx, hard_frame_pointer_rtx)); + aarch64_emit_stack_tie (hard_frame_pointer_rtx); } aarch64_save_callee_saves (saved_regs_offset, R0_REGNUM, R30_REGNUM, @@ -10248,7 +10258,7 @@ aarch64_expand_epilogue (bool for_sibcall) || cfun->calls_alloca || crtl->calls_eh_return) { - emit_insn (gen_stack_tie (stack_pointer_rtx, stack_pointer_rtx)); + aarch64_emit_stack_tie (stack_pointer_rtx); need_barrier_p = false; } @@ -10287,7 +10297,7 @@ aarch64_expand_epilogue (bool for_sibcall) callee_adjust != 0, &cfi_ops); if (need_barrier_p) - emit_insn (gen_stack_tie (stack_pointer_rtx, stack_pointer_rtx)); + aarch64_emit_stack_tie (stack_pointer_rtx); if (callee_adjust != 0) aarch64_pop_regs (reg1, reg2, callee_adjust, &cfi_ops); diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md index 25f7905c6a0..01cf989641f 100644 --- a/gcc/config/aarch64/aarch64.md +++ b/gcc/config/aarch64/aarch64.md @@ -7325,10 +7325,11 @@ (define_insn "stack_tie" [(set (mem:BLK (scratch)) - (unspec:BLK [(match_operand:DI 0 "register_operand" "rk") - (match_operand:DI 1 "register_operand" "rk")] + (unspec:BLK [(reg:DI SP_REGNUM) + (match_operand:DI 0 "register_operand" "rk") + (match_operand:DI 1 "const_int_operand")] UNSPEC_PRLG_STK))] - "" + "REGNO (operands[0]) == INTVAL (operands[1])" "" [(set_attr "length" "0")] )