| 1 | /* Check if TLSDESC relocation preserves %rdi, %rsi and %rbx. |
|---|
| 2 | Copyright (C) 2024 Free Software Foundation, Inc. |
|---|
| 3 | This file is part of the GNU C Library. |
|---|
| 4 | |
|---|
| 5 | The GNU C Library is free software; you can redistribute it and/or |
|---|
| 6 | modify it under the terms of the GNU Lesser General Public |
|---|
| 7 | License as published by the Free Software Foundation; either |
|---|
| 8 | version 2.1 of the License, or (at your option) any later version. |
|---|
| 9 | |
|---|
| 10 | The GNU C Library is distributed in the hope that it will be useful, |
|---|
| 11 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
|---|
| 12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|---|
| 13 | Lesser General Public License for more details. |
|---|
| 14 | |
|---|
| 15 | You should have received a copy of the GNU Lesser General Public |
|---|
| 16 | License along with the GNU C Library; if not, see |
|---|
| 17 | <http://www.gnu.org/licenses/>. */ |
|---|
| 18 | |
|---|
| 19 | #include <sysdep.h> |
|---|
| 20 | |
|---|
| 21 | /* On AVX512 machines, OFFSET == 40 caused _dl_tlsdesc_dynamic_xsavec |
|---|
| 22 | to clobber %rdi, %rsi and %rbx. On Intel AVX CPUs, the state size |
|---|
| 23 | is 960 bytes and this test didn't fail. It may be due to the unused |
|---|
| 24 | last 128 bytes. On AMD AVX CPUs, the state size is 832 bytes and |
|---|
| 25 | this test might fail without the fix. */ |
|---|
| 26 | #ifndef OFFSET |
|---|
| 27 | # define OFFSET 40 |
|---|
| 28 | #endif |
|---|
| 29 | |
|---|
| 30 | .text |
|---|
| 31 | .p2align 4 |
|---|
| 32 | .globl apply_tls |
|---|
| 33 | .type apply_tls, @function |
|---|
| 34 | apply_tls: |
|---|
| 35 | cfi_startproc |
|---|
| 36 | _CET_ENDBR |
|---|
| 37 | pushq %rbp |
|---|
| 38 | cfi_def_cfa_offset (16) |
|---|
| 39 | cfi_offset (6, -16) |
|---|
| 40 | movdqu (%RDI_LP), %xmm0 |
|---|
| 41 | lea tls_var1@TLSDESC(%rip), %RAX_LP |
|---|
| 42 | mov %RSP_LP, %RBP_LP |
|---|
| 43 | cfi_def_cfa_register (6) |
|---|
| 44 | /* Align stack to 64 bytes. */ |
|---|
| 45 | and $-64, %RSP_LP |
|---|
| 46 | sub $OFFSET, %RSP_LP |
|---|
| 47 | pushq %rbx |
|---|
| 48 | /* Set %ebx to 0xbadbeef. */ |
|---|
| 49 | movl $0xbadbeef, %ebx |
|---|
| 50 | movl $0xbadbeef, %esi |
|---|
| 51 | movq %rdi, saved_rdi(%rip) |
|---|
| 52 | movq %rsi, saved_rsi(%rip) |
|---|
| 53 | call *tls_var1@TLSCALL(%RAX_LP) |
|---|
| 54 | /* Check if _dl_tlsdesc_dynamic preserves %rdi, %rsi and %rbx. */ |
|---|
| 55 | cmpq saved_rdi(%rip), %rdi |
|---|
| 56 | jne L(hlt) |
|---|
| 57 | cmpq saved_rsi(%rip), %rsi |
|---|
| 58 | jne L(hlt) |
|---|
| 59 | cmpl $0xbadbeef, %ebx |
|---|
| 60 | jne L(hlt) |
|---|
| 61 | add %fs:0, %RAX_LP |
|---|
| 62 | movups %xmm0, 32(%RAX_LP) |
|---|
| 63 | movdqu 16(%RDI_LP), %xmm1 |
|---|
| 64 | mov %RAX_LP, %RBX_LP |
|---|
| 65 | movups %xmm1, 48(%RAX_LP) |
|---|
| 66 | lea 32(%RBX_LP), %RAX_LP |
|---|
| 67 | pop %rbx |
|---|
| 68 | leave |
|---|
| 69 | cfi_def_cfa (7, 8) |
|---|
| 70 | ret |
|---|
| 71 | L(hlt): |
|---|
| 72 | hlt |
|---|
| 73 | cfi_endproc |
|---|
| 74 | .size apply_tls, .-apply_tls |
|---|
| 75 | .hidden tls_var1 |
|---|
| 76 | .globl tls_var1 |
|---|
| 77 | .section .tbss,"awT" ,@nobits |
|---|
| 78 | .align 16 |
|---|
| 79 | .type tls_var1, @object |
|---|
| 80 | .size tls_var1, 3200 |
|---|
| 81 | tls_var1: |
|---|
| 82 | .zero 3200 |
|---|
| 83 | .local saved_rdi |
|---|
| 84 | .comm saved_rdi,8,8 |
|---|
| 85 | .local saved_rsi |
|---|
| 86 | .comm saved_rsi,8,8 |
|---|
| 87 | .section .note.GNU-stack,"" ,@progbits |
|---|
| 88 | |
|---|
