public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: Adhemerval Zanella <adhemerval.zanella@linaro.org>
To: libc-alpha@sourceware.org
Subject: [PATCH v5 04/22] i386: Fix Race conditions in pthread cancellation [BZ#12683]
Date: Mon, 10 Apr 2023 16:58:49 -0300	[thread overview]
Message-ID: <20230410195907.4123869-5-adhemerval.zanella@linaro.org> (raw)
In-Reply-To: <20230410195907.4123869-1-adhemerval.zanella@linaro.org>

The syscall bridge uses the old int80 instruction because by using
the optimized vDSO symbol the resulting PC value for an interrupted
syscall points to an adress outside the expected markers in
__syscall_cancel_arch.  It has been discussed in LKML [1] on how
kernel could help userland to accomplish it, but afaik discussion
has stalled.

Also, sysenter should not be used directly by libc since its calling
convention is set by the kernel depending of the underlying x86 chip
(check kernel commit 30bfa7b3488bfb1bb75c9f50a5fcac1832970c60).

Checked on i686-linux-gnu.

[1] https://lkml.org/lkml/2016/3/8/1105
---
 sysdeps/unix/sysv/linux/i386/syscall_cancel.S | 103 ++++++++++++++++++
 1 file changed, 103 insertions(+)
 create mode 100644 sysdeps/unix/sysv/linux/i386/syscall_cancel.S

diff --git a/sysdeps/unix/sysv/linux/i386/syscall_cancel.S b/sysdeps/unix/sysv/linux/i386/syscall_cancel.S
new file mode 100644
index 0000000000..7d1560e18f
--- /dev/null
+++ b/sysdeps/unix/sysv/linux/i386/syscall_cancel.S
@@ -0,0 +1,103 @@
+/* Cancellable syscall wrapper.  Linux/i686 version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <sysdep.h>
+#include <descr-const.h>
+
+/* long int [eax] __syscall_cancel_arch (int *cancelhandling [SP],
+					 long int nr   [SP+4],
+					 long int arg1 [SP+8],
+					 long int arg2 [SP+12],
+					 long int arg3 [SP+16],
+					 long int arg4 [SP+20],
+					 long int arg5 [SP+24],
+					 long int arg6 [SP+28])  */
+
+ENTRY (__syscall_cancel_arch)
+	pushl %ebp
+	cfi_def_cfa_offset (8)
+	cfi_offset (ebp, -8)
+	pushl %edi
+	cfi_def_cfa_offset (12)
+	cfi_offset (edi, -12)
+	pushl %esi
+	cfi_def_cfa_offset (16)
+	cfi_offset (esi, -16)
+	pushl %ebx
+	cfi_def_cfa_offset (20)
+	cfi_offset (ebx, -20)
+
+	.global __syscall_cancel_arch_start
+__syscall_cancel_arch_start:
+
+	/* if (*cancelhandling & CANCELED_BITMASK)
+	     __syscall_do_cancel()  */
+	testb	$TCB_CANCELED_BITMASK, (%eax)
+	jne     1f
+
+	/* Issue a 6 argument syscall, the nr [%eax] being the syscall
+	   number.  */
+	movl    24(%esp), %eax
+	movl    28(%esp), %ebx
+	movl    32(%esp), %ecx
+	movl    36(%esp), %edx
+	movl    40(%esp), %esi
+	movl    44(%esp), %edi
+	movl    48(%esp), %ebp
+
+	/* We can not use the vDSO helper for syscall (__kernel_vsyscall)
+	   because the returned PC from kernel will point to the vDSO page
+	   instead of the expected __syscall_cancel_arch_{start,end}
+	   marks.  */
+	int	$128
+
+	.global __syscall_cancel_arch_end
+__syscall_cancel_arch_end:
+
+	popl %ebx
+	cfi_restore (ebx)
+	cfi_def_cfa_offset (16)
+	popl %esi
+	cfi_restore (esi)
+	cfi_def_cfa_offset (12)
+	popl %edi
+	cfi_restore (edi)
+	cfi_def_cfa_offset (8)
+	popl %ebp
+	cfi_restore (ebp)
+	cfi_def_cfa_offset (4)
+        ret
+
+1:
+	/* Although the __syscall_do_cancel do not return, we need to stack
+	   being set correctly for unwind.  */
+	popl %ebx
+	cfi_restore (ebx)
+	cfi_def_cfa_offset (16)
+	popl %esi
+	cfi_restore (esi)
+	cfi_def_cfa_offset (12)
+	popl %edi
+	cfi_restore (edi)
+	cfi_def_cfa_offset (8)
+	popl %ebp
+	cfi_restore (ebp)
+	cfi_def_cfa_offset (4)
+	jmp __syscall_do_cancel
+
+END (__syscall_cancel_arch)
-- 
2.34.1


  parent reply	other threads:[~2023-04-10 19:59 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-04-10 19:58 [PATCH v5 00/22] " Adhemerval Zanella
2023-04-10 19:58 ` [PATCH v5 01/22] nptl: " Adhemerval Zanella
2023-04-10 19:58 ` [PATCH v5 02/22] x86_64: " Adhemerval Zanella
2023-04-10 19:58 ` [PATCH v5 03/22] x32: " Adhemerval Zanella
2023-04-10 19:58 ` Adhemerval Zanella [this message]
2023-04-10 19:58 ` [PATCH v5 05/22] aarch64: " Adhemerval Zanella
2023-04-10 19:58 ` [PATCH v5 06/22] arm: " Adhemerval Zanella
2023-04-10 19:58 ` [PATCH v5 07/22] powerpc: " Adhemerval Zanella
2023-04-14 15:27   ` Paul E Murphy
2023-04-27 11:13     ` Adhemerval Zanella Netto
2023-04-10 19:58 ` [PATCH v5 08/22] sparc: " Adhemerval Zanella
2023-04-10 20:46 ` [PATCH v5 09/22] s390: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 10/22] riscv: " Adhemerval Zanella
2023-04-11  6:35     ` Xi Ruoyao
2023-04-11  7:01       ` Xi Ruoyao
2023-04-11 13:49         ` Adhemerval Zanella Netto
2023-04-10 20:46   ` [PATCH v5 11/22] arc: Fix Race conditions in pthread cancellation [BZ #12683] Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 12/22] ia64: Fix Race conditions in pthread cancellation [BZ#12683] Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 13/22] sh: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 14/22] nios2: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 15/22] microblaze: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 16/22] hppa: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 17/22] m68k: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 18/22] alpha: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 19/22] csky: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 20/22] mips: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 21/22] or1k: " Adhemerval Zanella
2023-04-10 20:46   ` [PATCH v5 22/22] loongarch: " Adhemerval Zanella
2023-04-11  6:19     ` Xi Ruoyao
2023-04-11 12:00       ` caiyinyu
2023-04-11 12:21         ` caiyinyu
2023-04-11 12:55         ` Xi Ruoyao
2023-04-11 13:54           ` Adhemerval Zanella Netto
2023-04-11 13:56   ` [PATCH v5 09/22] s390: " Stefan Liebler

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230410195907.4123869-5-adhemerval.zanella@linaro.org \
    --to=adhemerval.zanella@linaro.org \
    --cc=libc-alpha@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).