Attachment 2927 Details for Bug 2142 – Work around clock_gettime kernel bug on Linux x32

[patch] Work around clock_gettime kernel bug on Linux x32

0001-Work-around-clock_gettime-kernel-bug-on-Linux-x32.patch (text/plain), 1.09 KB, created by Colin Watson on 2017-01-04 01:48:31 AEDT

(hide)

Description:

Filename:

MIME Type:

Creator: Colin Watson

Created: 2017-01-04 01:48:31 AEDT

Size: 1.09 KB

patch

obsolete

>From e60388551d4e219673f2d90e749bf1fa63c93383 Mon Sep 17 00:00:00 2001
>From: Colin Watson <cjwatson@debian.org>
>Date: Tue, 3 Jan 2017 14:01:56 +0000
>Subject: [PATCH] Work around clock_gettime kernel bug on Linux x32
>
>On Linux x32, the clock_gettime VDSO currently falls back to the x86-64
>syscall, so allow that as well as its x32 sibling.
>---
> sandbox-seccomp-filter.c | 9 +++++++++
> 1 file changed, 9 insertions(+)
>
>diff --git a/sandbox-seccomp-filter.c b/sandbox-seccomp-filter.c
>index 2e1ed2c..62c578d 100644
>--- a/sandbox-seccomp-filter.c
>+++ b/sandbox-seccomp-filter.c
>@@ -137,6 +137,15 @@ static const struct sock_filter preauth_insns[] = {
> #endif
> #ifdef __NR_clock_gettime
> 	SC_ALLOW(clock_gettime),
>+# if defined(__x86_64__) && defined(__ILP32__)
>+	/* On Linux x32, the clock_gettime VDSO currently falls back to the
>+	 * x86-64 syscall (see https://bugs.debian.org/849923), so allow
>+	 * that too.
>+	 */
>+	BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K,
>+	    __NR_clock_gettime & ~__X32_SYSCALL_BIT, 0, 1),
>+	BPF_STMT(BPF_RET+BPF_K, SECCOMP_RET_ALLOW),
>+# endif
> #endif
> #ifdef __NR_close
> 	SC_ALLOW(close),
>-- 
>2.7.4
>

Actions: View | Diff

Attachments on bug 2142: 2328 | 2563 | 2927 | 2962