The Linux kernel currently does not clear the direction flag before
calling a signal handler, whereas the x86/x86-64 ABI requires that.
Linux had this behavior/bug forever, but this becomes a real problem
with gcc version 4.3, which assumes that the direction flag is
correctly cleared at the entry of a function.
This patches changes the setup_frame() functions to clear the
direction before entering the signal handler.
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: H. Peter Anvin <hpa@zytor.com>
        regs->ss = __USER32_DS;
 
        set_fs(USER_DS);
-       regs->flags &= ~X86_EFLAGS_TF;
+       regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF);
        if (test_thread_flag(TIF_SINGLESTEP))
                ptrace_notify(SIGTRAP);
 
        regs->ss = __USER32_DS;
 
        set_fs(USER_DS);
-       regs->flags &= ~X86_EFLAGS_TF;
+       regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF);
        if (test_thread_flag(TIF_SINGLESTEP))
                ptrace_notify(SIGTRAP);
 
 
         * The tracer may want to single-step inside the
         * handler too.
         */
-       regs->flags &= ~TF_MASK;
+       regs->flags &= ~(TF_MASK | X86_EFLAGS_DF);
        if (test_thread_flag(TIF_SINGLESTEP))
                ptrace_notify(SIGTRAP);
 
         * The tracer may want to single-step inside the
         * handler too.
         */
-       regs->flags &= ~TF_MASK;
+       regs->flags &= ~(TF_MASK | X86_EFLAGS_DF);
        if (test_thread_flag(TIF_SINGLESTEP))
                ptrace_notify(SIGTRAP);
 
 
           see include/asm-x86_64/uaccess.h for details. */
        set_fs(USER_DS);
 
-       regs->flags &= ~X86_EFLAGS_TF;
+       regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF);
        if (test_thread_flag(TIF_SINGLESTEP))
                ptrace_notify(SIGTRAP);
 #ifdef DEBUG_SIG