[PATCH] Uml support: reorganize PTRACE_SYSEMU support
authorBodo Stroesser <bstroesser@fujitsu-siemens.com>
Sat, 3 Sep 2005 22:57:19 +0000 (15:57 -0700)
committerLinus Torvalds <torvalds@evo.osdl.org>
Mon, 5 Sep 2005 07:06:20 +0000 (00:06 -0700)
With this patch, we change the way we handle switching from PTRACE_SYSEMU to
PTRACE_{SINGLESTEP,SYSCALL}, to free TIF_SYSCALL_EMU from double use as a
preparation for PTRACE_SYSEMU_SINGLESTEP extension, without changing the
behavior of the host kernel.

Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
arch/i386/kernel/entry.S
arch/i386/kernel/ptrace.c

index b389e5f..9a47723 100644 (file)
@@ -339,12 +339,18 @@ syscall_trace_entry:
        xorl %edx,%edx
        call do_syscall_trace
        cmpl $0, %eax
-       jne syscall_exit                # ret != 0 -> running under PTRACE_SYSEMU,
+       jne syscall_skip                # ret != 0 -> running under PTRACE_SYSEMU,
                                        # so must skip actual syscall
        movl ORIG_EAX(%esp), %eax
        cmpl $(nr_syscalls), %eax
        jnae syscall_call
        jmp syscall_exit
+syscall_skip:
+       cli                             # make sure we don't miss an interrupt
+                                       # setting need_resched or sigpending
+                                       # between sampling and the iret
+       movl TI_flags(%ebp), %ecx
+       jmp work_pending
 
        # perform syscall exit tracing
        ALIGN
index 5b569dc..18642f0 100644 (file)
@@ -515,21 +515,14 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
                ret = -EIO;
                if (!valid_signal(data))
                        break;
-               /* If we came here with PTRACE_SYSEMU and now continue with
-                * PTRACE_SYSCALL, entry.S used to intercept the syscall return.
-                * But it shouldn't!
-                * So we don't clear TIF_SYSCALL_EMU, which is always unused in
-                * this special case, to remember, we came from SYSEMU. That
-                * flag will be cleared by do_syscall_trace().
-                */
                if (request == PTRACE_SYSEMU) {
                        set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
-               } else if (request == PTRACE_CONT) {
-                       clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
-               }
-               if (request == PTRACE_SYSCALL) {
+                       clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+               } else if (request == PTRACE_SYSCALL) {
                        set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+                       clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
                } else {
+                       clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
                        clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
                }
                child->exit_code = data;
@@ -558,8 +551,7 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
                ret = -EIO;
                if (!valid_signal(data))
                        break;
-               /*See do_syscall_trace to know why we don't clear
-                * TIF_SYSCALL_EMU.*/
+               clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
                clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
                set_singlestep(child);
                child->exit_code = data;
@@ -694,7 +686,7 @@ void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, int error_code)
 __attribute__((regparm(3)))
 int do_syscall_trace(struct pt_regs *regs, int entryexit)
 {
-       int is_sysemu, is_systrace, is_singlestep, ret = 0;
+       int is_sysemu, is_singlestep, ret = 0;
        /* do the secure computing check first */
        secure_computing(regs->orig_eax);
 
@@ -705,25 +697,13 @@ int do_syscall_trace(struct pt_regs *regs, int entryexit)
                goto out;
 
        is_sysemu = test_thread_flag(TIF_SYSCALL_EMU);
-       is_systrace = test_thread_flag(TIF_SYSCALL_TRACE);
        is_singlestep = test_thread_flag(TIF_SINGLESTEP);
 
-       /* We can detect the case of coming from PTRACE_SYSEMU and now running
-        * with PTRACE_SYSCALL or PTRACE_SINGLESTEP, by TIF_SYSCALL_EMU being
-        * set additionally.
-        * If so let's reset the flag and return without action (no singlestep
-        * nor syscall tracing, since no actual step has been executed).
-        */
-       if (is_sysemu && (is_systrace || is_singlestep)) {
-               clear_thread_flag(TIF_SYSCALL_EMU);
-               goto out;
-       }
-
        /* Fake a debug trap */
-       if (test_thread_flag(TIF_SINGLESTEP))
+       if (is_singlestep)
                send_sigtrap(current, regs, 0);
 
-       if (!is_systrace && !is_sysemu)
+       if (!test_thread_flag(TIF_SYSCALL_TRACE) && !is_sysemu)
                goto out;
 
        /* the 0x80 provides a way for the tracing parent to distinguish
@@ -745,5 +725,10 @@ int do_syscall_trace(struct pt_regs *regs, int entryexit)
        if (unlikely(current->audit_context) && !entryexit)
                audit_syscall_entry(current, AUDIT_ARCH_I386, regs->orig_eax,
                                    regs->ebx, regs->ecx, regs->edx, regs->esi);
-       return ret;
+       if (ret == 0)
+               return 0;
+
+       if (unlikely(current->audit_context))
+               audit_syscall_exit(current, AUDITSC_RESULT(regs->eax), regs->eax);
+       return 1;
 }