As I said, I greatly misread asm code, -report_syscall_exit() can be
called without utrace_resume/utrace_get_signal. We need the stack of
stop events. Not the stack in fact, array.
Incomplete. The code needs cleanup/refactoring.
In particular, ptrace_report_clone() case is still not right. We
1. I assume that, on every arch, we alway check TIF_NOTIFY_RESUME
and TIF_SIGPENDING before TIF_SYSCALL_TRACE, right?
No, these are only checked when you would otherwise return to user mode.
IOW, can I assume that syscall_trace_leave() should not be called
On 09/14, Roland McGrath wrote:
But can't we move -reporting from utrace to engine? This saves a word
in task_struct. Yes, this adds the same word to engine, but we should
optimize the common untraced case.
That will be optimized properly when we move struct utrace out of task_struct.
On 09/14, Roland McGrath wrote:
This code has no comments whatsoever. I know you intend much more cleanup
to come anyway, but this can make it hard even to review the code and give
advice on the directions you are taking.
Yes. Will try to fix later.
I don't understand why you need the