Skip to contentSkip to conversation

i386 TIF flags for debug regs and io bitmap in ctxsw (v2)

PATCH Chuck Ebbert 2 participants 2 messages Jul 25, 2006

lore
Share & tools
lore
#0 lore raw

In-Reply-To: <[email protected]>

Quoted 24 lines · Stephane Eranian

I would like to follow-up the TIF flags and especially on the rules of inheritance. It appears the TIF flags are copied from parent to child task systematically on copy_process. Then they are adjusted in copy_threads() or sub-functions.

The TIF_IO_BITMAP is checked in copy_threads() with the following code:

int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
unsigned long unused,
struct task_struct * p, struct pt_regs * regs)

{
      ....
      if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
              p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL);
              if (!p->thread.io_bitmap_ptr) {
                      p->thread.io_bitmap_max = 0;
                      return -ENOMEM;
              }
              memcpy(p->thread.io_bitmap_ptr, tsk->thread.io_bitmap_ptr,
                      IO_BITMAP_BYTES);
              set_tsk_thread_flag(p, TIF_IO_BITMAP);
      }
Quoted 2 lines

I would think that the set_tsk_thread_flag() is extraneous.

Yeah. But harmless.

Quoted 2 lines

As for TIF_DEBUG, my patch is not clearing it. I don't think you can have HW breakpoints be inherited from one task to the other.

Looks like the debug regs get copied on fork and only cleared on exec in flush_thread(). So this should be OK. Please doublecheck.

(The new TIF_DEBUG flag went into 2.8.18-rc, in case you didn't notice.)

Signature · 1 line
Chuck
#1 lore raw

Chuck,

Quoted 7 lines · Chuck Ebbert
Quoted 3 lines

As for TIF_DEBUG, my patch is not clearing it. I don't think you can have HW breakpoints be inherited from one task to the other.

Looks like the debug regs get copied on fork and only cleared on exec in flush_thread(). So this should be OK. Please doublecheck.

How is this supposed to work? You can set debug registers via ptrace(). So A is ptracing B and sets up breakpoints in B. Now, if B forks C, C inherits the breakpoints of B. But what about the ptracing and re-parenting that is associated with this? If A does not know about C, I wonder how this could be work and be useful?

Quoted 1 line

(The new TIF_DEBUG flag went into 2.8.18-rc, in case you didn't notice.)

I did not keep track but thanks or this update. That's great!

Signature · 1 line
-Stephane