Re: [RFC PATCH v4 4/6] uretprobes: return probe entry, prepare uretprobe

[Oleg Nesterov <oleg@redhat.com>]

On 03/04, Anton Arapov wrote:
>
> @@ -1085,6 +1093,7 @@ static int xol_add_vma(struct xol_area *area)
>  {
>  	struct mm_struct *mm = current->mm;
>  	int ret = -EALREADY;
> +	uprobe_opcode_t insn = UPROBE_SWBP_INSN;
>
>  	down_write(&mm->mmap_sem);
>  	if (mm->uprobes_state.xol_area)
> @@ -1106,6 +1115,13 @@ static int xol_add_vma(struct xol_area *area)
>  	smp_wmb();	/* pairs with get_xol_area() */
>  	mm->uprobes_state.xol_area = area;
>  	ret = 0;
> +
> +	/*
> +	 * If we reached this place, we did allocate a new area. We want
> +	 * pre-alloc a slot for the return probes here.
> +	 */
> +	xol_get_insn_slot(&insn);

Just change get_xol_area() to do set_bit(0, bitmap) and copy_to_page(page, int3)
(extacted from xol_get_insn_slot().

> @@ -1485,8 +1538,11 @@ static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp)
>  		*is_swbp = -EFAULT;
>  	}
>  
> -	if (!uprobe && test_and_clear_bit(MMF_RECALC_UPROBES, &mm->flags))
> +	utask = get_utask();
> +	if (!uprobe && hlist_empty(&utask->return_uprobes) &&
> +	    test_and_clear_bit(MMF_RECALC_UPROBES, &mm->flags)) {
>  		mmf_recalc_uprobes(mm);

Wait, I was wrong. We should not clear MMF_* if another thread has
->return_uprobes. Perhaps we should change uprobe_pre_sstep_notifier()
instead...

>  	down_read(&uprobe->register_rwsem);
>  	for (uc = uprobe->consumers; uc; uc = uc->next) {
> -		int rc = uc->handler(uc, regs);
> +		if (uc->handler)
> +			rc = uc->handler(uc, regs);
> +
> +		if (uc->rp_handler)
> +			prepare_uretprobe(uprobe, regs); /* put bp at return */

Again, this is not right. We should do this only once after the main
loop.

Oleg.