Re: [PATCH v4 49/66] bpf: Remove VMA linked list

[Liam Howlett]

* Vlastimil Babka <vbabka@xxxxxxx> [220119 12:05]:
> On 12/1/21 15:30, Liam Howlett wrote:
> > From: "Liam R. Howlett" <Liam.Howlett@xxxxxxxxxx>
> > 
> > Use vma_next() and remove reference to the start of the linked list
> > 
> > Signed-off-by: Liam R. Howlett <Liam.Howlett@xxxxxxxxxx>
> > ---
> >  kernel/bpf/task_iter.c | 21 ++++-----------------
> >  1 file changed, 4 insertions(+), 17 deletions(-)
> > 
> > diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
> > index b48750bfba5a..2d964743f1e6 100644
> > --- a/kernel/bpf/task_iter.c
> > +++ b/kernel/bpf/task_iter.c
> > @@ -299,8 +299,8 @@ struct bpf_iter_seq_task_vma_info {
> >  };
> >  
> >  enum bpf_task_vma_iter_find_op {
> > -	task_vma_iter_first_vma,   /* use mm->mmap */
> > -	task_vma_iter_next_vma,    /* use curr_vma->vm_next */
> > +	task_vma_iter_first_vma,   /* use find_vma() with addr 0 */
> > +	task_vma_iter_next_vma,    /* use vma_next() with curr_vma */
> >  	task_vma_iter_find_vma,    /* use find_vma() to find next vma */
> >  };
> >  
> > @@ -400,24 +400,11 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
> >  
> >  	switch (op) {
> >  	case task_vma_iter_first_vma:
> > -		curr_vma = curr_task->mm->mmap;
> > +		curr_vma = find_vma(curr_task->mm, 0);
> >  		break;
> >  	case task_vma_iter_next_vma:
> > -		curr_vma = curr_vma->vm_next;
> > -		break;
> >  	case task_vma_iter_find_vma:
> > -		/* We dropped mmap_lock so it is necessary to use find_vma
> > -		 * to find the next vma. This is similar to the  mechanism
> > -		 * in show_smaps_rollup().
> > -		 */
> > -		curr_vma = find_vma(curr_task->mm, info->prev_vm_end - 1);
> > -		/* case 1) and 4.2) above just use curr_vma */
> > -
> > -		/* check for case 2) or case 4.1) above */
> > -		if (curr_vma &&
> > -		    curr_vma->vm_start == info->prev_vm_start &&
> > -		    curr_vma->vm_end == info->prev_vm_end)
> > -			curr_vma = curr_vma->vm_next;
> > +		curr_vma = find_vma(curr_task->mm, curr_vma->vm_end);
> 
> Are you sure curr_vma is valid here and we can read its vm_end? Because I
> have no idea, but lots of doubts.

I am not sure.  I'm going to drop this patch and take the more
conservative but safer v3 version.

> 
> >  		break;
> >  	}
> >  	if (!curr_vma) {
> 
>