Re: [PATCH v3 2/4] virtio_test: use random length scatterlists to test descriptor chain

[Guo Zhi]

On 2022/7/11 23:35, Eugenio Perez Martin wrote:
> On Sat, Jul 9, 2022 at 4:28 AM Guo Zhi <qtxuning1999@xxxxxxxxxxx> wrote:
> > Prior implementation only use one descriptor for each io event, which
> > does't test code of descriptor chain. More importantly, one descriptor
> > will not use indirect feature even indirect feature is specified. Use
> > random length scatterlists here to test descriptor chain.
> >
> > Signed-off-by: Guo Zhi <qtxuning1999@xxxxxxxxxxx>
> > ---
> > v3:
> > - drop fda270fcd virtio_test: move magic number in code as defined constant
> > ---
> >   tools/virtio/virtio_test.c | 11 ++++++++---
> >   1 file changed, 8 insertions(+), 3 deletions(-)
> >
> > diff --git a/tools/virtio/virtio_test.c b/tools/virtio/virtio_test.c
> > index 95f78b311..1408a4a20 100644
> > --- a/tools/virtio/virtio_test.c
> > +++ b/tools/virtio/virtio_test.c
> > @@ -20,6 +20,7 @@
> >   #include "../../drivers/vhost/test.h"
> >
> >   #define RANDOM_BATCH -1
> > +#define MAX_SG_FRAGS 8UL
> >
> >   /* Unused */
> >   void *__kmalloc_fake, *__kfree_ignore_start, *__kfree_ignore_end;
> > @@ -169,7 +170,8 @@ static void wait_for_interrupt(struct vdev_info *dev)
> >   static void run_test(struct vdev_info *dev, struct vq_info *vq,
> >                       bool delayed, int batch, int reset_n, int bufs)
> >   {
> > -       struct scatterlist sl;
> > +       struct scatterlist sg[MAX_SG_FRAGS];
> > +       int sg_size = 0;
> >          long started = 0, completed = 0, next_reset = reset_n;
> >          long completed_before, started_before;
> >          int r, test = 1;
> > @@ -194,8 +196,11 @@ static void run_test(struct vdev_info *dev, struct vq_info *vq,
> >
> >                          while (started < bufs &&
> >                                 (started - completed) < batch) {
> > -                               sg_init_one(&sl, dev->buf, dev->buf_size);
> > -                               r = virtqueue_add_outbuf(vq->vq, &sl, 1,
> > +                               sg_size = random() % (MAX_SG_FRAGS - 1) + 1;
> I'm wondering if it would be simpler to reuse batch randomness here,
> and make sg_size = MIN(started - completed, MAX_SG_FRAGS). Vhost test
> should go faster because the longer chains, and I guess we should hit
> a good range of chain lengths with the batch tail anyway.
>
> Thanks!

IMHO, if we reuse batch randomness here, the random length of 
scatterlist only appears when --batch=random selected.

Otherwise, when we have to specify the batch size(eg, 256), the 
scatterlist( as well as the descriptor chain len) will not be randomed. 
So I propose decouple the randomness of batch size and descriptor chain 
length.

If we have to achive better performance for vhost_test, just enlarge the 
MAX_SG_FRAGS is ok.

> > +                               sg_init_table(sg, sg_size);
> > +                               for (int i = 0; i < sg_size; ++i)
> > +                                       sg_set_buf(&sg[i], dev->buf + i, 0x1);
> > +                               r = virtqueue_add_outbuf(vq->vq, sg, sg_size,
> >                                                           dev->buf + started,
> >                                                           GFP_ATOMIC);
> >                                  if (unlikely(r != 0)) {
> > --
> > 2.17.1