[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c72573d7e0c2165eaa92618a2c1c912d7c8117d6.camel@nvidia.com>
Date: Wed, 19 Jun 2024 17:03:56 +0000
From: Dragos Tatulea <dtatulea@...dia.com>
To: "eperezma@...hat.com" <eperezma@...hat.com>
CC: "linux-rdma@...r.kernel.org" <linux-rdma@...r.kernel.org>,
"xuanzhuo@...ux.alibaba.com" <xuanzhuo@...ux.alibaba.com>,
"virtualization@...ts.linux.dev" <virtualization@...ts.linux.dev>, Tariq
Toukan <tariqt@...dia.com>, "linux-kernel@...r.kernel.org"
<linux-kernel@...r.kernel.org>, Cosmin Ratiu <cratiu@...dia.com>,
"si-wei.liu@...cle.com" <si-wei.liu@...cle.com>, "jasowang@...hat.com"
<jasowang@...hat.com>, "mst@...hat.com" <mst@...hat.com>, Saeed Mahameed
<saeedm@...dia.com>, "leon@...nel.org" <leon@...nel.org>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH vhost 08/23] vdpa/mlx5: Clear and reinitialize software VQ
data on reset
On Wed, 2024-06-19 at 13:28 +0200, Eugenio Perez Martin wrote:
> On Mon, Jun 17, 2024 at 5:08 PM Dragos Tatulea <dtatulea@...dia.com> wrote:
> >
> > The hardware VQ configuration is mirrored by data in struct
> > mlx5_vdpa_virtqueue . Instead of clearing just a few fields at reset,
> > fully clear the struct and initialize with the appropriate default
> > values.
> >
> > As clear_vqs_ready() is used only during reset, get rid of it.
> >
> > Signed-off-by: Dragos Tatulea <dtatulea@...dia.com>
> > Reviewed-by: Cosmin Ratiu <cratiu@...dia.com>
>
> Acked-by: Eugenio Pérez <eperezma@...hat.com>
>
> > ---
> > drivers/vdpa/mlx5/net/mlx5_vnet.c | 16 +++-------------
> > 1 file changed, 3 insertions(+), 13 deletions(-)
> >
> > diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> > index c8b5c87f001d..de013b5a2815 100644
> > --- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
> > +++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> > @@ -2941,18 +2941,6 @@ static void teardown_vq_resources(struct mlx5_vdpa_net *ndev)
> > ndev->setup = false;
> > }
> >
> > -static void clear_vqs_ready(struct mlx5_vdpa_net *ndev)
> > -{
> > - int i;
> > -
> > - for (i = 0; i < ndev->mvdev.max_vqs; i++) {
> > - ndev->vqs[i].ready = false;
> > - ndev->vqs[i].modified_fields = 0;
> > - }
> > -
> > - ndev->mvdev.cvq.ready = false;
> > -}
> > -
> > static int setup_cvq_vring(struct mlx5_vdpa_dev *mvdev)
> > {
> > struct mlx5_control_vq *cvq = &mvdev->cvq;
> > @@ -3035,12 +3023,14 @@ static int mlx5_vdpa_compat_reset(struct vdpa_device *vdev, u32 flags)
> > down_write(&ndev->reslock);
> > unregister_link_notifier(ndev);
> > teardown_vq_resources(ndev);
> > - clear_vqs_ready(ndev);
> > + init_mvqs(ndev);
>
> Nitpick / suggestion if you have to send a v2. The init_mvqs function
> name sounds like it can allocate stuff that needs to be released to
> me. But I'm very bad at naming :). Maybe something like
> "mvqs_set_defaults" or similar?
Makes sense. I think I will call it mvqs_reset / reset_mvqs to keep things
consistent.
Thanks,
Dragos
>
> > +
> > if (flags & VDPA_RESET_F_CLEAN_MAP)
> > mlx5_vdpa_destroy_mr_resources(&ndev->mvdev);
> > ndev->mvdev.status = 0;
> > ndev->mvdev.suspended = false;
> > ndev->cur_num_vqs = MLX5V_DEFAULT_VQ_COUNT;
> > + ndev->mvdev.cvq.ready = false;
> > ndev->mvdev.cvq.received_desc = 0;
> > ndev->mvdev.cvq.completed_desc = 0;
> > memset(ndev->event_cbs, 0, sizeof(*ndev->event_cbs) * (mvdev->max_vqs + 1));
> >
> > --
> > 2.45.1
> >
>
Powered by blists - more mailing lists