On Sat, 5 Dec 2020 02:43:15 +0200 Vladimir Oltean wrote:
> Currently ocelot_set_rx_mode calls ocelot_mact_learn directly, which has
> a very nice ocelot_mact_wait_for_completion at the end. Introduced in
> commit 639c1b2625af ("net: mscc: ocelot: Register poll timeout should be
> wall time not attempts"), this function uses readx_poll_timeout which
> triggers a lot of lockdep warnings and is also dangerous to use from
> atomic context, potentially leading to lockups and panics.
>
> Steen Hegelund added a poll timeout of 100 ms for checking the MAC
> table, a duration which is clearly absurd to poll in atomic context.
> So we need to defer the MAC table access to process context, which we do
> via a dynamically allocated workqueue which contains all there is to
> know about the MAC table operation it has to do.
>
> Signed-off-by: Vladimir Oltean <vladimir.olt...@nxp.com>
> Reviewed-by: Florian Fainelli <f.faine...@gmail.com>
> ---
> Changes in v3:
> - Dropped Fixes tag and retargeted to net-next
> - Dropped get_device/put_device since they don't offer real protection
> - Now allocating a private ordered workqueue which is drained on unbind
> to avoid accessing freed memory
>
> Changes in v2:
> - Added Fixes tag (it won't backport that far, but anyway)
> - Using get_device and put_device to avoid racing with unbind
>
> drivers/net/ethernet/mscc/ocelot.c | 5 ++
> drivers/net/ethernet/mscc/ocelot_net.c | 81 +++++++++++++++++++++++++-
> include/soc/mscc/ocelot.h | 2 +
> 3 files changed, 85 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/ethernet/mscc/ocelot.c
> b/drivers/net/ethernet/mscc/ocelot.c
> index abea8dd2b0cb..b9626eec8db6 100644
> --- a/drivers/net/ethernet/mscc/ocelot.c
> +++ b/drivers/net/ethernet/mscc/ocelot.c
> @@ -1513,6 +1513,10 @@ int ocelot_init(struct ocelot *ocelot)
> if (!ocelot->stats_queue)
> return -ENOMEM;
>
> + ocelot->owq = alloc_ordered_workqueue("ocelot-owq", WQ_MEM_RECLAIM);
Why MEM_RECLAIM ?
> + if (!ocelot->owq)
> + return -ENOMEM;
I don't think you can pass NULL to destroy_workqueue() so IDK how this
code does error handling (freeing of ocelot->stats_queue if owq fails).
> INIT_LIST_HEAD(&ocelot->multicast);
> INIT_LIST_HEAD(&ocelot->pgids);
> ocelot_mact_init(ocelot);
> @@ -1619,6 +1623,7 @@ void ocelot_deinit(struct ocelot *ocelot)
> {
> cancel_delayed_work(&ocelot->stats_work);
> destroy_workqueue(ocelot->stats_queue);
> + destroy_workqueue(ocelot->owq);
> mutex_destroy(&ocelot->stats_lock);
> }
> +static int ocelot_enqueue_mact_action(struct ocelot *ocelot,
> + const struct ocelot_mact_work_ctx *ctx)
> +{
> + struct ocelot_mact_work_ctx *w = kmalloc(sizeof(*w), GFP_ATOMIC);
> +
> + if (!w)
> + return -ENOMEM;
> +
> + memcpy(w, ctx, sizeof(*w));
kmemdup()?
