On Mon, Nov 10, 2025 at 02:35:41AM -0800, Dipayaan Roy wrote:
> Implement .ndo_tx_timeout for MANA so any stalled TX queue can be detected
> and a device-controlled port reset for all queues can be scheduled to a
> ordered workqueue. The reset for all queues on stall detection is
> recomended by hardware team.
>
> The change introduces a single ordered workqueue
> ("mana_per_port_queue_reset_wq") with WQ_UNBOUND | WQ_MEM_RECLAIM and
> queues exactly one work_struct per port onto it.
I see that this goes some way to addressing Jakub's feedback
on the commit message in his review of v2. But I this paragraph
isn't adding much in it's current form. It seems to me some
explanation of why why WQ_UNBOUND and WQ_MEM_RECLAIM are used is
appropriate.
[1] https://lore.kernel.org/all/[email protected]/
> Reviewed-by: Pavan Chebbi <[email protected]>
> Reviewed-by: Haiyang Zhang <[email protected]>
> Signed-off-by: Dipayaan Roy <[email protected]>
> ---
> Changes in v3:
> -Fixed commit meesage, removed rtnl_trylock and added
> disable_work_sync, fixed mana_queue_reset_work, and few
> cosmetics.
> Changes in v2:
> -Fixed cosmetic changes.
> ---
> ---
> drivers/net/ethernet/microsoft/mana/mana_en.c | 78 ++++++++++++++++++-
> include/net/mana/gdma.h | 7 +-
> include/net/mana/mana.h | 7 ++
> 3 files changed, 90 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c
> b/drivers/net/ethernet/microsoft/mana/mana_en.c
> index cccd5b63cee6..636df3b066c5 100644
> --- a/drivers/net/ethernet/microsoft/mana/mana_en.c
> +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
> @@ -298,6 +298,42 @@ static int mana_get_gso_hs(struct sk_buff *skb)
> return gso_hs;
> }
>
> +static void mana_per_port_queue_reset_work_handler(struct work_struct *work)
> +{
> + struct mana_queue_reset_work *reset_queue_work =
> + container_of(work, struct mana_queue_reset_work, work);
> +
> + struct mana_port_context *apc = container_of(reset_queue_work,
> + struct mana_port_context,
> + queue_reset_work);
> + struct net_device *ndev = apc->ndev;
> + int err;
> +
> + rtnl_lock();
> +
> + /* Pre-allocate buffers to prevent failure in mana_attach later */
> + err = mana_pre_alloc_rxbufs(apc, ndev->mtu, apc->num_queues);
> + if (err) {
> + netdev_err(ndev, "Insufficient memory for reset post tx stall
> detection\n");
> + goto out;
> + }
> +
> + err = mana_detach(ndev, false);
> + if (err) {
> + netdev_err(ndev, "mana_detach failed: %d\n", err);
> + goto dealloc_pre_rxbufs;
> + }
> +
> + err = mana_attach(ndev);
> + if (err)
> + netdev_err(ndev, "mana_attach failed: %d\n", err);
> +
> +dealloc_pre_rxbufs:
> + mana_pre_dealloc_rxbufs(apc);
> +out:
> + rtnl_unlock();
> +}
> +
> netdev_tx_t mana_start_xmit(struct sk_buff *skb, struct net_device *ndev)
> {
> enum mana_tx_pkt_format pkt_fmt = MANA_SHORT_PKT_FMT;
> @@ -802,6 +838,23 @@ static int mana_change_mtu(struct net_device *ndev, int
> new_mtu)
> return err;
> }
>
> +static void mana_tx_timeout(struct net_device *netdev, unsigned int txqueue)
> +{
> + struct mana_port_context *apc = netdev_priv(netdev);
> + struct mana_context *ac = apc->ac;
> + struct gdma_context *gc = ac->gdma_dev->gdma_context;
> +
> + /* Already in service, hence tx queue reset is not required.*/
> + if (gc->in_service)
> + return;
> +
> + /* Note: If there are pending queue reset work for this port(apc),
> + * subsequent request queued up from here are ignored. This is because
> + * we are using the same work instance per port(apc).
> + */
> + queue_work(ac->per_port_queue_reset_wq, &apc->queue_reset_work.work);
> +}
> +
> static int mana_shaper_set(struct net_shaper_binding *binding,
> const struct net_shaper *shaper,
> struct netlink_ext_ack *extack)
> @@ -884,7 +937,9 @@ static const struct net_device_ops mana_devops = {
> .ndo_bpf = mana_bpf,
> .ndo_xdp_xmit = mana_xdp_xmit,
> .ndo_change_mtu = mana_change_mtu,
> - .net_shaper_ops = &mana_shaper_ops,
> + .ndo_tx_timeout = mana_tx_timeout,
> + .net_shaper_ops = &mana_shaper_ops,
> +
> };
>
> static void mana_cleanup_port_context(struct mana_port_context *apc)
> @@ -3244,6 +3299,7 @@ static int mana_probe_port(struct mana_context *ac, int
> port_idx,
> ndev->min_mtu = ETH_MIN_MTU;
> ndev->needed_headroom = MANA_HEADROOM;
> ndev->dev_port = port_idx;
> + ndev->watchdog_timeo = 15 * HZ;
> SET_NETDEV_DEV(ndev, gc->dev);
>
> netif_set_tso_max_size(ndev, GSO_MAX_SIZE);
> @@ -3283,6 +3339,10 @@ static int mana_probe_port(struct mana_context *ac,
> int port_idx,
>
> debugfs_create_u32("current_speed", 0400, apc->mana_port_debugfs,
> &apc->speed);
>
> + /* Initialize the per port queue reset work.*/
> + INIT_WORK(&apc->queue_reset_work.work,
> + mana_per_port_queue_reset_work_handler);
> +
I think it would make more sense to move this to before the call to
register_netdev(), which is a few lines above this hunk.
I suppose that because a watchdog timeout is involved, it won't happen in
practice, but in theory could fire ndo_tx_timeout before INIT_WORK is
called, resulting in access to the work queue before it is initialised.
> return 0;
>
> free_indir:
> @@ -3488,6 +3548,15 @@ int mana_probe(struct gdma_dev *gd, bool resuming)
> if (ac->num_ports > MAX_PORTS_IN_MANA_DEV)
> ac->num_ports = MAX_PORTS_IN_MANA_DEV;
>
> + ac->per_port_queue_reset_wq =
> + alloc_ordered_workqueue("mana_per_port_queue_reset_wq",
> + WQ_UNBOUND | WQ_MEM_RECLAIM);
> + if (!ac->per_port_queue_reset_wq) {
> + dev_err(dev, "Failed to allocate per port queue reset
> workqueue\n");
> + err = -ENOMEM;
> + goto out;
> + }
> +
> if (!resuming) {
> for (i = 0; i < ac->num_ports; i++) {
> err = mana_probe_port(ac, i, &ac->ports[i]);
It is not strictly related to this patch, but the lines above the hunk
below look like this:
apc = netdev_priv(ndev);
if (!ndev) {
if (i == 0)
dev_err(dev, "No net device to remove\n");
If ndev is null then the call to netdev_priv() will result in a
NULL pointer dereference. So I think it should be moved
to after the check for !ndev.
> @@ -3557,6 +3626,8 @@ void mana_remove(struct gdma_dev *gd, bool suspending)
> goto out;
> }
>
> + disable_work_sync(&apc->queue_reset_work.work);
> +
> /* All cleanup actions should stay after rtnl_lock(), otherwise
> * other functions may access partially cleaned up data.
> */
Comments on code flagged by Claude Code with
https://github.com/masoncl/review-prompts/
