Re: [PATCH 3/4] net: mhi_net: Implement runtime PM support

[Mayank Rana <mayank.rana@oss.qualcomm.com>]

+ Chris

Do we need to consider updating client driver (net/qrtr/mhi.c) ?

Regards,
Mayank

On 12/1/2025 4:43 AM, Krishna Chaitanya Chundru wrote:
> Add runtime power management support to the mhi_net driver to align with
> the updated MHI framework, which expects runtime PM to be enabled by client
> drivers. This ensures the controller remains active during data transfers
> and can autosuspend when idle.
> 
> The driver now uses pm_runtime_get() and pm_runtime_put() around
> transmit, receive, and buffer refill operations. Runtime PM is initialized
> during probe with autosuspend enabled and a 100ms delay. The device is
> marked with pm_runtime_no_callbacks() to notify PM framework that there
> are no callbacks for this driver.
> 
> Signed-off-by: Krishna Chaitanya Chundru <krishna.chundru@oss.qualcomm.com>
> ---
>   drivers/net/mhi_net.c | 13 +++++++++++++
>   1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/net/mhi_net.c b/drivers/net/mhi_net.c
> index ae169929a9d8e449b5a427993abf68e8d032fae2..c5c697f29e69e9bc874b6cfff4699de12a4af114 100644
> --- a/drivers/net/mhi_net.c
> +++ b/drivers/net/mhi_net.c
> @@ -9,6 +9,7 @@
>   #include <linux/mod_devicetable.h>
>   #include <linux/module.h>
>   #include <linux/netdevice.h>
> +#include <linux/pm_runtime.h>
>   #include <linux/skbuff.h>
>   #include <linux/u64_stats_sync.h>
>   
> @@ -76,6 +77,7 @@ static netdev_tx_t mhi_ndo_xmit(struct sk_buff *skb, struct net_device *ndev)
>   	struct mhi_device *mdev = mhi_netdev->mdev;
>   	int err;
>   
> +	pm_runtime_get(&mdev->dev);
>   	err = mhi_queue_skb(mdev, DMA_TO_DEVICE, skb, skb->len, MHI_EOT);
>   	if (unlikely(err)) {
>   		net_err_ratelimited("%s: Failed to queue TX buf (%d)\n",
> @@ -94,6 +96,7 @@ static netdev_tx_t mhi_ndo_xmit(struct sk_buff *skb, struct net_device *ndev)
>   	u64_stats_inc(&mhi_netdev->stats.tx_dropped);
>   	u64_stats_update_end(&mhi_netdev->stats.tx_syncp);
>   
> +	pm_runtime_put_autosuspend(&mdev->dev);
>   	return NETDEV_TX_OK;
>   }
>   
> @@ -261,6 +264,7 @@ static void mhi_net_ul_callback(struct mhi_device *mhi_dev,
>   	}
>   	u64_stats_update_end(&mhi_netdev->stats.tx_syncp);
>   
> +	pm_runtime_put_autosuspend(&mdev->dev);
>   	if (netif_queue_stopped(ndev) && !mhi_queue_is_full(mdev, DMA_TO_DEVICE))
>   		netif_wake_queue(ndev);
>   }
> @@ -277,6 +281,7 @@ static void mhi_net_rx_refill_work(struct work_struct *work)
>   
>   	size = mhi_netdev->mru ? mhi_netdev->mru : READ_ONCE(ndev->mtu);
>   
> +	pm_runtime_get_sync(&mdev->dev);
>   	while (!mhi_queue_is_full(mdev, DMA_FROM_DEVICE)) {
>   		skb = netdev_alloc_skb(ndev, size);
>   		if (unlikely(!skb))
> @@ -296,6 +301,7 @@ static void mhi_net_rx_refill_work(struct work_struct *work)
>   		cond_resched();
>   	}
>   
> +	pm_runtime_put_autosuspend(&mdev->dev);
>   	/* If we're still starved of rx buffers, reschedule later */
>   	if (mhi_get_free_desc_count(mdev, DMA_FROM_DEVICE) == mhi_netdev->rx_queue_sz)
>   		schedule_delayed_work(&mhi_netdev->rx_refill, HZ / 2);
> @@ -362,12 +368,19 @@ static int mhi_net_probe(struct mhi_device *mhi_dev,
>   
>   	SET_NETDEV_DEV(ndev, &mhi_dev->dev);
>   
> +	pm_runtime_no_callbacks(&mhi_dev->dev);
> +	devm_pm_runtime_set_active_enabled(&mhi_dev->dev);
> +	pm_runtime_set_autosuspend_delay(&mhi_dev->dev, 100);
> +	pm_runtime_use_autosuspend(&mhi_dev->dev);
> +	pm_runtime_get(&mhi_dev->dev);
>   	err = mhi_net_newlink(mhi_dev, ndev);
>   	if (err) {
>   		free_netdev(ndev);
> +		pm_runtime_put_autosuspend(&mhi_dev->dev);
>   		return err;
>   	}
>   
> +	pm_runtime_put_autosuspend(&mhi_dev->dev);
>   	return 0;
>   }
>   
>