Re: [PATCH] Revert "drm/amdkfd: SMI report dropped event count"

[Philip Yang <yangp@xxxxxxx>]

On 2024-10-21 13:46, Alex Deucher wrote:

> This reverts commit a3ab2d45b9887ee609cd3bea39f668236935774c.
>
> The userspace side for this code is not ready yet so revert
> for now.
>
> Signed-off-by: Alex Deucher <alexander.deucher@xxxxxxx>
> Cc: Philip Yang <Philip.Yang@xxxxxxx>
>     
    Reviewed-by: Philip Yang<Philip.Yang@xxxxxxx>
> ---
>  drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 27 +++------------------
>  include/uapi/linux/kfd_ioctl.h              |  6 -----
>  2 files changed, 4 insertions(+), 29 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
> index c8d67d62ca3f..9b8169761ec5 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
> @@ -42,7 +42,6 @@ struct kfd_smi_client {
>  	struct rcu_head rcu;
>  	pid_t pid;
>  	bool suser;
> -	u32 drop_count;
>  };
>  
>  #define KFD_MAX_KFIFO_SIZE	8192
> @@ -104,28 +103,12 @@ static ssize_t kfd_smi_ev_read(struct file *filep, char __user *user,
>  	}
>  	to_copy = min(size, to_copy);
>  	ret = kfifo_out(&client->fifo, buf, to_copy);
> +	spin_unlock(&client->lock);
>  	if (ret <= 0) {
> -		spin_unlock(&client->lock);
>  		ret = -EAGAIN;
>  		goto ret_err;
>  	}
>  
> -	if (client->drop_count) {
> -		char msg[KFD_SMI_EVENT_MSG_SIZE];
> -		int len;
> -
> -		len = snprintf(msg, sizeof(msg), "%x ", KFD_SMI_EVENT_DROPPED_EVENT);
> -		len += snprintf(msg + len, sizeof(msg) - len,
> -				KFD_EVENT_FMT_DROPPED_EVENT(ktime_get_boottime_ns(),
> -				client->pid, client->drop_count));
> -		if (kfifo_avail(&client->fifo) >= len) {
> -			kfifo_in(&client->fifo, msg, len);
> -			client->drop_count = 0;
> -		}
> -	}
> -
> -	spin_unlock(&client->lock);
> -
>  	ret = copy_to_user(user, buf, to_copy);
>  	if (ret) {
>  		ret = -EFAULT;
> @@ -199,15 +182,13 @@ static void add_event_to_kfifo(pid_t pid, struct kfd_node *dev,
>  	list_for_each_entry_rcu(client, &dev->smi_clients, list) {
>  		if (!kfd_smi_ev_enabled(pid, client, smi_event))
>  			continue;
> -
>  		spin_lock(&client->lock);
> -		if (!client->drop_count && kfifo_avail(&client->fifo) >= len) {
> +		if (kfifo_avail(&client->fifo) >= len) {
>  			kfifo_in(&client->fifo, event_msg, len);
>  			wake_up_all(&client->wait_queue);
>  		} else {
> -			client->drop_count++;
> -			pr_debug("smi_event(EventID: %u): no space left drop_count %d\n",
> -				 smi_event, client->drop_count);
> +			pr_debug("smi_event(EventID: %u): no space left\n",
> +					smi_event);
>  		}
>  		spin_unlock(&client->lock);
>  	}
> diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h
> index 7afd66d45313..fa9f9846b88e 100644
> --- a/include/uapi/linux/kfd_ioctl.h
> +++ b/include/uapi/linux/kfd_ioctl.h
> @@ -530,7 +530,6 @@ enum kfd_smi_event {
>  	KFD_SMI_EVENT_QUEUE_EVICTION = 9,
>  	KFD_SMI_EVENT_QUEUE_RESTORE = 10,
>  	KFD_SMI_EVENT_UNMAP_FROM_GPU = 11,
> -	KFD_SMI_EVENT_DROPPED_EVENT = 12,
>  
>  	/*
>  	 * max event number, as a flag bit to get events from all processes,
> @@ -611,7 +610,6 @@ struct kfd_ioctl_smi_events_args {
>   *    rw: 'W' for write page fault, 'R' for read page fault
>   *    rescheduled: 'R' if the queue restore failed and rescheduled to try again
>   *    error_code: migrate failure error code, 0 if no error
> - *    drop_count: how many events dropped when fifo is full
>   */
>  #define KFD_EVENT_FMT_UPDATE_GPU_RESET(reset_seq_num, reset_cause)\
>  		"%x %s\n", (reset_seq_num), (reset_cause)
> @@ -647,10 +645,6 @@ struct kfd_ioctl_smi_events_args {
>  		"%lld -%d @%lx(%lx) %x %d\n", (ns), (pid), (addr), (size),\
>  		(node), (unmap_trigger)
>  
> -#define KFD_EVENT_FMT_DROPPED_EVENT(ns, pid, drop_count)\
> -		"%lld -%d %d\n", (ns), (pid), (drop_count)
> -
> -
>  /**************************************************************************************************
>   * CRIU IOCTLs (Checkpoint Restore In Userspace)
>   *
>