Re: [PATCH] Revert "drm/dp_mst: Remove single tx msg restriction."

[Harry Wentland]

On 2020-04-23 12:42 p.m., Lyude Paul wrote:
> This reverts commit 6bb0942e8f46863a745489cce27efe5be2a3885e.
> 
> Unfortunately it would appear that the rumors we've heard of sideband
> message interleaving not being very well supported are true. On the
> Lenovo ThinkPad Thunderbolt 3 dock that I have, interleaved messages
> appear to just get dropped:
> 
>   [drm:drm_dp_mst_wait_tx_reply [drm_kms_helper]] timedout msg send
>   00000000571ddfd0 2 1
>   [dp_mst] txmsg cur_offset=2 cur_len=2 seqno=1 state=SENT path_msg=1 dst=00
>   [dp_mst] 	type=ENUM_PATH_RESOURCES contents:
>   [dp_mst] 		port=2
> 
> DP descriptor for this hub:
>   OUI 90-cc-24 dev-ID SYNA3  HW-rev 1.0 SW-rev 3.12 quirks 0x0008
> 
> It would seem like as well that this is a somewhat well known issue in
> the field. From section 5.4.2 of the DisplayPort 2.0 specification:
> 
>   There are MST Sink/Branch devices in the field that do not handle
>   interleaved message transactions.
> 
>   To facilitate message transaction handling by downstream devices, an
>   MST Source device shall generate message transactions in an atomic
>   manner (i.e., the MST Source device shall not concurrently interleave
>   multiple message transactions). Therefore, an MST Source device shall
>   clear the Message_Sequence_No value in the Sideband_MSG_Header to 0.
> 
>   MST Source devices that support field policy updates by way of
>   software should update the policy to forego the generation of
>   interleaved message transactions.
> 
> This is a bit disappointing, as features like HDCP require that we send
> a sideband request every ~2 seconds for each active stream. However,
> there isn't really anything in the specification that allows us to
> accurately probe for interleaved messages.
> 
> If it ends up being that we -really- need this in the future, we might
> be able to whitelist hubs where interleaving is known to work-or maybe
> try some sort of heuristics. But for now, let's just play it safe and
> not use it.
> 

Sounds like the DP spec would need an addition bit to indicate actual
support of interleaved messages by the RX.

Acked-by: Harry Wentland <harry.wentland@xxxxxxx>

Harry

> Signed-off-by: Lyude Paul <lyude@xxxxxxxxxx>
> Fixes: 6bb0942e8f46 ("drm/dp_mst: Remove single tx msg restriction.")
> Cc: Wayne Lin <Wayne.Lin@xxxxxxx>
> Cc: Sean Paul <seanpaul@xxxxxxxxxxxx>
> ---
>  drivers/gpu/drm/drm_dp_mst_topology.c | 14 ++++++++++++--
>  include/drm/drm_dp_mst_helper.h       |  5 +++++
>  2 files changed, 17 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> index 21f10ceb3d6c..03a1496f6120 100644
> --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> @@ -1205,6 +1205,8 @@ static int drm_dp_mst_wait_tx_reply(struct drm_dp_mst_branch *mstb,
>  		    txmsg->state == DRM_DP_SIDEBAND_TX_SENT) {
>  			mstb->tx_slots[txmsg->seqno] = NULL;
>  		}
> +		mgr->is_waiting_for_dwn_reply = false;
> +
>  	}
>  out:
>  	if (unlikely(ret == -EIO) && drm_debug_enabled(DRM_UT_DP)) {
> @@ -1214,6 +1216,7 @@ static int drm_dp_mst_wait_tx_reply(struct drm_dp_mst_branch *mstb,
>  	}
>  	mutex_unlock(&mgr->qlock);
>  
> +	drm_dp_mst_kick_tx(mgr);
>  	return ret;
>  }
>  
> @@ -2789,9 +2792,11 @@ static void process_single_down_tx_qlock(struct drm_dp_mst_topology_mgr *mgr)
>  	ret = process_single_tx_qlock(mgr, txmsg, false);
>  	if (ret == 1) {
>  		/* txmsg is sent it should be in the slots now */
> +		mgr->is_waiting_for_dwn_reply = true;
>  		list_del(&txmsg->next);
>  	} else if (ret) {
>  		DRM_DEBUG_KMS("failed to send msg in q %d\n", ret);
> +		mgr->is_waiting_for_dwn_reply = false;
>  		list_del(&txmsg->next);
>  		if (txmsg->seqno != -1)
>  			txmsg->dst->tx_slots[txmsg->seqno] = NULL;
> @@ -2831,7 +2836,8 @@ static void drm_dp_queue_down_tx(struct drm_dp_mst_topology_mgr *mgr,
>  		drm_dp_mst_dump_sideband_msg_tx(&p, txmsg);
>  	}
>  
> -	if (list_is_singular(&mgr->tx_msg_downq))
> +	if (list_is_singular(&mgr->tx_msg_downq) &&
> +	    !mgr->is_waiting_for_dwn_reply)
>  		process_single_down_tx_qlock(mgr);
>  	mutex_unlock(&mgr->qlock);
>  }
> @@ -3823,6 +3829,7 @@ static int drm_dp_mst_handle_down_rep(struct drm_dp_mst_topology_mgr *mgr)
>  	mutex_lock(&mgr->qlock);
>  	txmsg->state = DRM_DP_SIDEBAND_TX_RX;
>  	mstb->tx_slots[seqno] = NULL;
> +	mgr->is_waiting_for_dwn_reply = false;
>  	mutex_unlock(&mgr->qlock);
>  
>  	wake_up_all(&mgr->tx_waitq);
> @@ -3830,6 +3837,9 @@ static int drm_dp_mst_handle_down_rep(struct drm_dp_mst_topology_mgr *mgr)
>  	return 0;
>  
>  out_clear_reply:
> +	mutex_lock(&mgr->qlock);
> +	mgr->is_waiting_for_dwn_reply = false;
> +	mutex_unlock(&mgr->qlock);
>  	if (msg)
>  		memset(msg, 0, sizeof(struct drm_dp_sideband_msg_rx));
>  out:
> @@ -4683,7 +4693,7 @@ static void drm_dp_tx_work(struct work_struct *work)
>  	struct drm_dp_mst_topology_mgr *mgr = container_of(work, struct drm_dp_mst_topology_mgr, tx_work);
>  
>  	mutex_lock(&mgr->qlock);
> -	if (!list_empty(&mgr->tx_msg_downq))
> +	if (!list_empty(&mgr->tx_msg_downq) && !mgr->is_waiting_for_dwn_reply)
>  		process_single_down_tx_qlock(mgr);
>  	mutex_unlock(&mgr->qlock);
>  }
> diff --git a/include/drm/drm_dp_mst_helper.h b/include/drm/drm_dp_mst_helper.h
> index 2d7c26592c05..96bcf33c03d3 100644
> --- a/include/drm/drm_dp_mst_helper.h
> +++ b/include/drm/drm_dp_mst_helper.h
> @@ -592,6 +592,11 @@ struct drm_dp_mst_topology_mgr {
>  	 */
>  	bool payload_id_table_cleared : 1;
>  
> +	/**
> +	 * @is_waiting_for_dwn_reply: whether we're waiting for a down reply.
> +	 */
> +	bool is_waiting_for_dwn_reply : 1;
> +
>  	/**
>  	 * @mst_primary: Pointer to the primary/first branch device.
>  	 */
>