[AMD Official Use Only - General]
Hi Lyude,
Much appreciated!
Will update and send out another version.
Thanks,
Wayne Lin
> -----Original Message-----
> From: Lyude Paul <ly...@redhat.com>
> Sent: Wednesday, June 7, 2023 6:03 AM
> To: Lin, Wayne <wayne....@amd.com>; amd-gfx@lists.freedesktop.org
> Cc: ville.syrj...@linux.intel.com; jani.nik...@intel.com; imre.d...@intel.com;
> Wentland, Harry <harry.wentl...@amd.com>; Zuo, Jerry
> <jerry....@amd.com>; sta...@vger.kernel.org
> Subject: Re: [PATCH] drm/dp_mst: Clear MSG_RDY flag before sending new
> message
>
> My apologies for taking so long to reply to this. I'll be honest - the last
> MST
> fiasco seriously burned me out big time, and I had to take a big step back
> from
> working on MST for a while. Comments below
>
> On Wed, 2023-05-31 at 12:00 +0800, Wayne Lin wrote:
> > [Why]
> > The sequence for collecting down_reply from source perspective should
> > be:
> >
> > Request_n->repeat (get partial reply of Request_n->clear message ready
> > flag to ack DPRX that the message is received) till all partial
> > replies for Request_n are received->new Request_n+1.
> >
> > Now there is chance that drm_dp_mst_hpd_irq() will fire new down
> > request in the tx queue when the down reply is incomplete. Source is
> > restricted to generate interveleaved message transactions so we should
> > avoid it.
> >
> > Also, while assembling partial reply packets, reading out DPCD
> > DOWN_REP Sideband MSG buffer + clearing DOWN_REP_MSG_RDY flag
> should
> > be wrapped up as a complete operation for reading out a reply packet.
> > Kicking off a new request before clearing DOWN_REP_MSG_RDY flag might
> > be risky. e.g. If the reply of the new request has overwritten the
> > DPRX DOWN_REP Sideband MSG buffer before source writing one to clear
> > DOWN_REP_MSG_RDY flag, source then unintentionally flushes the reply
> > for the new request. Should handle the up request in the same way.
> >
> > [How]
> > Separete drm_dp_mst_hpd_irq() into 2 steps. After acking the MST IRQ
> > event, driver calls drm_dp_mst_hpd_irq_send_new_request() and might
> > trigger drm_dp_mst_kick_tx() only when there is no on going message
> > transaction.
> >
> > Changes since v1:
> > * Reworked on review comments received
> > -> Adjust the fix to let driver explicitly kick off new down request
> > when mst irq event is handled and acked
> > -> Adjust the commit message
> >
> > Changes since v2:
> > * Adjust the commit message
> > * Adjust the naming of the divided 2 functions and add a new input
> > parameter "ack".
> > * Adjust code flow as per review comments.
> >
> > Signed-off-by: Wayne Lin <wayne....@amd.com>
> > Cc: sta...@vger.kernel.org
> > ---
> > .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 33 +++++++++-----
> --
> > drivers/gpu/drm/display/drm_dp_mst_topology.c | 39
> +++++++++++++++++--
> > drivers/gpu/drm/i915/display/intel_dp.c | 7 ++--
> > drivers/gpu/drm/nouveau/dispnv50/disp.c | 12 ++++--
> > include/drm/display/drm_dp_mst_helper.h | 7 +++-
> > 5 files changed, 70 insertions(+), 28 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> > b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> > index d5cec03eaa8d..597c3368bcfb 100644
> > --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> > +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> > @@ -3236,6 +3236,7 @@ static void dm_handle_mst_sideband_msg(struct
> > amdgpu_dm_connector *aconnector) {
> > u8 esi[DP_PSR_ERROR_STATUS - DP_SINK_COUNT_ESI] = { 0 };
> > u8 dret;
> > + u8 ack;
> > bool new_irq_handled = false;
> > int dpcd_addr;
> > int dpcd_bytes_to_read;
> > @@ -3265,34 +3266,36 @@ static void
> dm_handle_mst_sideband_msg(struct amdgpu_dm_connector *aconnector)
> > process_count < max_process_count) {
> > u8 retry;
> > dret = 0;
> > + ack = 0;
> >
> > process_count++;
> >
> > DRM_DEBUG_DRIVER("ESI %02x %02x %02x\n", esi[0],
> esi[1], esi[2]);
> > /* handle HPD short pulse irq */
> > if (aconnector->mst_mgr.mst_state)
> > - drm_dp_mst_hpd_irq(
> > - &aconnector->mst_mgr,
> > - esi,
> > - &new_irq_handled);
> > + drm_dp_mst_hpd_irq_handle_event(&aconnector-
> >mst_mgr,
> > + esi,
> > + &ack,
> > + &new_irq_handled);
> >
> > if (new_irq_handled) {
> > /* ACK at DPCD to notify down stream */
> > - const int ack_dpcd_bytes_to_write =
> > - dpcd_bytes_to_read - 1;
> > -
> > for (retry = 0; retry < 3; retry++) {
> > - u8 wret;
> > -
> > - wret = drm_dp_dpcd_write(
> > - &aconnector->dm_dp_aux.aux,
> > - dpcd_addr + 1,
> > - &esi[1],
> > - ack_dpcd_bytes_to_write);
> > - if (wret == ack_dpcd_bytes_to_write)
> > + ssize_t wret;
> > +
> > + wret = drm_dp_dpcd_writeb(&aconnector-
> >dm_dp_aux.aux,
> > + dpcd_addr + 1,
> > + ack);
> > + if (wret == 1)
> > break;
> > }
> >
> > + if (retry == 3) {
> > + DRM_ERROR("Failed to ack MST event.\n");
> > + return;
> > + }
> > +
> > +
> drm_dp_mst_hpd_irq_send_new_request(&aconnector->mst_mgr);
> > /* check if there is new irq to be handled */
> > dret = drm_dp_dpcd_read(
> > &aconnector->dm_dp_aux.aux,
> > diff --git a/drivers/gpu/drm/display/drm_dp_mst_topology.c
> > b/drivers/gpu/drm/display/drm_dp_mst_topology.c
> > index 38dab76ae69e..13165e764709 100644
> > --- a/drivers/gpu/drm/display/drm_dp_mst_topology.c
> > +++ b/drivers/gpu/drm/display/drm_dp_mst_topology.c
> > @@ -4053,9 +4053,10 @@ static int drm_dp_mst_handle_up_req(struct
> > drm_dp_mst_topology_mgr *mgr) }
> >
> > /**
> > - * drm_dp_mst_hpd_irq() - MST hotplug IRQ notify
> > + * drm_dp_mst_hpd_irq_handle_event() - MST hotplug IRQ handle MST
> > + event
> > * @mgr: manager to notify irq for.
> > * @esi: 4 bytes from SINK_COUNT_ESI
> > + * @ack: flags of events to ack
> > * @handled: whether the hpd interrupt was consumed or not
> > *
> > * This should be called from the driver when it detects a short IRQ,
> > @@ -4063,7 +4064,8 @@ static int drm_dp_mst_handle_up_req(struct
> drm_dp_mst_topology_mgr *mgr)
> > * topology manager will process the sideband messages received as a result
> > * of this.
>
> We've gotta update the documentation here to reflect the changes being
> made.
> Otherwise I think this looks good. Mind sending out a new version and I will
> confirm this works on nouveau's side before giving an RB
>
> (That might take a day or two, I accidentally discovered nouveau's MST
> support regressed a bit while I was away :( )
>
> > */
> > -int drm_dp_mst_hpd_irq(struct drm_dp_mst_topology_mgr *mgr, u8 *esi,
> > bool *handled)
> > +int drm_dp_mst_hpd_irq_handle_event(struct drm_dp_mst_topology_mgr
> *mgr, const u8 *esi,
> > + u8 *ack, bool *handled)
> > {
> > int ret = 0;
> > int sc;
> > @@ -4078,18 +4080,47 @@ int drm_dp_mst_hpd_irq(struct
> drm_dp_mst_topology_mgr *mgr, u8 *esi, bool *handl
> > if (esi[1] & DP_DOWN_REP_MSG_RDY) {
> > ret = drm_dp_mst_handle_down_rep(mgr);
> > *handled = true;
> > + *ack |= DP_DOWN_REP_MSG_RDY;
> > }
> >
> > if (esi[1] & DP_UP_REQ_MSG_RDY) {
> > ret |= drm_dp_mst_handle_up_req(mgr);
> > *handled = true;
> > + *ack |= DP_UP_REQ_MSG_RDY;
> > }
> >
> > - drm_dp_mst_kick_tx(mgr);
> > return ret;
> > }
> > -EXPORT_SYMBOL(drm_dp_mst_hpd_irq);
> > +EXPORT_SYMBOL(drm_dp_mst_hpd_irq_handle_event);
> > +
> > +/**
> > + * drm_dp_mst_hpd_irq_send_new_request() - MST hotplug IRQ kick off
> > +new request
> > + * @mgr: manager to notify irq for.
> > + *
> > + * This should be called from the driver when mst irq event is
> > +handled
> > + * and acked. Note that new down request should only be sent when
> > + * previous message transaction is completed. Source is not supposed
> > +to generate
> > + * interleaved message transactions.
> > + */
> > +void drm_dp_mst_hpd_irq_send_new_request(struct
> > +drm_dp_mst_topology_mgr *mgr) {
> > + struct drm_dp_sideband_msg_tx *txmsg;
> > + bool kick = true;
> >
> > + mutex_lock(&mgr->qlock);
> > + txmsg = list_first_entry_or_null(&mgr->tx_msg_downq,
> > + struct drm_dp_sideband_msg_tx,
> next);
> > + /* If last transaction is not completed yet*/
> > + if (!txmsg ||
> > + txmsg->state == DRM_DP_SIDEBAND_TX_START_SEND ||
> > + txmsg->state == DRM_DP_SIDEBAND_TX_SENT)
> > + kick = false;
> > + mutex_unlock(&mgr->qlock);
> > +
> > + if (kick)
> > + drm_dp_mst_kick_tx(mgr);
> > +}
> > +EXPORT_SYMBOL(drm_dp_mst_hpd_irq_send_new_request);
> > /**
> > * drm_dp_mst_detect_port() - get connection status for an MST port
> > * @connector: DRM connector for this port diff --git
> > a/drivers/gpu/drm/i915/display/intel_dp.c
> > b/drivers/gpu/drm/i915/display/intel_dp.c
> > index 4bec8cd7979f..f24602887015 100644
> > --- a/drivers/gpu/drm/i915/display/intel_dp.c
> > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
> > @@ -4062,9 +4062,7 @@ intel_dp_mst_hpd_irq(struct intel_dp *intel_dp,
> > u8 *esi, u8 *ack) {
> > bool handled = false;
> >
> > - drm_dp_mst_hpd_irq(&intel_dp->mst_mgr, esi, &handled);
> > - if (handled)
> > - ack[1] |= esi[1] & (DP_DOWN_REP_MSG_RDY |
> DP_UP_REQ_MSG_RDY);
> > + drm_dp_mst_hpd_irq_handle_event(&intel_dp->mst_mgr, esi,
> &ack[1],
> > +&handled);
> >
> > if (esi[1] & DP_CP_IRQ) {
> > intel_hdcp_handle_cp_irq(intel_dp->attached_connector);
> > @@ -4139,6 +4137,9 @@ intel_dp_check_mst_status(struct intel_dp
> > *intel_dp)
> >
> > if (!intel_dp_ack_sink_irq_esi(intel_dp, ack))
> > drm_dbg_kms(&i915->drm, "Failed to ack ESI\n");
> > +
> > + if (ack[1] & (DP_DOWN_REP_MSG_RDY |
> DP_UP_REQ_MSG_RDY))
> > + drm_dp_mst_hpd_irq_send_new_request(&intel_dp-
> >mst_mgr);
> > }
> >
> > return link_ok;
> > diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c
> > b/drivers/gpu/drm/nouveau/dispnv50/disp.c
> > index 9b6824f6b9e4..b2d9978e88a8 100644
> > --- a/drivers/gpu/drm/nouveau/dispnv50/disp.c
> > +++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c
> > @@ -1357,6 +1357,7 @@ nv50_mstm_service(struct nouveau_drm *drm,
> > bool handled = true, ret = true;
> > int rc;
> > u8 esi[8] = {};
> > + u8 ack;
> >
> > while (handled) {
> > rc = drm_dp_dpcd_read(aux, DP_SINK_COUNT_ESI, esi, 8);
> @@ -1365,16
> > +1366,19 @@ nv50_mstm_service(struct nouveau_drm *drm,
> > break;
> > }
> >
> > - drm_dp_mst_hpd_irq(&mstm->mgr, esi, &handled);
> > + ack = 0;
> > + drm_dp_mst_hpd_irq_handle_event(&mstm->mgr, esi, &ack,
> &handled);
> > if (!handled)
> > break;
> >
> > - rc = drm_dp_dpcd_write(aux, DP_SINK_COUNT_ESI + 1,
> &esi[1],
> > - 3);
> > - if (rc != 3) {
> > + rc = drm_dp_dpcd_writeb(aux, DP_SINK_COUNT_ESI + 1,
> ack);
> > +
> > + if (rc != 1) {
> > ret = false;
> > break;
> > }
> > +
> > + drm_dp_mst_hpd_irq_send_new_request(&mstm->mgr);
> > }
> >
> > if (!ret)
> > diff --git a/include/drm/display/drm_dp_mst_helper.h
> > b/include/drm/display/drm_dp_mst_helper.h
> > index 32c764fb9cb5..40e855c8407c 100644
> > --- a/include/drm/display/drm_dp_mst_helper.h
> > +++ b/include/drm/display/drm_dp_mst_helper.h
> > @@ -815,8 +815,11 @@ void drm_dp_mst_topology_mgr_destroy(struct
> > drm_dp_mst_topology_mgr *mgr); bool drm_dp_read_mst_cap(struct
> > drm_dp_aux *aux, const u8 dpcd[DP_RECEIVER_CAP_SIZE]); int
> > drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr
> *mgr,
> > bool mst_state);
> >
> > -int drm_dp_mst_hpd_irq(struct drm_dp_mst_topology_mgr *mgr, u8 *esi,
> > bool *handled);
> > -
> > +int drm_dp_mst_hpd_irq_handle_event(struct drm_dp_mst_topology_mgr
> *mgr,
> > + const u8 *esi,
> > + u8 *ack,
> > + bool *handled);
> > +void drm_dp_mst_hpd_irq_send_new_request(struct
> > +drm_dp_mst_topology_mgr *mgr);
> >
> > int
> > drm_dp_mst_detect_port(struct drm_connector *connector,
>
> --
> Cheers,
> Lyude Paul (she/her)
> Software Engineer at Red Hat
