[PATCH 1/2] drm: Update MST First Link Slot Information Based on Encoding Format
Lyude Paul
lyude at redhat.com
Tue Aug 31 20:19:08 UTC 2021
On Tue, 2021-08-31 at 19:44 +0000, Zuo, Jerry wrote:
> [AMD Official Use Only]
>
> > -----Original Message-----
> > From: Lyude Paul <lyude at redhat.com>
> > Sent: August 30, 2021 4:00 PM
> > To: Zuo, Jerry <Jerry.Zuo at amd.com>; dri-devel at lists.freedesktop.org
> > Cc: Wentland, Harry <Harry.Wentland at amd.com>; Kazlauskas, Nicholas
> > <Nicholas.Kazlauskas at amd.com>; Lin, Wayne <Wayne.Lin at amd.com>
> > Subject: Re: [PATCH 1/2] drm: Update MST First Link Slot Information Based
> > on Encoding Format
> >
> > On Fri, 2021-08-27 at 19:43 -0400, Fangzhi Zuo wrote:
> > > 8b/10b encoding format requires to reserve the first slot for
> > > recording metadata. Real data transmission starts from the second
> > > slot, with a total of available 63 slots available.
> > >
> > > In 128b/132b encoding format, metadata is transmitted separately in
> > > LLCP packet before MTP. Real data transmission starts from the first
> > > slot, with a total of 64 slots available.
> > >
> > > Signed-off-by: Fangzhi Zuo <Jerry.Zuo at amd.com>
> > > ---
> > > drivers/gpu/drm/drm_dp_mst_topology.c | 27
> > > ++++++++++++++++++++-------
> > > include/drm/drm_dp_mst_helper.h | 9 +++++++++
> > > 2 files changed, 29 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c
> > > b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > index 86d13d6bc463..30544801d2b8 100644
> > > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > @@ -3370,7 +3370,7 @@ int drm_dp_update_payload_part1(struct
> > > drm_dp_mst_topology_mgr *mgr)
> > > struct drm_dp_payload req_payload;
> > > struct drm_dp_mst_port *port;
> > > int i, j;
> > > - int cur_slots = 1;
> > > + int cur_slots = mgr->start_slot;
> > > bool skip;
> > >
> > > mutex_lock(&mgr->payload_lock); @@ -4323,7 +4323,7 @@ int
> > > drm_dp_find_vcpi_slots(struct drm_dp_mst_topology_mgr *mgr,
> > > num_slots = DIV_ROUND_UP(pbn, mgr->pbn_div);
> > >
> > > /* max. time slots - one slot for MTP header */
> > > - if (num_slots > 63)
> > > + if (num_slots > mgr->total_avail_slots)
> > > return -ENOSPC;
> > > return num_slots;
> > > }
> > > @@ -4335,7 +4335,7 @@ static int drm_dp_init_vcpi(struct
> > > drm_dp_mst_topology_mgr *mgr,
> > > int ret;
> > >
> > > /* max. time slots - one slot for MTP header */
> > > - if (slots > 63)
> > > + if (slots > mgr->total_avail_slots)
> > > return -ENOSPC;
> > >
> > > vcpi->pbn = pbn;
> > > @@ -4509,6 +4509,17 @@ int drm_dp_atomic_release_vcpi_slots(struct
> > > drm_atomic_state *state,
> > > }
> > > EXPORT_SYMBOL(drm_dp_atomic_release_vcpi_slots);
> > >
> > > +void drm_dp_mst_update_encoding_cap(struct
> > drm_dp_mst_topology_mgr
> > > +*mgr,
> > > uint8_t link_encoding_cap)
> > > +{
> > > + if (link_encoding_cap == DP_CAP_ANSI_128B132B) {
> > > + mgr->total_avail_slots = 64;
> > > + mgr->start_slot = 0;
> > > + }
> > > + DRM_DEBUG_KMS("%s encoding format determined\n",
> > > + (link_encoding_cap == DP_CAP_ANSI_128B132B) ?
> > > "128b/132b" : "8b/10b");
> > > +}
> > > +EXPORT_SYMBOL(drm_dp_mst_update_encoding_cap);
> > > +
> >
> > This seems to be missing kdocs, can you fix that?
> >
> > Also - I'm not convinced this is all of the work we have to do, as there's
> > no
> > locking taking place here in this function. If we're changing the number
> > of
> > available VCPI slots that we have, we need to be able to factor that into
> > the
> > atomic check logic which means that we can't rely on mgr->* for any kind
> > of
> > data that isn't guaranteed to remain consistent throughout the lifetime of
> > the
> > driver or topology. (Note that some of the old MST code didn't follow this
> > logic, so I wouldn't be surprised if there's still exceptions to this we
> > need to
> > clean up).
> >
> > Note that I still expect we'll have to keep some sort of track of the
> > current
> > total slot count in the topology mgr, but that should be reflecting the
> > currently programmed state and not be relied on from our atomic check.
> >
>
> Thanks Lyude for your comments.
>
> Seems I should keep existing code to keep track of current slot status in
> mgr.
> That information is getting updated each time when topology change detected.
> That slot information saved in mgr is a sort of static, and could only be
> used
> for debug purpose to track what is the current encoding format.
Ahh - in that case maybe we should try getting rid of it (as we could just use
the atomic state for debugging output anyway).
>
> > IMHO - the correct way we should go about adding support for this is to
> > add
> > something into drm_dp_mst_topology_state and integrate this into the
> > atomic check helpers.
>
> The slot information should also be added into drm_dp_mst_topology_state to
> reflect the real-time slot status.
>
> I'd like to confirm the best place to get slot count info. updated.
> Should the update be done within &drm_mode_config_funcs. atomic_check(),
> before new stream is created, OR
> should be updated within drm_dp_mst_atomic_check() ?
drm_dp_mst_atomic_check() would be better, since I imagine that we have to do
this for all drivers using MST
>
> The updated slot count will be used in drm_dp_mst_atomic_check() to check
> slot limit,
> and in drm_dp_update_payload_part1() as initial cur_slots.
>
> >
> > > /**
> > > * drm_dp_mst_allocate_vcpi() - Allocate a virtual channel
> > > * @mgr: manager for this port
> > > @@ -4540,8 +4551,8 @@ bool drm_dp_mst_allocate_vcpi(struct
> > > drm_dp_mst_topology_mgr *mgr,
> > >
> > > ret = drm_dp_init_vcpi(mgr, &port->vcpi, pbn, slots);
> > > if (ret) {
> > > - drm_dbg_kms(mgr->dev, "failed to init vcpi slots=%d
> > > max=63 ret=%d\n",
> > > - DIV_ROUND_UP(pbn, mgr->pbn_div), ret);
> > > + drm_dbg_kms(mgr->dev, "failed to init vcpi slots=%d
> > > +max=%d
> > > ret=%d\n",
> > > + DIV_ROUND_UP(pbn, mgr->pbn_div), mgr-
> > > > total_avail_slots, ret);
> > > drm_dp_mst_topology_put_port(port);
> > > goto out;
> > > }
> > > @@ -5228,7 +5239,7 @@
> > drm_dp_mst_atomic_check_vcpi_alloc_limit(struct
> > > drm_dp_mst_topology_mgr *mgr,
> > > struct
> > > drm_dp_mst_topology_state
> > > *mst_state)
> > > {
> > > struct drm_dp_vcpi_allocation *vcpi;
> > > - int avail_slots = 63, payload_count = 0;
> > > + int avail_slots = mgr->total_avail_slots, payload_count = 0;
> > >
> > > list_for_each_entry(vcpi, &mst_state->vcpis, next) {
> > > /* Releasing VCPI is always OK-even if the port is
> > > gone */ @@ -5257,7 +5268,7 @@
> > > drm_dp_mst_atomic_check_vcpi_alloc_limit(struct
> > > drm_dp_mst_topology_mgr *mgr,
> > > }
> > > }
> > > drm_dbg_atomic(mgr->dev, "[MST MGR:%p] mst state %p VCPI
> > > avail=%d used=%d\n",
> > > - mgr, mst_state, avail_slots, 63 - avail_slots);
> > > + mgr, mst_state, avail_slots,
> > > +mgr->total_avail_slots -
> > > avail_slots);
> > >
> > > return 0;
> > > }
> > > @@ -5529,6 +5540,8 @@ int drm_dp_mst_topology_mgr_init(struct
> > > drm_dp_mst_topology_mgr *mgr,
> > > if (!mgr->proposed_vcpis)
> > > return -ENOMEM;
> > > set_bit(0, &mgr->payload_mask);
> > > + mgr->total_avail_slots = 63;
> > > + mgr->start_slot = 1;
> > >
> > > mst_state = kzalloc(sizeof(*mst_state), GFP_KERNEL);
> > > if (mst_state == NULL)
> > > diff --git a/include/drm/drm_dp_mst_helper.h
> > > b/include/drm/drm_dp_mst_helper.h index ddb9231d0309..eac5ce48f214
> > > 100644
> > > --- a/include/drm/drm_dp_mst_helper.h
> > > +++ b/include/drm/drm_dp_mst_helper.h
> > > @@ -661,6 +661,15 @@ struct drm_dp_mst_topology_mgr {
> > > */
> > > int pbn_div;
> > >
> > > + /**
> > > + * @total_avail_slots: available slots for data transmission
> > > + */
> > > + u8 total_avail_slots;
> > > + /**
> > > + * @start_slot: first slot index for data transmission
> > > + */
> > > + u8 start_slot;
> > > +
> > > /**
> > > * @funcs: Atomic helper callbacks
> > > */
> >
> > --
> > Cheers,
> > Lyude Paul (she/her)
> > Software Engineer at Red Hat
>
--
Cheers,
Lyude Paul (she/her)
Software Engineer at Red Hat
More information about the dri-devel
mailing list