drm/msm/mdp5: atomic smp + hwpipe assignment
cover drm/msm/mdp5: atomic smp + hwpipe assignment
In the process of getting kms fence fd (and related gpu egl fence fd extension) working with real hardware (using android drm-hwc) I found a lot of issues w/ plane support. Perhaps that is to be expected the first time getting a userspace that actually makes heavy use of overlay planes working. So some of the approaches here may be useful to other drivers when they cross that bridge.
The mdp5 hw, in particular variants that have "SMP" (Shared Memory Pool, ie. shared FIFO blocks that are allocated to the various hw pipes) have a lot of constraints that a generic userspace is not aware of. Such as not being able to arbitrarily change scanout width or pixel format on an active pipe. In addition, not all pipes can support all features (YUV, rotation, and in some cases, scaling).
Additionally, during the initial conversion to atomic, we did not handle SMP block allocation in plane->atomic_check(), leading to cases where an update could fail in plane->atomic_update() due to failure to allocate SMP blocks.
To solve this:
(1) dynamically assign hw pipes to planes at ->atomic_check() based on the required plane capabilities
(2) handle SMP block allocation in ->atomic_check(). When SMP block allocation needs to change, re-assign a new hw pipe
To do this, we need to track the SMP block allocation, and hw pipe assignment in a global atomic state object. This way, we ensure that multiple updates from different threads which touch shared state (the hw pipe assignemnt or SMP block allocation) are properly serialized (via state_lock), and atomic test-only updates, or commits that fail in atomic_check() (or deadlock/backoff scenarios) are properly un- wound by simply throwing away the new state objects.
The global state is handled in the same way as plane/crtc/etc state. On mdp5_get_state(), the state_lock is acquired as part of the atomic acquire_ctx, and the current state is copied. At that point changes to the copied state are permitted. New plane state can reference new hw pipes, since the assignment of hw pipes is tracked in the global state. At any point the new state can be thrown away (along with the various new plane state objects that reference it as part of the new drm_atomic_state). And if the new state is committed, the new global state is swapped with old state, in the same way per-plane/ crtc/connector state is handled, preserving consistency between global and per-plane state.
Rob Clark (5):
drm/msm/mdp5: introduce mdp5_hw_pipe
drm/msm: subclass drm_atomic_state drm/msm/mdp5: add skeletal mdp5_state drm/msm/mdp5: dynamically assign hw pipes to planes drm/msm/mdp5: handle SMP block allocations "atomically"
drivers/gpu/drm/msm/Makefile | 1 + drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c | 2 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 180 +++++++++++++++----- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 33 +++- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 132 +++++++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 54 ++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 202 +++++++++------------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c | 270 ++++++++++-------------------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h | 66 ++++++-- drivers/gpu/drm/msm/msm_atomic.c | 31 ++++ drivers/gpu/drm/msm/msm_drv.c | 3 + drivers/gpu/drm/msm/msm_drv.h | 3 + drivers/gpu/drm/msm/msm_kms.h | 14 ++ 13 files changed, 629 insertions(+), 362 deletions(-) create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h
Signature · 1 line
2.7.4
PATCH 1/5 drm/msm/mdp5: introduce mdp5_hw_pipe
Split out the hardware pipe specifics from mdp5_plane. To start, the hw pipes are statically assigned to planes, but next step is to assign the hw pipes during plane->atomic_check() based on requested caps (scaling, YUV, etc). And then hw pipe re-assignment if required if required SMP blocks changes.
--- drivers/gpu/drm/msm/Makefile | 1 + drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 126 +++++++++++++++++++----------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 43 ++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 39 +++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 66 +++++++--------- 6 files changed, 197 insertions(+), 85 deletions(-) create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h
drivers/gpu/drm/msm/Makefile +1 -0
@@ -37,6 +37,7 @@ msm-y := \ mdp/mdp5/mdp5_irq.o \ mdp/mdp5/mdp5_mdss.o \ mdp/mdp5/mdp5_kms.o \ + mdp/mdp5/mdp5_pipe.o \ mdp/mdp5/mdp5_plane.o \ mdp/mdp5/mdp5_smp.o \ msm_atomic.o \
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +81 -45
172 lines not on this page · raw message
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +5 -2
@@ -24,6 +24,7 @@#include "mdp5_cfg.h" /* must be included before mdp5.xml.h */ #include "mdp5.xml.h" #include "mdp5_ctl.h" +#include "mdp5_pipe.h" #include "mdp5_smp.h" struct mdp5_kms {
@@ -33,6 +34,9 @@ struct mdp5_kms {struct platform_device *pdev; +unsigned num_hwpipes; +struct mdp5_hw_pipe *hwpipes[16 ]; +struct mdp5_cfg_handler *cfg;uint32_t caps;/* MDP capabilities (MDP_CAP_XXX bits) */
@@ -207,8 +211,7 @@ void mdp5_plane_complete_commit(struct drm_plane *plane, struct drm_plane_state *state);enum mdp5_pipe mdp5_plane_pipe (struct drm_plane *plane);struct drm_plane *mdp5_plane_init (struct drm_device *dev, -enum mdp5_pipe pipe,bool private_plane, -uint32_t reg_offset,uint32_t caps); +struct mdp5_hw_pipe *hwpipe,bool primary);uint32_t mdp5_crtc_vblank (struct drm_crtc *crtc);
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c new file +43 -0
43 lines not on this page · raw message
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h new file +39 -0
@@ -0,0 +1,39 @@ +/* +* Copyright (C) 2016 Red Hat +* Author: Rob Clark <[email protected]> +* +* This program is free software; you can redistribute it and/or modify it +* under the terms of the GNU General Public License version 2 as published by +* the Free Software Foundation. +* +* This program is distributed in the hope that it will be useful, but WITHOUT +* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or +* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for +* more details. +* +* You should have received a copy of the GNU General Public License along with +* this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#ifndef __MDP5_PIPE_H__ +#define __MDP5_PIPE_H__ + +/* represents a hw pipe, which is dynamically assigned to a plane */ +struct mdp5_hw_pipe { +int idx; + +const char *name; +enum mdp5_pipe pipe; + +spinlock_t pipe_lock;/* protect REG_MDP5_PIPE_* registers */ +uint32_t reg_offset; +uint32_t caps; + +uint32_t flush_mask;/* used to commit pipe registers */ +}; + +struct mdp5_hw_pipe *mdp5_pipe_init (enum mdp5_pipe pipe, +uint32_t reg_offset,uint32_t caps); +void mdp5_pipe_destroy (struct mdp5_hw_pipe *hwpipe); + +#endif /* __MDP5_PIPE_H__ */
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +28 -38
190 lines not on this page · raw message
Signature · 6 lines
2.7.4 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH 2/5 drm/msm: subclass drm_atomic_state
In reply to Rob ClarkThis will give the kms backends a slot to stash their own hw specific global state.
--- drivers/gpu/drm/msm/msm_atomic.c | 31 +++++++++++++++++++++++++++++++ drivers/gpu/drm/msm/msm_drv.c | 3 +++ drivers/gpu/drm/msm/msm_drv.h | 3 +++ drivers/gpu/drm/msm/msm_kms.h | 14 ++++++++++++++ 4 files changed, 51 insertions(+)
drivers/gpu/drm/msm/msm_atomic.c +31 -0
@@ -241,6 +241,10 @@ int msm_atomic_commit(struct drm_device *dev,drm_atomic_helper_swap_state (state, true); +/* swap driver private state while still holding state_lock */ +if (to_kms_state (state)->state) + priv->kms->funcs->swap_state (priv->kms, state); +/* * Everything below can be run asynchronously without the need to grab * any modeset locks at all under one conditions: It must be guaranteed
@@ -271,3 +275,30 @@ error:drm_atomic_helper_cleanup_planes (dev, state);return ret; } + +struct drm_atomic_state *msm_atomic_state_alloc (struct drm_device *dev) +{ +struct msm_kms_state *state =kzalloc (sizeof (*state),GFP_KERNEL ); + +if (!state ||drm_atomic_state_init (dev, &state->base) <0 ) { +kfree (state); +return NULL ; + } + +return &state->base; +} + +void msm_atomic_state_clear (struct drm_atomic_state *s) +{ +struct msm_kms_state *state =to_kms_state (s); +drm_atomic_state_default_clear (&state->base); +kfree (state->state); + state->state =NULL ; +} + +void msm_atomic_state_free (struct drm_atomic_state *state) +{ +kfree (to_kms_state (state)->state); +drm_atomic_state_default_release (state); +kfree (state); +}
drivers/gpu/drm/msm/msm_drv.c +3 -0
@@ -46,6 +46,9 @@ static const struct drm_mode_config_funcs mode_config_funcs = { .output_poll_changed = msm_fb_output_poll_changed, .atomic_check = msm_atomic_check, .atomic_commit = msm_atomic_commit, + .atomic_state_alloc = msm_atomic_state_alloc, + .atomic_state_clear = msm_atomic_state_clear, + .atomic_state_free = msm_atomic_state_free, }; int msm_register_address_space(struct drm_device *dev,
drivers/gpu/drm/msm/msm_drv.h +3 -0
@@ -179,6 +179,9 @@ int msm_atomic_check(struct drm_device *dev, struct drm_atomic_state *state);int msm_atomic_commit (struct drm_device *dev,struct drm_atomic_state *state,bool nonblock); +struct drm_atomic_state *msm_atomic_state_alloc (struct drm_device *dev); +void msm_atomic_state_clear (struct drm_atomic_state *state); +void msm_atomic_state_free (struct drm_atomic_state *state);int msm_register_address_space (struct drm_device *dev,struct msm_gem_address_space *aspace);
drivers/gpu/drm/msm/msm_kms.h +14 -0
@@ -40,6 +40,8 @@ struct msm_kms_funcs {irqreturn_t (*irq)(struct msm_kms *kms);int (*enable_vblank)(struct msm_kms *kms,struct drm_crtc *crtc);void (*disable_vblank)(struct msm_kms *kms,struct drm_crtc *crtc); +/* swap global atomic state: */ +void (*swap_state)(struct msm_kms *kms,struct drm_atomic_state *state);/* modeset, bracketing atomic_commit(): */ void (*prepare_commit)(struct msm_kms *kms,struct drm_atomic_state *state);void (*complete_commit)(struct msm_kms *kms,struct drm_atomic_state *state);
@@ -65,6 +67,18 @@ struct msm_kms {int irq; }; +/** +* Subclass of drm_atomic_state, to allow kms backend to have driver +* private global state. The kms backend can do whatever it wants +* with the ->state ptr. On ->atomic_state_clear() the ->state ptr +* is kfree'd and set back to NULL. +*/ +struct msm_kms_state { +struct drm_atomic_state base; +void *state; +}; +#define to_kms_state (x) container_of(x, struct msm_kms_state, base) +static inline void msm_kms_init (struct msm_kms *kms,const struct msm_kms_funcs *funcs) {
Signature · 1 line
2.7.4
PATCH 3/5 drm/msm/mdp5: add skeletal mdp5_state
In reply to a message not in the archiveAdd basic state duplication/apply mechanism. Following commits will move actual global hw state into this.
The state_lock allows multiple concurrent updates to proceed as long as they don't both try to alter global state. The ww_mutex mechanism will trigger backoff in case of deadlock between multiple threads trying to update state.
--- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 43 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 22 +++++++++++++++++ 2 files changed, 65 insertions(+)
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +43 -0
@@ -72,6 +72,39 @@ static int mdp5_hw_init(struct msm_kms *kms)return 0 ; } +struct mdp5_state *mdp5_get_state (struct drm_atomic_state *s) +{ +struct msm_drm_private *priv = s->dev->dev_private; +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (priv->kms)); +struct msm_kms_state *state =to_kms_state (s); +struct mdp5_state *new_state; +int ret; + +if (state->state) +return state->state; + + ret =drm_modeset_lock (&mdp5_kms->state_lock, s->acquire_ctx); +if (ret) +return ERR_PTR (ret); + + new_state =kmalloc (sizeof (*mdp5_kms->state),GFP_KERNEL ); +if (!new_state) +return ERR_PTR (-ENOMEM ); + +/* Copy state: */ +/* TODO */ + + state->state = new_state; + +return new_state; +} + +static void mdp5_swap_state (struct msm_kms *kms,struct drm_atomic_state *state) +{ +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms)); +swap (to_kms_state (state)->state, mdp5_kms->state); +} +static void mdp5_prepare_commit (struct msm_kms *kms,struct drm_atomic_state *state) {struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms));
@@ -140,6 +173,7 @@ static const struct mdp_kms_funcs kms_funcs = { .irq = mdp5_irq, .enable_vblank = mdp5_enable_vblank, .disable_vblank = mdp5_disable_vblank, + .swap_state = mdp5_swap_state, .prepare_commit = mdp5_prepare_commit, .complete_commit = mdp5_complete_commit, .wait_for_crtc_commit_done = mdp5_wait_for_crtc_commit_done,
@@ -645,6 +679,8 @@ static void mdp5_destroy(struct platform_device *pdev)if (mdp5_kms->rpm_enabled)pm_runtime_disable (&pdev->dev); + +kfree (mdp5_kms->state); } static int construct_pipes(struct mdp5_kms *mdp5_kms, int cnt,
@@ -729,6 +765,13 @@ static int mdp5_init(struct platform_device *pdev, struct drm_device *dev) mdp5_kms->dev = dev; mdp5_kms->pdev = pdev; +drm_modeset_lock_init (&mdp5_kms->state_lock); + mdp5_kms->state =kzalloc (sizeof (*mdp5_kms->state),GFP_KERNEL ); +if (!mdp5_kms->state) { + ret = -ENOMEM ; + goto fail; + } + mdp5_kms->mmio =msm_ioremap (pdev,"mdp_phys" ,"MDP5" );if (IS_ERR (mdp5_kms->mmio)) { ret =PTR_ERR (mdp5_kms->mmio);
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +22 -0
@@ -27,6 +27,8 @@#include "mdp5_pipe.h" #include "mdp5_smp.h" +struct mdp5_state ; +struct mdp5_kms {struct mdp_kms base;
@@ -40,6 +42,11 @@ struct mdp5_kms {struct mdp5_cfg_handler *cfg;uint32_t caps;/* MDP capabilities (MDP_CAP_XXX bits) */ +/** +* Global atomic state. Do not access directly, use mdp5_get_state() +*/ +struct mdp5_state *state; +struct drm_modeset_lock state_lock;/* mapper-id used to request GEM buffer mapped for scanout: */ int id;
@@ -69,6 +76,21 @@ struct mdp5_kms { };#define to_mdp5_kms (x) container_of(x, struct mdp5_kms, base) +/* Global atomic state for tracking resources that are shared across +* multiple kms objects (planes/crtcs/etc). +* +* For atomic updates which require modifying global state, +*/ +struct mdp5_state { +uint32_t dummy; +}; + +struct mdp5_state *__must_check +mdp5_get_state (struct drm_atomic_state *s); + +/* Atomic plane state. Subclasses the base drm_plane_state in order to +* track assigned hwpipe and hw specific state. +*/ struct mdp5_plane_state {struct drm_plane_state base;
Signature · 6 lines
2.7.4 _______________________________________________ Freedreno mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/freedreno
PATCH 4/5 drm/msm/mdp5: dynamically assign hw pipes to planes
In reply to Rob Clark(re)assign the hw pipes to planes based on required caps, and to handle situations where we could not modify an in-use plane (ie. SMP block reallocation).
This means all planes advertise the superset of formats and properties. Userspace must (as always) use atomic TEST_ONLY step for atomic updates, as not all planes may be available for use on every frame.
The mapping of hwpipe to plane is stored in mdp5_state, so that state updates are atomically committed in the same way that plane/etc state updates are managed. This is needed because the mdp5_plane_state keeps a pointer to the hwpipe, and we don't want global state to become out of sync with the plane state if an atomic update fails, we hit deadlock/ backoff scenario, etc. The use of state_lock keeps multiple parallel updates which both re-assign hwpipes properly serialized.
--- drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c | 2 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 4 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 70 +++++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 10 ++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 158 ++++++++++++++++-------------- 6 files changed, 171 insertions(+), 80 deletions(-)
drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +1 -1
@@ -407,7 +407,7 @@ static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,for (i =0 ; i < cnt; i++) { pstates[i].state->stage =STAGE_BASE + i + base;DBG ("%s: assign pipe %s on stage=%d" , crtc->name, -pipe2name (mdp5_plane_pipe (pstates[i].plane)), + pstates[i].plane->name, pstates[i].state->stage); }
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +2 -2
@@ -92,7 +92,7 @@ struct mdp5_state *mdp5_get_state(struct drm_atomic_state *s)return ERR_PTR (-ENOMEM );/* Copy state: */ -/* TODO */ + new_state->hwpipe = mdp5_kms->state->hwpipe; state->state = new_state;
@@ -377,7 +377,7 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)struct drm_plane *plane;struct drm_crtc *crtc; - plane =mdp5_plane_init (dev, mdp5_kms->hwpipes[i], primary); + plane =mdp5_plane_init (dev, primary);if (IS_ERR (plane)) { ret =PTR_ERR (plane);dev_err (dev->dev,"failed to construct plane %d (%d)\n" , i, ret);
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +4 -3
@@ -82,7 +82,7 @@ struct mdp5_kms {* For atomic updates which require modifying global state, */ struct mdp5_state { -uint32_t dummy; +struct mdp5_hw_pipe_state hwpipe; }; struct mdp5_state *__must_check
@@ -94,6 +94,8 @@ mdp5_get_state(struct drm_atomic_state *s);struct mdp5_plane_state {struct drm_plane_state base; +struct mdp5_hw_pipe *hwpipe; +/* aligned with property */ uint8_t premultiplied;uint8_t zpos;
@@ -232,8 +234,7 @@ uint32_t mdp5_plane_get_flush(struct drm_plane *plane);void mdp5_plane_complete_commit (struct drm_plane *plane,struct drm_plane_state *state);enum mdp5_pipe mdp5_plane_pipe (struct drm_plane *plane); -struct drm_plane *mdp5_plane_init (struct drm_device *dev, -struct mdp5_hw_pipe *hwpipe,bool primary); +struct drm_plane *mdp5_plane_init (struct drm_device *dev,bool primary);uint32_t mdp5_crtc_vblank (struct drm_crtc *crtc);
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +70 -0
76 lines not on this page · raw message
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h +10 -0
@@ -32,6 +32,16 @@ struct mdp5_hw_pipe {uint32_t flush_mask;/* used to commit pipe registers */ }; +/* global atomic state of assignment between pipes and planes: */ +struct mdp5_hw_pipe_state { +struct drm_plane *hwpipe_to_plane[16 ]; +}; + +struct mdp5_hw_pipe *__must_check +mdp5_pipe_assign (struct drm_atomic_state *s,struct drm_plane *plane, +uint32_t caps); +void mdp5_pipe_release (struct drm_atomic_state *s,struct mdp5_hw_pipe *hwpipe); +struct mdp5_hw_pipe *mdp5_pipe_init (enum mdp5_pipe pipe,uint32_t reg_offset,uint32_t caps);void mdp5_pipe_destroy (struct mdp5_hw_pipe *hwpipe);
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +84 -74
261 lines not on this page · raw message
Signature · 6 lines
2.7.4 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH 5/5 drm/msm/mdp5: handle SMP block allocations "atomically"
In reply to Rob ClarkPreviously, SMP block allocation was not checked in the plane's atomic_check() fxn, so we could fail allocation SMP block allocation at atomic_update() time. Re-work the block allocation to request blocks during atomic_check(), but not update the hw until committing the atomic update.
Since SMP blocks allocated at atomic_check() time, we need to manage the SMP state as part of mdp5_state (global atomic state). This actually ends up significantly simplifying the SMP management, as the SMP module does not need to manage the intermediate state between assigning new blocks before setting flush bits and releasing old blocks after vblank. (The SMP registers and SMP allocation is not double-buffered, so newly allocated blocks need to be updated in kms->prepare_commit() released blocks in kms->complete_commit().)
--- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 11 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 1 + drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 21 ++- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 62 ++----- drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c | 270 ++++++++++-------------------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h | 66 ++++++-- 7 files changed, 193 insertions(+), 245 deletions(-)
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +10 -1
@@ -93,6 +93,8 @@ struct mdp5_state *mdp5_get_state(struct drm_atomic_state *s)/* Copy state: */ new_state->hwpipe = mdp5_kms->state->hwpipe; +if (mdp5_kms->smp) + new_state->smp = mdp5_kms->state->smp; state->state = new_state;
@@ -108,7 +110,11 @@ static void mdp5_swap_state(struct msm_kms *kms, struct drm_atomic_state *state)static void mdp5_prepare_commit (struct msm_kms *kms,struct drm_atomic_state *state) {struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms)); +mdp5_enable (mdp5_kms); + +if (mdp5_kms->smp) +mdp5_smp_prepare_commit (mdp5_kms->smp, &mdp5_kms->state->smp); }static void mdp5_complete_commit (struct msm_kms *kms,struct drm_atomic_state *state)
@@ -121,6 +127,9 @@ static void mdp5_complete_commit(struct msm_kms *kms, struct drm_atomic_state *sfor_each_plane_in_state (state, plane, plane_state, i)mdp5_plane_complete_commit (plane, plane_state); +if (mdp5_kms->smp) +mdp5_smp_complete_commit (mdp5_kms->smp, &mdp5_kms->state->smp); +mdp5_disable (mdp5_kms); }
@@ -825,7 +834,7 @@ static int mdp5_init(struct platform_device *pdev, struct drm_device *dev)* this section initializes the SMP: */ if (mdp5_kms->caps &MDP_CAP_SMP ) { - mdp5_kms->smp =mdp5_smp_init (mdp5_kms->dev, &config->hw->smp); + mdp5_kms->smp =mdp5_smp_init (mdp5_kms, &config->hw->smp);if (IS_ERR (mdp5_kms->smp)) { ret =PTR_ERR (mdp5_kms->smp); mdp5_kms->smp =NULL ;
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +1 -0
@@ -83,6 +83,7 @@ struct mdp5_kms {*/ struct mdp5_state {struct mdp5_hw_pipe_state hwpipe; +struct mdp5_smp_state smp; }; struct mdp5_state *__must_check
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +20 -1
@@ -18,7 +18,7 @@#include "mdp5_kms.h" struct mdp5_hw_pipe *mdp5_pipe_assign (struct drm_atomic_state *s, -struct drm_plane *plane,uint32_t caps) +struct drm_plane *plane,uint32_t caps,uint32_t blkcfg) {struct msm_drm_private *priv = s->dev->dev_private;struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (priv->kms));
@@ -63,6 +63,18 @@ struct mdp5_hw_pipe *mdp5_pipe_assign(struct drm_atomic_state *s,if (!hwpipe)return ERR_PTR (-ENOMEM ); +if (mdp5_kms->smp) { +int ret; + +DBG ("%s: alloc SMP blocks" , hwpipe->name); + ret =mdp5_smp_assign (mdp5_kms->smp, &state->smp, + hwpipe->pipe, blkcfg); +if (ret) +return ERR_PTR (-ENOMEM ); + + hwpipe->blkcfg = blkcfg; + } +DBG ("%s: assign to plane %s for caps %x" , hwpipe->name, plane->name, caps); new_state->hwpipe_to_plane[hwpipe->idx] = plane;
@@ -72,6 +84,8 @@ struct mdp5_hw_pipe *mdp5_pipe_assign(struct drm_atomic_state *s,void mdp5_pipe_release (struct drm_atomic_state *s,struct mdp5_hw_pipe *hwpipe) { +struct msm_drm_private *priv = s->dev->dev_private; +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (priv->kms));struct mdp5_state *state =mdp5_get_state (s);struct mdp5_hw_pipe_state *new_state = &state->hwpipe;
@@ -84,6 +98,11 @@ void mdp5_pipe_release(struct drm_atomic_state *s, struct mdp5_hw_pipe *hwpipe)DBG ("%s: release from plane %s" , hwpipe->name, new_state->hwpipe_to_plane[hwpipe->idx]->name); +if (mdp5_kms->smp) { +DBG ("%s: free SMP blocks" , hwpipe->name); +mdp5_smp_release (mdp5_kms->smp, &state->smp, hwpipe->pipe); + } + new_state->hwpipe_to_plane[hwpipe->idx] =NULL ; }
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h +6 -1
@@ -30,6 +30,11 @@ struct mdp5_hw_pipe {uint32_t caps;uint32_t flush_mask;/* used to commit pipe registers */ + +/* number of smp blocks per plane, ie: +* nblks_y | (nblks_u << 8) | (nblks_v << 16) +*/ +uint32_t blkcfg; };/* global atomic state of assignment between pipes and planes: */
@@ -39,7 +44,7 @@ struct mdp5_hw_pipe_state {struct mdp5_hw_pipe *__must_check mdp5_pipe_assign (struct drm_atomic_state *s,struct drm_plane *plane, -uint32_t caps); +uint32_t caps,uint32_t blkcfg);void mdp5_pipe_release (struct drm_atomic_state *s,struct mdp5_hw_pipe *hwpipe); struct mdp5_hw_pipe *mdp5_pipe_init(enum mdp5_pipe pipe,
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +13 -49
94 lines not on this page · raw message
drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c +87 -183
397 lines not on this page · raw message
drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h +56 -10
79 lines not on this page · raw message
Signature · 6 lines
2.7.4 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Archit Taneja
Reviewed-byRe: [PATCH 3/5] drm/msm/mdp5: add skeletal mdp5_state
In reply to a message not in the archiveQuoted 76 lines · Rob Clark
Add basic state duplication/apply mechanism. Following commits will move actual global hw state into this.
The
state_lockallows multiple concurrent updates to proceed as long as they don't both try to alter global state. Theww_mutexmechanism will trigger backoff in case of deadlock between multiple threads trying to update state.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 43 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 22 +++++++++++++++++ 2 files changed, 65 insertions(+)drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +36
-0@@ -72,6 +72,39 @@ static int mdp5_hw_init(struct msm_kms *kms)return 0 ; } +struct mdp5_state *mdp5_get_state (struct drm_atomic_state *s) +{ +struct msm_drm_private *priv = s->dev->dev_private; +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (priv->kms)); +struct msm_kms_state *state =to_kms_state (s); +struct mdp5_state *new_state; +int ret; + +if (state->state) +return state->state; + + ret =drm_modeset_lock (&mdp5_kms->state_lock, s->acquire_ctx); +if (ret) +return ERR_PTR (ret); + + new_state =kmalloc (sizeof (*mdp5_kms->state),GFP_KERNEL ); +if (!new_state) +return ERR_PTR (-ENOMEM ); + +/* Copy state: */ +/* TODO */ + + state->state = new_state; + +return new_state; +} + +static void mdp5_swap_state (struct msm_kms *kms,struct drm_atomic_state *state) +{ +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms)); +swap (to_kms_state (state)->state, mdp5_kms->state); +} +static void mdp5_prepare_commit (struct msm_kms *kms,struct drm_atomic_state *state) {struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms));@@ -140,6 +173,7 @@ static const struct mdp_kms_funcs kms_funcs = { .irq = mdp5_irq, .enable_vblank = mdp5_enable_vblank, .disable_vblank = mdp5_disable_vblank, + .swap_state = mdp5_swap_state, .prepare_commit = mdp5_prepare_commit, .complete_commit = mdp5_complete_commit, .wait_for_crtc_commit_done = mdp5_wait_for_crtc_commit_done,@@ -645,6 +679,8 @@ static void mdp5_destroy(struct platform_device *pdev)if (mdp5_kms->rpm_enabled)pm_runtime_disable (&pdev->dev); + +kfree (mdp5_kms->state); } static int construct_pipes(struct mdp5_kms *mdp5_kms, int cnt,
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c static int mdp5_init(struct platform_device *pdev, struct drm_device *dev) · line 774
Quoted 10 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +7
-0@@ -729,6 +765,13 @@ static int mdp5_init(struct platform_device *pdev, struct drm_device *dev) mdp5_kms->dev = dev; mdp5_kms->pdev = pdev; +drm_modeset_lock_init (&mdp5_kms->state_lock); + mdp5_kms->state =kzalloc (sizeof (*mdp5_kms->state),GFP_KERNEL ); +if (!mdp5_kms->state) { + ret = -ENOMEM ; + goto fail; + } +
This would probably be better in mdp5_kms_init() since it's intializing kms stuff, and not hw resources. Otherwise:
Quoted 51 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +0
-0mdp5_kms->mmio =msm_ioremap (pdev,"mdp_phys" ,"MDP5" );if (IS_ERR (mdp5_kms->mmio)) { ret =PTR_ERR (mdp5_kms->mmio);drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +22
-0@@ -27,6 +27,8 @@#include "mdp5_pipe.h" #include "mdp5_smp.h" +struct mdp5_state ; +struct mdp5_kms {struct mdp_kms base;@@ -40,6 +42,11 @@ struct mdp5_kms {struct mdp5_cfg_handler *cfg;uint32_t caps;/* MDP capabilities (MDP_CAP_XXX bits) */ +/** +* Global atomic state. Do not access directly, use mdp5_get_state() +*/ +struct mdp5_state *state; +struct drm_modeset_lock state_lock;/* mapper-id used to request GEM buffer mapped for scanout: */ int id;@@ -69,6 +76,21 @@ struct mdp5_kms { };#define to_mdp5_kms (x) container_of(x, struct mdp5_kms, base) +/* Global atomic state for tracking resources that are shared across +* multiple kms objects (planes/crtcs/etc). +* +* For atomic updates which require modifying global state, +*/ +struct mdp5_state { +uint32_t dummy; +}; + +struct mdp5_state *__must_check +mdp5_get_state (struct drm_atomic_state *s); + +/* Atomic plane state. Subclasses the base drm_plane_state in order to +* track assigned hwpipe and hw specific state. +*/ struct mdp5_plane_state {struct drm_plane_state base;
Signature · 6 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project _______________________________________________ Freedreno mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/freedreno
Re: [PATCH 4/5] drm/msm/mdp5: dynamically assign hw pipes to planes
In reply to Rob ClarkHi,
Minor comments below. LGTM otherwise.
Quoted 66 lines · Rob Clark
(re)assign the hw pipes to planes based on required caps, and to handle situations where we could not modify an in-use plane (ie. SMP block reallocation).
This means all planes advertise the superset of formats and properties. Userspace must (as always) use atomic
TEST_ONLYstep for atomic updates, as not all planes may be available for use on every frame.The mapping of hwpipe to plane is stored in
mdp5_state, so that state updates are atomically committed in the same way that plane/etc state updates are managed. This is needed because themdp5_plane_statekeeps a pointer to the hwpipe, and we don't want global state to become out of sync with the plane state if an atomic update fails, we hit deadlock/ backoff scenario, etc. The use ofstate_lockkeeps multiple parallel updates which both re-assign hwpipes properly serialized.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c | 2 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 4 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 70 +++++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 10 ++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 158 ++++++++++++++++-------------- 6 files changed, 171 insertions(+), 80 deletions(-)drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c +1
-1@@ -407,7 +407,7 @@ static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,for (i =0 ; i < cnt; i++) { pstates[i].state->stage =STAGE_BASE + i + base;DBG ("%s: assign pipe %s on stage=%d" , crtc->name, -pipe2name (mdp5_plane_pipe (pstates[i].plane)), + pstates[i].plane->name, pstates[i].state->stage); }drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +2
-2@@ -92,7 +92,7 @@ struct mdp5_state *mdp5_get_state(struct drm_atomic_state *s)return ERR_PTR (-ENOMEM );/* Copy state: */ -/* TODO */ + new_state->hwpipe = mdp5_kms->state->hwpipe; state->state = new_state;@@ -377,7 +377,7 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)struct drm_plane *plane;struct drm_crtc *crtc; - plane =mdp5_plane_init (dev, mdp5_kms->hwpipes[i], primary); + plane =mdp5_plane_init (dev, primary);if (IS_ERR (plane)) { ret =PTR_ERR (plane);dev_err (dev->dev,"failed to construct plane %d (%d)\n" , i, ret);drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +0
-0
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h struct mdp5_kms { · line 85
Quoted 5 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +1
-1@@ -82,7 +82,7 @@ struct mdp5_kms {* For atomic updates which require modifying global state, */ struct mdp5_state { -uint32_t dummy; +struct mdp5_hw_pipe_state hwpipe;
I was wondering if we could rename the above as hwpipes/hwpipe_state/hwpipe_map, because it gets a confusing since variables of the type mdp5_hw_pipe are also named hwpipe.
Quoted 30 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +3
-2}; struct mdp5_state *__must_check@@ -94,6 +94,8 @@ mdp5_get_state(struct drm_atomic_state *s);struct mdp5_plane_state {struct drm_plane_state base; +struct mdp5_hw_pipe *hwpipe; +/* aligned with property */ uint8_t premultiplied;uint8_t zpos;@@ -232,8 +234,7 @@ uint32_t mdp5_plane_get_flush(struct drm_plane *plane);void mdp5_plane_complete_commit (struct drm_plane *plane,struct drm_plane_state *state);enum mdp5_pipe mdp5_plane_pipe (struct drm_plane *plane); -struct drm_plane *mdp5_plane_init (struct drm_device *dev, -struct mdp5_hw_pipe *hwpipe,bool primary); +struct drm_plane *mdp5_plane_init (struct drm_device *dev,bool primary);uint32_t mdp5_crtc_vblank (struct drm_crtc *crtc);drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +0
-0@@ -17,6 +17,76 @@#include "mdp5_kms.h"
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c line 29
Quoted 10 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +10
-0+struct mdp5_hw_pipe *mdp5_pipe_assign (struct drm_atomic_state *s, +struct drm_plane *plane,uint32_t caps) +{ +struct msm_drm_private *priv = s->dev->dev_private; +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (priv->kms)); +struct mdp5_state *state =mdp5_get_state (s); +struct mdp5_hw_pipe_state *old_state, *new_state; +struct mdp5_hw_pipe *hwpipe =NULL ; +int i; +
Can we assign "state = mdp5_get_state(s);" here instead for clarity? Since the func does more than just getting the mdp5_state pointer?
Quoted 6 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +6
-0+if (IS_ERR (state)) +return ERR_CAST (state); + +/* grab old_state after mdp5_get_state(), since now we hold lock: */ + old_state = &mdp5_kms->state->hwpipe; + new_state = &state->hwpipe;
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c line 45
Quoted 10 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +10
-0+ +for (i =0 ; i < mdp5_kms->num_hwpipes; i++) { +struct mdp5_hw_pipe *cur = mdp5_kms->hwpipes[i]; + +/* skip if already in-use.. check both new and old state, +* since we cannot immediately re-use a pipe that is +* released in the current update in some cases: +* (1) mdp5 has SMP (non-double-buffered) +* (2) hw pipe previously assigned to different CRTC +* (vblanks might not be aligned)
For 1), we have non-SMP SoCs (a.k.a 8996), and for 2) we would have the info whether the hwpipe is changing crtcs in old and new states. I guess we could leave this as an optimization for the future.
Quoted 52 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c +44
-0+*/ +if (new_state->hwpipe_to_plane[cur->idx] || + old_state->hwpipe_to_plane[cur->idx]) +continue ; + +/* skip if doesn't support some required caps: */ +if (caps & ~cur->caps) +continue ; + +/* possible candidate, take the one with the +* fewest unneeded caps bits set: +*/ +if (!hwpipe || (hweight_long (cur->caps & ~caps) < +hweight_long (hwpipe->caps & ~caps))) + hwpipe = cur; + } + +if (!hwpipe) +return ERR_PTR (-ENOMEM ); + +DBG ("%s: assign to plane %s for caps %x" , + hwpipe->name, plane->name, caps); + new_state->hwpipe_to_plane[hwpipe->idx] = plane; + +return hwpipe; +} + +void mdp5_pipe_release (struct drm_atomic_state *s,struct mdp5_hw_pipe *hwpipe) +{ +struct mdp5_state *state =mdp5_get_state (s); +struct mdp5_hw_pipe_state *new_state = &state->hwpipe; + +if (!hwpipe) +return ; + +if (WARN_ON (!new_state->hwpipe_to_plane[hwpipe->idx])) +return ; + +DBG ("%s: release from plane %s" , hwpipe->name, + new_state->hwpipe_to_plane[hwpipe->idx]->name); + + new_state->hwpipe_to_plane[hwpipe->idx] =NULL ; +} +void mdp5_pipe_destroy (struct mdp5_hw_pipe *hwpipe) {kfree (hwpipe);drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h +0
-0
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h struct mdp5_hw_pipe { · line 37
Quoted 6 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h +3
-0@@ -32,6 +32,16 @@ struct mdp5_hw_pipe {uint32_t flush_mask;/* used to commit pipe registers */ }; +/* global atomic state of assignment between pipes and planes: */ +struct mdp5_hw_pipe_state { +struct drm_plane *hwpipe_to_plane[16 ];
We could use SSPP_MAX here instead of 16. We would need to move its definition from mdp5_crtc.c to mdp5_kms.h, though.
Quoted 167 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h +7
-0+}; + +struct mdp5_hw_pipe *__must_check +mdp5_pipe_assign (struct drm_atomic_state *s,struct drm_plane *plane, +uint32_t caps); +void mdp5_pipe_release (struct drm_atomic_state *s,struct mdp5_hw_pipe *hwpipe); +struct mdp5_hw_pipe *mdp5_pipe_init (enum mdp5_pipe pipe,uint32_t reg_offset,uint32_t caps);void mdp5_pipe_destroy (struct mdp5_hw_pipe *hwpipe);drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +48
-52@@ -22,8 +22,6 @@struct mdp5_plane {struct drm_plane base; -struct mdp5_hw_pipe *hwpipe; -uint32_t nformats;uint32_t formats[32 ]; };@@ -63,12 +61,6 @@ static void mdp5_plane_destroy(struct drm_plane *plane)static void mdp5_plane_install_rotation_property (struct drm_device *dev,struct drm_plane *plane) { -struct mdp5_plane *mdp5_plane =to_mdp5_plane (plane); - -if (!(mdp5_plane->hwpipe->caps &MDP_PIPE_CAP_HFLIP ) && - !(mdp5_plane->hwpipe->caps &MDP_PIPE_CAP_VFLIP )) -return ; - drm_plane_create_rotation_property(plane, DRM_ROTATE_0, DRM_ROTATE_0 |@@ -184,6 +176,8 @@ mdp5_plane_atomic_print_state(struct drm_printer *p, {struct mdp5_plane_state *pstate =to_mdp5_plane_state (state); +drm_printf (p,"\thwpipe=%s\n" , pstate->hwpipe ? + pstate->hwpipe->name :"(null)" );drm_printf (p,"\tpremultiplied=%u\n" , pstate->premultiplied);drm_printf (p,"\tzpos=%u\n" , pstate->zpos);drm_printf (p,"\talpha=%u\n" , pstate->alpha);@@ -237,10 +231,12 @@ mdp5_plane_duplicate_state(struct drm_plane *plane)static void mdp5_plane_destroy_state (struct drm_plane *plane,struct drm_plane_state *state) { +struct mdp5_plane_state *pstate =to_mdp5_plane_state (state); +if (state->fb)drm_framebuffer_unreference (state->fb); -kfree (to_mdp5_plane_state (state)); +kfree (pstate); }static const struct drm_plane_funcs mdp5_plane_funcs = {@@ -285,70 +281,81 @@ static void mdp5_plane_cleanup_fb(struct drm_plane *plane,static int mdp5_plane_atomic_check (struct drm_plane *plane,struct drm_plane_state *state) { -struct mdp5_plane *mdp5_plane =to_mdp5_plane (plane); +struct mdp5_plane_state *mdp5_state =to_mdp5_plane_state (state);struct drm_plane_state *old_state = plane->state; -const struct mdp_format *format; -bool vflip, hflip; +bool new_hwpipe = false; +uint32_t caps =0 ;DBG ("%s: check (%d -> %d)" , plane->name,plane_enabled (old_state),plane_enabled (state)); +/* We don't allow faster-than-vblank updates.. if we did add this +* some day, we would need to disallow in cases where hwpipe +* changes +*/ +if (WARN_ON (to_mdp5_plane_state (old_state)->pending)) +return -EBUSY ; +if (plane_enabled (state)) {unsigned int rotation; +const struct mdp_format *format; format =to_mdp_format (msm_framebuffer_format (state->fb)); -if (MDP_FORMAT_IS_YUV (format) && - !pipe_supports_yuv (mdp5_plane->hwpipe->caps)) { -DBG ("Pipe doesn't support YUV\n" ); - -return -EINVAL ; - } - -if (!(mdp5_plane->hwpipe->caps &MDP_PIPE_CAP_SCALE ) && - (((state->src_w >>16 ) != state->crtc_w) || - ((state->src_h >>16 ) != state->crtc_h))) { -DBG ("Pipe doesn't support scaling (%dx%d -> %dx%d)\n" , - state->src_w >>16 , state->src_h >>16 , - state->crtc_w, state->crtc_h); +if (MDP_FORMAT_IS_YUV (format)) + caps |=MDP_PIPE_CAP_SCALE |MDP_PIPE_CAP_CSC ; -return -EINVAL ; - } +if (((state->src_w >>16 ) != state->crtc_w) || + ((state->src_h >>16 ) != state->crtc_h)) + caps |=MDP_PIPE_CAP_SCALE ; rotation =drm_rotation_simplify (state->rotation,DRM_ROTATE_0 |DRM_REFLECT_X |DRM_REFLECT_Y ); - hflip = !!(rotation &DRM_REFLECT_X ); - vflip = !!(rotation &DRM_REFLECT_Y ); - -if ((vflip && !(mdp5_plane->hwpipe->caps &MDP_PIPE_CAP_VFLIP )) || - (hflip && !(mdp5_plane->hwpipe->caps &MDP_PIPE_CAP_HFLIP ))) { -DBG ("Pipe doesn't support flip\n" ); - -return -EINVAL ; +if (rotation &DRM_REFLECT_X ) + caps |=MDP_PIPE_CAP_HFLIP ; + +if (rotation &DRM_REFLECT_Y ) + caps |=MDP_PIPE_CAP_VFLIP ; + +/* (re)allocate hw pipe if we don't have one or caps-mismatch: */ +if (!mdp5_state->hwpipe || (caps & ~mdp5_state->hwpipe->caps)) + new_hwpipe = true; + +if (plane_enabled (old_state)) { +bool full_modeset = false; +if (state->fb->pixel_format != old_state->fb->pixel_format) { +DBG ("%s: pixel_format change!" , plane->name); + full_modeset = true; + } +if (state->src_w != old_state->src_w) { +DBG ("%s: src_w change!" , plane->name); + full_modeset = true; + } +if (full_modeset) { +/* cannot change SMP block allocation during +* scanout: +*/ +if (get_kms (plane)->smp) + new_hwpipe = true; + } } - } -if (plane_enabled (state) &&plane_enabled (old_state)) { -/* we cannot change SMP block configuration during scanout: */ -bool full_modeset = false; -if (state->fb->pixel_format != old_state->fb->pixel_format) { -DBG ("%s: pixel_format change!" , plane->name); - full_modeset = true; - } -if (state->src_w != old_state->src_w) { -DBG ("%s: src_w change!" , plane->name); - full_modeset = true; - } -if (to_mdp5_plane_state (old_state)->pending) { -DBG ("%s: still pending!" , plane->name); - full_modeset = true; - } -if (full_modeset) {
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c static void mdp5_plane_cleanup_fb(struct drm_plane *plane, · line 351
Quoted 10 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +7
-3-struct drm_crtc_state *crtc_state = -drm_atomic_get_crtc_state (state->state, state->crtc); - crtc_state->mode_changed = true; +/* (re)assign hwpipe if needed, otherwise keep old one: */ +if (new_hwpipe) { +/* TODO maybe we want to re-assign hwpipe sometimes +* in cases when we no-longer need some caps to make +* it available for other planes? +*/ +struct mdp5_hw_pipe *hwpipe = mdp5_state->hwpipe;
Maybe hwpipe above could be curr_hwpipe or old_hwpipe?
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c static void mdp5_plane_cleanup_fb(struct drm_plane *plane, · line 360
Quoted 9 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +7
-0+ mdp5_state->hwpipe = +mdp5_pipe_assign (state->state, plane, caps); +if (IS_ERR (mdp5_state->hwpipe)) { +DBG ("%s: failed to assign hwpipe!" , plane->name); +return PTR_ERR (mdp5_state->hwpipe); + } +mdp5_pipe_release (state->state, hwpipe); } }
Thanks,
Archit
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Re: [PATCH 1/5] drm/msm/mdp5: introduce mdp5_hw_pipe
In reply to Rob ClarkQuoted 62 lines · Rob Clark
Split out the hardware pipe specifics from
mdp5_plane. To start, the hw pipes are statically assigned to planes, but next step is to assign the hw pipes duringplane->atomic_check()based on requested caps (scaling, YUV, etc). And then hw pipe re-assignment if required if required SMP blocks changes.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/Makefile | 1 + drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 126 +++++++++++++++++++----------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 43 ++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 39 +++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 66 +++++++--------- 6 files changed, 197 insertions(+), 85 deletions(-) create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.hdrivers/gpu/drm/msm/Makefile +1
-0@@ -37,6 +37,7 @@ msm-y := \ mdp/mdp5/mdp5_irq.o \ mdp/mdp5/mdp5_mdss.o \ mdp/mdp5/mdp5_kms.o \ + mdp/mdp5/mdp5_pipe.o \ mdp/mdp5/mdp5_plane.o \ mdp/mdp5/mdp5_smp.o \ msm_atomic.o \drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +4
-9@@ -119,6 +119,10 @@ static void mdp5_kms_destroy(struct msm_kms *kms) {struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms));struct msm_gem_address_space *aspace = mdp5_kms->aspace; +int i; + +for (i =0 ; i < mdp5_kms->num_hwpipes; i++) +mdp5_pipe_destroy (mdp5_kms->hwpipes[i]);if (aspace) { aspace->mmu->funcs->detach(aspace->mmu,@@ -323,15 +327,6 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)static int modeset_init (struct mdp5_kms *mdp5_kms) { -static const enum mdp5_pipe rgb_planes[] = { -SSPP_RGB0 ,SSPP_RGB1 ,SSPP_RGB2 ,SSPP_RGB3 , - }; -static const enum mdp5_pipe vig_planes[] = { -SSPP_VIG0 ,SSPP_VIG1 ,SSPP_VIG2 ,SSPP_VIG3 , - }; -static const enum mdp5_pipe dma_planes[] = { -SSPP_DMA0 ,SSPP_DMA1 , - };struct drm_device *dev = mdp5_kms->dev;struct msm_drm_private *priv = dev->dev_private;const struct mdp5_cfg_hw *hw_cfg;
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c static int modeset_init(struct mdp5_kms *mdp5_kms) · line 340
Quoted 9 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +4
-2@@ -339,58 +334,34 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) hw_cfg =mdp5_cfg_get_hw_config (mdp5_kms->cfg); -/* construct CRTCs and their private planes: */ -for (i =0 ; i < hw_cfg->pipe_rgb.count; i++) { +/* Construct planes equaling the number of hw pipes, and CRTCs +* for the N layer-mixers (LM). The first N planes become primary +* planes for the CRTCs, with the remainder as overlay planes: +*/
Jfyi, we might need to change this a bit in the future. It'll be better to get the max number of displays connected on our platform via parsing DT, etc, and calculate CRTCs based on that, and not number of layermixers. Maybe add a couple for writeback too. This way, we get the right number of CRTCs, and we don't rely on #LMs, since we can have 2 per crtc in the future.
Signature · 6 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 3/5] drm/msm/mdp5: add skeletal mdp5_state
In reply to Archit TanejaQuoted 89 lines · Archit Taneja
Quoted 86 lines · Rob Clark
Add basic state duplication/apply mechanism. Following commits will move actual global hw state into this.
The
state_lockallows multiple concurrent updates to proceed as long as they don't both try to alter global state. Theww_mutexmechanism will trigger backoff in case of deadlock between multiple threads trying to update state.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 43 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 22 +++++++++++++++++ 2 files changed, 65 insertions(+)drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +33
-0@@ -72,6 +72,39 @@ static int mdp5_hw_init(struct msm_kms *kms)return 0 ; } +struct mdp5_state *mdp5_get_state (struct drm_atomic_state *s) +{ +struct msm_drm_private *priv = s->dev->dev_private; +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (priv->kms)); +struct msm_kms_state *state =to_kms_state (s); +struct mdp5_state *new_state; +int ret; + +if (state->state) +return state->state; + + ret =drm_modeset_lock (&mdp5_kms->state_lock, s->acquire_ctx); +if (ret) +return ERR_PTR (ret); + + new_state =kmalloc (sizeof (*mdp5_kms->state),GFP_KERNEL ); +if (!new_state) +return ERR_PTR (-ENOMEM ); + +/* Copy state: */ +/* TODO */ + + state->state = new_state; + +return new_state; +} + +static void mdp5_swap_state (struct msm_kms *kms,struct drm_atomic_state *state) +{ +struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms)); +swap (to_kms_state (state)->state, mdp5_kms->state); +} +static void mdp5_prepare_commit (struct msm_kms *kms,struct drm_atomic_state *state){ struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms));drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +1
-0@@ -140,6 +173,7 @@ static const struct mdp_kms_funcs kms_funcs = { .irq = mdp5_irq, .enable_vblank = mdp5_enable_vblank, .disable_vblank = mdp5_disable_vblank, + .swap_state = mdp5_swap_state, .prepare_commit = mdp5_prepare_commit, .complete_commit = mdp5_complete_commit, .wait_for_crtc_commit_done =
mdp5_wait_for_crtc_commit_done,drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +2
-0@@ -645,6 +679,8 @@ static void mdp5_destroy(struct platform_device *pdev)if (mdp5_kms->rpm_enabled)pm_runtime_disable (&pdev->dev); + +kfree (mdp5_kms->state); } static int construct_pipes(struct mdp5_kms *mdp5_kms, int cnt,@@ -729,6 +765,13 @@ static int mdp5_init(struct platform_device *pdev, struct drm_device *dev) mdp5_kms->dev = dev; mdp5_kms->pdev = pdev;
Quoted 12 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c static int mdp5_init(struct platform_device *pdev, struct drm_device *dev) · line 774
Quoted 7 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +7
-0+drm_modeset_lock_init (&mdp5_kms->state_lock); + mdp5_kms->state =kzalloc (sizeof (*mdp5_kms->state),GFP_KERNEL ); +if (!mdp5_kms->state) { + ret = -ENOMEM ; + goto fail; + } +This would probably be better in
mdp5_kms_init()since it's intializing kms stuff, and
not hw resources. Otherwise:
it ends up needing to be before mdp5_smp_init() unfortunately, so things can be setup properly for hw that has reserved blocks.
BR,
-R
Quoted 53 lines
Reviewed-by Archit Taneja <[email protected]>Quoted 49 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +0
-0mdp5_kms->mmio =msm_ioremap (pdev,"mdp_phys" ,"MDP5" );if (IS_ERR (mdp5_kms->mmio)) {ret = PTR_ERR(mdp5_kms->mmio);drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +7
-0@@ -27,6 +27,8 @@#include "mdp5_pipe.h" #include "mdp5_smp.h" +struct mdp5_state ; +struct mdp5_kms {struct mdp_kms base;@@ -40,6 +42,11 @@ struct mdp5_kms {struct mdp5_cfg_handler *cfg;uint32_t caps;/* MDP capabilities (MDP_CAP_XXX bits) */ +/** +* Global atomic state. Do not access directly, use mdp5_get_state() +*/ +struct mdp5_state *state; +struct drm_modeset_lock state_lock;/* mapper-id used to request GEM buffer mapped for scanout: */ int id;drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +15
-0@@ -69,6 +76,21 @@ struct mdp5_kms { };#define to_mdp5_kms (x) container_of(x, struct mdp5_kms, base) +/* Global atomic state for tracking resources that are shared across +* multiple kms objects (planes/crtcs/etc). +* +* For atomic updates which require modifying global state, +*/ +struct mdp5_state { +uint32_t dummy; +}; + +struct mdp5_state *__must_check +mdp5_get_state (struct drm_atomic_state *s); + +/* Atomic plane state. Subclasses the base drm_plane_state in order to +* track assigned hwpipe and hw specific state. +*/
Quoted 8 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h struct mdp5_kms { · line 96
Quoted 4 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h +0
-0struct mdp5_plane_state {struct drm_plane_state base;--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 1/5] drm/msm/mdp5: introduce mdp5_hw_pipe
In reply to a message not in the archiveQuoted 72 lines · Archit Taneja
Quoted 69 lines · Rob Clark
Split out the hardware pipe specifics from
mdp5_plane. To start, the hw pipes are statically assigned to planes, but next step is to assign the hw pipes duringplane->atomic_check()based on requested caps (scaling, YUV, etc). And then hw pipe re-assignment if required if required SMP blocks changes.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/Makefile | 1 + drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 126 +++++++++++++++++++----------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 43 ++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 39 +++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 66 +++++++--------- 6 files changed, 197 insertions(+), 85 deletions(-) create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.hdrivers/gpu/drm/msm/Makefile +1
-0@@ -37,6 +37,7 @@ msm-y := \ mdp/mdp5/mdp5_irq.o \ mdp/mdp5/mdp5_mdss.o \ mdp/mdp5/mdp5_kms.o \ + mdp/mdp5/mdp5_pipe.o \ mdp/mdp5/mdp5_plane.o \ mdp/mdp5/mdp5_smp.o \ msm_atomic.o \drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +4
-9@@ -119,6 +119,10 @@ static void mdp5_kms_destroy(struct msm_kms *kms) {struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms));struct msm_gem_address_space *aspace = mdp5_kms->aspace; +int i; + +for (i =0 ; i < mdp5_kms->num_hwpipes; i++) +mdp5_pipe_destroy (mdp5_kms->hwpipes[i]);if (aspace) { aspace->mmu->funcs->detach(aspace->mmu,@@ -323,15 +327,6 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)static int modeset_init (struct mdp5_kms *mdp5_kms) { -static const enum mdp5_pipe rgb_planes[] = { -SSPP_RGB0 ,SSPP_RGB1 ,SSPP_RGB2 ,SSPP_RGB3 , - }; -static const enum mdp5_pipe vig_planes[] = { -SSPP_VIG0 ,SSPP_VIG1 ,SSPP_VIG2 ,SSPP_VIG3 , - }; -static const enum mdp5_pipe dma_planes[] = { -SSPP_DMA0 ,SSPP_DMA1 , - };struct drm_device *dev = mdp5_kms->dev;struct msm_drm_private *priv = dev->dev_private;const struct mdp5_cfg_hw *hw_cfg;drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +0
-0@@ -339,58 +334,34 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) hw_cfg =mdp5_cfg_get_hw_config (mdp5_kms->cfg);
Quoted 14 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c static int modeset_init(struct mdp5_kms *mdp5_kms) · line 340
Quoted 6 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +4
-2-/* construct CRTCs and their private planes: */ -for (i =0 ; i < hw_cfg->pipe_rgb.count; i++) { +/* Construct planes equaling the number of hw pipes, and CRTCs +* for the N layer-mixers (LM). The first N planes become primary +* planes for the CRTCs, with the remainder as overlay planes: +*/ Jfyi, we might need to change this a bit in the future. It'll be better to get the max number of displays connected on our platform via parsing DT, etc, and calculate CRTCs based on that, and not number of layermixers. Maybe add a couple for writeback too. This way, we get the right number of CRTCs, and we don't rely on #LMs, since we can have 2 per crtc in the future.
I *guess* when we get to that stage, we'll dynamically assign LM's too, in a similar way as hwpipe. And I suppose we could also put a cap on # of crtc's based on # of encoders?
BR,
-R
Quoted 5 lines
Reviewed-by Archit Taneja <[email protected]>--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno
Re: [PATCH 1/5] drm/msm/mdp5: introduce mdp5_hw_pipe
Quoted 88 lines · Rob Clark
Quoted 72 lines · Archit Taneja
Quoted 69 lines · Rob Clark
Split out the hardware pipe specifics from
mdp5_plane. To start, the hw pipes are statically assigned to planes, but next step is to assign the hw pipes duringplane->atomic_check()based on requested caps (scaling, YUV, etc). And then hw pipe re-assignment if required if required SMP blocks changes.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/Makefile | 1 + drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c | 126 +++++++++++++++++++----------- drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h | 7 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c | 43 ++++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h | 39 +++++++++ drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 66 +++++++--------- 6 files changed, 197 insertions(+), 85 deletions(-) create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c create mode 100644 drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.hdrivers/gpu/drm/msm/Makefile +1
-0@@ -37,6 +37,7 @@ msm-y := \ mdp/mdp5/mdp5_irq.o \ mdp/mdp5/mdp5_mdss.o \ mdp/mdp5/mdp5_kms.o \ + mdp/mdp5/mdp5_pipe.o \ mdp/mdp5/mdp5_plane.o \ mdp/mdp5/mdp5_smp.o \ msm_atomic.o \drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +4
-9@@ -119,6 +119,10 @@ static void mdp5_kms_destroy(struct msm_kms *kms) {struct mdp5_kms *mdp5_kms =to_mdp5_kms (to_mdp_kms (kms));struct msm_gem_address_space *aspace = mdp5_kms->aspace; +int i; + +for (i =0 ; i < mdp5_kms->num_hwpipes; i++) +mdp5_pipe_destroy (mdp5_kms->hwpipes[i]);if (aspace) { aspace->mmu->funcs->detach(aspace->mmu,@@ -323,15 +327,6 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)static int modeset_init (struct mdp5_kms *mdp5_kms) { -static const enum mdp5_pipe rgb_planes[] = { -SSPP_RGB0 ,SSPP_RGB1 ,SSPP_RGB2 ,SSPP_RGB3 , - }; -static const enum mdp5_pipe vig_planes[] = { -SSPP_VIG0 ,SSPP_VIG1 ,SSPP_VIG2 ,SSPP_VIG3 , - }; -static const enum mdp5_pipe dma_planes[] = { -SSPP_DMA0 ,SSPP_DMA1 , - };struct drm_device *dev = mdp5_kms->dev;struct msm_drm_private *priv = dev->dev_private;const struct mdp5_cfg_hw *hw_cfg;drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +0
-0@@ -339,58 +334,34 @@ static int modeset_init(struct mdp5_kms *mdp5_kms) hw_cfg =mdp5_cfg_get_hw_config (mdp5_kms->cfg);Quoted 14 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c static int modeset_init(struct mdp5_kms *mdp5_kms) · line 340
Quoted 6 lines
drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c +4
-2-/* construct CRTCs and their private planes: */ -for (i =0 ; i < hw_cfg->pipe_rgb.count; i++) { +/* Construct planes equaling the number of hw pipes, and CRTCs +* for the N layer-mixers (LM). The first N planes become primary +* planes for the CRTCs, with the remainder as overlay planes: +*/ Jfyi, we might need to change this a bit in the future. It'll be better to get the max number of displays connected on our platform via parsing DT, etc, and calculate CRTCs based on that, and not number of layermixers. Maybe add a couple for writeback too. This way, we get the right number of CRTCs, and we don't rely on #LMs, since we can have 2 per crtc in the future.
Quoted 3 lines
I *guess* when we get to that stage, we'll dynamically assign LM's too, in a similar way as hwpipe. And I suppose we could also put a cap on # of crtc's based on # of encoders?
Yeah. Currently, we end up creating 2 encoders for each DSI instance, i.e, 1 for command mode, and another for video mode. Only one can be used at a time. If we adjust for that, then I guess # of crtcs should equal to # of encoders.
Archit
Quoted 14 lines
BR,
-RQuoted 5 lines
Reviewed-by Archit Taneja <[email protected]>--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project--
To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Signature · 6 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel