1
0
Fork 0
mutter-performance-source/src/backends/native/meta-kms.c

807 lines
21 KiB
C
Raw Normal View History

backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
/*
* Copyright (C) 2018 Red Hat
kms: Make GSource ready by default When testing a laptop with intel and DisplayLink devices, attempting to set the DL output as the only active output resulted in GNOME/Wayland freezing. The main event loop was running fine, but nothing on screen would get updated once the DL output become the only one. This patch fixes that issue. DisplayLink USB 3 devices use an out-of-tree kernel DRM driver called EVDI. EVDI can sometimes fail drmModePageFlip(). For me, the flip fails reliably when hotplugging the DL dock and when changing display configuration to DL only. Mutter has a workaround for failing flips, it just calls drmModeSetCrtc() and that succeeds. What does not work reliably in the fallback path is Mutter keeping track of the pageflip. Since drmModePageFlip() failed, there will not be a pageflip event coming and instead Mutter queues a callback in its stead. When you have more than one output, some other output repainting will attempt to swap buffers and calls wait_for_pending_flips() which has the side-effect of dispatching any queued flip callbacks. With multiple outputs, you don't get stuck (unless they all fail the exact same way at the same time?). When you have only one output, it cannot proceed to repaint and buffer swap because the pageflip is not marked complete yet. Nothing dispatches the flip callback, leading to the freeze. The flip callback is intended to be an idle callback, implemented with a GSource. It is supposed to be called as soon as execution returns to the main event loop. The setup of the GSource is incomplete, so it will never dispatch. Fix the GSource setup by setting its ready-time to be always in the past. That gets it dispatched on the next cycle of the main event loop. This is now the default behavior for all sources created by meta_kms_add_source_in_impl(). Sources that need a delay continue to do that by overriding the ready-time explicitly. An alternative solution could have been to implement GSource prepare and check callbacks returning TRUE. However, since meta_kms_add_source_in_impl() is used by flip retry code as well, and that code needs a delay through the ready-time, I was afraid I might break the flip retry code. Hence I decided to use ready-time instead. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1209
2020-04-17 11:50:13 +00:00
* Copyright 2020 DisplayLink (UK) Ltd.
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 of the
* License, or (at your option) any later version.
*
* 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, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
* 02111-1307, USA.
*/
#include "config.h"
#include "backends/native/meta-kms-private.h"
#include "backends/native/meta-backend-native.h"
#include "backends/native/meta-kms-device-private.h"
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
#include "backends/native/meta-kms-impl.h"
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
#include "backends/native/meta-kms-update-private.h"
#include "backends/native/meta-udev.h"
#include "cogl/cogl.h"
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
#include "meta-private-enum-types.h"
/**
* SECTION:kms
* @short description: KMS abstraction
* @title: KMS abstraction
*
* The KMS abstraction consists of various building blocks for helping out with
* interacting with the various drm API's, enabling users to use a
* transactional API, aiming to hide all interaction with the underlying APIs.
*
* The subsystem defines two separate contexts, the "main" context, and the
* "impl" context. The main context is the context of which mutter as a whole
* runs in. It uses the main GLib main loop and main context and always runs in
* the main thread.
*
* The impl context is where all underlying API is being executed. While in the
* current state, it always runs in the main thread, the aim is to be able to
* execute the impl context in a dedicated thread.
*
* The public facing MetaKms API is always assumed to be executed from the main
* context.
*
* The KMS abstraction consists of the following public components:
*
* #MetaKms:
*
* Main entry point; used by the native backend to create devices, post updates
* etc.
*
* #MetaKmsDevice:
*
* A device (usually /dev/dri/cardN, where N being a number). Used to get KMS
* objects, such as connectors, CRTCs, planes, as well as basic meta data such
* as device path etc.
*
* #MetaKmsCrtc:
*
* Represents a CRTC. It manages a representation of the current CRTC state,
* including current mode, coordinates, possible clones.
*
* #MetaKmsConnector:
*
* Represents a connector, e.g. a display port connection. It also manages a
* representation of the current state, including meta data such as physical
* dimension of the connected, available modes, EDID, tile info etc. It also
* contains helper functions for configuration, as well as methods for adding
* configuration to a transaction (See #MetaKmsUpdate).
*
* #MetaKmsPlane:
*
* Represents a hardware plane. A plane is used to define the content of what
* should be presented on a CRTC. Planes can either be primary planes, used as
* a backdrop for CRTCs, overlay planes, and cursor planes.
*
* #MetaKmsMode:
*
* Represents a mode a CRTC and connector can be configured with.
* Represents both modes directly derived from the devices, as well as
* fall back modes when the CRTC supports scaling.
*
* #MetaKmsUpdate:
*
* A KMS transaction object, meant to be processed potentially atomically when
* posted. An update consists of plane assignments, mode sets and KMS object
* property entries. The user adds updates to the object, and then posts it via
* MetaKms. It will then be processed by the MetaKms backend (See
* #MetaKmsImpl), potentially atomically. Each #MetaKmsUpdate deals with
* updating a single device.
*
*
* There are also these private objects, without public facing API:
*
* #MetaKmsImpl:
*
* The KMS impl context object, managing things in the impl context.
*
* #MetaKmsImplDevice:
*
* An object linked to a #MetaKmsDevice, but where it is executed in the impl
* context. It takes care of the updating of the various KMS object (CRTC,
* connector, ..) states.
*
* This is an abstract type, with currently #MetaKmsImplDeviceSimple,
* implementing mode setting and page flipping using legacy DRM API.
*
* #MetaKmsPageFlip:
*
* A object representing a page flip. It's created when a page flip is queued,
* and contains information necessary to provide feedback to the one requesting
* the page flip.
*
*/
enum
{
RESOURCES_CHANGED,
N_SIGNALS
};
static int signals[N_SIGNALS];
typedef struct _MetaKmsCallbackData
{
MetaKmsCallback callback;
gpointer user_data;
GDestroyNotify user_data_destroy;
} MetaKmsCallbackData;
typedef struct _MetaKmsSimpleImplSource
{
GSource source;
MetaKms *kms;
} MetaKmsSimpleImplSource;
typedef struct _MetaKmsFdImplSource
{
GSource source;
gpointer fd_tag;
MetaKms *kms;
MetaKmsImplTaskFunc dispatch;
gpointer user_data;
} MetaKmsFdImplSource;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
struct _MetaKms
{
GObject parent;
MetaKmsFlags flags;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
MetaBackend *backend;
gulong hotplug_handler_id;
gulong removed_handler_id;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
MetaKmsImpl *impl;
gboolean in_impl_task;
gboolean waiting_for_impl_task;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
GList *devices;
GList *pending_updates;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
GList *pending_callbacks;
guint callback_source_id;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
};
G_DEFINE_TYPE (MetaKms, meta_kms, G_TYPE_OBJECT)
static void
meta_kms_add_pending_update (MetaKms *kms,
MetaKmsUpdate *update)
{
kms->pending_updates = g_list_prepend (kms->pending_updates, update);
}
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
MetaKmsUpdate *
meta_kms_ensure_pending_update (MetaKms *kms,
MetaKmsDevice *device)
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
{
MetaKmsUpdate *update;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
update = meta_kms_get_pending_update (kms, device);
if (update)
return update;
update = meta_kms_update_new (device);
meta_kms_add_pending_update (kms, update);
return update;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
}
MetaKmsUpdate *
meta_kms_get_pending_update (MetaKms *kms,
MetaKmsDevice *device)
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
{
GList *l;
for (l = kms->pending_updates; l; l = l->next)
{
MetaKmsUpdate *update = l->data;
if (meta_kms_update_get_device (update) == device)
return update;
}
return NULL;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
}
static MetaKmsUpdate *
meta_kms_take_pending_update (MetaKms *kms,
MetaKmsDevice *device)
{
GList *l;
for (l = kms->pending_updates; l; l = l->next)
{
MetaKmsUpdate *update = l->data;
if (meta_kms_update_get_device (update) == device)
{
kms->pending_updates = g_list_delete_link (kms->pending_updates, l);
return update;
}
}
return NULL;
}
typedef struct
{
MetaKmsUpdate *update;
MetaKmsUpdateFlag flags;
} PostUpdateData;
static gpointer
meta_kms_process_update_in_impl (MetaKmsImpl *impl,
gpointer user_data,
GError **error)
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
{
PostUpdateData *data = user_data;
MetaKmsUpdate *update = data->update;
MetaKmsFeedback *feedback;
feedback = meta_kms_impl_process_update (impl, data->update, data->flags);
meta_kms_device_predict_states_in_impl (meta_kms_update_get_device (update),
update);
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
return feedback;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
}
MetaKmsFeedback *
meta_kms_post_pending_update_sync (MetaKms *kms,
MetaKmsDevice *device,
MetaKmsUpdateFlag flags)
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
{
MetaKmsUpdate *update;
PostUpdateData data;
MetaKmsFeedback *feedback;
GList *result_listeners;
GList *l;
COGL_TRACE_BEGIN_SCOPED (MetaKmsPostUpdateSync,
"KMS (post update)");
update = meta_kms_take_pending_update (kms, device);
if (!update)
return NULL;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
meta_kms_update_lock (update);
data = (PostUpdateData) {
.update = update,
.flags = flags,
};
feedback = meta_kms_run_impl_task_sync (kms,
meta_kms_process_update_in_impl,
&data,
NULL);
result_listeners = meta_kms_update_take_result_listeners (update);
if (feedback->error &&
flags & META_KMS_UPDATE_FLAG_PRESERVE_ON_ERROR)
{
GList *l;
meta_kms_update_unlock (update);
for (l = feedback->failed_planes; l; l = l->next)
{
MetaKmsPlane *plane = l->data;
meta_kms_update_drop_plane_assignment (update, plane);
}
meta_kms_update_drop_defunct_page_flip_listeners (update);
meta_kms_add_pending_update (kms, update);
}
else
{
meta_kms_update_free (update);
}
for (l = result_listeners; l; l = l->next)
{
MetaKmsResultListener *listener = l->data;
meta_kms_result_listener_notify (listener, feedback);
meta_kms_result_listener_free (listener);
}
g_list_free (result_listeners);
return feedback;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
}
static gpointer
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
meta_kms_discard_pending_page_flips_in_impl (MetaKmsImpl *impl,
gpointer user_data,
GError **error)
{
meta_kms_impl_discard_pending_page_flips (impl);
return GINT_TO_POINTER (TRUE);
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
}
void
meta_kms_discard_pending_page_flips (MetaKms *kms)
{
meta_kms_run_impl_task_sync (kms,
meta_kms_discard_pending_page_flips_in_impl,
NULL,
NULL);
}
static gpointer
meta_kms_notify_modes_set_in_impl (MetaKmsImpl *impl,
gpointer user_data,
GError **error)
{
meta_kms_impl_notify_modes_set (impl);
return GINT_TO_POINTER (TRUE);
}
void
meta_kms_notify_modes_set (MetaKms *kms)
{
meta_kms_run_impl_task_sync (kms,
meta_kms_notify_modes_set_in_impl,
NULL,
NULL);
}
static void
meta_kms_callback_data_free (MetaKmsCallbackData *callback_data)
{
if (callback_data->user_data_destroy)
callback_data->user_data_destroy (callback_data->user_data);
g_free (callback_data);
}
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
static int
flush_callbacks (MetaKms *kms)
{
GList *l;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
int callback_count = 0;
meta_assert_not_in_kms_impl (kms);
g_clear_handle_id (&kms->callback_source_id, g_source_remove);
for (l = kms->pending_callbacks; l; l = l->next)
{
MetaKmsCallbackData *callback_data = l->data;
callback_data->callback (kms, callback_data->user_data);
meta_kms_callback_data_free (callback_data);
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
callback_count++;
}
g_list_free (kms->pending_callbacks);
kms->pending_callbacks = NULL;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
return callback_count;
}
static gboolean
callback_idle (gpointer user_data)
{
MetaKms *kms = user_data;
flush_callbacks (kms);
kms->callback_source_id = 0;
return G_SOURCE_REMOVE;
}
void
meta_kms_queue_callback (MetaKms *kms,
MetaKmsCallback callback,
gpointer user_data,
GDestroyNotify user_data_destroy)
{
MetaKmsCallbackData *callback_data;
callback_data = g_new0 (MetaKmsCallbackData, 1);
*callback_data = (MetaKmsCallbackData) {
.callback = callback,
.user_data = user_data,
.user_data_destroy = user_data_destroy,
};
kms->pending_callbacks = g_list_append (kms->pending_callbacks,
callback_data);
if (!kms->callback_source_id)
kms->callback_source_id = g_idle_add (callback_idle, kms);
}
gpointer
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
meta_kms_run_impl_task_sync (MetaKms *kms,
MetaKmsImplTaskFunc func,
gpointer user_data,
GError **error)
{
gpointer ret;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
kms->in_impl_task = TRUE;
kms->waiting_for_impl_task = TRUE;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
ret = func (kms->impl, user_data, error);
kms->waiting_for_impl_task = FALSE;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
kms->in_impl_task = FALSE;
return ret;
}
static gboolean
simple_impl_source_dispatch (GSource *source,
GSourceFunc callback,
gpointer user_data)
{
MetaKmsSimpleImplSource *simple_impl_source =
(MetaKmsSimpleImplSource *) source;
MetaKms *kms = simple_impl_source->kms;
gboolean ret;
kms->in_impl_task = TRUE;
ret = callback (user_data);
kms->in_impl_task = FALSE;
return ret;
}
static GSourceFuncs simple_impl_source_funcs = {
.dispatch = simple_impl_source_dispatch,
};
GSource *
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
meta_kms_add_source_in_impl (MetaKms *kms,
GSourceFunc func,
gpointer user_data,
GDestroyNotify user_data_destroy)
{
GSource *source;
MetaKmsSimpleImplSource *simple_impl_source;
meta_assert_in_kms_impl (kms);
source = g_source_new (&simple_impl_source_funcs,
sizeof (MetaKmsSimpleImplSource));
simple_impl_source = (MetaKmsSimpleImplSource *) source;
simple_impl_source->kms = kms;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
g_source_set_callback (source, func, user_data, user_data_destroy);
kms: Make GSource ready by default When testing a laptop with intel and DisplayLink devices, attempting to set the DL output as the only active output resulted in GNOME/Wayland freezing. The main event loop was running fine, but nothing on screen would get updated once the DL output become the only one. This patch fixes that issue. DisplayLink USB 3 devices use an out-of-tree kernel DRM driver called EVDI. EVDI can sometimes fail drmModePageFlip(). For me, the flip fails reliably when hotplugging the DL dock and when changing display configuration to DL only. Mutter has a workaround for failing flips, it just calls drmModeSetCrtc() and that succeeds. What does not work reliably in the fallback path is Mutter keeping track of the pageflip. Since drmModePageFlip() failed, there will not be a pageflip event coming and instead Mutter queues a callback in its stead. When you have more than one output, some other output repainting will attempt to swap buffers and calls wait_for_pending_flips() which has the side-effect of dispatching any queued flip callbacks. With multiple outputs, you don't get stuck (unless they all fail the exact same way at the same time?). When you have only one output, it cannot proceed to repaint and buffer swap because the pageflip is not marked complete yet. Nothing dispatches the flip callback, leading to the freeze. The flip callback is intended to be an idle callback, implemented with a GSource. It is supposed to be called as soon as execution returns to the main event loop. The setup of the GSource is incomplete, so it will never dispatch. Fix the GSource setup by setting its ready-time to be always in the past. That gets it dispatched on the next cycle of the main event loop. This is now the default behavior for all sources created by meta_kms_add_source_in_impl(). Sources that need a delay continue to do that by overriding the ready-time explicitly. An alternative solution could have been to implement GSource prepare and check callbacks returning TRUE. However, since meta_kms_add_source_in_impl() is used by flip retry code as well, and that code needs a delay through the ready-time, I was afraid I might break the flip retry code. Hence I decided to use ready-time instead. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1209
2020-04-17 11:50:13 +00:00
g_source_set_ready_time (source, 0);
g_source_attach (source, g_main_context_get_thread_default ());
return source;
}
static gboolean
meta_kms_fd_impl_source_check (GSource *source)
{
MetaKmsFdImplSource *fd_impl_source = (MetaKmsFdImplSource *) source;
return g_source_query_unix_fd (source, fd_impl_source->fd_tag) & G_IO_IN;
}
static gboolean
meta_kms_fd_impl_source_dispatch (GSource *source,
GSourceFunc callback,
gpointer user_data)
{
MetaKmsFdImplSource *fd_impl_source = (MetaKmsFdImplSource *) source;
MetaKms *kms = fd_impl_source->kms;
gpointer ret;
GError *error = NULL;
kms->in_impl_task = TRUE;
ret = fd_impl_source->dispatch (kms->impl,
fd_impl_source->user_data,
&error);
kms->in_impl_task = FALSE;
if (!GPOINTER_TO_INT (ret))
{
g_warning ("Failed to dispatch fd source: %s", error->message);
g_error_free (error);
}
return G_SOURCE_CONTINUE;
}
static GSourceFuncs fd_impl_source_funcs = {
NULL,
meta_kms_fd_impl_source_check,
meta_kms_fd_impl_source_dispatch
};
GSource *
meta_kms_register_fd_in_impl (MetaKms *kms,
int fd,
MetaKmsImplTaskFunc dispatch,
gpointer user_data)
{
GSource *source;
MetaKmsFdImplSource *fd_impl_source;
meta_assert_in_kms_impl (kms);
source = g_source_new (&fd_impl_source_funcs, sizeof (MetaKmsFdImplSource));
fd_impl_source = (MetaKmsFdImplSource *) source;
fd_impl_source->dispatch = dispatch;
fd_impl_source->user_data = user_data;
fd_impl_source->kms = kms;
fd_impl_source->fd_tag = g_source_add_unix_fd (source, fd,
G_IO_IN | G_IO_ERR);
g_source_attach (source, g_main_context_get_thread_default ());
return source;
}
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
gboolean
meta_kms_in_impl_task (MetaKms *kms)
{
return kms->in_impl_task;
}
gboolean
meta_kms_is_waiting_for_impl_task (MetaKms *kms)
{
return kms->waiting_for_impl_task;
}
typedef struct _UpdateStatesData
{
const char *device_path;
uint32_t crtc_id;
uint32_t connector_id;
} UpdateStatesData;
static MetaKmsUpdateChanges
meta_kms_update_states_in_impl (MetaKms *kms,
UpdateStatesData *update_data)
{
MetaKmsUpdateChanges changes = META_KMS_UPDATE_CHANGE_NONE;
GList *l;
COGL_TRACE_BEGIN_SCOPED (MetaKmsUpdateStates,
"KMS (update states)");
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
meta_assert_in_kms_impl (kms);
if (!kms->devices)
return META_KMS_UPDATE_CHANGE_NO_DEVICES;
for (l = kms->devices; l; l = l->next)
{
MetaKmsDevice *kms_device = META_KMS_DEVICE (l->data);
const char *kms_device_path = meta_kms_device_get_path (kms_device);
if (update_data->device_path &&
g_strcmp0 (kms_device_path, update_data->device_path) != 0)
continue;
if (update_data->crtc_id > 0 &&
!meta_kms_device_find_crtc_in_impl (kms_device, update_data->crtc_id))
continue;
if (update_data->connector_id > 0 &&
!meta_kms_device_find_connector_in_impl (kms_device,
update_data->connector_id))
continue;
changes |=
meta_kms_device_update_states_in_impl (kms_device,
update_data->crtc_id,
update_data->connector_id);
}
return changes;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
}
static gpointer
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
update_states_in_impl (MetaKmsImpl *impl,
gpointer user_data,
GError **error)
{
UpdateStatesData *data = user_data;
MetaKms *kms = meta_kms_impl_get_kms (impl);
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 20:36:41 +00:00
return GUINT_TO_POINTER (meta_kms_update_states_in_impl (kms, data));
}
static MetaKmsUpdateChanges
meta_kms_update_states_sync (MetaKms *kms,
GUdevDevice *udev_device)
{
UpdateStatesData data = {};
gpointer ret;
if (udev_device)
{
data.device_path = g_udev_device_get_device_file (udev_device);
data.crtc_id =
CLAMP (g_udev_device_get_property_as_int (udev_device, "CRTC"),
0, UINT32_MAX);
data.connector_id =
CLAMP (g_udev_device_get_property_as_int (udev_device, "CONNECTOR"),
0, UINT32_MAX);
}
ret = meta_kms_run_impl_task_sync (kms, update_states_in_impl, &data, NULL);
return GPOINTER_TO_UINT (ret);
}
static void
handle_hotplug_event (MetaKms *kms,
GUdevDevice *udev_device,
MetaKmsUpdateChanges changes)
{
changes |= meta_kms_update_states_sync (kms, udev_device);
if (changes != META_KMS_UPDATE_CHANGE_NONE)
g_signal_emit (kms, signals[RESOURCES_CHANGED], 0, changes);
}
void
meta_kms_resume (MetaKms *kms)
{
handle_hotplug_event (kms, NULL, META_KMS_UPDATE_CHANGE_FULL);
}
static void
on_udev_hotplug (MetaUdev *udev,
GUdevDevice *udev_device,
MetaKms *kms)
{
handle_hotplug_event (kms, udev_device, META_KMS_UPDATE_CHANGE_NONE);
}
static void
on_udev_device_removed (MetaUdev *udev,
GUdevDevice *device,
MetaKms *kms)
{
handle_hotplug_event (kms, NULL, META_KMS_UPDATE_CHANGE_NONE);
}
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
MetaBackend *
meta_kms_get_backend (MetaKms *kms)
{
return kms->backend;
}
GList *
meta_kms_get_devices (MetaKms *kms)
{
return kms->devices;
}
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
MetaKmsDevice *
meta_kms_create_device (MetaKms *kms,
const char *path,
MetaKmsDeviceFlag flags,
GError **error)
{
MetaKmsDevice *device;
if (kms->flags & META_KMS_FLAG_NO_MODE_SETTING)
flags |= META_KMS_DEVICE_FLAG_NO_MODE_SETTING;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
device = meta_kms_device_new (kms, path, flags, error);
if (!device)
return NULL;
kms->devices = g_list_append (kms->devices, device);
return device;
}
MetaKms *
meta_kms_new (MetaBackend *backend,
MetaKmsFlags flags,
GError **error)
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
{
MetaBackendNative *backend_native = META_BACKEND_NATIVE (backend);
MetaUdev *udev = meta_backend_native_get_udev (backend_native);
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
MetaKms *kms;
kms = g_object_new (META_TYPE_KMS, NULL);
kms->flags = flags;
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
kms->backend = backend;
kms->impl = meta_kms_impl_new (kms);
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
if (!kms->impl)
{
g_object_unref (kms);
return NULL;
}
if (!(flags & META_KMS_FLAG_NO_MODE_SETTING))
{
kms->hotplug_handler_id =
g_signal_connect (udev, "hotplug", G_CALLBACK (on_udev_hotplug), kms);
}
kms->removed_handler_id =
g_signal_connect (udev, "device-removed",
G_CALLBACK (on_udev_device_removed), kms);
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
return kms;
}
static gpointer
prepare_shutdown_in_impl (MetaKmsImpl *impl,
gpointer user_data,
GError **error)
{
meta_kms_impl_prepare_shutdown (impl);
return GINT_TO_POINTER (TRUE);
}
void
meta_kms_prepare_shutdown (MetaKms *kms)
{
meta_kms_run_impl_task_sync (kms, prepare_shutdown_in_impl, NULL, NULL);
flush_callbacks (kms);
}
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
static void
meta_kms_finalize (GObject *object)
{
MetaKms *kms = META_KMS (object);
MetaBackendNative *backend_native = META_BACKEND_NATIVE (kms->backend);
MetaUdev *udev = meta_backend_native_get_udev (backend_native);
GList *l;
for (l = kms->pending_callbacks; l; l = l->next)
meta_kms_callback_data_free (l->data);
g_list_free (kms->pending_callbacks);
g_clear_handle_id (&kms->callback_source_id, g_source_remove);
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
g_list_free_full (kms->devices, g_object_unref);
g_clear_signal_handler (&kms->hotplug_handler_id, udev);
g_clear_signal_handler (&kms->removed_handler_id, udev);
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
G_OBJECT_CLASS (meta_kms_parent_class)->finalize (object);
}
static void
meta_kms_init (MetaKms *kms)
{
}
static void
meta_kms_class_init (MetaKmsClass *klass)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
object_class->finalize = meta_kms_finalize;
signals[RESOURCES_CHANGED] =
g_signal_new ("resources-changed",
G_TYPE_FROM_CLASS (klass),
G_SIGNAL_RUN_LAST,
0,
NULL, NULL, NULL,
G_TYPE_NONE, 1,
META_TYPE_KMS_UPDATE_CHANGES);
backends/native: Add basic KMS abstraction building blocks The intention with KMS abstraction is to hide away accessing the drm functions behind an API that allows us to have different kind of KMS implementations, including legacy non-atomic and atomic. The intention is also that the code interacting with the drm device should be able to be run in a different thread than the main thread. This means that we need to make sure that all drm*() API usage must only occur from within tasks that eventually can be run in the dedicated thread. The idea here is that MetaKms provides a outward facing API other places of mutter can use (e.g. MetaGpuKms and friends), while MetaKmsImpl is an internal implementation that only gets interacted with via "tasks" posted via the MetaKms object. These tasks will in the future potentially be run on the dedicated KMS thread. Initially, we don't create any new threads. Likewise, MetaKmsDevice is a outward facing representation of a KMS device, while MetaKmsImplDevice is the corresponding implementation, which only runs from within the MetaKmsImpl tasks. This commit only moves opening and closing the device to this new API, while leaking the fd outside of the impl enclosure, effectively making the isolation for drm*() calls pointless. This, however, is necessary to allow gradual porting of drm interaction, and eventually the file descriptor in MetaGpuKms will be removed. For now, it's harmless, since everything still run in the main thread. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-01-29 09:24:44 +00:00
}