sungmg/mutter-performance-source
1
0
Fork 0
Code Activity
mutter-performance-source/src/backends/native/meta-kms-impl-simple.c

1128 lines
36 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) 2018-2019 Red Hat
kms-impl/simple: Fix page_flip_data ref leak on fallback If drmModePageFlip() or custom_page_flip_func fails, process_page_flip() was forgetting to undo the ref taken for that call. This would leak page_flip_data. The reference counting works like this: - when created, ref count is 1 - when calling drmModePageFlip, ref count is increased to 2 - new: if flip failed, ref count is decreased back to 1 - if calling schedule_retry_page_flip(), it takes a ref internally - if calling mode_set_fallback(), it takes a ref internally - all return FALSE paths have an explicit unref - return TRUE path has an explicit unref This issue was found by code inspection and while debugging an unrelated issue with debug prints sprinkled around. I am not aware of any end-user visible issues being fixed by this, as the leak is small and probably very rare. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1209
2020-04-17 11:50:13 +00:00
* Copyright (C) 2019-2020 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-impl-simple.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 <errno.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 <gbm.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 <xf86drmMode.h>
#include "backends/native/meta-kms-connector.h"
#include "backends/native/meta-kms-crtc.h"
#include "backends/native/meta-kms-device-private.h"
#include "backends/native/meta-kms-page-flip-private.h"
#include "backends/native/meta-kms-plane.h"
#include "backends/native/meta-kms-private.h"
#include "backends/native/meta-kms-update-private.h"
#include "backends/native/meta-kms-utils.h"
typedef struct _CachedModeSet
{
GList *connectors;
drmModeModeInfo *drm_mode;
} CachedModeSet;
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 _MetaKmsImplSimple
{
MetaKmsImpl parent;
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
GSource *mode_set_fallback_feedback_source;
GList *mode_set_fallback_page_flip_datas;
GList *pending_page_flip_retries;
GSource *retry_page_flips_source;
GList *postponed_page_flip_datas;
GList *postponed_mode_set_fallback_datas;
GHashTable *cached_mode_sets;
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 (MetaKmsImplSimple, meta_kms_impl_simple,
META_TYPE_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
static void
flush_postponed_page_flip_datas (MetaKmsImplSimple *impl_simple);
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
MetaKmsImplSimple *
meta_kms_impl_simple_new (MetaKms *kms,
GError **error)
{
return g_object_new (META_TYPE_KMS_IMPL_SIMPLE,
"kms", kms,
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 gboolean
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
process_connector_property (MetaKmsImpl *impl,
MetaKmsUpdate *update,
gpointer update_entry,
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
{
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
MetaKmsConnectorProperty *connector_property = update_entry;
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
MetaKmsConnector *connector = connector_property->connector;
MetaKmsDevice *device = meta_kms_connector_get_device (connector);
MetaKmsImplDevice *impl_device = meta_kms_device_get_impl_device (device);
int fd;
int ret;
fd = meta_kms_impl_device_get_fd (impl_device);
ret = drmModeObjectSetProperty (fd,
meta_kms_connector_get_id (connector),
DRM_MODE_OBJECT_CONNECTOR,
connector_property->prop_id,
connector_property->value);
if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
"Failed to set connector %u property %u: %s",
meta_kms_connector_get_id (connector),
connector_property->prop_id,
g_strerror (-ret));
return FALSE;
}
return TRUE;
}
static gboolean
process_plane_property (MetaKmsImpl *impl,
MetaKmsPlane *plane,
MetaKmsProperty *prop,
GError **error)
{
MetaKmsDevice *device = meta_kms_plane_get_device (plane);
MetaKmsImplDevice *impl_device = meta_kms_device_get_impl_device (device);
int fd;
int ret;
fd = meta_kms_impl_device_get_fd (impl_device);
ret = drmModeObjectSetProperty (fd,
meta_kms_plane_get_id (plane),
DRM_MODE_OBJECT_PLANE,
prop->prop_id,
prop->value);
if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
"Failed to set plane %u property %u: %s",
meta_kms_plane_get_id (plane),
prop->prop_id,
g_strerror (-ret));
return FALSE;
}
return TRUE;
}
static CachedModeSet *
cached_mode_set_new (GList *connectors,
const drmModeModeInfo *drm_mode)
{
CachedModeSet *cached_mode_set;
cached_mode_set = g_new0 (CachedModeSet, 1);
*cached_mode_set = (CachedModeSet) {
.connectors = g_list_copy (connectors),
.drm_mode = g_memdup (drm_mode, sizeof *drm_mode),
};
return cached_mode_set;
}
static void
cached_mode_set_free (CachedModeSet *cached_mode_set)
{
g_list_free (cached_mode_set->connectors);
g_free (cached_mode_set->drm_mode);
g_free (cached_mode_set);
}
static void
fill_connector_ids_array (GList *connectors,
uint32_t **out_connectors,
int *out_n_connectors)
{
GList *l;
int i;
*out_n_connectors = g_list_length (connectors);
*out_connectors = g_new0 (uint32_t, *out_n_connectors);
i = 0;
for (l = connectors; l; l = l->next)
{
MetaKmsConnector *connector = l->data;
(*out_connectors)[i++] = meta_kms_connector_get_id (connector);
}
}
static gboolean
process_mode_set (MetaKmsImpl *impl,
MetaKmsUpdate *update,
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
gpointer update_entry,
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
GError **error)
{
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
MetaKmsModeSet *mode_set = update_entry;
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
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (impl);
kms-impl/simple: Removing extra semi colon https://gitlab.gnome.org/GNOME/mutter/merge_requests/634
2019-06-24 08:35:47 +00:00
MetaKmsCrtc *crtc = mode_set->crtc;
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
MetaKmsDevice *device = meta_kms_crtc_get_device (crtc);
MetaKmsImplDevice *impl_device = meta_kms_device_get_impl_device (device);
g_autofree uint32_t *connectors = NULL;
int n_connectors;
MetaKmsPlaneAssignment *plane_assignment;
uint32_t x, y;
uint32_t fb_id;
int fd;
int ret;
crtc = mode_set->crtc;
if (mode_set->drm_mode)
{
GList *l;
fill_connector_ids_array (mode_set->connectors,
&connectors,
&n_connectors);
kms/update: Add helper to get primary plane assignment Replaces a private static function, with no functional changes made to it. https://gitlab.gnome.org/GNOME/mutter/merge_requests/826
2019-10-04 09:48:46 +00:00
plane_assignment = meta_kms_update_get_primary_plane_assignment (update,
crtc);
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
if (!plane_assignment)
{
g_set_error (error, G_IO_ERROR, G_IO_ERROR_FAILED,
"Missing primary plane assignment for legacy mode set on CRTC %u",
meta_kms_crtc_get_id (crtc));
return FALSE;
}
x = meta_fixed_16_to_int (plane_assignment->src_rect.x);
y = meta_fixed_16_to_int (plane_assignment->src_rect.y);
for (l = plane_assignment->plane_properties; l; l = l->next)
{
MetaKmsProperty *prop = l->data;
if (!process_plane_property (impl, plane_assignment->plane,
prop, error))
return FALSE;
}
fb_id = plane_assignment->fb_id;
}
else
{
x = y = 0;
n_connectors = 0;
connectors = NULL;
fb_id = 0;
}
fd = meta_kms_impl_device_get_fd (impl_device);
ret = drmModeSetCrtc (fd,
meta_kms_crtc_get_id (crtc),
fb_id,
x, y,
connectors, n_connectors,
mode_set->drm_mode);
if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
kms-impl-simple: Include mode name in error message https://gitlab.gnome.org/GNOME/mutter/merge_requests/1007
2020-01-14 10:23:04 +00:00
"Failed to set mode %s on CRTC %u: %s",
mode_set->drm_mode ? mode_set->drm_mode->name : "off",
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_crtc_get_id (crtc),
g_strerror (-ret));
return FALSE;
}
if (mode_set->drm_mode)
{
g_hash_table_replace (impl_simple->cached_mode_sets,
crtc,
cached_mode_set_new (mode_set->connectors,
mode_set->drm_mode));
}
else
{
g_hash_table_remove (impl_simple->cached_mode_sets, crtc);
}
return TRUE;
}
monitor-manager/kms: Use KMS abstraction to get and set CRTC gamma Still doesn't synchronize with frame drawing, but no point in doing that until gamma is managed by mutter itself and not gnome-settings-daemon. https://gitlab.gnome.org/GNOME/mutter/merge_requests/634
2019-05-03 17:20:21 +00:00
static gboolean
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
process_crtc_gamma (MetaKmsImpl *impl,
MetaKmsUpdate *update,
gpointer update_entry,
GError **error)
monitor-manager/kms: Use KMS abstraction to get and set CRTC gamma Still doesn't synchronize with frame drawing, but no point in doing that until gamma is managed by mutter itself and not gnome-settings-daemon. https://gitlab.gnome.org/GNOME/mutter/merge_requests/634
2019-05-03 17:20:21 +00:00
{
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
MetaKmsCrtcGamma *gamma = update_entry;
monitor-manager/kms: Use KMS abstraction to get and set CRTC gamma Still doesn't synchronize with frame drawing, but no point in doing that until gamma is managed by mutter itself and not gnome-settings-daemon. https://gitlab.gnome.org/GNOME/mutter/merge_requests/634
2019-05-03 17:20:21 +00:00
MetaKmsCrtc *crtc = gamma->crtc;
MetaKmsDevice *device = meta_kms_crtc_get_device (crtc);
MetaKmsImplDevice *impl_device = meta_kms_device_get_impl_device (device);
int fd;
int ret;
fd = meta_kms_impl_device_get_fd (impl_device);
ret = drmModeCrtcSetGamma (fd, meta_kms_crtc_get_id (crtc),
gamma->size,
gamma->red,
gamma->green,
gamma->blue);
if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
"drmModeCrtcSetGamma on CRTC %u failed: %s",
meta_kms_crtc_get_id (crtc),
g_strerror (-ret));
return FALSE;
}
return 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
static gboolean
is_timestamp_earlier_than (uint64_t ts1,
uint64_t ts2)
{
if (ts1 == ts2)
return FALSE;
else
return ts2 - ts1 < UINT64_MAX / 2;
}
typedef struct _RetryPageFlipData
{
MetaKmsCrtc *crtc;
uint32_t fb_id;
MetaKmsPageFlipData *page_flip_data;
float refresh_rate;
uint64_t retry_time_us;
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
MetaKmsCustomPageFlipFunc custom_page_flip_func;
gpointer custom_page_flip_user_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
} RetryPageFlipData;
static void
retry_page_flip_data_free (RetryPageFlipData *retry_page_flip_data)
{
g_assert (!retry_page_flip_data->page_flip_data);
g_free (retry_page_flip_data);
}
kms-impl-simple: Handle mode set race conditions gracefully If we end up trying to do a mode set on a DRM state that has already changed behind our back without us yet having seen the hotplug event we may fail with `EINVAL`. Since the renderer layer doesn't handle mode set failure, it'll still try to page flip later on, which will then also fail. When failing, it'll try to look up the cached mode set in order to retry the mode set later on, as is needed to handle other error conditions. However, if the mode set prior to the page flip failed, we won't cache the mode set, and the page flip error handling code will get confused. Instead of asserting that a page flip always has a valid cached mode set ready to look up, handle it being missing more gracefully by failing to mode set. It is expected that things will correct themself as there should be a hotplug event waiting around the the corner, to reconfigure the monitor configuration setting new modes. Fixes: https://gitlab.gnome.org/GNOME/mutter/issues/917 https://gitlab.gnome.org/GNOME/mutter/merge_requests/1007
2020-01-14 10:16:09 +00:00
static CachedModeSet *
get_cached_mode_set (MetaKmsImplSimple *impl_simple,
MetaKmsCrtc *crtc)
{
return g_hash_table_lookup (impl_simple->cached_mode_sets, crtc);
}
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 float
get_cached_crtc_refresh_rate (MetaKmsImplSimple *impl_simple,
MetaKmsCrtc *crtc)
{
CachedModeSet *cached_mode_set;
cached_mode_set = g_hash_table_lookup (impl_simple->cached_mode_sets,
crtc);
g_assert (cached_mode_set);
return meta_calculate_drm_mode_refresh_rate (cached_mode_set->drm_mode);
}
static gboolean
retry_page_flips (gpointer user_data)
{
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (user_data);
uint64_t now_us;
GList *l;
meta_assert_in_kms_impl (meta_kms_impl_get_kms (META_KMS_IMPL (impl_simple)));
now_us = g_source_get_time (impl_simple->retry_page_flips_source);
l = impl_simple->pending_page_flip_retries;
while (l)
{
RetryPageFlipData *retry_page_flip_data = l->data;
MetaKmsCrtc *crtc = retry_page_flip_data->crtc;
MetaKmsDevice *device = meta_kms_crtc_get_device (crtc);
MetaKmsImplDevice *impl_device = meta_kms_device_get_impl_device (device);
GList *l_next = l->next;
int fd;
int ret;
MetaKmsPageFlipData *page_flip_data;
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
MetaKmsCustomPageFlipFunc custom_page_flip_func;
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
if (is_timestamp_earlier_than (now_us,
retry_page_flip_data->retry_time_us))
{
l = l_next;
continue;
}
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
custom_page_flip_func = retry_page_flip_data->custom_page_flip_func;
if (custom_page_flip_func)
{
ret = custom_page_flip_func (retry_page_flip_data->custom_page_flip_user_data,
retry_page_flip_data->page_flip_data);
}
else
{
fd = meta_kms_impl_device_get_fd (impl_device);
ret = drmModePageFlip (fd,
meta_kms_crtc_get_id (crtc),
retry_page_flip_data->fb_id,
DRM_MODE_PAGE_FLIP_EVENT,
retry_page_flip_data->page_flip_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
if (ret == -EBUSY)
{
float refresh_rate;
refresh_rate = get_cached_crtc_refresh_rate (impl_simple, crtc);
retry_page_flip_data->retry_time_us +=
(uint64_t) (G_USEC_PER_SEC / refresh_rate);
l = l_next;
continue;
}
impl_simple->pending_page_flip_retries =
g_list_remove_link (impl_simple->pending_page_flip_retries, l);
page_flip_data = g_steal_pointer (&retry_page_flip_data->page_flip_data);
if (ret != 0)
{
g_autoptr (GError) error = NULL;
g_set_error (&error, G_IO_ERROR, g_io_error_from_errno (-ret),
"drmModePageFlip on CRTC %u failed: %s",
meta_kms_crtc_get_id (crtc),
g_strerror (-ret));
if (!g_error_matches (error,
G_IO_ERROR,
G_IO_ERROR_PERMISSION_DENIED))
g_critical ("Failed to page flip: %s", error->message);
meta_kms_page_flip_data_discard_in_impl (page_flip_data, error);
}
retry_page_flip_data_free (retry_page_flip_data);
l = l_next;
}
if (impl_simple->pending_page_flip_retries)
{
GList *l;
uint64_t earliest_retry_time_us = 0;
for (l = impl_simple->pending_page_flip_retries; l; l = l->next)
{
RetryPageFlipData *retry_page_flip_data = l->data;
if (l == impl_simple->pending_page_flip_retries ||
is_timestamp_earlier_than (retry_page_flip_data->retry_time_us,
earliest_retry_time_us))
earliest_retry_time_us = retry_page_flip_data->retry_time_us;
}
g_source_set_ready_time (impl_simple->retry_page_flips_source,
earliest_retry_time_us);
return G_SOURCE_CONTINUE;
}
else
{
g_clear_pointer (&impl_simple->retry_page_flips_source,
g_source_unref);
flush_postponed_page_flip_datas (impl_simple);
return G_SOURCE_REMOVE;
}
}
static void
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
schedule_retry_page_flip (MetaKmsImplSimple *impl_simple,
MetaKmsCrtc *crtc,
uint32_t fb_id,
float refresh_rate,
MetaKmsPageFlipData *page_flip_data,
MetaKmsCustomPageFlipFunc custom_page_flip_func,
gpointer custom_page_flip_user_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
{
RetryPageFlipData *retry_page_flip_data;
uint64_t now_us;
uint64_t retry_time_us;
now_us = g_get_monotonic_time ();
retry_time_us = now_us + (uint64_t) (G_USEC_PER_SEC / refresh_rate);
retry_page_flip_data = g_new0 (RetryPageFlipData, 1);
*retry_page_flip_data = (RetryPageFlipData) {
.crtc = crtc,
.fb_id = fb_id,
.page_flip_data = meta_kms_page_flip_data_ref (page_flip_data),
.refresh_rate = refresh_rate,
.retry_time_us = retry_time_us,
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
.custom_page_flip_func = custom_page_flip_func,
.custom_page_flip_user_data = custom_page_flip_user_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
};
if (!impl_simple->retry_page_flips_source)
{
MetaKms *kms = meta_kms_impl_get_kms (META_KMS_IMPL (impl_simple));
GSource *source;
source = meta_kms_add_source_in_impl (kms, retry_page_flips,
impl_simple, NULL);
g_source_set_ready_time (source, retry_time_us);
impl_simple->retry_page_flips_source = source;
}
else
{
GList *l;
for (l = impl_simple->pending_page_flip_retries; l; l = l->next)
{
RetryPageFlipData *pending_retry_page_flip_data = l->data;
uint64_t pending_retry_time_us =
pending_retry_page_flip_data->retry_time_us;
if (is_timestamp_earlier_than (retry_time_us, pending_retry_time_us))
{
g_source_set_ready_time (impl_simple->retry_page_flips_source,
retry_time_us);
break;
}
}
}
impl_simple->pending_page_flip_retries =
g_list_append (impl_simple->pending_page_flip_retries,
retry_page_flip_data);
}
static void
invoke_page_flip_datas (GList *page_flip_datas,
MetaPageFlipDataFeedbackFunc func)
{
g_list_foreach (page_flip_datas, (GFunc) func, NULL);
}
static void
clear_page_flip_datas (GList **page_flip_datas)
{
g_list_free_full (*page_flip_datas,
(GDestroyNotify) meta_kms_page_flip_data_unref);
*page_flip_datas = NULL;
}
static gboolean
mode_set_fallback_feedback_idle (gpointer user_data)
{
MetaKmsImplSimple *impl_simple = user_data;
g_clear_pointer (&impl_simple->mode_set_fallback_feedback_source,
g_source_unref);
kms-impl/simple: Fix meta_set_fallback_feedback_idle Presumably this function is supposed to be like meta_kms_impl_simple_handle_page_flip_callback() but the condition in the if-statement is inverted. Fix the inversion to make these two functions look alike. This is part 2 of 2 fixing a complete desktop freeze when drmModePageFlip() fails with EINVAL and the fallback to drmModeSetCrtc() succeeds but the success is not registered correctly as completed "flip". The freeze occurs under wait_for_pending_flips() which calls down into meta_kms_impl_device_dispatch() which ends up poll()'ing the DRM fd even though drmModeSetCrtc() will not produce a DRM event, hence the poll() never returns. The freeze was observed when hotplugging a DisplayLink dock for the first time on Ubuntu 19.10. This patch makes meta_set_fallback_feedback_idle() actually end up calling into notify_view_crtc_presented() which decrements secondary_gpu_state->pending_flips so that wait_for_pending_flips() can finish. CC stable: gnome-3-34 https://gitlab.gnome.org/GNOME/mutter/merge_requests/953
2019-11-29 12:15:32 +00:00
if (impl_simple->pending_page_flip_retries)
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
{
impl_simple->postponed_mode_set_fallback_datas =
g_steal_pointer (&impl_simple->mode_set_fallback_page_flip_datas);
}
else
{
invoke_page_flip_datas (impl_simple->mode_set_fallback_page_flip_datas,
meta_kms_page_flip_data_mode_set_fallback_in_impl);
clear_page_flip_datas (&impl_simple->mode_set_fallback_page_flip_datas);
}
return G_SOURCE_REMOVE;
}
static gboolean
mode_set_fallback (MetaKmsImplSimple *impl_simple,
MetaKmsUpdate *update,
MetaKmsPageFlip *page_flip,
MetaKmsPlaneAssignment *plane_assignment,
MetaKmsPageFlipData *page_flip_data,
GError **error)
{
MetaKms *kms = meta_kms_impl_get_kms (META_KMS_IMPL (impl_simple));
MetaKmsCrtc *crtc = page_flip->crtc;
MetaKmsDevice *device = meta_kms_crtc_get_device (crtc);
MetaKmsImplDevice *impl_device = meta_kms_device_get_impl_device (device);
CachedModeSet *cached_mode_set;
g_autofree uint32_t *connectors = NULL;
int n_connectors;
uint32_t x, y;
int fd;
int ret;
cached_mode_set = g_hash_table_lookup (impl_simple->cached_mode_sets,
crtc);
kms-impl-simple: Handle lack of cached mode set in flip fallback When a page flip fails with a certain error code, we've treated this as a hint that page flipping is broken and we should try to use mode setting instead. On some drivers, it seems that this error is also reported when there was no mode set, which means we'll have no cached mode set to use in the fallback. The lack of prior mode set tends to happen when we hit a race when the DRM objects change before we have the time to process a hotplug event. Handle the lack a missing mode set in the flip fallback path, with the assumption that we'll get a hotplug event that'll fix things up for us eventually. Closes: https://gitlab.gnome.org/GNOME/mutter/issues/917
2020-01-17 07:28:17 +00:00
if (!cached_mode_set)
{
g_set_error (error, G_IO_ERROR, G_IO_ERROR_FAILED,
"Missing mode set for page flip fallback");
return FALSE;
}
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
fill_connector_ids_array (cached_mode_set->connectors,
&connectors,
&n_connectors);
x = meta_fixed_16_to_int (plane_assignment->src_rect.x);
y = meta_fixed_16_to_int (plane_assignment->src_rect.y);
fd = meta_kms_impl_device_get_fd (impl_device);
ret = drmModeSetCrtc (fd,
meta_kms_crtc_get_id (crtc),
plane_assignment->fb_id,
x, y,
connectors, n_connectors,
cached_mode_set->drm_mode);
if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
"drmModeSetCrtc mode '%s' on CRTC %u failed: %s",
cached_mode_set->drm_mode->name,
meta_kms_crtc_get_id (crtc),
g_strerror (-ret));
return FALSE;
}
if (!impl_simple->mode_set_fallback_feedback_source)
{
GSource *source;
source = meta_kms_add_source_in_impl (kms,
mode_set_fallback_feedback_idle,
impl_simple,
NULL);
impl_simple->mode_set_fallback_feedback_source = source;
}
impl_simple->mode_set_fallback_page_flip_datas =
g_list_prepend (impl_simple->mode_set_fallback_page_flip_datas,
meta_kms_page_flip_data_ref (page_flip_data));
return TRUE;
}
static gboolean
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
process_page_flip (MetaKmsImpl *impl,
MetaKmsUpdate *update,
gpointer update_entry,
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
{
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
MetaKmsPageFlip *page_flip = update_entry;
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
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (impl);
MetaKmsCrtc *crtc;
MetaKmsDevice *device;
MetaKmsImplDevice *impl_device;
MetaKmsPlaneAssignment *plane_assignment;
MetaKmsPageFlipData *page_flip_data;
MetaKmsCustomPageFlipFunc custom_page_flip_func;
int fd;
int ret;
crtc = page_flip->crtc;
kms/update: Add helper to get primary plane assignment Replaces a private static function, with no functional changes made to it. https://gitlab.gnome.org/GNOME/mutter/merge_requests/826
2019-10-04 09:48:46 +00:00
plane_assignment = meta_kms_update_get_primary_plane_assignment (update,
crtc);
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
page_flip_data = meta_kms_page_flip_data_new (impl,
crtc,
page_flip->feedback,
page_flip->user_data);
device = meta_kms_crtc_get_device (crtc);
impl_device = meta_kms_device_get_impl_device (device);
fd = meta_kms_impl_device_get_fd (impl_device);
custom_page_flip_func = page_flip->custom_page_flip_func;
if (custom_page_flip_func)
{
ret = custom_page_flip_func (page_flip->custom_page_flip_user_data,
meta_kms_page_flip_data_ref (page_flip_data));
}
else
{
ret = drmModePageFlip (fd,
meta_kms_crtc_get_id (crtc),
plane_assignment->fb_id,
DRM_MODE_PAGE_FLIP_EVENT,
meta_kms_page_flip_data_ref (page_flip_data));
}
kms-impl/simple: Fix page_flip_data ref leak on fallback If drmModePageFlip() or custom_page_flip_func fails, process_page_flip() was forgetting to undo the ref taken for that call. This would leak page_flip_data. The reference counting works like this: - when created, ref count is 1 - when calling drmModePageFlip, ref count is increased to 2 - new: if flip failed, ref count is decreased back to 1 - if calling schedule_retry_page_flip(), it takes a ref internally - if calling mode_set_fallback(), it takes a ref internally - all return FALSE paths have an explicit unref - return TRUE path has an explicit unref This issue was found by code inspection and while debugging an unrelated issue with debug prints sprinkled around. I am not aware of any end-user visible issues being fixed by this, as the leak is small and probably very rare. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1209
2020-04-17 11:50:13 +00:00
if (ret != 0)
meta_kms_page_flip_data_unref (page_flip_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
if (ret == -EBUSY)
{
kms-impl-simple: Handle mode set race conditions gracefully If we end up trying to do a mode set on a DRM state that has already changed behind our back without us yet having seen the hotplug event we may fail with `EINVAL`. Since the renderer layer doesn't handle mode set failure, it'll still try to page flip later on, which will then also fail. When failing, it'll try to look up the cached mode set in order to retry the mode set later on, as is needed to handle other error conditions. However, if the mode set prior to the page flip failed, we won't cache the mode set, and the page flip error handling code will get confused. Instead of asserting that a page flip always has a valid cached mode set ready to look up, handle it being missing more gracefully by failing to mode set. It is expected that things will correct themself as there should be a hotplug event waiting around the the corner, to reconfigure the monitor configuration setting new modes. Fixes: https://gitlab.gnome.org/GNOME/mutter/issues/917 https://gitlab.gnome.org/GNOME/mutter/merge_requests/1007
2020-01-14 10:16:09 +00:00
CachedModeSet *cached_mode_set;
cached_mode_set = get_cached_mode_set (impl_simple, crtc);
if (cached_mode_set)
{
drmModeModeInfo *drm_mode;
float refresh_rate;
drm_mode = cached_mode_set->drm_mode;
refresh_rate = meta_calculate_drm_mode_refresh_rate (drm_mode);
schedule_retry_page_flip (impl_simple,
crtc,
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
plane_assignment ?
plane_assignment->fb_id : 0,
kms-impl-simple: Handle mode set race conditions gracefully If we end up trying to do a mode set on a DRM state that has already changed behind our back without us yet having seen the hotplug event we may fail with `EINVAL`. Since the renderer layer doesn't handle mode set failure, it'll still try to page flip later on, which will then also fail. When failing, it'll try to look up the cached mode set in order to retry the mode set later on, as is needed to handle other error conditions. However, if the mode set prior to the page flip failed, we won't cache the mode set, and the page flip error handling code will get confused. Instead of asserting that a page flip always has a valid cached mode set ready to look up, handle it being missing more gracefully by failing to mode set. It is expected that things will correct themself as there should be a hotplug event waiting around the the corner, to reconfigure the monitor configuration setting new modes. Fixes: https://gitlab.gnome.org/GNOME/mutter/issues/917 https://gitlab.gnome.org/GNOME/mutter/merge_requests/1007
2020-01-14 10:16:09 +00:00
refresh_rate,
kms: Use custom page flip function when retrying failed flips When using its EGLStream-based presentation path with the proprietary NVIDIA driver, mutter will use a different function to process page flips - custom_egl_stream_page_flip. If that fails due to an EBUSY error, it will attempt to retry the flip. However, when retrying, it unconditionally uses the libdrm-based path. In practice, this causes a segfault when attempting to access plane_assignments->fb_id, since plane_assignments will be NULL in the EGLStream case. The issue can be reproduced reliably by VT-switching away from GNOME and back again while an EGL application is running. This patch has mutter also use the custom page flip function when retrying the failed flip. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1375
2020-07-17 20:38:55 +00:00
page_flip_data,
custom_page_flip_func,
page_flip->custom_page_flip_user_data);
kms-impl-simple: Handle mode set race conditions gracefully If we end up trying to do a mode set on a DRM state that has already changed behind our back without us yet having seen the hotplug event we may fail with `EINVAL`. Since the renderer layer doesn't handle mode set failure, it'll still try to page flip later on, which will then also fail. When failing, it'll try to look up the cached mode set in order to retry the mode set later on, as is needed to handle other error conditions. However, if the mode set prior to the page flip failed, we won't cache the mode set, and the page flip error handling code will get confused. Instead of asserting that a page flip always has a valid cached mode set ready to look up, handle it being missing more gracefully by failing to mode set. It is expected that things will correct themself as there should be a hotplug event waiting around the the corner, to reconfigure the monitor configuration setting new modes. Fixes: https://gitlab.gnome.org/GNOME/mutter/issues/917 https://gitlab.gnome.org/GNOME/mutter/merge_requests/1007
2020-01-14 10:16:09 +00:00
}
else
{
g_set_error (error, G_IO_ERROR, G_IO_ERROR_FAILED,
"Page flip of %u failed, and no mode set available",
meta_kms_crtc_get_id (crtc));
meta_kms_page_flip_data_unref (page_flip_data);
return FALSE;
}
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
}
else if (ret == -EINVAL)
{
if (!mode_set_fallback (impl_simple,
update,
page_flip,
plane_assignment,
page_flip_data,
error))
{
meta_kms_page_flip_data_unref (page_flip_data);
return FALSE;
}
}
else if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
"drmModePageFlip on CRTC %u failed: %s",
meta_kms_crtc_get_id (crtc),
g_strerror (-ret));
meta_kms_page_flip_data_unref (page_flip_data);
return FALSE;
}
meta_kms_page_flip_data_unref (page_flip_data);
return TRUE;
}
static void
discard_page_flip (MetaKmsImpl *impl,
MetaKmsUpdate *update,
MetaKmsPageFlip *page_flip)
{
MetaKmsCrtc *crtc;
MetaKmsPageFlipData *page_flip_data;
crtc = page_flip->crtc;
page_flip_data = meta_kms_page_flip_data_new (impl,
crtc,
page_flip->feedback,
page_flip->user_data);
meta_kms_page_flip_data_discard_in_impl (page_flip_data, NULL);
meta_kms_page_flip_data_unref (page_flip_data);
}
static gboolean
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
process_entries (MetaKmsImpl *impl,
MetaKmsUpdate *update,
GList *entries,
gboolean (* func) (MetaKmsImpl *impl,
MetaKmsUpdate *update,
gpointer entry_data,
GError **error),
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
{
GList *l;
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
for (l = entries; l; l = l->next)
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
{
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
if (!func (impl, update, l->data, error))
return FALSE;
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
}
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
return TRUE;
}
monitor-manager/kms: Use KMS abstraction to get and set CRTC gamma Still doesn't synchronize with frame drawing, but no point in doing that until gamma is managed by mutter itself and not gnome-settings-daemon. https://gitlab.gnome.org/GNOME/mutter/merge_requests/634
2019-05-03 17:20:21 +00:00
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
static gboolean
process_cursor_plane_assignment (MetaKmsImpl *impl,
MetaKmsUpdate *update,
MetaKmsPlaneAssignment *plane_assignment,
GError **error)
{
MetaKmsPlane *plane;
MetaKmsDevice *device;
MetaKmsImplDevice *impl_device;
int fd;
plane = plane_assignment->plane;
device = meta_kms_plane_get_device (plane);
impl_device = meta_kms_device_get_impl_device (device);
fd = meta_kms_impl_device_get_fd (impl_device);
if (!(plane_assignment->flags & META_KMS_ASSIGN_PLANE_FLAG_FB_UNCHANGED))
{
int width, height;
kms/plane-assignment: Add API to set cursor hotspot metadata The transactional KMS API has been modelled after atomic KMS. Atomic KMS currently doesn't support forwarding cursor hotspot metadata, thus it was left out of the transactional KMS API having the user set the simply create a plane assigment with the cursor sprite assigned to a cursor plane using regular coordinates. This, however, proved to be inadequate for virtual machines using "seamless mouse mode" where they rely on the cursor position to correspond to the actual cursor position of the virtual machine, not the cursor plane. In effect, this caused cursor positions to look "shifted". Fix this by adding back the hotspot metadata, right now as a optional field to the plane assignment. In the legacy KMS implementation, this is translated into drmModeSetCursor2() just as before, while still falling back to drmModeSetCursor() with the plane coordinates, if either there was no hotspot set, or if drmModeSetCursor2() failed. Eventually, the atomic KMS API will learn about hotspots, but when adding our own atomic KMS backend to the transacitonal KMS API, we must until then still fall back to legacy KMS for virtual machines. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1136
2020-03-19 08:01:03 +00:00
int ret = -1;
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
width = meta_fixed_16_to_int (plane_assignment->dst_rect.width);
height = meta_fixed_16_to_int (plane_assignment->dst_rect.height);
kms/plane-assignment: Add API to set cursor hotspot metadata The transactional KMS API has been modelled after atomic KMS. Atomic KMS currently doesn't support forwarding cursor hotspot metadata, thus it was left out of the transactional KMS API having the user set the simply create a plane assigment with the cursor sprite assigned to a cursor plane using regular coordinates. This, however, proved to be inadequate for virtual machines using "seamless mouse mode" where they rely on the cursor position to correspond to the actual cursor position of the virtual machine, not the cursor plane. In effect, this caused cursor positions to look "shifted". Fix this by adding back the hotspot metadata, right now as a optional field to the plane assignment. In the legacy KMS implementation, this is translated into drmModeSetCursor2() just as before, while still falling back to drmModeSetCursor() with the plane coordinates, if either there was no hotspot set, or if drmModeSetCursor2() failed. Eventually, the atomic KMS API will learn about hotspots, but when adding our own atomic KMS backend to the transacitonal KMS API, we must until then still fall back to legacy KMS for virtual machines. https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1136
2020-03-19 08:01:03 +00:00
if (plane_assignment->cursor_hotspot.is_valid)
{
ret = drmModeSetCursor2 (fd, meta_kms_crtc_get_id (plane_assignment->crtc),
plane_assignment->fb_id,
width, height,
plane_assignment->cursor_hotspot.x,
plane_assignment->cursor_hotspot.y);
}
if (ret != 0)
{
ret = drmModeSetCursor (fd, meta_kms_crtc_get_id (plane_assignment->crtc),
plane_assignment->fb_id,
width, height);
}
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
if (ret != 0)
{
g_set_error (error, G_IO_ERROR, g_io_error_from_errno (-ret),
"drmModeSetCursor failed: %s", g_strerror (-ret));
return FALSE;
}
}
drmModeMoveCursor (fd,
meta_kms_crtc_get_id (plane_assignment->crtc),
meta_fixed_16_to_int (plane_assignment->dst_rect.x),
meta_fixed_16_to_int (plane_assignment->dst_rect.y));
return TRUE;
}
static gboolean
process_plane_assignment (MetaKmsImpl *impl,
MetaKmsUpdate *update,
MetaKmsPlaneAssignment *plane_assignment,
MetaKmsPlaneFeedback **plane_feedback)
{
MetaKmsPlane *plane;
MetaKmsPlaneType plane_type;
GError *error = NULL;
plane = plane_assignment->plane;
plane_type = meta_kms_plane_get_plane_type (plane);
switch (plane_type)
{
case META_KMS_PLANE_TYPE_PRIMARY:
/* Handled as part of the mode-set and page flip. */
return TRUE;
case META_KMS_PLANE_TYPE_CURSOR:
if (!process_cursor_plane_assignment (impl, update,
plane_assignment,
&error))
{
*plane_feedback =
meta_kms_plane_feedback_new_take_error (plane,
plane_assignment->crtc,
g_steal_pointer (&error));
return FALSE;
}
else
{
return TRUE;
}
case META_KMS_PLANE_TYPE_OVERLAY:
error = g_error_new_literal (G_IO_ERROR, G_IO_ERROR_FAILED,
"Overlay planes cannot be assigned");
*plane_feedback =
meta_kms_plane_feedback_new_take_error (plane,
plane_assignment->crtc,
g_steal_pointer (&error));
return TRUE;
}
g_assert_not_reached ();
}
static GList *
process_plane_assignments (MetaKmsImpl *impl,
MetaKmsUpdate *update)
{
GList *failed_planes = NULL;
GList *l;
for (l = meta_kms_update_get_plane_assignments (update); l; l = l->next)
{
MetaKmsPlaneAssignment *plane_assignment = l->data;
MetaKmsPlaneFeedback *plane_feedback;
if (!process_plane_assignment (impl, update, plane_assignment,
&plane_feedback))
failed_planes = g_list_prepend (failed_planes, plane_feedback);
}
return failed_planes;
}
static GList *
generate_all_failed_feedbacks (MetaKmsUpdate *update)
{
GList *failed_planes = NULL;
GList *l;
for (l = meta_kms_update_get_plane_assignments (update); l; l = l->next)
{
MetaKmsPlaneAssignment *plane_assignment = l->data;
MetaKmsPlane *plane;
MetaKmsPlaneType plane_type;
MetaKmsPlaneFeedback *plane_feedback;
plane = plane_assignment->plane;
plane_type = meta_kms_plane_get_plane_type (plane);
switch (plane_type)
{
case META_KMS_PLANE_TYPE_PRIMARY:
continue;
case META_KMS_PLANE_TYPE_CURSOR:
case META_KMS_PLANE_TYPE_OVERLAY:
break;
}
plane_feedback =
meta_kms_plane_feedback_new_take_error (plane_assignment->plane,
plane_assignment->crtc,
g_error_new (G_IO_ERROR,
G_IO_ERROR_FAILED,
"Discarded"));
failed_planes = g_list_prepend (failed_planes, plane_feedback);
}
return failed_planes;
}
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
static MetaKmsFeedback *
meta_kms_impl_simple_process_update (MetaKmsImpl *impl,
MetaKmsUpdate *update)
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
{
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
GError *error = NULL;
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
GList *failed_planes;
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
GList *l;
monitor-manager/kms: Use KMS abstraction to get and set CRTC gamma Still doesn't synchronize with frame drawing, but no point in doing that until gamma is managed by mutter itself and not gnome-settings-daemon. https://gitlab.gnome.org/GNOME/mutter/merge_requests/634
2019-05-03 17:20:21 +00:00
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
meta_assert_in_kms_impl (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
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
if (!process_entries (impl,
update,
meta_kms_update_get_connector_properties (update),
process_connector_property,
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
&error))
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
goto err_planes_not_assigned;
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
if (!process_entries (impl,
update,
meta_kms_update_get_mode_sets (update),
process_mode_set,
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
&error))
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
goto err_planes_not_assigned;
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
if (!process_entries (impl,
update,
meta_kms_update_get_crtc_gammas (update),
process_crtc_gamma,
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
&error))
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
goto err_planes_not_assigned;
failed_planes = process_plane_assignments (impl, update);
if (failed_planes)
{
g_set_error (&error, G_IO_ERROR, G_IO_ERROR_FAILED,
"Failed to assign one or more planes");
goto err_planes_assigned;
}
kms-impl-simple: Add helper to process update entries This avoids some loop iteration boiler plate. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 10:30:18 +00:00
if (!process_entries (impl,
update,
meta_kms_update_get_page_flips (update),
process_page_flip,
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
&error))
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
goto err_planes_assigned;
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
kms: Make update processing return direct feedback The current API as all synchronous, so they can be made to return feedback immediately. This will be needed for the cursor renderer which needs to know whether it should fall back to OpenGL cursor rendering. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:14:36 +00:00
return meta_kms_feedback_new_passed ();
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
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
err_planes_not_assigned:
failed_planes = generate_all_failed_feedbacks (update);
err_planes_assigned:
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
for (l = meta_kms_update_get_page_flips (update); l; l = l->next)
{
MetaKmsPageFlip *page_flip = l->data;
backends/native: Fall back to compositing if direct scanout failed Even when a direct client buffer has a compatible format, stride and modifier for direct scanout, drmModePageFlip() may still fail sometimes. From testing, it has been observed that it may seemingly randomly fail with ENOSPC, where all subsequent attempts later on the same CRTC failing with EBUSY. Handle this by falling back to flipping after having composited a full frame again. Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/1410
2020-09-10 15:41:06 +00:00
if (page_flip->flags & META_KMS_PAGE_FLIP_FLAG_NO_DISCARD_FEEDBACK)
continue;
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
discard_page_flip (impl, update, page_flip);
}
kms/impl-simple: Add cursor plane processing A cursor plane can now be assigned, and for the simple KMS implementation, it'll translate into drmModeSetCursor() and drmModeMoveCursor() calls. When assignments failed, the cursor planes that failed to be assigned are communicated via the feedback object. https://gitlab.gnome.org/GNOME/mutter/merge_requests/930
2019-11-08 23:17:33 +00:00
return meta_kms_feedback_new_failed (failed_planes, 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
}
static void
flush_postponed_page_flip_datas (MetaKmsImplSimple *impl_simple)
{
invoke_page_flip_datas (impl_simple->postponed_page_flip_datas,
meta_kms_page_flip_data_flipped_in_impl);
clear_page_flip_datas (&impl_simple->postponed_page_flip_datas);
invoke_page_flip_datas (impl_simple->postponed_mode_set_fallback_datas,
meta_kms_page_flip_data_mode_set_fallback_in_impl);
clear_page_flip_datas (&impl_simple->postponed_mode_set_fallback_datas);
}
static void
meta_kms_impl_simple_handle_page_flip_callback (MetaKmsImpl *impl,
MetaKmsPageFlipData *page_flip_data)
{
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (impl);
if (impl_simple->pending_page_flip_retries)
{
impl_simple->postponed_page_flip_datas =
g_list_append (impl_simple->postponed_page_flip_datas,
backends: Plug MetaKmsPageFlipData leak If the page flip is postponed, keep a ref to it, but still unref it on the page flip callback anyways. Fix suggested by Jonas Ådahl. https://gitlab.gnome.org/GNOME/mutter/merge_requests/873
2019-10-21 16:55:12 +00:00
meta_kms_page_flip_data_ref (page_flip_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
}
else
{
meta_kms_page_flip_data_flipped_in_impl (page_flip_data);
}
backends: Plug MetaKmsPageFlipData leak If the page flip is postponed, keep a ref to it, but still unref it on the page flip callback anyways. Fix suggested by Jonas Ådahl. https://gitlab.gnome.org/GNOME/mutter/merge_requests/873
2019-10-21 16:55:12 +00:00
meta_kms_page_flip_data_unref (page_flip_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 void
meta_kms_impl_simple_discard_pending_page_flips (MetaKmsImpl *impl)
{
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (impl);
GList *l;
if (!impl_simple->pending_page_flip_retries)
return;
for (l = impl_simple->pending_page_flip_retries; l; l = l->next)
{
RetryPageFlipData *retry_page_flip_data = l->data;
MetaKmsPageFlipData *page_flip_data;
page_flip_data = g_steal_pointer (&retry_page_flip_data->page_flip_data);
meta_kms_page_flip_data_discard_in_impl (page_flip_data, NULL);
retry_page_flip_data_free (retry_page_flip_data);
}
g_clear_pointer (&impl_simple->pending_page_flip_retries, g_list_free);
g_clear_pointer (&impl_simple->retry_page_flips_source,
g_source_destroy);
}
kms: Process impl idle callbacks before pre dispatch flush mode_set_fallback() schedules a call to mode_set_fallback_feedback_idle(), but it is possible for Mutter to repaint before the idle callbacks are dispatched. If that happens, mode_set_fallback_feedback_idle() does not get called before Mutter enters wait_for_pending_flips(), leading to a deadlock. Add the needed interfaces so that meta_kms_device_dispatch_sync() can flush all the implementation idle callbacks before it checks if any "events" are available. This prevents the deadlock by ensuring mode_set_fallback_feedback_idle() does get called before potentially waiting for actual DRM events. Presumably this call would not be needed if the implementation was running in its own thread, since it would eventually dispatch its idle callbacks before going to sleep polling on the DRM fd. This call might even be unnecessary overhead in that case, synchronizing with the implementation thread needlessly. But the thread does not exist yet, so this is needed for now. This is part 1 of 2 fixing a complete desktop freeze when drmModePageFlip() fails with EINVAL and the fallback to drmModeSetCrtc() succeeds but the success is not registered correctly as completed "flip". The freeze occurs under wait_for_pending_flips() which calls down into meta_kms_impl_device_dispatch() which ends up poll()'ing the DRM fd even though drmModeSetCrtc() will not produce a DRM event, hence the poll() never returns. The freeze was observed when hotplugging a DisplayLink dock for the first time on Ubuntu 19.10. CC stable: gnome-3-34 https://gitlab.gnome.org/GNOME/mutter/merge_requests/953
2019-11-29 10:50:30 +00:00
static void
meta_kms_impl_simple_dispatch_idle (MetaKmsImpl *impl)
{
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (impl);
if (impl_simple->mode_set_fallback_feedback_source)
mode_set_fallback_feedback_idle (impl_simple);
}
kms-impl-simple: Add fake cursor planes if no real ones Non-atomic drivers may support drmModeSetCursor() even if no cursor plane is advertised. To deal with this, add a fake cursor plane for every CRTC when using MetaKmsImplSimple. This will eventually be translated to drmModeSetCursor() calls without any explicit cursor plane usage. Fixes: https://gitlab.gnome.org/GNOME/mutter/issues/1058 https://gitlab.gnome.org/GNOME/mutter/merge_requests/1079
2020-02-21 11:06:28 +00:00
static void
meta_kms_impl_simple_notify_device_created (MetaKmsImpl *impl,
MetaKmsDevice *device)
{
GList *l;
for (l = meta_kms_device_get_crtcs (device); l; l = l->next)
{
MetaKmsCrtc *crtc = l->data;
MetaKmsPlane *plane;
plane = meta_kms_device_get_cursor_plane_for (device, crtc);
if (plane)
continue;
meta_kms_device_add_fake_plane_in_impl (device,
META_KMS_PLANE_TYPE_CURSOR,
crtc);
}
}
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 void
meta_kms_impl_simple_finalize (GObject *object)
{
MetaKmsImplSimple *impl_simple = META_KMS_IMPL_SIMPLE (object);
g_list_free_full (impl_simple->pending_page_flip_retries,
(GDestroyNotify) retry_page_flip_data_free);
g_list_free_full (impl_simple->postponed_page_flip_datas,
(GDestroyNotify) meta_kms_page_flip_data_unref);
g_list_free_full (impl_simple->postponed_mode_set_fallback_datas,
(GDestroyNotify) meta_kms_page_flip_data_unref);
g_clear_pointer (&impl_simple->mode_set_fallback_feedback_source,
g_source_destroy);
g_hash_table_destroy (impl_simple->cached_mode_sets);
G_OBJECT_CLASS (meta_kms_impl_simple_parent_class)->finalize (object);
}
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_impl_simple_init (MetaKmsImplSimple *impl_simple)
{
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
impl_simple->cached_mode_sets =
g_hash_table_new_full (NULL,
NULL,
NULL,
(GDestroyNotify) cached_mode_set_free);
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_impl_simple_class_init (MetaKmsImplSimpleClass *klass)
{
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
GObjectClass *object_class = G_OBJECT_CLASS (klass);
MetaKmsImplClass *impl_class = META_KMS_IMPL_CLASS (klass);
object_class->finalize = meta_kms_impl_simple_finalize;
impl_class->process_update = meta_kms_impl_simple_process_update;
impl_class->handle_page_flip_callback = meta_kms_impl_simple_handle_page_flip_callback;
impl_class->discard_pending_page_flips = meta_kms_impl_simple_discard_pending_page_flips;
kms: Process impl idle callbacks before pre dispatch flush mode_set_fallback() schedules a call to mode_set_fallback_feedback_idle(), but it is possible for Mutter to repaint before the idle callbacks are dispatched. If that happens, mode_set_fallback_feedback_idle() does not get called before Mutter enters wait_for_pending_flips(), leading to a deadlock. Add the needed interfaces so that meta_kms_device_dispatch_sync() can flush all the implementation idle callbacks before it checks if any "events" are available. This prevents the deadlock by ensuring mode_set_fallback_feedback_idle() does get called before potentially waiting for actual DRM events. Presumably this call would not be needed if the implementation was running in its own thread, since it would eventually dispatch its idle callbacks before going to sleep polling on the DRM fd. This call might even be unnecessary overhead in that case, synchronizing with the implementation thread needlessly. But the thread does not exist yet, so this is needed for now. This is part 1 of 2 fixing a complete desktop freeze when drmModePageFlip() fails with EINVAL and the fallback to drmModeSetCrtc() succeeds but the success is not registered correctly as completed "flip". The freeze occurs under wait_for_pending_flips() which calls down into meta_kms_impl_device_dispatch() which ends up poll()'ing the DRM fd even though drmModeSetCrtc() will not produce a DRM event, hence the poll() never returns. The freeze was observed when hotplugging a DisplayLink dock for the first time on Ubuntu 19.10. CC stable: gnome-3-34 https://gitlab.gnome.org/GNOME/mutter/merge_requests/953
2019-11-29 10:50:30 +00:00
impl_class->dispatch_idle = meta_kms_impl_simple_dispatch_idle;
kms-impl-simple: Add fake cursor planes if no real ones Non-atomic drivers may support drmModeSetCursor() even if no cursor plane is advertised. To deal with this, add a fake cursor plane for every CRTC when using MetaKmsImplSimple. This will eventually be translated to drmModeSetCursor() calls without any explicit cursor plane usage. Fixes: https://gitlab.gnome.org/GNOME/mutter/issues/1058 https://gitlab.gnome.org/GNOME/mutter/merge_requests/1079
2020-02-21 11:06:28 +00:00
impl_class->notify_device_created = meta_kms_impl_simple_notify_device_created;
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
}
Copy permalink