mutter/src/backends/native/meta-output-kms.c

418 lines
13 KiB
C
Raw Normal View History

/* -*- mode: C; c-file-style: "gnu"; indent-tabs-mode: nil; -*- */
/*
* Copyright (C) 2013-2017 Red Hat
* Copyright (C) 2018 DisplayLink (UK) Ltd.
*
* 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-output-kms.h"
#include <errno.h>
#include <stdlib.h>
#include <string.h>
#include "backends/meta-crtc.h"
#include "backends/native/meta-kms-connector.h"
#include "backends/native/meta-kms-utils.h"
#include "backends/native/meta-crtc-kms.h"
#include "meta-default-modes.h"
#define SYNC_TOLERANCE 0.01 /* 1 percent */
typedef struct _MetaOutputKms
{
MetaOutput parent;
MetaKmsConnector *kms_connector;
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 16:36:41 -04:00
} MetaOutputKms;
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 16:36:41 -04:00
MetaKmsConnector *
meta_output_kms_get_kms_connector (MetaOutput *output)
{
MetaOutputKms *output_kms = output->driver_private;
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 16:36:41 -04:00
return output_kms->kms_connector;
}
void
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 16:36:41 -04:00
meta_output_kms_set_underscan (MetaOutput *output,
MetaKmsUpdate *kms_update)
{
MetaOutputKms *output_kms = output->driver_private;
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 16:36:41 -04:00
if (!output->supports_underscanning)
return;
if (meta_output_is_underscanning (output))
{
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 16:36:41 -04:00
MetaCrtc *crtc;
MetaCrtcConfig *crtc_config;
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 16:36:41 -04:00
uint64_t hborder, vborder;
crtc = meta_output_get_assigned_crtc (output);
crtc_config = crtc->config;
hborder = MIN (128, (uint64_t) round (crtc_config->mode->width * 0.05));
vborder = MIN (128, (uint64_t) round (crtc_config->mode->height * 0.05));
g_debug ("Setting underscan of connector %s to %" G_GUINT64_FORMAT " x %" G_GUINT64_FORMAT,
meta_kms_connector_get_name (output_kms->kms_connector),
hborder, vborder);
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 16:36:41 -04:00
meta_kms_connector_set_underscanning (output_kms->kms_connector,
kms_update,
hborder,
vborder);
}
else
{
g_debug ("Unsetting underscan of connector %s",
meta_kms_connector_get_name (output_kms->kms_connector));
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 16:36:41 -04:00
meta_kms_connector_unset_underscanning (output_kms->kms_connector,
kms_update);
}
}
uint32_t
meta_output_kms_get_connector_id (MetaOutput *output)
{
MetaOutputKms *output_kms = output->driver_private;
return meta_kms_connector_get_id (output_kms->kms_connector);
}
void
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 16:36:41 -04:00
meta_output_kms_set_power_save_mode (MetaOutput *output,
uint64_t dpms_state,
MetaKmsUpdate *kms_update)
{
MetaOutputKms *output_kms = output->driver_private;
g_debug ("Setting DPMS state of connector %s to %" G_GUINT64_FORMAT,
meta_kms_connector_get_name (output_kms->kms_connector),
dpms_state);
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 16:36:41 -04:00
meta_kms_connector_update_set_dpms_state (output_kms->kms_connector,
kms_update,
dpms_state);
}
gboolean
meta_output_kms_can_clone (MetaOutput *output,
MetaOutput *other_output)
{
MetaOutputKms *output_kms = output->driver_private;
MetaOutputKms *other_output_kms = other_output->driver_private;
return meta_kms_connector_can_clone (output_kms->kms_connector,
other_output_kms->kms_connector);
}
GBytes *
meta_output_kms_read_edid (MetaOutput *output)
{
MetaOutputKms *output_kms = output->driver_private;
const MetaKmsConnectorState *connector_state;
GBytes *edid_data;
connector_state =
meta_kms_connector_get_current_state (output_kms->kms_connector);
edid_data = connector_state->edid_data;
if (!edid_data)
return NULL;
return g_bytes_new_from_bytes (edid_data, 0, g_bytes_get_size (edid_data));
}
static void
meta_output_destroy_notify (MetaOutput *output)
{
MetaOutputKms *output_kms;
output_kms = output->driver_private;
g_slice_free (MetaOutputKms, output_kms);
}
static void
add_common_modes (MetaOutput *output,
MetaGpuKms *gpu_kms)
{
const drmModeModeInfo *drm_mode;
MetaCrtcMode *crtc_mode;
GPtrArray *array;
float refresh_rate;
unsigned i;
unsigned max_hdisplay = 0;
unsigned max_vdisplay = 0;
float max_refresh_rate = 0.0;
for (i = 0; i < output->n_modes; i++)
{
drm_mode = output->modes[i]->driver_private;
refresh_rate = meta_calculate_drm_mode_refresh_rate (drm_mode);
max_hdisplay = MAX (max_hdisplay, drm_mode->hdisplay);
max_vdisplay = MAX (max_vdisplay, drm_mode->vdisplay);
max_refresh_rate = MAX (max_refresh_rate, refresh_rate);
}
max_refresh_rate = MAX (max_refresh_rate, 60.0);
max_refresh_rate *= (1 + SYNC_TOLERANCE);
array = g_ptr_array_new ();
if (max_hdisplay > max_vdisplay)
{
for (i = 0; i < G_N_ELEMENTS (meta_default_landscape_drm_mode_infos); i++)
{
drm_mode = &meta_default_landscape_drm_mode_infos[i];
refresh_rate = meta_calculate_drm_mode_refresh_rate (drm_mode);
if (drm_mode->hdisplay > max_hdisplay ||
drm_mode->vdisplay > max_vdisplay ||
refresh_rate > max_refresh_rate)
continue;
crtc_mode = meta_gpu_kms_get_mode_from_drm_mode (gpu_kms,
drm_mode);
g_ptr_array_add (array, crtc_mode);
}
}
else
{
for (i = 0; i < G_N_ELEMENTS (meta_default_portrait_drm_mode_infos); i++)
{
drm_mode = &meta_default_portrait_drm_mode_infos[i];
refresh_rate = meta_calculate_drm_mode_refresh_rate (drm_mode);
if (drm_mode->hdisplay > max_hdisplay ||
drm_mode->vdisplay > max_vdisplay ||
refresh_rate > max_refresh_rate)
continue;
crtc_mode = meta_gpu_kms_get_mode_from_drm_mode (gpu_kms,
drm_mode);
g_ptr_array_add (array, crtc_mode);
}
}
output->modes = g_renew (MetaCrtcMode *, output->modes,
output->n_modes + array->len);
memcpy (output->modes + output->n_modes, array->pdata,
array->len * sizeof (MetaCrtcMode *));
output->n_modes += array->len;
g_ptr_array_free (array, TRUE);
}
static int
compare_modes (const void *one,
const void *two)
{
MetaCrtcMode *a = *(MetaCrtcMode **) one;
MetaCrtcMode *b = *(MetaCrtcMode **) two;
if (a->width != b->width)
return a->width > b->width ? -1 : 1;
if (a->height != b->height)
return a->height > b->height ? -1 : 1;
if (a->refresh_rate != b->refresh_rate)
return a->refresh_rate > b->refresh_rate ? -1 : 1;
return g_strcmp0 (b->name, a->name);
}
static gboolean
init_output_modes (MetaOutput *output,
MetaGpuKms *gpu_kms,
GError **error)
{
MetaOutputKms *output_kms = output->driver_private;
const MetaKmsConnectorState *connector_state;
int i;
connector_state =
meta_kms_connector_get_current_state (output_kms->kms_connector);
output->preferred_mode = NULL;
output->n_modes = connector_state->n_modes;
output->modes = g_new0 (MetaCrtcMode *, output->n_modes);
for (i = 0; i < connector_state->n_modes; i++)
{
drmModeModeInfo *drm_mode = &connector_state->modes[i];
MetaCrtcMode *crtc_mode;
crtc_mode = meta_gpu_kms_get_mode_from_drm_mode (gpu_kms, drm_mode);
output->modes[i] = crtc_mode;
if (drm_mode->type & DRM_MODE_TYPE_PREFERRED)
output->preferred_mode = output->modes[i];
}
/* FIXME: MSC feature bit? */
/* Presume that if the output supports scaling, then we have
* a panel fitter capable of adjusting any mode to suit.
*/
if (connector_state->has_scaling)
add_common_modes (output, gpu_kms);
if (!output->modes)
{
g_set_error (error, G_IO_ERROR, G_IO_ERROR_FAILED,
"No modes available");
return FALSE;
}
qsort (output->modes, output->n_modes,
sizeof (MetaCrtcMode *), compare_modes);
if (!output->preferred_mode)
output->preferred_mode = output->modes[0];
return TRUE;
}
MetaOutput *
meta_create_kms_output (MetaGpuKms *gpu_kms,
MetaKmsConnector *kms_connector,
MetaOutput *old_output,
GError **error)
{
MetaGpu *gpu = META_GPU (gpu_kms);
uint32_t connector_id;
uint32_t gpu_id;
MetaOutput *output;
MetaOutputKms *output_kms;
const MetaKmsConnectorState *connector_state;
GArray *crtcs;
GList *l;
gpu_id = meta_gpu_kms_get_id (gpu_kms);
connector_id = meta_kms_connector_get_id (kms_connector);
output = g_object_new (META_TYPE_OUTPUT,
"id", ((uint64_t) gpu_id << 32) | connector_id,
"gpu", gpu,
NULL);
output_kms = g_slice_new0 (MetaOutputKms);
output->driver_private = output_kms;
output->driver_notify = (GDestroyNotify) meta_output_destroy_notify;
output->name = g_strdup (meta_kms_connector_get_name (kms_connector));
output_kms->kms_connector = kms_connector;
connector_state = meta_kms_connector_get_current_state (kms_connector);
output->panel_orientation_transform =
connector_state->panel_orientation_transform;
if (meta_monitor_transform_is_rotated (output->panel_orientation_transform))
{
output->width_mm = connector_state->height_mm;
output->height_mm = connector_state->width_mm;
}
else
{
output->width_mm = connector_state->width_mm;
output->height_mm = connector_state->height_mm;
}
if (!init_output_modes (output, gpu_kms, error))
{
g_object_unref (output);
return NULL;
}
crtcs = g_array_new (FALSE, FALSE, sizeof (MetaCrtc *));
for (l = meta_gpu_get_crtcs (gpu); l; l = l->next)
{
MetaCrtc *crtc = l->data;
MetaKmsCrtc *kms_crtc = meta_crtc_kms_get_kms_crtc (crtc);
uint32_t crtc_idx;
crtc_idx = meta_kms_crtc_get_idx (kms_crtc);
if (connector_state->common_possible_crtcs & (1 << crtc_idx))
g_array_append_val (crtcs, crtc);
}
output->n_possible_crtcs = crtcs->len;
output->possible_crtcs = (MetaCrtc **) g_array_free (crtcs, FALSE);
if (connector_state->current_crtc_id)
{
for (l = meta_gpu_get_crtcs (gpu); l; l = l->next)
{
MetaCrtc *crtc = l->data;
if (meta_crtc_get_id (crtc) == connector_state->current_crtc_id)
{
MetaOutputAssignment output_assignment;
if (old_output)
{
output_assignment = (MetaOutputAssignment) {
.is_primary = meta_output_is_primary (old_output),
.is_presentation = meta_output_is_presentation (old_output),
};
}
else
{
output_assignment = (MetaOutputAssignment) {
.is_primary = FALSE,
.is_presentation = FALSE,
};
}
meta_output_assign_crtc (output, crtc, &output_assignment);
break;
}
}
}
else
{
meta_output_unassign_crtc (output);
}
output->suggested_x = connector_state->suggested_x;
output->suggested_y = connector_state->suggested_y;
output->hotplug_mode_update = connector_state->hotplug_mode_update;
backend/native: Add and use transactional KMS API This commit introduces, and makes use of, a transactional API used for setting up KMS state, later to be applied, potentially atomically. From an API point of view, so is always the case, but in the current implementation, it still uses legacy drmMode* API to apply the state non-atomically. The API consists of various buliding blocks: * MetaKmsUpdate - a set of configuration changes, the higher level handle for handing over configuration to the impl backend. It's used to set mode, assign framebuffers to planes, queue page flips and set connector properties. * MetaKmsPlaneAssignment - the assignment of a framebuffer to a plane. Currently used to map a framebuffer to the primary plane of a CRTC. In the legacy KMS implementation, the plane assignment is used to derive the framebuffer used for mode setting and page flipping. This also means various high level changes: State, excluding configuring the cursor plane and creating/destroying DRM framebuffer handles, are applied in the end of a clutter frame, in one go. From an API point of view, this is done atomically, but as mentioned, only the non-atomic implementation exists so far. From MetaRendererNative's point of view, a page flip now initially always succeeds; the handling of EBUSY errors are done asynchronously in the MetaKmsImpl backend (still by retrying at refresh rate, but postponing flip callbacks instead of manipulating the frame clock). Handling of falling back to mode setting instead of page flipping is notified after the fact by a more precise page flip feedback API. EGLStream based page flipping relies on the impl backend not being atomic, as the page flipping is done in the EGLStream backend (e.g. nvidia driver). It uses a 'custom' page flip queueing method, keeping the EGLStream logic inside meta-renderer-native.c. Page flip handling is moved to meta-kms-impl-device.c from meta-gpu-kms.c. It goes via an extra idle callback before reaching meta-renderer-native.c to make sure callbacks are invoked outside of the impl context. While dummy power save page flipping is kept in meta-renderer-native.c, the EBUSY handling is moved to meta-kms-impl-simple.c. Instead of freezing the frame clock, actual page flip callbacks are postponed until all EBUSY retries have either succeeded or failed due to some other error than EBUSY. This effectively inhibits new frames to be drawn, meaning we won't stall waiting on the file descriptor for pending page flips. https://gitlab.gnome.org/GNOME/mutter/issues/548 https://gitlab.gnome.org/GNOME/mutter/merge_requests/525
2019-04-04 16:36:41 -04:00
output->supports_underscanning =
meta_kms_connector_is_underscanning_supported (kms_connector);
meta_output_parse_edid (output, connector_state->edid_data);
output->connector_type = meta_kms_connector_get_connector_type (kms_connector);
output->tile_info = connector_state->tile_info;
/* FIXME: backlight is a very driver specific thing unfortunately,
every DDX does its own thing, and the dumb KMS API does not include it.
For example, xf86-video-intel has a list of paths to probe in /sys/class/backlight
(one for each major HW maker, and then some).
We can't do the same because we're not root.
It might be best to leave backlight out of the story and rely on the setuid
helper in gnome-settings-daemon.
*/
output->backlight_min = 0;
output->backlight_max = 0;
return output;
}