Call drm_client_setup() to run the kernel's default client setup
for DRM. Set fbdev_probe in struct drm_driver, so that the client
setup can start the common fbdev client.
The bochs driver specifies a preferred color mode of 32. As this
is the default if no format has been given, leave it out entirely.
v5:
- select DRM_CLIENT_SELECTION
v4:
- rebase on top of GEM-SHMEM
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240924071734.98201-52-tzimmermann@suse.de
Opt for devm_ioremap_wc() over devm_ioremap() when mapping the framebuffer.
Using devm_ioremap() results in the VA being mapped with PAT=UC-, which
considerably slows down drm_fb_memcpy(). In contrast, devm_ioremap_wc()
maps the VA with PAT set to WC, leading to better performance on platforms
where access to UC memory is much slower than WC memory.
Here's the performance data measured in a guest on the physical machine
"Sapphire Rapids XCC".
With host KVM honors guest PAT memory types, the effective memory type
for this framebuffer range is
- WC when devm_ioremap_wc() is used
- UC- when devm_ioremap() is used.
The data presented is an average from 10 execution runs.
Cycles: Avg cycles of executed bochs_primary_plane_helper_atomic_update()
from VM boot to GDM show up
Cnt: Avg cnt of executed bochs_primary_plane_helper_atomic_update()
from VM boot to GDM show up
T: Avg time of each bochs_primary_plane_helper_atomic_update().
-------------------------------------------------
| | devm_ioremap() | devm_ioremap_wc() |
|------------|----------------|-------------------|
| Cycles | 211.545M | 0.157M |
|------------|----------------|-------------------|
| Cnt | 142 | 1917 |
|------------|----------------|-------------------|
| T | 0.1748s | 0.0004s |
-------------------------------------------------
Note:
Following the rebase to [3], the previously reported GDM failure on the
VGA device [1] can no longer be reproduced, thanks to the memory management
improvements made in [2]. Despite this, I have proceeded to submit this
patch because of the noticeable performance improvements it provides.
Reported-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Closes: https://lore.kernel.org/all/87jzfutmfc.fsf@redhat.com/#t
Cc: Sean Christopherson <seanjc@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Kevin Tian <kevin.tian@intel.com>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Yan Zhao <yan.y.zhao@intel.com>
Link: https://lore.kernel.org/all/87jzfutmfc.fsf@redhat.com/#t [1]
Link: https://patchwork.freedesktop.org/series/138086 [2]
Link: https://gitlab.freedesktop.org/drm/misc/kernel/-/tree/drm-misc-next [3]
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Tested-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20240909131643.28915-1-yan.y.zhao@intel.com
For each mode, test the required memory against the available video
memory. Filters out modes that do not fit into display memory.
Also remove the old test against the 4 MiB limit. It is now obsolete
and did not necessarily produce correct results.
v2:
- fix __udivdi3 linker error (kernel test robot)
- fix vdisplay and hdisplay usage
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240902105546.792625-10-tzimmermann@suse.de
Replace GEM VRAM with GEM SHMEM in bochs. The new memory manager
stores buffer objects in system memory. Makes the driver's memory
management more reliably.
Most of the changes are hidden in external helpers that allocate
buffers. Replacing DRM_GEM_VRAM_DRIVER with DRM_GEM_SHMEM_DRIVER_OPS
swaps these. With GEM VRAM, the video memory was updated directly by
the DRM client. The biggest change within bochs is in atomic_update,
which now updates video memory via memcpy() from the BO in system
memory. Shadow-plane helpers maintaining the pointers to the buffer's
data, so bochs doesn't have to. The update is triggered by each page
flip's call to the framebuffer's dirty helper. The driver supports
damage clipping to minimize memcpy() overhead.
The advantage of GEM SHMEM is that it makes memory management
more reliable. Given DRM's double buffering during page flips, the
minimum amount of video memory is three times the maximum consumption
in some pathological cases. For example, if the maximum size of a GEM
buffer is 1920x1080-32 (i.e., 32-bit FullHD), the buffer size is
8 MiB. Display hardware has to provide at lease 24 MiB to reliably
page flip such configurations. This cannot always be guaranteed and
bochs already contains code to rule out <4 MiB configurations. With
GEM SHMEM, only 8 MiB of video memory are required for the given
example. Unsupported modes can be sorted out easily.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240902105546.792625-9-tzimmermann@suse.de
Remove the simple display pipeline in favor of the regular atomic
helpers in bochs. The simple-pipe helpers are considered deprecated
in DRM.
This effectivly inlines the simple-pipe code for plane and CRTC
support. Instead of a single update helper, there's now a mode-set
helper for the CRTC and an update helper for the plane. The encoder
changes type from NONE ot VIRTUAL.
Removing simple-pipe helpers from bochs will allow for related
cleanups in GEM VRAM helpers.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240902105546.792625-8-tzimmermann@suse.de
Implement a read function for struct drm_edid and read the EDID data
with drm_edit_read_custom(). Update the connector data accordingly.
The EDID data comes from the emulator itself and the connector stores
a copy in its EDID property. The drm_edid field in struct bochs_device
is therefore not required. Remove it.
If qemu provides no EDID data, install default display modes as before.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240902105546.792625-3-tzimmermann@suse.de
Add support for the drm_panic module, which displays a user-friendly
message to the screen when a kernel panic occurs.
v8:
* Replace get_scanout_buffer() with drm_panic_set_buffer()
(Thomas Zimmermann)
v9:
* Revert to using get_scanout_buffer() (Sima)
* move get_scanout_buffer() to plane helper functions (Thomas Zimmermann)
v12:
* Use array for map and pitch in struct drm_scanout_buffer
to support multi-planar format later. (Thomas Zimmermann)
Signed-off-by: Jocelyn Falempe <jfalempe@redhat.com>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20240409163432.352518-7-jfalempe@redhat.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Jyri Sarha <jyri.sarha@iki.fi>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20231102165640.3307820-20-u.kleine-koenig@pengutronix.de
