Digi
/
meta-digi
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

weston: recover ccimx6/6qp suspend/resume workaround from DEY 3.0 

We initially created this patch when adding XWayland support for the ccimx6 and
ccimx6qp back in dey-3.0-r2.1 (commit eda2f4177e), but we ended up removing
it in dey-4.0-r1.beta1 (commit 526cb3cae0). Even though the blank screen
issue isn't nearly as reproducible as it was back when it was initially
discovered (it used to have a 66% chance of happening), it can still occur
every now and then, especially on the ccimx6 when using an LVDS display.

Adapt the patch to weston 10.0.5.imx and apply it to fix the issue.

https://onedigi.atlassian.net/browse/DEL-7236

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
This commit is contained in:
Gabriel Valcazar 2026-03-23 18:05:12 +01:00
parent 5407d4a206
commit 1ca5a27f51
2 changed files with 93 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

88
meta-digi-dey/dynamic-layers/freescale-layer/recipes-graphics/wayland/weston-10.0.5.imx/0001-libweston-g2d-renderer-try-re-adjusting-fb-if-the-FB.patch Normal file
View File

@ -0,0 +1,88 @@
From: Gabriel Valcazar <gabriel.valcazar@digi.com>
Date: Thu, 17 Sep 2020 13:16:04 +0200
Subject: [PATCH] libweston: g2d-renderer: try re-adjusting fb if the
FBIOPAN_DISPLAY ioctl fails
By default, the g2d renderer works with 3 buffers, and uses the FBIOPAN_DISPLAY
ioctl to have the kernel switch between them every time the output has to be
shown. Because of this, when the renderer is initialized, it will increase the
framebuffer's yres_virtual parameter so it is 3 times the size of yres
(vertical resolution).
However, when the system goes through a suspend/resume iteration, the
framebuffer's yres_virtual parameter will change back to its original value,
causing the FBIOPAN_DISPLAY ioctl to fail when using the 2nd and 3rd buffers,
which is 2 out of 3 times whenever something changes in the output. This has
three direct effects:
* Constant "FBIOPAN_DISPLAY failed" messages in the weston log
* Choppy framerate (due to the renderer only being able to show 1/3 of the
total output frames)
* A 2/3 chance that the desktop will not show after resuming from suspend until
there's movement on the screen
To avoid this, whenever a FBIOPAN_DISPLAY ioctl fails, check if the
yres_virtual attribute read from the kernel is different from the one saved in
the renderer and, if so, update it once again and retry the ioctl. This adds
some additional overhead in case of a FBIOPAN_DISPLAY failure, but these have
only been observed in the specific suspend/resume scenario, not in an average
use case.
The only drawback to this workaround is that there's a 1/3 chance that the
display will go blank for a very small period of time when resuming from
suspend. This is due to the framebuffer being reconfigured and, although
undesireable, is much less bothersome than the original issue.
NOTE FOR DEY 5.0: although this patch's description was accurate back when it
was first introduced in DEY 3.0-r2.1, the reproducibility in DEY 5.0 is much
smaller, but the workaround remains the same.
Upstream-Status: Inappropriate [DEY specific]
https://jira.digi.com/browse/DEL-7236
Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
---
libweston/renderer-g2d/g2d-renderer.c | 23 ++++++++++++++++++++++-
1 file changed, 22 insertions(+), 1 deletion(-)
diff --git a/libweston/renderer-g2d/g2d-renderer.c b/libweston/renderer-g2d/g2d-renderer.c
index 3e2d0791..d390ea17 100644
--- a/libweston/renderer-g2d/g2d-renderer.c
+++ b/libweston/renderer-g2d/g2d-renderer.c
@@ -456,13 +456,34 @@ g2d_flip_surface(struct weston_output *output)
{
int i;
struct g2d_output_state *go = get_output_state(output);
+ struct fb_var_screeninfo aux_varinfo;
go->fb_info.varinfo.yoffset = go->activebuffer * go->fb_info.y_resolution;
if(ioctl(go->fb_info.fb_fd, FBIOPAN_DISPLAY, &(go->fb_info.varinfo)) < 0)
{
- weston_log("FBIOPAN_DISPLAY Failed\n");
+ /* Check if yres_virtual has changed (it happens on suspend/resume) */
+ if (ioctl(go->fb_info.fb_fd, FBIOGET_VSCREENINFO, &aux_varinfo) < 0) {
+ weston_log("FBIOGET_VSCREENINFO Failed\n");
+ goto out;
+ }
+
+ /* If yres_virtual has changed, adjust it and try flipping the surface again */
+ if (aux_varinfo.yres_virtual != go->fb_info.varinfo.yres_virtual) {
+ aux_varinfo.yres_virtual = aux_varinfo.yres * go->nNumBuffers;
+ if (ioctl(go->fb_info.fb_fd, FBIOPUT_VSCREENINFO, &aux_varinfo) < 0) {
+ weston_log("FBIOPUT_VSCREENINFO Failed\n");
+ goto out;
+ }
+
+ if(ioctl(go->fb_info.fb_fd, FBIOPAN_DISPLAY, &(go->fb_info.varinfo)) < 0) {
+ weston_log("FBIOPAN_DISPLAY Failed\n");
+ }
+ } else {
+ weston_log("FBIOPAN_DISPLAY Failed\n");
+ }
}
+out:
for (i = 0; i < go->clone_display_num; i++)
{
go->mirror_fb_info[i].varinfo.yoffset = go->activebuffer * go->mirror_fb_info[i].y_resolution;

5
meta-digi-dey/dynamic-layers/freescale-layer/recipes-graphics/wayland/weston_10.0.5.imx.bbappend
View File

@ -1,3 +1,8 @@
# Copyright (C) 2026, Digi International Inc.
FILESEXTRAPATHS:prepend := "${THISDIR}/${BP}:"
SRCREV = "5223a3c86177709d25f86a96622c0829da955a0e"
SRC_URI += " \
file://0001-libweston-g2d-renderer-try-re-adjusting-fb-if-the-FB.patch \
"