[PATCH 1/2] drm/client: Clear cmdline modes before the 1024x768 clone pass

[Hugo Sales]

drm_client_target_cloned() first tries to clone with each connector's
cmdline mode. That pass fills modes[] as it goes and stops at the first
enabled connector without a cmdline mode, or ends with modes that don't
match. Either way the earlier slots keep their cmdline modes

The 1024x768 pass that follows only writes a slot when the connector has
a matching mode, and relies on an empty slot to spot a connector it
can't clone. A connector with a mode from video= and no 1024x768 mode
keeps its cmdline mode instead, so the function returns true with modes
that differ. drm_client_target_preferred() then never runs, and on a
single-CRTC device drm_client_pick_crtcs() leaves one of the connectors
without a CRTC

Free the modes from the cmdline pass before looking for 1024x768

Fixes: 1d42bbc8f7f9 ("drm/fbdev: fix cloning on fbcon")
Signed-off-by: Hugo Sales <hugo@xxxxxxx>
---
Sashiko flagged this while reviewing my fbdev probe series [1]. The bug
predates that series, so I'm sending the fix on its own. I haven't hit
it on hardware. The KUnit test in the next patch fails without this fix
and passes with it

DRM KUnit on UML passes 672/672, also with KASAN and with kmemleak.
kmemleak reports six objects leaked by drm_bridge tests, which it also
reports on the base without this series

[1] https://lore.kernel.org/all/20261024165355.2782145-1-hugo@xxxxxxx/
---
 drivers/gpu/drm/drm_client_modeset.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/gpu/drm/drm_client_modeset.c b/drivers/gpu/drm/drm_client_modeset.c
index 0080a8e9556a5..5e45beea66438 100644
--- a/drivers/gpu/drm/drm_client_modeset.c
+++ b/drivers/gpu/drm/drm_client_modeset.c
@@ -343,6 +343,8 @@ static bool drm_client_target_cloned(struct drm_device *dev,
 		return true;
 	}
 
+	modes_destroy(dev, modes, connector_count);
+
 	/* try and find a 1024x768 mode on each connector */
 	can_clone = true;
 	dmt_mode = drm_mode_find_dmt(dev, 1024, 768, 60, false);
-- 
2.56.0