On 06/04/2022 05:51, Stephen Boyd wrote:
Quoting Dmitry Baryshkov (2022-04-05 16:45:48)
Add calls to drm_bridge_add()/drm_bridge_remove() for the internal HDMI
bridges. This fixes the following warning.
[ 2.195003] ------------[ cut here ]------------
Usually this line is left out
[ 2.195044] WARNING: CPU: 0 PID: 1 at kernel/locking/mutex.c:579 __mutex_lock+0x840/0x9f4
And the timestamp is removed
Ack for both notes, will fix in v2
[ 2.198774] DEBUG_LOCKS_WARN_ON(lock->magic != lock)
[ 2.198786] Modules linked in:
[ 2.211868] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.18.0-rc1-00002-g3054695a0d27-dirty #55
[ 2.214671] Hardware name: Generic DT based system
[ 2.223265] unwind_backtrace from show_stack+0x10/0x14
[ 2.228036] show_stack from dump_stack_lvl+0x58/0x70
[ 2.233159] dump_stack_lvl from __warn+0xc8/0x1e8
[ 2.238367] __warn from warn_slowpath_fmt+0x78/0xa8
[ 2.243054] warn_slowpath_fmt from __mutex_lock+0x840/0x9f4
[ 2.248174] __mutex_lock from mutex_lock_nested+0x1c/0x24
[ 2.253818] mutex_lock_nested from drm_bridge_hpd_enable+0x2c/0x84
[ 2.259116] drm_bridge_hpd_enable from msm_hdmi_modeset_init+0xc0/0x21c
[ 2.265279] msm_hdmi_modeset_init from mdp4_kms_init+0x53c/0x90c
[ 2.272223] mdp4_kms_init from msm_drm_bind+0x514/0x698
[ 2.278212] msm_drm_bind from try_to_bring_up_aggregate_device+0x160/0x1bc
I'd probably cut it off here.
ack
Is there any Fixes tag for this? Still seems worthwhile to have one even
if this is a lockdep warning.
I thought about this before sending v1, but ended up not doing so. Each
of these changes is not atomic. A call to drm_bridge_add() without final
drm_bridge_remove() in the core msm code would leave dangling pointers
in the drm core. A drm_bridge_remove() is not sensible without
converting _all_ users.
So there are two alternatives:
- leave this patch series as is w/o a Fixes tag
- squash four patches into a single patch and add 'Fixes: a3376e3ec81c
("drm/msm: convert to drm_bridge")' tag
What would you prefer?
--
With best wishes
Dmitry
