regulator: dummy: force synchronous probing
authorChristian Eggers <ceggers@arri.de>
Tue, 11 Mar 2025 09:18:02 +0000 (10:18 +0100)
committerMark Brown <broonie@kernel.org>
Tue, 11 Mar 2025 13:11:11 +0000 (13:11 +0000)
commit8619909b38eeebd3e60910158d7d68441fc954e9
treec5b445899e7f18d9a5b338a9e6eed80c37ab1b93
parent1c81a8c78ae653f3a21cde0f37a91f1b22b7d2fb
regulator: dummy: force synchronous probing

Sometimes I get a NULL pointer dereference at boot time in kobject_get()
with the following call stack:

anatop_regulator_probe()
 devm_regulator_register()
  regulator_register()
   regulator_resolve_supply()
    kobject_get()

By placing some extra BUG_ON() statements I could verify that this is
raised because probing of the 'dummy' regulator driver is not completed
('dummy_regulator_rdev' is still NULL).

In the JTAG debugger I can see that dummy_regulator_probe() and
anatop_regulator_probe() can be run by different kernel threads
(kworker/u4:*).  I haven't further investigated whether this can be
changed or if there are other possibilities to force synchronization
between these two probe routines.  On the other hand I don't expect much
boot time penalty by probing the 'dummy' regulator synchronously.

Cc: stable@vger.kernel.org
Fixes: 259b93b21a9f ("regulator: Set PROBE_PREFER_ASYNCHRONOUS for drivers that existed in 4.14")
Signed-off-by: Christian Eggers <ceggers@arri.de>
Link: https://patch.msgid.link/20250311091803.31026-1-ceggers@arri.de
Signed-off-by: Mark Brown <broonie@kernel.org>
drivers/regulator/dummy.c