Hi Linus,
I've just hit the issue you've described in this PR:
On Sat, Nov 23, 2024 at 05:23:26PM +0100, Linus Walleij wrote:
- The second issue is more sneaky: a recent fixup patch to one
of the rc:s (I think -rc4) fixed some error path bugs in
the AW9523 driver, then a patch to the regular devel is
improving the use of devres so the fixed errorpath fixes
things broken.
I have been applying the following fixup patch for -next to work:
And realized that that proposed fixup:
diff --git a/drivers/pinctrl/pinctrl-aw9523.c b/drivers/pinctrl/pinctrl-aw9523.c
index ebd590a3cec6..90059b0d20e5 100644
--- a/drivers/pinctrl/pinctrl-aw9523.c
+++ b/drivers/pinctrl/pinctrl-aw9523.c
@@ -983,11 +983,8 @@ static int aw9523_probe(struct i2c_client *client)
lockdep_set_subclass(&awi->i2c_lock,
i2c_adapter_depth(client->adapter));
pdesc = devm_kzalloc(dev, sizeof(*pdesc), GFP_KERNEL);
- if (!pdesc) {
- ret = -ENOMEM;
- goto err_disable_vregs;
- }
-
+ if (!pdesc)
+ return -ENOMEM;
ret = aw9523_hw_init(awi);
if (ret)
return ret;
This can be folded in as an "evil merge" or applied separately on
top, your pick.
Is effectively a revert of one of the commits that are part of this PR:
pinctrl: aw9523: add missing mutex_destroy
Would it make more sense to just re-do this PR without the offending
commit? I understand that this is a fairly small fixup, but I'm
concerned that this will just create confusion later on...
--
Thanks,
Sasha
