Hello,
> Problem is still present in 6.3-rc1 . Bisecting, the first bad commit
> is d6fb6ee1820c .
There is a small timeframe where the (now deleted) struct
acpi_device::driver is different from acpi_device::dev->driver. That's
because the latter is assigned to in drivers/base/dd.c's really_probe()
before the acpi probe function is called, while the former is assigned
only later.
So acpi_drv->ops.add(acpi_dev) is called with the driver set while that
was not true before d6fb6ee1820c.
To test if this is the culprit, can you please try a 5.19 kernel plus
the following patch?:
diff --git a/drivers/acpi/bus.c b/drivers/acpi/bus.c
index e2db1bdd9dd2..657a2520a3e1 100644
--- a/drivers/acpi/bus.c
+++ b/drivers/acpi/bus.c
@@ -1022,6 +1022,8 @@ static int acpi_device_probe(struct device *dev)
struct acpi_driver *acpi_drv = to_acpi_driver(dev->driver);
int ret;
+ acpi_dev->driver = acpi_drv;
+
if (acpi_dev->handler && !acpi_is_pnp_device(acpi_dev))
return -EINVAL;
@@ -1032,8 +1034,6 @@ static int acpi_device_probe(struct device *dev)
if (ret)
return ret;
- acpi_dev->driver = acpi_drv;
-
pr_debug("Driver [%s] successfully bound to device [%s]\n",
acpi_drv->name, acpi_dev->pnp.bus_id);
If the above patch breaks 5.19 that's the trail we have to follow. That
this is timing-related is strange though.
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | https://www.pengutronix.de/ |
Attachment:
signature.asc
Description: PGP signature
