The interrupt mask is enabled before any potential failure points in
the driver, which can leave a failure path where we exit with
interrupts enabled but the device not live. This causes an infinite
stream of interrupts on an Apple M1 Pro laptop on USB-C.
Add a failure label that's used post enabling interrupts, where we
mask them again before returning an error.
Suggested-by: Sven Peter <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
---
drivers/usb/typec/tipd/core.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c
index 6d27a5b5e3ca..7ffcda94d323 100644
--- a/drivers/usb/typec/tipd/core.c
+++ b/drivers/usb/typec/tipd/core.c
@@ -761,12 +761,12 @@ static int tps6598x_probe(struct i2c_client *client)
ret = tps6598x_read32(tps, TPS_REG_STATUS, &status);
if (ret < 0)
- return ret;
+ goto err_clear_mask;
trace_tps6598x_status(status);
ret = tps6598x_read32(tps, TPS_REG_SYSTEM_CONF, &conf);
if (ret < 0)
- return ret;
+ goto err_clear_mask;
/*
* This fwnode has a "compatible" property, but is never populated as a
@@ -855,7 +855,8 @@ static int tps6598x_probe(struct i2c_client *client)
usb_role_switch_put(tps->role_sw);
err_fwnode_put:
fwnode_handle_put(fwnode);
-
+err_clear_mask:
+ tps6598x_write64(tps, TPS_REG_INT_MASK1, 0);
return ret;
}
--
2.34.1
--
Jens Axboe
On Tue, Feb 15, 2022 at 11:22:04AM -0700, Jens Axboe wrote:
> The interrupt mask is enabled before any potential failure points in
> the driver, which can leave a failure path where we exit with
> interrupts enabled but the device not live. This causes an infinite
> stream of interrupts on an Apple M1 Pro laptop on USB-C.
>
> Add a failure label that's used post enabling interrupts, where we
> mask them again before returning an error.
>
> Suggested-by: Sven Peter <[email protected]>
> Signed-off-by: Jens Axboe <[email protected]>
Should this be marked as a fix?
Reviewed-by: Heikki Krogerus <[email protected]>
> ---
> drivers/usb/typec/tipd/core.c | 7 ++++---
> 1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c
> index 6d27a5b5e3ca..7ffcda94d323 100644
> --- a/drivers/usb/typec/tipd/core.c
> +++ b/drivers/usb/typec/tipd/core.c
> @@ -761,12 +761,12 @@ static int tps6598x_probe(struct i2c_client *client)
>
> ret = tps6598x_read32(tps, TPS_REG_STATUS, &status);
> if (ret < 0)
> - return ret;
> + goto err_clear_mask;
> trace_tps6598x_status(status);
>
> ret = tps6598x_read32(tps, TPS_REG_SYSTEM_CONF, &conf);
> if (ret < 0)
> - return ret;
> + goto err_clear_mask;
>
> /*
> * This fwnode has a "compatible" property, but is never populated as a
> @@ -855,7 +855,8 @@ static int tps6598x_probe(struct i2c_client *client)
> usb_role_switch_put(tps->role_sw);
> err_fwnode_put:
> fwnode_handle_put(fwnode);
> -
> +err_clear_mask:
> + tps6598x_write64(tps, TPS_REG_INT_MASK1, 0);
> return ret;
> }
--
heikki
On 2/17/22 1:30 AM, Heikki Krogerus wrote:
> On Tue, Feb 15, 2022 at 11:22:04AM -0700, Jens Axboe wrote:
>> The interrupt mask is enabled before any potential failure points in
>> the driver, which can leave a failure path where we exit with
>> interrupts enabled but the device not live. This causes an infinite
>> stream of interrupts on an Apple M1 Pro laptop on USB-C.
>>
>> Add a failure label that's used post enabling interrupts, where we
>> mask them again before returning an error.
>>
>> Suggested-by: Sven Peter <[email protected]>
>> Signed-off-by: Jens Axboe <[email protected]>
>
> Should this be marked as a fix?
Maybe, I can try and dig out the offending commit. From some quick
checking, it didn't come in with recent fixes, so it's probably been
there since support got added. Maybe we just mark it stable?
> Reviewed-by: Heikki Krogerus <[email protected]>
Thanks! Greg, will you pick this up?
--
Jens Axboe
On Thu, Feb 17, 2022 at 06:51:27AM -0700, Jens Axboe wrote:
> On 2/17/22 1:30 AM, Heikki Krogerus wrote:
> > On Tue, Feb 15, 2022 at 11:22:04AM -0700, Jens Axboe wrote:
> >> The interrupt mask is enabled before any potential failure points in
> >> the driver, which can leave a failure path where we exit with
> >> interrupts enabled but the device not live. This causes an infinite
> >> stream of interrupts on an Apple M1 Pro laptop on USB-C.
> >>
> >> Add a failure label that's used post enabling interrupts, where we
> >> mask them again before returning an error.
> >>
> >> Suggested-by: Sven Peter <[email protected]>
> >> Signed-off-by: Jens Axboe <[email protected]>
> >
> > Should this be marked as a fix?
>
> Maybe, I can try and dig out the offending commit. From some quick
> checking, it didn't come in with recent fixes, so it's probably been
> there since support got added. Maybe we just mark it stable?
>
> > Reviewed-by: Heikki Krogerus <[email protected]>
>
> Thanks! Greg, will you pick this up?
Will do, thanks.
greg k-h