Message ID | 20211031073231.13780-2-lars@metafoo.de (mailing list archive) |
---|---|
State | Accepted |
Headers | show |
Series | [1/2] iio: iio_alloc_device(): Free device correctly on error | expand |
On 10/31/21 9:54 AM, Andy Shevchenko wrote: > > > On Sunday, October 31, 2021, Lars-Peter Clausen <lars@metafoo.de > <mailto:lars@metafoo.de>> wrote: > > Once device_initialize() has been called on a struct device the > device must > be freed by decreasing the reference count rather than directly > freeing the > underlying memory. > > This is so that any additional state and resources associated with the > device get properly freed. > > In this particular case there are no additional resources > associated with > the device and no additional reference count. So there is no > resource leak > or use-after-free by freeing the struct device directly > > But in order to follow best practices and avoid accidental future > breakage > use put_device() instead of kfree() to free the device when an error > occurs. > > Signed-off-by: Lars-Peter Clausen <lars@metafoo.de > <mailto:lars@metafoo.de>> > --- > drivers/iio/industrialio-trigger.c | 12 +++++------- > 1 file changed, 5 insertions(+), 7 deletions(-) > > diff --git a/drivers/iio/industrialio-trigger.c > b/drivers/iio/industrialio-trigger.c > index 93990ff1dfe3..d566e8d4a14b 100644 > --- a/drivers/iio/industrialio-trigger.c > +++ b/drivers/iio/industrialio-trigger.c > @@ -480,7 +480,7 @@ static void iio_trig_release(struct device > *device) > struct iio_trigger *trig = to_iio_trigger(device); > int i; > > - if (trig->subirq_base) { > + if (trig->subirq_base > 0) { > > > > >= ? I don't know. 0 is not supposed to be a valid irq number. And we kzalloc() the struct, so if it hasn't been explicitly initialized we'd get 0. The way the code is at the moment we'd never end up here without calling irq_alloc_descs(), so it is either a valid irq or a negative error code and I can see why you might want to use >= for consistency and symmetry. > > for (i = 0; i < CONFIG_IIO_CONSUMERS_PER_TRIGGER; > i++) { > irq_modify_status(trig->subirq_base + i, > IRQ_NOAUTOEN, > @@ -541,11 +541,11 @@ struct iio_trigger > *viio_trigger_alloc(struct device *parent, > CONFIG_IIO_CONSUMERS_PER_TRIGGER, > 0); > if (trig->subirq_base < 0) > - goto free_trig; > + goto err_put_trig; > > trig->name = kvasprintf(GFP_KERNEL, fmt, vargs); > if (trig->name == NULL) > - goto free_descs; > + goto err_put_trig; > > trig->subirq_chip.name <http://subirq_chip.name> = trig->name; > trig->subirq_chip.irq_mask = &iio_trig_subirqmask; > @@ -559,10 +559,8 @@ struct iio_trigger *viio_trigger_alloc(struct > device *parent, > > return trig; > > -free_descs: > - irq_free_descs(trig->subirq_base, > CONFIG_IIO_CONSUMERS_PER_TRIGGER); > -free_trig: > - kfree(trig); > +err_put_trig: > + put_device(&trig->dev); > return NULL; > } > > -- > 2.20.1 > > > > -- > With Best Regards, > Andy Shevchenko > >
On Sun, Oct 31, 2021 at 11:15 AM Lars-Peter Clausen <lars@metafoo.de> wrote: > On 10/31/21 9:54 AM, Andy Shevchenko wrote: > > On Sunday, October 31, 2021, Lars-Peter Clausen <lars@metafoo.de > > <mailto:lars@metafoo.de>> wrote: ... > > - if (trig->subirq_base) { > > + if (trig->subirq_base > 0) { > > > > >= ? > > I don't know. 0 is not supposed to be a valid irq number. And we > kzalloc() the struct, so if it hasn't been explicitly initialized we'd > get 0. But it will change the behaviour of the code. >=0 is the opposite of replacing < 0. > The way the code is at the moment we'd never end up here without calling > irq_alloc_descs(), so it is either a valid irq or a negative error code > and I can see why you might want to use >= for consistency and symmetry. Right! (But on some architectures and cases 0 might be a valid vIRQ)
On Sun, 31 Oct 2021 15:00:38 +0200 Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > On Sun, Oct 31, 2021 at 11:15 AM Lars-Peter Clausen <lars@metafoo.de> wrote: > > On 10/31/21 9:54 AM, Andy Shevchenko wrote: > > > On Sunday, October 31, 2021, Lars-Peter Clausen <lars@metafoo.de > > > <mailto:lars@metafoo.de>> wrote: > > ... > > > > - if (trig->subirq_base) { > > > + if (trig->subirq_base > 0) { > > > > > > >= ? > > > > I don't know. 0 is not supposed to be a valid irq number. And we > > kzalloc() the struct, so if it hasn't been explicitly initialized we'd > > get 0. > > But it will change the behaviour of the code. > >=0 is the opposite of replacing < 0. > > > > The way the code is at the moment we'd never end up here without calling > > irq_alloc_descs(), so it is either a valid irq or a negative error code > > and I can see why you might want to use >= for consistency and symmetry. > > Right! > > (But on some architectures and cases 0 might be a valid vIRQ) > Given I'm fairly sure this will be after any other irqs we should be fine but I don't think it would be a problem to allow 0. If that's fine with both of you I can just change it to >= 0 whilst applying, or Lars can do a v2 when has time. Thanks, Jonathan
diff --git a/drivers/iio/industrialio-trigger.c b/drivers/iio/industrialio-trigger.c index 93990ff1dfe3..d566e8d4a14b 100644 --- a/drivers/iio/industrialio-trigger.c +++ b/drivers/iio/industrialio-trigger.c @@ -480,7 +480,7 @@ static void iio_trig_release(struct device *device) struct iio_trigger *trig = to_iio_trigger(device); int i; - if (trig->subirq_base) { + if (trig->subirq_base > 0) { for (i = 0; i < CONFIG_IIO_CONSUMERS_PER_TRIGGER; i++) { irq_modify_status(trig->subirq_base + i, IRQ_NOAUTOEN, @@ -541,11 +541,11 @@ struct iio_trigger *viio_trigger_alloc(struct device *parent, CONFIG_IIO_CONSUMERS_PER_TRIGGER, 0); if (trig->subirq_base < 0) - goto free_trig; + goto err_put_trig; trig->name = kvasprintf(GFP_KERNEL, fmt, vargs); if (trig->name == NULL) - goto free_descs; + goto err_put_trig; trig->subirq_chip.name = trig->name; trig->subirq_chip.irq_mask = &iio_trig_subirqmask; @@ -559,10 +559,8 @@ struct iio_trigger *viio_trigger_alloc(struct device *parent, return trig; -free_descs: - irq_free_descs(trig->subirq_base, CONFIG_IIO_CONSUMERS_PER_TRIGGER); -free_trig: - kfree(trig); +err_put_trig: + put_device(&trig->dev); return NULL; }
Once device_initialize() has been called on a struct device the device must be freed by decreasing the reference count rather than directly freeing the underlying memory. This is so that any additional state and resources associated with the device get properly freed. In this particular case there are no additional resources associated with the device and no additional reference count. So there is no resource leak or use-after-free by freeing the struct device directly But in order to follow best practices and avoid accidental future breakage use put_device() instead of kfree() to free the device when an error occurs. Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> --- drivers/iio/industrialio-trigger.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)