From: Bartosz Golaszewski <[email protected]>
When an interrupt is requested, a procfs directory is created under
"/proc/irq/<irqnum>/<label>" where <label> is the string passed to one of
the request_irq() variants.
What follows is that the string must not contain the "/" character or
the procfs mkdir operation will fail. We don't have such constraints for
GPIO consumer labels which are used verbatim as interrupt labels for
GPIO irqs. We must therefore sanitize the consumer string before
requesting the interrupt.
Let's replace all "/" with "-".
Cc: [email protected]
Reported-by: Stefan Wahren <[email protected]>
Closes: https://lore.kernel.org/linux-gpio/[email protected]/
Signed-off-by: Bartosz Golaszewski <[email protected]>
---
drivers/gpio/gpiolib-cdev.c | 32 +++++++++++++++++++++++++++-----
1 file changed, 27 insertions(+), 5 deletions(-)
diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
index f384fa278764..8b5e8e92cbb5 100644
--- a/drivers/gpio/gpiolib-cdev.c
+++ b/drivers/gpio/gpiolib-cdev.c
@@ -1083,10 +1083,20 @@ static u32 gpio_v2_line_config_debounce_period(struct gpio_v2_line_config *lc,
return 0;
}
+static inline char *make_irq_label(const char *orig)
+{
+ return kstrdup_and_replace(orig, '/', '-', GFP_KERNEL);
+}
+
+static inline void free_irq_label(const char *label)
+{
+ kfree(label);
+}
+
static void edge_detector_stop(struct line *line)
{
if (line->irq) {
- free_irq(line->irq, line);
+ free_irq_label(free_irq(line->irq, line));
line->irq = 0;
}
@@ -1110,6 +1120,7 @@ static int edge_detector_setup(struct line *line,
unsigned long irqflags = 0;
u64 eflags;
int irq, ret;
+ char *label;
eflags = edflags & GPIO_V2_LINE_EDGE_FLAGS;
if (eflags && !kfifo_initialized(&line->req->events)) {
@@ -1146,11 +1157,17 @@ static int edge_detector_setup(struct line *line,
IRQF_TRIGGER_RISING : IRQF_TRIGGER_FALLING;
irqflags |= IRQF_ONESHOT;
+ label = make_irq_label(line->req->label);
+ if (!label)
+ return -ENOMEM;
+
/* Request a thread to read the events */
ret = request_threaded_irq(irq, edge_irq_handler, edge_irq_thread,
- irqflags, line->req->label, line);
- if (ret)
+ irqflags, label, line);
+ if (ret) {
+ free_irq_label(label);
return ret;
+ }
line->irq = irq;
return 0;
@@ -1973,7 +1990,7 @@ static void lineevent_free(struct lineevent_state *le)
blocking_notifier_chain_unregister(&le->gdev->device_notifier,
&le->device_unregistered_nb);
if (le->irq)
- free_irq(le->irq, le);
+ free_irq_label(free_irq(le->irq, le));
if (le->desc)
gpiod_free(le->desc);
kfree(le->label);
@@ -2114,6 +2131,7 @@ static int lineevent_create(struct gpio_device *gdev, void __user *ip)
int fd;
int ret;
int irq, irqflags = 0;
+ char *label;
if (copy_from_user(&eventreq, ip, sizeof(eventreq)))
return -EFAULT;
@@ -2198,12 +2216,16 @@ static int lineevent_create(struct gpio_device *gdev, void __user *ip)
if (ret)
goto out_free_le;
+ label = make_irq_label(le->label);
+ if (!label)
+ goto out_free_le;
+
/* Request a thread to read the events */
ret = request_threaded_irq(irq,
lineevent_irq_handler,
lineevent_irq_thread,
irqflags,
- le->label,
+ label,
le);
if (ret)
goto out_free_le;
--
2.40.1
On Wed, Mar 20, 2024 at 01:59:44PM +0100, Bartosz Golaszewski wrote:
> From: Bartosz Golaszewski <[email protected]>
>
> When an interrupt is requested, a procfs directory is created under
> "/proc/irq/<irqnum>/<label>" where <label> is the string passed to one of
> the request_irq() variants.
>
> What follows is that the string must not contain the "/" character or
> the procfs mkdir operation will fail. We don't have such constraints for
> GPIO consumer labels which are used verbatim as interrupt labels for
> GPIO irqs. We must therefore sanitize the consumer string before
> requesting the interrupt.
>
As previously mentioned, it would be nice for that constraint to be
documented in the irq header to help prevent others falling into the
same trap. Else a short comment in the code here to explain the
need for sanitization.
> Let's replace all "/" with "-".
>
I actually prefer the ":" you originally suggested, as it more clearly
indicates a tier separation, whereas a hyphen is commonly used for
multi-word names. And as the hyphen is more commonly used the sanitized
name is more likely to conflict.
> Cc: [email protected]
> Reported-by: Stefan Wahren <[email protected]>
> Closes: https://lore.kernel.org/linux-gpio/[email protected]/
> Signed-off-by: Bartosz Golaszewski <[email protected]>
> ---
> drivers/gpio/gpiolib-cdev.c | 32 +++++++++++++++++++++++++++-----
> 1 file changed, 27 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
> index f384fa278764..8b5e8e92cbb5 100644
> --- a/drivers/gpio/gpiolib-cdev.c
> +++ b/drivers/gpio/gpiolib-cdev.c
> @@ -1083,10 +1083,20 @@ static u32 gpio_v2_line_config_debounce_period(struct gpio_v2_line_config *lc,
> return 0;
> }
>
> +static inline char *make_irq_label(const char *orig)
> +{
> + return kstrdup_and_replace(orig, '/', '-', GFP_KERNEL);
> +}
> +
> +static inline void free_irq_label(const char *label)
> +{
> + kfree(label);
> +}
> +
> static void edge_detector_stop(struct line *line)
> {
> if (line->irq) {
> - free_irq(line->irq, line);
> + free_irq_label(free_irq(line->irq, line));
> line->irq = 0;
> }
>
> @@ -1110,6 +1120,7 @@ static int edge_detector_setup(struct line *line,
> unsigned long irqflags = 0;
> u64 eflags;
> int irq, ret;
> + char *label;
>
> eflags = edflags & GPIO_V2_LINE_EDGE_FLAGS;
> if (eflags && !kfifo_initialized(&line->req->events)) {
> @@ -1146,11 +1157,17 @@ static int edge_detector_setup(struct line *line,
> IRQF_TRIGGER_RISING : IRQF_TRIGGER_FALLING;
> irqflags |= IRQF_ONESHOT;
>
> + label = make_irq_label(line->req->label);
> + if (!label)
> + return -ENOMEM;
> +
> /* Request a thread to read the events */
> ret = request_threaded_irq(irq, edge_irq_handler, edge_irq_thread,
> - irqflags, line->req->label, line);
> - if (ret)
> + irqflags, label, line);
> + if (ret) {
> + free_irq_label(label);
> return ret;
> + }
>
> line->irq = irq;
> return 0;
> @@ -1973,7 +1990,7 @@ static void lineevent_free(struct lineevent_state *le)
> blocking_notifier_chain_unregister(&le->gdev->device_notifier,
> &le->device_unregistered_nb);
> if (le->irq)
> - free_irq(le->irq, le);
> + free_irq_label(free_irq(le->irq, le));
> if (le->desc)
> gpiod_free(le->desc);
> kfree(le->label);
> @@ -2114,6 +2131,7 @@ static int lineevent_create(struct gpio_device *gdev, void __user *ip)
> int fd;
> int ret;
> int irq, irqflags = 0;
> + char *label;
>
> if (copy_from_user(&eventreq, ip, sizeof(eventreq)))
> return -EFAULT;
> @@ -2198,12 +2216,16 @@ static int lineevent_create(struct gpio_device *gdev, void __user *ip)
> if (ret)
> goto out_free_le;
>
> + label = make_irq_label(le->label);
> + if (!label)
> + goto out_free_le;
> +
Need to set ret = -ENOMEM before the goto, else you will return 0.
Cheers,
Kent.
> /* Request a thread to read the events */
> ret = request_threaded_irq(irq,
> lineevent_irq_handler,
> lineevent_irq_thread,
> irqflags,
> - le->label,
> + label,
> le);
> if (ret)
> goto out_free_le;
> --
> 2.40.1
>
On Fri, Mar 22, 2024 at 08:46:50AM +0100, Bartosz Golaszewski wrote:
> On Fri, Mar 22, 2024 at 2:30 AM Kent Gibson <[email protected]> wrote:
> >
> > On Wed, Mar 20, 2024 at 01:59:44PM +0100, Bartosz Golaszewski wrote:
> > > From: Bartosz Golaszewski <[email protected]>
> > >
> > > Let's replace all "/" with "-".
> > >
> >
> > I actually prefer the ":" you originally suggested, as it more clearly
> > indicates a tier separation, whereas a hyphen is commonly used for
> > multi-word names. And as the hyphen is more commonly used the sanitized
> > name is more likely to conflict.
> >
>
> Sounds good, will do.
> > >
> > > + label = make_irq_label(le->label);
> > > + if (!label)
> > > + goto out_free_le;
> > > +
> >
> > Need to set ret = -ENOMEM before the goto, else you will return 0.
> >
>
> Eek, right, thanks.
Smatch has a warning about this, btw.
drivers/gpio/gpiolib-cdev.c:2221 lineevent_create() warn: missing error code 'ret'
The other warning here is:
drivers/gpio/gpiolib-cdev.c:2269 lineevent_create() warn: 'irq' from request_threaded_irq() not released on lines: 2258.
regards,
dan carpenter
On Fri, Mar 22, 2024 at 2:30 AM Kent Gibson <[email protected]> wrote:
>
> On Wed, Mar 20, 2024 at 01:59:44PM +0100, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <[email protected]>
> >
> > Let's replace all "/" with "-".
> >
>
> I actually prefer the ":" you originally suggested, as it more clearly
> indicates a tier separation, whereas a hyphen is commonly used for
> multi-word names. And as the hyphen is more commonly used the sanitized
> name is more likely to conflict.
>
Sounds good, will do.
> >
> > + label = make_irq_label(le->label);
> > + if (!label)
> > + goto out_free_le;
> > +
>
> Need to set ret = -ENOMEM before the goto, else you will return 0.
>
Eek, right, thanks.
Bart
On Fri, Mar 22, 2024 at 12:31:36PM +0300, Dan Carpenter wrote:
> On Fri, Mar 22, 2024 at 08:46:50AM +0100, Bartosz Golaszewski wrote:
> > On Fri, Mar 22, 2024 at 2:30 AM Kent Gibson <[email protected]> wrote:
> > >
> > > On Wed, Mar 20, 2024 at 01:59:44PM +0100, Bartosz Golaszewski wrote:
> > > > From: Bartosz Golaszewski <[email protected]>
> > > >
> > > > Let's replace all "/" with "-".
> > > >
> > >
> > > I actually prefer the ":" you originally suggested, as it more clearly
> > > indicates a tier separation, whereas a hyphen is commonly used for
> > > multi-word names. And as the hyphen is more commonly used the sanitized
> > > name is more likely to conflict.
> > >
> >
> > Sounds good, will do.
> > > >
> > > > + label = make_irq_label(le->label);
> > > > + if (!label)
> > > > + goto out_free_le;
> > > > +
> > >
> > > Need to set ret = -ENOMEM before the goto, else you will return 0.
> > >
> >
> > Eek, right, thanks.
>
> Smatch has a warning about this, btw.
> drivers/gpio/gpiolib-cdev.c:2221 lineevent_create() warn: missing error code 'ret'
>
And that triggered a "what the hell does that mean" warning in my
wetware error parser ;-).
That could be better worded - it isn't "missing", it hasn't been
appropriately set. So maybe "unset error code"?
> The other warning here is:
> drivers/gpio/gpiolib-cdev.c:2269 lineevent_create() warn: 'irq' from request_threaded_irq() not released on lines: 2258.
>
Looks like a false positive to me - as per the comment in the code, that path
(2258) results in lineevent_release() being called and that releases the irq.
Cheers,
Kent.
On Fri, Mar 22, 2024 at 07:54:19PM +0800, Kent Gibson wrote:
> On Fri, Mar 22, 2024 at 12:31:36PM +0300, Dan Carpenter wrote:
> > On Fri, Mar 22, 2024 at 08:46:50AM +0100, Bartosz Golaszewski wrote:
> > > On Fri, Mar 22, 2024 at 2:30 AM Kent Gibson <[email protected]> wrote:
> > > >
> > > > On Wed, Mar 20, 2024 at 01:59:44PM +0100, Bartosz Golaszewski wrote:
> > > > > From: Bartosz Golaszewski <[email protected]>
> > > > >
> > > > > Let's replace all "/" with "-".
> > > > >
> > > >
> > > > I actually prefer the ":" you originally suggested, as it more clearly
> > > > indicates a tier separation, whereas a hyphen is commonly used for
> > > > multi-word names. And as the hyphen is more commonly used the sanitized
> > > > name is more likely to conflict.
> > > >
> > >
> > > Sounds good, will do.
> > > > >
> > > > > + label = make_irq_label(le->label);
> > > > > + if (!label)
> > > > > + goto out_free_le;
> > > > > +
> > > >
> > > > Need to set ret = -ENOMEM before the goto, else you will return 0.
> > > >
> > >
> > > Eek, right, thanks.
> >
> > Smatch has a warning about this, btw.
> > drivers/gpio/gpiolib-cdev.c:2221 lineevent_create() warn: missing error code 'ret'
> >
>
> And that triggered a "what the hell does that mean" warning in my
> wetware error parser ;-).
>
> That could be better worded - it isn't "missing", it hasn't been
> appropriately set. So maybe "unset error code"?
>
It's kind of a pain to change the warning message after the fact because
then they show up as new warnings for everyone... I maybe should poll
people to see if they care about the hassel of it.
> > The other warning here is:
> > drivers/gpio/gpiolib-cdev.c:2269 lineevent_create() warn: 'irq' from request_threaded_irq() not released on lines: 2258.
> >
>
> Looks like a false positive to me - as per the comment in the code, that path
> (2258) results in lineevent_release() being called and that releases the irq.
Ah right. Thanks!
regards,
dan carpenter