On Wednesday, February 10, 2021 1:36:00 PM CET Rafael J. Wysocki wrote:
> On Wed, Feb 10, 2021 at 12:51 PM Andy Shevchenko
> <andriy.shevche...@linux.intel.com> wrote:
> >
> > We allow to read the single value as a first element in the array.
> > Unfortunately the counting doesn't work in this case and we can't
> > call fwnode_property_count_*() API without getting an error.
> 
> It would be good to mention what the symptom of the issue is here.
> 
> > Modify acpi_data_prop_read() to always try the single value read
> > and thus allow counting the single value as an array of 1 element.
> >
> > Reported-by: Calvin Johnson <calvin.john...@oss.nxp.com>
> > Signed-off-by: Andy Shevchenko <andriy.shevche...@linux.intel.com>
> 
> This is a bug fix, so it should go in before the cleanups in this series IMO.
> 
> Also it looks like stable@vger material.
> 
> > ---
> >  drivers/acpi/property.c | 6 ++++--
> >  1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/acpi/property.c b/drivers/acpi/property.c
> > index 236316ee0e25..d6100585fceb 100644
> > --- a/drivers/acpi/property.c
> > +++ b/drivers/acpi/property.c
> > @@ -913,12 +913,14 @@ static int acpi_data_prop_read(const struct 
> > acpi_device_data *data,
> >         const union acpi_object *items;
> >         int ret;
> >
> > -       if (val && nval == 1) {
> > +       /* Try to read as a single value first */
> > +       if (!val || nval == 1) {
> >                 ret = acpi_data_prop_read_single(data, propname, proptype, 
> > val);
> 
> This returns -EINVAL if val is NULL.
> 
> >                 if (ret >= 0)
> > -                       return ret;
> > +                       return val ? ret : 1;
> 
> So val cannot be NULL here.
> 
> >         }
> >
> > +       /* It's not the single value, get an array instead */
> >         ret = acpi_data_get_property_array(data, propname, ACPI_TYPE_ANY, 
> > &obj);
> >         if (ret)
> >                 return ret;
> > --
> 
> To me, acpi_fwnode_property_read_string_array() needs to special-case
> val == NULL and nval == 0.

Well, scratch this.

Something like the patch below (untested) should be sufficient to address this
if I'm not mistaken.

---
 drivers/acpi/property.c |   17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

Index: linux-pm/drivers/acpi/property.c
===================================================================
--- linux-pm.orig/drivers/acpi/property.c
+++ linux-pm/drivers/acpi/property.c
@@ -787,14 +787,14 @@ static int acpi_data_prop_read_single(co
        const union acpi_object *obj;
        int ret;
 
-       if (!val)
-               return -EINVAL;
-
        if (proptype >= DEV_PROP_U8 && proptype <= DEV_PROP_U64) {
                ret = acpi_data_get_property(data, propname, ACPI_TYPE_INTEGER, 
&obj);
                if (ret)
                        return ret;
 
+               if (!val)
+                       return 1;
+
                switch (proptype) {
                case DEV_PROP_U8:
                        if (obj->integer.value > U8_MAX)
@@ -820,7 +820,8 @@ static int acpi_data_prop_read_single(co
                if (ret)
                        return ret;
 
-               *(char **)val = obj->string.pointer;
+               if (val)
+                       *(char **)val = obj->string.pointer;
 
                return 1;
        } else {
@@ -928,10 +929,16 @@ static int acpi_data_prop_read(const str
        const union acpi_object *items;
        int ret;
 
-       if (val && nval == 1) {
+       if (nval == 1) {
                ret = acpi_data_prop_read_single(data, propname, proptype, val);
                if (ret >= 0)
                        return ret;
+
+               /*
+                * Reading this property as a single-value one failed, but its
+                * value may still be represented as one-element array, so
+                * continue.
+                */
        }
 
        ret = acpi_data_get_property_array(data, propname, ACPI_TYPE_ANY, &obj);




Reply via email to