of: of_property_read_string return -ENODATA when !length

When the length of the string is zero of_property_read_string should
return -ENODATA according to the description of the function.

However, of_property_read_string doesn't check prop->length. If
prop->length is zero, return -ENODATA.

Without this patch the following command in u-boot:

fdt set /chosen/node property-name

results in of_property_read_string returning -EILSEQ when attempting to
read property-name. With this patch, it returns -ENODATA as expected.

Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220416003028.1315268-1-sstabellini@kernel.org

diff --git a/drivers/of/property.c b/drivers/of/property.c
index 8e90071..84903da 100644 (file)
--- a/drivers/of/property.c
+++ b/drivers/of/property.c
@@ -431,6 +431,9 @@ EXPORT_SYMBOL_GPL(of_property_read_variable_u64_array);
  * property does not have a value, and -EILSEQ if the string is not
  * null-terminated within the length of the property data.
  *
+ * Note that the empty string "" has length of 1, thus -ENODATA cannot
+ * be interpreted as an empty string.
+ *
  * The out_string pointer is modified only if a valid string can be decoded.
  */
 int of_property_read_string(const struct device_node *np, const char *propname,
@@ -439,7 +442,7 @@ int of_property_read_string(const struct device_node *np, const char *propname,
        const struct property *prop = of_find_property(np, propname, NULL);
        if (!prop)
                return -EINVAL;
-       if (!prop->value)
+       if (!prop->length)
                return -ENODATA;
        if (strnlen(prop->value, prop->length) >= prop->length)
                return -EILSEQ;