On Thu, 20 Feb 2025 14:54:53 -0600
David Lechner [off-list ref] wrote:

On 2/10/25 4:33 PM, David Lechner wrote:

quoted

Replace bitmap array access with bitmap_write.

Accessing the bitmap array directly is not recommended and now there is
a helper function that can be used.

Reviewed-by: Linus Walleij <redacted>
Signed-off-by: David Lechner <dlechner@baylibre.com>
---
 drivers/iio/resolver/ad2s1210.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/iio/resolver/ad2s1210.c b/drivers/iio/resolver/ad2s1210.c
index 7f18df790157f1e411fb70de193a49f0677c999f..04879e6d538bce664469c5f6759d8b1cedea16e9 100644
--- a/drivers/iio/resolver/ad2s1210.c
+++ b/drivers/iio/resolver/ad2s1210.c

@@ -46,6 +46,7 @@
  */
 
 #include <linux/bitfield.h>
+#include <linux/bitmap.h>
 #include <linux/bits.h>
 #include <linux/cleanup.h>
 #include <linux/clk.h>

@@ -180,7 +181,7 @@ static int ad2s1210_set_mode(struct ad2s1210_state *st, enum ad2s1210_mode mode)
 	if (!gpios)
 		return mode == st->fixed_mode ? 0 : -EOPNOTSUPP;
 
-	bitmap[0] = mode;
+	bitmap_write(bitmap, mode, 0, 2);
 
 	return gpiod_multi_set_value_cansleep(gpios, bitmap);
 }

@@ -1470,7 +1471,7 @@ static int ad2s1210_setup_gpios(struct ad2s1210_state *st)
 			return dev_err_probe(dev, -EINVAL,
 				      "requires exactly 2 resolution-gpios\n");
 
-		bitmap[0] = st->resolution;
+		bitmap_write(bitmap, st->resolution, 0, 2);
 
 		ret = gpiod_multi_set_value_cansleep(resolution_gpios, bitmap);
 		if (ret < 0)
  

There is actually a bug here pointed out in a similar patch. bitmap_write()
only modifies the bitmap, so this introduces an unintialized use bug. [1]
Here, we only use the bits that we set, so runtime behavior would not actually
be buggy but still best to fully initialize the memory.

I'm a bit surprised that my local compiler and iio/testing both didn't catch that
since GCC 14 caught it in the other driver.

[1]: https://lore.kernel.org/linux-gpio/20250217132152.29d86d6c@jic23-huawei/T/#m3163d2c5db5b7376504d8ad6f23716f1119de761 (local)

The fix is simple, we can zero-initialize the bitmap.

Please send this as a fix patch on top as I'd rather not unwind my tree
for just this and the patch is already pushed out on what is mostly a non
rebasing branch.