iio: trx-rf: ad9088: fix __counted_by misuse in ad9088_clk_register()

clk_hw_onecell_data::hws is a flexible array annotated with
__counted_by(num). Using hws[num++] to append clocks is, in theory,
correct from a sequence point perspective (the post-increment in a
subscript is well-defined in C), but it triggers -Wsequence-point
warnings with recent GCC versions. More importantly, it violates the
__counted_by contract: the array is accessed at index num before the
counter is incremented, triggering runtime UBSAN bounds checks.

Fix this by indexing hws with the source enum directly (which already
indexes into phy->clks[]) and setting clk_data->num once upfront in
ad9088_probe().

Signed-off-by: Nuno Sá <nuno.sa@analog.com>

Author: nunojsa

drivers/iio/trx-rf/ad9088/ad9088.c

@@ -1222,7 +1222,7 @@ static int ad9088_clk_register(struct ad9088_phy *phy, const char *name,
 		return ret;
 
 	phy->clks[source] = clk_priv->hw.clk;
-	phy->clk_data->hws[phy->clk_data->num++] = &clk_priv->hw;
+	phy->clk_data->hws[source] = &clk_priv->hw;
 
 	return 0;
 }
@@ -5215,6 +5215,11 @@ static int ad9088_probe(struct spi_device *spi)
 	if (!phy->clk_data)
 		return -ENOMEM;
 
+	/*
+	 * Even though we already allocate NUM_AD9088_CLKS, RX_SAMPL_CLK_LINK2 is still
+	 * not being registered, hence the -1.
+	 */
+	phy->clk_data->num = NUM_AD9088_CLKS - 1;
 	ret = ad9088_clk_register(phy, "-rx_sampl_clk", __clk_get_name(dev_clk), NULL,
 				  CLK_GET_RATE_NOCACHE | CLK_IGNORE_UNUSED, RX_SAMPL_CLK);
 	if (ret)