Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 21 Jan 2022 16:19:40 -0600
From:   Rob Herring <robh@kernel.org>
To:     Peter Geis <pgwipeout@gmail.com>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [BUG] device_property_read_u16 reads out only zero
Message-ID: <YesxfEDbK/qFizFG@robh.at.kernel.org>
References: <CAMdYzYquceSBrOsvO8rW9wmJA_RO=HSwv_waVoS=0hsP414T-A@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAMdYzYquceSBrOsvO8rW9wmJA_RO=HSwv_waVoS=0hsP414T-A@mail.gmail.com>
Precedence: bulk

On Fri, Jan 14, 2022 at 05:48:52PM -0500, Peter Geis wrote:
> Good Evening,
> 
> I seem to have come across a strange bug with drivers/base/property.c
> while expanding the new cyttsp5 driver to handle device-tree
> overrides.
> 
> With the property:
> touchscreen-size-x = <1863>;
> The following code:
> u32 test_u32 = 32; /* canary to catch writing a zero */
> u16 test_u16 = 16; /* canary to catch writing a zero */
> int ret;
> 
> ret = device_property_read_u32(ts->dev, "touchscreen-size-x", &test_u32);
> if(ret)
> dev_err(ts->dev, "read_u32 failed ret: %d\n", ret);
> 
> ret = device_property_read_u16(ts->dev, "touchscreen-size-x", &test_u16);
> if(ret)
> dev_err(ts->dev, "read_u16 failed ret: %d\n", ret);
> 
> dev_err(ts->dev, "read_u32: %d, read_u16: %d\n", test_u32, test_u16);
> 
> returns the following:
> [    1.010876] cyttsp5 5-0024: read_u32: 1863, read_u16: 0
> 
> This was as of 5.16-rc8, using the
> gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu compiler.
> I honestly am at a loss here, any insight you can provide here would
> be appreciated.

The property "touchscreen-size-x" is a u32. Calling 
device_property_read_u16 is an error though one is not returned here. 
You get 0 because that is what the first 2 bytes of the property 
contain. DT data is big-endian.

I suspect making this a hard error would break some users, but we could 
try a WARN.

Rob