Re: [PATCH 05/10] backlight: qcom-wled: Fix off-by-one maximum with default num_strings

[Daniel Thompson <daniel.thompson@linaro.org>]

On Tue, Oct 05, 2021 at 01:44:35PM +0200, Marijn Suijten wrote:
> On 2021-10-05 11:53:12, Daniel Thompson wrote:
> > On Tue, Oct 05, 2021 at 11:38:43AM +0100, Daniel Thompson wrote:
> > > On Tue, Oct 05, 2021 at 12:06:06PM +0200, Marijn Suijten wrote:
> > > > On 2021-10-05 10:19:47, Daniel Thompson wrote:
> > > > > On Mon, Oct 04, 2021 at 09:27:36PM +0200, Marijn Suijten wrote:
> > > > > > When not specifying num-strings in the DT the default is used, but +1 is
> > > > > > added to it which turns wled3 into 4 and wled4/5 into 5 strings instead
> > > > > > of 3 and 4 respectively, causing out of bounds reads and register
> > > > > > read/writes.  This +1 exists for a deficiency in the DT parsing code,
> > > > > > and is simply omitted entirely - solving this oob issue - by allowing
> > > > > > one extra iteration of the wled_var_cfg function parsing this particular
> > > > > > property.
> > > > > > 
> > > > > > Fixes: 93c64f1ea1e8 ("leds: add Qualcomm PM8941 WLED driver")
> > > > > > Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
> > > > > > Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@somainline.org>
> > > > > > ---
> > > > > >  drivers/video/backlight/qcom-wled.c | 8 +++-----
> > > > > >  1 file changed, 3 insertions(+), 5 deletions(-)
> > > > > > 
> > > > > > diff --git a/drivers/video/backlight/qcom-wled.c b/drivers/video/backlight/qcom-wled.c
> > > > > > index 27e8949c7922..66ce77ee3099 100644
> > > > > > --- a/drivers/video/backlight/qcom-wled.c
> > > > > > +++ b/drivers/video/backlight/qcom-wled.c
> > > > > > @@ -1255,17 +1255,17 @@ static const struct wled_var_cfg wled5_ovp_cfg = {
> > > > > >  
> > > > > >  static u32 wled3_num_strings_values_fn(u32 idx)
> > > > > >  {
> > > > > > -	return idx + 1;
> > > > > > +	return idx;
> > > > > >  }
> > > > > >  
> > > > > >  static const struct wled_var_cfg wled3_num_strings_cfg = {
> > > > > >  	.fn = wled3_num_strings_values_fn,
> > > > > > -	.size = 3,
> > > > > > +	.size = 4, /* [0, 3] */
> > > > > 
> > > > > 0 is not a valid value for this property.
> > > > 
> > > > These comments represent the possible loop iterations the DT "cfg
> > > > parser" runs through, starting at j=0 and running up until and including
> > > > j=3.  Should I make that more clear or omit these comments entirely?
> > > 
> > > The role of wled3_num_strings_values_fn() is to enumerate the list of
> > > legal values for the property [ 1, 2, 3 ]. Your changes cause the
> > > enumeration to include a non-legal value so that you can have an
> > > identity mapping between the symbol and the enumerate value.
> > > 
> > > An alternative approach would be to leave the enumeration logic
> > > alone but set the num_string default to UINT_MAX in all cases:
> > > 
> > > -	cfg->num_strings = cfg->num_strings + 1;
> > > +	if (cfg->num_strings == UINT_MAX)
> > > +		cfg->num_strings = 
> > 
> > Oops... looks like I missed the cfg->max_string_count here.
> > 
> > 
> > > +	else
> > > +               /* Convert from enumerated to numeric form */
> > > +		cfg->num_strings = wled3_num_strings_values_fn(
> > > +						cfg->num_strings);
> > 
> > 
> > PS the alternative option is not to treat num-strings as an enumerated
> >    value at all and just read it directly without using wled_values()...
> 
> I much prefer doing that instead of trying to wrangle enumeration
> parsing around integer values that are supposed to be used as-is.  After
> all this variable is already named to set the `+ 1` override currently,
> and `qcom,enabled_strings` has "custom" handling as well.  I'll extend
> the validation to ensure num_strings>=1 too.

Great.


> In addition, and this needs some investigation on the dt-bindings side
> too, it might be beneficial to make both properties mutually exclusive.
> When specifying qcom,enabled_strings it makes little sense to also
> provide qcom,num_strings and we want the former to take precedence.

If we are designing a "fix" for that then my view is that if both are
passed then num-strings should take precedence because it is an
explicit statement about the number of strings where enabled_strings
is implicit. In other words, if num-strings <= len(enabled_strings) then
we should do what we are told, otherwise report error.


> At that point one might ask why qcom,num_strings remains at all when
> DT can use qcom,enabled_strings instead.  We will supposedly have to
> keep backwards compatibility with DTs in mind so none of this can be
> removed or made mutually exclusive from a driver standpoint, that all
> has to be done in dt-bindings yaml to be enforced on checked-in DTs.

So... perhaps I made a make offering a Reviewed-by: to a patch
that allows len(enabled-strings) to have precedence. If anything
currently uses enabled-strings then it *will* be 4 cells long and
is relying on num-strings to ensure the right things happens ;-) .

We'd like that case to keep working so we must allow num-strings to have
precedence. In other words, when you add the new code, please put it at
the end of the function!


Daniel.


> 
> - Marijn