[RFC] soc: fsl: guts: handle devm_kstrdup() failure

devm_kstrdup() may return NULL if internal allocation failed.
soc_dev_attr.machine  should be checked (although its only use
in pr_info() would be safe even with a NULL). Therefor
in the unlikely case of allocation failure, fsl_guts_probe() returns
-ENOMEM as this allocating failing is an indication of something
more serious going wrong at system level.

As  machine  is from the device tree which I assume to be RO - if
that assumption is always correct - a better alternative would be
to use devm_kstrdup_const() here. That would then simply copy the
reference to the RO data and not perform any allocation at all.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Fixes: a6fc3b698130 ("soc: fsl: add GUTS driver for QorIQ platforms")
---

Problem located by experimental coccinelle script

Patch was compile tested with: multi_v7_defconfig (implies FSL_GUTS=y)

Patch is against 4.20-rc4 (localversion-next is next-20181130)

 drivers/soc/fsl/guts.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

On Sun, Dec 2, 2018 at 3:07 AM Nicholas Mc Guire <hofrat@osadl.org> wrote:
>
> devm_kstrdup() may return NULL if internal allocation failed.
> soc_dev_attr.machine  should be checked (although its only use
> in pr_info() would be safe even with a NULL). Therefor
> in the unlikely case of allocation failure, fsl_guts_probe() returns
> -ENOMEM as this allocating failing is an indication of something
> more serious going wrong at system level.
>
> As  machine  is from the device tree which I assume to be RO - if
> that assumption is always correct - a better alternative would be
> to use devm_kstrdup_const() here. That would then simply copy the
> reference to the RO data and not perform any allocation at all.

I think your assumption is correct.  Do you want to send a new and
better version?  :)

>
> Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> Fixes: a6fc3b698130 ("soc: fsl: add GUTS driver for QorIQ platforms")
> ---
>
> Problem located by experimental coccinelle script
>
> Patch was compile tested with: multi_v7_defconfig (implies FSL_GUTS=y)
>
> Patch is against 4.20-rc4 (localversion-next is next-20181130)
>
>  drivers/soc/fsl/guts.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/soc/fsl/guts.c b/drivers/soc/fsl/guts.c
> index 302e0c8..a0c751b 100644
> --- a/drivers/soc/fsl/guts.c
> +++ b/drivers/soc/fsl/guts.c
> @@ -156,8 +156,11 @@ static int fsl_guts_probe(struct platform_device *pdev)
>         if (of_property_read_string(root, "model", &machine))
>                 of_property_read_string_index(root, "compatible", 0, &machine);
>         of_node_put(root);
> -       if (machine)
> +       if (machine) {
>                 soc_dev_attr.machine = devm_kstrdup(dev, machine, GFP_KERNEL);
> +               if (!soc_dev_attr.machine)
> +                       return -ENOMEM;
> +       }
>
>         svr = fsl_guts_get_svr();
>         soc_die = fsl_soc_die_match(svr, fsl_soc_die);
> --
> 2.1.4
>

On Wed, Dec 05, 2018 at 02:42:55PM -0600, Li Yang wrote:
> On Sun, Dec 2, 2018 at 3:07 AM Nicholas Mc Guire <hofrat@osadl.org> wrote:
> >
> > devm_kstrdup() may return NULL if internal allocation failed.
> > soc_dev_attr.machine  should be checked (although its only use
> > in pr_info() would be safe even with a NULL). Therefor
> > in the unlikely case of allocation failure, fsl_guts_probe() returns
> > -ENOMEM as this allocating failing is an indication of something
> > more serious going wrong at system level.
> >
> > As  machine  is from the device tree which I assume to be RO - if
> > that assumption is always correct - a better alternative would be
> > to use devm_kstrdup_const() here. That would then simply copy the
> > reference to the RO data and not perform any allocation at all.
> 
> I think your assumption is correct.  Do you want to send a new and
> better version?  :)

The issue was actually more general that I did not find a reliable
method to assure that some object is *always* RO. Even for device
tree data it was not clear to me if there could be systems where
it is not RO.
Anyway - will send the version using devm_kstrdup_const() then.

thx!
hofrat

> 
> >
> > Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> > Fixes: a6fc3b698130 ("soc: fsl: add GUTS driver for QorIQ platforms")
> > ---
> >
> > Problem located by experimental coccinelle script
> >
> > Patch was compile tested with: multi_v7_defconfig (implies FSL_GUTS=y)
> >
> > Patch is against 4.20-rc4 (localversion-next is next-20181130)
> >
> >  drivers/soc/fsl/guts.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/soc/fsl/guts.c b/drivers/soc/fsl/guts.c
> > index 302e0c8..a0c751b 100644
> > --- a/drivers/soc/fsl/guts.c
> > +++ b/drivers/soc/fsl/guts.c
> > @@ -156,8 +156,11 @@ static int fsl_guts_probe(struct platform_device *pdev)
> >         if (of_property_read_string(root, "model", &machine))
> >                 of_property_read_string_index(root, "compatible", 0, &machine);
> >         of_node_put(root);
> > -       if (machine)
> > +       if (machine) {
> >                 soc_dev_attr.machine = devm_kstrdup(dev, machine, GFP_KERNEL);
> > +               if (!soc_dev_attr.machine)
> > +                       return -ENOMEM;
> > +       }
> >
> >         svr = fsl_guts_get_svr();
> >         soc_die = fsl_soc_die_match(svr, fsl_soc_die);
> > --
> > 2.1.4
> >