Re: [PATCH v2] proc/sysctl: add shared variables for range check

[Matteo Croce]

On April 17, 2019 12:17:41 PM GMT+09:00, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Wed, 17 Apr 2019 02:59:43 +0200 Matteo Croce <mcroce@xxxxxxxxxx>
> wrote:
> 
> > In the sysctl code the proc_dointvec_minmax() function is often used
> to
> > validate the user supplied value between an allowed range. This
> function
> > uses the extra1 and extra2 members from struct ctl_table as minimum
> and
> > maximum allowed value.
> > 
> > On sysctl handler declaration, in every source file there are some
> readonly
> > variables containing just an integer which address is assigned to
> the
> > extra1 and extra2 members, so the sysctl range is enforced.
> > 
> > The special values 0, 1 and INT_MAX are very often used as range
> boundary,
> > leading duplication of variables like zero=0, one=1, int_max=INT_MAX
> in
> > different source files:
> > 
> >     $ git grep -E '\.extra[12].*&(zero|one|int_max)\b' |wc -l
> >     245
> > 
> > This patch adds three const variables for the most commonly used
> values,
> > and use them instead of creating a local one for every object file.
> > 
> 
> Nice.  A few thoughts:
> 
> > --- a/arch/s390/appldata/appldata_base.c
> > +++ b/arch/s390/appldata/appldata_base.c
> > @@ -220,15 +220,13 @@ appldata_timer_handler(struct ctl_table *ctl,
> int write,
> >  			   void __user *buffer, size_t *lenp, loff_t *ppos)
> >  {
> >  	int timer_active = appldata_timer_active;
> > -	int zero = 0;
> > -	int one = 1;
> >  	int rc;
> >  	struct ctl_table ctl_entry = {
> >  		.procname	= ctl->procname,
> >  		.data		= &timer_active,
> >  		.maxlen		= sizeof(int),
> > -		.extra1		= &zero,
> > -		.extra2		= &one,
> > +		.extra1		= (void *)&sysctl_zero,
> > +		.extra2		= (void *)&sysctl_one,
> 
> The casts are ugly, and by casting away constness they introduce the
> risk that some errant could could change the value of 0, 1 and
> INT_MAX!
> Maybe - perhaps trying to do that would cause a segv but still,
> they're ugly.
> 
> A proper fix would require changing extra1 and extra2 to const void *.
> 
> Perhaps that would be unfeasibly extensive?
> 

Hi Andrew,

I agree that the casts are ugly, but the "casts discards const qualifier" is way more ugly, so I have no choice.

I though about declaring extra1,2 as const, I quickly checked for code which write into these pointers and I found none, but I only looked for one, two and int_max values.

We could do a deeper search to see if other values are safe to turn to const.

> > ...
> >
> > --- a/fs/proc/proc_sysctl.c
> > +++ b/fs/proc/proc_sysctl.c
> > @@ -21,6 +21,11 @@ static const struct inode_operations
> proc_sys_inode_operations;
> >  static const struct file_operations proc_sys_dir_file_operations;
> >  static const struct inode_operations proc_sys_dir_operations;
> >  
> > +/* shared constants to be used in various sysctls */
> > +const int sysctl_zero = 0;
> > +const int sysctl_one = 1;
> > +const int sysctl_int_max = INT_MAX;
> 
> Don't these require EXPORT_SYMBOL()?

Yes, for kernel modules, as the kbuild bot just pointed out.

Regards,
-- 
Matteo Croce
per aspera ad upstream