Re: [PATCH] drm/i915: Use %p for pointer formatting

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

On 2026-09-18 14:00:15 [+0200], Sebastian Brzezinka wrote:
> Hi Sebastian,

Sebastian :)

> > --- a/drivers/gpu/drm/i915/i915_debugfs.c
> > +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> > @@ -179,7 +179,7 @@ i915_debugfs_describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
> >  	struct i915_vma *vma;
> >  	int pin_count = 0;
> >  
> > -	seq_printf(m, "%pK: %c%c%c %8zdKiB %02x %02x %s%s%s",
> > +	seq_printf(m, "%p: %c%c%c %8zdKiB %02x %02x %s%s%s",
> Please, correct me if I'm wrong, but i915_debugfs_describe_obj() 
> prints into a debugfs file read by userspace exactly the case where
> Documentation/core-api/printk-formats.rst recommends %pK over %p.

I should probably get rid of this, too.
So we do have %p and %pK by now and both return hashed pointer unless
overridden. K can have a security policy which can only work in a user
context. There is a different override switch which does not make things
easier. The vast majority of users are gone.

> >If (and only if) you are printing addresses as a content of a virtual file in
> >e.g. procfs or sysfs (using e.g. seq_printf(), not printk()) read by a
> >userspace process, use the %pK modifier described below instead of %p or %px.
> 
> And it was addedby 2563a4524feb, which explicitly wants to
> guard it with kptr_restrict(i mean that it's intentionally stronger).

Please look at the timeline:
pre v3.9-rc4, %p prints the bare pointer
v3.9-rc4  2563a4524febe ("drm/i915: restrict kernel address leak in debugfs")

(%p -> %pK the leak in i915/debugfs is closed, %p still prints the bare
pointer).

v4.15-rc2 ad67b74d2469d ("printk: hash addresses printed with %p")

now %p does not leak the pointer, too. 

Sebastian