Re: BUG FIX: [PATCH RFC v3] memstick_check() memleak in kernel 6.1.0+ introduced pre 4.17

[Greg KH]

On Sat, Apr 01, 2023 at 11:18:19AM +0200, Greg KH wrote:
> On Sat, Apr 01, 2023 at 08:33:36AM +0200, Greg KH wrote:
> > On Sat, Apr 01, 2023 at 08:28:07AM +0200, Greg KH wrote:
> > > On Sat, Apr 01, 2023 at 08:23:26AM +0200, Mirsad Goran Todorovac wrote:
> > > > > This patch is implying that anyone who calls "dev_set_name()" also has
> > > > > to do this hack, which shouldn't be the case at all.
> > > > > 
> > > > > thanks,
> > > > > 
> > > > > greg k-h
> > > > 
> > > > This is my best guess. Unless there is dev_free_name() or kobject_free_name(), I don't
> > > > see a more sensible way to patch this up.
> > > 
> > > In sleeping on this, I think this has to move to the driver core.  I
> > > don't understand why we haven't seen this before, except maybe no one
> > > has really noticed before (i.e. we haven't had good leak detection tools
> > > that run with removable devices?)
> > > 
> > > Anyway, let me see if I can come up with something this weekend, give me
> > > a chance...
> > 
> > Wait, no, this already should be handled by the kobject core, look at
> > kobject_cleanup(), at the bottom.  So your change should be merely
> > duplicating the logic there that already runs when the struct device is
> > freed, right?
> > 
> > So I don't understand why your change works, odd.  I need more coffee...
> 
> I think you got half of the change correctly.  This init code is a maze
> of twisty passages, let me take your patch and tweak it a bit into
> something that I think should work.  This looks to be only a memstick
> issue, not a driver core issue (which makes me feel better.)

Oops, forgot the patch.  Can you try this change here and let me know if
that solves the problem or not?  I have compile-tested it only, so I
have no idea if it works.

If this does work, I'll make up a "real" function to replace the
horrible dev.kobj.name mess that a driver would have to do here as it
shouldn't be required that a driver author knows the internals of the
driver core that well...

thanks,

greg k-h

--------------------


diff --git a/drivers/memstick/core/memstick.c b/drivers/memstick/core/memstick.c
index bf7667845459..bbfaf6536903 100644
--- a/drivers/memstick/core/memstick.c
+++ b/drivers/memstick/core/memstick.c
@@ -410,6 +410,7 @@ static struct memstick_dev *memstick_alloc_card(struct memstick_host *host)
 	return card;
 err_out:
 	host->card = old_card;
+	kfree_const(card->dev.kobj.name);
 	kfree(card);
 	return NULL;
 }
@@ -468,8 +469,10 @@ static void memstick_check(struct work_struct *work)
 				put_device(&card->dev);
 				host->card = NULL;
 			}
-		} else
+		} else {
+			kfree_const(card->dev.kobj.name);
 			kfree(card);
+		}
 	}
 
 out_power_off: