Re: [PATCH] scsi: core: Correct wrong kfree() usage for `kobj->name`

[Tzung-Bi Shih <tzungbi@kernel.org>]

On Fri, Jan 16, 2026 at 11:50:15AM +0100, Greg KH wrote:
> On Fri, Jan 16, 2026 at 09:37:07AM +0000, Tzung-Bi Shih wrote:
> > On Fri, Jan 16, 2026 at 10:00:11AM +0100, Greg KH wrote:
> > > On Fri, Jan 16, 2026 at 08:13:59AM +0000, Tzung-Bi Shih wrote:
> > > > `kobj->name` should be freed by kfree_const()[1][2].  Correct it.
> > > > 
> > > > [1] https://elixir.bootlin.com/linux/v6.18/source/lib/kasprintf.c#L41
> > > > [2] https://elixir.bootlin.com/linux/v6.18/source/lib/kobject.c#L695
> > > > 
> > > > Cc: stable@vger.kernel.org
> > > > Fixes: b49493f99690 ("Fix a memory leak in scsi_host_dev_release()")
> > > > Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
> > > > ---
> > > >  drivers/scsi/hosts.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
> > > > index e047747d4ecf..50ec782cf9f4 100644
> > > > --- a/drivers/scsi/hosts.c
> > > > +++ b/drivers/scsi/hosts.c
> > > > @@ -373,7 +373,7 @@ static void scsi_host_dev_release(struct device *dev)
> > > >  		 * name as well as the proc dir structure are leaked.
> > > >  		 */
> > > >  		scsi_proc_hostdir_rm(shost->hostt);
> > > > -		kfree(dev_name(&shost->shost_dev));
> > > > +		kfree_const(dev_name(&shost->shost_dev));
> > > 
> > > Shouldn't the struct device name be freed by the driver core for this
> > > device when it goes out of scope?  Why is it being manually freed here
> > > at all?
> > 
> > Ah, correct.  I think the following patch is what it really needs:
> > 
> > diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
> > index 1b3fbd328277..e3362f445f93 100644
> > --- a/drivers/scsi/hosts.c
> > +++ b/drivers/scsi/hosts.c
> > @@ -373,7 +373,6 @@ static void scsi_host_dev_release(struct device *dev)
> >  		 * name as well as the proc dir structure are leaked.
> >  		 */
> >  		scsi_proc_hostdir_rm(shost->hostt);
> > -		kfree(dev_name(&shost->shost_dev));
> >  	}
> > 
> >  	kfree(shost->shost_data);
> > @@ -548,11 +547,7 @@ struct Scsi_Host *scsi_host_alloc(const struct scsi_host_template *sht, int priv
> >  		goto fail;
> >  	return shost;
> >   fail:
> > -	/*
> > -	 * Host state is still SHOST_CREATED and that is enough to release
> > -	 * ->shost_gendev. scsi_host_dev_release() will free
> > -	 * dev_name(&shost->shost_dev).
> > -	 */
> > +	put_device(&shost->shost_dev);
> >  	put_device(&shost->shost_gendev);
> > 
> >  	return NULL;

The patch doesn't work well.  It can cause an underflow on the reference
count of `&shost->shost_gendev`.  [3] is a more appropriate fix.

> Can you test this to verify that the leak you were seeing is actually
> now handled?

To clarify, the patch wasn't motivated by the leak.  But I can reproduce
the leak by reverting b49493f99690, manual fault injection, rebinding the
driver, and kmemleak.  [3] is tested by the scenario.

[3] https://lore.kernel.org/all/20260117193221.152540-1-tzungbi@kernel.org/