linux-kernel - Re: [PATCH] irqdomain: remove debugfs_file from struct irq

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <b5739c15db3d009556abcf9704984dab@kernel.org>
Date:   Thu, 18 Feb 2021 08:47:15 +0000
From:   Marc Zyngier <maz@...nel.org>
To:     Greg KH <gregkh@...uxfoundation.org>
Cc:     Michael Walle <michael@...le.cc>, linux-kernel@...r.kernel.org,
        tglx@...utronix.de
Subject: Re: [PATCH] irqdomain: remove debugfs_file from struct irq_domain

On 2021-02-18 08:38, Greg KH wrote:
> On Thu, Feb 18, 2021 at 09:27:09AM +0100, Michael Walle wrote:
>> Am 2021-02-18 08:31, schrieb Greg KH:
>> > On Wed, Feb 17, 2021 at 09:50:38PM +0000, Marc Zyngier wrote:
>> > > On Wed, 17 Feb 2021 20:10:50 +0000,
>> > > Michael Walle <michael@...le.cc> wrote:
>> > > >
>> > > > Am 2021-02-17 21:02, schrieb Marc Zyngier:
>> > > > > On 2021-02-17 19:57, Michael Walle wrote:
>> > > > >> Hi Greg,
>> > > > >>
>> > > > >>> There's no need to keep around a dentry pointer to a simple file that
>> > > > >>> debugfs itself can look up when we need to remove it from the system.
>> > > > >>> So simplify the code by deleting the variable and cleaning up the
>> > > > >>> logic
>> > > > >>> around the debugfs file.
>> > > > >>
>> > > > >> This will generate the following oops on my board (arm64,
>> > > > >> freescale/fsl-ls1028a-kontron-sl28-var3-ads2.dts). In debugfs_lookup()
>> > > > >> debugfs_mount is NULL.
>> > > > >
>> > > > > That's odd. I gave it a go yesterday, and nothing blew up.
>> > > > > Which makes me wonder whether I had the debug stuff enabled
>> > > > > the first place...
>> > > > >
>> > > > > I've dropped the patch from -next for now until I figure it out
>> > > > > (probably tomorrow).
>> > > >
>> > > > Mh, maybe its my .config, I've attached it. I also noticed that
>> > > > the board boots just fine in our kernel-ci [1].
>> > >
>> > > I reproduced here. I had disabled GENERIC_IRQ_DEBUGFS for obscure
>> > > reasons, and it caught fire as I re-enabled it.
>> > >
>> > > Adding this fixes it for me:
>> > >
>> > > diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c
>> > > index 367ff1c35f75..d8a14cf1a7b6 100644
>> > > --- a/kernel/irq/irqdomain.c
>> > > +++ b/kernel/irq/irqdomain.c
>> > > @@ -1904,7 +1904,8 @@ static void debugfs_add_domain_dir(struct
>> > > irq_domain *d)
>> > >
>> > >  static void debugfs_remove_domain_dir(struct irq_domain *d)
>> > >  {
>> > > -	debugfs_remove(debugfs_lookup(d->name, domain_dir));
>> > > +	if (domain_dir)
>> > > +		debugfs_remove(debugfs_lookup(d->name, domain_dir));
>> > >  }
>> > >
>> > >  void __init irq_domain_debugfs_init(struct dentry *root)
>> > >
>> > >
>> > > Could you please check whether it works for you?
>> >
>> > Can you try this debugfs core change instead?  Callers to debugfs should
>> > not have to do the above type of checking as debugfs should be much more
>> > robust than that.
>> >
>> > thanks,
>> >
>> > greg k-h
>> >
>> >
>> > diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
>> > index 2fcf66473436..5aa798f52b6e 100644
>> > --- a/fs/debugfs/inode.c
>> > +++ b/fs/debugfs/inode.c
>> > @@ -297,7 +297,7 @@ struct dentry *debugfs_lookup(const char *name,
>> > struct dentry *parent)
>> >  {
>> >  	struct dentry *dentry;
>> >
>> > -	if (IS_ERR(parent))
>> > +	if (IS_ERR_OR_NULL(name) || IS_ERR(parent))
>> >  		return NULL;
>> >
>> >  	if (!parent)
>> 
>> This doesn't work. name is not NULL when it is called.
> 
> Ok, but it is a good check we need to make anyway, I'll add it to a
> patch series :)
> 
>> What has to happen before debugfs_lookup() can be called? Looks like
>> someone has to initialize the static debugfs_mount, first.
> 
> Wow, wait, you are removing a debugfs file _before_ debugfs is even
> initialized?  Didn't expect that, ok, let me go try this again...

Yeah, that's a poor man's rename (file being deleted and re-created).

         M.
-- 
Jazz is not dead. It just smells funny...