linux-kernel - Re: [PATCH v2 3/4] pstore/blk: Include zone in pstore_device

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 for Android: free password hash cracker in your pocket

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

Message-ID: <202106160737.B0B8882B@keescook>
Date:   Wed, 16 Jun 2021 07:41:52 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Christoph Hellwig <hch@....de>
Cc:     linux-kernel@...r.kernel.org, Al Viro <viro@...iv.linux.org.uk>,
        gmpy.liaowx@...il.com, Anton Vorontsov <anton@...msg.org>,
        Colin Cross <ccross@...roid.com>,
        Tony Luck <tony.luck@...el.com>,
        Jonathan Corbet <corbet@....net>,
        Miquel Raynal <miquel.raynal@...tlin.com>,
        Richard Weinberger <richard@....at>,
        Vignesh Raghavendra <vigneshr@...com>,
        linux-doc@...r.kernel.org, linux-mtd@...ts.infradead.org,
        linux-block@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH v2 3/4] pstore/blk: Include zone in pstore_device_info

On Wed, Jun 16, 2021 at 06:02:47AM +0200, Christoph Hellwig wrote:
> > +#define verify_size(name, alignsize, enabled) {				\
> > +		long _##name_;						\
> > +		if (enabled)						\
> > +			_##name_ = check_size(name, alignsize);		\
> > +		else							\
> > +			_##name_ = 0;					\
> > +		/* synchronize visible module parameters to result. */	\
> > +		name = _##name_ / 1024;					\
> > +		dev->zone.name = _##name_;				\
> > +	}
> 
> The formatting here looks weird between the two-tab indent and the
> opening brace on the macro definition line.

I can adjust that, sure.

> 
> > -	if (!dev || !dev->total_size || !dev->read || !dev->write) {
> > +	if (!dev || !dev->zone.total_size || !dev->zone.read || !dev->zone.write) {
> >  		if (!dev)
> > -			pr_err("NULL device info\n");
> > +			pr_err("NULL pstore_device_info\n");
> >  		else {
> > -			if (!dev->total_size)
> > +			if (!dev->zone.total_size)
> >  				pr_err("zero sized device\n");
> > -			if (!dev->read)
> > +			if (!dev->zone.read)
> >  				pr_err("no read handler for device\n");
> > -			if (!dev->write)
> > +			if (!dev->zone.write)
> >  				pr_err("no write handler for device\n");
> >  		}
> 
> This still looks odd to me.  Why not the somewhat more verbose but
> much more obvious:
> 
> 	if (!dev) {
> 		pr_err("NULL pstore_device_info\n");
> 		return -EINVAL;
> 	}
> 	if (!dev->zone.total_size) {
> 		pr_err("zero sized device\n");
> 		return -EINVAL;
> 	}
> 	...

Will do.

> > -	dev.total_size = i_size_read(I_BDEV(psblk_file->f_mapping->host)->bd_inode);
> > +	dev->zone.total_size = i_size_read(I_BDEV(psblk_file->f_mapping->host)->bd_inode);
> 
> This is starting to be unreadable long.  A local variable for the inode
> might be nice, as that can also be used in the ISBLK check above.

Fair enough; will change.

> > +	if (!pstore_device_info && best_effort && blkdev[0]) {
> > +		struct pstore_device_info *best_effort_dev;
> > +
> > +		best_effort_dev = kzalloc(sizeof(*best_effort_dev), GFP_KERNEL);
> > +		if (!best_effort) {
> > +			ret = -ENOMEM;
> > +			goto unlock;
> > +		}
> > +		best_effort_dev->zone.read = psblk_generic_blk_read;
> > +		best_effort_dev->zone.write = psblk_generic_blk_write;
> > +
> > +		ret = __register_pstore_blk(best_effort_dev,
> > +					    early_boot_devpath(blkdev));
> > +		if (ret)
> > +			kfree(best_effort_dev);
> > +		else
> > +			pr_info("attached %s (%zu) (no dedicated panic_write!)\n",
> > +				blkdev, best_effort_dev->zone.total_size);
> 
> Maybe split this into a little helper?
> 
> > +	/* Unregister and free the best_effort device. */
> > +	if (psblk_file) {
> > +		struct pstore_device_info *dev = pstore_device_info;
> > +
> > +		__unregister_pstore_device(dev);
> > +		kfree(dev);
> > +		fput(psblk_file);
> > +		psblk_file = NULL;
> >  	}
> 
> Same.

I guess? I don't feel strongly one way or another.

> 
> > +	/* If we've been asked to unload, unregister any registered device. */
> > +	if (pstore_device_info)
> > +		__unregister_pstore_device(pstore_device_info);
> 
> Won't this double unregister pstore_device_info?

No, __unregister_pstore_device() will NULL pstore_device_info.

> 
> >  struct pstore_device_info {
> > -	unsigned long total_size;
> >  	unsigned int flags;
> > -	pstore_zone_read_op read;
> > -	pstore_zone_write_op write;
> > -	pstore_zone_erase_op erase;
> > -	pstore_zone_write_op panic_write;
> > +	struct pstore_zone_info zone;
> >  };
> 
> Given that flags is only used inside of __register_pstore_device
> why not kill this struct and just pass it explicitly?

Because of the mess pstore's internal APIs used to be. :) It's likely
other things will get added here in the future, and I don't want to
have to repeat the kind of argument passing games that used to exist in
this code.

-- 
Kees Cook