Received-SPF: pass (google.com: domain of linux-kernel+bounces-72538-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Tue, 20 Feb 2024 09:22:45 +0100
From: Christoph Hellwig <hch@lst.de>
To: Dave Chinner <david@fromorbit.com>
Cc: John Garry <john.g.garry@oracle.com>, axboe@kernel.dk,
	kbusch@kernel.org, hch@lst.de, sagi@grimberg.me, jejb@linux.ibm.com,
	martin.petersen@oracle.com, djwong@kernel.org,
	viro@zeniv.linux.org.uk, brauner@kernel.org, dchinner@redhat.com,
	jack@suse.cz, linux-block@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-nvme@lists.infradead.org,
	linux-fsdevel@vger.kernel.org, tytso@mit.edu, jbongio@google.com,
	linux-scsi@vger.kernel.org, ojaswin@linux.ibm.com,
	linux-aio@kvack.org, linux-btrfs@vger.kernel.org,
	io-uring@vger.kernel.org, nilay@linux.ibm.com,
	ritesh.list@gmail.com
Subject: Re: [PATCH v4 05/11] block: Add core atomic write support
Message-ID: <20240220082245.GB13785@lst.de>
References: <20240219130109.341523-1-john.g.garry@oracle.com> <20240219130109.341523-6-john.g.garry@oracle.com> <ZdPdHzNAVb5hqlkY@dread.disaster.area>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZdPdHzNAVb5hqlkY@dread.disaster.area>
User-Agent: Mutt/1.5.17 (2007-11-01)

On Tue, Feb 20, 2024 at 09:58:39AM +1100, Dave Chinner wrote:
> > +	lim->atomic_write_hw_max_sectors = 0;
> > +	lim->atomic_write_max_sectors = 0;
> > +	lim->atomic_write_hw_boundary_sectors = 0;
> > +	lim->atomic_write_hw_unit_min_sectors = 0;
> > +	lim->atomic_write_unit_min_sectors = 0;
> > +	lim->atomic_write_hw_unit_max_sectors = 0;
> > +	lim->atomic_write_unit_max_sectors = 0;
> >  }
> 
> Seems to me this function would do better to just
> 
> 	memset(lim, 0, sizeof(*lim));
> 
> and then set all the non-zero fields.

. which the caller already has done :)  In the block tree this
function looks completely different now and relies on the caller
provided zeroing.

> > +void blk_queue_atomic_write_max_bytes(struct request_queue *q,
> > +				      unsigned int bytes)
> > +{
> > +	q->limits.atomic_write_hw_max_sectors = bytes >> SECTOR_SHIFT;
> > +	blk_atomic_writes_update_limits(q);
> > +}
> > +EXPORT_SYMBOL(blk_queue_atomic_write_max_bytes);
> 
> Ok, so this can silently set a limit that is different to what the
> caller asked to have set?
> 
> How is the caller supposed to find this out if the smaller limit
> that was set is not compatible with their configuration?
> 
> i.e. shouldn't this return an error if the requested size cannot
> be set exactly as specified?

That's how the blk limits all work.  The driver provides the hardware
capabilities for a given value, and the block layer ensures it
works with other limits imposed by the block layer or other parts
of the device limits.