Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 21 Apr 2020 01:04:05 -0700
From:   Christoph Hellwig <hch@infradead.org>
To:     Dave Chinner <david@fromorbit.com>
Cc:     Ritesh Harjani <riteshh@linux.ibm.com>,
        bugzilla-daemon@bugzilla.kernel.org, linux-ext4@vger.kernel.org,
        Theodore Ts'o <tytso@mit.edu>, Jan Kara <jack@suse.cz>,
        "Darrick J. Wong" <darrick.wong@oracle.com>,
        linux-fsdevel@vger.kernel.org
Subject: Re: [Bug 207367] Accraid / aptec / Microsemi / ext4 / larger then
 16TB
Message-ID: <20200421080405.GA4149@infradead.org>
References: <bug-207367-13602@https.bugzilla.kernel.org/>
 <bug-207367-13602-zdl9QZH6DN@https.bugzilla.kernel.org/>
 <20200421042039.BF8074C046@d06av22.portsmouth.uk.ibm.com>
 <20200421050850.GB27860@dread.disaster.area>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200421050850.GB27860@dread.disaster.area>
Sender: linux-ext4-owner@vger.kernel.org
Precedence: bulk

On Tue, Apr 21, 2020 at 03:08:50PM +1000, Dave Chinner wrote:
> > FYI - I do see that bmap() is also used by below APIs/subsystem.
> > Not sure if any of subsystems mentioned below may still fail later
> > if the underlying FS moved to iomap_bmap() interface or for
> > any existing callers of iomap_bmap() :-
> > 
> > 1. mm/page-io.c (generic_swapfile_activate() func)
> 
> Filesystems using iomap infrastructure should be providing
> aops->swap_activate() to map swapfile extents via
> iomap_swapfile_activate() (e.g. see xfs_iomap_swapfile_activate()),
> not using generic_swapfile_activate().

And we also need to eventually phase generic_swapfile_activate out,
maybe by having a version with a get_blocks callback for the non-iomap
case.

> > 4. fs/jbd2/journal.c
> 
> Broken on filesystems where the journal file might be placed beyond
> a 32 bit block number, iomap_bmap() just makes that obvious. Needs
> fixing.

I think this wants to use iomap, as that would solve all the problems.

> And you missed the MD bitmap code uses bmap() to map it's bitmap
> storage file, which means that is broken is the bitmap file is on a
> filesystem/block device > 16TB, too...

This probably needs to use the in-kernel direct I/O interface, just
as it is planned for cachefiles.