Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 28 Mar 2023 13:58:09 +1100
From:   Dave Chinner <david@fromorbit.com>
To:     "Darrick J. Wong" <djwong@kernel.org>
Cc:     Christoph Hellwig <hch@infradead.org>,
        "yebin (H)" <yebin10@huawei.com>, Ye Bin <yebin@huaweicloud.com>,
        linux-xfs@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] xfs: fix BUG_ON in xfs_getbmap()
Message-ID: <20230328025809.GC3223426@dread.disaster.area>
References: <20230327140218.4154709-1-yebin@huaweicloud.com>
 <20230327151524.GC16180@frogsfrogsfrogs>
 <64224406.5090106@huawei.com>
 <20230328014328.GG16180@frogsfrogsfrogs>
 <ZCJHSsqk4SJEDOTC@infradead.org>
 <20230328020341.GH16180@frogsfrogsfrogs>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230328020341.GH16180@frogsfrogsfrogs>
Precedence: bulk

On Mon, Mar 27, 2023 at 07:03:41PM -0700, Darrick J. Wong wrote:
> On Mon, Mar 27, 2023 at 06:47:54PM -0700, Christoph Hellwig wrote:
> > On Mon, Mar 27, 2023 at 06:43:28PM -0700, Darrick J. Wong wrote:
> > > <shrug> Seeing as the data fork mappings can change the instant the
> > > ILOCK drops, I'm not /that/ worried about users seeing a delalloc
> > > mapping even if the user requested a flush.  The results are already
> > > obsolete when they get to userspace, unless the application software has
> > > found another means to lock out access to the file.
> > 
> > That is true, but then again the users asked to not see delalloc
> > mappings, so we really shouldn't report one, right?
> 
> Yeah, I suppose so.  I wonder how many programs there are out there that
> don't pass in BMV_IF_DELALLOC /and/ can't handle that?  But I suppose
> taking MMAP_EXCL is good enough to shut up the obvious assertion vector.

Why not just skip it? Take the flush completion as being a
point-in-time snapshot where there are no delalloc extents, and if
any new ones have been created racily, just skip them as being
"after" the flush and so don't get reported...

> The COW implementation probably ought to be doing the flush too.

Yup, and then just skip any delalloc extents found after that, too.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com