xfs
[Top] [All Lists]

Re: [PATCH 01/15] xfs: update mount options documentation

To: Eric Sandeen <sandeen@xxxxxxxxxxx>
Subject: Re: [PATCH 01/15] xfs: update mount options documentation
From: Ben Myers <bpm@xxxxxxx>
Date: Fri, 28 Jun 2013 15:46:25 -0500
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <51CCF26C.1030908@xxxxxxxxxxx>
References: <1372313099-8121-1-git-send-email-david@xxxxxxxxxxxxx> <1372313099-8121-2-git-send-email-david@xxxxxxxxxxxxx> <20130627144814.GM20932@xxxxxxx> <20130627190831.GN20932@xxxxxxx> <51CCF26C.1030908@xxxxxxxxxxx>
User-agent: Mutt/1.5.20 (2009-06-14)
Eric,

On Thu, Jun 27, 2013 at 10:18:20PM -0400, Eric Sandeen wrote:
> On 6/27/13 3:08 PM, Ben Myers wrote:
> > Hey Dave,
> > 
> > On Thu, Jun 27, 2013 at 09:48:14AM -0500, Ben Myers wrote:
> >> On Thu, Jun 27, 2013 at 04:04:45PM +1000, Dave Chinner wrote:
> >>> From: Dave Chinner <dchinner@xxxxxxxxxx>
> >>>
> >>> Because it's horribly out of date.
> >>>
> >>> And mark various deprecated options as deprecated and give them a
> >>> removal date.
> >>>
> >>> Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> >>> Reviewed-by: Mark Tinguely <tinguely@xxxxxxx>
> >>
> >> Regarding removal of these mount options and sysctls:  initially these all 
> >> look
> >> pretty reasonable but we need to be very careful here.  I've read some
> >> discussions on lkml that seem to suggest that such interfaces which have 
> >> been
> >> exported to userspace shouldn't be removed at all.  Not that I want to keep
> >> around a bunch of worn out interfaces...
> >>
> >> Applied.
> > 
> > On second thought... Not pushed.
> > 
> > I'm going to hold off on pushing this one to oss for now because I'm just 
> > not
> > comfortable with it yet.  I can pull this in sans the removal notices if you
> > want.  Lets discuss whether the removal of deprecated mount options and 
> > sysctls
> > is acceptable before announcing an intention to remove them.  I'm trending 
> > no,
> > but I can be flexible if this really is ok.
> 
> There's a long history of deprecating mount options - indeed, even entire
> filesystems (smbfs)!  Here's a sampling:
> 
> 5e2a4b25da232a2f4ce264a4b2ae113d0b2a799c btrfs: deprecate subvolrootid mount 
> option
> 1b359204901e182b27aa9da432095cbe2cfd2512 cifs: remove support for deprecated 
> "forcedirectio" and "strictcache" mount options
> 67c1f5295150eb86d065d57b4515a472ecbf008f cifs: add deprecation warning to 
> sockopt=TCP_NODELAY option
> 43829731dd372d04d6706c51052b9dabab9ca356 workqueue: deprecate 
> flush[_delayed]_work_sync()
> 09983b2fab80fa037b1dcf9a11de5a70df59ef7f cifs: add deprecation warnings to 
> strictcache and forcedirectio
> 4d61cd6ec764368689fab3bd19e78d76c1e6b176 cifs: add a deprecation warning to 
> CIFS_IOC_CHECKUMOUNT ioctl
> f70486055ee351158bd6999f3965ad378b52c694 ext4: try to deprecate noacl and 
> noxattr_user mount options
> 87f26807e91102f2526d59c0232bf020479c8d0c ext4: remove deprecation warnings 
> for minix_df and grpid
> 789b4588da40cf572ef982bdc5d590ec1b0386fe cifs: warn about impending 
> deprecation of legacy MultiuserMount code
> 93b8a5854f247138e401471a9c3b82ccb62ff608 xfs: remove the deprecated 
> nodelaylog option
> 8bc4392a1e50f346e97f8777aaefd9cfc3d45c9f cifs: warn about deprecation of 
> /proc/fs/cifs/OplockEnabled interface
> 4113c4caa4f355b8ff8b7ff0510c29c9d00d30b3 ext4: remove deprecated oldalloc
> 242d621964dd8641df53f7d51d4c6ead655cc5a6 xfs: deprecate the nodelaylog mount 
> option
> fbc854027c91fa2813ae7f9de43cc0b5c1119f41 ext3: remove deprecated oldalloc
> be8f684d73d8d916847e996bf69cef14352872c6 oom: make deprecated use of oom_adj 
> more verbose
> 49b28684fdba2c84a3b8e54aaa0faa9ce2e4f140 nfsd: Remove deprecated nfsctl 
> system call and related code.
> 52ca0e84b05595cf74f1ff772b3f9807256b1b27 hugetlbfs: lessen the impact of a 
> deprecation warning
> e52eec13cd6b7f30ab19081b387813e03e592ae5 SYSFS: Allow boot time switching 
> between deprecated and modern sysfs layout
> 1e1405673e4e40a94ed7620553eb440a21040402 nfsd: allow deprecated interface to 
> be compiled out.
> c67874f942e30039442d925b03793e0a46ddcddd nfsd: formally deprecate legacy nfsd 
> syscall interface
> 51b1bd2ace1595b72956224deda349efa880b693 oom: deprecate oom_adj tunable
> a4432345352c2be157ed844603147ac2c82f209c NFSv41: Deprecate 
> nfs_client->cl_minorversion
> 437ca0fda3b442dff9e591581b5e1ffdfec24660 ext4: deprecate obsoleted mount 
> options
> ed58f8027945f1cf415bfe3805e1fa3fe8ed9edf fs/smbfs/inode.c: fix warning 
> message deprecating smbfs
> c773633916c66f8362ca01983d97bd33e35b743f deprecate smbfs in favour of cifs
> 8e9073ed027771bcdee4033eb900a3c09ac90a19 Deprecate a.out ELF interpreters
> e10a4437cb37c85f2df95432025b392d98aac2aa [PATCH] Remove final references to 
> deprecated "MAP_ANON" page protection flag
> f1ddcaf3393b7a3871809b97fae90fac841a1f39 [CRYPTO] api: Remove deprecated 
> interface
> 1ec320afdc9552c92191d5f89fcd1ebe588334ca [PATCH] add process_session() helper 
> routine: deprecate old field

RIP smbfs.

I'm cool with the deprecation of mount options.  What if we were to completely
remove the deprecated option in such a way that mount returns EINVAL when
someone uses it?  If we remove a sysctl that some application expects to be
there, haven't we broken the application?  That seems like it could be an
important issue, but maybe that's not Dave's intent here.

I took a peek at a few of the above.  The ones I saw seem to print a message
'deprecated mount option is being ignored'...  That seems to be the way to go.

> Marking as deprecated, and putting back into play if people REALLY care?
> That's been done too:
> 
> 87f26807e91102f2526d59c0232bf020479c8d0c ext4: remove deprecation warnings 
> for minix_df and grpid
> 
> > I'm thinking of the 3.3 glusterfs and 3.8 pulseaudio reakeage.  And I would
> > really like to have a nice holiday weekend. ;)
> 
> Is "reakage" a freudian slip?  :)

Yeah, there was supposed to be a 'b' on the front. ;)

> I dunno about PA,

See:
https://lkml.org/lkml/2012/12/23/75

That story made /.
        
> but the gluster thing wasn't related to deprecation.

There is more than one way to break userspace, right?  My concern is that the
removal of deprecated options and sysctls could be one of them.  Anyway, I'll
check out the doc Dave pointed to and come back to this.

Thanks,
        Ben

<Prev in Thread] Current Thread [Next in Thread>