在 2014年9月25日，下午3:59，Dave Chinner <david@xxxxxxxxxxxxx> 写道：
> On Wed, Sep 24, 2014 at 07:01:05PM +0200, Jan Kara wrote:
>> On Wed 24-09-14 09:26:34, Christoph Hellwig wrote:
>>> On Wed, Sep 24, 2014 at 06:25:07PM +0200, Jan Kara wrote:
>>>> On Wed 24-09-14 22:04:30, Li Xi wrote:
>>>>> This patch adds ioctl interface for setting/getting project of ext4.
>>>> The patch looks good to me. I was just wondering whether it won't be
>>>> useful to add an ioctl() which isn't ext4 specific. We could just extend
>>>> ->setattr() to allow setting of project ID (most filesystems would just
>>>> return -EOPNOTSUPP but ext4 and xfs could do the right thing) and then call
>>>> ->setattr from the generic ioctl. That way userspace won't have to care
>>>> about filesystem type when setting project ID... What do others think?
>>> Absolutely. In general I also wonder why this patch doesn't implement
>>> the full XFS API. Maybe there is a reason it was considered and
>>> rejected, but it would be helpful to document why.
>> Do you mean full get/setfsxattr API?
> That's a good start.
> The bigger issue in my mind is that we already have a fully featured
> quota API that supports project quotas and userspace tools available
> that manipulate it. xfstests already uses those tools and API
> for testing project quotas.
> This whole patchset reinvents all the quota APIs, and will require
> adding support in userspace, and hence require re-inventing all the
> test infrastructure we already have because it won't be compatible
> with the existing project quota test code.
>> That basically contains project ID,
>> flags (those that are currently get/set with FS_IOC_GETFLAGS/SETFLAGS), and
>> extent size hint right?
> It's a different set of flag definitions. We translate the interface
> XFS_XFLAG_* values to/from the inode on-disk XFS_DIFLAG_* inode values, just
> like we translate the VFS FS_*_FL flags that get passed through the
> FS_IOC_GETFLAGS/SETFLAGS ioctl.
>> That seems workable and it would also make setting
>> of PROJINHERIT flag fs agnostic. Only we would have to create some generic
>> flags namespace and merge into that ext4 flags and have a translation
>> function for the old ext4 flags.
> The XFS_XFLAGS_* are already filesystem agnostic - they are flags
> that are only used for interfacing with userspace and hence only
> exist at the ioctl copy in/out layer.....
>> Also I'm afraid we may quickly run out of
>> 32 available flags in xflags so we'd need to extend that. But all this
>> seems to be doable.
> The struct fsxattr was designed to be extensible - it has unused
> padding and enough space in the flags field to allow us to
> conditionally use that padding….
I was mostly working on the semantics of inherit flag on these patches. And
I didn’t realized that the interface differences would bother you so much. Sorry
I agree that we should choose a good interface. It would be good that it is
general so that it works well for all file systems. I agree that adding an
ext4 specific ioctl() is far from the best choice. I am willing to change it to
any general interface. A general ioctl() sounds good to me. Extend setattr()
and getattr() for project ID sounds even better, since project ID looks like
UID/GID. And general xattr name is another choice. But it is might be a little
bit confusing if we use xattr actually, since we are not saving project ID as
extended attribute on Ext4. Any choice is fine with me, as long as the
implementation won't introduce nasty codes or inconsistent design.
However, the problem is, I do not quite understand why we should keep
the interface exactly the same with XFS. It would be good if we can. But
as far as I can see, it seems hard. XFS uses a lot interfaces which are
not so standard and used by other file systems. For example, struct
fsxattr is not used by other file systems at all except XFS. I am not sure
why we should introduce this into Ext4 if there are a lot of other better
ways. I would be happy to change to XFS interfaces, if it is general.
However, I don’t think it is general enough.
I know xfstest is using the existing project quota interfaces of XFS. And
maybe there are some applications which are using them too. But
keeping the interfaces exactly the same with XFS would cost so much
effort that I’d like to get enough reasons before start working on it. Is it
really necessary? I am not so sure. It is so easy to change user
space applications comparing to changing a weird interfaces. For
example, I think it won’t cost even more than a day to add xfstest
support for new Ext4 project quota. And since project quota is far from
a widely used feature, I don’t think there is much compatibility problems
for existing applications. And If the new project interface are general
enough, there won’t be any compatibility problems for new applications
I might missed something important. So please let me know if you have
- Li Xi