[fuse] How to trigger writeback?

[fuse] How to trigger writeback?

[Nikolaus Rath]

Hi,

How do I trigger writeback from a (writeback-cache enabled) FUSE
filesystem?

Background: I am emulating the reflink ioctl with a magic xattr, but for
that to work the filesystem needs to have all the data that is supposed
to be in the copied file.


Best,
Nikolaus

-- 
GPG Fingerprint: ED31 791B 2C5C 1613 AF38 8B8A D113 FCAC 3C4E 599F

             »Time flies like an arrow, fruit flies like a Banana.«

Re: [fuse] How to trigger writeback?

[Miklos Szeredi]

On Mon, Sep 24, 2018 at 10:02 AM, Nikolaus Rath <Nikolaus@rath.org> wrote:
> Hi,
>
> How do I trigger writeback from a (writeback-cache enabled) FUSE
> filesystem?

No way, currently (well, triggering fsync(2)/fdatasync(2) on the
filesystem would work, but such recursion is considered highly
hackish).

I'm working on new interfaces that will allow, among other things,
triggering sync from the server.

> Background: I am emulating the reflink ioctl with a magic xattr, but for
> that to work the filesystem needs to have all the data that is supposed
> to be in the copied file.

Have a look at the copyfile patch on the for-next branch of

  git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse.git

Adding reflink support based on that should be pretty simple (and it
takes care of flushing dirty pages).

Thanks,
Miklos

Re: [fuse-devel] [fuse] How to trigger writeback?

[Nikolaus Rath]

On Sep 24 2018, Miklos Szeredi <miklos@szeredi.hu> wrote:
> On Mon, Sep 24, 2018 at 10:02 AM, Nikolaus Rath <Nikolaus@rath.org> wrote:
>> Hi,
>>
>> How do I trigger writeback from a (writeback-cache enabled) FUSE
>> filesystem?
>
> No way, currently (well, triggering fsync(2)/fdatasync(2) on the
> filesystem would work, but such recursion is considered highly
> hackish).


You mean sync(2)/syncfs(2), not fsync(), right? Unless my manpages are
out of date, fsync() only works on a specific inode.


Best,
-Nikolaus

-- 
GPG Fingerprint: ED31 791B 2C5C 1613 AF38 8B8A D113 FCAC 3C4E 599F

             »Time flies like an arrow, fruit flies like a Banana.«

Re: [fuse-devel] [fuse] How to trigger writeback?

[Miklos Szeredi]

On Tue, Sep 25, 2018 at 10:30 AM, Nikolaus Rath <Nikolaus@rath.org> wrote:
> On Sep 24 2018, Miklos Szeredi <miklos@szeredi.hu> wrote:
>> On Mon, Sep 24, 2018 at 10:02 AM, Nikolaus Rath <Nikolaus@rath.org> wrote:
>>> Hi,
>>>
>>> How do I trigger writeback from a (writeback-cache enabled) FUSE
>>> filesystem?
>>
>> No way, currently (well, triggering fsync(2)/fdatasync(2) on the
>> filesystem would work, but such recursion is considered highly
>> hackish).
>
>
> You mean sync(2)/syncfs(2), not fsync(), right? Unless my manpages are
> out of date, fsync() only works on a specific inode.

I would've thought you only need writeback of a specific inode for reflink, no?

Thanks,
Miklos

Re: [fuse-devel] [fuse] How to trigger writeback?

[Nikolaus Rath]

On Sep 25 2018, Miklos Szeredi <miklos@szeredi.hu> wrote:
> On Tue, Sep 25, 2018 at 10:30 AM, Nikolaus Rath <Nikolaus@rath.org> wrote:
>> On Sep 24 2018, Miklos Szeredi <miklos@szeredi.hu> wrote:
>>> On Mon, Sep 24, 2018 at 10:02 AM, Nikolaus Rath <Nikolaus@rath.org> wrote:
>>>> Hi,
>>>>
>>>> How do I trigger writeback from a (writeback-cache enabled) FUSE
>>>> filesystem?
>>>
>>> No way, currently (well, triggering fsync(2)/fdatasync(2) on the
>>> filesystem would work, but such recursion is considered highly
>>> hackish).
>>
>>
>> You mean sync(2)/syncfs(2), not fsync(), right? Unless my manpages are
>> out of date, fsync() only works on a specific inode.
>
> I would've thought you only need writeback of a specific inode for reflink, no?

Aeh, yeah, I did not mention that. Since I couldn't implement real
reflink support but have to go through xattrs, my "reflink" also
supports reflinking entire directory trees. I guess I could still issue
a bunch of fsyncs(), but I think syncfs() is nicer (I can even move it
into the client program that sets the magic xattr so that the filesystem
does not call into itself).


Best,
-Nikolaus

-- 
GPG Fingerprint: ED31 791B 2C5C 1613 AF38 8B8A D113 FCAC 3C4E 599F

             »Time flies like an arrow, fruit flies like a Banana.«