On Thu, Jul 22, 2021 at 09:10:23AM -0700, Daeho Jeong wrote:
> On Thu, Jul 22, 2021 at 7:34 AM Eric Biggers <[email protected]> wrote:
> >
> > On Thu, Jul 22, 2021 at 01:42:05AM -0700, Daeho Jeong wrote:
> > > On Wed, Jul 21, 2021 at 6:56 PM Eric Biggers <[email protected]> wrote:
> > > >
> > > > On Wed, Jul 21, 2021 at 06:40:00PM -0700, Daeho Jeong wrote:
> > > > > On Wed, Jul 21, 2021 at 6:15 PM Eric Biggers <[email protected]> 
> > > > > wrote:
> > > > > >
> > > > > > On Wed, Jul 21, 2021 at 06:04:22PM -0700, Daeho Jeong wrote:
> > > > > > > >
> > > > > > > > How f2fs stores the mapping information doesn't matter.  That's 
> > > > > > > > an
> > > > > > > > implementation detail that shouldn't be exposed to userspace.  
> > > > > > > > The only thing
> > > > > > > > that should be exposed is the actual mapping, and for that it 
> > > > > > > > seems natural to
> > > > > > > > report the physical blocks first.
> > > > > > > >
> > > > > > > > There is no perfect solution for how to handle the remaining 
> > > > > > > > logical blocks,
> > > > > > > > given that the fiemap API was not designed for compressed 
> > > > > > > > files, but I think we
> > > > > > > > should just go with extending the length of the last compressed 
> > > > > > > > extent in the
> > > > > > > > cluster to cover the remaining logical blocks, i.e.:
> > > > > > > >
> > > > > > > >   [0..31]: 2683128..2683159 flag(0x1009) -> merged, encoded, 
> > > > > > > > last_extent
> > > > > > > >
> > > > > > > > That's what btrfs does on compressed files.
> > > > > > > >
> > > > > > > > - Eric
> > > > > > >
> > > > > > > I also agree that that's an implementation detail that shouldn't 
> > > > > > > be
> > > > > > > exposed to userspace.
> > > > > > >
> > > > > > > I want to make it more clear for better appearance.
> > > > > > >
> > > > > > > Do you think we have to remove "unwritten" information below? I 
> > > > > > > also
> > > > > > > think it might be unnecessary information for the user.
> > > > > > > [0..31]: 2683128..2683159 flag(0x1009) -> merged, encoded, 
> > > > > > > last_extent
> > > > > > > (unwritten?)
> > > > > >
> > > > > > FIEMAP_EXTENT_UNWRITTEN already has a specific meaning; see
> > > > > > Documentation/filesystems/fiemap.rst.  It means that the data is 
> > > > > > all zeroes, and
> > > > > > the disk space is preallocated but the data hasn't been written to 
> > > > > > disk yet.
> > > > > >
> > > > > > In this case, the data is *not* necessarily all zeroes.  So I think
> > > > > > FIEMAP_EXTENT_UNWRITTEN shouldn't be used here.
> > > > > >
> > > > > > > Do you want f2fs to print out the info on a cluster basis, even 
> > > > > > > when
> > > > > > > the user asks for one block information?
> > > > > > > Like
> > > > > > > If the user asks for the info of [8..15], f2fs will return the 
> > > > > > > info of [0..31]?
> > > > > >
> > > > > > Yes, since that's how FS_IOC_FIEMAP is supposed to work; see
> > > > > > Documentation/filesystems/fiemap.rst:
> > > > > >
> > > > > >         All offsets and lengths are in bytes and mirror those on 
> > > > > > disk.  It is
> > > > > >         valid for an extents logical offset to start before the 
> > > > > > request or its
> > > > > >         logical length to extend past the request.
> > > > > >
> > > > > > (That being said, the f2fs compression+encryption tests I've 
> > > > > > written don't
> > > > > > exercise this case; they only map the whole file at once.)
> > > > > >
> > > > > > - Eric
> > > > >
> > > > > My last question is.
> > > > > How about a discontinuous cluster like [0..31] maps to discontinuous
> > > > > three blocks like physical address 0x4, 0x14 and 0x24.
> > > > > I think we have to return three extents for the one logical region
> > > > > like the below. What do you think?
> > > > > [0..31] -> 0x4 (merged, encoded)
> > > > > [0..31] -> 0x14 (merged, encoded)
> > > > > [0..31] -> 0x24 (merged, encoded, last_extent)
> > > >
> > > > No, please don't do that.  struct fiemap_extent only has a single 
> > > > length field,
> > > > not separate lengths for fe_logical and fe_physical, so with your 
> > > > proposal there
> > > > would be no way to know how many physical blocks to take from each 
> > > > extent.  It
> > > > would be reporting the same part of the file in contradictory ways.
> > > >
> > > > Like I suggested originally, I think this case should be reported like:
> > > >
> > > >        fe_logical=0    fe_physical=16384  length=4096
> > > >        fe_logical=4096 fe_physical=81920  length=4096
> > > >        fe_logical=8192 fe_physical=147456 length=8192
> > > >
> > >
> > > Hi Eric,
> > >
> > > I think we need to separate one more extent in the example to figure
> > > out that the 4 block cluster turned into 3 compressed blocks.
> > >
> > >        fe_logical=0    fe_physical=16384  length=4096
> > >        fe_logical=4096 fe_physical=81920  length=4096
> > >        fe_logical=8192 fe_physical=147456 length=4096
> > >        fe_logical=12288 fe_physical=0 length=4096
> > >
> > > What do you think?
> > >
> >
> > [+linux-f2fs-devel, not sure why the list was dropped...]
> >
> > I think that would work too, but what flags would you use in the last entry?
> >
> > - Eric
> 
> I think it matches with FIEMAP_EXTENT_UNWRITTEN. Otherwise, we should
> shorten the last extent like below.
> 
>         fe_logical=0    fe_physical=16384  length=4096
>         fe_logical=4096 fe_physical=81920  length=4096
>         fe_logical=8192 fe_physical=147456 length=4096
> 
> 
>   Unwritten extent - the extent is allocated but its data has not been
>   initialized.  This indicates the extent's data will be all zero if read
>   through the filesystem but the contents are undefined if read directly from
>   the device.

Well, as I said before, using UNWRITTEN isn't appropriate because it indicates
that the data is all zeroes, which in this case it is not.  Similarly, reporting
a hole isn't appropriate because it also indicates that the data is all zeroes
and also that it has no space allocated on-disk at all.

I think we should just over-report the physical length of the last extent in the
cluster, which is what btrfs does...

- Eric


_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Reply via email to