Re: [PATCH v4 2/2] btrfs-progs: cmds: Add subcommand that dumps file extents

[Sidong Yang <realwakka@gmail.com>]

On Fri, Aug 20, 2021 at 06:16:18AM +0800, Qu Wenruo wrote:
> 
> 
> On 2021/8/19 下午11:27, Sidong Yang wrote:
> > On Thu, Aug 19, 2021 at 02:05:52PM +0800, Qu Wenruo wrote:
> > > 
> > > 
> > > On 2021/8/19 下午2:03, Qu Wenruo wrote:
> > > > 
> > > > 
> > > > On 2021/8/18 上午8:38, Sidong Yang wrote:
> > > > > On Tue, Aug 17, 2021 at 03:30:22PM +0200, David Sterba wrote:
> > > > > > On Sun, Jul 18, 2021 at 06:46:01AM +0000, Sidong Yang wrote:
> > > > > > > This patch adds an subcommand in inspect-internal. It dumps file
> > > > > > > extents of
> > > > > > > the file that user provided. It helps to show the internal information
> > > > > > > about file extents comprise the file.
> > > > > > 
> > > > > > Do you have an example of the output? That's the most interesting part.
> > > > > > Thanks.
> > > > > 
> > > > > Thanks for reply.
> > > > > This is an example of the output below.
> > > > > 
> > > > > # ./btrfs inspect-internal dump-file-extent /mnt/test1
> > > > > type = regular, start = 2097152, len = 3227648, disk_bytenr = 0,
> > > > > disk_num_bytes = 0, offset = 0, compression = none
> > > > > type = regular, start = 5324800, len = 16728064, disk_bytenr = 0,
> > > > > disk_num_bytes = 0, offset = 0, compression = none
> > > > > type = regular, start = 22052864, len = 8486912, disk_bytenr = 0,
> > > > > disk_num_bytes = 0, offset = 0, compression = none
> > > > > type = regular, start = 30572544, len = 36540416, disk_bytenr = 0,
> > > > > disk_num_bytes = 0, offset = 0, compression = none
> > > > > type = regular, start = 67112960, len = 5299630080, disk_bytenr = 0,
> > > > > disk_num_bytes = 0, offset = 0, compression = none
> > > > 
> > > > Could you give an example which includes both real (non-hole) extents
> > > > and real extents (better to include regular, compressed, preallocated
> > > > and inline).
> > > 
> > > Tons of typos... I mean to include both holes (like the existing
> > > example) and non-holes extents...
> > 
> > Sorry, I had no idea about holes. But I found some test code in
> > xfstests. It helpes me to make hole in file.
> > 
> > xfs_io -c "fpunch 96K 32K" /mnt/a/foobar
> > xfs_io -c "fpunch 64K 128K" /mnt/a/foobar
> > 
> > and the example is below.
> > 
> > # ./btrfs inspect dump-file-extent /mnt/a/foobar
> > type = regular, start = 0, len = 98304, disk_bytenr = 21651456,
> > disk_num_bytes = 4096, offset = 0, compression = zstd
> > type = regular, start = 98304, len = 32768, disk_bytenr = 0,
> > disk_num_bytes = 0, offset = 0, compression = none
> > 
> > I'm afaid that I understand your request correctly. Is it what you want?
> 
> This example is much better.
> 
> But still, for holes, things like
> disk_bytenr/disk_num_bytes/offset/compression makes no sense and can be
> skipped.

Okay. If disk_bytenr is zero, other information after this would be
skipped in next version.
> 
> Furthermore, for hole/prealloc they need extra type other than "regular"

For now, It already has prealloc type. but no hole type. All the types
are inline/prealloc/regular/hole? If so, disk_bytenr should be checked
before print it's type.

Thanks,
Sidong
> 
> Thanks,
> Qu
> > > 
> > > > 
> > > > Currently the output only contains holes, and for holes, a lot of
> > > > members makes no sense, like disk_bytenr/disk_num_bytes/offset (even it
> > > > can be non-zero) and compression.
> > > > 
> > > > Thanks,
> > > > Qu
> > > > 
> > > > > 
> > > > > Thanks,
> > > > > Sidong
> > > > >