Re: [PATCH 2/9] libxfs: break out the fsop geometry manpage

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Fri, Jun 14, 2019 at 05:00:30PM -0500, Eric Sandeen wrote:
> On 6/7/19 2:29 PM, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> > 
> > Break out the fs geometry ioctl into a separate manpage so that we can
> > document how it works.
> > 
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > ---
> >  man/man2/ioctl_xfs_fsop_geometry.2 |  214 ++++++++++++++++++++++++++++++++++++
> >  man/man3/xfsctl.3                  |   11 +-
> >  2 files changed, 221 insertions(+), 4 deletions(-)
> >  create mode 100644 man/man2/ioctl_xfs_fsop_geometry.2
> > 
> > 
> > diff --git a/man/man2/ioctl_xfs_fsop_geometry.2 b/man/man2/ioctl_xfs_fsop_geometry.2
> > new file mode 100644
> > index 00000000..4045e03b
> > --- /dev/null
> > +++ b/man/man2/ioctl_xfs_fsop_geometry.2
> > @@ -0,0 +1,214 @@
> > +.\" Copyright (c) 2019, Oracle.  All rights reserved.
> > +.\"
> > +.\" %%%LICENSE_START(GPLv2+_DOC_FULL)
> > +.\" SPDX-License-Identifier: GPL-2.0+
> > +.\" %%%LICENSE_END
> > +.TH IOCTL-XFS-FSOP-GEOMETRY 2 2019-04-11 "XFS"
> > +.SH NAME
> > +ioctl_xfs_fsop_geometry \- report XFS filesystem shape
> > +.SH SYNOPSIS
> > +.br
> > +.B #include <xfs/xfs_fs.h>
> > +.PP
> > +.BI "int ioctl(int " fd ", XFS_IOC_FSOP_GEOMETRY, struct xfs_fsop_geometry *" arg );
> > +.PP
> 
> ditto on the .br?
> 
> also xfs_fsop_geometry isn't a thing, is it?  xfs_fsop_geom?
> 
> > +.BI "int ioctl(int " fd ", XFS_IOC_FSOP_GEOMETRY_V1, struct xfs_fsop_geometry_v1 *" arg );
> 
> and xfs_fsop_geom_v1

Fixed.

> > +.SH DESCRIPTION
> > +Report the storage space parameters that influence allocation decisions in
> > +this XFS filesystem.
> > +This information is conveyed in a structure of the following form:
> > +.PP
> > +.in +4n
> > +.nf
> > +struct xfs_fsop_geom {
> > +	__u32         blocksize;
> > +	__u32         rtextsize;
> > +	__u32         agblocks;
> > +	__u32         agcount;
> > +	__u32         logblocks;
> > +	__u32         sectsize;
> > +	__u32         inodesize;
> > +	__u32         imaxpct;
> > +	__u64         datablocks;
> > +	__u64         rtblocks;
> > +	__u64         rtextents;
> > +	__u64         logstart;
> > +	unsigned char uuid[16];
> > +	__u32         sunit;
> > +	__u32         swidth;
> > +	__s32         version;
> > +	__u32         flags;
> > +	__u32         logsectsize;
> > +	__u32         rtsectsize;
> > +	__u32         dirblocksize;
> > +	/* struct xfs_fsop_geom_v1 stops here. */
> > +
> > +	__u32         logsunit;
> > +};
> > +.fi
> > +.in
> > +.PP
> > +.I blocksize
> > +is the size of a fundamental filesystem block, in bytes.
> > +.PP
> > +.I rtextsize
> > +is the size of an extent on the realtime volume, in bytes.
> > +.PP
> > +.I agblocks
> > +is the size of an allocation group, in units of filesystem blocks.
> > +.PP
> > +.I agcount
> > +is the number of allocation groups in the filesystem.
> > +.PP
> > +.I logblocks
> > +is the size of the log, in units of filesystem blocks.
> > +.PP
> > +.I sectsize
> > +is the smallest amount of data that can be written to the data device
> > +atomically, in bytes.
> > +.PP
> > +.I inodesize
> > +is the size of an inode record, in bytes.
> > +.PP
> > +.I imaxpct
> > +is the maximum percentage of the filesystem that can be allocated to inode
> > +record blocks.
> > +.PP
> > +.I datablocks
> > +is the size of the data device, in units of filesystem blocks.
> > +.PP
> > +.I rtblocks
> > +is the size of the realtime device, in units of filesystem blocks.
> > +.PP
> > +.I rtextents
> > +is the number of extents that can be allocated on the realtime device.
> > +This ought to be
> > +.RB "( " rtblocks " * " blocksize " ) / " rtextsize .
> 
> ought to be?  I'm not sure what that means?

Dropped.

> > +.PP
> > +.I logstart
> > +tells the start of the log, in units of filesystem blocks.
> 
> contains?

I'll change it to 'is' here.

> > +If the filesystem has an external log, this will be zero.
> > +.PP
> > +.I uuid
> > +is the universal unique identifier of the filesystem.
> > +.PP
> > +.I sunit
> > +is what the filesystem has been told is the size of a RAID stripe unit on the
> > +underlying data device, in filesystem blocks.
> > +.PP
> > +.I swidth
> > +is what the filesystem has been told is the width of a RAID stripe on the
> > +underlying data device, in units of RAID stripe units.
> > +.PP
> > +.I version
> > +is the version of this structure.
> > +This value will be XFS_FSOP_GEOM_VERSION.
> > +.PP
> > +.I flags
> > +tell us what features are enabled on the filesystem.
> > +This field can be any combination of the following:
> > +.RS 0.4i
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_ATTR
> > +Extended attributes are present.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_NLINK
> > +This filesystem supports up to 2^32 links.
> 
> and if not it supports what?
> 
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_QUOTA
> > +Quotas are enabled.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_IALIGN
> > +Inodes are aligned for better performance.
> 
> aligned to ____ ? (worth trying to explain?)

Aligned to ... inode cluster buffer size, which we've never really
documented outside of "RTFS"? :)

> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_DALIGN
> > +Data blocks are aligned for better performance.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_SHARED
> > +Unused.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_EXTFLG
> > +Filesystem supports unwritten extents.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_DIRV2
> > +Directories maintain free space data for better performance.
> 
> I don't remember all the v1->v2 differences but in other places
> you simply say "contains version 2 format" or whatnot, so that might
> suffice here.  Also can't do v1 since forever, anyway.  ;)

Fixed.

> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_LOGV2
> > +Log uses the V2 format.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_SECTOR
> > +The log device has a sector size larger than 512 bytes.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_ATTR2
> > +Filesystem contains V2 extended attributes.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_PROJID32
> > +Project IDs can be as large as 2^32.
> 
> otherwise 2^16?


yes.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_DIRV2CI
> > +Case-insensitive lookups are supported on directories.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_LAZYSB
> > +On-disk superblock counters are updated only at unmount time.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_V5SB
> > +Metadata blocks are self describing and contain checksums.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_FTYPE
> > +Directories cache inode types in directory entries.
> 
> s/cache/contain/

Ok.

> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_FINOBT
> > +Filesystem maintains an index of free inodes.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_SPINODES
> > +Filesystem tries harder to allocate inodes when free space is fragmented.
> 
> Filesystem may allocate discontinuous inode chunks when free space is fragmented?

Ok.

> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_RMAPBT
> > +Filesystem stores reverse mappings of blocks to owners.
> > +.TP
> > +.B XFS_FSOP_GEOM_FLAGS_REFLINK
> > +Filesystem supports sharing blocks.
> 
> between inodes?

Between files; will fix.

> > +.RE
> > +
> > +.PD 1
> > +.PP
> > +.I logsectsize
> > +is the smallest amount of data that can be written to the log device atomically,
> > +in bytes.
> 
> oh hello, we're back to the structure members again!
> 
> I wonder if it'd be better to move the flag details down below all the structure descriptions.

Yeah, probably.  I'll do that for all the new manpages since otherwise
this becomes a mess.

--D

> > +.PP
> > +.I rtsectsize
> > +is the smallest amount of data that can be written to the realtime device
> > +atomically, in bytes.
> > +.PP
> > +.I dirblocksize
> > +is the size of directory blocks, in bytes.
> > +.PP
> > +.I logsunit
> > +is what the filesystem has been told is the size of a RAID stripe unit on the
> > +underlying log device, in filesystem blocks.
> > +This field is meaningful only if the flag
> > +.B  XFS_FSOP_GEOM_FLAGS_LOGV2
> > +is set.
> > +.SH RETURN VALUE
> > +On error, \-1 is returned, and
> > +.I errno
> > +is set to indicate the error.
> > +.PP
> > +.SH ERRORS
> > +Error codes can be one of, but are not limited to, the following:
> > +.TP
> > +.B EFAULT
> > +The kernel was not able to copy into the userspace buffer.
> > +.TP
> > +.B EFSBADCRC
> > +Metadata checksum validation failed while performing the query.
> > +.TP
> > +.B EFSCORRUPTED
> > +Metadata corruption was encountered while performing the query.
> > +.TP
> > +.B EIO
> > +An I/O error was encountered while performing the query.
> > +.SH CONFORMING TO
> > +This API is specific to XFS filesystem on the Linux kernel.
> > +.SH SEE ALSO
> > +.BR ioctl (2)
> > diff --git a/man/man3/xfsctl.3 b/man/man3/xfsctl.3
> > index 2992b5be..1237eac6 100644
> > --- a/man/man3/xfsctl.3
> > +++ b/man/man3/xfsctl.3
> > @@ -479,6 +479,12 @@ the kernel, except no output count parameter is used (should
> >  be initialized to zero).
> >  An error is returned if the inode number is invalid.
> >  
> > +.TP
> > +.B XFS_IOC_FSGEOMETRY
> > +See
> > +.BR ioctl_xfs_fsop_geometry (2)
> > +for more information.
> > +
> >  .PP
> >  .nf
> >  .B XFS_IOC_THAW
> > @@ -494,10 +500,6 @@ An error is returned if the inode number is invalid.
> >  These interfaces are used to implement various filesystem internal
> >  operations on XFS filesystems.
> >  For
> > -.B XFS_IOC_FSGEOMETRY
> > -(get filesystem mkfs time information), the output structure is of type
> > -.BR struct xfs_fsop_geom .
> > -For
> >  .B XFS_FS_COUNTS
> >  (get filesystem dynamic global information), the output structure is of type
> >  .BR xfs_fsop_counts_t .
> > @@ -506,6 +508,7 @@ as they are not of general use to applications.
> >  
> >  .SH SEE ALSO
> >  .BR ioctl_xfs_fsgetxattr (2),
> > +.BR ioctl_xfs_fsop_geometry (2),
> >  .BR fstatfs (2),
> >  .BR statfs (2),
> >  .BR xfs (5),
> >