[PATCH 00/17] mkfs: sanitise input parameters
Jan Ťulák
jtulak at redhat.com
Fri Jun 19 06:01:49 CDT 2015
Hi all,
I picked up this patch set Dave Chinner wrote back in 2013 to clean
mkfs a bit. It is just part of what I want to do, and still work in progress.
I updated it to the current codebase (for-next branch) and fixed issues
the original set was causing - currently, this set goes through
./check -n quick with the same result, as clean for-next branch.
Two changes against the original set are left in separate patches:
"mkfs: move spinodes crc check" is just a small cosmetic thing that doesn't
fit into the existing patches, but "mkfs fix: handling of files" is a fix
for the last patch from Dave about files. I put this fix standalone for now
to make sure the changes in it are reviewed. As I'm learning about xfs,
I could misunderstood something. :-)
Bellow is the original message from Dave:
> Hi folks,
>␣
> This is still a work in progress, but is compelte enough to get
> feedback on the general structure. The problem being solved here is
> that mkfs does a terrible job of input validation from the command
> line, has huge amounts of repeated code in the sub options
> processing loops and has many, many unnecessary variable for
> tracking simply things like whether a parameter was specified.
>␣
> This patchset introduces a parameter table structure that is used to
> define the parameters and their constraints. Things like minimum and
> maximum valid values, default values, conflicting options, etc are
> all contained within the table, so all the "policy" is found in a
> single place.
>␣
> This greatly reduces the complexity of the option parsing loop. It
> doesn't remove all the complexity (yet) because many of the options
> have special cases or more complex conflicts than I've yet added
> support for. The idea is, however, that all of the sub-option
> parameter setup will eventually end up being implemented as a
> generic loop as the parameter structure will hold all the
> information about in the input parameters.
>␣
> To get there, the parameter table still needs more work - it needs
> to hold the value/string for the parameter, and we need to reference
> those in the code.
>␣
> The flow on effect of this is that we can remove the many, many
> individual variables and start passing the option structures to
> functions rather than avoiding using functions because passing so
> many variables is messy and nasty. IOWs, it lays the groundwork for
> factoring xfs_mkfs.c into something more than a bunch of spagetti...
>␣
> Anyway, have a look and see what you think about progress so far.
>␣
> FWIW, the first patch is following up on the multi-disk discussion
> Christoph and I had, and the last patch in the series covers all the
> issues that arose with "-d file" and treating files like block
> devices....
>␣
> Cheers,
>␣
> Dave.
Cheers,
Jan
Dave Chinner (15):
xfsprogs: use common code for multi-disk detection
mkfs: sanitise ftype parameter values.
mkfs: Sanitise the superblock feature macros
mkfs: validate all input values
mkfs: factor boolean option parsing
mkfs: validate logarithmic parameters sanely
mkfs: structify input parameter passing
mkfs: getbool is redundant
mkfs: use getnum_checked for all ranged parameters
mkfs: add respecification detection to generic parsing
mkfs: table based parsing for converted parameters
mkfs: merge getnum
mkfs: encode conflicts into parsing table
mkfs: add string options to generic parsing
mkfs: don't treat files as though they are block devices
Jan Ťulák (2):
mkfs fix: handling of files
mkfs: move spinodes crc check
include/Makefile | 8 +-
include/xfs_mkfs.h | 73 ++
libxfs/init.c | 7 +
libxfs/linux.c | 11 +-
man/man8/mkfs.xfs.8 | 26 +-
mkfs/Makefile | 2 +-
mkfs/proto.c | 56 +-
mkfs/xfs_mkfs.c | 1873 ++++++++++++++++++++++++++++++---------------------
mkfs/xfs_mkfs.h | 89 ---
repair/xfs_repair.c | 45 +-
10 files changed, 1305 insertions(+), 885 deletions(-)
create mode 100644 include/xfs_mkfs.h
delete mode 100644 mkfs/xfs_mkfs.h
--
2.1.0
More information about the xfs
mailing list