Re: [PATCH 2/3] misc: fix fallocate commands that need the unshare switch

Re: [PATCH 2/3] misc: fix fallocate commands that need the unshare switch

[Christoph Hellwig]

On Fri, Oct 14, 2016 at 11:13:02AM -0700, Darrick J. Wong wrote:
> We re-added the UNSHARE flag to fallocate, so go make sure that all
> the unshare tests actually check that the xfs_io falloc command
> supports that flag.

We'll need to improve _require_xfs_io_command to actually run the
command for this to work, otherwise this will fail if the fs supports
reflink but not unshared once xfs_io supports -u.

The poster child would be btrfs, and I would have added some output
here if btrfs support in xfstests wasn't completely broken at this
point.

Well, added Ccs and some output anyway in this case..

root@testvm:~/xfstests# mkfs.btrfs -f /dev/vdb
Btrfs v3.17
See http://btrfs.wiki.kernel.org for more information.

Turning ON incompat feature 'extref': increased hardlink limit per file
to 65536
[  243.592803] BTRFS: device fsid 364c6d95-eaae-43c9-b6bd-139e86206519
devid 1 transid 3 /dev/vdb
fs created label (null) on /dev/vdb
	nodesize 16384 leafsize 16384 sectorsize 4096 size 10.00GiB
root@testvm:~/xfstests# ./check -g quick
[  246.101606] BTRFS info (device vdb): disk space caching is enabled
[  246.102260] BTRFS info (device vdb): flagging fs with big metadata
feature
[  246.106123] BTRFS info (device vdb): creating UUID tree
FSTYP         -- btrfs
PLATFORM      -- Linux/x86_64 testvm 4.8.0+
MKFS_OPTIONS  -- /dev/vdb
MOUNT_OPTIONS -- /dev/vdb /mnt/scratch

our local _scratch_mkfs routine ...
Error: /dev/vdb is mounted
check: failed to mkfs $SCRATCH_DEV using specified options
Passed all 0 tests

Re: [PATCH 2/3] misc: fix fallocate commands that need the unshare switch

[Christoph Hellwig]

On Sat, Oct 15, 2016 at 10:03:03AM -0700, Christoph Hellwig wrote:
> The poster child would be btrfs, and I would have added some output
> here if btrfs support in xfstests wasn't completely broken at this
> point.
> 
> Well, added Ccs and some output anyway in this case..

Turns out the btrfs failure was my stupidity, sorry.

I can reproduce the issue I was going to originally show (which was
actually pointed out by Eric for a different fallocate flag check
I wanted to add), here is the diff of the output files when running
generic/156 on btrfs with your patch:

--- tests/generic/156.out	2016-03-29 13:59:30.411720622 +0000
+++ /root/xfstests/results//generic/156.out.bad	2016-10-16 06:15:27.118776421 +0000
@@ -2,8 +2,13 @@
 Create the original file blocks
 Create the reflink copies
 funshare part of a file
+fallocate: Operation not supported
 funshare some of the copies
+fallocate: Operation not supported
+fallocate: Operation not supported
 funshare the rest of the files
+fallocate: Operation not supported
+fallocate: Operation not supported
 Rewrite the original file
 free blocks after reflinking is in range
 free blocks after nocow'ing some copies is in range

So what we really need an enhanced falloc tester that checks that
the tested subcommand is actually implemented on the given file system.
(And we already need something like that for -k on NFS)

Re: [PATCH 2/3] misc: fix fallocate commands that need the unshare switch

[Darrick J. Wong]

On Sat, Oct 15, 2016 at 11:18:37PM -0700, Christoph Hellwig wrote:
> On Sat, Oct 15, 2016 at 10:03:03AM -0700, Christoph Hellwig wrote:
> > The poster child would be btrfs, and I would have added some output
> > here if btrfs support in xfstests wasn't completely broken at this
> > point.
> > 
> > Well, added Ccs and some output anyway in this case..
> 
> Turns out the btrfs failure was my stupidity, sorry.
> 
> I can reproduce the issue I was going to originally show (which was
> actually pointed out by Eric for a different fallocate flag check
> I wanted to add), here is the diff of the output files when running
> generic/156 on btrfs with your patch:

Somehow it totally escaped my notice that I could just do:

_require_xfs_io_command "funshare"

and that would suffice for excluding things like btrfs.  Will
rework.

--D

> 
> --- tests/generic/156.out	2016-03-29 13:59:30.411720622 +0000
> +++ /root/xfstests/results//generic/156.out.bad	2016-10-16 06:15:27.118776421 +0000
> @@ -2,8 +2,13 @@
>  Create the original file blocks
>  Create the reflink copies
>  funshare part of a file
> +fallocate: Operation not supported
>  funshare some of the copies
> +fallocate: Operation not supported
> +fallocate: Operation not supported
>  funshare the rest of the files
> +fallocate: Operation not supported
> +fallocate: Operation not supported
>  Rewrite the original file
>  free blocks after reflinking is in range
>  free blocks after nocow'ing some copies is in range
> 
> So what we really need an enhanced falloc tester that checks that
> the tested subcommand is actually implemented on the given file system.
> (And we already need something like that for -k on NFS)
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html