lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 24 Jun 2019 10:05:15 -0700
From:   "Darrick J. Wong" <>
To:     "Theodore Ts'o" <>
Cc:     Christoph Hellwig <>,
        Eryu Guan <>,
        Gabriel Krisman Bertazi <>,,,
        "Lakshmipathi.G" <>
Subject: Re: Removing the shared class of tests

On Mon, Jun 24, 2019 at 09:07:30AM -0400, Theodore Ts'o wrote:
> On Mon, Jun 24, 2019 at 12:16:10AM -0700, Christoph Hellwig wrote:
> > 
> > As for the higher level question?  The shared tests always confused the
> > heck out of me.  generic with the right feature checks seem like a much
> > better idea.
> Agreed.  I've sent out a patch series to bring the number of patches
> in shared down to four.  Here's what's left:
> shared/002 --- needs a feature test to somehow determine whether a
> 	file system supports thousads of xattrs in a file (currently
> 	on btrfs and xfs)

I don't know of a good way to do that other than trying it.

> shared/011 --- needs some way of determining that a file system
> 	supports cgroup-aware writeback (currently enabled only for
> 	ext4 and btrfs).  Do we consider lack of support of
> 	cgroup-aware writeback a bug?  If so, maybe it doesn't need a
> 	feature test at all?

...but for the ones that do, we need a test to make sure the reported
accounting values aren't totally off in the stratosphere.

I wonder, could we add a _require_scratch_cgroupwb that would assign a
new cgroup, try to write a fixed amount of data (~64k) and then _notrun
if the cgroup write back thing reported zero bytes written?

> shared/032 --- needs a feature test to determine whether or not a file
> 	system's mkfs supports detection of "foreign file systems".
> 	e.g., whether or not it warns if you try overwrite a file
> 	system w/o another file system.  (Currently enabled by xfs and
> 	btrfs; it doesn't work for ext[234] because e2fsprogs, because
> 	I didn't want to break existing shell scripts, only warns when
> 	it is used interactively.  We could a way to force it to be
> 	activated it points out this tests is fundamentally testing
> 	implementation choices of the userspace utilities of a file
> 	system.  Does it belong in xfstests?   : ¯\_(ツ)_/¯ )
> shared/289 --- contains ext4, xfs, and btrfs mechanisms for
> 	determining blocks which are unallocated.  Has hard-coded
> 	invocations to dumpe2fs, xfs_db, and /bin/btrfs.

Huh?  shared/289 looks like a pure ext* test to me....

# Copyright (c) 2012 Red Hat, Inc.  All Rights Reserved.
# FS QA Test No. 289
# Test overhead & df output for extN filesystems


> These don't have obvious solutions.  We could maybe add a _notrun if
> adding the thousands of xattrs fails with an ENOSPC or related error
> (f2fs uses something else).
> Maybe we just move shared/011 and move it generic/ w/o a feature test.
> Maybe we remove shared/032 altogether, since for e2fsprogs IMHO
> the right place to put it is the regression test in e2fsprogs --- but
> I know xfs has a different test philosophy for xfsprogs; and tha begs
> the question of what to do for mkfs.btrfs.

<shrug> I'm fine with leaving the test there for xfs since that's where
we put all the xfsprogs tests anyway. :)


> And maybe we just split up shared/289 to three different tests in
> ext4/, xfs/, and btrfs/, since it would make the test script much
> simpler to understand?
> What do people think?
> 						- Ted

Powered by blists - more mailing lists