On Mon, Sep 15, 2025 at 4:40 PM Eric Biggers <[email protected]> wrote:
>
> On Mon, Sep 15, 2025 at 12:04:51PM +0200, Jan Prusakowski wrote:
> > common/encrypt: Do not run _verify_ciphertext_for_encryption_policy on
> > compressed FS
>
> A better title would be something like
> "common/encrypt: Explicitly set the test file to uncompressed".
Agreed, will fix that in v3.
> > @@ -790,6 +790,13 @@ _do_verify_ciphertext_for_encryption_policy()
> > _set_encpolicy $dir $keyspec $set_encpolicy_args -f $policy_flags
> > for src in $tmp.testfile_*; do
> > dst=$dir/${src##*.}
> > + # To make sure the test file is not compressed we create an
> > empty one
> > + # and disable compression first (F2FS won't allow resetting
> > the
> > + # compression flag if the file has data already in it).
> > + touch $dst
> > + if lsattr $dst | grep -qE ".+c.+ $dst" ; then
> > + chattr -c $dst
> > + fi
> > cp $src $dst
> > inode=$(stat -c %i $dst)
> > blocklist=$(_get_ciphertext_block_list $dst)
>
> Is adding 'm' (FS_NOCOMP_FL) needed too? If not, why does it exist?
In my setup files created have FS_COMPR_FL set from the start. Just clearing
FS_COMPR_FL appears to help as now all the tests using
_verify_ciphertext_for_encryption_policy
pass when I run them on f2fs with "-o compress_extension=*".
Do you think we should add 'm' (FS_NOCOMP_FL) as well just in case some other FS
behaves differently? Do you have any ideas on what other FS I should check?
> Also, have you verified that the tests that use this function still pass
> on both ext4 and f2fs?
Yes, there are no regressions on ext4 and f2fs. I checked generic/369,
generic/548, generic/549, generic/550, generic/582, generic/583, generic/584,
generic/592, generic/602, generic/693 and generic/739.
All now pass except generic/369 which is skipped due to no support for
hardware-wrapped inline encryption keys).
Kind regards,
Jan
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
