Re: [PATCH] fstests Fix build errors on kernels before v5.1

On 23/7/26 03:49, Lukas Herbolt wrote:
> On 2026-07-22 11:35, Andrey Albershteyn wrote:
>> On 2026-07-22 07:22:58, Lukas Herbolt wrote:
>>> The FAN_REPORT_FID and fanotify_event_info_fid were introduced
>>> in v5.1. Currently there is no check if the kernel supports it
>>> making xfstests build to fail on older kernels.
>>>
>>> Fixes: 29d120e41c3d ("generic: test fsnotify filesystem error reporting")
>>>
>>> Signed-off-by: Lukas Herbolt <[email protected]>
>>
>> Looks good to me
>> Reviewed-by: Andrey Albershteyn <[email protected]>
>>
>>> ---
>>>  configure.ac          |  1 +
>>>  include/builddefs.in  |  1 +
>>>  m4/package_libcdev.m4 | 14 ++++++++++++++
>>>  src/Makefile          |  6 +++++-
>>>  4 files changed, 21 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/configure.ac b/configure.ac
>>> index 441f543c6089..feb24ddd8854 100644
>>> --- a/configure.ac
>>> +++ b/configure.ac
>>> @@ -74,6 +74,7 @@ AC_NEED_INTERNAL_XFS_IOC_EXCHANGE_RANGE
>>>  AC_HAVE_FICLONE
>>>  AC_HAVE_TRIVIAL_AUTO_VAR_INIT
>>>  AC_HAVE_FILE_GETATTR
>>> +AC_HAVE_FANOTIFY_FID
>>>
>>>  AC_CHECK_FUNCS([renameat2])
>>>  AC_CHECK_FUNCS([reallocarray])
>>> diff --git a/include/builddefs.in b/include/builddefs.in
>>> index 708d75b24d76..8e9e2e3b4113 100644
>>> --- a/include/builddefs.in
>>> +++ b/include/builddefs.in
>>> @@ -75,6 +75,7 @@ HAVE_RLIMIT_NOFILE = @have_rlimit_nofile@
>>>  NEED_INTERNAL_XFS_IOC_EXCHANGE_RANGE = @need_internal_xfs_ioc_exchange_range@
>>>  HAVE_FICLONE = @have_ficlone@
>>>  HAVE_FILE_GETATTR = @have_file_getattr@
>>> +HAVE_FANOTIFY_FID = @have_fanotify_fid@
>>>
>>>  GCCFLAGS = -std=gnu11 -funsigned-char -fno-strict-aliasing -Wall
>>>  SANITIZER_CFLAGS += @autovar_init_cflags@
>>> diff --git a/m4/package_libcdev.m4 b/m4/package_libcdev.m4
>>> index 17f57f427410..340cdad3faf9 100644
>>> --- a/m4/package_libcdev.m4
>>> +++ b/m4/package_libcdev.m4
>>> @@ -102,3 +102,17 @@ AC_DEFUN([AC_HAVE_FILE_GETATTR],
>>>         AC_MSG_RESULT(yes)],[AC_MSG_RESULT(no)])
>>>      AC_SUBST(have_file_getattr)
>>>    ])
>>> +
>>> +# Check if we have fanotify FID support (Linux 5.1+)
>>> +AC_DEFUN([AC_HAVE_FANOTIFY_FID],
>>> +  [ AC_MSG_CHECKING([for fanotify FID support])
>>> +    AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
>>> +#include <sys/fanotify.h>
>>> +    ]], [[
>>> +       int x = FAN_REPORT_FID;
>>> +       struct fanotify_event_info_fid fid;
>>> +       (void)x; (void)fid;
>>> +    ]])],[have_fanotify_fid=yes
>>> +       AC_MSG_RESULT(yes)],[AC_MSG_RESULT(no)])
>>> +    AC_SUBST(have_fanotify_fid)
>>> +  ])
>>> diff --git a/src/Makefile b/src/Makefile
>>> index 31ac43b215e9..847fb168181e 100644
>>> --- a/src/Makefile
>>> +++ b/src/Makefile
>>> @@ -36,7 +36,7 @@ LINUX_TARGETS = xfsctl bstat t_mtab getdevicesize preallo_rw_pattern_reader \
>>>      fscrypt-crypt-util bulkstat_null_ocount splice-test chprojid_fail \
>>>      detached_mounts_propagation ext4_resize t_readdir_3 splice2pipe \
>>>      uuid_ioctl t_snapshot_deleted_subvolume fiemap-fault min_dio_alignment \
>>> -    rw_hint fs-monitor
>>> +    rw_hint
>>>
>>>  EXTRA_EXECS = dmerror fill2attr fill2fs fill2fs_check scaleread.sh \
>>>            btrfs_crc32c_forged_name.py popdir.pl popattr.py \
>>> @@ -107,6 +107,10 @@ ifeq ($(HAVE_FICLONE),yes)
>>>       TARGETS += t_reflink_read_race
>>>  endif
>>>
>>> +ifeq ($(HAVE_FANOTIFY_FID),yes)
>>> +     TARGETS += fs-monitor
>> ...
>>>       LINUX_TARGETS += t_btrfs_received_uuid_ioctl
>>
>> hah just noticed that fstests "supports" different targets (darwin,
>> freebsd), not sure if it works at all
> 
> Ugh! Indeed that should be LINUX_TARGETS += fs-monitor
> I will send v2

I think this was mainly to allow for IRIX back when fstests ran there, the 
last of which were removed by 

bc2568f9 fstests: remove IRIX test program open_unlink

A lot of the more recent programs in TARGETS look like they wouldn't apply 
to anything outside of linux, i.e. t_ext4_dax_journal_corruption and 
t_ext4_dax_inline_corruption.

Given the use of configure to decide what to compile in it may be clearer
if this was just simplified to use TARGETS in the long term.