Some stat-related tests are TBROK on XFS due to invalid mkfs argument

[sandeen]

lstat03 and stat04 both fail when LTP_FS_DEV_TYPE is set to XFS (and presumably other fs types) because they try to run mkfs.xfs with i.e. -b 1024 which is valid for mkfs.ext4, but not for mkfs.xfs (which uses mkfs.xfs -b size=1024) or necessarily any other version of mkfs for different filesystems.

I'd send a patch, but I'm not sure about the philosophy behind some of these tests - lstat03 probably does require a specific block size, as it wants to see a different block size vs tmpfs (?). For stat04, I don't think it matters at all.

Restricting both of these tests to ext4 may be the simplest path forward, as is done in statx05.c and elsewhere?

[metan-ucw]

I guess that we should add a helper function to the test library that would produce the right string per filesystem. Something as char *tst_mkfs_bsize(const char *fs_type, int size) into include/tst_mkfs.h that would prepare the right string per a filesystem (and return NULL if no such support is available). So that we can call it in the test setup() in lstat03.c and stat04.c and adjust the test accordingly. Once that is done we can as well turn on the .all_filesystems flag for the tests so that the tests iterate over all supported filesystems.

The philosophy of these tests is to make sure that the data for the file and link are different to make sure kernel picks up the right information. If the block sizes are the same for both the file and link we cannot detect a breakage if link is followed/not followed correctly. I guess that these tests are still worth to be executed without the blocksize flag on filesystems that does not support setting it since that would still guard against returning wrong values by a mistake. So we should skip passing the blocksize parameter when filesystem does not support it.

As for statx05.c the situation is different there does not seem to be any other filesystem that LTP supports for testing that supports this API.

[sandeen]

re: statx05.c which tests STATX_ATTR_ENCRYPTED, at least f2fs also supports this so I'd expect it to run there as well, but that's a different issue. :)
I was just using it as an example of a restricted test, but perhaps it's the wrong example here if it's incorrectly doing so.