Re: [mm/readahead] a0b99df1aa: xfstests.xfs.421.fail

[Matthew Wilcox]

On Mon, Feb 21, 2022 at 04:02:18PM +0800, kernel test robot wrote:
> commit: a0b99df1aa37d714eb80be5fb54efd56c88a3336 ("mm/readahead: Add large folio readahead")

> xfs/420	- output mismatch (see /lkp/benchmarks/xfstests/results//xfs/420.out.bad)
>     --- tests/xfs/420.out	2022-02-17 11:55:00.000000000 +0000
>     +++ /lkp/benchmarks/xfstests/results//xfs/420.out.bad	2022-02-20 20:34:22.430378506 +0000
>     @@ -13,9 +13,7 @@
>      Seek holes and data in file2
>      Whence	Result
>      DATA	0
>     -HOLE	131072
>     -DATA	196608
>     -HOLE	262144
>     +HOLE	524288

Confirm this test now fails.  I don't think it's actually a bug,
though.  I think the test is now using larger pages to cache the
file, and it fails to report that there's a hole in the file.
Maybe there actually isn't a hole in the file any more; using
larger pages to cache the file means we'll now write more data
than we used to.

Adding XFS people for their thoughts.

Complete output:

$ diff -u ../ktest/tests/xfstests/tests/xfs/420.out ktest-out/xfstests/xfs/420.out.bad
--- ../ktest/tests/xfstests/tests/xfs/420.out	2021-07-05 15:49:45.539887305 -0400
+++ ktest-out/xfstests/xfs/420.out.bad	2022-02-21 08:14:40.000000000 -0500
@@ -13,9 +13,7 @@
 Seek holes and data in file2
 Whence	Result
 DATA	0
-HOLE	131072
-DATA	196608
-HOLE	262144
+HOLE	524288
 Compare files
 c2803804acc9936eef8aab42c119bfac  SCRATCH_MNT/test-420/file1
 017c08a9320aad844ce86aa9631afb98  SCRATCH_MNT/test-420/file2
@@ -28,9 +26,7 @@
 Seek holes and data in file2
 Whence	Result
 DATA	0
-HOLE	131072
-DATA	196608
-HOLE	262144
+HOLE	524288
 Compare files
 c2803804acc9936eef8aab42c119bfac  SCRATCH_MNT/test-420/file1
 017c08a9320aad844ce86aa9631afb98  SCRATCH_MNT/test-420/file2

So the file checksums are right, which means I didn't break the COW
functionality.  But we're no longer reporting a hole at 128k.