Hi Konstantin, I found three problems with the NTFS (de)compression code. The first two are simple inefficiencies; there's no need to kmap these pages. The third looks like a potential data corruption (description in the patch). I've marked the third one for backport. I haven't tested any of this; it could be that my analysis is wrong. Matthew Wilcox (Oracle) (3): ntfs: Do not kmap pages used for reading from disk ntfs: Do not kmap page cache pages for compression ntfs: Do not overwrite uptodate pages fs/ntfs3/frecord.c | 50 +++++++++++++++++++++++++++------------------- 1 file changed, 29 insertions(+), 21 deletions(-) -- 2.47.2
