On Wed, Apr 05, 2023 at 08:42:34AM -0700, Christoph Hellwig wrote: > On Wed, Apr 05, 2023 at 08:40:22AM -0700, Darrick J. Wong wrote: > > <shrug> Welllll... if someone presents a strong case for adopting the > > utf8 casefolding feature that f2fs and ext4 added some ways back, I > > could be persuaded to import that, bugs and all, into XFS. However, > > given all the weird problems I've uncovered with "ascii"-ci, I'm going > > to be very hardnosed about adding test cases and making sure /all/ the > > tooling works properly. > > You'll love this paper: > > https://www.usenix.org/conference/fast23/presentation/basu I know. I stick to my earlier statements about "I wouldn't ever enable this feature on any computer I use..." and "...not a tarpit that I ever want to visit in XFS." At one point I had wired up xfs_scrub to complain about filenames that map to the same casefolded utf8 names to warn syadmins that this could be used in some sort of unicode casefolding attack. I pushed it back on my patch stack and ran it against /home today and got a bunch of stuff like this from the kernel source tree: Warning: inode 4187068442 (31/26318874): Case-folded Unicode name "ip6t_hl.h" in directory could be confused with "ip6t_HL.h". (unicrash.c line 614) Warning: inode 4032422946 (30/5891106): Case-folded Unicode name "ipt_ecn.h" in directory could be confused with "ipt_ECN.h". (unicrash.c line 614) Warning: inode 4032422946 (30/5891106): Case-folded Unicode name "ipt_ttl.h" in directory could be confused with "ipt_TTL.h". (unicrash.c line 614) Warning: inode 2285477942 (17/3776566): Case-folded Unicode name "xt_hl.c" in directory could be confused with "xt_HL.c". (unicrash.c line 614) Warning: inode 2285477942 (17/3776566): Case-folded Unicode name "xt_tcpmss.c" in directory could be confused with "xt_TCPMSS.c". (unicrash.c line 614) Warning: inode 3627924489 (27/4045833): Case-folded Unicode name "xt_tcpmss.h" in directory could be confused with "xt_TCPMSS.h". (unicrash.c line 614) Warning: inode 3763353714 (28/5257330): Case-folded Unicode name "ip6t_hl.h" in directory could be confused with "ip6t_HL.h". (unicrash.c line 614) Warning: inode 3763353714 (28/5257330): Case-folded Unicode name ".ip6t_hl.h.cmd" in directory could be confused with ".ip6t_HL.h.cmd". (unicrash.c line 614) Warning: inode 7042717 (0/7042717): Case-folded Unicode name ".ipt_ecn.h.cmd" in directory could be confused with ".ipt_ECN.h.cmd". (unicrash.c line 614) Warning: inode 7042717 (0/7042717): Case-folded Unicode name ".ipt_ttl.h.cmd" in directory could be confused with ".ipt_TTL.h.cmd". (unicrash.c line 614) Warning: inode 7042718 (0/7042718): Case-folded Unicode name "ip6t_hl.h" in directory could be confused with "ip6t_HL.h". (unicrash.c line 614) Warning: inode 7042718 (0/7042718): Case-folded Unicode name ".ip6t_hl.h.cmd" in directory could be confused with ".ip6t_HL.h.cmd". (unicrash.c line 614) Warning: inode 406880264 (3/4227080): Case-folded Unicode name "Z6.0+pooncelock+pooncelock+pombonce.litmus" in directory could be confused with "Z6.0+pooncelock+poonceLock+pombonce.litmus". (unicrash.c line 614) Yuck. I never sent this patch to linux-xfs because XFS doesn't do casefolding so who cares. The xtables stuff is easy to spot, but that last one took some staring before I even figured out what was different between the two names -- lock vs Lock. --D
