Note build warnings (when running normal sparse checks during compile) looks like need to resolve endian errors so that this is sure to work in big endian, not just little endian boxes e.g. make C=1 M=fs/cifs modules CF=-D__CHECK_ENDIAN__ CHECK fs/cifs/cifs_unicode.c fs/cifs/cifs_unicode.c:508:34: warning: incorrect type in assignment (different base types) fs/cifs/cifs_unicode.c:508:34: expected restricted __le16 [assigned] [usertype] dst_char fs/cifs/cifs_unicode.c:508:34: got unsigned short [unsigned] [short] [usertype] <noident> fs/cifs/cifs_unicode.c:517:42: warning: incorrect type in assignment (different base types) fs/cifs/cifs_unicode.c:517:42: expected restricted __le16 [assigned] [usertype] dst_char fs/cifs/cifs_unicode.c:517:42: got unsigned short [unsigned] [short] [usertype] <noident> fs/cifs/cifs_unicode.c:523:42: warning: incorrect type in assignment (different base types) fs/cifs/cifs_unicode.c:523:42: expected restricted __le16 [assigned] [usertype] dst_char fs/cifs/cifs_unicode.c:523:42: got unsigned short [unsigned] [short] [usertype] <noident> fs/cifs/cifs_unicode.c:526:42: warning: incorrect type in assignment (different base types) fs/cifs/cifs_unicode.c:526:42: expected restricted __le16 [assigned] [usertype] dst_char fs/cifs/cifs_unicode.c:526:42: got unsigned short [unsigned] [short] [usertype] <noident> On Wed, Feb 4, 2015 at 1:23 AM, Nakajima Akira <nakajima.akira@xxxxxxxxxxxx> wrote: > Garbled characters happen by using surrogate pair for filename. > (replace each 1 character to ??) > > This causes > -- inode number duplication (when ?? and ?? exist) . > -- can't remove file, directory. > -- can't operate under directory. > -- auto-remount with noserverino. > > > [Steps to Reproduce] > client# touch $(echo -e '\xf0\x9d\x9f\xa3') > client# touch $(echo -e '\xf0\x9d\x9f\xa4') > client# ls -li > You see same inode number, same filename(=?? and ??) . > > > > [BUG description] > Some functions do not consider about surrogate pair (and IVS). > > cifs_utf16_bytes() > -- return incorrect length, because of not considering about surrogate pair. > This causes problem about SymbolicLink with surrogate pair filename. > > cifs_mapchar() > -- not considering about surrogate pair. > > cifs_from_utf16() > -- not convert surrogate pair from SMB response (from UTF-16 to UTF-8) > , then ls shows garbled characters. > > cifsConvertToUTF16() > -- not convert surrogate pair when mapchars(SFM/SFU). -- Thanks, Steve -- To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
