In the Linux kernel, the following vulnerability has been resolved: libfs: fix infinite directory reads for offset dir After we switch tmpfs dir operations from simplediroperations to simpleoffsetdiroperations, every rename happened will fill new dentry to dest dir's maple tree(&SHMEMI(inode)->diroffsets->mt) with a free key starting with octx->newxoffset, and then set newxoffset equals to free key + 1. This will lead to infinite readdir combine with rename happened at the same time, which fail generic/736 in xfstests(detail show as below). 1. create 5000 files(1 2 3...) under one dir 2. call readdir(man 3 readdir) once, and get one entry 3. rename(entry, "TEMPFILE"), then rename("TEMPFILE", entry) 4. loop 2~3, until readdir return nothing or we loop too many times(tmpfs break test with the second condition) We choose the same logic what commit 9b378f6ad48cf ("btrfs: fix infinite directory reads") to fix it, record the lastindex when we open dir, and do not emit the entry which index >= lastindex. The file->privatedata now used in offset dir can use directly to do this, and we also update the lastindex when we llseek the dir file. [brauner: only update lastindex after seek when offset is zero like Jan suggested]