"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nubifs: ubifs_symlink: Fix memleak of inode->i_link in error path\n\nFor error handling path in ubifs_symlink(), inode will be marked as\nbad first, then iput() is invoked. If inode->i_link is initialized by\nfscrypt_encrypt_symlink() in encryption scenario, inode->i_link won't\nbe freed by callchain ubifs_free_inode -> fscrypt_free_inode in error\nhandling path, because make_bad_inode() has changed 'inode->i_mode' as\n'S_IFREG'.\nFollowing kmemleak is easy to be reproduced by injecting error in\nubifs_jnl_update() when doing symlink in encryption scenario:\n unreferenced object 0xffff888103da3d98 (size 8):\n comm \"ln\", pid 1692, jiffies 4294914701 (age 12.045s)\n backtrace:\n kmemdup+0x32/0x70\n __fscrypt_encrypt_symlink+0xed/0x1c0\n ubifs_symlink+0x210/0x300 [ubifs]\n vfs_symlink+0x216/0x360\n do_symlinkat+0x11a/0x190\n do_syscall_64+0x3b/0xe0\nThere are two ways fixing it:\n 1. Remove make_bad_inode() in error handling path. We can do that\n because ubifs_evict_inode() will do same processes for good\n symlink inode and bad symlink inode, for inode->i_nlink checking\n is before is_bad_inode().\n 2. Free inode->i_link before marking inode bad.\nMethod 2 is picked, it has less influence, personally, I think."
