In the Linux kernel, the following vulnerability has been resolved:
dm btree remove: assign new_root only when removal succeeds
removeraw() in dmbtreeremove() may fail due to IO read error (e.g. read the content of origin block fails during shadowing), and the value of shadowspine::root is uninitialized, but the uninitialized value is still assign to newroot in the end of dmbtree_remove().
For dm-thin, the value of pmd->detailsroot or pmd->root will become an uninitialized value, so if trying to read detailsinfo tree again out-of-bound memory may occur as showed below:
general protection fault, probably for non-canonical address 0x3fdcb14c8d7520 CPU: 4 PID: 515 Comm: dmsetup Not tainted 5.13.0-rc6 Hardware name: QEMU Standard PC RIP: 0010:metadatallloadie+0x14/0x30 Call Trace: smmetadatacountismorethanone+0xb9/0xe0 dmtmshadowblock+0x52/0x1c0 shadowstep+0x59/0xf0 removeraw+0xb2/0x170 dmbtreeremove+0xf4/0x1c0 dmpooldeletethindevice+0xc3/0x140 poolmessage+0x218/0x2b0 targetmessage+0x251/0x290 ctlioctl+0x1c4/0x4d0 dmctlioctl+0xe/0x20 _x64sysioctl+0x7b/0xb0 dosyscall64+0x40/0xb0 entrySYSCALL64afterhwframe+0x44/0xae
Fixing it by only assign new_root when removal succeeds