CVE-2021-47118

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> pid: take a reference when initializing `cad_pid`<br /> <br /> During boot, kernel_init_freeable() initializes `cad_pid` to the init<br /> task&amp;#39;s struct pid. Later on, we may change `cad_pid` via a sysctl, and<br /> when this happens proc_do_cad_pid() will increment the refcount on the<br /> new pid via get_pid(), and will decrement the refcount on the old pid<br /> via put_pid(). As we never called get_pid() when we initialized<br /> `cad_pid`, we decrement a reference we never incremented, can therefore<br /> free the init task&amp;#39;s struct pid early. As there can be dangling<br /> references to the struct pid, we can later encounter a use-after-free<br /> (e.g. when delivering signals).<br /> <br /> This was spotted when fuzzing v5.13-rc3 with Syzkaller, but seems to<br /> have been around since the conversion of `cad_pid` to struct pid in<br /> commit 9ec52099e4b8 ("[PATCH] replace cad_pid by a struct pid") from the<br /> pre-KASAN stone age of v2.6.19.<br /> <br /> Fix this by getting a reference to the init task&amp;#39;s struct pid when we<br /> assign it to `cad_pid`.<br /> <br /> Full KASAN splat below.<br /> <br /> ==================================================================<br /> BUG: KASAN: use-after-free in ns_of_pid include/linux/pid.h:153 [inline]<br /> BUG: KASAN: use-after-free in task_active_pid_ns+0xc0/0xc8 kernel/pid.c:509<br /> Read of size 4 at addr ffff23794dda0004 by task syz-executor.0/273<br /> <br /> CPU: 1 PID: 273 Comm: syz-executor.0 Not tainted 5.12.0-00001-g9aef892b2d15 #1<br /> Hardware name: linux,dummy-virt (DT)<br /> Call trace:<br /> ns_of_pid include/linux/pid.h:153 [inline]<br /> task_active_pid_ns+0xc0/0xc8 kernel/pid.c:509<br /> do_notify_parent+0x308/0xe60 kernel/signal.c:1950<br /> exit_notify kernel/exit.c:682 [inline]<br /> do_exit+0x2334/0x2bd0 kernel/exit.c:845<br /> do_group_exit+0x108/0x2c8 kernel/exit.c:922<br /> get_signal+0x4e4/0x2a88 kernel/signal.c:2781<br /> do_signal arch/arm64/kernel/signal.c:882 [inline]<br /> do_notify_resume+0x300/0x970 arch/arm64/kernel/signal.c:936<br /> work_pending+0xc/0x2dc<br /> <br /> Allocated by task 0:<br /> slab_post_alloc_hook+0x50/0x5c0 mm/slab.h:516<br /> slab_alloc_node mm/slub.c:2907 [inline]<br /> slab_alloc mm/slub.c:2915 [inline]<br /> kmem_cache_alloc+0x1f4/0x4c0 mm/slub.c:2920<br /> alloc_pid+0xdc/0xc00 kernel/pid.c:180<br /> copy_process+0x2794/0x5e18 kernel/fork.c:2129<br /> kernel_clone+0x194/0x13c8 kernel/fork.c:2500<br /> kernel_thread+0xd4/0x110 kernel/fork.c:2552<br /> rest_init+0x44/0x4a0 init/main.c:687<br /> arch_call_rest_init+0x1c/0x28<br /> start_kernel+0x520/0x554 init/main.c:1064<br /> 0x0<br /> <br /> Freed by task 270:<br /> slab_free_hook mm/slub.c:1562 [inline]<br /> slab_free_freelist_hook+0x98/0x260 mm/slub.c:1600<br /> slab_free mm/slub.c:3161 [inline]<br /> kmem_cache_free+0x224/0x8e0 mm/slub.c:3177<br /> put_pid.part.4+0xe0/0x1a8 kernel/pid.c:114<br /> put_pid+0x30/0x48 kernel/pid.c:109<br /> proc_do_cad_pid+0x190/0x1b0 kernel/sysctl.c:1401<br /> proc_sys_call_handler+0x338/0x4b0 fs/proc/proc_sysctl.c:591<br /> proc_sys_write+0x34/0x48 fs/proc/proc_sysctl.c:617<br /> call_write_iter include/linux/fs.h:1977 [inline]<br /> new_sync_write+0x3ac/0x510 fs/read_write.c:518<br /> vfs_write fs/read_write.c:605 [inline]<br /> vfs_write+0x9c4/0x1018 fs/read_write.c:585<br /> ksys_write+0x124/0x240 fs/read_write.c:658<br /> __do_sys_write fs/read_write.c:670 [inline]<br /> __se_sys_write fs/read_write.c:667 [inline]<br /> __arm64_sys_write+0x78/0xb0 fs/read_write.c:667<br /> __invoke_syscall arch/arm64/kernel/syscall.c:37 [inline]<br /> invoke_syscall arch/arm64/kernel/syscall.c:49 [inline]<br /> el0_svc_common.constprop.1+0x16c/0x388 arch/arm64/kernel/syscall.c:129<br /> do_el0_svc+0xf8/0x150 arch/arm64/kernel/syscall.c:168<br /> el0_svc+0x28/0x38 arch/arm64/kernel/entry-common.c:416<br /> el0_sync_handler+0x134/0x180 arch/arm64/kernel/entry-common.c:432<br /> el0_sync+0x154/0x180 arch/arm64/kernel/entry.S:701<br /> <br /> The buggy address belongs to the object at ffff23794dda0000<br /> which belongs to the cache pid of size 224<br /> The buggy address is located 4 bytes inside of<br /> 224-byte region [ff<br /> ---truncated---