On Tue, Apr 16, 2019 at 01:29:59PM -0700, Zubin Mithra wrote:
Hello,
Syzkaller has triggered a UAF when fuzzing a 4.19 kernel with the following stacktrace.
Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0xc8/0x129 lib/dump_stack.c:113 print_address_description+0x67/0x230 mm/kasan/report.c:256 kasan_report_error mm/kasan/report.c:354 [inline] kasan_report+0x24e/0x28c mm/kasan/report.c:412 get_link fs/namei.c:1152 [inline] trailing_symlink+0x593/0x677 fs/namei.c:2326 path_lookupat.isra.35+0x413/0x5d1 fs/namei.c:2382 filename_lookup.part.50+0xe1/0x1b7 fs/namei.c:2411 filename_lookup fs/namei.c:2405 [inline] user_path_at_empty+0x59/0x6c fs/namei.c:2677 user_path include/linux/namei.h:62 [inline] do_mount+0x15c/0x17a4 fs/namespace.c:2773 ksys_mount+0x98/0xcc fs/namespace.c:3052 __do_sys_mount fs/namespace.c:3066 [inline] __se_sys_mount fs/namespace.c:3063 [inline] __x64_sys_mount+0xd0/0xdb fs/namespace.c:3063 do_syscall_64+0xf8/0x133 arch/x86/entry/common.c:291 entry_SYSCALL_64_after_hwframe+0x49/0xbe
Allocated by task 8112: set_track mm/kasan/kasan.c:460 [inline] kasan_kmalloc+0x85/0x93 mm/kasan/kasan.c:553 slab_post_alloc_hook+0x31/0x55 mm/slab.h:444 slab_alloc_node mm/slub.c:2706 [inline] slab_alloc mm/slub.c:2714 [inline] __kmalloc_track_caller+0x100/0x148 mm/slub.c:4290 kstrdup+0x39/0x63 mm/util.c:56 bpf_symlink+0x26/0xf4 kernel/bpf/inode.c:356 vfs_symlink2+0xfc/0x12b fs/namei.c:4238 do_symlinkat+0x14a/0x1d5 fs/namei.c:4271 do_syscall_64+0xf8/0x133 arch/x86/entry/common.c:291 entry_SYSCALL_64_after_hwframe+0x49/0xbe
Freed by task 8116: set_track mm/kasan/kasan.c:460 [inline] __kasan_slab_free+0x100/0x122 mm/kasan/kasan.c:521 slab_free_hook mm/slub.c:1371 [inline] slab_free_freelist_hook+0x9a/0xed mm/slub.c:1398 slab_free mm/slub.c:2953 [inline] kfree+0x177/0x212 mm/slub.c:3906 bpf_evict_inode+0x80/0x107 kernel/bpf/inode.c:565 evict+0x30b/0x4ce fs/inode.c:558 iput_final fs/inode.c:1550 [inline] iput+0x541/0x551 fs/inode.c:1576 do_unlinkat+0x2fc/0x403 fs/namei.c:4180 do_syscall_64+0xf8/0x133 arch/x86/entry/common.c:291 entry_SYSCALL_64_after_hwframe+0x49/0xbe
Could the following patch be applied to 4.19.y? 1da6c4d9140c ("bpf: fix use after free in bpf_evict_inode")
Tests run:
- Chrome OS tryjobs
- Syzkaller reproducer
I've queued it up, thanks again for all these tests!
-- Thanks, Sasha