On Tue, Mar 12, 2019 at 11:07:53AM -0700, Zubin Mithra wrote:

Hello,

Syzkaller has triggered a stack OOB read when fuzzing a 4.4 kernel with the following stacktrace.

Call Trace: [<ffffffff81cb9fad>] __dump_stack lib/dump_stack.c:15 [inline] [<ffffffff81cb9fad>] dump_stack+0xc1/0x124 lib/dump_stack.c:51 [<ffffffff815275e9>] print_address_description mm/kasan/report.c:199 [inline] [<ffffffff815275e9>] kasan_report_error mm/kasan/report.c:285 [inline] [<ffffffff815275e9>] kasan_report.part.2.cold.3+0x447/0x4ec mm/kasan/report.c:310 [<ffffffff814eca2e>] kasan_report mm/kasan/report.c:328 [inline] [<ffffffff814eca2e>] __asan_report_load1_noabort+0x2e/0x30 mm/kasan/report.c:328 [<ffffffff81cd54d1>] strnlen+0xc1/0xd0 lib/string.c:498 [<ffffffff81cdd6ac>] string.isra.4+0x4c/0x250 lib/vsprintf.c:518 [<ffffffff81ce27da>] vsnprintf+0x42a/0x18c0 lib/vsprintf.c:1904 [<ffffffff8114fd73>] __request_module+0x153/0x7a0 kernel/kmod.c:146 [<ffffffff82986521>] find_inlist_lock.constprop.15+0x111/0x210 net/bridge/netfilter/ebtables.c:347 [<ffffffff8298a942>] find_table_lock net/bridge/netfilter/ebtables.c:356 [inline] [<ffffffff8298a942>] do_ebt_get_ctl+0x152/0x570 net/bridge/netfilter/ebtables.c:1531 [<ffffffff823e8bc5>] nf_sockopt net/netfilter/nf_sockopt.c:103 [inline] [<ffffffff823e8bc5>] nf_getsockopt+0x75/0xd0 net/netfilter/nf_sockopt.c:121 [<ffffffff8261d60d>] ip_getsockopt+0x12d/0x170 net/ipv4/ip_sockglue.c:1533 [<ffffffff826411cd>] tcp_getsockopt+0x8d/0xe0 net/ipv4/tcp.c:3040 [<ffffffff82250c0f>] sock_common_getsockopt+0x9f/0xe0 net/core/sock.c:2652 [<ffffffff8224e45d>] SYSC_getsockopt net/socket.c:1811 [inline] [<ffffffff8224e45d>] SyS_getsockopt+0x14d/0x230 net/socket.c:1793 [<ffffffff82a5f3cf>] tracesys_phase2+0x90/0x95

Could the following patch be applied to v4.4.y? The patch is present in v4.9.y.

• b301f2538759 ("netfilter: x_tables: enforce nul-terminated table name from getsockopt GET_ENTRIES")

Tests run:

• Chrome OS tryjobs
• Syzkaller reproducer

Now queued up, thanks.

greg k-h