Re: UBSAN: Undefined error in time.h signed integer overflow

From: Thomas Gleixner
Date: Wed Sep 06 2017 - 04:07:42 EST


On Tue, 5 Sep 2017, John Stultz wrote:
> On Tue, Sep 5, 2017 at 9:30 PM, Shankara Pailoor <sp3485@xxxxxxxxxxxx> wrote:
> > Hi,
> >
> > I encountered this bug while fuzzing linux kernel 4.13-rc7 with syzkaller.
> >
> > ================================================================================
> > UBSAN: Undefined behaviour in ./include/linux/time.h:233:27
> > signed integer overflow:
> > 8391720337152500783 * 1000000000 cannot be represented in type 'long int'
> > CPU: 0 PID: 31798 Comm: syz-executor2 Not tainted 4.13.0-rc7 #2
> > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
> > Ubuntu-1.8.2-1ubuntu1 04/01/2014
> > Call Trace:
> > __dump_stack lib/dump_stack.c:16 [inline]
> > dump_stack+0xf7/0x1ae lib/dump_stack.c:52
> > ubsan_epilogue+0x12/0x8f lib/ubsan.c:164
> > handle_overflow+0x21e/0x292 lib/ubsan.c:195
> > __ubsan_handle_mul_overflow+0x2a/0x3e lib/ubsan.c:219
> > timespec_to_ns include/linux/time.h:233 [inline]
> > posix_cpu_timer_set+0xb5c/0xf20 kernel/time/posix-cpu-timers.c:686
> > do_timer_settime+0x1f4/0x390 kernel/time/posix-timers.c:890
> > SYSC_timer_settime kernel/time/posix-timers.c:916 [inline]
> > SyS_timer_settime+0xea/0x170 kernel/time/posix-timers.c:902
> > entry_SYSCALL_64_fastpath+0x18/0xad
> > RIP: 0033:0x451e59
> > RSP: 002b:00007fb62af4fc08 EFLAGS: 00000216 ORIG_RAX: 00000000000000df
> > RAX: ffffffffffffffda RBX: 0000000000718000 RCX: 0000000000451e59
> > RDX: 0000000020006000 RSI: 0000000000000000 RDI: 0000000000000000
> > RBP: 0000000000000046 R08: 0000000000000000 R09: 0000000000000000
> > R10: 0000000020003fe0 R11: 0000000000000216 R12: 00000000004be920
> > R13: 00000000ffffffff R14: ffffffffffffffff R15: 0000000000000000
> > ================================================================================
>
> Looks similar to the issue Thomas fixed here:
> https://patchwork.kernel.org/patch/9799827/
>
>
> Thomas: Should we change timespec_to_ns() to use the same transition
> internally rather then trying to track all the callers?

Probably.

Thanks,

tglx