Re: [PATCH] Reset task stack state in bringup_cpu()

[Mark Rutland <mark.rutland@arm.com>]

On Tue, Nov 16, 2021 at 11:31:40AM -0500, Qian Cai wrote:
> On Mon, Nov 15, 2021 at 11:33:10AM +0000, Mark Rutland wrote:
> > To hot unplug a CPU, the idle task on that CPU calls a few layers of C
> > code before finally leaving the kernel. When KASAN is in use, poisoned
> > shadow is left around for each of the active stack frames, and when
> > shadow call stacks are in use. When shadow call stacks are in use the
> > task's SCS SP is left pointing at an arbitrary point within the task's
> > shadow call stack.
> > 
> > When an offlines CPU is hotlpugged back into the kernel, this stale
> > state can adversely affect the newly onlined CPU. Stale KASAN shadow can
> > alias new stackframes and result in bogus KASAN warnings. A stale SCS SP
> > is effectively a memory leak, and prevents a portion of the shadow call
> > stack being used. Across a number of hotplug cycles the task's entire
> > shadow call stack can become unusable.
> > 
> > We previously fixed the KASAN issue in commit:
> > 
> >   e1b77c92981a5222 ("sched/kasan: remove stale KASAN poison after hotplug")
> > 
> > In commit:
> > 
> >   f1a0a376ca0c4ef1 ("sched/core: Initialize the idle task with preemption disabled")
> > 
> > ... we broke both KASAN and SCS, with SCS being fixed up in commit:
> > 
> >   63acd42c0d4942f7 ("sched/scs: Reset the shadow stack when idle_task_exit")
> > 
> > ... but as this runs in the context of the idle task being offlines it's
> > potentially fragile.
> > 
> > Fix both of these consistently and more robustly by resetting the SCS SP
> > and KASAN shadow immediately before we online a CPU. This ensures the
> > idle task always has a consistent state, and removes the need to do so
> > when initializing an idle task or when unplugging an idle task.
> > 
> > I've tested this with both GCC and clang, with reelvant options enabled,
> > offlining and online CPUs with:
> > 
> > | while true; do
> > |   for C in /sys/devices/system/cpu/cpu*/online; do
> > |     echo 0 > $C;
> > |     echo 1 > $C;
> > |   done
> > | done
> > 
> > Link: https://lore.kernel.org/lkml/20211012083521.973587-1-woodylin@google.com/
> > Link: https://lore.kernel.org/linux-arm-kernel/YY9ECKyPtDbD9q8q@qian-HP-Z2-SFF-G5-Workstation/
> > Fixes: 1a0a376ca0c4ef1 ("sched/core: Initialize the idle task with preemption disabled")
> > Reported-by: Qian Cai <quic_qiancai@quicinc.com>
> > Signed-off-by: Mark Rutland <mark.rutland@arm.com>
> 
> Thanks for fixing this quickly, Mark. Triggering an user-after-free in
> user namespace but don't think it is related. I'll investigate that
> first since it is blocking the rest of regression testing.

Cool; are you happy to provide a Tested-by tag for this patch? :)

Thanks,
Mark.