On Tue, May 28, 2019 at 07:32:28PM +0200, Peter Zijlstra wrote:
> On Tue, May 28, 2019 at 04:32:24PM +0100, Will Deacon wrote:
> > On Tue, May 28, 2019 at 04:01:03PM +0200, Peter Zijlstra wrote:
> > > On Tue, May 28, 2019 at 08:31:29PM +0800, Young Xiao wrote:
> > > > When a kthread calls call_usermodehelper() the steps are:
> > > > 1. allocate current->mm
> > > > 2. load_elf_binary()
> > > > 3. populate current->thread.regs
> > > >
> > > > While doing this, interrupts are not disabled. If there is a perf
> > > > interrupt in the middle of this process (i.e. step 1 has completed
> > > > but not yet reached to step 3) and if perf tries to read userspace
> > > > regs, kernel oops.
> >
> > This seems to be because pt_regs(current) gives NULL for kthreads on Power.
>
> 'funny' thing that, perf_sample_regs_user() seems to assume that
> anything with current->mm is in fact a user task, and that assumption is
> just plain wrong, consider use_mm().
Tagnentially, it looks like that assumption is made elsewhere, and could
do with a more general cleanup. IIUC, the following are suspect:
* kmemleak's scan_should_stop()
* x86's __kernel_fpu_begin()
* arm64's arch_dup_task_struct()
It's probably worth an is_thread(task) helper so that those can be
written in an obviously correct way.
> So I'm thinking the right thing to do here is something like the below;
> umh should get PF_KTHREAD cleared when it passes exec(). And this should
> also fix the power splat I'm thinking.
>
> ---
>
> diff --git a/kernel/events/core.c b/kernel/events/core.c
> index abbd4b3b96c2..9929404b6eb9 100644
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -5923,7 +5923,7 @@ static void perf_sample_regs_user(struct perf_regs
> *regs_user,
> if (user_mode(regs)) {
> regs_user->abi = perf_reg_abi(current);
> regs_user->regs = regs;
> - } else if (current->mm) {
> + } else if (!(current->flags & PF_KTHREAD) && current->mm) {
Wouldn't !PF_KTHREAD imply current->mm anyhow?
Thanks,
Mark.
