Re: [RFC][PATCH] x86/mm: Sync all vmalloc mappings before text_poke()

[Steven Rostedt]

On Wed, 29 Apr 2020 08:28:54 -0400
Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:

> On Wed, 29 Apr 2020 12:59:41 +0200
> Joerg Roedel <jroedel@xxxxxxx> wrote:
> 
> > >   use_temporary_mm() {
> > >     switch_mm_irqs_off() {
> > >       load_new_mm_cr3() {
> > >         write_cr3() <<--- Lock up!    
> > 
> > I don't see how it could lock up in write_cr3(), at least on bare-metal.
> > What is the environment this happens, 32 or 64 bit, in a VM or
> > bare-metal?  
> 
> 64 bit bare-metal. In fact, it wasn't reproducible on a VM (according
> to Tzvetomir, who was just using a Fedora kernel). I only tried it on
> bare-metal.
> 
> > 
> > I think it is more likely that your lockup is actually a page-fault
> > loop, where the #PF handler does not map the faulting address correctly.  
> 
> Sounds reasonable.
> 
> > 
> > But I have to look closer into how text_poke() works before I can say
> > more.
> > 
> > Btw, in case it happens on x86-64, does it also happen without
> > vmalloc-stacks?  
> 
> Just tried it out with !CONFIG_VMAP_STACKS and it still locks up :-/

Talking with Mathieu about this on IRC, he pointed out that my code does
have a vzalloc() that is called:

in trace_pid_write()

	pid_list->pids = vzalloc((pid_list->pid_max + 7) >> 3);

This is done when -P1,2 is on the trace-cmd command line.

-- Steve