On Wed, 18 Jun 2025 16:13:45 +0200
Peter Zijlstra <pet...@infradead.org> wrote:
> > diff --git a/include/linux/entry-common.h b/include/linux/entry-common.h
> > index f94f3fdf15fc..6e850c9d3f0c 100644
> > --- a/include/linux/entry-common.h
> > +++ b/include/linux/entry-common.h
> > @@ -12,6 +12,7 @@
> > #include <linux/resume_user_mode.h>
> > #include <linux/tick.h>
> > #include <linux/kmsan.h>
> > +#include <linux/unwind_deferred.h>
> >
> > #include <asm/entry-common.h>
> > #include <asm/syscall.h>
> > @@ -362,6 +363,7 @@ static __always_inline void exit_to_user_mode(void)
> > lockdep_hardirqs_on_prepare();
> > instrumentation_end();
> >
> > + unwind_exit_to_user_mode();
>
> So I was expecting this to do the actual unwind, and was about to go
> yell this is the wrong place for that.
>
> But this is not that. Perhaps find a better name like:
> unwind_clear_cache() or so?
Sure.
How about unwind_reset_info()?
As it's not going to just clear the cache but also reset the trace info
(like the timestamp and such).
>
> > user_enter_irqoff();
> > arch_exit_to_user_mode();
> > lockdep_hardirqs_on(CALLER_ADDR0);
>
>
> > diff --git a/include/linux/unwind_deferred_types.h
> > b/include/linux/unwind_deferred_types.h
> > index aa32db574e43..db5b54b18828 100644
> > --- a/include/linux/unwind_deferred_types.h
> > +++ b/include/linux/unwind_deferred_types.h
> > @@ -2,8 +2,13 @@
> > #ifndef _LINUX_UNWIND_USER_DEFERRED_TYPES_H
> > #define _LINUX_UNWIND_USER_DEFERRED_TYPES_H
> >
> > +struct unwind_cache {
> > + unsigned int nr_entries;
> > + unsigned long entries[];
> > +};
> > +
> > struct unwind_task_info {
> > - unsigned long *entries;
> > + struct unwind_cache *cache;
> > };
> >
> > #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */
> > diff --git a/kernel/unwind/deferred.c b/kernel/unwind/deferred.c
> > index 0bafb95e6336..e3913781c8c6 100644
> > --- a/kernel/unwind/deferred.c
> > +++ b/kernel/unwind/deferred.c
> > @@ -24,6 +24,7 @@
> > int unwind_deferred_trace(struct unwind_stacktrace *trace)
> > {
> > struct unwind_task_info *info = ¤t->unwind_info;
> > + struct unwind_cache *cache;
> >
> > /* Should always be called from faultable context */
> > might_fault();
> > @@ -31,17 +32,30 @@ int unwind_deferred_trace(struct unwind_stacktrace
> > *trace)
> > if (current->flags & PF_EXITING)
> > return -EINVAL;
> >
> > - if (!info->entries) {
> > - info->entries = kmalloc_array(UNWIND_MAX_ENTRIES, sizeof(long),
> > - GFP_KERNEL);
> > - if (!info->entries)
> > + if (!info->cache) {
> > + info->cache = kzalloc(struct_size(cache, entries,
> > UNWIND_MAX_ENTRIES),
> > + GFP_KERNEL);
>
> And now you're one 'long' larger than a page. Surely that's a crap size
> for an allocator?
Bah, Ingo suggested to put the counter in the allocation and I didn't think
about the size going over the page. Good catch!
Since it can make one per task, it may be good to make this into a
kmemcache.
-- Steve
