Re: [RFC][PATCH v2 11/11] context_tracking,x86: Fix text_poke_sync() vs NOHZ_FULL

[Marcelo Tosatti <mtosatti@redhat.com>]

Peter,

static __always_inline void arch_exit_to_user_mode(void)
{
        mds_user_clear_cpu_buffers();
}

/**
 * mds_user_clear_cpu_buffers - Mitigation for MDS and TAA vulnerability
 *
 * Clear CPU buffers if the corresponding static key is enabled
 */
static __always_inline void mds_user_clear_cpu_buffers(void)
{
        if (static_branch_likely(&mds_user_clear))
                mds_clear_cpu_buffers();
}

We were discussing how to perform objtool style validation 
that no code after the check for 

> +             /* NMI happens here and must still do/finish CT_WORK_n */
> +             sync_core();

But after the discussion with you, it seems doing the TLB checking 
and (also sync_core) checking very late/very early on exit/entry 
makes things easier to review.

Can then use a single atomic variable with USER/KERNEL state and cmpxchg
loops.

On Wed, Sep 29, 2021 at 05:17:34PM +0200, Peter Zijlstra wrote:
> Use the new context_tracking infrastructure to avoid disturbing
> userspace tasks when we rewrite kernel code.
> 
> XXX re-audit the entry code to make sure only the context_tracking
> static_branch is before hitting this code.
> 
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> ---
>  arch/x86/include/asm/sync_core.h |    2 ++
>  arch/x86/kernel/alternative.c    |    8 +++++++-
>  include/linux/context_tracking.h |    1 +
>  kernel/context_tracking.c        |   12 ++++++++++++
>  4 files changed, 22 insertions(+), 1 deletion(-)
> 
> --- a/arch/x86/include/asm/sync_core.h
> +++ b/arch/x86/include/asm/sync_core.h
> @@ -87,6 +87,8 @@ static inline void sync_core(void)
>  	 */
>  	iret_to_self();
>  }
> +#define sync_core sync_core
> +
>  
>  /*
>   * Ensure that a core serializing instruction is issued before returning
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -18,6 +18,7 @@
>  #include <linux/mmu_context.h>
>  #include <linux/bsearch.h>
>  #include <linux/sync_core.h>
> +#include <linux/context_tracking.h>
>  #include <asm/text-patching.h>
>  #include <asm/alternative.h>
>  #include <asm/sections.h>
> @@ -924,9 +925,14 @@ static void do_sync_core(void *info)
>  	sync_core();
>  }
>  
> +static bool do_sync_core_cond(int cpu, void *info)
> +{
> +	return !context_tracking_set_cpu_work(cpu, CT_WORK_SYNC);
> +}
> +
>  void text_poke_sync(void)
>  {
> -	on_each_cpu(do_sync_core, NULL, 1);
> +	on_each_cpu_cond(do_sync_core_cond, do_sync_core, NULL, 1);
>  }
>  
>  struct text_poke_loc {
> --- a/include/linux/context_tracking.h
> +++ b/include/linux/context_tracking.h
> @@ -11,6 +11,7 @@
>  
>  enum ct_work {
>  	CT_WORK_KLP = 1,
> +	CT_WORK_SYNC = 2,
>  };
>  
>  /*
> --- a/kernel/context_tracking.c
> +++ b/kernel/context_tracking.c
> @@ -51,6 +51,10 @@ static __always_inline void context_trac
>  	__this_cpu_dec(context_tracking.recursion);
>  }
>  
> +#ifndef sync_core
> +static inline void sync_core(void) { }
> +#endif
> +
>  /* CT_WORK_n, must be noinstr, non-blocking, NMI safe and deal with spurious calls */
>  static noinstr void ct_exit_user_work(struct context_tracking *ct)
>  {
> @@ -64,6 +68,14 @@ static noinstr void ct_exit_user_work(struct
>  		arch_atomic_andnot(CT_WORK_KLP, &ct->work);
>  	}
>  
> +	if (work & CT_WORK_SYNC) {
> +		/* NMI happens here and must still do/finish CT_WORK_n */
> +		sync_core();
> +
> +		smp_mb__before_atomic();
> +		arch_atomic_andnot(CT_WORK_SYNC, &ct->work);
> +	}
> +
>  	smp_mb__before_atomic();
>  	arch_atomic_andnot(CT_SEQ_WORK, &ct->seq);
>  }
> 
> 
>