Re: [PATCH v2 05/20] KVM: s390: Add helper functions for fault handling

[Claudio Imbrenda <imbrenda@linux.ibm.com>]

On Thu, 18 Sep 2025 16:41:56 +0200
Alexander Gordeev <agordeev@linux.ibm.com> wrote:

> On Wed, Sep 10, 2025 at 08:07:31PM +0200, Claudio Imbrenda wrote:
> > Add some helper functions for handling multiple guest faults at the
> > same time.
> > 
> > This will be needed for VSIE, where a nested guest access also needs to
> > access all the page tables that map it.
> > 
> > Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> > ---
> >  arch/s390/kvm/gaccess.h  | 14 ++++++++++
> >  arch/s390/kvm/kvm-s390.c | 44 +++++++++++++++++++++++++++++++
> >  arch/s390/kvm/kvm-s390.h | 56 ++++++++++++++++++++++++++++++++++++++++
> >  3 files changed, 114 insertions(+)
> > 
> > diff --git a/arch/s390/kvm/gaccess.h b/arch/s390/kvm/gaccess.h
> > index 3fde45a151f2..9c82f7460821 100644
> > --- a/arch/s390/kvm/gaccess.h
> > +++ b/arch/s390/kvm/gaccess.h
> > @@ -457,4 +457,18 @@ int kvm_s390_check_low_addr_prot_real(struct kvm_vcpu *vcpu, unsigned long gra);
> >  int kvm_s390_shadow_fault(struct kvm_vcpu *vcpu, struct gmap *shadow,
> >  			  unsigned long saddr, unsigned long *datptr);
> >  
> > +static inline int __kvm_s390_faultin_read_gpa(struct kvm *kvm, struct guest_fault *f, gpa_t gaddr,
> > +					      unsigned long *val)
> > +{
> > +	phys_addr_t phys_addr;
> > +	int rc;
> > +
> > +	rc = __kvm_s390_faultin_gfn(kvm, f, gpa_to_gfn(gaddr), false);  
> 
> Why not a "typical" flow (below)?
> 
> > +	if (!rc) {
> > +		phys_addr = PFN_PHYS(f->pfn) | offset_in_page(gaddr);
> > +		*val = *(unsigned long *)phys_to_virt(phys_addr);
> > +	}
> > +	return rc;  
> 
> 	if (rc)
> 		return rc;
> 
> 	phys_addr = pfn_to_phys(f->pfn) | offset_in_page(gaddr);
> 	*val = *(unsigned long *)phys_to_virt(phys_addr);
> 
> 	return 0;
> 

I swear, at some point it made more sense in the way I had written it :D

I'll fix it

> > +}
> > +
> >  #endif /* __KVM_S390_GACCESS_H */
> > diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
> > index 61aa64886c36..af8a62abec48 100644
> > --- a/arch/s390/kvm/kvm-s390.c
> > +++ b/arch/s390/kvm/kvm-s390.c
> > @@ -4858,6 +4858,50 @@ static void kvm_s390_assert_primary_as(struct kvm_vcpu *vcpu)
> >  		current->thread.gmap_int_code, current->thread.gmap_teid.val);
> >  }
> >  
> > +/**
> > + * __kvm_s390_faultin_gfn() - fault in and pin a guest address
> > + * @kvm: the guest
> > + * @guest_fault: will be filled with information on the pin operation
> > + * @gfn: guest frame
> > + * @wr: if true indicates a write access
> > + *
> > + * Fault in and pin a guest address using absolute addressing, and without
> > + * marking the page referenced.
> > + *
> > + * Context: Called with mm->mmap_lock in read mode.
> > + *
> > + * Return:
> > + * * 0 in case of success,
> > + * * -EFAULT if reading using the virtual address failed,
> > + * * -EINTR if a signal is pending,
> > + * * -EAGAIN if FOLL_NOWAIT was specified, but IO is needed
> > + * * PGM_ADDRESSING if the guest address lies outside of guest memory.
> > + */
> > +int __kvm_s390_faultin_gfn(struct kvm *kvm, struct guest_fault *guest_fault, gfn_t gfn, bool wr)
> > +{
> > +	struct kvm_memory_slot *slot;
> > +	kvm_pfn_t pfn;
> > +	int foll;
> > +
> > +	foll = wr ? FOLL_WRITE : 0;
> > +	slot = gfn_to_memslot(kvm, gfn);
> > +	pfn = __kvm_faultin_pfn(slot, gfn, foll, &guest_fault->writable, &guest_fault->page);
> > +	if (is_noslot_pfn(pfn))
> > +		return PGM_ADDRESSING;
> > +	if (is_sigpending_pfn(pfn))
> > +		return -EINTR;
> > +	if (pfn == KVM_PFN_ERR_NEEDS_IO)
> > +		return -EAGAIN;
> > +	if (is_error_pfn(pfn))
> > +		return -EFAULT;
> > +
> > +	guest_fault->pfn = pfn;
> > +	guest_fault->gfn = gfn;
> > +	guest_fault->write_attempt = wr;
> > +	guest_fault->valid = true;
> > +	return 0;
> > +}
> > +
> >  /*
> >   * __kvm_s390_handle_dat_fault() - handle a dat fault for the gmap of a vcpu
> >   * @vcpu: the vCPU whose gmap is to be fixed up
> > diff --git a/arch/s390/kvm/kvm-s390.h b/arch/s390/kvm/kvm-s390.h
> > index c44fe0c3a097..dabcf65f58ff 100644
> > --- a/arch/s390/kvm/kvm-s390.h
> > +++ b/arch/s390/kvm/kvm-s390.h
> > @@ -22,6 +22,15 @@
> >  
> >  #define KVM_S390_UCONTROL_MEMSLOT (KVM_USER_MEM_SLOTS + 0)
> >  
> > +struct guest_fault {
> > +	gfn_t gfn;		/* Guest frame */
> > +	kvm_pfn_t pfn;		/* Host PFN */
> > +	struct page *page;	/* Host page */
> > +	bool writable;		/* Mapping is writable */
> > +	bool write_attempt;	/* Write access attempted */
> > +	bool valid;		/* This entry contains valid data */
> > +};
> > +
> >  static inline void kvm_s390_fpu_store(struct kvm_run *run)
> >  {
> >  	fpu_stfpc(&run->s.regs.fpc);
> > @@ -464,12 +473,59 @@ int kvm_s390_cpus_from_pv(struct kvm *kvm, u16 *rc, u16 *rrc);
> >  int __kvm_s390_handle_dat_fault(struct kvm_vcpu *vcpu, gfn_t gfn, gpa_t gaddr, unsigned int flags);
> >  int __kvm_s390_mprotect_many(struct gmap *gmap, gpa_t gpa, u8 npages, unsigned int prot,
> >  			     unsigned long bits);
> > +int __kvm_s390_faultin_gfn(struct kvm *kvm, struct guest_fault *f, gfn_t gfn, bool wr);
> >  
> >  static inline int kvm_s390_handle_dat_fault(struct kvm_vcpu *vcpu, gpa_t gaddr, unsigned int flags)
> >  {
> >  	return __kvm_s390_handle_dat_fault(vcpu, gpa_to_gfn(gaddr), gaddr, flags);
> >  }
> >  
> > +static inline void release_faultin_multiple(struct kvm *kvm, struct guest_fault *guest_faults,
> > +					    int n, bool ignore)
> > +{
> > +	int i;
> > +
> > +	for (i = 0; i < n; i++) {
> > +		kvm_release_faultin_page(kvm, guest_faults[i].page, ignore,
> > +					 guest_faults[i].write_attempt);
> > +		guest_faults[i].page = NULL;
> > +	}
> > +}
> > +
> > +static inline bool __kvm_s390_multiple_faults_need_retry(struct kvm *kvm, unsigned long seq,
> > +							 struct guest_fault *guest_faults, int n,
> > +							 bool unsafe)
> > +{
> > +	int i;
> > +
> > +	for (i = 0; i < n; i++) {
> > +		if (!guest_faults[i].valid)
> > +			continue;
> > +		if ((unsafe && mmu_invalidate_retry_gfn_unsafe(kvm, seq, guest_faults[i].gfn)) ||
> > +		    (!unsafe && mmu_invalidate_retry_gfn(kvm, seq, guest_faults[i].gfn))) {
> > +			release_faultin_multiple(kvm, guest_faults, n, true);
> > +			return true;
> > +		}
> > +	}
> > +	return false;
> > +}
> > +
> > +static inline int __kvm_s390_faultin_gfn_range(struct kvm *kvm, struct guest_fault *guest_faults,
> > +					       gfn_t start, int n_pages, bool write_attempt)
> > +{
> > +	int i, rc = 0;
> > +
> > +	for (i = 0; !rc && i < n_pages; i++)
> > +		rc = __kvm_s390_faultin_gfn(kvm, guest_faults + i, start + i, write_attempt);
> > +	return rc;
> > +}
> > +
> > +#define release_faultin_array(kvm, array, ignore) \
> > +	release_faultin_multiple(kvm, array, ARRAY_SIZE(array), ignore)
> > +
> > +#define __kvm_s390_fault_array_needs_retry(kvm, seq, array, unsafe) \
> > +	__kvm_s390_multiple_faults_need_retry(kvm, seq, array, ARRAY_SIZE(array), unsafe)
> > +
> >  /* implemented in diag.c */
> >  int kvm_s390_handle_diag(struct kvm_vcpu *vcpu);
> >  
> > -- 
> > 2.51.0
> >