Re: [PATCH v4 12/16] x86/virt/tdx: Add helpers to allow for pre-allocating pages

["Edgecombe, Rick P" <rick.p.edgecombe@intel.com>]

On Wed, 2025-11-26 at 11:40 +0800, Binbin Wu wrote:
> 
> On 11/21/2025 8:51 AM, Rick Edgecombe wrote:
> > In the KVM fault path page, tables and private pages need to be
> "In the KVM fault path page, tables ..." should be
> "In the KVM fault path, page tables ..."

Right, thanks.

> 
> > installed under a spin lock. This means that the operations around
> > installing PAMT pages for them will not be able to allocate pages.
> > 
> [...]
> > @@ -141,7 +142,46 @@ int tdx_guest_keyid_alloc(void);
> >    u32 tdx_get_nr_guest_keyids(void);
> >    void tdx_guest_keyid_free(unsigned int keyid);
> >    
> > -int tdx_pamt_get(struct page *page);
> > +int tdx_dpamt_entry_pages(void);
> > +
> > +/*
> > + * Simple structure for pre-allocating Dynamic
> > + * PAMT pages outside of locks.
> 
> It's not just for Dynamic PAMT pages, but also external page table pages.

I'll adjust.

> 
> > + */
> > +struct tdx_prealloc {
> > +	struct list_head page_list;
> > +	int cnt;
> > +};
> > +
> > +static inline struct page *get_tdx_prealloc_page(struct tdx_prealloc *prealloc)
> > +{
> > +	struct page *page;
> > +
> > +	page = list_first_entry_or_null(&prealloc->page_list, struct page, lru);
> > +	if (page) {
> > +		list_del(&page->lru);
> > +		prealloc->cnt--;
> > +	}
> > +
> > +	return page;
> > +}
> > +
> > +static inline int topup_tdx_prealloc_page(struct tdx_prealloc *prealloc, unsigned int min_size)
> > +{
> > +	while (prealloc->cnt < min_size) {
> > +		struct page *page = alloc_page(GFP_KERNEL_ACCOUNT);
> > +
> > +		if (!page)
> > +			return -ENOMEM;
> > +
> > +		list_add(&page->lru, &prealloc->page_list);
> > +		prealloc->cnt++;
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +int tdx_pamt_get(struct page *page, struct tdx_prealloc *prealloc);
> >    void tdx_pamt_put(struct page *page);
> >    
> >    struct page *tdx_alloc_page(void);
> > @@ -219,6 +259,7 @@ static inline int tdx_enable(void)  { return -ENODEV; }
> >    static inline u32 tdx_get_nr_guest_keyids(void) { return 0; }
> >    static inline const char *tdx_dump_mce_info(struct mce *m) { return NULL; }
> >    static inline const struct tdx_sys_info *tdx_get_sysinfo(void) { return NULL; }
> > +static inline int tdx_dpamt_entry_pages(void) { return 0; }
> >    #endif	/* CONFIG_INTEL_TDX_HOST */
> >    
> >    #ifdef CONFIG_KEXEC_CORE
> > diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c
> > index 260bb0e6eb44..61a058a8f159 100644
> > --- a/arch/x86/kvm/vmx/tdx.c
> > +++ b/arch/x86/kvm/vmx/tdx.c
> > @@ -1644,23 +1644,34 @@ static int tdx_mem_page_add(struct kvm *kvm, gfn_t gfn, enum pg_level level,
> >    
> >    static void *tdx_alloc_external_fault_cache(struct kvm_vcpu *vcpu)
> >    {
> > -	struct vcpu_tdx *tdx = to_tdx(vcpu);
> > +	struct page *page = get_tdx_prealloc_page(&to_tdx(vcpu)->prealloc);
> >    
> > -	return kvm_mmu_memory_cache_alloc(&tdx->mmu_external_spt_cache);
> > +	if (WARN_ON_ONCE(!page))
> > +		return (void *)__get_free_page(GFP_ATOMIC | __GFP_ACCOUNT);
> 
> kvm_mmu_memory_cache_alloc() calls BUG_ON() if the atomic allocation failed.
> Do we want to follow?

Ignoring, per your other comment. But we don't want to add BUG_ON()s in general.
> 
> > +
> > +	return page_address(page);
> >    }
> >    
> >    static int tdx_topup_external_fault_cache(struct kvm_vcpu *vcpu, unsigned int cnt)
> >    {
> > -	struct vcpu_tdx *tdx = to_tdx(vcpu);
> > +	struct tdx_prealloc *prealloc = &to_tdx(vcpu)->prealloc;
> > +	int min_fault_cache_size;
> >    
> > -	return kvm_mmu_topup_memory_cache(&tdx->mmu_external_spt_cache, cnt);
> > +	/* External page tables */
> > +	min_fault_cache_size = cnt;
> > +	/* Dynamic PAMT pages (if enabled) */
> > +	min_fault_cache_size += tdx_dpamt_entry_pages() * PT64_ROOT_MAX_LEVEL;
> 
> Is the value PT64_ROOT_MAX_LEVEL intended, since dynamic PAMT pages are only
> needed for 4KB level?

I'm not sure I follow. We need DPAMT backing for each S-EPT page table.