Merge tag 'efi-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mfleming/efi into x86/efi

Pull EFI virtual mapping changes from Matt Fleming:

* New static EFI runtime services virtual mapping layout which is
groundwork for kexec support on EFI. (Borislav Petkov)

Signed-off-by: Ingo Molnar <[e-mail]>

Documentation/kernel-parameters.txt

···
 	edd=		[EDD]
 			Format: {"off" | "on" | "skip[mbr]"}
 
+	efi=		[EFI]
+			Format: { "old_map" }
+			old_map [X86-64]: switch to the old ioremap-based EFI
+			runtime services mapping. 32-bit still uses this one by
+			default.
+
 	efi_no_storage_paranoia [EFI; X86]
 			Using this parameter you can use more than 50% of
 			your efi variable storage. Use this parameter only if

Documentation/x86/x86_64/mm.txt

···
 Current X86-64 implementations only support 40 bits of address space,
 but we support up to 46 bits. This expands into MBZ space in the page tables.
 
+->trampoline_pgd:
+
+We map EFI runtime services in the aforementioned PGD in the virtual
+range of 64Gb (arbitrarily set, can be raised if needed)
+
+0xffffffef00000000 - 0xffffffff00000000
+
 -Andi Kleen, Jul 2004

arch/x86/include/asm/efi.h

···
 #ifndef _ASM_X86_EFI_H
 #define _ASM_X86_EFI_H
 
+/*
+ * We map the EFI regions needed for runtime services non-contiguously,
+ * with preserved alignment on virtual addresses starting from -4G down
+ * for a total max space of 64G. This way, we provide for stable runtime
+ * services addresses across kernels so that a kexec'd kernel can still
+ * use them.
+ *
+ * This is the main reason why we're doing stable VA mappings for RT
+ * services.
+ *
+ * This flag is used in conjuction with a chicken bit called
+ * "efi=old_map" which can be used as a fallback to the old runtime
+ * services mapping method in case there's some b0rkage with a
+ * particular EFI implementation (haha, it is hard to hold up the
+ * sarcasm here...).
+ */
+#define EFI_OLD_MEMMAP		EFI_ARCH_1
+
 #ifdef CONFIG_X86_32
 
 #define EFI_LOADER_SIGNATURE	"EL32"
···
 	efi_call6((f), (u64)(a1), (u64)(a2), (u64)(a3),		\
 		  (u64)(a4), (u64)(a5), (u64)(a6))
 
+#define _efi_call_virtX(x, f, ...)					\
+({									\
+	efi_status_t __s;						\
+									\
+	efi_sync_low_kernel_mappings();					\
+	preempt_disable();						\
+	__s = efi_call##x((void *)efi.systab->runtime->f, __VA_ARGS__);	\
+	preempt_enable();						\
+	__s;								\
+})
+
 #define efi_call_virt0(f)				\
-	efi_call0((efi.systab->runtime->f))
-#define efi_call_virt1(f, a1)					\
-	efi_call1((efi.systab->runtime->f), (u64)(a1))
-#define efi_call_virt2(f, a1, a2)					\
-	efi_call2((efi.systab->runtime->f), (u64)(a1), (u64)(a2))
-#define efi_call_virt3(f, a1, a2, a3)					\
-	efi_call3((efi.systab->runtime->f), (u64)(a1), (u64)(a2), \
-		  (u64)(a3))
-#define efi_call_virt4(f, a1, a2, a3, a4)				\
-	efi_call4((efi.systab->runtime->f), (u64)(a1), (u64)(a2), \
-		  (u64)(a3), (u64)(a4))
-#define efi_call_virt5(f, a1, a2, a3, a4, a5)				\
-	efi_call5((efi.systab->runtime->f), (u64)(a1), (u64)(a2), \
-		  (u64)(a3), (u64)(a4), (u64)(a5))
-#define efi_call_virt6(f, a1, a2, a3, a4, a5, a6)			\
-	efi_call6((efi.systab->runtime->f), (u64)(a1), (u64)(a2), \
-		  (u64)(a3), (u64)(a4), (u64)(a5), (u64)(a6))
+	_efi_call_virtX(0, f)
+#define efi_call_virt1(f, a1)				\
+	_efi_call_virtX(1, f, (u64)(a1))
+#define efi_call_virt2(f, a1, a2)			\
+	_efi_call_virtX(2, f, (u64)(a1), (u64)(a2))
+#define efi_call_virt3(f, a1, a2, a3)			\
+	_efi_call_virtX(3, f, (u64)(a1), (u64)(a2), (u64)(a3))
+#define efi_call_virt4(f, a1, a2, a3, a4)		\
+	_efi_call_virtX(4, f, (u64)(a1), (u64)(a2), (u64)(a3), (u64)(a4))
+#define efi_call_virt5(f, a1, a2, a3, a4, a5)		\
+	_efi_call_virtX(5, f, (u64)(a1), (u64)(a2), (u64)(a3), (u64)(a4), (u64)(a5))
+#define efi_call_virt6(f, a1, a2, a3, a4, a5, a6)	\
+	_efi_call_virtX(6, f, (u64)(a1), (u64)(a2), (u64)(a3), (u64)(a4), (u64)(a5), (u64)(a6))
 
 extern void __iomem *efi_ioremap(unsigned long addr, unsigned long size,
 				 u32 type, u64 attribute);
···
 
 extern int add_efi_memmap;
 extern unsigned long x86_efi_facility;
+extern struct efi_scratch efi_scratch;
 extern void efi_set_executable(efi_memory_desc_t *md, bool executable);
 extern int efi_memblock_x86_reserve_range(void);
 extern void efi_call_phys_prelog(void);
 extern void efi_call_phys_epilog(void);
 extern void efi_unmap_memmap(void);
 extern void efi_memory_uc(u64 addr, unsigned long size);
+extern void __init efi_map_region(efi_memory_desc_t *md);
+extern void efi_sync_low_kernel_mappings(void);
+extern void efi_setup_page_tables(void);
+extern void __init old_map_region(efi_memory_desc_t *md);
 
 #ifdef CONFIG_EFI
 

arch/x86/include/asm/pgtable_types.h

···
  */
 extern pte_t *lookup_address(unsigned long address, unsigned int *level);
 extern phys_addr_t slow_virt_to_phys(void *__address);
-
+extern int kernel_map_pages_in_pgd(pgd_t *pgd, u64 pfn, unsigned long address,
+				   unsigned numpages, unsigned long page_flags);
 #endif	/* !__ASSEMBLY__ */
 
 #endif /* _ASM_X86_PGTABLE_DEFS_H */

arch/x86/mm/pageattr.c

···
  */
 struct cpa_data {
 	unsigned long	*vaddr;
+	pgd_t		*pgd;
 	pgprot_t	mask_set;
 	pgprot_t	mask_clr;
 	int		numpages;
···
 	return prot;
 }
 
-/*
- * Lookup the page table entry for a virtual address. Return a pointer
- * to the entry and the level of the mapping.
- *
- * Note: We return pud and pmd either when the entry is marked large
- * or when the present bit is not set. Otherwise we would return a
- * pointer to a nonexisting mapping.
- */
-pte_t *lookup_address(unsigned long address, unsigned int *level)
+static pte_t *__lookup_address_in_pgd(pgd_t *pgd, unsigned long address,
+				      unsigned int *level)
 {
-	pgd_t *pgd = pgd_offset_k(address);
 	pud_t *pud;
 	pmd_t *pmd;
 
···
 
 	return pte_offset_kernel(pmd, address);
 }
+
+/*
+ * Lookup the page table entry for a virtual address. Return a pointer
+ * to the entry and the level of the mapping.
+ *
+ * Note: We return pud and pmd either when the entry is marked large
+ * or when the present bit is not set. Otherwise we would return a
+ * pointer to a nonexisting mapping.
+ */
+pte_t *lookup_address(unsigned long address, unsigned int *level)
+{
+        return __lookup_address_in_pgd(pgd_offset_k(address), address, level);
+}
 EXPORT_SYMBOL_GPL(lookup_address);
+
+static pte_t *_lookup_address_cpa(struct cpa_data *cpa, unsigned long address,
+				  unsigned int *level)
+{
+        if (cpa->pgd)
+		return __lookup_address_in_pgd(cpa->pgd + pgd_index(address),
+					       address, level);
+
+        return lookup_address(address, level);
+}
 
 /*
  * This is necessary because __pa() does not work on some
···
 	 * Check for races, another CPU might have split this page
 	 * up already:
 	 */
-	tmp = lookup_address(address, &level);
+	tmp = _lookup_address_cpa(cpa, address, &level);
 	if (tmp != kpte)
 		goto out_unlock;
 
···
 }
 
 static int
-__split_large_page(pte_t *kpte, unsigned long address, struct page *base)
+__split_large_page(struct cpa_data *cpa, pte_t *kpte, unsigned long address,
+		   struct page *base)
 {
 	pte_t *pbase = (pte_t *)page_address(base);
 	unsigned long pfn, pfninc = 1;
···
 	 * Check for races, another CPU might have split this page
 	 * up for us already:
 	 */
-	tmp = lookup_address(address, &level);
+	tmp = _lookup_address_cpa(cpa, address, &level);
 	if (tmp != kpte) {
 		spin_unlock(&pgd_lock);
 		return 1;
···
 	return 0;
 }
 
-static int split_large_page(pte_t *kpte, unsigned long address)
+static int split_large_page(struct cpa_data *cpa, pte_t *kpte,
+			    unsigned long address)
 {
 	struct page *base;
 
···
 	if (!base)
 		return -ENOMEM;
 
-	if (__split_large_page(kpte, address, base))
+	if (__split_large_page(cpa, kpte, address, base))
 		__free_page(base);
 
+	return 0;
+}
+
+static bool try_to_free_pte_page(pte_t *pte)
+{
+	int i;
+
+	for (i = 0; i < PTRS_PER_PTE; i++)
+		if (!pte_none(pte[i]))
+			return false;
+
+	free_page((unsigned long)pte);
+	return true;
+}
+
+static bool try_to_free_pmd_page(pmd_t *pmd)
+{
+	int i;
+
+	for (i = 0; i < PTRS_PER_PMD; i++)
+		if (!pmd_none(pmd[i]))
+			return false;
+
+	free_page((unsigned long)pmd);
+	return true;
+}
+
+static bool unmap_pte_range(pmd_t *pmd, unsigned long start, unsigned long end)
+{
+	pte_t *pte = pte_offset_kernel(pmd, start);
+
+	while (start < end) {
+		set_pte(pte, __pte(0));
+
+		start += PAGE_SIZE;
+		pte++;
+	}
+
+	if (try_to_free_pte_page((pte_t *)pmd_page_vaddr(*pmd))) {
+		pmd_clear(pmd);
+		return true;
+	}
+	return false;
+}
+
+static void __unmap_pmd_range(pud_t *pud, pmd_t *pmd,
+			      unsigned long start, unsigned long end)
+{
+	if (unmap_pte_range(pmd, start, end))
+		if (try_to_free_pmd_page((pmd_t *)pud_page_vaddr(*pud)))
+			pud_clear(pud);
+}
+
+static void unmap_pmd_range(pud_t *pud, unsigned long start, unsigned long end)
+{
+	pmd_t *pmd = pmd_offset(pud, start);
+
+	/*
+	 * Not on a 2MB page boundary?
+	 */
+	if (start & (PMD_SIZE - 1)) {
+		unsigned long next_page = (start + PMD_SIZE) & PMD_MASK;
+		unsigned long pre_end = min_t(unsigned long, end, next_page);
+
+		__unmap_pmd_range(pud, pmd, start, pre_end);
+
+		start = pre_end;
+		pmd++;
+	}
+
+	/*
+	 * Try to unmap in 2M chunks.
+	 */
+	while (end - start >= PMD_SIZE) {
+		if (pmd_large(*pmd))
+			pmd_clear(pmd);
+		else
+			__unmap_pmd_range(pud, pmd, start, start + PMD_SIZE);
+
+		start += PMD_SIZE;
+		pmd++;
+	}
+
+	/*
+	 * 4K leftovers?
+	 */
+	if (start < end)
+		return __unmap_pmd_range(pud, pmd, start, end);
+
+	/*
+	 * Try again to free the PMD page if haven't succeeded above.
+	 */
+	if (!pud_none(*pud))
+		if (try_to_free_pmd_page((pmd_t *)pud_page_vaddr(*pud)))
+			pud_clear(pud);
+}
+
+static void unmap_pud_range(pgd_t *pgd, unsigned long start, unsigned long end)
+{
+	pud_t *pud = pud_offset(pgd, start);
+
+	/*
+	 * Not on a GB page boundary?
+	 */
+	if (start & (PUD_SIZE - 1)) {
+		unsigned long next_page = (start + PUD_SIZE) & PUD_MASK;
+		unsigned long pre_end	= min_t(unsigned long, end, next_page);
+
+		unmap_pmd_range(pud, start, pre_end);
+
+		start = pre_end;
+		pud++;
+	}
+
+	/*
+	 * Try to unmap in 1G chunks?
+	 */
+	while (end - start >= PUD_SIZE) {
+
+		if (pud_large(*pud))
+			pud_clear(pud);
+		else
+			unmap_pmd_range(pud, start, start + PUD_SIZE);
+
+		start += PUD_SIZE;
+		pud++;
+	}
+
+	/*
+	 * 2M leftovers?
+	 */
+	if (start < end)
+		unmap_pmd_range(pud, start, end);
+
+	/*
+	 * No need to try to free the PUD page because we'll free it in
+	 * populate_pgd's error path
+	 */
+}
+
+static int alloc_pte_page(pmd_t *pmd)
+{
+	pte_t *pte = (pte_t *)get_zeroed_page(GFP_KERNEL | __GFP_NOTRACK);
+	if (!pte)
+		return -1;
+
+	set_pmd(pmd, __pmd(__pa(pte) | _KERNPG_TABLE));
+	return 0;
+}
+
+static int alloc_pmd_page(pud_t *pud)
+{
+	pmd_t *pmd = (pmd_t *)get_zeroed_page(GFP_KERNEL | __GFP_NOTRACK);
+	if (!pmd)
+		return -1;
+
+	set_pud(pud, __pud(__pa(pmd) | _KERNPG_TABLE));
+	return 0;
+}
+
+static void populate_pte(struct cpa_data *cpa,
+			 unsigned long start, unsigned long end,
+			 unsigned num_pages, pmd_t *pmd, pgprot_t pgprot)
+{
+	pte_t *pte;
+
+	pte = pte_offset_kernel(pmd, start);
+
+	while (num_pages-- && start < end) {
+
+		/* deal with the NX bit */
+		if (!(pgprot_val(pgprot) & _PAGE_NX))
+			cpa->pfn &= ~_PAGE_NX;
+
+		set_pte(pte, pfn_pte(cpa->pfn >> PAGE_SHIFT, pgprot));
+
+		start	 += PAGE_SIZE;
+		cpa->pfn += PAGE_SIZE;
+		pte++;
+	}
+}
+
+static int populate_pmd(struct cpa_data *cpa,
+			unsigned long start, unsigned long end,
+			unsigned num_pages, pud_t *pud, pgprot_t pgprot)
+{
+	unsigned int cur_pages = 0;
+	pmd_t *pmd;
+
+	/*
+	 * Not on a 2M boundary?
+	 */
+	if (start & (PMD_SIZE - 1)) {
+		unsigned long pre_end = start + (num_pages << PAGE_SHIFT);
+		unsigned long next_page = (start + PMD_SIZE) & PMD_MASK;
+
+		pre_end   = min_t(unsigned long, pre_end, next_page);
+		cur_pages = (pre_end - start) >> PAGE_SHIFT;
+		cur_pages = min_t(unsigned int, num_pages, cur_pages);
+
+		/*
+		 * Need a PTE page?
+		 */
+		pmd = pmd_offset(pud, start);
+		if (pmd_none(*pmd))
+			if (alloc_pte_page(pmd))
+				return -1;
+
+		populate_pte(cpa, start, pre_end, cur_pages, pmd, pgprot);
+
+		start = pre_end;
+	}
+
+	/*
+	 * We mapped them all?
+	 */
+	if (num_pages == cur_pages)
+		return cur_pages;
+
+	while (end - start >= PMD_SIZE) {
+
+		/*
+		 * We cannot use a 1G page so allocate a PMD page if needed.
+		 */
+		if (pud_none(*pud))
+			if (alloc_pmd_page(pud))
+				return -1;
+
+		pmd = pmd_offset(pud, start);
+
+		set_pmd(pmd, __pmd(cpa->pfn | _PAGE_PSE | massage_pgprot(pgprot)));
+
+		start	  += PMD_SIZE;
+		cpa->pfn  += PMD_SIZE;
+		cur_pages += PMD_SIZE >> PAGE_SHIFT;
+	}
+
+	/*
+	 * Map trailing 4K pages.
+	 */
+	if (start < end) {
+		pmd = pmd_offset(pud, start);
+		if (pmd_none(*pmd))
+			if (alloc_pte_page(pmd))
+				return -1;
+
+		populate_pte(cpa, start, end, num_pages - cur_pages,
+			     pmd, pgprot);
+	}
+	return num_pages;
+}
+
+static int populate_pud(struct cpa_data *cpa, unsigned long start, pgd_t *pgd,
+			pgprot_t pgprot)
+{
+	pud_t *pud;
+	unsigned long end;
+	int cur_pages = 0;
+
+	end = start + (cpa->numpages << PAGE_SHIFT);
+
+	/*
+	 * Not on a Gb page boundary? => map everything up to it with
+	 * smaller pages.
+	 */
+	if (start & (PUD_SIZE - 1)) {
+		unsigned long pre_end;
+		unsigned long next_page = (start + PUD_SIZE) & PUD_MASK;
+
+		pre_end   = min_t(unsigned long, end, next_page);
+		cur_pages = (pre_end - start) >> PAGE_SHIFT;
+		cur_pages = min_t(int, (int)cpa->numpages, cur_pages);
+
+		pud = pud_offset(pgd, start);
+
+		/*
+		 * Need a PMD page?
+		 */
+		if (pud_none(*pud))
+			if (alloc_pmd_page(pud))
+				return -1;
+
+		cur_pages = populate_pmd(cpa, start, pre_end, cur_pages,
+					 pud, pgprot);
+		if (cur_pages < 0)
+			return cur_pages;
+
+		start = pre_end;
+	}
+
+	/* We mapped them all? */
+	if (cpa->numpages == cur_pages)
+		return cur_pages;
+
+	pud = pud_offset(pgd, start);
+
+	/*
+	 * Map everything starting from the Gb boundary, possibly with 1G pages
+	 */
+	while (end - start >= PUD_SIZE) {
+		set_pud(pud, __pud(cpa->pfn | _PAGE_PSE | massage_pgprot(pgprot)));
+
+		start	  += PUD_SIZE;
+		cpa->pfn  += PUD_SIZE;
+		cur_pages += PUD_SIZE >> PAGE_SHIFT;
+		pud++;
+	}
+
+	/* Map trailing leftover */
+	if (start < end) {
+		int tmp;
+
+		pud = pud_offset(pgd, start);
+		if (pud_none(*pud))
+			if (alloc_pmd_page(pud))
+				return -1;
+
+		tmp = populate_pmd(cpa, start, end, cpa->numpages - cur_pages,
+				   pud, pgprot);
+		if (tmp < 0)
+			return cur_pages;
+
+		cur_pages += tmp;
+	}
+	return cur_pages;
+}
+
+/*
+ * Restrictions for kernel page table do not necessarily apply when mapping in
+ * an alternate PGD.
+ */
+static int populate_pgd(struct cpa_data *cpa, unsigned long addr)
+{
+	pgprot_t pgprot = __pgprot(_KERNPG_TABLE);
+	bool allocd_pgd = false;
+	pgd_t *pgd_entry;
+	pud_t *pud = NULL;	/* shut up gcc */
+	int ret;
+
+	pgd_entry = cpa->pgd + pgd_index(addr);
+
+	/*
+	 * Allocate a PUD page and hand it down for mapping.
+	 */
+	if (pgd_none(*pgd_entry)) {
+		pud = (pud_t *)get_zeroed_page(GFP_KERNEL | __GFP_NOTRACK);
+		if (!pud)
+			return -1;
+
+		set_pgd(pgd_entry, __pgd(__pa(pud) | _KERNPG_TABLE));
+		allocd_pgd = true;
+	}
+
+	pgprot_val(pgprot) &= ~pgprot_val(cpa->mask_clr);
+	pgprot_val(pgprot) |=  pgprot_val(cpa->mask_set);
+
+	ret = populate_pud(cpa, addr, pgd_entry, pgprot);
+	if (ret < 0) {
+		unmap_pud_range(pgd_entry, addr,
+				addr + (cpa->numpages << PAGE_SHIFT));
+
+		if (allocd_pgd) {
+			/*
+			 * If I allocated this PUD page, I can just as well
+			 * free it in this error path.
+			 */
+			pgd_clear(pgd_entry);
+			free_page((unsigned long)pud);
+		}
+		return ret;
+	}
+	cpa->numpages = ret;
 	return 0;
 }
 
 static int __cpa_process_fault(struct cpa_data *cpa, unsigned long vaddr,
 			       int primary)
 {
+	if (cpa->pgd)
+		return populate_pgd(cpa, vaddr);
+
 	/*
 	 * Ignore all non primary paths.
 	 */
···
 	else
 		address = *cpa->vaddr;
 repeat:
-	kpte = lookup_address(address, &level);
+	kpte = _lookup_address_cpa(cpa, address, &level);
 	if (!kpte)
 		return __cpa_process_fault(cpa, address, primary);
 
···
 	/*
 	 * We have to split the large page:
 	 */
-	err = split_large_page(kpte, address);
+	err = split_large_page(cpa, kpte, address);
 	if (!err) {
 		/*
 	 	 * Do a global flush tlb after splitting the large page
···
 	struct cpa_data cpa;
 	int ret, cache, checkalias;
 	unsigned long baddr = 0;
+
+	memset(&cpa, 0, sizeof(cpa));
 
 	/*
 	 * Check, if we are requested to change a not supported
···
 {
 	unsigned long tempaddr = (unsigned long) page_address(page);
 	struct cpa_data cpa = { .vaddr = &tempaddr,
+				.pgd = NULL,
 				.numpages = numpages,
 				.mask_set = __pgprot(_PAGE_PRESENT | _PAGE_RW),
 				.mask_clr = __pgprot(0),
···
 {
 	unsigned long tempaddr = (unsigned long) page_address(page);
 	struct cpa_data cpa = { .vaddr = &tempaddr,
+				.pgd = NULL,
 				.numpages = numpages,
 				.mask_set = __pgprot(0),
 				.mask_clr = __pgprot(_PAGE_PRESENT | _PAGE_RW),
···
 #endif /* CONFIG_HIBERNATION */
 
 #endif /* CONFIG_DEBUG_PAGEALLOC */
+
+int kernel_map_pages_in_pgd(pgd_t *pgd, u64 pfn, unsigned long address,
+			    unsigned numpages, unsigned long page_flags)
+{
+	int retval = -EINVAL;
+
+	struct cpa_data cpa = {
+		.vaddr = &address,
+		.pfn = pfn,
+		.pgd = pgd,
+		.numpages = numpages,
+		.mask_set = __pgprot(0),
+		.mask_clr = __pgprot(0),
+		.flags = 0,
+	};
+
+	if (!(__supported_pte_mask & _PAGE_NX))
+		goto out;
+
+	if (!(page_flags & _PAGE_NX))
+		cpa.mask_clr = __pgprot(_PAGE_NX);
+
+	cpa.mask_set = __pgprot(_PAGE_PRESENT | page_flags);
+
+	retval = __change_page_attr_set_clr(&cpa, 0);
+	__flush_tlb_all();
+
+out:
+	return retval;
+}
 
 /*
  * The testcases use internal knowledge of the implementation that shouldn't

arch/x86/platform/efi/efi.c

···
  *	Bibo Mao <[e-mail]>
  *	Chandramouli Narayanan <[e-mail]>
  *	Huang Ying <[e-mail]>
+ * Copyright (C) 2013 SuSE Labs
+ *	Borislav Petkov <[e-mail]> - runtime services VA mapping
  *
  * Copied from efi_32.c to eliminate the duplicated code between EFI
  * 32/64 support code. --ying 2007-10-26
···
 #include <asm/x86_init.h>
 #include <asm/rtc.h>
 
-#define EFI_DEBUG	1
+#define EFI_DEBUG
 
 #define EFI_MIN_RESERVE 5120
 
···
 	return 0;
 }
 
-#if EFI_DEBUG
 static void __init print_efi_memmap(void)
 {
+#ifdef EFI_DEBUG
 	efi_memory_desc_t *md;
 	void *p;
 	int i;
···
 			md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT),
 			(md->num_pages >> (20 - EFI_PAGE_SHIFT)));
 	}
-}
 #endif  /*  EFI_DEBUG  */
+}
 
 void __init efi_reserve_boot_services(void)
 {
···
 		x86_platform.set_wallclock = efi_set_rtc_mmss;
 	}
 #endif
-
-#if EFI_DEBUG
 	print_efi_memmap();
-#endif
 }
 
 void __init efi_late_init(void)
···
 	set_memory_uc(addr, npages);
 }
 
+void __init old_map_region(efi_memory_desc_t *md)
+{
+	u64 start_pfn, end_pfn, end;
+	unsigned long size;
+	void *va;
+
+	start_pfn = PFN_DOWN(md->phys_addr);
+	size	  = md->num_pages << PAGE_SHIFT;
+	end	  = md->phys_addr + size;
+	end_pfn   = PFN_UP(end);
+
+	if (pfn_range_is_mapped(start_pfn, end_pfn)) {
+		va = __va(md->phys_addr);
+
+		if (!(md->attribute & EFI_MEMORY_WB))
+			efi_memory_uc((u64)(unsigned long)va, size);
+	} else
+		va = efi_ioremap(md->phys_addr, size,
+				 md->type, md->attribute);
+
+	md->virt_addr = (u64) (unsigned long) va;
+	if (!va)
+		pr_err("ioremap of 0x%llX failed!\n",
+		       (unsigned long long)md->phys_addr);
+}
+
 /*
  * This function will switch the EFI runtime services to virtual mode.
- * Essentially, look through the EFI memmap and map every region that
- * has the runtime attribute bit set in its memory descriptor and update
- * that memory descriptor with the virtual address obtained from ioremap().
- * This enables the runtime services to be called without having to
+ * Essentially, we look through the EFI memmap and map every region that
+ * has the runtime attribute bit set in its memory descriptor into the
+ * ->trampoline_pgd page table using a top-down VA allocation scheme.
+ *
+ * The old method which used to update that memory descriptor with the
+ * virtual address obtained from ioremap() is still supported when the
+ * kernel is booted with efi=old_map on its command line. Same old
+ * method enabled the runtime services to be called without having to
  * thunk back into physical mode for every invocation.
+ *
+ * The new method does a pagetable switch in a preemption-safe manner
+ * so that we're in a different address space when calling a runtime
+ * function. For function arguments passing we do copy the PGDs of the
+ * kernel page table into ->trampoline_pgd prior to each call.
  */
 void __init efi_enter_virtual_mode(void)
 {
 	efi_memory_desc_t *md, *prev_md = NULL;
-	efi_status_t status;
+	void *p, *new_memmap = NULL;
 	unsigned long size;
-	u64 end, systab, start_pfn, end_pfn;
-	void *p, *va, *new_memmap = NULL;
+	efi_status_t status;
+	u64 end, systab;
 	int count = 0;
 
 	efi.systab = NULL;
···
 	 * We don't do virtual mode, since we don't do runtime services, on
 	 * non-native EFI
 	 */
-
 	if (!efi_is_native()) {
 		efi_unmap_memmap();
 		return;
···
 			continue;
 		}
 		prev_md = md;
+
 	}
 
 	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
···
 				continue;
 		}
 
+		efi_map_region(md);
+
 		size = md->num_pages << EFI_PAGE_SHIFT;
 		end = md->phys_addr + size;
-
-		start_pfn = PFN_DOWN(md->phys_addr);
-		end_pfn = PFN_UP(end);
-		if (pfn_range_is_mapped(start_pfn, end_pfn)) {
-			va = __va(md->phys_addr);
-
-			if (!(md->attribute & EFI_MEMORY_WB))
-				efi_memory_uc((u64)(unsigned long)va, size);
-		} else
-			va = efi_ioremap(md->phys_addr, size,
-					 md->type, md->attribute);
-
-		md->virt_addr = (u64) (unsigned long) va;
-
-		if (!va) {
-			pr_err("ioremap of 0x%llX failed!\n",
-			       (unsigned long long)md->phys_addr);
-			continue;
-		}
 
 		systab = (u64) (unsigned long) efi_phys.systab;
 		if (md->phys_addr <= systab && systab < end) {
 			systab += md->virt_addr - md->phys_addr;
+
 			efi.systab = (efi_system_table_t *) (unsigned long) systab;
 		}
+
 		new_memmap = krealloc(new_memmap,
 				      (count + 1) * memmap.desc_size,
 				      GFP_KERNEL);
+		if (!new_memmap)
+			goto err_out;
+
 		memcpy(new_memmap + (count * memmap.desc_size), md,
 		       memmap.desc_size);
 		count++;
 	}
 
 	BUG_ON(!efi.systab);
+
+	efi_setup_page_tables();
+	efi_sync_low_kernel_mappings();
 
 	status = phys_efi_set_virtual_address_map(
 		memmap.desc_size * count,
···
 	efi.query_variable_info = virt_efi_query_variable_info;
 	efi.update_capsule = virt_efi_update_capsule;
 	efi.query_capsule_caps = virt_efi_query_capsule_caps;
-	if (__supported_pte_mask & _PAGE_NX)
+
+	if (efi_enabled(EFI_OLD_MEMMAP) && (__supported_pte_mask & _PAGE_NX))
 		runtime_code_page_mkexec();
 
 	kfree(new_memmap);
···
 			 EFI_VARIABLE_BOOTSERVICE_ACCESS |
 			 EFI_VARIABLE_RUNTIME_ACCESS,
 			 0, NULL);
+
+	return;
+
+ err_out:
+	pr_err("Error reallocating memory, EFI runtime non-functional!\n");
 }
 
 /*
···
 	return EFI_SUCCESS;
 }
 EXPORT_SYMBOL_GPL(efi_query_variable_store);
+
+static int __init parse_efi_cmdline(char *str)
+{
+	if (*str == '=')
+		str++;
+
+	if (!strncmp(str, "old_map", 7))
+		set_bit(EFI_OLD_MEMMAP, &x86_efi_facility);
+
+	return 0;
+}
+early_param("efi", parse_efi_cmdline);

arch/x86/platform/efi/efi_32.c

···
  * claim EFI runtime service handler exclusively and to duplicate a memory in
  * low memory space say 0 - 3G.
  */
-
 static unsigned long efi_rt_eflags;
+
+void efi_sync_low_kernel_mappings(void) {}
+void efi_setup_page_tables(void) {}
+
+void __init efi_map_region(efi_memory_desc_t *md)
+{
+	old_map_region(md);
+}
 
 void efi_call_phys_prelog(void)
 {

arch/x86/platform/efi/efi_64.c

···
 #include <asm/efi.h>
 #include <asm/cacheflush.h>
 #include <asm/fixmap.h>
+#include <asm/realmode.h>
 
 static pgd_t *save_pgd __initdata;
 static unsigned long efi_flags __initdata;
+
+/*
+ * We allocate runtime services regions bottom-up, starting from -4G, i.e.
+ * 0xffff_ffff_0000_0000 and limit EFI VA mapping space to 64G.
+ */
+static u64 efi_va	= -4 * (1UL << 30);
+#define EFI_VA_END	(-68 * (1UL << 30))
+
+/*
+ * Scratch space used for switching the pagetable in the EFI stub
+ */
+struct efi_scratch {
+	u64 r15;
+	u64 prev_cr3;
+	pgd_t *efi_pgt;
+	bool use_pgd;
+};
 
 static void __init early_code_mapping_set_exec(int executable)
 {
···
 	int pgd;
 	int n_pgds;
 
+	if (!efi_enabled(EFI_OLD_MEMMAP))
+		return;
+
 	early_code_mapping_set_exec(1);
 	local_irq_save(efi_flags);
 
···
 	 */
 	int pgd;
 	int n_pgds = DIV_ROUND_UP((max_pfn << PAGE_SHIFT) , PGDIR_SIZE);
+
+	if (!efi_enabled(EFI_OLD_MEMMAP))
+		return;
+
 	for (pgd = 0; pgd < n_pgds; pgd++)
 		set_pgd(pgd_offset_k(pgd * PGDIR_SIZE), save_pgd[pgd]);
 	kfree(save_pgd);
 	__flush_tlb_all();
 	local_irq_restore(efi_flags);
 	early_code_mapping_set_exec(0);
+}
+
+/*
+ * Add low kernel mappings for passing arguments to EFI functions.
+ */
+void efi_sync_low_kernel_mappings(void)
+{
+	unsigned num_pgds;
+	pgd_t *pgd = (pgd_t *)__va(real_mode_header->trampoline_pgd);
+
+	if (efi_enabled(EFI_OLD_MEMMAP))
+		return;
+
+	num_pgds = pgd_index(MODULES_END - 1) - pgd_index(PAGE_OFFSET);
+
+	memcpy(pgd + pgd_index(PAGE_OFFSET),
+		init_mm.pgd + pgd_index(PAGE_OFFSET),
+		sizeof(pgd_t) * num_pgds);
+}
+
+void efi_setup_page_tables(void)
+{
+	efi_scratch.efi_pgt = (pgd_t *)(unsigned long)real_mode_header->trampoline_pgd;
+
+	if (!efi_enabled(EFI_OLD_MEMMAP))
+		efi_scratch.use_pgd = true;
+}
+
+static void __init __map_region(efi_memory_desc_t *md, u64 va)
+{
+	pgd_t *pgd = (pgd_t *)__va(real_mode_header->trampoline_pgd);
+	unsigned long pf = 0, size;
+	u64 end;
+
+	if (!(md->attribute & EFI_MEMORY_WB))
+		pf |= _PAGE_PCD;
+
+	size = md->num_pages << PAGE_SHIFT;
+	end  = va + size;
+
+	if (kernel_map_pages_in_pgd(pgd, md->phys_addr, va, md->num_pages, pf))
+		pr_warn("Error mapping PA 0x%llx -> VA 0x%llx!\n",
+			   md->phys_addr, va);
+}
+
+void __init efi_map_region(efi_memory_desc_t *md)
+{
+	unsigned long size = md->num_pages << PAGE_SHIFT;
+	u64 pa = md->phys_addr;
+
+	if (efi_enabled(EFI_OLD_MEMMAP))
+		return old_map_region(md);
+
+	/*
+	 * Make sure the 1:1 mappings are present as a catch-all for b0rked
+	 * firmware which doesn't update all internal pointers after switching
+	 * to virtual mode and would otherwise crap on us.
+	 */
+	__map_region(md, md->phys_addr);
+
+	efi_va -= size;
+
+	/* Is PA 2M-aligned? */
+	if (!(pa & (PMD_SIZE - 1))) {
+		efi_va &= PMD_MASK;
+	} else {
+		u64 pa_offset = pa & (PMD_SIZE - 1);
+		u64 prev_va = efi_va;
+
+		/* get us the same offset within this 2M page */
+		efi_va = (efi_va & PMD_MASK) + pa_offset;
+
+		if (efi_va > prev_va)
+			efi_va -= PMD_SIZE;
+	}
+
+	if (efi_va < EFI_VA_END) {
+		pr_warn(FW_WARN "VA address range overflow!\n");
+		return;
+	}
+
+	/* Do the VA map */
+	__map_region(md, efi_va);
+	md->virt_addr = efi_va;
 }
 
 void __iomem *__init efi_ioremap(unsigned long phys_addr, unsigned long size,

arch/x86/platform/efi/efi_stub_64.S

···
 	mov %rsi, %cr0;			\
 	mov (%rsp), %rsp
 
+	/* stolen from gcc */
+	.macro FLUSH_TLB_ALL
+	movq %r15, efi_scratch(%rip)
+	movq %r14, efi_scratch+8(%rip)
+	movq %cr4, %r15
+	movq %r15, %r14
+	andb $0x7f, %r14b
+	movq %r14, %cr4
+	movq %r15, %cr4
+	movq efi_scratch+8(%rip), %r14
+	movq efi_scratch(%rip), %r15
+	.endm
+
+	.macro SWITCH_PGT
+	cmpb $0, efi_scratch+24(%rip)
+	je 1f
+	movq %r15, efi_scratch(%rip)		# r15
+	# save previous CR3
+	movq %cr3, %r15
+	movq %r15, efi_scratch+8(%rip)		# prev_cr3
+	movq efi_scratch+16(%rip), %r15		# EFI pgt
+	movq %r15, %cr3
+	1:
+	.endm
+
+	.macro RESTORE_PGT
+	cmpb $0, efi_scratch+24(%rip)
+	je 2f
+	movq efi_scratch+8(%rip), %r15
+	movq %r15, %cr3
+	movq efi_scratch(%rip), %r15
+	FLUSH_TLB_ALL
+	2:
+	.endm
+
 ENTRY(efi_call0)
 	SAVE_XMM
 	subq $32, %rsp
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $32, %rsp
 	RESTORE_XMM
 	ret
···
 	SAVE_XMM
 	subq $32, %rsp
 	mov  %rsi, %rcx
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $32, %rsp
 	RESTORE_XMM
 	ret
···
 	SAVE_XMM
 	subq $32, %rsp
 	mov  %rsi, %rcx
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $32, %rsp
 	RESTORE_XMM
 	ret
···
 	subq $32, %rsp
 	mov  %rcx, %r8
 	mov  %rsi, %rcx
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $32, %rsp
 	RESTORE_XMM
 	ret
···
 	mov %r8, %r9
 	mov %rcx, %r8
 	mov %rsi, %rcx
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $32, %rsp
 	RESTORE_XMM
 	ret
···
 	mov %r8, %r9
 	mov %rcx, %r8
 	mov %rsi, %rcx
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $48, %rsp
 	RESTORE_XMM
 	ret
···
 	mov %r8, %r9
 	mov %rcx, %r8
 	mov %rsi, %rcx
+	SWITCH_PGT
 	call *%rdi
+	RESTORE_PGT
 	addq $48, %rsp
 	RESTORE_XMM
 	ret
 ENDPROC(efi_call6)
+
+	.data
+ENTRY(efi_scratch)
+	.fill 3,8,0
+	.byte 0

include/linux/efi.h

···
 #define EFI_RUNTIME_SERVICES	3	/* Can we use runtime services? */
 #define EFI_MEMMAP		4	/* Can we use EFI memory map? */
 #define EFI_64BIT		5	/* Is the firmware 64-bit? */
+#define EFI_ARCH_1		6	/* First arch-specific bit */
 
 #ifdef CONFIG_EFI
 # ifdef CONFIG_X86