arm64: stable UEFI mappings for kexec

Currently, swapper_pg_dir and idmap_pg_dir share the init_mm mm_struct instance. To allow the introduction of other pg_dir instances, for instance, for UEFI's mapping of Runtime Services, make the struct_mm instance an explicit argument that gets passed down to the pmd and pte instantiation functions. Note that the consumers (pmd_populate/pgd_populate) of the mm_struct argument don't actually inspect it, but let's fix it for correctness' sake.

Signed-off-by Ard Biesheuvel <[email protected]>

For UEFI, we need to install the memory mappings used for Runtime Services in a dedicated set of page tables. Add create_pgd_mapping(), which allows us to allocate and install those page table entries early. This also adds a 'map_xn' option, that creates regions with the PXN and UXN bits set.

Signed-off-by Ard Biesheuvel <[email protected]>

---
 arch/arm64/include/asm/mmu.h |  3 +++
 arch/arm64/mm/mmu.c          | 28 ++++++++++++++++++++--------
 2 files changed, 23 insertions(+), 8 deletions(-)

arch/arm64/include/asm/mmu.h +0 -0

index c2f006c48bdb..bcf166043a8b 100644

arch/arm64/include/asm/mmu.h +0 -0

 #endif

arch/arm64/mm/mmu.c +1 -1

index 7eaa6a8c8467..f7d17a5a1f56 100644
@@ -157,7 +157,7 @@ static void __init alloc_init_pte(pmd_t *pmd, unsigned long addr,
 
 static void __init alloc_init_pmd(struct mm_struct *mm, pud_t *pud,
 				  unsigned long addr, unsigned long end,
-				  phys_addr_t phys, int map_io)
+				  phys_addr_t phys, int map_io, int map_xn)
 {
 	pmd_t *pmd;
 	unsigned long next;

arch/arm64/mm/mmu.c +1 -1

@@ -203,7 +206,7 @@ static void __init alloc_init_pmd(struct mm_struct *mm, pud_t *pud,
 
 static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
 				  unsigned long addr, unsigned long end,
-				  unsigned long phys, int map_io)
+				  unsigned long phys, int map_io, int map_xn)
 {
 	pud_t *pud;
 	unsigned long next;

arch/arm64/mm/mmu.c +14 -5

 		    ((addr | next | phys) & ~PUD_MASK) == 0) {
 			pud_t old_pud = *pud;
 			set_pud(pud, __pud(phys | PROT_SECT_NORMAL_EXEC));

@@ -239,7 +242,8 @@ static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
 				flush_tlb_all();
 			}
 		} else {
-			alloc_init_pmd(mm, pud, addr, next, phys, map_io);
+			alloc_init_pmd(mm, pud, addr, next, phys, map_io,
+				       map_xn);
 		}
 		phys += next - addr;
 	} while (pud++, addr = next, addr != end);

@@ -251,7 +255,7 @@ static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
  */
 static void __init __create_mapping(struct mm_struct *mm, pgd_t *pgd,
 				    phys_addr_t phys, unsigned long virt,
-				    phys_addr_t size, int map_io)
+				    phys_addr_t size, int map_io, int map_xn)
 {
 	unsigned long addr, length, end, next;
 

@@ -261,7 +265,7 @@ static void __init __create_mapping(struct mm_struct *mm, pgd_t *pgd,
 	end = addr + length;
 	do {
 		next = pgd_addr_end(addr, end);
-		alloc_init_pud(mm, pgd, addr, next, phys, map_io);
+		alloc_init_pud(mm, pgd, addr, next, phys, map_io, map_xn);
 		phys += next - addr;
 	} while (pgd++, addr = next, addr != end);
 }

@@ -275,7 +279,7 @@ static void __init create_mapping(phys_addr_t phys, unsigned long virt,
 		return;
 	}
 	__create_mapping(&init_mm, pgd_offset_k(virt & PAGE_MASK), phys, virt,
-			 size, 0);
+			 size, 0, 0);
 }
 
 void __init create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io)

@@ -285,7 +289,15 @@ void __init create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io)
 		return;
 	}
 	__create_mapping(&init_mm, &idmap_pg_dir[pgd_index(addr)],
-			 addr, addr, size, map_io);
+			 addr, addr, size, map_io, 0);
+}
+
+void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
+			       unsigned long virt, phys_addr_t size,
+			       int map_io, int map_xn)
+{
+	__create_mapping(mm, pgd_offset(mm, virt), phys, virt, size, map_io,
+			 map_xn);
 }
 
 static void __init map_mem(void)

Signature · 1 line

1.8.3.2

Quoted 7 lines · Ard Biesheuvel

For UEFI, we need to install the memory mappings used for Runtime Services in a dedicated set of page tables. Add create_pgd_mapping(), which allows us to allocate and install those page table entries early. This also adds a 'map_xn' option, that creates regions with the PXN and UXN bits set.

Signed-off-by Ard Biesheuvel <[email protected]>

Hi Ard,

Some comments below...

Quoted 10 lines

---
 arch/arm64/include/asm/mmu.h |  3 +++
 arch/arm64/mm/mmu.c          | 28 ++++++++++++++++++++--------
 2 files changed, 23 insertions(+), 8 deletions(-)

arch/arm64/include/asm/mmu.h +0 -0

index c2f006c48bdb..bcf166043a8b 100644

arch/arm64/include/asm/mmu.h extern void __iomem *early_io_map(phys_addr_t phys, unsigned long virt); · line 38

Quoted 6 lines

arch/arm64/include/asm/mmu.h +3 -0

@@ -33,5 +33,8 @@ extern void __iomem *early_io_map(phys_addr_t phys, unsigned long virt);
 extern void init_mem_pgprot(void);
 /* create an identity mapping for memory (or io if map_io is true) */
 extern void create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io);
+extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
+                            unsigned long virt, phys_addr_t size,
+                            int map_io, int map_xn);

I really don't like these map_io and map_xn parameters. Further down we have logic "if (map_io)...", which is out of place, and doesn't match the normal style of mapping page table entries.

Could we not instead have something like:
extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
unsigned long virt, phys_addr_t size, pgprot_t prot_sect, pgprot_t prot_pte);

Then we can remove all the conditional logic for map_io, map_xn?

Quoted 16 lines

arch/arm64/include/asm/mmu.h +0 -0

 #endif

arch/arm64/mm/mmu.c +1 -1

index 7eaa6a8c8467..f7d17a5a1f56 100644
@@ -157,7 +157,7 @@ static void __init alloc_init_pte(pmd_t *pmd, unsigned long addr,
 
 static void __init alloc_init_pmd(struct mm_struct *mm, pud_t *pud,
                                unsigned long addr, unsigned long end,
-                               phys_addr_t phys, int map_io)
+                               phys_addr_t phys, int map_io, int map_xn)
 {
      pmd_t *pmd;
      unsigned long next;

arch/arm64/mm/mmu.c static void __init alloc_init_pmd(struct mm_struct *mm, pud_t *pud, · line 175

Quoted 9 lines

arch/arm64/mm/mmu.c +3 -0

@@ -167,6 +167,9 @@ static void __init alloc_init_pmd(struct mm_struct *mm, pud_t *pud,
      if (map_io) {
              prot_sect = PROT_SECT_DEVICE_nGnRE;
              prot_pte = __pgprot(PROT_DEVICE_nGnRE);
+     } else if (map_xn) {
+             prot_sect = PROT_SECT_NORMAL;
+             prot_pte = PAGE_KERNEL;
      } else {
              prot_sect = PROT_SECT_NORMAL_EXEC;
              prot_pte = PAGE_KERNEL_EXEC;

Ideally, this if block should be completely removed.

Quoted 11 lines

arch/arm64/mm/mmu.c +1 -1

@@ -203,7 +206,7 @@ static void __init alloc_init_pmd(struct mm_struct *mm, pud_t *pud,
 
 static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
                                unsigned long addr, unsigned long end,
-                               unsigned long phys, int map_io)
+                               unsigned long phys, int map_io, int map_xn)
 {
      pud_t *pud;
      unsigned long next;

@@ -221,7 +224,7 @@ static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
              /*

arch/arm64/mm/mmu.c static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd, · line 227

Quoted 4 lines

arch/arm64/mm/mmu.c +1 -1

               * For 4K granule only, attempt to put down a 1GB block
               */
-             if (!map_io && (PAGE_SHIFT == 12) &&
+             if (!map_io && !map_xn && (PAGE_SHIFT == 12) &&

Presumably the !map_io and !map_xn tests are here because of the PROT_SECT_NORMAL_EXEC below? Is there another reason why a pud mapping would be unsuitable for these?

Quoted 58 lines

arch/arm64/mm/mmu.c +14 -6

                  ((addr | next | phys) & ~PUD_MASK) == 0) {
                      pud_t old_pud = *pud;
                      set_pud(pud, __pud(phys | PROT_SECT_NORMAL_EXEC));

@@ -239,7 +242,8 @@ static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
                              flush_tlb_all();
                      }
              } else {
-                     alloc_init_pmd(mm, pud, addr, next, phys, map_io);
+                     alloc_init_pmd(mm, pud, addr, next, phys, map_io,
+                                    map_xn);
              }
              phys += next - addr;
      } while (pud++, addr = next, addr != end);

@@ -251,7 +255,7 @@ static void __init alloc_init_pud(struct mm_struct *mm, pgd_t *pgd,
  */
 static void __init __create_mapping(struct mm_struct *mm, pgd_t *pgd,
                                  phys_addr_t phys, unsigned long virt,
-                                 phys_addr_t size, int map_io)
+                                 phys_addr_t size, int map_io, int map_xn)
 {
      unsigned long addr, length, end, next;
 

@@ -261,7 +265,7 @@ static void __init __create_mapping(struct mm_struct *mm, pgd_t *pgd,
      end = addr + length;
      do {
              next = pgd_addr_end(addr, end);
-             alloc_init_pud(mm, pgd, addr, next, phys, map_io);
+             alloc_init_pud(mm, pgd, addr, next, phys, map_io, map_xn);
              phys += next - addr;
      } while (pgd++, addr = next, addr != end);
 }

@@ -275,7 +279,7 @@ static void __init create_mapping(phys_addr_t phys, unsigned long virt,
              return;
      }
      __create_mapping(&init_mm, pgd_offset_k(virt & PAGE_MASK), phys, virt,
-                      size, 0);
+                      size, 0, 0);
 }
 
 void __init create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io)

@@ -285,7 +289,15 @@ void __init create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io)
              return;
      }
      __create_mapping(&init_mm, &idmap_pg_dir[pgd_index(addr)],
-                      addr, addr, size, map_io);
+                      addr, addr, size, map_io, 0);
+}
+
+void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
+                            unsigned long virt, phys_addr_t size,
+                            int map_io, int map_xn)
+{
+     __create_mapping(mm, pgd_offset(mm, virt), phys, virt, size, map_io,
+                      map_xn);
 }
 
 static void __init map_mem(void)
--

Quoted 2 lines

1.8.3.2

Cheers,
--
Steve

In order to support kexec, the kernel needs to be able to deal with the state of the UEFI firmware after SetVirtualAddressMap() has been called. To avoid having separate code paths for non-kexec and kexec, let's move the call to SetVirtualAddressMap() to the stub: this will guarantee us that it will only be called once (since the stub is not executed during kexec), and ensures that the UEFI state is identical between kexec and normal boot.

This implies that the layout of the virtual mapping needs to be created by the stub as well. All regions are rounded up to a naturally aligned multiple of 64 KB (for compatibility with 64k pages kernels) and recorded in the UEFI memory map. The kernel proper reads those values and installs the mappings in a dedicated set of page tables that are swapped in during UEFI Runtime Services calls.

Signed-off-by Ard Biesheuvel <[email protected]>

---
 arch/arm64/include/asm/efi.h       |  19 +++-
 arch/arm64/kernel/efi.c            | 205 +++++++++++++++++++------------------
 drivers/firmware/efi/libstub/fdt.c | 104 ++++++++++++++++++-
 3 files changed, 224 insertions(+), 104 deletions(-)

arch/arm64/kernel/efi.c +15 -0

@@ -69,9 +74,36 @@ static void __init efi_setup_idmap(void)
        }
 }
 
+/*
+ * Translate a EFI virtual address into a physical address: this is necessary,
+ * as some data members of the EFI system table are virtually remapped after
+ * SetVirtualAddressMap() has been called.
+ */
+static phys_addr_t __init efi_to_phys(unsigned long addr)
+{
+       efi_memory_desc_t *md;
+
+       for_each_efi_memory_desc(&memmap, md) {
+               if (!(md->attribute & EFI_MEMORY_RUNTIME))
+                       continue;
+               if (md->virt_addr == 0)
+                       /* no virtual mapping has been installed by the stub */
+                       break;

Quoted 8 lines · Ard Biesheuvel

In order to support kexec, the kernel needs to be able to deal with the state of the UEFI firmware after SetVirtualAddressMap() has been called. To avoid having separate code paths for non-kexec and kexec, let's move the call to SetVirtualAddressMap() to the stub: this will guarantee us that it will only be called once (since the stub is not executed during kexec), and ensures that the UEFI state is identical between kexec and normal boot.

Quoted 8 lines

This implies that the layout of the virtual mapping needs to be created by the stub as well. All regions are rounded up to a naturally aligned multiple of 64 KB (for compatibility with 64k pages kernels) and recorded in the UEFI memory map. The kernel proper reads those values and installs the mappings in a dedicated set of page tables that are swapped in during UEFI Runtime Services calls.

Signed-off-by Ard Biesheuvel <[email protected]>

A questions grounded in my ignorance below...

Quoted 30 lines

---
 arch/arm64/include/asm/efi.h       |  19 +++-
 arch/arm64/kernel/efi.c            | 205 +++++++++++++++++++------------------
 drivers/firmware/efi/libstub/fdt.c | 104 ++++++++++++++++++-
 3 files changed, 224 insertions(+), 104 deletions(-)

arch/arm64/kernel/efi.c +20 -0

@@ -69,9 +74,36 @@ static void __init efi_setup_idmap(void)
        }
 }
 
+/*
+ * Translate a EFI virtual address into a physical address: this is necessary,
+ * as some data members of the EFI system table are virtually remapped after
+ * SetVirtualAddressMap() has been called.
+ */
+static phys_addr_t __init efi_to_phys(unsigned long addr)
+{
+       efi_memory_desc_t *md;
+
+       for_each_efi_memory_desc(&memmap, md) {
+               if (!(md->attribute & EFI_MEMORY_RUNTIME))
+                       continue;
+               if (md->virt_addr == 0)
+                       /* no virtual mapping has been installed by the stub */
+                       break;
+               if (md->virt_addr < addr &&
+                   (addr - md->virt_addr) < (md->num_pages << EFI_PAGE_SHIFT))
+                       return md->phys_addr + addr - md->virt_addr;
+       }
+

arch/arm64/kernel/efi.c static void __init efi_setup_idmap(void) · line 101

Quoted 5 lines

arch/arm64/kernel/efi.c +5 -0

+       WARN_ONCE(1, "UEFI virtual mapping incomplete or missing -- no entry found for 0x%lx\n",
+                 addr);
+       return addr;
+}
+

Is function applicable to all EFI implementations? Should it be in the common EFI code?

How much of this patch series is (theoretically) applicable to aarch32?

In order to support kexec, the kernel needs to be able to deal with the state of the UEFI firmware after SetVirtualAddressMap() has been called. To avoid having separate code paths for non-kexec and kexec, let's move the call to SetVirtualAddressMap() to the stub: this will guarantee us that it will only be called once (since the stub is not executed during kexec), and ensures that the UEFI state is identical between kexec and normal boot.

This implies that the layout of the virtual mapping needs to be created by the stub as well. All regions are rounded up to a naturally aligned multiple of 64 KB (for compatibility with 64k pages kernels) and recorded in the UEFI memory map. The kernel proper reads those values and installs the mappings in a dedicated set of page tables that are swapped in during UEFI Runtime Services calls.

Signed-off-by Ard Biesheuvel <[email protected]>

---
 arch/arm64/include/asm/efi.h       |  19 +++-
 arch/arm64/kernel/efi.c            | 205 +++++++++++++++++++------------------
 drivers/firmware/efi/libstub/fdt.c | 104 ++++++++++++++++++-
 3 files changed, 224 insertions(+), 104 deletions(-)

arch/arm64/include/asm/efi.h +15 -4

index a34fd3b12e2b..d752e5480096 100644
@@ -12,23 +12,32 @@ extern void efi_idmap_init(void);
 #define efi_idmap_init()
 #endif
 
+void efi_load_rt_mapping(void);
+void efi_unload_rt_mapping(void);
+
 #define efi_call_virt(f, ...)						\
 ({									\
-	efi_##f##_t *__f = efi.systab->runtime->f;			\
+	efi_##f##_t *__f;						\
 	efi_status_t __s;						\
 									\
-	kernel_neon_begin();						\
+	kernel_neon_begin();	/* disables preemption */		\
+	efi_load_rt_mapping();						\
+	__f = efi.systab->runtime->f;					\
 	__s = __f(__VA_ARGS__);						\
+	efi_unload_rt_mapping();					\
 	kernel_neon_end();						\
 	__s;								\
 })
 
 #define __efi_call_virt(f, ...)						\
 ({									\
-	efi_##f##_t *__f = efi.systab->runtime->f;			\
+	efi_##f##_t *__f;						\
 									\
-	kernel_neon_begin();						\
+	kernel_neon_begin();	/* disables preemption */		\
+	efi_load_rt_mapping();						\
+	__f = efi.systab->runtime->f;					\
 	__f(__VA_ARGS__);						\
+	efi_unload_rt_mapping();					\
 	kernel_neon_end();						\
 })
 

@@ -44,4 +53,6 @@ extern void efi_idmap_init(void);
 
 #define efi_call_early(f, ...) sys_table_arg->boottime->f(__VA_ARGS__)
 
+#define EFI_VIRTMAP		EFI_ARCH_1
+
 #endif /* _ASM_EFI_H */

arch/arm64/kernel/efi.c +19 -2

index baab9344a32b..324398c03acd 100644
@@ -20,16 +20,21 @@
 #include <linux/of_fdt.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
+#include <linux/mm_types.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+#include <linux/spinlock.h>
+#include <linux/atomic.h>
 
 #include <asm/cacheflush.h>
 #include <asm/efi.h>
 #include <asm/tlbflush.h>
+#include <asm/pgtable.h>
 #include <asm/mmu_context.h>
+#include <asm/mmu.h>
 
 struct efi_memory_map memmap;
 
-static efi_runtime_services_t *runtime;
-
 static u64 efi_system_table;
 
 static int uefi_debug __initdata;

@@ -69,9 +74,36 @@ static void __init efi_setup_idmap(void)
 	}
 }
 
+/*
+ * Translate a EFI virtual address into a physical address: this is necessary,
+ * as some data members of the EFI system table are virtually remapped after
+ * SetVirtualAddressMap() has been called.
+ */
+static phys_addr_t __init efi_to_phys(unsigned long addr)
+{
+	efi_memory_desc_t *md;
+
+	for_each_efi_memory_desc(&memmap, md) {
+		if (!(md->attribute & EFI_MEMORY_RUNTIME))
+			continue;

arch/arm64/kernel/efi.c +6 -1

+	return addr;
+}
+
 static int __init uefi_init(void)
 {
 	efi_char16_t *c16;
+	void *config_tables;
+	u64 table_size;
 	char vendor[100] = "unknown";
 	int i, retval;
 

@@ -99,7 +131,7 @@ static int __init uefi_init(void)
 			efi.systab->hdr.revision & 0xffff);
 
 	/* Show what we know for posterity */
-	c16 = early_memremap(efi.systab->fw_vendor,
+	c16 = early_memremap(efi_to_phys(efi.systab->fw_vendor),
 			     sizeof(vendor));
 	if (c16) {
 		for (i = 0; i < (int) sizeof(vendor) - 1 && *c16; ++i)

arch/arm64/kernel/efi.c +69 -94

+
+	retval = efi_config_parse_tables(config_tables,
+					 efi.systab->nr_tables, NULL);
 
+	early_memunmap(config_tables, table_size);
 out:
 	early_memunmap(efi.systab,  sizeof(efi_system_table_t));
 	return retval;

@@ -319,60 +357,64 @@ void __init efi_init(void)
 	reserve_regions();
 }
 
+static pgd_t efi_pgd[PTRS_PER_PGD] __page_aligned_bss;
+
+static struct mm_struct efi_mm = {
+	.mm_rb			= RB_ROOT,
+	.pgd			= efi_pgd,
+	.mm_users		= ATOMIC_INIT(2),
+	.mm_count		= ATOMIC_INIT(1),
+	.mmap_sem		= __RWSEM_INITIALIZER(efi_mm.mmap_sem),
+	.page_table_lock	= __SPIN_LOCK_UNLOCKED(efi_mm.page_table_lock),
+	.mmlist			= LIST_HEAD_INIT(efi_mm.mmlist),
+	INIT_MM_CONTEXT(efi_mm)
+};
+
 void __init efi_idmap_init(void)
 {
+	efi_memory_desc_t *md;
+
 	if (!efi_enabled(EFI_BOOT))
 		return;
 
 	/* boot time idmap_pg_dir is incomplete, so fill in missing parts */
 	efi_setup_idmap();
-}
-
-static int __init remap_region(efi_memory_desc_t *md, void **new)
-{
-	u64 paddr, vaddr, npages, size;
-
-	paddr = md->phys_addr;
-	npages = md->num_pages;
-	memrange_efi_to_native(&paddr, &npages);
-	size = npages << PAGE_SHIFT;
-
-	if (is_normal_ram(md))
-		vaddr = (__force u64)ioremap_cache(paddr, size);
-	else
-		vaddr = (__force u64)ioremap(paddr, size);
 
-	if (!vaddr) {
-		pr_err("Unable to remap 0x%llx pages @ %p\n",
-		       npages, (void *)paddr);
-		return 0;
-	}
+	for_each_efi_memory_desc(&memmap, md) {
+		u64 paddr, npages, size;
 
-	/* adjust for any rounding when EFI and system pagesize differs */
-	md->virt_addr = vaddr + (md->phys_addr - paddr);
+		if (!(md->attribute & EFI_MEMORY_RUNTIME))
+			continue;
+		if (md->virt_addr == 0)
+			/* no virtual mapping has been installed by the stub */
+			return;
 
-	if (uefi_debug)
-		pr_info("  EFI remap 0x%012llx => %p\n",
-			md->phys_addr, (void *)md->virt_addr);
+		paddr = md->phys_addr;
+		npages = md->num_pages;
+		memrange_efi_to_native(&paddr, &npages);
+		size = npages << PAGE_SHIFT;
 
-	memcpy(*new, md, memmap.desc_size);
-	*new += memmap.desc_size;
+		if (uefi_debug)
+			pr_info("  EFI remap 0x%012llx => %p\n",
+				md->phys_addr, (void *)md->virt_addr);
 
-	return 1;
+		/*
+		 * Only regions of type EFI_RUNTIME_SERVICES_CODE need to be
+		 * executable, everything else can be mapped with the XN bits
+		 * set. Unfortunately, we cannot map those code regions write
+		 * protect, as they may contain read-write .data sections as
+		 * well.
+		 */
+		create_pgd_mapping(&efi_mm, paddr, md->virt_addr, size,
+				   !is_normal_ram(md),
+				   md->type != EFI_RUNTIME_SERVICES_CODE);
+	}
+	set_bit(EFI_VIRTMAP, &efi.flags);
 }
 
-/*
- * Switch UEFI from an identity map to a kernel virtual map
- */
 static int __init arm64_enter_virtual_mode(void)
 {
-	efi_memory_desc_t *md;
-	phys_addr_t virtmap_phys;
-	void *virtmap, *virt_md;
-	efi_status_t status;
 	u64 mapsize;
-	int count = 0;
-	unsigned long flags;
 
 	if (!efi_enabled(EFI_MEMMAP))
 		return 0;

@@ -393,79 +435,28 @@ static int __init arm64_enter_virtual_mode(void)
 
 	efi.memmap = &memmap;
 
-	/* Map the runtime regions */
-	virtmap = kmalloc(mapsize, GFP_KERNEL);
-	if (!virtmap) {
-		pr_err("Failed to allocate EFI virtual memmap\n");
-		return -1;
-	}
-	virtmap_phys = virt_to_phys(virtmap);
-	virt_md = virtmap;
-
-	for_each_efi_memory_desc(&memmap, md) {
-		if (!(md->attribute & EFI_MEMORY_RUNTIME))
-			continue;
-		if (!remap_region(md, &virt_md))
-			goto err_unmap;
-		++count;
-	}
-
-	efi.systab = (__force void *)efi_lookup_mapped_addr(efi_system_table);
+	efi.systab = (__force void *)ioremap_cache(efi_system_table,
+						   sizeof(efi_system_table_t));
 	if (!efi.systab) {
-		/*
-		 * If we have no virtual mapping for the System Table at this
-		 * point, the memory map doesn't cover the physical offset where
-		 * it resides. This means the System Table will be inaccessible
-		 * to Runtime Services themselves once the virtual mapping is
-		 * installed.
-		 */
-		pr_err("Failed to remap EFI System Table -- buggy firmware?\n");
-		goto err_unmap;
+		pr_err("Failed to remap EFI System Table\n");
+		return -1;
 	}
 	set_bit(EFI_SYSTEM_TABLES, &efi.flags);
 
-	local_irq_save(flags);
-	cpu_switch_mm(idmap_pg_dir, &init_mm);
-
-	/* Call SetVirtualAddressMap with the physical address of the map */
-	runtime = efi.systab->runtime;
-	efi.set_virtual_address_map = runtime->set_virtual_address_map;
-
-	status = efi.set_virtual_address_map(count * memmap.desc_size,
-					     memmap.desc_size,
-					     memmap.desc_version,
-					     (efi_memory_desc_t *)virtmap_phys);
-	cpu_set_reserved_ttbr0();
-	flush_tlb_all();
-	local_irq_restore(flags);
-
-	kfree(virtmap);
-
 	free_boot_services();
 
-	if (status != EFI_SUCCESS) {
-		pr_err("Failed to set EFI virtual address map! [%lx]\n",
-			status);
+	if (!efi_enabled(EFI_VIRTMAP)) {
+		pr_err("No UEFI virtual mapping was installed -- runtime services will not be available\n");
 		return -1;
 	}
 
 	/* Set up runtime services function pointers */
-	runtime = efi.systab->runtime;
 	efi_native_runtime_setup();
 	set_bit(EFI_RUNTIME_SERVICES, &efi.flags);
 
 	efi.runtime_version = efi.systab->hdr.revision;
 
 	return 0;
-
-err_unmap:
-	/* unmap all mappings that succeeded: there are 'count' of those */
-	for (virt_md = virtmap; count--; virt_md += memmap.desc_size) {
-		md = virt_md;
-		iounmap((__force void __iomem *)md->virt_addr);
-	}
-	kfree(virtmap);
-	return -1;
 }
 early_initcall(arm64_enter_virtual_mode);
 

@@ -482,3 +473,21 @@ static int __init arm64_dmi_init(void)
 	return 0;
 }
 core_initcall(arm64_dmi_init);
+
+static void efi_set_pgd(struct mm_struct *mm)
+{
+	cpu_switch_mm(mm->pgd, mm);
+	flush_tlb_all();
+	if (icache_is_aivivt())
+		__flush_icache_all();
+}
+
+void efi_load_rt_mapping(void)
+{
+	efi_set_pgd(&efi_mm);
+}
+
+void efi_unload_rt_mapping(void)
+{
+	efi_set_pgd(current->active_mm);
+}

drivers/firmware/efi/libstub/fdt.c +87 -2

index c846a9608cbd..9a5f6e54a423 100644
@@ -168,6 +168,68 @@ fdt_set_fail:
 #endif
 
 /*
+ * This is the base address at which to start allocating virtual memory ranges
+ * for UEFI Runtime Services. This is a userland range so that we can use any
+ * allocation we choose, and eliminate the risk of a conflict after kexec.
+ */
+#define EFI_RT_VIRTUAL_BASE	0x40000000
+
+static void update_memory_map(efi_memory_desc_t *memory_map,
+			      unsigned long map_size, unsigned long desc_size,
+			      int *count)
+{
+	u64 efi_virt_base = EFI_RT_VIRTUAL_BASE;
+	union {
+		efi_memory_desc_t entry;
+		u8 pad[desc_size];
+	} *p, *q, tmp;
+	int i = map_size / desc_size;
+
+	p = (void *)memory_map;
+	for (q = p; i >= 0; i--, q++) {
+		u64 paddr, size;
+
+		if (!(q->entry.attribute & EFI_MEMORY_RUNTIME))
+			continue;
+
+		/*
+		 * Swap the entries around so that all EFI_MEMORY_RUNTIME
+		 * entries bubble to the top. This will allow us to reuse the
+		 * table as input to SetVirtualAddressMap().
+		 */
+		if (q != p) {
+			tmp = *p;
+			*p = *q;
+			*q = tmp;
+		}
+
+		/*
+		 * Make the mapping compatible with 64k pages: this allows
+		 * a 4k page size kernel to kexec a 64k page size kernel and
+		 * vice versa.
+		 */
+		paddr = round_down(p->entry.phys_addr, SZ_64K);
+		size = round_up(p->entry.num_pages * EFI_PAGE_SIZE +
+				p->entry.phys_addr - paddr, SZ_64K);
+
+		/*
+		 * Avoid wasting memory on PTEs by choosing a virtual base that
+		 * is compatible with section mappings if this region has the
+		 * appropriate size and physical alignment. (Sections are 2 MB
+		 * on 4k granule kernels)
+		 */
+		if (IS_ALIGNED(p->entry.phys_addr, SZ_2M) && size >= SZ_2M)
+			efi_virt_base = round_up(efi_virt_base, SZ_2M);
+
+		p->entry.virt_addr = efi_virt_base + p->entry.phys_addr - paddr;
+		efi_virt_base += size;
+
+		++p;
+		++*count;
+	}
+}
+
+/*
  * Allocate memory for a new FDT, then add EFI, commandline, and
  * initrd related fields to the FDT.  This routine increases the
  * FDT allocation size until the allocated memory is large

@@ -196,6 +258,7 @@ efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table,
 	efi_memory_desc_t *memory_map;
 	unsigned long new_fdt_size;
 	efi_status_t status;
+	int runtime_entry_count = 0;
 
 	/*
 	 * Estimate size of new FDT, and allocate memory for it. We

@@ -248,12 +311,49 @@ efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table,
 		}
 	}
 
+	/*
+	 * Update the memory map with virtual addresses, and reorder the entries
+	 * so that we can pass it straight into SetVirtualAddressMap()
+	 */
+	update_memory_map(memory_map, map_size, desc_size,
+			  &runtime_entry_count);
+
 	/* Now we are ready to exit_boot_services.*/
 	status = sys_table->boottime->exit_boot_services(handle, mmap_key);
 
+	if (status == EFI_SUCCESS) {
+		efi_set_virtual_address_map_t *svam;
+
+		/* Install the new virtual address map */
+		svam = sys_table->runtime->set_virtual_address_map;
+		status = svam(runtime_entry_count * desc_size, desc_size,
+			      desc_ver, memory_map);
 
-	if (status == EFI_SUCCESS)
-		return status;
+		/*
+		 * We are beyond the point of no return here, so if the call to
+		 * SetVirtualAddressMap() failed, we need to signal that to the
+		 * incoming kernel but proceed normally otherwise.
+		 */
+		if (status != EFI_SUCCESS) {
+			int i;
+
+			/*
+			 * Set the virtual address field of all

drivers/firmware/efi/libstub/fdt.c +4 -0

+			}
+		}
+		return EFI_SUCCESS;
+	}
 
 	pr_efi_err(sys_table, "Exit boot services failed.\n");
 

Signature · 6 lines

1.8.3.2

--
To unsubscribe from this list: send the line "unsubscribe linux-efi" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hi, Ard

The approach looks good, thanks for the patchset.

Please see a few code comments inline..

Quoted 118 lines

In order to support kexec, the kernel needs to be able to deal with the state of the UEFI firmware after SetVirtualAddressMap() has been called. To avoid having separate code paths for non-kexec and kexec, let's move the call to SetVirtualAddressMap() to the stub: this will guarantee us that it will only be called once (since the stub is not executed during kexec), and ensures that the UEFI state is identical between kexec and normal boot.

This implies that the layout of the virtual mapping needs to be created by the stub as well. All regions are rounded up to a naturally aligned multiple of 64 KB (for compatibility with 64k pages kernels) and recorded in the UEFI memory map. The kernel proper reads those values and installs the mappings in a dedicated set of page tables that are swapped in during UEFI Runtime Services calls.

Signed-off-by Ard Biesheuvel <[email protected]>

---
 arch/arm64/include/asm/efi.h       |  19 +++-
 arch/arm64/kernel/efi.c            | 205 +++++++++++++++++++------------------
 drivers/firmware/efi/libstub/fdt.c | 104 ++++++++++++++++++-
 3 files changed, 224 insertions(+), 104 deletions(-)

arch/arm64/include/asm/efi.h +15 -4

index a34fd3b12e2b..d752e5480096 100644
@@ -12,23 +12,32 @@ extern void efi_idmap_init(void);
 #define efi_idmap_init()
 #endif
 
+void efi_load_rt_mapping(void);
+void efi_unload_rt_mapping(void);
+
 #define efi_call_virt(f, ...)                                                \
 ({                                                                   \
-     efi_##f##_t *__f = efi.systab->runtime->f;                      \
+     efi_##f##_t *__f;                                               \
      efi_status_t __s;                                               \
                                                                      \
-     kernel_neon_begin();                                            \
+     kernel_neon_begin();    /* disables preemption */               \
+     efi_load_rt_mapping();                                          \
+     __f = efi.systab->runtime->f;                                   \
      __s = __f(__VA_ARGS__);                                         \
+     efi_unload_rt_mapping();                                        \
      kernel_neon_end();                                              \
      __s;                                                            \
 })
 
 #define __efi_call_virt(f, ...)                                              \
 ({                                                                   \
-     efi_##f##_t *__f = efi.systab->runtime->f;                      \
+     efi_##f##_t *__f;                                               \
                                                                      \
-     kernel_neon_begin();                                            \
+     kernel_neon_begin();    /* disables preemption */               \
+     efi_load_rt_mapping();                                          \
+     __f = efi.systab->runtime->f;                                   \
      __f(__VA_ARGS__);                                               \
+     efi_unload_rt_mapping();                                        \
      kernel_neon_end();                                              \
 })
 

@@ -44,4 +53,6 @@ extern void efi_idmap_init(void);
 
 #define efi_call_early(f, ...) sys_table_arg->boottime->f(__VA_ARGS__)
 
+#define EFI_VIRTMAP          EFI_ARCH_1
+
 #endif /* _ASM_EFI_H */

arch/arm64/kernel/efi.c +23 -2

index baab9344a32b..324398c03acd 100644
@@ -20,16 +20,21 @@
 #include <linux/of_fdt.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
+#include <linux/mm_types.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+#include <linux/spinlock.h>
+#include <linux/atomic.h>
 
 #include <asm/cacheflush.h>
 #include <asm/efi.h>
 #include <asm/tlbflush.h>
+#include <asm/pgtable.h>
 #include <asm/mmu_context.h>
+#include <asm/mmu.h>
 
 struct efi_memory_map memmap;
 
-static efi_runtime_services_t *runtime;
-
 static u64 efi_system_table;
 
 static int uefi_debug __initdata;

@@ -69,9 +74,36 @@ static void __init efi_setup_idmap(void)
      }
 }
 
+/*
+ * Translate a EFI virtual address into a physical address: this is necessary,
+ * as some data members of the EFI system table are virtually remapped after
+ * SetVirtualAddressMap() has been called.
+ */
+static phys_addr_t __init efi_to_phys(unsigned long addr)
+{
+     efi_memory_desc_t *md;
+
+     for_each_efi_memory_desc(&memmap, md) {
+             if (!(md->attribute & EFI_MEMORY_RUNTIME))
+                     continue;
+             if (md->virt_addr == 0)
+                     /* no virtual mapping has been installed by the stub */
+                     break;
+             if (md->virt_addr < addr &&

arch/arm64/kernel/efi.c static void __init efi_setup_idmap(void) · line 98

Quoted 6 lines

arch/arm64/kernel/efi.c +6 -0

+                 (addr - md->virt_addr) < (md->num_pages << EFI_PAGE_SHIFT))
+                     return md->phys_addr + addr - md->virt_addr;
+     }
+
+     WARN_ONCE(1, "UEFI virtual mapping incomplete or missing -- no entry found for 0x%lx\n",
+               addr);

die here instead of warn looks better, addr could be anything which does not match the virt maps.

Quoted 21 lines

arch/arm64/kernel/efi.c +6 -1

+     return addr;
+}
+
 static int __init uefi_init(void)
 {
      efi_char16_t *c16;
+     void *config_tables;
+     u64 table_size;
      char vendor[100] = "unknown";
      int i, retval;
 

@@ -99,7 +131,7 @@ static int __init uefi_init(void)
                      efi.systab->hdr.revision & 0xffff);
 
      /* Show what we know for posterity */
-     c16 = early_memremap(efi.systab->fw_vendor,
+     c16 = early_memremap(efi_to_phys(efi.systab->fw_vendor),
                           sizeof(vendor));
      if (c16) {
              for (i = 0; i < (int) sizeof(vendor) - 1 && *c16; ++i)

arch/arm64/kernel/efi.c static int __init uefi_init(void) · line 149

Quoted 7 lines

arch/arm64/kernel/efi.c +3 -1

@@ -112,8 +144,14 @@ static int __init uefi_init(void)
              efi.systab->hdr.revision >> 16,
              efi.systab->hdr.revision & 0xffff, vendor);
 
-     retval = efi_config_init(NULL);
+     table_size = sizeof(efi_config_table_64_t) * efi.systab->nr_tables;
+     config_tables = early_memremap(efi_to_phys(efi.systab->tables),
+                                    table_size);

There should be error handling for early_memremap.

Quoted 337 lines

arch/arm64/kernel/efi.c +69 -94

+
+     retval = efi_config_parse_tables(config_tables,
+                                      efi.systab->nr_tables, NULL);
 
+     early_memunmap(config_tables, table_size);
 out:
      early_memunmap(efi.systab,  sizeof(efi_system_table_t));
      return retval;

@@ -319,60 +357,64 @@ void __init efi_init(void)
      reserve_regions();
 }
 
+static pgd_t efi_pgd[PTRS_PER_PGD] __page_aligned_bss;
+
+static struct mm_struct efi_mm = {
+     .mm_rb                  = RB_ROOT,
+     .pgd                    = efi_pgd,
+     .mm_users               = ATOMIC_INIT(2),
+     .mm_count               = ATOMIC_INIT(1),
+     .mmap_sem               = __RWSEM_INITIALIZER(efi_mm.mmap_sem),
+     .page_table_lock        = __SPIN_LOCK_UNLOCKED(efi_mm.page_table_lock),
+     .mmlist                 = LIST_HEAD_INIT(efi_mm.mmlist),
+     INIT_MM_CONTEXT(efi_mm)
+};
+
 void __init efi_idmap_init(void)
 {
+     efi_memory_desc_t *md;
+
      if (!efi_enabled(EFI_BOOT))
              return;
 
      /* boot time idmap_pg_dir is incomplete, so fill in missing parts */
      efi_setup_idmap();
-}
-
-static int __init remap_region(efi_memory_desc_t *md, void **new)
-{
-     u64 paddr, vaddr, npages, size;
-
-     paddr = md->phys_addr;
-     npages = md->num_pages;
-     memrange_efi_to_native(&paddr, &npages);
-     size = npages << PAGE_SHIFT;
-
-     if (is_normal_ram(md))
-             vaddr = (__force u64)ioremap_cache(paddr, size);
-     else
-             vaddr = (__force u64)ioremap(paddr, size);
 
-     if (!vaddr) {
-             pr_err("Unable to remap 0x%llx pages @ %p\n",
-                    npages, (void *)paddr);
-             return 0;
-     }
+     for_each_efi_memory_desc(&memmap, md) {
+             u64 paddr, npages, size;
 
-     /* adjust for any rounding when EFI and system pagesize differs */
-     md->virt_addr = vaddr + (md->phys_addr - paddr);
+             if (!(md->attribute & EFI_MEMORY_RUNTIME))
+                     continue;
+             if (md->virt_addr == 0)
+                     /* no virtual mapping has been installed by the stub */
+                     return;
 
-     if (uefi_debug)
-             pr_info("  EFI remap 0x%012llx => %p\n",
-                     md->phys_addr, (void *)md->virt_addr);
+             paddr = md->phys_addr;
+             npages = md->num_pages;
+             memrange_efi_to_native(&paddr, &npages);
+             size = npages << PAGE_SHIFT;
 
-     memcpy(*new, md, memmap.desc_size);
-     *new += memmap.desc_size;
+             if (uefi_debug)
+                     pr_info("  EFI remap 0x%012llx => %p\n",
+                             md->phys_addr, (void *)md->virt_addr);
 
-     return 1;
+             /*
+              * Only regions of type EFI_RUNTIME_SERVICES_CODE need to be
+              * executable, everything else can be mapped with the XN bits
+              * set. Unfortunately, we cannot map those code regions write
+              * protect, as they may contain read-write .data sections as
+              * well.
+              */
+             create_pgd_mapping(&efi_mm, paddr, md->virt_addr, size,
+                                !is_normal_ram(md),
+                                md->type != EFI_RUNTIME_SERVICES_CODE);
+     }
+     set_bit(EFI_VIRTMAP, &efi.flags);
 }
 
-/*
- * Switch UEFI from an identity map to a kernel virtual map
- */
 static int __init arm64_enter_virtual_mode(void)
 {
-     efi_memory_desc_t *md;
-     phys_addr_t virtmap_phys;
-     void *virtmap, *virt_md;
-     efi_status_t status;
      u64 mapsize;
-     int count = 0;
-     unsigned long flags;
 
      if (!efi_enabled(EFI_MEMMAP))
              return 0;

@@ -393,79 +435,28 @@ static int __init arm64_enter_virtual_mode(void)
 
      efi.memmap = &memmap;
 
-     /* Map the runtime regions */
-     virtmap = kmalloc(mapsize, GFP_KERNEL);
-     if (!virtmap) {
-             pr_err("Failed to allocate EFI virtual memmap\n");
-             return -1;
-     }
-     virtmap_phys = virt_to_phys(virtmap);
-     virt_md = virtmap;
-
-     for_each_efi_memory_desc(&memmap, md) {
-             if (!(md->attribute & EFI_MEMORY_RUNTIME))
-                     continue;
-             if (!remap_region(md, &virt_md))
-                     goto err_unmap;
-             ++count;
-     }
-
-     efi.systab = (__force void *)efi_lookup_mapped_addr(efi_system_table);
+     efi.systab = (__force void *)ioremap_cache(efi_system_table,
+                                                sizeof(efi_system_table_t));
      if (!efi.systab) {
-             /*
-              * If we have no virtual mapping for the System Table at this
-              * point, the memory map doesn't cover the physical offset where
-              * it resides. This means the System Table will be inaccessible
-              * to Runtime Services themselves once the virtual mapping is
-              * installed.
-              */
-             pr_err("Failed to remap EFI System Table -- buggy firmware?\n");
-             goto err_unmap;
+             pr_err("Failed to remap EFI System Table\n");
+             return -1;
      }
      set_bit(EFI_SYSTEM_TABLES, &efi.flags);
 
-     local_irq_save(flags);
-     cpu_switch_mm(idmap_pg_dir, &init_mm);
-
-     /* Call SetVirtualAddressMap with the physical address of the map */
-     runtime = efi.systab->runtime;
-     efi.set_virtual_address_map = runtime->set_virtual_address_map;
-
-     status = efi.set_virtual_address_map(count * memmap.desc_size,
-                                          memmap.desc_size,
-                                          memmap.desc_version,
-                                          (efi_memory_desc_t *)virtmap_phys);
-     cpu_set_reserved_ttbr0();
-     flush_tlb_all();
-     local_irq_restore(flags);
-
-     kfree(virtmap);
-
      free_boot_services();
 
-     if (status != EFI_SUCCESS) {
-             pr_err("Failed to set EFI virtual address map! [%lx]\n",
-                     status);
+     if (!efi_enabled(EFI_VIRTMAP)) {
+             pr_err("No UEFI virtual mapping was installed -- runtime services will not be available\n");
              return -1;
      }
 
      /* Set up runtime services function pointers */
-     runtime = efi.systab->runtime;
      efi_native_runtime_setup();
      set_bit(EFI_RUNTIME_SERVICES, &efi.flags);
 
      efi.runtime_version = efi.systab->hdr.revision;
 
      return 0;
-
-err_unmap:
-     /* unmap all mappings that succeeded: there are 'count' of those */
-     for (virt_md = virtmap; count--; virt_md += memmap.desc_size) {
-             md = virt_md;
-             iounmap((__force void __iomem *)md->virt_addr);
-     }
-     kfree(virtmap);
-     return -1;
 }
 early_initcall(arm64_enter_virtual_mode);
 

@@ -482,3 +473,21 @@ static int __init arm64_dmi_init(void)
      return 0;
 }
 core_initcall(arm64_dmi_init);
+
+static void efi_set_pgd(struct mm_struct *mm)
+{
+     cpu_switch_mm(mm->pgd, mm);
+     flush_tlb_all();
+     if (icache_is_aivivt())
+             __flush_icache_all();
+}
+
+void efi_load_rt_mapping(void)
+{
+     efi_set_pgd(&efi_mm);
+}
+
+void efi_unload_rt_mapping(void)
+{
+     efi_set_pgd(current->active_mm);
+}

drivers/firmware/efi/libstub/fdt.c +91 -2

index c846a9608cbd..9a5f6e54a423 100644
@@ -168,6 +168,68 @@ fdt_set_fail:
 #endif
 
 /*
+ * This is the base address at which to start allocating virtual memory ranges
+ * for UEFI Runtime Services. This is a userland range so that we can use any
+ * allocation we choose, and eliminate the risk of a conflict after kexec.
+ */
+#define EFI_RT_VIRTUAL_BASE  0x40000000
+
+static void update_memory_map(efi_memory_desc_t *memory_map,
+                           unsigned long map_size, unsigned long desc_size,
+                           int *count)
+{
+     u64 efi_virt_base = EFI_RT_VIRTUAL_BASE;
+     union {
+             efi_memory_desc_t entry;
+             u8 pad[desc_size];
+     } *p, *q, tmp;
+     int i = map_size / desc_size;
+
+     p = (void *)memory_map;
+     for (q = p; i >= 0; i--, q++) {
+             u64 paddr, size;
+
+             if (!(q->entry.attribute & EFI_MEMORY_RUNTIME))
+                     continue;
+
+             /*
+              * Swap the entries around so that all EFI_MEMORY_RUNTIME
+              * entries bubble to the top. This will allow us to reuse the
+              * table as input to SetVirtualAddressMap().
+              */
+             if (q != p) {
+                     tmp = *p;
+                     *p = *q;
+                     *q = tmp;
+             }
+
+             /*
+              * Make the mapping compatible with 64k pages: this allows
+              * a 4k page size kernel to kexec a 64k page size kernel and
+              * vice versa.
+              */
+             paddr = round_down(p->entry.phys_addr, SZ_64K);
+             size = round_up(p->entry.num_pages * EFI_PAGE_SIZE +
+                             p->entry.phys_addr - paddr, SZ_64K);
+
+             /*
+              * Avoid wasting memory on PTEs by choosing a virtual base that
+              * is compatible with section mappings if this region has the
+              * appropriate size and physical alignment. (Sections are 2 MB
+              * on 4k granule kernels)
+              */
+             if (IS_ALIGNED(p->entry.phys_addr, SZ_2M) && size >= SZ_2M)
+                     efi_virt_base = round_up(efi_virt_base, SZ_2M);
+
+             p->entry.virt_addr = efi_virt_base + p->entry.phys_addr - paddr;
+             efi_virt_base += size;
+
+             ++p;
+             ++*count;
+     }
+}
+
+/*
  * Allocate memory for a new FDT, then add EFI, commandline, and
  * initrd related fields to the FDT.  This routine increases the
  * FDT allocation size until the allocated memory is large

@@ -196,6 +258,7 @@ efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table,
      efi_memory_desc_t *memory_map;
      unsigned long new_fdt_size;
      efi_status_t status;
+     int runtime_entry_count = 0;
 
      /*
       * Estimate size of new FDT, and allocate memory for it. We

@@ -248,12 +311,49 @@ efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table,
              }
      }
 
+     /*
+      * Update the memory map with virtual addresses, and reorder the entries
+      * so that we can pass it straight into SetVirtualAddressMap()
+      */
+     update_memory_map(memory_map, map_size, desc_size,
+                       &runtime_entry_count);
+
      /* Now we are ready to exit_boot_services.*/
      status = sys_table->boottime->exit_boot_services(handle, mmap_key);
 
+     if (status == EFI_SUCCESS) {
+             efi_set_virtual_address_map_t *svam;
+
+             /* Install the new virtual address map */
+             svam = sys_table->runtime->set_virtual_address_map;
+             status = svam(runtime_entry_count * desc_size, desc_size,
+                           desc_ver, memory_map);
 
-     if (status == EFI_SUCCESS)
-             return status;
+             /*
+              * We are beyond the point of no return here, so if the call to
+              * SetVirtualAddressMap() failed, we need to signal that to the
+              * incoming kernel but proceed normally otherwise.
+              */
+             if (status != EFI_SUCCESS) {
+                     int i;
+
+                     /*
+                      * Set the virtual address field of all
+                      * EFI_MEMORY_RUNTIME entries to 0. This will signal
+                      * the incoming kernel that no virtual translation has
+                      * been installed.
+                      */

drivers/firmware/efi/libstub/fdt.c efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table, · line 351

Quoted 6 lines

drivers/firmware/efi/libstub/fdt.c +6 -0

+                     for (i = 0; i < map_size; i += desc_size) {
+                             efi_memory_desc_t *p;
+
+                             p = (efi_memory_desc_t *)((u8 *)memory_map + i);
+                             if (!(p->attribute & EFI_MEMORY_RUNTIME))
+                                     break;

should it be continue instead of break?

drivers/firmware/efi/libstub/fdt.c efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table, · line 352

Quoted 1 line

drivers/firmware/efi/libstub/fdt.c +1 -0

+                             p->virt_addr = 0;

In X86 kernel panics in case set_virtual_address_map, Matt mentioned that it's not reasonable to continue. But for arm64 I'm not sure it's same case.

Quoted 14 lines

drivers/firmware/efi/libstub/fdt.c +4 -1

+                     }
+             }
+             return EFI_SUCCESS;
+     }
 
      pr_efi_err(sys_table, "Exit boot services failed.\n");
 
--

1.8.3.2

--
To unsubscribe from this list: send the line "unsubscribe linux-efi" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Quoted 11 lines

Quoted 11 lines

arch/arm64/kernel/efi.c +11 -0

+static phys_addr_t __init efi_to_phys(unsigned long addr)
+{
+     efi_memory_desc_t *md;
+
+     for_each_efi_memory_desc(&memmap, md) {
+             if (!(md->attribute & EFI_MEMORY_RUNTIME))
+                     continue;
+             if (md->virt_addr == 0)
+                     /* no virtual mapping has been installed by the stub */
+                     break;
+             if (md->virt_addr < addr &&

Quoted 9 lines

arch/arm64/kernel/efi.c static void __init efi_setup_idmap(void) · line 98

Quoted 6 lines

arch/arm64/kernel/efi.c +6 -0

+                 (addr - md->virt_addr) < (md->num_pages << EFI_PAGE_SHIFT))
+                     return md->phys_addr + addr - md->virt_addr;
+     }
+
+     WARN_ONCE(1, "UEFI virtual mapping incomplete or missing -- no entry found for 0x%lx\n",
+               addr);

die here instead of warn looks better, addr could be anything which does not match the virt maps.

The general idea was (and below as well) that SetVirtualAddressMap() is called from the stub, at a point where we can't easily panic/error out in a meaningful way, as we have just called ExitBootServices() so there is no console etc. So instead, I zero out all the virt_addr fields, which indicates that no virtual mapping is installed. That does not necessarily mean (on arm64 at least) that we will not be able to bring up ACPI etc and moan in a way that could catch someone's attention rather than only on the console.

I chose a WARN_ONCE() because it gives a nice fat stack dump in the kernel log, but as the RT services regions and fw_vendor and config_table are still id mapped, I don't see a compelling reason to panic right here.

Quoted 27 lines

Quoted 27 lines

      /* Now we are ready to exit_boot_services.*/
      status = sys_table->boottime->exit_boot_services(handle, mmap_key);

drivers/firmware/efi/libstub/fdt.c +21 -2

+     if (status == EFI_SUCCESS) {
+             efi_set_virtual_address_map_t *svam;
+
+             /* Install the new virtual address map */
+             svam = sys_table->runtime->set_virtual_address_map;
+             status = svam(runtime_entry_count * desc_size, desc_size,
+                           desc_ver, memory_map);
 
-     if (status == EFI_SUCCESS)
-             return status;
+             /*
+              * We are beyond the point of no return here, so if the call to
+              * SetVirtualAddressMap() failed, we need to signal that to the
+              * incoming kernel but proceed normally otherwise.
+              */
+             if (status != EFI_SUCCESS) {
+                     int i;
+
+                     /*
+                      * Set the virtual address field of all
+                      * EFI_MEMORY_RUNTIME entries to 0. This will signal
+                      * the incoming kernel that no virtual translation has
+                      * been installed.
+                      */

Quoted 9 lines

drivers/firmware/efi/libstub/fdt.c efi_status_t allocate_new_fdt_and_exit_boot(efi_system_table_t *sys_table, · line 351

Quoted 6 lines

drivers/firmware/efi/libstub/fdt.c +6 -0

+                     for (i = 0; i < map_size; i += desc_size) {
+                             efi_memory_desc_t *p;
+
+                             p = (efi_memory_desc_t *)((u8 *)memory_map + i);
+                             if (!(p->attribute & EFI_MEMORY_RUNTIME))
+                                     break;

should it be continue instead of break?

The memory map is sorted, so we can break as soon as we encounter the first one without the attribute set.

Quoted 6 lines

Quoted 2 lines

+ p->virt_addr = 0;

In X86 kernel panics in case set_virtual_address_map, Matt mentioned that it's not reasonable to continue. But for arm64 I'm not sure it's same case.

As I said, perhaps it is not reasonable, or perhaps it is, but panicking here is maybe not the most productive thing to do. I could add a printk() before calling exitbootservices() perhaps, saying 'if you can read this line, and nothing more, we may have crashed in/after SetVirtualAddressMap()'

Now that we have moved the call to SetVirtualAddressMap() to the stub, UEFI has no use for the ID map, so we can drop the code that installs ID mappings for UEFI memory regions.

Signed-off-by Ard Biesheuvel <[email protected]>

---
 arch/arm64/include/asm/efi.h |  4 ++--
 arch/arm64/include/asm/mmu.h |  2 --
 arch/arm64/kernel/efi.c      | 26 +-------------------------
 arch/arm64/kernel/setup.c    |  2 +-
 arch/arm64/mm/mmu.c          | 10 ----------
 5 files changed, 4 insertions(+), 40 deletions(-)

arch/arm64/include/asm/efi.h +2 -2

index d752e5480096..09854df5a2a1 100644
@@ -6,10 +6,10 @@
 
 #ifdef CONFIG_EFI
 extern void efi_init(void);
-extern void efi_idmap_init(void);
+extern void efi_virtmap_init(void);
 #else
 #define efi_init()
-#define efi_idmap_init()
+#define efi_virtmap_init()
 #endif
 
 void efi_load_rt_mapping(void);

arch/arm64/include/asm/mmu.h +0 -2

index bcf166043a8b..df33cf02300a 100644
@@ -31,8 +31,6 @@ extern void paging_init(void);
 extern void setup_mm_for_reboot(void);
 extern void __iomem *early_io_map(phys_addr_t phys, unsigned long virt);
 extern void init_mem_pgprot(void);
-/* create an identity mapping for memory (or io if map_io is true) */
-extern void create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io);
 extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
                               unsigned long virt, phys_addr_t size,
                               int map_io, int map_xn);

arch/arm64/kernel/efi.c +1 -25

index cbff3e8eff11..02bd7e877f0b 100644
@@ -53,27 +53,6 @@ static int __init is_normal_ram(efi_memory_desc_t *md)
        return 0;
 }
 
-static void __init efi_setup_idmap(void)
-{
-       struct memblock_region *r;
-       efi_memory_desc_t *md;
-       u64 paddr, npages, size;
-
-       for_each_memblock(memory, r)
-               create_id_mapping(r->base, r->size, 0);
-
-       /* map runtime io spaces */
-       for_each_efi_memory_desc(&memmap, md) {
-               if (!(md->attribute & EFI_MEMORY_RUNTIME) || is_normal_ram(md))
-                       continue;
-               paddr = md->phys_addr;
-               npages = md->num_pages;
-               memrange_efi_to_native(&paddr, &npages);
-               size = npages << PAGE_SHIFT;
-               create_id_mapping(paddr, size, 1);
-       }
-}
-
 /*
  * Translate a EFI virtual address into a physical address: this is necessary,
  * as some data members of the EFI system table are virtually remapped after

@@ -251,16 +230,13 @@ static struct mm_struct efi_mm = {
        INIT_MM_CONTEXT(efi_mm)
 };
 
-void __init efi_idmap_init(void)
+void __init efi_virtmap_init(void)
 {
        efi_memory_desc_t *md;
 
        if (!efi_enabled(EFI_BOOT))
                return;
 
-       /* boot time idmap_pg_dir is incomplete, so fill in missing parts */
-       efi_setup_idmap();
-
        for_each_efi_memory_desc(&memmap, md) {
                u64 paddr, npages, size;
 

arch/arm64/kernel/setup.c +1 -1

index 2437196cc5d4..a6028490a28f 100644
@@ -392,7 +392,7 @@ void __init setup_arch(char **cmdline_p)
        paging_init();
        request_standard_resources();
 
-       efi_idmap_init();
+       efi_virtmap_init();
 
        unflatten_device_tree();
 

arch/arm64/mm/mmu.c +0 -11

index f7d17a5a1f56..19eae06aab81 100644
@@ -282,16 +282,6 @@ static void __init create_mapping(phys_addr_t phys, unsigned long virt,
                         size, 0, 0);
 }
 
-void __init create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io)
-{
-       if ((addr >> PGDIR_SHIFT) >= ARRAY_SIZE(idmap_pg_dir)) {
-               pr_warn("BUG: not creating id mapping for %pa\n", &addr);
-               return;
-       }
-       __create_mapping(&init_mm, &idmap_pg_dir[pgd_index(addr)],
-                        addr, addr, size, map_io, 0);
-}
-
 void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
                               unsigned long virt, phys_addr_t size,
                               int map_io, int map_xn)
--

1.8.3.2

Quoted 4 lines · Ard Biesheuvel

Now that we have moved the call to SetVirtualAddressMap() to the stub, UEFI has no use for the ID map, so we can drop the code that installs ID mappings for UEFI memory regions.

Quoted 1 line

Signed-off-by Ard Biesheuvel <[email protected]>

I have to say, this series makes me happy. :-)

This method will go a long way to catching UEFI implementations that do incorrect things after exitbootservices is called. I'm assuming that any attempt to access a region that boot services has not requested will get trapped by the kernel, correct?

Quoted 125 lines

---
 arch/arm64/include/asm/efi.h |  4 ++--
 arch/arm64/include/asm/mmu.h |  2 --
 arch/arm64/kernel/efi.c      | 26 +-------------------------
 arch/arm64/kernel/setup.c    |  2 +-
 arch/arm64/mm/mmu.c          | 10 ----------
 5 files changed, 4 insertions(+), 40 deletions(-)

arch/arm64/include/asm/efi.h +2 -2

index d752e5480096..09854df5a2a1 100644
@@ -6,10 +6,10 @@
 
 #ifdef CONFIG_EFI
 extern void efi_init(void);
-extern void efi_idmap_init(void);
+extern void efi_virtmap_init(void);
 #else
 #define efi_init()
-#define efi_idmap_init()
+#define efi_virtmap_init()
 #endif
 
 void efi_load_rt_mapping(void);

arch/arm64/include/asm/mmu.h +0 -2

index bcf166043a8b..df33cf02300a 100644
@@ -31,8 +31,6 @@ extern void paging_init(void);
 extern void setup_mm_for_reboot(void);
 extern void __iomem *early_io_map(phys_addr_t phys, unsigned long virt);
 extern void init_mem_pgprot(void);
-/* create an identity mapping for memory (or io if map_io is true) */
-extern void create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io);
 extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
                               unsigned long virt, phys_addr_t size,
                               int map_io, int map_xn);

arch/arm64/kernel/efi.c +1 -25

index cbff3e8eff11..02bd7e877f0b 100644
@@ -53,27 +53,6 @@ static int __init is_normal_ram(efi_memory_desc_t *md)
        return 0;
 }
 
-static void __init efi_setup_idmap(void)
-{
-       struct memblock_region *r;
-       efi_memory_desc_t *md;
-       u64 paddr, npages, size;
-
-       for_each_memblock(memory, r)
-               create_id_mapping(r->base, r->size, 0);
-
-       /* map runtime io spaces */
-       for_each_efi_memory_desc(&memmap, md) {
-               if (!(md->attribute & EFI_MEMORY_RUNTIME) || is_normal_ram(md))
-                       continue;
-               paddr = md->phys_addr;
-               npages = md->num_pages;
-               memrange_efi_to_native(&paddr, &npages);
-               size = npages << PAGE_SHIFT;
-               create_id_mapping(paddr, size, 1);
-       }
-}
-
 /*
  * Translate a EFI virtual address into a physical address: this is necessary,
  * as some data members of the EFI system table are virtually remapped after

@@ -251,16 +230,13 @@ static struct mm_struct efi_mm = {
        INIT_MM_CONTEXT(efi_mm)
 };
 
-void __init efi_idmap_init(void)
+void __init efi_virtmap_init(void)
 {
        efi_memory_desc_t *md;
 
        if (!efi_enabled(EFI_BOOT))
                return;
 
-       /* boot time idmap_pg_dir is incomplete, so fill in missing parts */
-       efi_setup_idmap();
-
        for_each_efi_memory_desc(&memmap, md) {
                u64 paddr, npages, size;
 

arch/arm64/kernel/setup.c +1 -1

index 2437196cc5d4..a6028490a28f 100644
@@ -392,7 +392,7 @@ void __init setup_arch(char **cmdline_p)
        paging_init();
        request_standard_resources();
 
-       efi_idmap_init();
+       efi_virtmap_init();
 
        unflatten_device_tree();
 

arch/arm64/mm/mmu.c +0 -11

index f7d17a5a1f56..19eae06aab81 100644
@@ -282,16 +282,6 @@ static void __init create_mapping(phys_addr_t phys, unsigned long virt,
                         size, 0, 0);
 }
 
-void __init create_id_mapping(phys_addr_t addr, phys_addr_t size, int map_io)
-{
-       if ((addr >> PGDIR_SHIFT) >= ARRAY_SIZE(idmap_pg_dir)) {
-               pr_warn("BUG: not creating id mapping for %pa\n", &addr);
-               return;
-       }
-       __create_mapping(&init_mm, &idmap_pg_dir[pgd_index(addr)],
-                        addr, addr, size, map_io, 0);
-}
-
 void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
                               unsigned long virt, phys_addr_t size,
                               int map_io, int map_xn)
--

1.8.3.2

Quoted 6 lines · Grant Likely

Quoted 4 lines · Ard Biesheuvel

Now that we have moved the call to SetVirtualAddressMap() to the stub, UEFI has no use for the ID map, so we can drop the code that installs ID mappings for UEFI memory regions.

Quoted 9 lines

Quoted 1 line

Signed-off-by Ard Biesheuvel <[email protected]>

I have to say, this series makes me happy. :-)

This method will go a long way to catching UEFI implementations that do incorrect things after exitbootservices is called. I'm assuming that any attempt to access a region that boot services has not requested will get trapped by the kernel, correct?

If we really want to catch firmware problems, we should probably wipe all boot services regions between the calls to ExitBootServices() and SetVirtualAddressMap(). Mark Salter's original approach here was fairly cautious here, i.e., reserving boot services regions until after the call to SetVirtualAddressMap(), but there is no point in doing that for kexec, that's why I removed it.