x86/mm: Add support to access boot related data in the clear
authorTom Lendacky <thomas.lendacky@amd.com>
Mon, 17 Jul 2017 21:10:16 +0000 (16:10 -0500)
committerIngo Molnar <mingo@kernel.org>
Tue, 18 Jul 2017 09:38:02 +0000 (11:38 +0200)
Boot data (such as EFI related data) is not encrypted when the system is
booted because UEFI/BIOS does not run with SME active. In order to access
this data properly it needs to be mapped decrypted.

Update early_memremap() to provide an arch specific routine to modify the
pagetable protection attributes before they are applied to the new
mapping. This is used to remove the encryption mask for boot related data.

Update memremap() to provide an arch specific routine to determine if RAM
remapping is allowed.  RAM remapping will cause an encrypted mapping to be
generated. By preventing RAM remapping, ioremap_cache() will be used
instead, which will provide a decrypted mapping of the boot related data.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Matt Fleming <matt@codeblueprint.co.uk>
Reviewed-by: Borislav Petkov <bp@suse.de>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Larry Woodman <lwoodman@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Toshimitsu Kani <toshi.kani@hpe.com>
Cc: kasan-dev@googlegroups.com
Cc: kvm@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Cc: linux-efi@vger.kernel.org
Cc: linux-mm@kvack.org
Link: http://lkml.kernel.org/r/81fb6b4117a5df6b9f2eda342f81bbef4b23d2e5.1500319216.git.thomas.lendacky@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/include/asm/io.h
arch/x86/mm/ioremap.c
include/linux/io.h
kernel/memremap.c
mm/early_ioremap.c

index 7afb0e2f07f40d69ccbe44a70baea52bc70ef685..09c5557b1454e5160af35c0b75502278c94a4cd5 100644 (file)
@@ -381,4 +381,9 @@ extern void arch_io_free_memtype_wc(resource_size_t start, resource_size_t size)
 #define arch_io_reserve_memtype_wc arch_io_reserve_memtype_wc
 #endif
 
+extern bool arch_memremap_can_ram_remap(resource_size_t offset,
+                                       unsigned long size,
+                                       unsigned long flags);
+#define arch_memremap_can_ram_remap arch_memremap_can_ram_remap
+
 #endif /* _ASM_X86_IO_H */
index 570201bbf442488ea62d0bb1a89e65d84683a2a5..8986b28689440be44ee80d44bdbc82188811d1a4 100644 (file)
@@ -13,6 +13,8 @@
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
 #include <linux/mmiotrace.h>
+#include <linux/mem_encrypt.h>
+#include <linux/efi.h>
 
 #include <asm/set_memory.h>
 #include <asm/e820/api.h>
@@ -21,6 +23,7 @@
 #include <asm/tlbflush.h>
 #include <asm/pgalloc.h>
 #include <asm/pat.h>
+#include <asm/setup.h>
 
 #include "physaddr.h"
 
@@ -417,6 +420,183 @@ void unxlate_dev_mem_ptr(phys_addr_t phys, void *addr)
        iounmap((void __iomem *)((unsigned long)addr & PAGE_MASK));
 }
 
+/*
+ * Examine the physical address to determine if it is an area of memory
+ * that should be mapped decrypted.  If the memory is not part of the
+ * kernel usable area it was accessed and created decrypted, so these
+ * areas should be mapped decrypted.
+ */
+static bool memremap_should_map_decrypted(resource_size_t phys_addr,
+                                         unsigned long size)
+{
+       /* Check if the address is outside kernel usable area */
+       switch (e820__get_entry_type(phys_addr, phys_addr + size - 1)) {
+       case E820_TYPE_RESERVED:
+       case E820_TYPE_ACPI:
+       case E820_TYPE_NVS:
+       case E820_TYPE_UNUSABLE:
+               return true;
+       default:
+               break;
+       }
+
+       return false;
+}
+
+/*
+ * Examine the physical address to determine if it is EFI data. Check
+ * it against the boot params structure and EFI tables and memory types.
+ */
+static bool memremap_is_efi_data(resource_size_t phys_addr,
+                                unsigned long size)
+{
+       u64 paddr;
+
+       /* Check if the address is part of EFI boot/runtime data */
+       if (!efi_enabled(EFI_BOOT))
+               return false;
+
+       paddr = boot_params.efi_info.efi_memmap_hi;
+       paddr <<= 32;
+       paddr |= boot_params.efi_info.efi_memmap;
+       if (phys_addr == paddr)
+               return true;
+
+       paddr = boot_params.efi_info.efi_systab_hi;
+       paddr <<= 32;
+       paddr |= boot_params.efi_info.efi_systab;
+       if (phys_addr == paddr)
+               return true;
+
+       if (efi_is_table_address(phys_addr))
+               return true;
+
+       switch (efi_mem_type(phys_addr)) {
+       case EFI_BOOT_SERVICES_DATA:
+       case EFI_RUNTIME_SERVICES_DATA:
+               return true;
+       default:
+               break;
+       }
+
+       return false;
+}
+
+/*
+ * Examine the physical address to determine if it is boot data by checking
+ * it against the boot params setup_data chain.
+ */
+static bool memremap_is_setup_data(resource_size_t phys_addr,
+                                  unsigned long size)
+{
+       struct setup_data *data;
+       u64 paddr, paddr_next;
+
+       paddr = boot_params.hdr.setup_data;
+       while (paddr) {
+               unsigned int len;
+
+               if (phys_addr == paddr)
+                       return true;
+
+               data = memremap(paddr, sizeof(*data),
+                               MEMREMAP_WB | MEMREMAP_DEC);
+
+               paddr_next = data->next;
+               len = data->len;
+
+               memunmap(data);
+
+               if ((phys_addr > paddr) && (phys_addr < (paddr + len)))
+                       return true;
+
+               paddr = paddr_next;
+       }
+
+       return false;
+}
+
+/*
+ * Examine the physical address to determine if it is boot data by checking
+ * it against the boot params setup_data chain (early boot version).
+ */
+static bool __init early_memremap_is_setup_data(resource_size_t phys_addr,
+                                               unsigned long size)
+{
+       struct setup_data *data;
+       u64 paddr, paddr_next;
+
+       paddr = boot_params.hdr.setup_data;
+       while (paddr) {
+               unsigned int len;
+
+               if (phys_addr == paddr)
+                       return true;
+
+               data = early_memremap_decrypted(paddr, sizeof(*data));
+
+               paddr_next = data->next;
+               len = data->len;
+
+               early_memunmap(data, sizeof(*data));
+
+               if ((phys_addr > paddr) && (phys_addr < (paddr + len)))
+                       return true;
+
+               paddr = paddr_next;
+       }
+
+       return false;
+}
+
+/*
+ * Architecture function to determine if RAM remap is allowed. By default, a
+ * RAM remap will map the data as encrypted. Determine if a RAM remap should
+ * not be done so that the data will be mapped decrypted.
+ */
+bool arch_memremap_can_ram_remap(resource_size_t phys_addr, unsigned long size,
+                                unsigned long flags)
+{
+       if (!sme_active())
+               return true;
+
+       if (flags & MEMREMAP_ENC)
+               return true;
+
+       if (flags & MEMREMAP_DEC)
+               return false;
+
+       if (memremap_is_setup_data(phys_addr, size) ||
+           memremap_is_efi_data(phys_addr, size) ||
+           memremap_should_map_decrypted(phys_addr, size))
+               return false;
+
+       return true;
+}
+
+/*
+ * Architecture override of __weak function to adjust the protection attributes
+ * used when remapping memory. By default, early_memremap() will map the data
+ * as encrypted. Determine if an encrypted mapping should not be done and set
+ * the appropriate protection attributes.
+ */
+pgprot_t __init early_memremap_pgprot_adjust(resource_size_t phys_addr,
+                                            unsigned long size,
+                                            pgprot_t prot)
+{
+       if (!sme_active())
+               return prot;
+
+       if (early_memremap_is_setup_data(phys_addr, size) ||
+           memremap_is_efi_data(phys_addr, size) ||
+           memremap_should_map_decrypted(phys_addr, size))
+               prot = pgprot_decrypted(prot);
+       else
+               prot = pgprot_encrypted(prot);
+
+       return prot;
+}
+
 #ifdef CONFIG_ARCH_USE_MEMREMAP_PROT
 /* Remap memory with encryption */
 void __init *early_memremap_encrypted(resource_size_t phys_addr,
index 2195d9ea4aaae0c054f04aab2da7cff851d2b997..32e30e8fb9db92cf1472c6188ff3310a610b3c5f 100644 (file)
@@ -157,6 +157,8 @@ enum {
        MEMREMAP_WB = 1 << 0,
        MEMREMAP_WT = 1 << 1,
        MEMREMAP_WC = 1 << 2,
+       MEMREMAP_ENC = 1 << 3,
+       MEMREMAP_DEC = 1 << 4,
 };
 
 void *memremap(resource_size_t offset, size_t size, unsigned long flags);
index 124bed776532d3d2e69f4079e3f2d23da193fd2e..9afdc434fb490a3384d847bc50647fa3dd3ab16a 100644 (file)
@@ -34,13 +34,24 @@ static void *arch_memremap_wb(resource_size_t offset, unsigned long size)
 }
 #endif
 
-static void *try_ram_remap(resource_size_t offset, size_t size)
+#ifndef arch_memremap_can_ram_remap
+static bool arch_memremap_can_ram_remap(resource_size_t offset, size_t size,
+                                       unsigned long flags)
+{
+       return true;
+}
+#endif
+
+static void *try_ram_remap(resource_size_t offset, size_t size,
+                          unsigned long flags)
 {
        unsigned long pfn = PHYS_PFN(offset);
 
        /* In the simple case just return the existing linear address */
-       if (pfn_valid(pfn) && !PageHighMem(pfn_to_page(pfn)))
+       if (pfn_valid(pfn) && !PageHighMem(pfn_to_page(pfn)) &&
+           arch_memremap_can_ram_remap(offset, size, flags))
                return __va(offset);
+
        return NULL; /* fallback to arch_memremap_wb */
 }
 
@@ -48,7 +59,8 @@ static void *try_ram_remap(resource_size_t offset, size_t size)
  * memremap() - remap an iomem_resource as cacheable memory
  * @offset: iomem resource start address
  * @size: size of remap
- * @flags: any of MEMREMAP_WB, MEMREMAP_WT and MEMREMAP_WC
+ * @flags: any of MEMREMAP_WB, MEMREMAP_WT, MEMREMAP_WC,
+ *               MEMREMAP_ENC, MEMREMAP_DEC
  *
  * memremap() is "ioremap" for cases where it is known that the resource
  * being mapped does not have i/o side effects and the __iomem
@@ -95,7 +107,7 @@ void *memremap(resource_size_t offset, size_t size, unsigned long flags)
                 * the requested range is potentially in System RAM.
                 */
                if (is_ram == REGION_INTERSECTS)
-                       addr = try_ram_remap(offset, size);
+                       addr = try_ram_remap(offset, size, flags);
                if (!addr)
                        addr = arch_memremap_wb(offset, size);
        }
index d7d30da754ba6fc6f03364fef80bb6bf09fb04b8..b1dd4a948fc0b3afc375964d23ebcb9f69eaafa3 100644 (file)
@@ -30,6 +30,13 @@ early_param("early_ioremap_debug", early_ioremap_debug_setup);
 
 static int after_paging_init __initdata;
 
+pgprot_t __init __weak early_memremap_pgprot_adjust(resource_size_t phys_addr,
+                                                   unsigned long size,
+                                                   pgprot_t prot)
+{
+       return prot;
+}
+
 void __init __weak early_ioremap_shutdown(void)
 {
 }
@@ -215,14 +222,19 @@ early_ioremap(resource_size_t phys_addr, unsigned long size)
 void __init *
 early_memremap(resource_size_t phys_addr, unsigned long size)
 {
-       return (__force void *)__early_ioremap(phys_addr, size,
-                                              FIXMAP_PAGE_NORMAL);
+       pgprot_t prot = early_memremap_pgprot_adjust(phys_addr, size,
+                                                    FIXMAP_PAGE_NORMAL);
+
+       return (__force void *)__early_ioremap(phys_addr, size, prot);
 }
 #ifdef FIXMAP_PAGE_RO
 void __init *
 early_memremap_ro(resource_size_t phys_addr, unsigned long size)
 {
-       return (__force void *)__early_ioremap(phys_addr, size, FIXMAP_PAGE_RO);
+       pgprot_t prot = early_memremap_pgprot_adjust(phys_addr, size,
+                                                    FIXMAP_PAGE_RO);
+
+       return (__force void *)__early_ioremap(phys_addr, size, prot);
 }
 #endif