Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / arch / arm64 / kernel / efi.c
diff --git a/kernel/arch/arm64/kernel/efi.c b/kernel/arch/arm64/kernel/efi.c
new file mode 100644 (file)
index 0000000..ab21e0d
--- /dev/null
@@ -0,0 +1,369 @@
+/*
+ * Extensible Firmware Interface
+ *
+ * Based on Extensible Firmware Interface Specification version 2.4
+ *
+ * Copyright (C) 2013, 2014 Linaro Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/atomic.h>
+#include <linux/dmi.h>
+#include <linux/efi.h>
+#include <linux/export.h>
+#include <linux/memblock.h>
+#include <linux/mm_types.h>
+#include <linux/bootmem.h>
+#include <linux/of.h>
+#include <linux/of_fdt.h>
+#include <linux/preempt.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+
+#include <asm/cacheflush.h>
+#include <asm/efi.h>
+#include <asm/tlbflush.h>
+#include <asm/mmu_context.h>
+#include <asm/mmu.h>
+#include <asm/pgtable.h>
+
+struct efi_memory_map memmap;
+
+static u64 efi_system_table;
+
+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)
+};
+
+static int uefi_debug __initdata;
+static int __init uefi_debug_setup(char *str)
+{
+       uefi_debug = 1;
+
+       return 0;
+}
+early_param("uefi_debug", uefi_debug_setup);
+
+static int __init is_normal_ram(efi_memory_desc_t *md)
+{
+       if (md->attribute & EFI_MEMORY_WB)
+               return 1;
+       return 0;
+}
+
+/*
+ * 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 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;
+       }
+       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;
+
+       efi.systab = early_memremap(efi_system_table,
+                                   sizeof(efi_system_table_t));
+       if (efi.systab == NULL) {
+               pr_warn("Unable to map EFI system table.\n");
+               return -ENOMEM;
+       }
+
+       set_bit(EFI_BOOT, &efi.flags);
+       set_bit(EFI_64BIT, &efi.flags);
+
+       /*
+        * Verify the EFI Table
+        */
+       if (efi.systab->hdr.signature != EFI_SYSTEM_TABLE_SIGNATURE) {
+               pr_err("System table signature incorrect\n");
+               retval = -EINVAL;
+               goto out;
+       }
+       if ((efi.systab->hdr.revision >> 16) < 2)
+               pr_warn("Warning: EFI system table version %d.%02d, expected 2.00 or greater\n",
+                       efi.systab->hdr.revision >> 16,
+                       efi.systab->hdr.revision & 0xffff);
+
+       /* Show what we know for posterity */
+       c16 = early_memremap(efi_to_phys(efi.systab->fw_vendor),
+                            sizeof(vendor));
+       if (c16) {
+               for (i = 0; i < (int) sizeof(vendor) - 1 && *c16; ++i)
+                       vendor[i] = c16[i];
+               vendor[i] = '\0';
+               early_memunmap(c16, sizeof(vendor));
+       }
+
+       pr_info("EFI v%u.%.02u by %s\n",
+               efi.systab->hdr.revision >> 16,
+               efi.systab->hdr.revision & 0xffff, vendor);
+
+       table_size = sizeof(efi_config_table_64_t) * efi.systab->nr_tables;
+       config_tables = early_memremap(efi_to_phys(efi.systab->tables),
+                                      table_size);
+
+       retval = efi_config_parse_tables(config_tables, efi.systab->nr_tables,
+                                        sizeof(efi_config_table_64_t), NULL);
+
+       early_memunmap(config_tables, table_size);
+out:
+       early_memunmap(efi.systab,  sizeof(efi_system_table_t));
+       return retval;
+}
+
+/*
+ * Return true for RAM regions we want to permanently reserve.
+ */
+static __init int is_reserve_region(efi_memory_desc_t *md)
+{
+       switch (md->type) {
+       case EFI_LOADER_CODE:
+       case EFI_LOADER_DATA:
+       case EFI_BOOT_SERVICES_CODE:
+       case EFI_BOOT_SERVICES_DATA:
+       case EFI_CONVENTIONAL_MEMORY:
+               return 0;
+       default:
+               break;
+       }
+       return is_normal_ram(md);
+}
+
+static __init void reserve_regions(void)
+{
+       efi_memory_desc_t *md;
+       u64 paddr, npages, size;
+
+       if (uefi_debug)
+               pr_info("Processing EFI memory map:\n");
+
+       for_each_efi_memory_desc(&memmap, md) {
+               paddr = md->phys_addr;
+               npages = md->num_pages;
+
+               if (uefi_debug) {
+                       char buf[64];
+
+                       pr_info("  0x%012llx-0x%012llx %s",
+                               paddr, paddr + (npages << EFI_PAGE_SHIFT) - 1,
+                               efi_md_typeattr_format(buf, sizeof(buf), md));
+               }
+
+               memrange_efi_to_native(&paddr, &npages);
+               size = npages << PAGE_SHIFT;
+
+               if (is_normal_ram(md))
+                       early_init_dt_add_memory_arch(paddr, size);
+
+               if (is_reserve_region(md)) {
+                       memblock_reserve(paddr, size);
+                       if (uefi_debug)
+                               pr_cont("*");
+               }
+
+               if (uefi_debug)
+                       pr_cont("\n");
+       }
+
+       set_bit(EFI_MEMMAP, &efi.flags);
+}
+
+void __init efi_init(void)
+{
+       struct efi_fdt_params params;
+
+       /* Grab UEFI information placed in FDT by stub */
+       if (!efi_get_fdt_params(&params, uefi_debug))
+               return;
+
+       efi_system_table = params.system_table;
+
+       memblock_reserve(params.mmap & PAGE_MASK,
+                        PAGE_ALIGN(params.mmap_size + (params.mmap & ~PAGE_MASK)));
+       memmap.phys_map = (void *)params.mmap;
+       memmap.map = early_memremap(params.mmap, params.mmap_size);
+       memmap.map_end = memmap.map + params.mmap_size;
+       memmap.desc_size = params.desc_size;
+       memmap.desc_version = params.desc_ver;
+
+       if (uefi_init() < 0)
+               return;
+
+       reserve_regions();
+       early_memunmap(memmap.map, params.mmap_size);
+}
+
+static bool __init efi_virtmap_init(void)
+{
+       efi_memory_desc_t *md;
+
+       for_each_efi_memory_desc(&memmap, md) {
+               u64 paddr, npages, size;
+               pgprot_t prot;
+
+               if (!(md->attribute & EFI_MEMORY_RUNTIME))
+                       continue;
+               if (md->virt_addr == 0)
+                       return false;
+
+               paddr = md->phys_addr;
+               npages = md->num_pages;
+               memrange_efi_to_native(&paddr, &npages);
+               size = npages << PAGE_SHIFT;
+
+               pr_info("  EFI remap 0x%016llx => %p\n",
+                       md->phys_addr, (void *)md->virt_addr);
+
+               /*
+                * Only regions of type EFI_RUNTIME_SERVICES_CODE need to be
+                * executable, everything else can be mapped with the XN bits
+                * set.
+                */
+               if (!is_normal_ram(md))
+                       prot = __pgprot(PROT_DEVICE_nGnRE);
+               else if (md->type == EFI_RUNTIME_SERVICES_CODE)
+                       prot = PAGE_KERNEL_EXEC;
+               else
+                       prot = PAGE_KERNEL;
+
+               create_pgd_mapping(&efi_mm, paddr, md->virt_addr, size, prot);
+       }
+       return true;
+}
+
+/*
+ * Enable the UEFI Runtime Services if all prerequisites are in place, i.e.,
+ * non-early mapping of the UEFI system table and virtual mappings for all
+ * EFI_MEMORY_RUNTIME regions.
+ */
+static int __init arm64_enable_runtime_services(void)
+{
+       u64 mapsize;
+
+       if (!efi_enabled(EFI_BOOT)) {
+               pr_info("EFI services will not be available.\n");
+               return -1;
+       }
+
+       if (efi_runtime_disabled()) {
+               pr_info("EFI runtime services will be disabled.\n");
+               return -1;
+       }
+
+       pr_info("Remapping and enabling EFI services.\n");
+
+       mapsize = memmap.map_end - memmap.map;
+       memmap.map = (__force void *)ioremap_cache((phys_addr_t)memmap.phys_map,
+                                                  mapsize);
+       if (!memmap.map) {
+               pr_err("Failed to remap EFI memory map\n");
+               return -1;
+       }
+       memmap.map_end = memmap.map + mapsize;
+       efi.memmap = &memmap;
+
+       efi.systab = (__force void *)ioremap_cache(efi_system_table,
+                                                  sizeof(efi_system_table_t));
+       if (!efi.systab) {
+               pr_err("Failed to remap EFI System Table\n");
+               return -1;
+       }
+       set_bit(EFI_SYSTEM_TABLES, &efi.flags);
+
+       if (!efi_virtmap_init()) {
+               pr_err("No UEFI virtual mapping was installed -- runtime services will not be available\n");
+               return -1;
+       }
+
+       /* Set up runtime services function pointers */
+       efi_native_runtime_setup();
+       set_bit(EFI_RUNTIME_SERVICES, &efi.flags);
+
+       efi.runtime_version = efi.systab->hdr.revision;
+
+       return 0;
+}
+early_initcall(arm64_enable_runtime_services);
+
+static int __init arm64_dmi_init(void)
+{
+       /*
+        * On arm64, DMI depends on UEFI, and dmi_scan_machine() needs to
+        * be called early because dmi_id_init(), which is an arch_initcall
+        * itself, depends on dmi_scan_machine() having been called already.
+        */
+       dmi_scan_machine();
+       if (dmi_available)
+               dmi_set_dump_stack_arch_desc();
+       return 0;
+}
+core_initcall(arm64_dmi_init);
+
+static void efi_set_pgd(struct mm_struct *mm)
+{
+       if (mm == &init_mm)
+               cpu_set_reserved_ttbr0();
+       else
+               cpu_switch_mm(mm->pgd, mm);
+
+       flush_tlb_all();
+       if (icache_is_aivivt())
+               __flush_icache_all();
+}
+
+void efi_virtmap_load(void)
+{
+       preempt_disable();
+       efi_set_pgd(&efi_mm);
+}
+
+void efi_virtmap_unload(void)
+{
+       efi_set_pgd(current->active_mm);
+       preempt_enable();
+}
+
+/*
+ * UpdateCapsule() depends on the system being shutdown via
+ * ResetSystem().
+ */
+bool efi_poweroff_required(void)
+{
+       return efi_enabled(EFI_RUNTIME_SERVICES);
+}