diff options
Diffstat (limited to 'arch/x86/boot/startup')
| -rw-r--r-- | arch/x86/boot/startup/Makefile | 30 | ||||
| -rw-r--r-- | arch/x86/boot/startup/efi-mixed.S | 253 | ||||
| -rw-r--r-- | arch/x86/boot/startup/gdt_idt.c | 71 | ||||
| -rw-r--r-- | arch/x86/boot/startup/la57toggle.S | 111 | ||||
| -rw-r--r-- | arch/x86/boot/startup/map_kernel.c | 217 | ||||
| -rw-r--r-- | arch/x86/boot/startup/sev-shared.c | 887 | ||||
| -rw-r--r-- | arch/x86/boot/startup/sev-startup.c | 368 | ||||
| -rw-r--r-- | arch/x86/boot/startup/sme.c | 575 |
8 files changed, 2512 insertions, 0 deletions
diff --git a/arch/x86/boot/startup/Makefile b/arch/x86/boot/startup/Makefile new file mode 100644 index 000000000000..b514f7e81332 --- /dev/null +++ b/arch/x86/boot/startup/Makefile @@ -0,0 +1,30 @@ +# SPDX-License-Identifier: GPL-2.0 + +KBUILD_AFLAGS += -D__DISABLE_EXPORTS +KBUILD_CFLAGS += -D__DISABLE_EXPORTS -mcmodel=small -fPIC \ + -Os -DDISABLE_BRANCH_PROFILING \ + $(DISABLE_STACKLEAK_PLUGIN) \ + -fno-stack-protector -D__NO_FORTIFY \ + -fno-jump-tables \ + -include $(srctree)/include/linux/hidden.h + +# disable ftrace hooks and LTO +KBUILD_CFLAGS := $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) +KBUILD_CFLAGS := $(filter-out $(CC_FLAGS_LTO),$(KBUILD_CFLAGS)) +KASAN_SANITIZE := n +KCSAN_SANITIZE := n +KMSAN_SANITIZE := n +UBSAN_SANITIZE := n +KCOV_INSTRUMENT := n + +obj-$(CONFIG_X86_64) += gdt_idt.o map_kernel.o +obj-$(CONFIG_AMD_MEM_ENCRYPT) += sme.o sev-startup.o + +lib-$(CONFIG_X86_64) += la57toggle.o +lib-$(CONFIG_EFI_MIXED) += efi-mixed.o + +# +# Disable objtool validation for all library code, which is intended +# to be linked into the decompressor or the EFI stub but not vmlinux +# +$(patsubst %.o,$(obj)/%.o,$(lib-y)): OBJECT_FILES_NON_STANDARD := y diff --git a/arch/x86/boot/startup/efi-mixed.S b/arch/x86/boot/startup/efi-mixed.S new file mode 100644 index 000000000000..e04ed99bc449 --- /dev/null +++ b/arch/x86/boot/startup/efi-mixed.S @@ -0,0 +1,253 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2014, 2015 Intel Corporation; author Matt Fleming + * + * Early support for invoking 32-bit EFI services from a 64-bit kernel. + * + * Because this thunking occurs before ExitBootServices() we have to + * restore the firmware's 32-bit GDT and IDT before we make EFI service + * calls. + * + * On the plus side, we don't have to worry about mangling 64-bit + * addresses into 32-bits because we're executing with an identity + * mapped pagetable and haven't transitioned to 64-bit virtual addresses + * yet. + */ + +#include <linux/linkage.h> +#include <asm/desc_defs.h> +#include <asm/msr.h> +#include <asm/page_types.h> +#include <asm/pgtable_types.h> +#include <asm/processor-flags.h> +#include <asm/segment.h> + + .text + .code32 +#ifdef CONFIG_EFI_HANDOVER_PROTOCOL +SYM_FUNC_START(efi32_stub_entry) + call 1f +1: popl %ecx + + /* Clear BSS */ + xorl %eax, %eax + leal (_bss - 1b)(%ecx), %edi + leal (_ebss - 1b)(%ecx), %ecx + subl %edi, %ecx + shrl $2, %ecx + cld + rep stosl + + add $0x4, %esp /* Discard return address */ + movl 8(%esp), %ebx /* struct boot_params pointer */ + jmp efi32_startup +SYM_FUNC_END(efi32_stub_entry) +#endif + +/* + * Called using a far call from __efi64_thunk() below, using the x86_64 SysV + * ABI (except for R8/R9 which are inaccessible to 32-bit code - EAX/EBX are + * used instead). EBP+16 points to the arguments passed via the stack. + * + * The first argument (EDI) is a pointer to the boot service or protocol, to + * which the remaining arguments are passed, each truncated to 32 bits. + */ +SYM_FUNC_START_LOCAL(efi_enter32) + /* + * Convert x86-64 SysV ABI params to i386 ABI + */ + pushl 32(%ebp) /* Up to 3 args passed via the stack */ + pushl 24(%ebp) + pushl 16(%ebp) + pushl %ebx /* R9 */ + pushl %eax /* R8 */ + pushl %ecx + pushl %edx + pushl %esi + + /* Disable paging */ + movl %cr0, %eax + btrl $X86_CR0_PG_BIT, %eax + movl %eax, %cr0 + + /* Disable long mode via EFER */ + movl $MSR_EFER, %ecx + rdmsr + btrl $_EFER_LME, %eax + wrmsr + + call *%edi + + /* We must preserve return value */ + movl %eax, %edi + + call efi32_enable_long_mode + + addl $32, %esp + movl %edi, %eax + lret +SYM_FUNC_END(efi_enter32) + + .code64 +SYM_FUNC_START(__efi64_thunk) + push %rbp + movl %esp, %ebp + push %rbx + + /* Move args #5 and #6 into 32-bit accessible registers */ + movl %r8d, %eax + movl %r9d, %ebx + + lcalll *efi32_call(%rip) + + pop %rbx + pop %rbp + RET +SYM_FUNC_END(__efi64_thunk) + + .code32 +SYM_FUNC_START_LOCAL(efi32_enable_long_mode) + movl %cr4, %eax + btsl $(X86_CR4_PAE_BIT), %eax + movl %eax, %cr4 + + movl $MSR_EFER, %ecx + rdmsr + btsl $_EFER_LME, %eax + wrmsr + + /* Disable interrupts - the firmware's IDT does not work in long mode */ + cli + + /* Enable paging */ + movl %cr0, %eax + btsl $X86_CR0_PG_BIT, %eax + movl %eax, %cr0 + ret +SYM_FUNC_END(efi32_enable_long_mode) + +/* + * This is the common EFI stub entry point for mixed mode. It sets up the GDT + * and page tables needed for 64-bit execution, after which it calls the + * common 64-bit EFI entrypoint efi_stub_entry(). + * + * Arguments: 0(%esp) image handle + * 4(%esp) EFI system table pointer + * %ebx struct boot_params pointer (or NULL) + * + * Since this is the point of no return for ordinary execution, no registers + * are considered live except for the function parameters. [Note that the EFI + * stub may still exit and return to the firmware using the Exit() EFI boot + * service.] + */ +SYM_FUNC_START_LOCAL(efi32_startup) + movl %esp, %ebp + + subl $8, %esp + sgdtl (%esp) /* Save GDT descriptor to the stack */ + movl 2(%esp), %esi /* Existing GDT pointer */ + movzwl (%esp), %ecx /* Existing GDT limit */ + inc %ecx /* Existing GDT size */ + andl $~7, %ecx /* Ensure size is multiple of 8 */ + + subl %ecx, %esp /* Allocate new GDT */ + andl $~15, %esp /* Realign the stack */ + movl %esp, %edi /* New GDT address */ + leal 7(%ecx), %eax /* New GDT limit */ + pushw %cx /* Push 64-bit CS (for LJMP below) */ + pushl %edi /* Push new GDT address */ + pushw %ax /* Push new GDT limit */ + + /* Copy GDT to the stack and add a 64-bit code segment at the end */ + movl $GDT_ENTRY(DESC_CODE64, 0, 0xfffff) & 0xffffffff, (%edi,%ecx) + movl $GDT_ENTRY(DESC_CODE64, 0, 0xfffff) >> 32, 4(%edi,%ecx) + shrl $2, %ecx + cld + rep movsl /* Copy the firmware GDT */ + lgdtl (%esp) /* Switch to the new GDT */ + + call 1f +1: pop %edi + + /* Record mixed mode entry */ + movb $0x0, (efi_is64 - 1b)(%edi) + + /* Set up indirect far call to re-enter 32-bit mode */ + leal (efi32_call - 1b)(%edi), %eax + addl %eax, (%eax) + movw %cs, 4(%eax) + + /* Disable paging */ + movl %cr0, %eax + btrl $X86_CR0_PG_BIT, %eax + movl %eax, %cr0 + + /* Set up 1:1 mapping */ + leal (pte - 1b)(%edi), %eax + movl $_PAGE_PRESENT | _PAGE_RW | _PAGE_PSE, %ecx + leal (_PAGE_PRESENT | _PAGE_RW)(%eax), %edx +2: movl %ecx, (%eax) + addl $8, %eax + addl $PMD_SIZE, %ecx + jnc 2b + + movl $PAGE_SIZE, %ecx + .irpc l, 0123 + movl %edx, \l * 8(%eax) + addl %ecx, %edx + .endr + addl %ecx, %eax + movl %edx, (%eax) + movl %eax, %cr3 + + call efi32_enable_long_mode + + /* Set up far jump to 64-bit mode (CS is already on the stack) */ + leal (efi_stub_entry - 1b)(%edi), %eax + movl %eax, 2(%esp) + + movl 0(%ebp), %edi + movl 4(%ebp), %esi + movl %ebx, %edx + ljmpl *2(%esp) +SYM_FUNC_END(efi32_startup) + +/* + * efi_status_t efi32_pe_entry(efi_handle_t image_handle, + * efi_system_table_32_t *sys_table) + */ +SYM_FUNC_START(efi32_pe_entry) + pushl %ebx // save callee-save registers + + /* Check whether the CPU supports long mode */ + movl $0x80000001, %eax // assume extended info support + cpuid + btl $29, %edx // check long mode bit + jnc 1f + leal 8(%esp), %esp // preserve stack alignment + xor %ebx, %ebx // no struct boot_params pointer + jmp efi32_startup // only ESP and EBX remain live +1: movl $0x80000003, %eax // EFI_UNSUPPORTED + popl %ebx + RET +SYM_FUNC_END(efi32_pe_entry) + +#ifdef CONFIG_EFI_HANDOVER_PROTOCOL + .org efi32_stub_entry + 0x200 + .code64 +SYM_FUNC_START_NOALIGN(efi64_stub_entry) + jmp efi_handover_entry +SYM_FUNC_END(efi64_stub_entry) +#endif + + .data + .balign 8 +SYM_DATA_START_LOCAL(efi32_call) + .long efi_enter32 - . + .word 0x0 +SYM_DATA_END(efi32_call) +SYM_DATA(efi_is64, .byte 1) + + .bss + .balign PAGE_SIZE +SYM_DATA_LOCAL(pte, .fill 6 * PAGE_SIZE, 1, 0) diff --git a/arch/x86/boot/startup/gdt_idt.c b/arch/x86/boot/startup/gdt_idt.c new file mode 100644 index 000000000000..a3112a69b06a --- /dev/null +++ b/arch/x86/boot/startup/gdt_idt.c @@ -0,0 +1,71 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include <linux/linkage.h> +#include <linux/types.h> + +#include <asm/desc.h> +#include <asm/init.h> +#include <asm/setup.h> +#include <asm/sev.h> +#include <asm/trapnr.h> + +/* + * Data structures and code used for IDT setup in head_64.S. The bringup-IDT is + * used until the idt_table takes over. On the boot CPU this happens in + * x86_64_start_kernel(), on secondary CPUs in start_secondary(). In both cases + * this happens in the functions called from head_64.S. + * + * The idt_table can't be used that early because all the code modifying it is + * in idt.c and can be instrumented by tracing or KASAN, which both don't work + * during early CPU bringup. Also the idt_table has the runtime vectors + * configured which require certain CPU state to be setup already (like TSS), + * which also hasn't happened yet in early CPU bringup. + */ +static gate_desc bringup_idt_table[NUM_EXCEPTION_VECTORS] __page_aligned_data; + +/* This may run while still in the direct mapping */ +void __head startup_64_load_idt(void *vc_handler) +{ + struct desc_ptr desc = { + .address = (unsigned long)rip_rel_ptr(bringup_idt_table), + .size = sizeof(bringup_idt_table) - 1, + }; + struct idt_data data; + gate_desc idt_desc; + + /* @vc_handler is set only for a VMM Communication Exception */ + if (vc_handler) { + init_idt_data(&data, X86_TRAP_VC, vc_handler); + idt_init_desc(&idt_desc, &data); + native_write_idt_entry((gate_desc *)desc.address, X86_TRAP_VC, &idt_desc); + } + + native_load_idt(&desc); +} + +/* + * Setup boot CPU state needed before kernel switches to virtual addresses. + */ +void __head startup_64_setup_gdt_idt(void) +{ + struct gdt_page *gp = rip_rel_ptr((void *)(__force unsigned long)&gdt_page); + void *handler = NULL; + + struct desc_ptr startup_gdt_descr = { + .address = (unsigned long)gp->gdt, + .size = GDT_SIZE - 1, + }; + + /* Load GDT */ + native_load_gdt(&startup_gdt_descr); + + /* New GDT is live - reload data segment registers */ + asm volatile("movl %%eax, %%ds\n" + "movl %%eax, %%ss\n" + "movl %%eax, %%es\n" : : "a"(__KERNEL_DS) : "memory"); + + if (IS_ENABLED(CONFIG_AMD_MEM_ENCRYPT)) + handler = rip_rel_ptr(vc_no_ghcb); + + startup_64_load_idt(handler); +} diff --git a/arch/x86/boot/startup/la57toggle.S b/arch/x86/boot/startup/la57toggle.S new file mode 100644 index 000000000000..370075b4d95b --- /dev/null +++ b/arch/x86/boot/startup/la57toggle.S @@ -0,0 +1,111 @@ +/* SPDX-License-Identifier: GPL-2.0 */ + +#include <linux/linkage.h> +#include <asm/segment.h> +#include <asm/boot.h> +#include <asm/msr.h> +#include <asm/processor-flags.h> + +/* + * This is the 32-bit trampoline that will be copied over to low memory. It + * will be called using the ordinary 64-bit calling convention from code + * running in 64-bit mode. + * + * Return address is at the top of the stack (might be above 4G). + * The first argument (EDI) contains the address of the temporary PGD level + * page table in 32-bit addressable memory which will be programmed into + * register CR3. + */ + + .section ".rodata", "a", @progbits +SYM_CODE_START(trampoline_32bit_src) + /* + * Preserve callee save 64-bit registers on the stack: this is + * necessary because the architecture does not guarantee that GPRs will + * retain their full 64-bit values across a 32-bit mode switch. + */ + pushq %r15 + pushq %r14 + pushq %r13 + pushq %r12 + pushq %rbp + pushq %rbx + + /* Preserve top half of RSP in a legacy mode GPR to avoid truncation */ + movq %rsp, %rbx + shrq $32, %rbx + + /* Switch to compatibility mode (CS.L = 0 CS.D = 1) via far return */ + pushq $__KERNEL32_CS + leaq 0f(%rip), %rax + pushq %rax + lretq + + /* + * The 32-bit code below will do a far jump back to long mode and end + * up here after reconfiguring the number of paging levels. First, the + * stack pointer needs to be restored to its full 64-bit value before + * the callee save register contents can be popped from the stack. + */ +.Lret: + shlq $32, %rbx + orq %rbx, %rsp + + /* Restore the preserved 64-bit registers */ + popq %rbx + popq %rbp + popq %r12 + popq %r13 + popq %r14 + popq %r15 + retq + + .code32 +0: + /* Disable paging */ + movl %cr0, %eax + btrl $X86_CR0_PG_BIT, %eax + movl %eax, %cr0 + + /* Point CR3 to the trampoline's new top level page table */ + movl %edi, %cr3 + + /* Set EFER.LME=1 as a precaution in case hypervsior pulls the rug */ + movl $MSR_EFER, %ecx + rdmsr + btsl $_EFER_LME, %eax + /* Avoid writing EFER if no change was made (for TDX guest) */ + jc 1f + wrmsr +1: + /* Toggle CR4.LA57 */ + movl %cr4, %eax + btcl $X86_CR4_LA57_BIT, %eax + movl %eax, %cr4 + + /* Enable paging again. */ + movl %cr0, %eax + btsl $X86_CR0_PG_BIT, %eax + movl %eax, %cr0 + + /* + * Return to the 64-bit calling code using LJMP rather than LRET, to + * avoid the need for a 32-bit addressable stack. The destination + * address will be adjusted after the template code is copied into a + * 32-bit addressable buffer. + */ +.Ljmp: ljmpl $__KERNEL_CS, $(.Lret - trampoline_32bit_src) +SYM_CODE_END(trampoline_32bit_src) + +/* + * This symbol is placed right after trampoline_32bit_src() so its address can + * be used to infer the size of the trampoline code. + */ +SYM_DATA(trampoline_ljmp_imm_offset, .word .Ljmp + 1 - trampoline_32bit_src) + + /* + * The trampoline code has a size limit. + * Make sure we fail to compile if the trampoline code grows + * beyond TRAMPOLINE_32BIT_CODE_SIZE bytes. + */ + .org trampoline_32bit_src + TRAMPOLINE_32BIT_CODE_SIZE diff --git a/arch/x86/boot/startup/map_kernel.c b/arch/x86/boot/startup/map_kernel.c new file mode 100644 index 000000000000..332dbe6688c4 --- /dev/null +++ b/arch/x86/boot/startup/map_kernel.c @@ -0,0 +1,217 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include <linux/init.h> +#include <linux/linkage.h> +#include <linux/types.h> +#include <linux/kernel.h> +#include <linux/pgtable.h> + +#include <asm/init.h> +#include <asm/sections.h> +#include <asm/setup.h> +#include <asm/sev.h> + +extern pmd_t early_dynamic_pgts[EARLY_DYNAMIC_PAGE_TABLES][PTRS_PER_PMD]; +extern unsigned int next_early_pgt; + +static inline bool check_la57_support(void) +{ + /* + * 5-level paging is detected and enabled at kernel decompression + * stage. Only check if it has been enabled there. + */ + if (!(native_read_cr4() & X86_CR4_LA57)) + return false; + + __pgtable_l5_enabled = 1; + pgdir_shift = 48; + ptrs_per_p4d = 512; + + return true; +} + +static unsigned long __head sme_postprocess_startup(struct boot_params *bp, + pmdval_t *pmd, + unsigned long p2v_offset) +{ + unsigned long paddr, paddr_end; + int i; + + /* Encrypt the kernel and related (if SME is active) */ + sme_encrypt_kernel(bp); + + /* + * Clear the memory encryption mask from the .bss..decrypted section. + * The bss section will be memset to zero later in the initialization so + * there is no need to zero it after changing the memory encryption + * attribute. + */ + if (sme_get_me_mask()) { + paddr = (unsigned long)rip_rel_ptr(__start_bss_decrypted); + paddr_end = (unsigned long)rip_rel_ptr(__end_bss_decrypted); + + for (; paddr < paddr_end; paddr += PMD_SIZE) { + /* + * On SNP, transition the page to shared in the RMP table so that + * it is consistent with the page table attribute change. + * + * __start_bss_decrypted has a virtual address in the high range + * mapping (kernel .text). PVALIDATE, by way of + * early_snp_set_memory_shared(), requires a valid virtual + * address but the kernel is currently running off of the identity + * mapping so use the PA to get a *currently* valid virtual address. + */ + early_snp_set_memory_shared(paddr, paddr, PTRS_PER_PMD); + + i = pmd_index(paddr - p2v_offset); + pmd[i] -= sme_get_me_mask(); + } + } + + /* + * Return the SME encryption mask (if SME is active) to be used as a + * modifier for the initial pgdir entry programmed into CR3. + */ + return sme_get_me_mask(); +} + +/* + * This code is compiled using PIC codegen because it will execute from the + * early 1:1 mapping of memory, which deviates from the mapping expected by the + * linker. Due to this deviation, taking the address of a global variable will + * produce an ambiguous result when using the plain & operator. Instead, + * rip_rel_ptr() must be used, which will return the RIP-relative address in + * the 1:1 mapping of memory. Kernel virtual addresses can be determined by + * subtracting p2v_offset from the RIP-relative address. + */ +unsigned long __head __startup_64(unsigned long p2v_offset, + struct boot_params *bp) +{ + pmd_t (*early_pgts)[PTRS_PER_PMD] = rip_rel_ptr(early_dynamic_pgts); + unsigned long physaddr = (unsigned long)rip_rel_ptr(_text); + unsigned long va_text, va_end; + unsigned long pgtable_flags; + unsigned long load_delta; + pgdval_t *pgd; + p4dval_t *p4d; + pudval_t *pud; + pmdval_t *pmd, pmd_entry; + bool la57; + int i; + + la57 = check_la57_support(); + + /* Is the address too large? */ + if (physaddr >> MAX_PHYSMEM_BITS) + for (;;); + + /* + * Compute the delta between the address I am compiled to run at + * and the address I am actually running at. + */ + phys_base = load_delta = __START_KERNEL_map + p2v_offset; + + /* Is the address not 2M aligned? */ + if (load_delta & ~PMD_MASK) + for (;;); + + va_text = physaddr - p2v_offset; + va_end = (unsigned long)rip_rel_ptr(_end) - p2v_offset; + + /* Include the SME encryption mask in the fixup value */ + load_delta += sme_get_me_mask(); + + /* Fixup the physical addresses in the page table */ + + pgd = rip_rel_ptr(early_top_pgt); + pgd[pgd_index(__START_KERNEL_map)] += load_delta; + + if (la57) { + p4d = (p4dval_t *)rip_rel_ptr(level4_kernel_pgt); + p4d[MAX_PTRS_PER_P4D - 1] += load_delta; + + pgd[pgd_index(__START_KERNEL_map)] = (pgdval_t)p4d | _PAGE_TABLE; + } + + level3_kernel_pgt[PTRS_PER_PUD - 2].pud += load_delta; + level3_kernel_pgt[PTRS_PER_PUD - 1].pud += load_delta; + + for (i = FIXMAP_PMD_TOP; i > FIXMAP_PMD_TOP - FIXMAP_PMD_NUM; i--) + level2_fixmap_pgt[i].pmd += load_delta; + + /* + * Set up the identity mapping for the switchover. These + * entries should *NOT* have the global bit set! This also + * creates a bunch of nonsense entries but that is fine -- + * it avoids problems around wraparound. + */ + + pud = &early_pgts[0]->pmd; + pmd = &early_pgts[1]->pmd; + next_early_pgt = 2; + + pgtable_flags = _KERNPG_TABLE_NOENC + sme_get_me_mask(); + + if (la57) { + p4d = &early_pgts[next_early_pgt++]->pmd; + + i = (physaddr >> PGDIR_SHIFT) % PTRS_PER_PGD; + pgd[i + 0] = (pgdval_t)p4d + pgtable_flags; + pgd[i + 1] = (pgdval_t)p4d + pgtable_flags; + + i = physaddr >> P4D_SHIFT; + p4d[(i + 0) % PTRS_PER_P4D] = (pgdval_t)pud + pgtable_flags; + p4d[(i + 1) % PTRS_PER_P4D] = (pgdval_t)pud + pgtable_flags; + } else { + i = (physaddr >> PGDIR_SHIFT) % PTRS_PER_PGD; + pgd[i + 0] = (pgdval_t)pud + pgtable_flags; + pgd[i + 1] = (pgdval_t)pud + pgtable_flags; + } + + i = physaddr >> PUD_SHIFT; + pud[(i + 0) % PTRS_PER_PUD] = (pudval_t)pmd + pgtable_flags; + pud[(i + 1) % PTRS_PER_PUD] = (pudval_t)pmd + pgtable_flags; + + pmd_entry = __PAGE_KERNEL_LARGE_EXEC & ~_PAGE_GLOBAL; + pmd_entry += sme_get_me_mask(); + pmd_entry += physaddr; + + for (i = 0; i < DIV_ROUND_UP(va_end - va_text, PMD_SIZE); i++) { + int idx = i + (physaddr >> PMD_SHIFT); + + pmd[idx % PTRS_PER_PMD] = pmd_entry + i * PMD_SIZE; + } + + /* + * Fixup the kernel text+data virtual addresses. Note that + * we might write invalid pmds, when the kernel is relocated + * cleanup_highmap() fixes this up along with the mappings + * beyond _end. + * + * Only the region occupied by the kernel image has so far + * been checked against the table of usable memory regions + * provided by the firmware, so invalidate pages outside that + * region. A page table entry that maps to a reserved area of + * memory would allow processor speculation into that area, + * and on some hardware (particularly the UV platform) even + * speculative access to some reserved areas is caught as an + * error, causing the BIOS to halt the system. + */ + + pmd = rip_rel_ptr(level2_kernel_pgt); + + /* invalidate pages before the kernel image */ + for (i = 0; i < pmd_index(va_text); i++) + pmd[i] &= ~_PAGE_PRESENT; + + /* fixup pages that are part of the kernel image */ + for (; i <= pmd_index(va_end); i++) + if (pmd[i] & _PAGE_PRESENT) + pmd[i] += load_delta; + + /* invalidate pages after the kernel image */ + for (; i < PTRS_PER_PMD; i++) + pmd[i] &= ~_PAGE_PRESENT; + + return sme_postprocess_startup(bp, pmd, p2v_offset); +} diff --git a/arch/x86/boot/startup/sev-shared.c b/arch/x86/boot/startup/sev-shared.c new file mode 100644 index 000000000000..7a706db87b93 --- /dev/null +++ b/arch/x86/boot/startup/sev-shared.c @@ -0,0 +1,887 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * AMD Encrypted Register State Support + * + * Author: Joerg Roedel <jroedel@suse.de> + * + * This file is not compiled stand-alone. It contains code shared + * between the pre-decompression boot code and the running Linux kernel + * and is included directly into both code-bases. + */ + +#include <asm/setup_data.h> + +#ifndef __BOOT_COMPRESSED +#define error(v) pr_err(v) +#define has_cpuflag(f) boot_cpu_has(f) +#else +#undef WARN +#define WARN(condition, format...) (!!(condition)) +#undef vc_forward_exception +#define vc_forward_exception(c) panic("SNP: Hypervisor requested exception\n") +#endif + +/* + * SVSM related information: + * During boot, the page tables are set up as identity mapped and later + * changed to use kernel virtual addresses. Maintain separate virtual and + * physical addresses for the CAA to allow SVSM functions to be used during + * early boot, both with identity mapped virtual addresses and proper kernel + * virtual addresses. + */ +struct svsm_ca *boot_svsm_caa __ro_after_init; +u64 boot_svsm_caa_pa __ro_after_init; + +/* + * Since feature negotiation related variables are set early in the boot + * process they must reside in the .data section so as not to be zeroed + * out when the .bss section is later cleared. + * + * GHCB protocol version negotiated with the hypervisor. + */ +static u16 ghcb_version __ro_after_init; + +/* Copy of the SNP firmware's CPUID page. */ +static struct snp_cpuid_table cpuid_table_copy __ro_after_init; + +/* + * These will be initialized based on CPUID table so that non-present + * all-zero leaves (for sparse tables) can be differentiated from + * invalid/out-of-range leaves. This is needed since all-zero leaves + * still need to be post-processed. + */ +static u32 cpuid_std_range_max __ro_after_init; +static u32 cpuid_hyp_range_max __ro_after_init; +static u32 cpuid_ext_range_max __ro_after_init; + +bool __init sev_es_check_cpu_features(void) +{ + if (!has_cpuflag(X86_FEATURE_RDRAND)) { + error("RDRAND instruction not supported - no trusted source of randomness available\n"); + return false; + } + + return true; +} + +void __head __noreturn +sev_es_terminate(unsigned int set, unsigned int reason) +{ + u64 val = GHCB_MSR_TERM_REQ; + + /* Tell the hypervisor what went wrong. */ + val |= GHCB_SEV_TERM_REASON(set, reason); + + /* Request Guest Termination from Hypervisor */ + sev_es_wr_ghcb_msr(val); + VMGEXIT(); + + while (true) + asm volatile("hlt\n" : : : "memory"); +} + +/* + * The hypervisor features are available from GHCB version 2 onward. + */ +u64 get_hv_features(void) +{ + u64 val; + + if (ghcb_version < 2) + return 0; + + sev_es_wr_ghcb_msr(GHCB_MSR_HV_FT_REQ); + VMGEXIT(); + + val = sev_es_rd_ghcb_msr(); + if (GHCB_RESP_CODE(val) != GHCB_MSR_HV_FT_RESP) + return 0; + + return GHCB_MSR_HV_FT_RESP_VAL(val); +} + +void snp_register_ghcb_early(unsigned long paddr) +{ + unsigned long pfn = paddr >> PAGE_SHIFT; + u64 val; + + sev_es_wr_ghcb_msr(GHCB_MSR_REG_GPA_REQ_VAL(pfn)); + VMGEXIT(); + + val = sev_es_rd_ghcb_msr(); + + /* If the response GPA is not ours then abort the guest */ + if ((GHCB_RESP_CODE(val) != GHCB_MSR_REG_GPA_RESP) || + (GHCB_MSR_REG_GPA_RESP_VAL(val) != pfn)) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_REGISTER); +} + +bool sev_es_negotiate_protocol(void) +{ + u64 val; + + /* Do the GHCB protocol version negotiation */ + sev_es_wr_ghcb_msr(GHCB_MSR_SEV_INFO_REQ); + VMGEXIT(); + val = sev_es_rd_ghcb_msr(); + + if (GHCB_MSR_INFO(val) != GHCB_MSR_SEV_INFO_RESP) + return false; + + if (GHCB_MSR_PROTO_MAX(val) < GHCB_PROTOCOL_MIN || + GHCB_MSR_PROTO_MIN(val) > GHCB_PROTOCOL_MAX) + return false; + + ghcb_version = min_t(size_t, GHCB_MSR_PROTO_MAX(val), GHCB_PROTOCOL_MAX); + + return true; +} + +static enum es_result verify_exception_info(struct ghcb *ghcb, struct es_em_ctxt *ctxt) +{ + u32 ret; + + ret = ghcb->save.sw_exit_info_1 & GENMASK_ULL(31, 0); + if (!ret) + return ES_OK; + + if (ret == 1) { + u64 info = ghcb->save.sw_exit_info_2; + unsigned long v = info & SVM_EVTINJ_VEC_MASK; + + /* Check if exception information from hypervisor is sane. */ + if ((info & SVM_EVTINJ_VALID) && + ((v == X86_TRAP_GP) || (v == X86_TRAP_UD)) && + ((info & SVM_EVTINJ_TYPE_MASK) == SVM_EVTINJ_TYPE_EXEPT)) { + ctxt->fi.vector = v; + + if (info & SVM_EVTINJ_VALID_ERR) + ctxt->fi.error_code = info >> 32; + + return ES_EXCEPTION; + } + } + + return ES_VMM_ERROR; +} + +static inline int svsm_process_result_codes(struct svsm_call *call) +{ + switch (call->rax_out) { + case SVSM_SUCCESS: + return 0; + case SVSM_ERR_INCOMPLETE: + case SVSM_ERR_BUSY: + return -EAGAIN; + default: + return -EINVAL; + } +} + +/* + * Issue a VMGEXIT to call the SVSM: + * - Load the SVSM register state (RAX, RCX, RDX, R8 and R9) + * - Set the CA call pending field to 1 + * - Issue VMGEXIT + * - Save the SVSM return register state (RAX, RCX, RDX, R8 and R9) + * - Perform atomic exchange of the CA call pending field + * + * - See the "Secure VM Service Module for SEV-SNP Guests" specification for + * details on the calling convention. + * - The calling convention loosely follows the Microsoft X64 calling + * convention by putting arguments in RCX, RDX, R8 and R9. + * - RAX specifies the SVSM protocol/callid as input and the return code + * as output. + */ +static __always_inline void svsm_issue_call(struct svsm_call *call, u8 *pending) +{ + register unsigned long rax asm("rax") = call->rax; + register unsigned long rcx asm("rcx") = call->rcx; + register unsigned long rdx asm("rdx") = call->rdx; + register unsigned long r8 asm("r8") = call->r8; + register unsigned long r9 asm("r9") = call->r9; + + call->caa->call_pending = 1; + + asm volatile("rep; vmmcall\n\t" + : "+r" (rax), "+r" (rcx), "+r" (rdx), "+r" (r8), "+r" (r9) + : : "memory"); + + *pending = xchg(&call->caa->call_pending, *pending); + + call->rax_out = rax; + call->rcx_out = rcx; + call->rdx_out = rdx; + call->r8_out = r8; + call->r9_out = r9; +} + +static int svsm_perform_msr_protocol(struct svsm_call *call) +{ + u8 pending = 0; + u64 val, resp; + + /* + * When using the MSR protocol, be sure to save and restore + * the current MSR value. + */ + val = sev_es_rd_ghcb_msr(); + + sev_es_wr_ghcb_msr(GHCB_MSR_VMPL_REQ_LEVEL(0)); + + svsm_issue_call(call, &pending); + + resp = sev_es_rd_ghcb_msr(); + + sev_es_wr_ghcb_msr(val); + + if (pending) + return -EINVAL; + + if (GHCB_RESP_CODE(resp) != GHCB_MSR_VMPL_RESP) + return -EINVAL; + + if (GHCB_MSR_VMPL_RESP_VAL(resp)) + return -EINVAL; + + return svsm_process_result_codes(call); +} + +static int svsm_perform_ghcb_protocol(struct ghcb *ghcb, struct svsm_call *call) +{ + struct es_em_ctxt ctxt; + u8 pending = 0; + + vc_ghcb_invalidate(ghcb); + + /* + * Fill in protocol and format specifiers. This can be called very early + * in the boot, so use rip-relative references as needed. + */ + ghcb->protocol_version = ghcb_version; + ghcb->ghcb_usage = GHCB_DEFAULT_USAGE; + + ghcb_set_sw_exit_code(ghcb, SVM_VMGEXIT_SNP_RUN_VMPL); + ghcb_set_sw_exit_info_1(ghcb, 0); + ghcb_set_sw_exit_info_2(ghcb, 0); + + sev_es_wr_ghcb_msr(__pa(ghcb)); + + svsm_issue_call(call, &pending); + + if (pending) + return -EINVAL; + + switch (verify_exception_info(ghcb, &ctxt)) { + case ES_OK: + break; + case ES_EXCEPTION: + vc_forward_exception(&ctxt); + fallthrough; + default: + return -EINVAL; + } + + return svsm_process_result_codes(call); +} + +enum es_result sev_es_ghcb_hv_call(struct ghcb *ghcb, + struct es_em_ctxt *ctxt, + u64 exit_code, u64 exit_info_1, + u64 exit_info_2) +{ + /* Fill in protocol and format specifiers */ + ghcb->protocol_version = ghcb_version; + ghcb->ghcb_usage = GHCB_DEFAULT_USAGE; + + ghcb_set_sw_exit_code(ghcb, exit_code); + ghcb_set_sw_exit_info_1(ghcb, exit_info_1); + ghcb_set_sw_exit_info_2(ghcb, exit_info_2); + + sev_es_wr_ghcb_msr(__pa(ghcb)); + VMGEXIT(); + + return verify_exception_info(ghcb, ctxt); +} + +static int __sev_cpuid_hv(u32 fn, int reg_idx, u32 *reg) +{ + u64 val; + + sev_es_wr_ghcb_msr(GHCB_CPUID_REQ(fn, reg_idx)); + VMGEXIT(); + val = sev_es_rd_ghcb_msr(); + if (GHCB_RESP_CODE(val) != GHCB_MSR_CPUID_RESP) + return -EIO; + + *reg = (val >> 32); + + return 0; +} + +static int __sev_cpuid_hv_msr(struct cpuid_leaf *leaf) +{ + int ret; + + /* + * MSR protocol does not support fetching non-zero subfunctions, but is + * sufficient to handle current early-boot cases. Should that change, + * make sure to report an error rather than ignoring the index and + * grabbing random values. If this issue arises in the future, handling + * can be added here to use GHCB-page protocol for cases that occur late + * enough in boot that GHCB page is available. + */ + if (cpuid_function_is_indexed(leaf->fn) && leaf->subfn) + return -EINVAL; + + ret = __sev_cpuid_hv(leaf->fn, GHCB_CPUID_REQ_EAX, &leaf->eax); + ret = ret ? : __sev_cpuid_hv(leaf->fn, GHCB_CPUID_REQ_EBX, &leaf->ebx); + ret = ret ? : __sev_cpuid_hv(leaf->fn, GHCB_CPUID_REQ_ECX, &leaf->ecx); + ret = ret ? : __sev_cpuid_hv(leaf->fn, GHCB_CPUID_REQ_EDX, &leaf->edx); + + return ret; +} + +static int __sev_cpuid_hv_ghcb(struct ghcb *ghcb, struct es_em_ctxt *ctxt, struct cpuid_leaf *leaf) +{ + u32 cr4 = native_read_cr4(); + int ret; + + ghcb_set_rax(ghcb, leaf->fn); + ghcb_set_rcx(ghcb, leaf->subfn); + + if (cr4 & X86_CR4_OSXSAVE) + /* Safe to read xcr0 */ + ghcb_set_xcr0(ghcb, xgetbv(XCR_XFEATURE_ENABLED_MASK)); + else + /* xgetbv will cause #UD - use reset value for xcr0 */ + ghcb_set_xcr0(ghcb, 1); + + ret = sev_es_ghcb_hv_call(ghcb, ctxt, SVM_EXIT_CPUID, 0, 0); + if (ret != ES_OK) + return ret; + + if (!(ghcb_rax_is_valid(ghcb) && + ghcb_rbx_is_valid(ghcb) && + ghcb_rcx_is_valid(ghcb) && + ghcb_rdx_is_valid(ghcb))) + return ES_VMM_ERROR; + + leaf->eax = ghcb->save.rax; + leaf->ebx = ghcb->save.rbx; + leaf->ecx = ghcb->save.rcx; + leaf->edx = ghcb->save.rdx; + + return ES_OK; +} + +static int sev_cpuid_hv(struct ghcb *ghcb, struct es_em_ctxt *ctxt, struct cpuid_leaf *leaf) +{ + return ghcb ? __sev_cpuid_hv_ghcb(ghcb, ctxt, leaf) + : __sev_cpuid_hv_msr(leaf); +} + +/* + * This may be called early while still running on the initial identity + * mapping. Use RIP-relative addressing to obtain the correct address + * while running with the initial identity mapping as well as the + * switch-over to kernel virtual addresses later. + */ +const struct snp_cpuid_table *snp_cpuid_get_table(void) +{ + return rip_rel_ptr(&cpuid_table_copy); +} + +/* + * The SNP Firmware ABI, Revision 0.9, Section 7.1, details the use of + * XCR0_IN and XSS_IN to encode multiple versions of 0xD subfunctions 0 + * and 1 based on the corresponding features enabled by a particular + * combination of XCR0 and XSS registers so that a guest can look up the + * version corresponding to the features currently enabled in its XCR0/XSS + * registers. The only values that differ between these versions/table + * entries is the enabled XSAVE area size advertised via EBX. + * + * While hypervisors may choose to make use of this support, it is more + * robust/secure for a guest to simply find the entry corresponding to the + * base/legacy XSAVE area size (XCR0=1 or XCR0=3), and then calculate the + * XSAVE area size using subfunctions 2 through 64, as documented in APM + * Volume 3, Rev 3.31, Appendix E.3.8, which is what is done here. + * + * Since base/legacy XSAVE area size is documented as 0x240, use that value + * directly rather than relying on the base size in the CPUID table. + * + * Return: XSAVE area size on success, 0 otherwise. + */ +static u32 __head snp_cpuid_calc_xsave_size(u64 xfeatures_en, bool compacted) +{ + const struct snp_cpuid_table *cpuid_table = snp_cpuid_get_table(); + u64 xfeatures_found = 0; + u32 xsave_size = 0x240; + int i; + + for (i = 0; i < cpuid_table->count; i++) { + const struct snp_cpuid_fn *e = &cpuid_table->fn[i]; + + if (!(e->eax_in == 0xD && e->ecx_in > 1 && e->ecx_in < 64)) + continue; + if (!(xfeatures_en & (BIT_ULL(e->ecx_in)))) + continue; + if (xfeatures_found & (BIT_ULL(e->ecx_in))) + continue; + + xfeatures_found |= (BIT_ULL(e->ecx_in)); + + if (compacted) + xsave_size += e->eax; + else + xsave_size = max(xsave_size, e->eax + e->ebx); + } + + /* + * Either the guest set unsupported XCR0/XSS bits, or the corresponding + * entries in the CPUID table were not present. This is not a valid + * state to be in. + */ + if (xfeatures_found != (xfeatures_en & GENMASK_ULL(63, 2))) + return 0; + + return xsave_size; +} + +static bool __head +snp_cpuid_get_validated_func(struct cpuid_leaf *leaf) +{ + const struct snp_cpuid_table *cpuid_table = snp_cpuid_get_table(); + int i; + + for (i = 0; i < cpuid_table->count; i++) { + const struct snp_cpuid_fn *e = &cpuid_table->fn[i]; + + if (e->eax_in != leaf->fn) + continue; + + if (cpuid_function_is_indexed(leaf->fn) && e->ecx_in != leaf->subfn) + continue; + + /* + * For 0xD subfunctions 0 and 1, only use the entry corresponding + * to the base/legacy XSAVE area size (XCR0=1 or XCR0=3, XSS=0). + * See the comments above snp_cpuid_calc_xsave_size() for more + * details. + */ + if (e->eax_in == 0xD && (e->ecx_in == 0 || e->ecx_in == 1)) + if (!(e->xcr0_in == 1 || e->xcr0_in == 3) || e->xss_in) + continue; + + leaf->eax = e->eax; + leaf->ebx = e->ebx; + leaf->ecx = e->ecx; + leaf->edx = e->edx; + + return true; + } + + return false; +} + +static void snp_cpuid_hv(struct ghcb *ghcb, struct es_em_ctxt *ctxt, struct cpuid_leaf *leaf) +{ + if (sev_cpuid_hv(ghcb, ctxt, leaf)) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_CPUID_HV); +} + +static int __head +snp_cpuid_postprocess(struct ghcb *ghcb, struct es_em_ctxt *ctxt, + struct cpuid_leaf *leaf) +{ + struct cpuid_leaf leaf_hv = *leaf; + + switch (leaf->fn) { + case 0x1: + snp_cpuid_hv(ghcb, ctxt, &leaf_hv); + + /* initial APIC ID */ + leaf->ebx = (leaf_hv.ebx & GENMASK(31, 24)) | (leaf->ebx & GENMASK(23, 0)); + /* APIC enabled bit */ + leaf->edx = (leaf_hv.edx & BIT(9)) | (leaf->edx & ~BIT(9)); + + /* OSXSAVE enabled bit */ + if (native_read_cr4() & X86_CR4_OSXSAVE) + leaf->ecx |= BIT(27); + break; + case 0x7: + /* OSPKE enabled bit */ + leaf->ecx &= ~BIT(4); + if (native_read_cr4() & X86_CR4_PKE) + leaf->ecx |= BIT(4); + break; + case 0xB: + leaf_hv.subfn = 0; + snp_cpuid_hv(ghcb, ctxt, &leaf_hv); + + /* extended APIC ID */ + leaf->edx = leaf_hv.edx; + break; + case 0xD: { + bool compacted = false; + u64 xcr0 = 1, xss = 0; + u32 xsave_size; + + if (leaf->subfn != 0 && leaf->subfn != 1) + return 0; + + if (native_read_cr4() & X86_CR4_OSXSAVE) + xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK); + if (leaf->subfn == 1) { + /* Get XSS value if XSAVES is enabled. */ + if (leaf->eax & BIT(3)) { + unsigned long lo, hi; + + asm volatile("rdmsr" : "=a" (lo), "=d" (hi) + : "c" (MSR_IA32_XSS)); + xss = (hi << 32) | lo; + } + + /* + * The PPR and APM aren't clear on what size should be + * encoded in 0xD:0x1:EBX when compaction is not enabled + * by either XSAVEC (feature bit 1) or XSAVES (feature + * bit 3) since SNP-capable hardware has these feature + * bits fixed as 1. KVM sets it to 0 in this case, but + * to avoid this becoming an issue it's safer to simply + * treat this as unsupported for SNP guests. + */ + if (!(leaf->eax & (BIT(1) | BIT(3)))) + return -EINVAL; + + compacted = true; + } + + xsave_size = snp_cpuid_calc_xsave_size(xcr0 | xss, compacted); + if (!xsave_size) + return -EINVAL; + + leaf->ebx = xsave_size; + } + break; + case 0x8000001E: + snp_cpuid_hv(ghcb, ctxt, &leaf_hv); + + /* extended APIC ID */ + leaf->eax = leaf_hv.eax; + /* compute ID */ + leaf->ebx = (leaf->ebx & GENMASK(31, 8)) | (leaf_hv.ebx & GENMASK(7, 0)); + /* node ID */ + leaf->ecx = (leaf->ecx & GENMASK(31, 8)) | (leaf_hv.ecx & GENMASK(7, 0)); + break; + default: + /* No fix-ups needed, use values as-is. */ + break; + } + + return 0; +} + +/* + * Returns -EOPNOTSUPP if feature not enabled. Any other non-zero return value + * should be treated as fatal by caller. + */ +int __head +snp_cpuid(struct ghcb *ghcb, struct es_em_ctxt *ctxt, struct cpuid_leaf *leaf) +{ + const struct snp_cpuid_table *cpuid_table = snp_cpuid_get_table(); + + if (!cpuid_table->count) + return -EOPNOTSUPP; + + if (!snp_cpuid_get_validated_func(leaf)) { + /* + * Some hypervisors will avoid keeping track of CPUID entries + * where all values are zero, since they can be handled the + * same as out-of-range values (all-zero). This is useful here + * as well as it allows virtually all guest configurations to + * work using a single SNP CPUID table. + * + * To allow for this, there is a need to distinguish between + * out-of-range entries and in-range zero entries, since the + * CPUID table entries are only a template that may need to be + * augmented with additional values for things like + * CPU-specific information during post-processing. So if it's + * not in the table, set the values to zero. Then, if they are + * within a valid CPUID range, proceed with post-processing + * using zeros as the initial values. Otherwise, skip + * post-processing and just return zeros immediately. + */ + leaf->eax = leaf->ebx = leaf->ecx = leaf->edx = 0; + + /* Skip post-processing for out-of-range zero leafs. */ + if (!(leaf->fn <= cpuid_std_range_max || + (leaf->fn >= 0x40000000 && leaf->fn <= cpuid_hyp_range_max) || + (leaf->fn >= 0x80000000 && leaf->fn <= cpuid_ext_range_max))) + return 0; + } + + return snp_cpuid_postprocess(ghcb, ctxt, leaf); +} + +/* + * Boot VC Handler - This is the first VC handler during boot, there is no GHCB + * page yet, so it only supports the MSR based communication with the + * hypervisor and only the CPUID exit-code. + */ +void __head do_vc_no_ghcb(struct pt_regs *regs, unsigned long exit_code) +{ + unsigned int subfn = lower_bits(regs->cx, 32); + unsigned int fn = lower_bits(regs->ax, 32); + u16 opcode = *(unsigned short *)regs->ip; + struct cpuid_leaf leaf; + int ret; + + /* Only CPUID is supported via MSR protocol */ + if (exit_code != SVM_EXIT_CPUID) + goto fail; + + /* Is it really a CPUID insn? */ + if (opcode != 0xa20f) + goto fail; + + leaf.fn = fn; + leaf.subfn = subfn; + + ret = snp_cpuid(NULL, NULL, &leaf); + if (!ret) + goto cpuid_done; + + if (ret != -EOPNOTSUPP) + goto fail; + + if (__sev_cpuid_hv_msr(&leaf)) + goto fail; + +cpuid_done: + regs->ax = leaf.eax; + regs->bx = leaf.ebx; + regs->cx = leaf.ecx; + regs->dx = leaf.edx; + + /* + * This is a VC handler and the #VC is only raised when SEV-ES is + * active, which means SEV must be active too. Do sanity checks on the + * CPUID results to make sure the hypervisor does not trick the kernel + * into the no-sev path. This could map sensitive data unencrypted and + * make it accessible to the hypervisor. + * + * In particular, check for: + * - Availability of CPUID leaf 0x8000001f + * - SEV CPUID bit. + * + * The hypervisor might still report the wrong C-bit position, but this + * can't be checked here. + */ + + if (fn == 0x80000000 && (regs->ax < 0x8000001f)) + /* SEV leaf check */ + goto fail; + else if ((fn == 0x8000001f && !(regs->ax & BIT(1)))) + /* SEV bit */ + goto fail; + + /* Skip over the CPUID two-byte opcode */ + regs->ip += 2; + + return; + +fail: + /* Terminate the guest */ + sev_es_terminate(SEV_TERM_SET_GEN, GHCB_SEV_ES_GEN_REQ); +} + +struct cc_setup_data { + struct setup_data header; + u32 cc_blob_address; +}; + +/* + * Search for a Confidential Computing blob passed in as a setup_data entry + * via the Linux Boot Protocol. + */ +static __head +struct cc_blob_sev_info *find_cc_blob_setup_data(struct boot_params *bp) +{ + struct cc_setup_data *sd = NULL; + struct setup_data *hdr; + + hdr = (struct setup_data *)bp->hdr.setup_data; + + while (hdr) { + if (hdr->type == SETUP_CC_BLOB) { + sd = (struct cc_setup_data *)hdr; + return (struct cc_blob_sev_info *)(unsigned long)sd->cc_blob_address; + } + hdr = (struct setup_data *)hdr->next; + } + + return NULL; +} + +/* + * Initialize the kernel's copy of the SNP CPUID table, and set up the + * pointer that will be used to access it. + * + * Maintaining a direct mapping of the SNP CPUID table used by firmware would + * be possible as an alternative, but the approach is brittle since the + * mapping needs to be updated in sync with all the changes to virtual memory + * layout and related mapping facilities throughout the boot process. + */ +static void __head setup_cpuid_table(const struct cc_blob_sev_info *cc_info) +{ + const struct snp_cpuid_table *cpuid_table_fw, *cpuid_table; + int i; + + if (!cc_info || !cc_info->cpuid_phys || cc_info->cpuid_len < PAGE_SIZE) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_CPUID); + + cpuid_table_fw = (const struct snp_cpuid_table *)cc_info->cpuid_phys; + if (!cpuid_table_fw->count || cpuid_table_fw->count > SNP_CPUID_COUNT_MAX) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_CPUID); + + cpuid_table = snp_cpuid_get_table(); + memcpy((void *)cpuid_table, cpuid_table_fw, sizeof(*cpuid_table)); + + /* Initialize CPUID ranges for range-checking. */ + for (i = 0; i < cpuid_table->count; i++) { + const struct snp_cpuid_fn *fn = &cpuid_table->fn[i]; + + if (fn->eax_in == 0x0) + cpuid_std_range_max = fn->eax; + else if (fn->eax_in == 0x40000000) + cpuid_hyp_range_max = fn->eax; + else if (fn->eax_in == 0x80000000) + cpuid_ext_range_max = fn->eax; + } +} + +static void __head svsm_pval_4k_page(unsigned long paddr, bool validate) +{ + struct svsm_pvalidate_call *pc; + struct svsm_call call = {}; + unsigned long flags; + u64 pc_pa; + int ret; + + /* + * This can be called very early in the boot, use native functions in + * order to avoid paravirt issues. + */ + flags = native_local_irq_save(); + + call.caa = svsm_get_caa(); + + pc = (struct svsm_pvalidate_call *)call.caa->svsm_buffer; + pc_pa = svsm_get_caa_pa() + offsetof(struct svsm_ca, svsm_buffer); + + pc->num_entries = 1; + pc->cur_index = 0; + pc->entry[0].page_size = RMP_PG_SIZE_4K; + pc->entry[0].action = validate; + pc->entry[0].ignore_cf = 0; + pc->entry[0].pfn = paddr >> PAGE_SHIFT; + + /* Protocol 0, Call ID 1 */ + call.rax = SVSM_CORE_CALL(SVSM_CORE_PVALIDATE); + call.rcx = pc_pa; + + ret = svsm_perform_call_protocol(&call); + if (ret) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_PVALIDATE); + + native_local_irq_restore(flags); +} + +static void __head pvalidate_4k_page(unsigned long vaddr, unsigned long paddr, + bool validate) +{ + int ret; + + if (snp_vmpl) { + svsm_pval_4k_page(paddr, validate); + } else { + ret = pvalidate(vaddr, RMP_PG_SIZE_4K, validate); + if (ret) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_PVALIDATE); + } +} + +/* + * Maintain the GPA of the SVSM Calling Area (CA) in order to utilize the SVSM + * services needed when not running in VMPL0. + */ +static bool __head svsm_setup_ca(const struct cc_blob_sev_info *cc_info) +{ + struct snp_secrets_page *secrets_page; + struct snp_cpuid_table *cpuid_table; + unsigned int i; + u64 caa; + + BUILD_BUG_ON(sizeof(*secrets_page) != PAGE_SIZE); + + /* + * Check if running at VMPL0. + * + * Use RMPADJUST (see the rmpadjust() function for a description of what + * the instruction does) to update the VMPL1 permissions of a page. If + * the guest is running at VMPL0, this will succeed and implies there is + * no SVSM. If the guest is running at any other VMPL, this will fail. + * Linux SNP guests only ever run at a single VMPL level so permission mask + * changes of a lesser-privileged VMPL are a don't-care. + * + * Use a rip-relative reference to obtain the proper address, since this + * routine is running identity mapped when called, both by the decompressor + * code and the early kernel code. + */ + if (!rmpadjust((unsigned long)rip_rel_ptr(&boot_ghcb_page), RMP_PG_SIZE_4K, 1)) + return false; + + /* + * Not running at VMPL0, ensure everything has been properly supplied + * for running under an SVSM. + */ + if (!cc_info || !cc_info->secrets_phys || cc_info->secrets_len != PAGE_SIZE) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_SECRETS_PAGE); + + secrets_page = (struct snp_secrets_page *)cc_info->secrets_phys; + if (!secrets_page->svsm_size) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_NO_SVSM); + + if (!secrets_page->svsm_guest_vmpl) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_SVSM_VMPL0); + + snp_vmpl = secrets_page->svsm_guest_vmpl; + + caa = secrets_page->svsm_caa; + + /* + * An open-coded PAGE_ALIGNED() in order to avoid including + * kernel-proper headers into the decompressor. + */ + if (caa & (PAGE_SIZE - 1)) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_SVSM_CAA); + + /* + * The CA is identity mapped when this routine is called, both by the + * decompressor code and the early kernel code. + */ + boot_svsm_caa = (struct svsm_ca *)caa; + boot_svsm_caa_pa = caa; + + /* Advertise the SVSM presence via CPUID. */ + cpuid_table = (struct snp_cpuid_table *)snp_cpuid_get_table(); + for (i = 0; i < cpuid_table->count; i++) { + struct snp_cpuid_fn *fn = &cpuid_table->fn[i]; + + if (fn->eax_in == 0x8000001f) + fn->eax |= BIT(28); + } + + return true; +} diff --git a/arch/x86/boot/startup/sev-startup.c b/arch/x86/boot/startup/sev-startup.c new file mode 100644 index 000000000000..0b7e3b950183 --- /dev/null +++ b/arch/x86/boot/startup/sev-startup.c @@ -0,0 +1,368 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * AMD Memory Encryption Support + * + * Copyright (C) 2019 SUSE + * + * Author: Joerg Roedel <jroedel@suse.de> + */ + +#define pr_fmt(fmt) "SEV: " fmt + +#include <linux/percpu-defs.h> +#include <linux/cc_platform.h> +#include <linux/printk.h> +#include <linux/mm_types.h> +#include <linux/set_memory.h> +#include <linux/memblock.h> +#include <linux/kernel.h> +#include <linux/mm.h> +#include <linux/cpumask.h> +#include <linux/efi.h> +#include <linux/io.h> +#include <linux/psp-sev.h> +#include <uapi/linux/sev-guest.h> + +#include <asm/init.h> +#include <asm/cpu_entry_area.h> +#include <asm/stacktrace.h> +#include <asm/sev.h> +#include <asm/sev-internal.h> +#include <asm/insn-eval.h> +#include <asm/fpu/xcr.h> +#include <asm/processor.h> +#include <asm/realmode.h> +#include <asm/setup.h> +#include <asm/traps.h> +#include <asm/svm.h> +#include <asm/smp.h> +#include <asm/cpu.h> +#include <asm/apic.h> +#include <asm/cpuid/api.h> +#include <asm/cmdline.h> + +/* For early boot hypervisor communication in SEV-ES enabled guests */ +struct ghcb boot_ghcb_page __bss_decrypted __aligned(PAGE_SIZE); + +/* + * Needs to be in the .data section because we need it NULL before bss is + * cleared + */ +struct ghcb *boot_ghcb __section(".data"); + +/* Bitmap of SEV features supported by the hypervisor */ +u64 sev_hv_features __ro_after_init; + +/* Secrets page physical address from the CC blob */ +u64 sev_secrets_pa __ro_after_init; + +/* For early boot SVSM communication */ +struct svsm_ca boot_svsm_ca_page __aligned(PAGE_SIZE); + +DEFINE_PER_CPU(struct svsm_ca *, svsm_caa); +DEFINE_PER_CPU(u64, svsm_caa_pa); + +/* + * Nothing shall interrupt this code path while holding the per-CPU + * GHCB. The backup GHCB is only for NMIs interrupting this path. + * + * Callers must disable local interrupts around it. + */ +noinstr struct ghcb *__sev_get_ghcb(struct ghcb_state *state) +{ + struct sev_es_runtime_data *data; + struct ghcb *ghcb; + + WARN_ON(!irqs_disabled()); + + data = this_cpu_read(runtime_data); + ghcb = &data->ghcb_page; + + if (unlikely(data->ghcb_active)) { + /* GHCB is already in use - save its contents */ + + if (unlikely(data->backup_ghcb_active)) { + /* + * Backup-GHCB is also already in use. There is no way + * to continue here so just kill the machine. To make + * panic() work, mark GHCBs inactive so that messages + * can be printed out. + */ + data->ghcb_active = false; + data->backup_ghcb_active = false; + + instrumentation_begin(); + panic("Unable to handle #VC exception! GHCB and Backup GHCB are already in use"); + instrumentation_end(); + } + + /* Mark backup_ghcb active before writing to it */ + data->backup_ghcb_active = true; + + state->ghcb = &data->backup_ghcb; + + /* Backup GHCB content */ + *state->ghcb = *ghcb; + } else { + state->ghcb = NULL; + data->ghcb_active = true; + } + + return ghcb; +} + +/* Include code shared with pre-decompression boot stage */ +#include "sev-shared.c" + +noinstr void __sev_put_ghcb(struct ghcb_state *state) +{ + struct sev_es_runtime_data *data; + struct ghcb *ghcb; + + WARN_ON(!irqs_disabled()); + + data = this_cpu_read(runtime_data); + ghcb = &data->ghcb_page; + + if (state->ghcb) { + /* Restore GHCB from Backup */ + *ghcb = *state->ghcb; + data->backup_ghcb_active = false; + state->ghcb = NULL; + } else { + /* + * Invalidate the GHCB so a VMGEXIT instruction issued + * from userspace won't appear to be valid. + */ + vc_ghcb_invalidate(ghcb); + data->ghcb_active = false; + } +} + +int svsm_perform_call_protocol(struct svsm_call *call) +{ + struct ghcb_state state; + unsigned long flags; + struct ghcb *ghcb; + int ret; + + /* + * This can be called very early in the boot, use native functions in + * order to avoid paravirt issues. + */ + flags = native_local_irq_save(); + + if (sev_cfg.ghcbs_initialized) + ghcb = __sev_get_ghcb(&state); + else if (boot_ghcb) + ghcb = boot_ghcb; + else + ghcb = NULL; + + do { + ret = ghcb ? svsm_perform_ghcb_protocol(ghcb, call) + : svsm_perform_msr_protocol(call); + } while (ret == -EAGAIN); + + if (sev_cfg.ghcbs_initialized) + __sev_put_ghcb(&state); + + native_local_irq_restore(flags); + + return ret; +} + +void __head +early_set_pages_state(unsigned long vaddr, unsigned long paddr, + unsigned long npages, enum psc_op op) +{ + unsigned long paddr_end; + u64 val; + + vaddr = vaddr & PAGE_MASK; + + paddr = paddr & PAGE_MASK; + paddr_end = paddr + (npages << PAGE_SHIFT); + + while (paddr < paddr_end) { + /* Page validation must be rescinded before changing to shared */ + if (op == SNP_PAGE_STATE_SHARED) + pvalidate_4k_page(vaddr, paddr, false); + + /* + * Use the MSR protocol because this function can be called before + * the GHCB is established. + */ + sev_es_wr_ghcb_msr(GHCB_MSR_PSC_REQ_GFN(paddr >> PAGE_SHIFT, op)); + VMGEXIT(); + + val = sev_es_rd_ghcb_msr(); + + if (GHCB_RESP_CODE(val) != GHCB_MSR_PSC_RESP) + goto e_term; + + if (GHCB_MSR_PSC_RESP_VAL(val)) + goto e_term; + + /* Page validation must be performed after changing to private */ + if (op == SNP_PAGE_STATE_PRIVATE) + pvalidate_4k_page(vaddr, paddr, true); + + vaddr += PAGE_SIZE; + paddr += PAGE_SIZE; + } + + return; + +e_term: + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_PSC); +} + +void __head early_snp_set_memory_private(unsigned long vaddr, unsigned long paddr, + unsigned long npages) +{ + /* + * This can be invoked in early boot while running identity mapped, so + * use an open coded check for SNP instead of using cc_platform_has(). + * This eliminates worries about jump tables or checking boot_cpu_data + * in the cc_platform_has() function. + */ + if (!(sev_status & MSR_AMD64_SEV_SNP_ENABLED)) + return; + + /* + * Ask the hypervisor to mark the memory pages as private in the RMP + * table. + */ + early_set_pages_state(vaddr, paddr, npages, SNP_PAGE_STATE_PRIVATE); +} + +void __head early_snp_set_memory_shared(unsigned long vaddr, unsigned long paddr, + unsigned long npages) +{ + /* + * This can be invoked in early boot while running identity mapped, so + * use an open coded check for SNP instead of using cc_platform_has(). + * This eliminates worries about jump tables or checking boot_cpu_data + * in the cc_platform_has() function. + */ + if (!(sev_status & MSR_AMD64_SEV_SNP_ENABLED)) + return; + + /* Ask hypervisor to mark the memory pages shared in the RMP table. */ + early_set_pages_state(vaddr, paddr, npages, SNP_PAGE_STATE_SHARED); +} + +/* + * Initial set up of SNP relies on information provided by the + * Confidential Computing blob, which can be passed to the kernel + * in the following ways, depending on how it is booted: + * + * - when booted via the boot/decompress kernel: + * - via boot_params + * + * - when booted directly by firmware/bootloader (e.g. CONFIG_PVH): + * - via a setup_data entry, as defined by the Linux Boot Protocol + * + * Scan for the blob in that order. + */ +static __head struct cc_blob_sev_info *find_cc_blob(struct boot_params *bp) +{ + struct cc_blob_sev_info *cc_info; + + /* Boot kernel would have passed the CC blob via boot_params. */ + if (bp->cc_blob_address) { + cc_info = (struct cc_blob_sev_info *)(unsigned long)bp->cc_blob_address; + goto found_cc_info; + } + + /* + * If kernel was booted directly, without the use of the + * boot/decompression kernel, the CC blob may have been passed via + * setup_data instead. + */ + cc_info = find_cc_blob_setup_data(bp); + if (!cc_info) + return NULL; + +found_cc_info: + if (cc_info->magic != CC_BLOB_SEV_HDR_MAGIC) + snp_abort(); + + return cc_info; +} + +static __head void svsm_setup(struct cc_blob_sev_info *cc_info) +{ + struct svsm_call call = {}; + int ret; + u64 pa; + + /* + * Record the SVSM Calling Area address (CAA) if the guest is not + * running at VMPL0. The CA will be used to communicate with the + * SVSM to perform the SVSM services. + */ + if (!svsm_setup_ca(cc_info)) + return; + + /* + * It is very early in the boot and the kernel is running identity + * mapped but without having adjusted the pagetables to where the + * kernel was loaded (physbase), so the get the CA address using + * RIP-relative addressing. + */ + pa = (u64)rip_rel_ptr(&boot_svsm_ca_page); + + /* + * Switch over to the boot SVSM CA while the current CA is still + * addressable. There is no GHCB at this point so use the MSR protocol. + * + * SVSM_CORE_REMAP_CA call: + * RAX = 0 (Protocol=0, CallID=0) + * RCX = New CA GPA + */ + call.caa = svsm_get_caa(); + call.rax = SVSM_CORE_CALL(SVSM_CORE_REMAP_CA); + call.rcx = pa; + ret = svsm_perform_call_protocol(&call); + if (ret) + sev_es_terminate(SEV_TERM_SET_LINUX, GHCB_TERM_SVSM_CA_REMAP_FAIL); + + boot_svsm_caa = (struct svsm_ca *)pa; + boot_svsm_caa_pa = pa; +} + +bool __head snp_init(struct boot_params *bp) +{ + struct cc_blob_sev_info *cc_info; + + if (!bp) + return false; + + cc_info = find_cc_blob(bp); + if (!cc_info) + return false; + + if (cc_info->secrets_phys && cc_info->secrets_len == PAGE_SIZE) + sev_secrets_pa = cc_info->secrets_phys; + else + return false; + + setup_cpuid_table(cc_info); + + svsm_setup(cc_info); + + /* + * The CC blob will be used later to access the secrets page. Cache + * it here like the boot kernel does. + */ + bp->cc_blob_address = (u32)(unsigned long)cc_info; + + return true; +} + +void __head __noreturn snp_abort(void) +{ + sev_es_terminate(SEV_TERM_SET_GEN, GHCB_SNP_UNSUPPORTED); +} diff --git a/arch/x86/boot/startup/sme.c b/arch/x86/boot/startup/sme.c new file mode 100644 index 000000000000..70ea1748c0a7 --- /dev/null +++ b/arch/x86/boot/startup/sme.c @@ -0,0 +1,575 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * AMD Memory Encryption Support + * + * Copyright (C) 2016 Advanced Micro Devices, Inc. + * + * Author: Tom Lendacky <thomas.lendacky@amd.com> + */ + +/* + * Since we're dealing with identity mappings, physical and virtual + * addresses are the same, so override these defines which are ultimately + * used by the headers in misc.h. + */ +#define __pa(x) ((unsigned long)(x)) +#define __va(x) ((void *)((unsigned long)(x))) + +/* + * Special hack: we have to be careful, because no indirections are + * allowed here, and paravirt_ops is a kind of one. As it will only run in + * baremetal anyway, we just keep it from happening. (This list needs to + * be extended when new paravirt and debugging variants are added.) + */ +#undef CONFIG_PARAVIRT +#undef CONFIG_PARAVIRT_XXL +#undef CONFIG_PARAVIRT_SPINLOCKS + +/* + * This code runs before CPU feature bits are set. By default, the + * pgtable_l5_enabled() function uses bit X86_FEATURE_LA57 to determine if + * 5-level paging is active, so that won't work here. USE_EARLY_PGTABLE_L5 + * is provided to handle this situation and, instead, use a variable that + * has been set by the early boot code. + */ +#define USE_EARLY_PGTABLE_L5 + +#include <linux/kernel.h> +#include <linux/mm.h> +#include <linux/mem_encrypt.h> +#include <linux/cc_platform.h> + +#include <asm/init.h> +#include <asm/setup.h> +#include <asm/sections.h> +#include <asm/coco.h> +#include <asm/sev.h> + +#define PGD_FLAGS _KERNPG_TABLE_NOENC +#define P4D_FLAGS _KERNPG_TABLE_NOENC +#define PUD_FLAGS _KERNPG_TABLE_NOENC +#define PMD_FLAGS _KERNPG_TABLE_NOENC + +#define PMD_FLAGS_LARGE (__PAGE_KERNEL_LARGE_EXEC & ~_PAGE_GLOBAL) + +#define PMD_FLAGS_DEC PMD_FLAGS_LARGE +#define PMD_FLAGS_DEC_WP ((PMD_FLAGS_DEC & ~_PAGE_LARGE_CACHE_MASK) | \ + (_PAGE_PAT_LARGE | _PAGE_PWT)) + +#define PMD_FLAGS_ENC (PMD_FLAGS_LARGE | _PAGE_ENC) + +#define PTE_FLAGS (__PAGE_KERNEL_EXEC & ~_PAGE_GLOBAL) + +#define PTE_FLAGS_DEC PTE_FLAGS +#define PTE_FLAGS_DEC_WP ((PTE_FLAGS_DEC & ~_PAGE_CACHE_MASK) | \ + (_PAGE_PAT | _PAGE_PWT)) + +#define PTE_FLAGS_ENC (PTE_FLAGS | _PAGE_ENC) + +struct sme_populate_pgd_data { + void *pgtable_area; + pgd_t *pgd; + + pmdval_t pmd_flags; + pteval_t pte_flags; + unsigned long paddr; + + unsigned long vaddr; + unsigned long vaddr_end; +}; + +/* + * This work area lives in the .init.scratch section, which lives outside of + * the kernel proper. It is sized to hold the intermediate copy buffer and + * more than enough pagetable pages. + * + * By using this section, the kernel can be encrypted in place and it + * avoids any possibility of boot parameters or initramfs images being + * placed such that the in-place encryption logic overwrites them. This + * section is 2MB aligned to allow for simple pagetable setup using only + * PMD entries (see vmlinux.lds.S). + */ +static char sme_workarea[2 * PMD_SIZE] __section(".init.scratch"); + +static void __head sme_clear_pgd(struct sme_populate_pgd_data *ppd) +{ + unsigned long pgd_start, pgd_end, pgd_size; + pgd_t *pgd_p; + + pgd_start = ppd->vaddr & PGDIR_MASK; + pgd_end = ppd->vaddr_end & PGDIR_MASK; + + pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1) * sizeof(pgd_t); + + pgd_p = ppd->pgd + pgd_index(ppd->vaddr); + + memset(pgd_p, 0, pgd_size); +} + +static pud_t __head *sme_prepare_pgd(struct sme_populate_pgd_data *ppd) +{ + pgd_t *pgd; + p4d_t *p4d; + pud_t *pud; + pmd_t *pmd; + + pgd = ppd->pgd + pgd_index(ppd->vaddr); + if (pgd_none(*pgd)) { + p4d = ppd->pgtable_area; + memset(p4d, 0, sizeof(*p4d) * PTRS_PER_P4D); + ppd->pgtable_area += sizeof(*p4d) * PTRS_PER_P4D; + set_pgd(pgd, __pgd(PGD_FLAGS | __pa(p4d))); + } + + p4d = p4d_offset(pgd, ppd->vaddr); + if (p4d_none(*p4d)) { + pud = ppd->pgtable_area; + memset(pud, 0, sizeof(*pud) * PTRS_PER_PUD); + ppd->pgtable_area += sizeof(*pud) * PTRS_PER_PUD; + set_p4d(p4d, __p4d(P4D_FLAGS | __pa(pud))); + } + + pud = pud_offset(p4d, ppd->vaddr); + if (pud_none(*pud)) { + pmd = ppd->pgtable_area; + memset(pmd, 0, sizeof(*pmd) * PTRS_PER_PMD); + ppd->pgtable_area += sizeof(*pmd) * PTRS_PER_PMD; + set_pud(pud, __pud(PUD_FLAGS | __pa(pmd))); + } + + if (pud_leaf(*pud)) + return NULL; + + return pud; +} + +static void __head sme_populate_pgd_large(struct sme_populate_pgd_data *ppd) +{ + pud_t *pud; + pmd_t *pmd; + + pud = sme_prepare_pgd(ppd); + if (!pud) + return; + + pmd = pmd_offset(pud, ppd->vaddr); + if (pmd_leaf(*pmd)) + return; + + set_pmd(pmd, __pmd(ppd->paddr | ppd->pmd_flags)); +} + +static void __head sme_populate_pgd(struct sme_populate_pgd_data *ppd) +{ + pud_t *pud; + pmd_t *pmd; + pte_t *pte; + + pud = sme_prepare_pgd(ppd); + if (!pud) + return; + + pmd = pmd_offset(pud, ppd->vaddr); + if (pmd_none(*pmd)) { + pte = ppd->pgtable_area; + memset(pte, 0, sizeof(*pte) * PTRS_PER_PTE); + ppd->pgtable_area += sizeof(*pte) * PTRS_PER_PTE; + set_pmd(pmd, __pmd(PMD_FLAGS | __pa(pte))); + } + + if (pmd_leaf(*pmd)) + return; + + pte = pte_offset_kernel(pmd, ppd->vaddr); + if (pte_none(*pte)) + set_pte(pte, __pte(ppd->paddr | ppd->pte_flags)); +} + +static void __head __sme_map_range_pmd(struct sme_populate_pgd_data *ppd) +{ + while (ppd->vaddr < ppd->vaddr_end) { + sme_populate_pgd_large(ppd); + + ppd->vaddr += PMD_SIZE; + ppd->paddr += PMD_SIZE; + } +} + +static void __head __sme_map_range_pte(struct sme_populate_pgd_data *ppd) +{ + while (ppd->vaddr < ppd->vaddr_end) { + sme_populate_pgd(ppd); + + ppd->vaddr += PAGE_SIZE; + ppd->paddr += PAGE_SIZE; + } +} + +static void __head __sme_map_range(struct sme_populate_pgd_data *ppd, + pmdval_t pmd_flags, pteval_t pte_flags) +{ + unsigned long vaddr_end; + + ppd->pmd_flags = pmd_flags; + ppd->pte_flags = pte_flags; + + /* Save original end value since we modify the struct value */ + vaddr_end = ppd->vaddr_end; + + /* If start is not 2MB aligned, create PTE entries */ + ppd->vaddr_end = ALIGN(ppd->vaddr, PMD_SIZE); + __sme_map_range_pte(ppd); + + /* Create PMD entries */ + ppd->vaddr_end = vaddr_end & PMD_MASK; + __sme_map_range_pmd(ppd); + + /* If end is not 2MB aligned, create PTE entries */ + ppd->vaddr_end = vaddr_end; + __sme_map_range_pte(ppd); +} + +static void __head sme_map_range_encrypted(struct sme_populate_pgd_data *ppd) +{ + __sme_map_range(ppd, PMD_FLAGS_ENC, PTE_FLAGS_ENC); +} + +static void __head sme_map_range_decrypted(struct sme_populate_pgd_data *ppd) +{ + __sme_map_range(ppd, PMD_FLAGS_DEC, PTE_FLAGS_DEC); +} + +static void __head sme_map_range_decrypted_wp(struct sme_populate_pgd_data *ppd) +{ + __sme_map_range(ppd, PMD_FLAGS_DEC_WP, PTE_FLAGS_DEC_WP); +} + +static unsigned long __head sme_pgtable_calc(unsigned long len) +{ + unsigned long entries = 0, tables = 0; + + /* + * Perform a relatively simplistic calculation of the pagetable + * entries that are needed. Those mappings will be covered mostly + * by 2MB PMD entries so we can conservatively calculate the required + * number of P4D, PUD and PMD structures needed to perform the + * mappings. For mappings that are not 2MB aligned, PTE mappings + * would be needed for the start and end portion of the address range + * that fall outside of the 2MB alignment. This results in, at most, + * two extra pages to hold PTE entries for each range that is mapped. + * Incrementing the count for each covers the case where the addresses + * cross entries. + */ + + /* PGDIR_SIZE is equal to P4D_SIZE on 4-level machine. */ + if (PTRS_PER_P4D > 1) + entries += (DIV_ROUND_UP(len, PGDIR_SIZE) + 1) * sizeof(p4d_t) * PTRS_PER_P4D; + entries += (DIV_ROUND_UP(len, P4D_SIZE) + 1) * sizeof(pud_t) * PTRS_PER_PUD; + entries += (DIV_ROUND_UP(len, PUD_SIZE) + 1) * sizeof(pmd_t) * PTRS_PER_PMD; + entries += 2 * sizeof(pte_t) * PTRS_PER_PTE; + + /* + * Now calculate the added pagetable structures needed to populate + * the new pagetables. + */ + + if (PTRS_PER_P4D > 1) + tables += DIV_ROUND_UP(entries, PGDIR_SIZE) * sizeof(p4d_t) * PTRS_PER_P4D; + tables += DIV_ROUND_UP(entries, P4D_SIZE) * sizeof(pud_t) * PTRS_PER_PUD; + tables += DIV_ROUND_UP(entries, PUD_SIZE) * sizeof(pmd_t) * PTRS_PER_PMD; + + return entries + tables; +} + +void __head sme_encrypt_kernel(struct boot_params *bp) +{ + unsigned long workarea_start, workarea_end, workarea_len; + unsigned long execute_start, execute_end, execute_len; + unsigned long kernel_start, kernel_end, kernel_len; + unsigned long initrd_start, initrd_end, initrd_len; + struct sme_populate_pgd_data ppd; + unsigned long pgtable_area_len; + unsigned long decrypted_base; + + /* + * This is early code, use an open coded check for SME instead of + * using cc_platform_has(). This eliminates worries about removing + * instrumentation or checking boot_cpu_data in the cc_platform_has() + * function. + */ + if (!sme_get_me_mask() || sev_status & MSR_AMD64_SEV_ENABLED) + return; + + /* + * Prepare for encrypting the kernel and initrd by building new + * pagetables with the necessary attributes needed to encrypt the + * kernel in place. + * + * One range of virtual addresses will map the memory occupied + * by the kernel and initrd as encrypted. + * + * Another range of virtual addresses will map the memory occupied + * by the kernel and initrd as decrypted and write-protected. + * + * The use of write-protect attribute will prevent any of the + * memory from being cached. + */ + + kernel_start = (unsigned long)rip_rel_ptr(_text); + kernel_end = ALIGN((unsigned long)rip_rel_ptr(_end), PMD_SIZE); + kernel_len = kernel_end - kernel_start; + + initrd_start = 0; + initrd_end = 0; + initrd_len = 0; +#ifdef CONFIG_BLK_DEV_INITRD + initrd_len = (unsigned long)bp->hdr.ramdisk_size | + ((unsigned long)bp->ext_ramdisk_size << 32); + if (initrd_len) { + initrd_start = (unsigned long)bp->hdr.ramdisk_image | + ((unsigned long)bp->ext_ramdisk_image << 32); + initrd_end = PAGE_ALIGN(initrd_start + initrd_len); + initrd_len = initrd_end - initrd_start; + } +#endif + + /* + * Calculate required number of workarea bytes needed: + * executable encryption area size: + * stack page (PAGE_SIZE) + * encryption routine page (PAGE_SIZE) + * intermediate copy buffer (PMD_SIZE) + * pagetable structures for the encryption of the kernel + * pagetable structures for workarea (in case not currently mapped) + */ + execute_start = workarea_start = (unsigned long)rip_rel_ptr(sme_workarea); + execute_end = execute_start + (PAGE_SIZE * 2) + PMD_SIZE; + execute_len = execute_end - execute_start; + + /* + * One PGD for both encrypted and decrypted mappings and a set of + * PUDs and PMDs for each of the encrypted and decrypted mappings. + */ + pgtable_area_len = sizeof(pgd_t) * PTRS_PER_PGD; + pgtable_area_len += sme_pgtable_calc(execute_end - kernel_start) * 2; + if (initrd_len) + pgtable_area_len += sme_pgtable_calc(initrd_len) * 2; + + /* PUDs and PMDs needed in the current pagetables for the workarea */ + pgtable_area_len += sme_pgtable_calc(execute_len + pgtable_area_len); + + /* + * The total workarea includes the executable encryption area and + * the pagetable area. The start of the workarea is already 2MB + * aligned, align the end of the workarea on a 2MB boundary so that + * we don't try to create/allocate PTE entries from the workarea + * before it is mapped. + */ + workarea_len = execute_len + pgtable_area_len; + workarea_end = ALIGN(workarea_start + workarea_len, PMD_SIZE); + + /* + * Set the address to the start of where newly created pagetable + * structures (PGDs, PUDs and PMDs) will be allocated. New pagetable + * structures are created when the workarea is added to the current + * pagetables and when the new encrypted and decrypted kernel + * mappings are populated. + */ + ppd.pgtable_area = (void *)execute_end; + + /* + * Make sure the current pagetable structure has entries for + * addressing the workarea. + */ + ppd.pgd = (pgd_t *)native_read_cr3_pa(); + ppd.paddr = workarea_start; + ppd.vaddr = workarea_start; + ppd.vaddr_end = workarea_end; + sme_map_range_decrypted(&ppd); + + /* Flush the TLB - no globals so cr3 is enough */ + native_write_cr3(__native_read_cr3()); + + /* + * A new pagetable structure is being built to allow for the kernel + * and initrd to be encrypted. It starts with an empty PGD that will + * then be populated with new PUDs and PMDs as the encrypted and + * decrypted kernel mappings are created. + */ + ppd.pgd = ppd.pgtable_area; + memset(ppd.pgd, 0, sizeof(pgd_t) * PTRS_PER_PGD); + ppd.pgtable_area += sizeof(pgd_t) * PTRS_PER_PGD; + + /* + * A different PGD index/entry must be used to get different + * pagetable entries for the decrypted mapping. Choose the next + * PGD index and convert it to a virtual address to be used as + * the base of the mapping. + */ + decrypted_base = (pgd_index(workarea_end) + 1) & (PTRS_PER_PGD - 1); + if (initrd_len) { + unsigned long check_base; + + check_base = (pgd_index(initrd_end) + 1) & (PTRS_PER_PGD - 1); + decrypted_base = max(decrypted_base, check_base); + } + decrypted_base <<= PGDIR_SHIFT; + + /* Add encrypted kernel (identity) mappings */ + ppd.paddr = kernel_start; + ppd.vaddr = kernel_start; + ppd.vaddr_end = kernel_end; + sme_map_range_encrypted(&ppd); + + /* Add decrypted, write-protected kernel (non-identity) mappings */ + ppd.paddr = kernel_start; + ppd.vaddr = kernel_start + decrypted_base; + ppd.vaddr_end = kernel_end + decrypted_base; + sme_map_range_decrypted_wp(&ppd); + + if (initrd_len) { + /* Add encrypted initrd (identity) mappings */ + ppd.paddr = initrd_start; + ppd.vaddr = initrd_start; + ppd.vaddr_end = initrd_end; + sme_map_range_encrypted(&ppd); + /* + * Add decrypted, write-protected initrd (non-identity) mappings + */ + ppd.paddr = initrd_start; + ppd.vaddr = initrd_start + decrypted_base; + ppd.vaddr_end = initrd_end + decrypted_base; + sme_map_range_decrypted_wp(&ppd); + } + + /* Add decrypted workarea mappings to both kernel mappings */ + ppd.paddr = workarea_start; + ppd.vaddr = workarea_start; + ppd.vaddr_end = workarea_end; + sme_map_range_decrypted(&ppd); + + ppd.paddr = workarea_start; + ppd.vaddr = workarea_start + decrypted_base; + ppd.vaddr_end = workarea_end + decrypted_base; + sme_map_range_decrypted(&ppd); + + /* Perform the encryption */ + sme_encrypt_execute(kernel_start, kernel_start + decrypted_base, + kernel_len, workarea_start, (unsigned long)ppd.pgd); + + if (initrd_len) + sme_encrypt_execute(initrd_start, initrd_start + decrypted_base, + initrd_len, workarea_start, + (unsigned long)ppd.pgd); + + /* + * At this point we are running encrypted. Remove the mappings for + * the decrypted areas - all that is needed for this is to remove + * the PGD entry/entries. + */ + ppd.vaddr = kernel_start + decrypted_base; + ppd.vaddr_end = kernel_end + decrypted_base; + sme_clear_pgd(&ppd); + + if (initrd_len) { + ppd.vaddr = initrd_start + decrypted_base; + ppd.vaddr_end = initrd_end + decrypted_base; + sme_clear_pgd(&ppd); + } + + ppd.vaddr = workarea_start + decrypted_base; + ppd.vaddr_end = workarea_end + decrypted_base; + sme_clear_pgd(&ppd); + + /* Flush the TLB - no globals so cr3 is enough */ + native_write_cr3(__native_read_cr3()); +} + +void __head sme_enable(struct boot_params *bp) +{ + unsigned int eax, ebx, ecx, edx; + unsigned long feature_mask; + unsigned long me_mask; + bool snp_en; + u64 msr; + + snp_en = snp_init(bp); + + /* Check for the SME/SEV support leaf */ + eax = 0x80000000; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + if (eax < 0x8000001f) + return; + +#define AMD_SME_BIT BIT(0) +#define AMD_SEV_BIT BIT(1) + + /* + * Check for the SME/SEV feature: + * CPUID Fn8000_001F[EAX] + * - Bit 0 - Secure Memory Encryption support + * - Bit 1 - Secure Encrypted Virtualization support + * CPUID Fn8000_001F[EBX] + * - Bits 5:0 - Pagetable bit position used to indicate encryption + */ + eax = 0x8000001f; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + /* Check whether SEV or SME is supported */ + if (!(eax & (AMD_SEV_BIT | AMD_SME_BIT))) + return; + + me_mask = 1UL << (ebx & 0x3f); + + /* Check the SEV MSR whether SEV or SME is enabled */ + sev_status = msr = native_rdmsrq(MSR_AMD64_SEV); + feature_mask = (msr & MSR_AMD64_SEV_ENABLED) ? AMD_SEV_BIT : AMD_SME_BIT; + + /* + * Any discrepancies between the presence of a CC blob and SNP + * enablement abort the guest. + */ + if (snp_en ^ !!(msr & MSR_AMD64_SEV_SNP_ENABLED)) + snp_abort(); + + /* Check if memory encryption is enabled */ + if (feature_mask == AMD_SME_BIT) { + if (!(bp->hdr.xloadflags & XLF_MEM_ENCRYPTION)) + return; + + /* + * No SME if Hypervisor bit is set. This check is here to + * prevent a guest from trying to enable SME. For running as a + * KVM guest the MSR_AMD64_SYSCFG will be sufficient, but there + * might be other hypervisors which emulate that MSR as non-zero + * or even pass it through to the guest. + * A malicious hypervisor can still trick a guest into this + * path, but there is no way to protect against that. + */ + eax = 1; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + if (ecx & BIT(31)) + return; + + /* For SME, check the SYSCFG MSR */ + msr = native_rdmsrq(MSR_AMD64_SYSCFG); + if (!(msr & MSR_AMD64_SYSCFG_MEM_ENCRYPT)) + return; + } + + sme_me_mask = me_mask; + physical_mask &= ~me_mask; + cc_vendor = CC_VENDOR_AMD; + cc_set_mask(me_mask); +} + +#ifdef CONFIG_MITIGATION_PAGE_TABLE_ISOLATION +/* Local version for startup code, which never operates on user page tables */ +__weak +pgd_t __pti_set_user_pgtbl(pgd_t *pgdp, pgd_t pgd) +{ + return pgd; +} +#endif |