x86/efi: Correct EFI identity mapping under 'efi=old_map' when KASLR is enabled

For EFI with the 'efi=old_map' kernel option specified, the kernel will panic
when KASLR is enabled:

  BUG: unable to handle kernel paging request at 000000007febd57e
  IP: 0x7febd57e
  PGD 1025a067
  PUD 0

  Oops: 0010 [#1] SMP
  Call Trace:
   efi_enter_virtual_mode()
   start_kernel()
   x86_64_start_reservations()
   x86_64_start_kernel()
   start_cpu()

The root cause is that the identity mapping is not built correctly
in the 'efi=old_map' case.

On 'nokaslr' kernels, PAGE_OFFSET is 0xffff880000000000 which is PGDIR_SIZE
aligned. We can borrow the PUD table from the direct mappings safely. Given a
physical address X, we have pud_index(X) == pud_index(__va(X)).

However, on KASLR kernels, PAGE_OFFSET is PUD_SIZE aligned. For a given physical
address X, pud_index(X) != pud_index(__va(X)). We can't just copy the PGD entry
from direct mapping to build identity mapping, instead we need to copy the
PUD entries one by one from the direct mapping.

Fix it.
Signed-off-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Bhupesh Sharma <bhsharma@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Young <dyoung@redhat.com>
Cc: Frank Ramsay <frank.ramsay@hpe.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Russ Anderson <rja@sgi.com>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-efi@vger.kernel.org
Link: http://lkml.kernel.org/r/20170526113652.21339-5-matt@codeblueprint.co.uk
[ Fixed and reworded the changelog and code comments to be more readable. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/platform/efi/efi_64.c

@@ -71,11 +71,13 @@ static void __init early_code_mapping_set_exec(int executable)
 
 pgd_t * __init efi_call_phys_prolog(void)
 {
-	unsigned long vaddress;
-	pgd_t *save_pgd;
+	unsigned long vaddr, addr_pgd, addr_p4d, addr_pud;
+	pgd_t *save_pgd, *pgd_k, *pgd_efi;
+	p4d_t *p4d, *p4d_k, *p4d_efi;
+	pud_t *pud;
 
 	int pgd;
-	int n_pgds;
+	int n_pgds, i, j;
 
 	if (!efi_enabled(EFI_OLD_MEMMAP)) {
 		save_pgd = (pgd_t *)read_cr3();
@@ -88,10 +90,49 @@ pgd_t * __init efi_call_phys_prolog(void)
 	n_pgds = DIV_ROUND_UP((max_pfn << PAGE_SHIFT), PGDIR_SIZE);
 	save_pgd = kmalloc_array(n_pgds, sizeof(*save_pgd), GFP_KERNEL);
 
+	/*
+	 * Build 1:1 identity mapping for efi=old_map usage. Note that
+	 * PAGE_OFFSET is PGDIR_SIZE aligned when KASLR is disabled, while
+	 * it is PUD_SIZE ALIGNED with KASLR enabled. So for a given physical
+	 * address X, the pud_index(X) != pud_index(__va(X)), we can only copy
+	 * PUD entry of __va(X) to fill in pud entry of X to build 1:1 mapping.
+	 * This means here we can only reuse the PMD tables of the direct mapping.
+	 */
 	for (pgd = 0; pgd < n_pgds; pgd++) {
-		save_pgd[pgd] = *pgd_offset_k(pgd * PGDIR_SIZE);
-		vaddress = (unsigned long)__va(pgd * PGDIR_SIZE);
-		set_pgd(pgd_offset_k(pgd * PGDIR_SIZE), *pgd_offset_k(vaddress));
+		addr_pgd = (unsigned long)(pgd * PGDIR_SIZE);
+		vaddr = (unsigned long)__va(pgd * PGDIR_SIZE);
+		pgd_efi = pgd_offset_k(addr_pgd);
+		save_pgd[pgd] = *pgd_efi;
+
+		p4d = p4d_alloc(&init_mm, pgd_efi, addr_pgd);
+		if (!p4d) {
+			pr_err("Failed to allocate p4d table!\n");
+			goto out;
+		}
+
+		for (i = 0; i < PTRS_PER_P4D; i++) {
+			addr_p4d = addr_pgd + i * P4D_SIZE;
+			p4d_efi = p4d + p4d_index(addr_p4d);
+
+			pud = pud_alloc(&init_mm, p4d_efi, addr_p4d);
+			if (!pud) {
+				pr_err("Failed to allocate pud table!\n");
+				goto out;
+			}
+
+			for (j = 0; j < PTRS_PER_PUD; j++) {
+				addr_pud = addr_p4d + j * PUD_SIZE;
+
+				if (addr_pud > (max_pfn << PAGE_SHIFT))
+					break;
+
+				vaddr = (unsigned long)__va(addr_pud);
+
+				pgd_k = pgd_offset_k(vaddr);
+				p4d_k = p4d_offset(pgd_k, vaddr);
+				pud[j] = *pud_offset(p4d_k, vaddr);
+			}
+		}
 	}
 out:
 	__flush_tlb_all();
@@ -104,8 +145,11 @@ void __init efi_call_phys_epilog(pgd_t *save_pgd)
 	/*
 	 * After the lock is released, the original page table is restored.
 	 */
-	int pgd_idx;
+	int pgd_idx, i;
 	int nr_pgds;
+	pgd_t *pgd;
+	p4d_t *p4d;
+	pud_t *pud;
 
 	if (!efi_enabled(EFI_OLD_MEMMAP)) {
 		write_cr3((unsigned long)save_pgd);
@@ -115,9 +159,28 @@ void __init efi_call_phys_epilog(pgd_t *save_pgd)
 
 	nr_pgds = DIV_ROUND_UP((max_pfn << PAGE_SHIFT) , PGDIR_SIZE);
 
-	for (pgd_idx = 0; pgd_idx < nr_pgds; pgd_idx++)
+	for (pgd_idx = 0; pgd_idx < nr_pgds; pgd_idx++) {
+		pgd = pgd_offset_k(pgd_idx * PGDIR_SIZE);
 		set_pgd(pgd_offset_k(pgd_idx * PGDIR_SIZE), save_pgd[pgd_idx]);
 
+		if (!(pgd_val(*pgd) & _PAGE_PRESENT))
+			continue;
+
+		for (i = 0; i < PTRS_PER_P4D; i++) {
+			p4d = p4d_offset(pgd,
+					 pgd_idx * PGDIR_SIZE + i * P4D_SIZE);
+
+			if (!(p4d_val(*p4d) & _PAGE_PRESENT))
+				continue;
+
+			pud = (pud_t *)p4d_page_vaddr(*p4d);
+			pud_free(&init_mm, pud);
+		}
+
+		p4d = (p4d_t *)pgd_page_vaddr(*pgd);
+		p4d_free(&init_mm, p4d);
+	}
+
 	kfree(save_pgd);
 
 	__flush_tlb_all();