[S390] noexec protection

This provides a noexec protection on s390 hardware. Our hardware does
not have any bits left in the pte for a hw noexec bit, so this is a
different approach using shadow page tables and a special addressing
mode that allows separate address spaces for code and data.

As a special feature of our "secondary-space" addressing mode, separate
page tables can be specified for the translation of data addresses
(storage operands) and instruction addresses. The shadow page table is
used for the instruction addresses and the standard page table for the
data addresses.
The shadow page table is linked to the standard page table by a pointer
in page->lru.next of the struct page corresponding to the page that
contains the standard page table (since page->private is not really
private with the pte_lock and the page table pages are not in the LRU
list).
Depending on the software bits of a pte, it is either inserted into
both page tables or just into the standard (data) page table. Pages of
a vma that does not have the VM_EXEC bit set get mapped only in the
data address space. Any try to execute code on such a page will cause a
page translation exception. The standard reaction to this is a SIGSEGV
with two exceptions: the two system call opcodes 0x0a77 (sys_sigreturn)
and 0x0aad (sys_rt_sigreturn) are allowed. They are stored by the
kernel to the signal stack frame. Unfortunately, the signal return
mechanism cannot be modified to use an SA_RESTORER because the
exception unwinding code depends on the system call opcode stored
behind the signal stack frame.

This feature requires that user space is executed in secondary-space
mode and the kernel in home-space mode, which means that the addressing
modes need to be switched and that the noexec protection only works
for user space.
After switching the addressing modes, we cannot use the mvcp/mvcs
instructions anymore to copy between kernel and user space. A new
mvcos instruction has been added to the z9 EC/BC hardware which allows
to copy between arbitrary address spaces, but on older hardware the
page tables need to be walked manually.

Signed-off-by: Gerald Schaefer <geraldsc@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

2 files changed, 372 insertions, 1 deletions

diff --git a/arch/s390/lib/uaccess_mvcos.c b/arch/s390/lib/uaccess_mvcos.c
index 78c48f88f5f..6d8772339d7 100644
--- a/arch/s390/lib/uaccess_mvcos.c
+++ b/arch/s390/lib/uaccess_mvcos.c
@@ -162,6 +162,44 @@ static size_t clear_user_mvcos(size_t size, void __user *to)
 	return size;
 }
 
+static size_t strnlen_user_mvcos(size_t count, const char __user *src)
+{
+	char buf[256];
+	int rc;
+	size_t done, len, len_str;
+
+	done = 0;
+	do {
+		len = min(count - done, (size_t) 256);
+		rc = uaccess.copy_from_user(len, src + done, buf);
+		if (unlikely(rc == len))
+			return 0;
+		len -= rc;
+		len_str = strnlen(buf, len);
+		done += len_str;
+	} while ((len_str == len) && (done < count));
+	return done + 1;
+}
+
+static size_t strncpy_from_user_mvcos(size_t count, const char __user *src,
+				      char *dst)
+{
+	int rc;
+	size_t done, len, len_str;
+
+	done = 0;
+	do {
+		len = min(count - done, (size_t) 4096);
+		rc = uaccess.copy_from_user(len, src + done, dst);
+		if (unlikely(rc == len))
+			return -EFAULT;
+		len -= rc;
+		len_str = strnlen(dst, len);
+		done += len_str;
+	} while ((len_str == len) && (done < count));
+	return done;
+}
+
 struct uaccess_ops uaccess_mvcos = {
 	.copy_from_user = copy_from_user_mvcos_check,
 	.copy_from_user_small = copy_from_user_std,
@@ -174,3 +212,18 @@ struct uaccess_ops uaccess_mvcos = {
 	.futex_atomic_op = futex_atomic_op_std,
 	.futex_atomic_cmpxchg = futex_atomic_cmpxchg_std,
 };
+
+#ifdef CONFIG_S390_SWITCH_AMODE
+struct uaccess_ops uaccess_mvcos_switch = {
+	.copy_from_user = copy_from_user_mvcos,
+	.copy_from_user_small = copy_from_user_mvcos,
+	.copy_to_user = copy_to_user_mvcos,
+	.copy_to_user_small = copy_to_user_mvcos,
+	.copy_in_user = copy_in_user_mvcos,
+	.clear_user = clear_user_mvcos,
+	.strnlen_user = strnlen_user_mvcos,
+	.strncpy_from_user = strncpy_from_user_mvcos,
+	.futex_atomic_op = futex_atomic_op_pt,
+	.futex_atomic_cmpxchg = futex_atomic_cmpxchg_pt,
+};
+#endif
diff --git a/arch/s390/lib/uaccess_pt.c b/arch/s390/lib/uaccess_pt.c
index 24ead559c7b..637192fa7c9 100644
--- a/arch/s390/lib/uaccess_pt.c
+++ b/arch/s390/lib/uaccess_pt.c
@@ -1,7 +1,8 @@
 /*
  *  arch/s390/lib/uaccess_pt.c
  *
- *  User access functions based on page table walks.
+ *  User access functions based on page table walks for enhanced
+ *  system layout without hardware support.
  *
  *    Copyright IBM Corp. 2006
  *    Author(s): Gerald Schaefer (gerald.schaefer@de.ibm.com)
@@ -134,6 +135,49 @@ fault:
 	goto retry;
 }
 
+/*
+ * Do DAT for user address by page table walk, return kernel address.
+ * This function needs to be called with current->mm->page_table_lock held.
+ */
+static inline unsigned long __dat_user_addr(unsigned long uaddr)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long pfn, ret;
+	pgd_t *pgd;
+	pmd_t *pmd;
+	pte_t *pte;
+	int rc;
+
+	ret = 0;
+retry:
+	pgd = pgd_offset(mm, uaddr);
+	if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
+		goto fault;
+
+	pmd = pmd_offset(pgd, uaddr);
+	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
+		goto fault;
+
+	pte = pte_offset_map(pmd, uaddr);
+	if (!pte || !pte_present(*pte))
+		goto fault;
+
+	pfn = pte_pfn(*pte);
+	if (!pfn_valid(pfn))
+		goto out;
+
+	ret = (pfn << PAGE_SHIFT) + (uaddr & (PAGE_SIZE - 1));
+out:
+	return ret;
+fault:
+	spin_unlock(&mm->page_table_lock);
+	rc = __handle_fault(mm, uaddr, 0);
+	spin_lock(&mm->page_table_lock);
+	if (rc)
+		goto out;
+	goto retry;
+}
+
 size_t copy_from_user_pt(size_t n, const void __user *from, void *to)
 {
 	size_t rc;
@@ -156,3 +200,277 @@ size_t copy_to_user_pt(size_t n, void __user *to, const void *from)
 	}
 	return __user_copy_pt((unsigned long) to, (void *) from, n, 1);
 }
+
+static size_t clear_user_pt(size_t n, void __user *to)
+{
+	long done, size, ret;
+
+	if (segment_eq(get_fs(), KERNEL_DS)) {
+		memset((void __kernel __force *) to, 0, n);
+		return 0;
+	}
+	done = 0;
+	do {
+		if (n - done > PAGE_SIZE)
+			size = PAGE_SIZE;
+		else
+			size = n - done;
+		ret = __user_copy_pt((unsigned long) to + done,
+				      &empty_zero_page, size, 1);
+		done += size;
+		if (ret)
+			return ret + n - done;
+	} while (done < n);
+	return 0;
+}
+
+static size_t strnlen_user_pt(size_t count, const char __user *src)
+{
+	char *addr;
+	unsigned long uaddr = (unsigned long) src;
+	struct mm_struct *mm = current->mm;
+	unsigned long offset, pfn, done, len;
+	pgd_t *pgd;
+	pmd_t *pmd;
+	pte_t *pte;
+	size_t len_str;
+
+	if (segment_eq(get_fs(), KERNEL_DS))
+		return strnlen((const char __kernel __force *) src, count) + 1;
+	done = 0;
+retry:
+	spin_lock(&mm->page_table_lock);
+	do {
+		pgd = pgd_offset(mm, uaddr);
+		if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
+			goto fault;
+
+		pmd = pmd_offset(pgd, uaddr);
+		if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
+			goto fault;
+
+		pte = pte_offset_map(pmd, uaddr);
+		if (!pte || !pte_present(*pte))
+			goto fault;
+
+		pfn = pte_pfn(*pte);
+		if (!pfn_valid(pfn)) {
+			done = -1;
+			goto out;
+		}
+
+		offset = uaddr & (PAGE_SIZE-1);
+		addr = (char *)(pfn << PAGE_SHIFT) + offset;
+		len = min(count - done, PAGE_SIZE - offset);
+		len_str = strnlen(addr, len);
+		done += len_str;
+		uaddr += len_str;
+	} while ((len_str == len) && (done < count));
+out:
+	spin_unlock(&mm->page_table_lock);
+	return done + 1;
+fault:
+	spin_unlock(&mm->page_table_lock);
+	if (__handle_fault(mm, uaddr, 0)) {
+		return 0;
+	}
+	goto retry;
+}
+
+static size_t strncpy_from_user_pt(size_t count, const char __user *src,
+				   char *dst)
+{
+	size_t n = strnlen_user_pt(count, src);
+
+	if (!n)
+		return -EFAULT;
+	if (n > count)
+		n = count;
+	if (segment_eq(get_fs(), KERNEL_DS)) {
+		memcpy(dst, (const char __kernel __force *) src, n);
+		if (dst[n-1] == '\0')
+			return n-1;
+		else
+			return n;
+	}
+	if (__user_copy_pt((unsigned long) src, dst, n, 0))
+		return -EFAULT;
+	if (dst[n-1] == '\0')
+		return n-1;
+	else
+		return n;
+}
+
+static size_t copy_in_user_pt(size_t n, void __user *to,
+			      const void __user *from)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long offset_from, offset_to, offset_max, pfn_from, pfn_to,
+		      uaddr, done, size;
+	unsigned long uaddr_from = (unsigned long) from;
+	unsigned long uaddr_to = (unsigned long) to;
+	pgd_t *pgd_from, *pgd_to;
+	pmd_t *pmd_from, *pmd_to;
+	pte_t *pte_from, *pte_to;
+	int write_user;
+
+	done = 0;
+retry:
+	spin_lock(&mm->page_table_lock);
+	do {
+		pgd_from = pgd_offset(mm, uaddr_from);
+		if (pgd_none(*pgd_from) || unlikely(pgd_bad(*pgd_from))) {
+			uaddr = uaddr_from;
+			write_user = 0;
+			goto fault;
+		}
+		pgd_to = pgd_offset(mm, uaddr_to);
+		if (pgd_none(*pgd_to) || unlikely(pgd_bad(*pgd_to))) {
+			uaddr = uaddr_to;
+			write_user = 1;
+			goto fault;
+		}
+
+		pmd_from = pmd_offset(pgd_from, uaddr_from);
+		if (pmd_none(*pmd_from) || unlikely(pmd_bad(*pmd_from))) {
+			uaddr = uaddr_from;
+			write_user = 0;
+			goto fault;
+		}
+		pmd_to = pmd_offset(pgd_to, uaddr_to);
+		if (pmd_none(*pmd_to) || unlikely(pmd_bad(*pmd_to))) {
+			uaddr = uaddr_to;
+			write_user = 1;
+			goto fault;
+		}
+
+		pte_from = pte_offset_map(pmd_from, uaddr_from);
+		if (!pte_from || !pte_present(*pte_from)) {
+			uaddr = uaddr_from;
+			write_user = 0;
+			goto fault;
+		}
+		pte_to = pte_offset_map(pmd_to, uaddr_to);
+		if (!pte_to || !pte_present(*pte_to) || !pte_write(*pte_to)) {
+			uaddr = uaddr_to;
+			write_user = 1;
+			goto fault;
+		}
+
+		pfn_from = pte_pfn(*pte_from);
+		if (!pfn_valid(pfn_from))
+			goto out;
+		pfn_to = pte_pfn(*pte_to);
+		if (!pfn_valid(pfn_to))
+			goto out;
+
+		offset_from = uaddr_from & (PAGE_SIZE-1);
+		offset_to = uaddr_from & (PAGE_SIZE-1);
+		offset_max = max(offset_from, offset_to);
+		size = min(n - done, PAGE_SIZE - offset_max);
+
+		memcpy((void *)(pfn_to << PAGE_SHIFT) + offset_to,
+		       (void *)(pfn_from << PAGE_SHIFT) + offset_from, size);
+		done += size;
+		uaddr_from += size;
+		uaddr_to += size;
+	} while (done < n);
+out:
+	spin_unlock(&mm->page_table_lock);
+	return n - done;
+fault:
+	spin_unlock(&mm->page_table_lock);
+	if (__handle_fault(mm, uaddr, write_user))
+		return n - done;
+	goto retry;
+}
+
+#define __futex_atomic_op(insn, ret, oldval, newval, uaddr, oparg)	\
+	asm volatile("0: l   %1,0(%6)\n"				\
+		     "1: " insn						\
+		     "2: cs  %1,%2,0(%6)\n"				\
+		     "3: jl  1b\n"					\
+		     "   lhi %0,0\n"					\
+		     "4:\n"						\
+		     EX_TABLE(0b,4b) EX_TABLE(2b,4b) EX_TABLE(3b,4b)	\
+		     : "=d" (ret), "=&d" (oldval), "=&d" (newval),	\
+		       "=m" (*uaddr)					\
+		     : "0" (-EFAULT), "d" (oparg), "a" (uaddr),		\
+		       "m" (*uaddr) : "cc" );
+
+int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
+{
+	int oldval = 0, newval, ret;
+
+	spin_lock(&current->mm->page_table_lock);
+	uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
+	if (!uaddr) {
+		spin_unlock(&current->mm->page_table_lock);
+		return -EFAULT;
+	}
+	get_page(virt_to_page(uaddr));
+	spin_unlock(&current->mm->page_table_lock);
+	switch (op) {
+	case FUTEX_OP_SET:
+		__futex_atomic_op("lr %2,%5\n",
+				  ret, oldval, newval, uaddr, oparg);
+		break;
+	case FUTEX_OP_ADD:
+		__futex_atomic_op("lr %2,%1\nar %2,%5\n",
+				  ret, oldval, newval, uaddr, oparg);
+		break;
+	case FUTEX_OP_OR:
+		__futex_atomic_op("lr %2,%1\nor %2,%5\n",
+				  ret, oldval, newval, uaddr, oparg);
+		break;
+	case FUTEX_OP_ANDN:
+		__futex_atomic_op("lr %2,%1\nnr %2,%5\n",
+				  ret, oldval, newval, uaddr, oparg);
+		break;
+	case FUTEX_OP_XOR:
+		__futex_atomic_op("lr %2,%1\nxr %2,%5\n",
+				  ret, oldval, newval, uaddr, oparg);
+		break;
+	default:
+		ret = -ENOSYS;
+	}
+	put_page(virt_to_page(uaddr));
+	*old = oldval;
+	return ret;
+}
+
+int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
+{
+	int ret;
+
+	spin_lock(&current->mm->page_table_lock);
+	uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
+	if (!uaddr) {
+		spin_unlock(&current->mm->page_table_lock);
+		return -EFAULT;
+	}
+	get_page(virt_to_page(uaddr));
+	spin_unlock(&current->mm->page_table_lock);
+	asm volatile("   cs   %1,%4,0(%5)\n"
+		     "0: lr   %0,%1\n"
+		     "1:\n"
+		     EX_TABLE(0b,1b)
+		     : "=d" (ret), "+d" (oldval), "=m" (*uaddr)
+		     : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
+		     : "cc", "memory" );
+	put_page(virt_to_page(uaddr));
+	return ret;
+}
+
+struct uaccess_ops uaccess_pt = {
+	.copy_from_user		= copy_from_user_pt,
+	.copy_from_user_small	= copy_from_user_pt,
+	.copy_to_user		= copy_to_user_pt,
+	.copy_to_user_small	= copy_to_user_pt,
+	.copy_in_user		= copy_in_user_pt,
+	.clear_user		= clear_user_pt,
+	.strnlen_user		= strnlen_user_pt,
+	.strncpy_from_user	= strncpy_from_user_pt,
+	.futex_atomic_op	= futex_atomic_op_pt,
+	.futex_atomic_cmpxchg	= futex_atomic_cmpxchg_pt,
+};