s390/uaccess: rework uaccess code - fix locking issues

The current uaccess code uses a page table walk in some circumstances,
e.g. in case of the in atomic futex operations or if running on old
hardware which doesn't support the mvcos instruction.

However it turned out that the page table walk code does not correctly
lock page tables when accessing page table entries.
In other words: a different cpu may invalidate a page table entry while
the current cpu inspects the pte. This may lead to random data corruption.

Adding correct locking however isn't trivial for all uaccess operations.
Especially copy_in_user() is problematic since that requires to hold at
least two locks, but must be protected against ABBA deadlock when a
different cpu also performs a copy_in_user() operation.

So the solution is a different approach where we change address spaces:

User space runs in primary address mode, or access register mode within
vdso code, like it currently already does.

The kernel usually also runs in home space mode, however when accessing
user space the kernel switches to primary or secondary address mode if
the mvcos instruction is not available or if a compare-and-swap (futex)
instruction on a user space address is performed.
KVM however is special, since that requires the kernel to run in home
address space while implicitly accessing user space with the sie
instruction.

So we end up with:

User space:
- runs in primary or access register mode
- cr1 contains the user asce
- cr7 contains the user asce
- cr13 contains the kernel asce

Kernel space:
- runs in home space mode
- cr1 contains the user or kernel asce
  -> the kernel asce is loaded when a uaccess requires primary or
     secondary address mode
- cr7 contains the user or kernel asce, (changed with set_fs())
- cr13 contains the kernel asce

In case of uaccess the kernel changes to:
- primary space mode in case of a uaccess (copy_to_user) and uses
  e.g. the mvcp instruction to access user space. However the kernel
  will stay in home space mode if the mvcos instruction is available
- secondary space mode in case of futex atomic operations, so that the
  instructions come from primary address space and data from secondary
  space

In case of kvm the kernel runs in home space mode, but cr1 gets switched
to contain the gmap asce before the sie instruction gets executed. When
the sie instruction is finished cr1 will be switched back to contain the
user asce.

A context switch between two processes will always load the kernel asce
for the next process in cr1. So the first exit to user space is a bit
more expensive (one extra load control register instruction) than before,
however keeps the code rather simple.

In sum this means there is no need to perform any error prone page table
walks anymore when accessing user space.

The patch seems to be rather large, however it mainly removes the
the page table walk code and restores the previously deleted "standard"
uaccess code, with a couple of changes.

The uaccess without mvcos mode can be enforced with the "uaccess_primary"
kernel parameter.
Reported-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/futex.h

 #ifndef _ASM_S390_FUTEX_H
 #define _ASM_S390_FUTEX_H
 
-#include <linux/futex.h>
 #include <linux/uaccess.h>
+#include <linux/futex.h>
+#include <asm/mmu_context.h>
 #include <asm/errno.h>
 
-int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr, u32 oldval, u32 newval);
-int __futex_atomic_op_inuser(int op, u32 __user *uaddr, int oparg, int *old);
+#define __futex_atomic_op(insn, ret, oldval, newval, uaddr, oparg)	\
+	asm volatile(							\
+		"   sacf  256\n"					\
+		"0: l     %1,0(%6)\n"					\
+		"1:"insn						\
+		"2: cs    %1,%2,0(%6)\n"				\
+		"3: jl    1b\n"						\
+		"   lhi   %0,0\n"					\
+		"4: sacf  768\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");
 
 static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
@@ -14,13 +27,37 @@ static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 	int cmp = (encoded_op >> 24) & 15;
 	int oparg = (encoded_op << 8) >> 20;
 	int cmparg = (encoded_op << 20) >> 20;
-	int oldval, ret;
+	int oldval = 0, newval, ret;
 
+	update_primary_asce(current);
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
 	pagefault_disable();
-	ret = __futex_atomic_op_inuser(op, uaddr, oparg, &oldval);
+	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;
+	}
 	pagefault_enable();
 
 	if (!ret) {
@@ -37,4 +74,23 @@ static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 	return ret;
 }
 
+static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+						u32 oldval, u32 newval)
+{
+	int ret;
+
+	update_primary_asce(current);
+	asm volatile(
+		"   sacf 256\n"
+		"0: cs   %1,%4,0(%5)\n"
+		"1: la   %0,0\n"
+		"2: sacf 768\n"
+		EX_TABLE(0b,2b) EX_TABLE(1b,2b)
+		: "=d" (ret), "+d" (oldval), "=m" (*uaddr)
+		: "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
+		: "cc", "memory");
+	*uval = oldval;
+	return ret;
+}
+
 #endif /* _ASM_S390_FUTEX_H */

arch/s390/include/asm/mmu_context.h

@@ -30,27 +30,33 @@ static inline int init_new_context(struct task_struct *tsk,
 
 #define destroy_context(mm)             do { } while (0)
 
-#ifndef CONFIG_64BIT
-#define LCTL_OPCODE "lctl"
-#else
-#define LCTL_OPCODE "lctlg"
-#endif
-
-static inline void update_user_asce(struct mm_struct *mm)
+static inline void update_user_asce(struct mm_struct *mm, int load_primary)
 {
 	pgd_t *pgd = mm->pgd;
 
 	S390_lowcore.user_asce = mm->context.asce_bits | __pa(pgd);
-	/* Load primary space page table origin. */
-	asm volatile(LCTL_OPCODE" 1,1,%0\n" : : "m" (S390_lowcore.user_asce));
+	if (load_primary)
+		__ctl_load(S390_lowcore.user_asce, 1, 1);
 	set_fs(current->thread.mm_segment);
 }
 
-static inline void clear_user_asce(struct mm_struct *mm)
+static inline void clear_user_asce(struct mm_struct *mm, int load_primary)
 {
 	S390_lowcore.user_asce = S390_lowcore.kernel_asce;
-	asm volatile(LCTL_OPCODE" 1,1,%0\n" : : "m" (S390_lowcore.user_asce));
-	asm volatile(LCTL_OPCODE" 7,7,%0\n" : : "m" (S390_lowcore.user_asce));
+
+	if (load_primary)
+		__ctl_load(S390_lowcore.user_asce, 1, 1);
+	__ctl_load(S390_lowcore.user_asce, 7, 7);
+}
+
+static inline void update_primary_asce(struct task_struct *tsk)
+{
+	unsigned long asce;
+
+	__ctl_store(asce, 1, 1);
+	if (asce != S390_lowcore.kernel_asce)
+		__ctl_load(S390_lowcore.kernel_asce, 1, 1);
+	set_tsk_thread_flag(tsk, TIF_ASCE);
 }
 
 static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
@@ -58,6 +64,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 {
 	int cpu = smp_processor_id();
 
+	update_primary_asce(tsk);
 	if (prev == next)
 		return;
 	if (MACHINE_HAS_TLB_LC)
@@ -66,10 +73,10 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/* Delay update_user_asce until all TLB flushes are done. */
 		set_tsk_thread_flag(tsk, TIF_TLB_WAIT);
 		/* Clear old ASCE by loading the kernel ASCE. */
-		clear_user_asce(next);
+		clear_user_asce(next, 0);
 	} else {
 		cpumask_set_cpu(cpu, mm_cpumask(next));
-		update_user_asce(next);
+		update_user_asce(next, 0);
 		if (next->context.flush_mm)
 			/* Flush pending TLBs */
 			__tlb_flush_mm(next);
@@ -94,7 +101,7 @@ static inline void finish_arch_post_lock_switch(void)
 		cpu_relax();
 
 	cpumask_set_cpu(smp_processor_id(), mm_cpumask(mm));
-	update_user_asce(mm);
+	update_user_asce(mm, 0);
 	if (mm->context.flush_mm)
 		__tlb_flush_mm(mm);
 	preempt_enable();

arch/s390/include/asm/switch_to.h

@@ -132,6 +132,7 @@ static inline void restore_access_regs(unsigned int *acrs)
 		update_cr_regs(next);					\
 	}								\
 	prev = __switch_to(prev,next);					\
+	update_primary_asce(current);					\
 } while (0)
 
 #define finish_arch_switch(prev) do {					     \

arch/s390/include/asm/thread_info.h

@@ -82,6 +82,7 @@ static inline struct thread_info *current_thread_info(void)
 #define TIF_SIGPENDING		2	/* signal pending */
 #define TIF_NEED_RESCHED	3	/* rescheduling necessary */
 #define TIF_TLB_WAIT		4	/* wait for TLB flush completion */
+#define TIF_ASCE		5	/* primary asce needs fixup / uaccess */
 #define TIF_PER_TRAP		6	/* deliver sigtrap on return to user */
 #define TIF_MCCK_PENDING	7	/* machine check handling is pending */
 #define TIF_SYSCALL_TRACE	8	/* syscall trace active */
@@ -99,6 +100,7 @@ static inline struct thread_info *current_thread_info(void)
 #define _TIF_SIGPENDING		(1<<TIF_SIGPENDING)
 #define _TIF_NEED_RESCHED	(1<<TIF_NEED_RESCHED)
 #define _TIF_TLB_WAIT		(1<<TIF_TLB_WAIT)
+#define _TIF_ASCE		(1<<TIF_ASCE)
 #define _TIF_PER_TRAP		(1<<TIF_PER_TRAP)
 #define _TIF_MCCK_PENDING	(1<<TIF_MCCK_PENDING)
 #define _TIF_SYSCALL_TRACE	(1<<TIF_SYSCALL_TRACE)

arch/s390/include/asm/uaccess.h

@@ -92,8 +92,6 @@ static inline unsigned long extable_fixup(const struct exception_table_entry *x)
 #define ARCH_HAS_SORT_EXTABLE
 #define ARCH_HAS_SEARCH_EXTABLE
 
-int __handle_fault(unsigned long, unsigned long, int);
-
 /**
  * __copy_from_user: - Copy a block of data from user space, with less checking.
  * @to:   Destination address, in kernel space.

arch/s390/kernel/asm-offsets.c

@@ -136,6 +136,7 @@ int main(void)
 	DEFINE(__LC_RESTART_FN, offsetof(struct _lowcore, restart_fn));
 	DEFINE(__LC_RESTART_DATA, offsetof(struct _lowcore, restart_data));
 	DEFINE(__LC_RESTART_SOURCE, offsetof(struct _lowcore, restart_source));
+	DEFINE(__LC_KERNEL_ASCE, offsetof(struct _lowcore, kernel_asce));
 	DEFINE(__LC_USER_ASCE, offsetof(struct _lowcore, user_asce));
 	DEFINE(__LC_INT_CLOCK, offsetof(struct _lowcore, int_clock));
 	DEFINE(__LC_MCCK_CLOCK, offsetof(struct _lowcore, mcck_clock));

arch/s390/kernel/entry.S

@@ -38,9 +38,9 @@ __PT_R14     =	__PT_GPRS + 56
 __PT_R15     =	__PT_GPRS + 60
 
 _TIF_WORK_SVC = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
-		 _TIF_MCCK_PENDING | _TIF_PER_TRAP )
+		 _TIF_MCCK_PENDING | _TIF_PER_TRAP | _TIF_ASCE)
 _TIF_WORK_INT = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
-		 _TIF_MCCK_PENDING)
+		 _TIF_MCCK_PENDING | _TIF_ASCE)
 _TIF_TRACE    = (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | _TIF_SECCOMP | \
 		 _TIF_SYSCALL_TRACEPOINT)
 _TIF_TRANSFER = (_TIF_MCCK_PENDING | _TIF_TLB_WAIT)
@@ -241,6 +241,8 @@ sysc_work:
 	jo	sysc_sigpending
 	tm	__TI_flags+3(%r12),_TIF_NOTIFY_RESUME
 	jo	sysc_notify_resume
+	tm	__TI_flags+3(%r12),_TIF_ASCE
+	jo	sysc_uaccess
 	j	sysc_return		# beware of critical section cleanup
 
 #
@@ -259,6 +261,14 @@ sysc_mcck_pending:
 	la	%r14,BASED(sysc_return)
 	br	%r1			# TIF bit will be cleared by handler
 
+#
+# _TIF_ASCE is set, load user space asce
+#
+sysc_uaccess:
+	ni	__TI_flags+3(%r12),255-_TIF_ASCE
+	lctl	%c1,%c1,__LC_USER_ASCE	# load primary asce
+	j	sysc_return
+
 #
 # _TIF_SIGPENDING is set, call do_signal
 #
@@ -522,6 +532,8 @@ io_work_tif:
 	jo	io_sigpending
 	tm	__TI_flags+3(%r12),_TIF_NOTIFY_RESUME
 	jo	io_notify_resume
+	tm	__TI_flags+3(%r12),_TIF_ASCE
+	jo	io_uaccess
 	j	io_return		# beware of critical section cleanup
 
 #
@@ -534,6 +546,14 @@ io_mcck_pending:
 	TRACE_IRQS_OFF
 	j	io_return
 
+#
+# _TIF_ASCE is set, load user space asce
+#
+io_uaccess:
+	ni	__TI_flags+3(%r12),255-_TIF_ASCE
+	lctl	%c1,%c1,__LC_USER_ASCE	# load primary asce
+	j	io_return
+
 #
 # _TIF_NEED_RESCHED is set, call schedule
 #

arch/s390/kernel/entry64.S

@@ -43,9 +43,9 @@ STACK_SIZE  = 1 << STACK_SHIFT
 STACK_INIT = STACK_SIZE - STACK_FRAME_OVERHEAD - __PT_SIZE
 
 _TIF_WORK_SVC = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
-		 _TIF_MCCK_PENDING | _TIF_PER_TRAP )
+		 _TIF_MCCK_PENDING | _TIF_PER_TRAP | _TIF_ASCE)
 _TIF_WORK_INT = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
-		 _TIF_MCCK_PENDING)
+		 _TIF_MCCK_PENDING | _TIF_ASCE)
 _TIF_TRACE    = (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | _TIF_SECCOMP | \
 		 _TIF_SYSCALL_TRACEPOINT)
 _TIF_TRANSFER = (_TIF_MCCK_PENDING | _TIF_TLB_WAIT)
@@ -275,6 +275,8 @@ sysc_work:
 	jo	sysc_sigpending
 	tm	__TI_flags+7(%r12),_TIF_NOTIFY_RESUME
 	jo	sysc_notify_resume
+	tm	__TI_flags+7(%r12),_TIF_ASCE
+	jo	sysc_uaccess
 	j	sysc_return		# beware of critical section cleanup
 
 #
@@ -291,6 +293,14 @@ sysc_mcck_pending:
 	larl	%r14,sysc_return
 	jg	s390_handle_mcck	# TIF bit will be cleared by handler
 
+#
+# _TIF_ASCE is set, load user space asce
+#
+sysc_uaccess:
+	ni	__TI_flags+7(%r12),255-_TIF_ASCE
+	lctlg	%c1,%c1,__LC_USER_ASCE		# load primary asce
+	j	sysc_return
+
 #
 # _TIF_SIGPENDING is set, call do_signal
 #
@@ -559,6 +569,8 @@ io_work_tif:
 	jo	io_sigpending
 	tm	__TI_flags+7(%r12),_TIF_NOTIFY_RESUME
 	jo	io_notify_resume
+	tm	__TI_flags+7(%r12),_TIF_ASCE
+	jo	io_uaccess
 	j	io_return		# beware of critical section cleanup
 
 #
@@ -570,6 +582,14 @@ io_mcck_pending:
 	TRACE_IRQS_OFF
 	j	io_return
 
+#
+# _TIF_ASCE is set, load user space asce
+#
+io_uaccess:
+	ni	__TI_flags+7(%r12),255-_TIF_ASCE
+	lctlg	%c1,%c1,__LC_USER_ASCE		# load primary asce
+	j	io_return
+
 #
 # _TIF_NEED_RESCHED is set, call schedule
 #

arch/s390/lib/Makefile

@@ -2,7 +2,7 @@
 # Makefile for s390-specific library files..
 #
 
-lib-y += delay.o string.o uaccess_pt.o uaccess_mvcos.o find.o
+lib-y += delay.o string.o uaccess.o find.o
 obj-$(CONFIG_32BIT) += div64.o qrnnd.o ucmpdi2.o mem32.o
 obj-$(CONFIG_64BIT) += mem64.o
 lib-$(CONFIG_SMP) += spinlock.o

arch/s390/lib/uaccess.h

-/*
- *    Copyright IBM Corp. 2007
- *
- */
-
-#ifndef __ARCH_S390_LIB_UACCESS_H
-#define __ARCH_S390_LIB_UACCESS_H
-
-unsigned long copy_from_user_pt(void *to, const void __user *from, unsigned long n);
-unsigned long copy_to_user_pt(void __user *to, const void *from, unsigned long n);
-unsigned long copy_in_user_pt(void __user *to, const void __user *from, unsigned long n);
-unsigned long clear_user_pt(void __user *to, unsigned long n);
-unsigned long strnlen_user_pt(const char __user *src, unsigned long count);
-long strncpy_from_user_pt(char *dst, const char __user *src, long count);
-
-#endif /* __ARCH_S390_LIB_UACCESS_H */

arch/s390/mm/fault.c

@@ -105,21 +105,24 @@ void bust_spinlocks(int yes)
  * Returns the address space associated with the fault.
  * Returns 0 for kernel space and 1 for user space.
  */
-static inline int user_space_fault(unsigned long trans_exc_code)
+static inline int user_space_fault(struct pt_regs *regs)
 {
+	unsigned long trans_exc_code;
+
 	/*
 	 * The lowest two bits of the translation exception
 	 * identification indicate which paging table was used.
 	 */
-	trans_exc_code &= 3;
-	if (trans_exc_code == 2)
-		/* Access via secondary space, set_fs setting decides */
+	trans_exc_code = regs->int_parm_long & 3;
+	if (trans_exc_code == 3) /* home space -> kernel */
+		return 0;
+	if (user_mode(regs))
+		return 1;
+	if (trans_exc_code == 2) /* secondary space -> set_fs */
 		return current->thread.mm_segment.ar4;
-	/*
-	 * Access via primary space or access register is from user space
-	 * and access via home space is from the kernel.
-	 */
-	return trans_exc_code != 3;
+	if (current->flags & PF_VCPU)
+		return 1;
+	return 0;
 }
 
 static inline void report_user_fault(struct pt_regs *regs, long signr)
@@ -171,7 +174,7 @@ static noinline void do_no_context(struct pt_regs *regs)
 	 * terminate things with extreme prejudice.
 	 */
 	address = regs->int_parm_long & __FAIL_ADDR_MASK;
-	if (!user_space_fault(regs->int_parm_long))
+	if (!user_space_fault(regs))
 		printk(KERN_ALERT "Unable to handle kernel pointer dereference"
 		       " at virtual kernel address %p\n", (void *)address);
 	else
@@ -291,7 +294,7 @@ static inline int do_exception(struct pt_regs *regs, int access)
 	 * user context.
 	 */
 	fault = VM_FAULT_BADCONTEXT;
-	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
+	if (unlikely(!user_space_fault(regs) || in_atomic() || !mm))
 		goto out;
 
 	address = trans_exc_code & __FAIL_ADDR_MASK;
@@ -423,30 +426,6 @@ void __kprobes do_dat_exception(struct pt_regs *regs)
 		do_fault_error(regs, fault);
 }
 
-int __handle_fault(unsigned long uaddr, unsigned long pgm_int_code, int write)
-{
-	struct pt_regs regs;
-	int access, fault;
-
-	/* Emulate a uaccess fault from kernel mode. */
-	regs.psw.mask = PSW_KERNEL_BITS | PSW_MASK_DAT | PSW_MASK_MCHECK;
-	if (!irqs_disabled())
-		regs.psw.mask |= PSW_MASK_IO | PSW_MASK_EXT;
-	regs.psw.addr = (unsigned long) __builtin_return_address(0);
-	regs.psw.addr |= PSW_ADDR_AMODE;
-	regs.int_code = pgm_int_code;
-	regs.int_parm_long = (uaddr & PAGE_MASK) | 2;
-	access = write ? VM_WRITE : VM_READ;
-	fault = do_exception(&regs, access);
-	/*
-	 * Since the fault happened in kernel mode while performing a uaccess
-	 * all we need to do now is emulating a fixup in case "fault" is not
-	 * zero.
-	 * For the calling uaccess functions this results always in -EFAULT.
-	 */
-	return fault ? -EFAULT : 0;
-}
-
 #ifdef CONFIG_PFAULT 
 /*
  * 'pfault' pseudo page faults routines.

arch/s390/mm/pgtable.c

@@ -54,7 +54,7 @@ static void __crst_table_upgrade(void *arg)
 	struct mm_struct *mm = arg;
 
 	if (current->active_mm == mm)
-		update_user_asce(mm);
+		update_user_asce(mm, 1);
 	__tlb_flush_local();
 }
 
@@ -108,7 +108,7 @@ void crst_table_downgrade(struct mm_struct *mm, unsigned long limit)
 	pgd_t *pgd;
 
 	if (current->active_mm == mm) {
-		clear_user_asce(mm);
+		clear_user_asce(mm, 1);
 		__tlb_flush_mm(mm);
 	}
 	while (mm->context.asce_limit > limit) {
@@ -134,7 +134,7 @@ void crst_table_downgrade(struct mm_struct *mm, unsigned long limit)
 		crst_table_free(mm, (unsigned long *) pgd);
 	}
 	if (current->active_mm == mm)
-		update_user_asce(mm);
+		update_user_asce(mm, 1);
 }
 #endif