[S390] cleanup lowcore access from program checks

Read all required fields for program checks from the lowcore in the
first level interrupt handler in entry[64].S. If the context that
caused the fault was enabled for interrupts we can now re-enable the
irqs in entry[64].S.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/kernel/asm-offsets.c

@@ -84,6 +84,7 @@ int main(void)
 	DEFINE(__LC_SVC_INT_CODE, offsetof(struct _lowcore, svc_code));
 	DEFINE(__LC_PGM_ILC, offsetof(struct _lowcore, pgm_ilc));
 	DEFINE(__LC_PGM_INT_CODE, offsetof(struct _lowcore, pgm_code));
+	DEFINE(__LC_TRANS_EXC_CODE, offsetof(struct _lowcore, trans_exc_code));
 	DEFINE(__LC_PER_ATMID, offsetof(struct _lowcore, per_perc_atmid));
 	DEFINE(__LC_PER_ADDRESS, offsetof(struct _lowcore, per_address));
 	DEFINE(__LC_PER_ACCESS_ID, offsetof(struct _lowcore, per_access_id));

arch/s390/kernel/entry.S

@@ -72,25 +72,9 @@ STACK_SIZE  = 1 << STACK_SHIFT
 	l	%r1,BASED(.Ltrace_irq_off_caller)
 	basr	%r14,%r1
 	.endm
-
-	.macro	TRACE_IRQS_CHECK_ON
-	tm	SP_PSW(%r15),0x03	# irqs enabled?
-	bz	BASED(0f)
-	TRACE_IRQS_ON
-0:
-	.endm
-
-	.macro	TRACE_IRQS_CHECK_OFF
-	tm	SP_PSW(%r15),0x03	# irqs enabled?
-	bz	BASED(0f)
-	TRACE_IRQS_OFF
-0:
-	.endm
 #else
 #define TRACE_IRQS_ON
 #define TRACE_IRQS_OFF
-#define TRACE_IRQS_CHECK_ON
-#define TRACE_IRQS_CHECK_OFF
 #endif
 
 #ifdef CONFIG_LOCKDEP
@@ -198,6 +182,12 @@ STACK_SIZE  = 1 << STACK_SHIFT
 	lpsw	\psworg			# back to caller
 	.endm
 
+	.macro REENABLE_IRQS
+	mvc	__SF_EMPTY(1,%r15),SP_PSW(%r15)
+	ni	__SF_EMPTY(%r15),0xbf
+	ssm	__SF_EMPTY(%r15)
+	.endm
+
 /*
  * Scheduler resume function, called by switch_to
  *  gpr2 = (task_struct *) prev
@@ -440,13 +430,11 @@ kernel_execve:
 	br	%r14
 	# execve succeeded.
 0:	stnsm	__SF_EMPTY(%r15),0xfc	# disable interrupts
-	TRACE_IRQS_OFF
 	l	%r15,__LC_KERNEL_STACK	# load ksp
 	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
 	l	%r9,__LC_THREAD_INFO
 	mvc	SP_PTREGS(__PT_SIZE,%r15),0(%r12)	# copy pt_regs
 	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
-	TRACE_IRQS_ON
 	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
 	l	%r1,BASED(.Lexecve_tail)
 	basr	%r14,%r1
@@ -483,9 +471,10 @@ pgm_check_handler:
 	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 pgm_no_vtime:
-	TRACE_IRQS_CHECK_OFF
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	l	%r3,__LC_PGM_ILC	# load program interruption code
+	l	%r4,__LC_TRANS_EXC_CODE
+	REENABLE_IRQS
 	la	%r8,0x7f
 	nr	%r8,%r3
 pgm_do_call:
@@ -495,7 +484,6 @@ pgm_do_call:
 	la	%r2,SP_PTREGS(%r15)	# address of register-save area
 	basr	%r14,%r7		# branch to interrupt-handler
 pgm_exit:
-	TRACE_IRQS_CHECK_ON
 	b	BASED(sysc_return)
 
 #
@@ -523,7 +511,6 @@ pgm_per_std:
 	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 pgm_no_vtime2:
-	TRACE_IRQS_CHECK_OFF
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	l	%r1,__TI_task(%r9)
 	tm	SP_PSW+1(%r15),0x01	# kernel per event ?
@@ -533,6 +520,8 @@ pgm_no_vtime2:
 	mvc	__THREAD_per+__PER_access_id(1,%r1),__LC_PER_ACCESS_ID
 	oi	__TI_flags+3(%r9),_TIF_SINGLE_STEP # set TIF_SINGLE_STEP
 	l	%r3,__LC_PGM_ILC	# load program interruption code
+	l	%r4,__LC_TRANS_EXC_CODE
+	REENABLE_IRQS
 	la	%r8,0x7f
 	nr	%r8,%r3 		# clear per-event-bit and ilc
 	be	BASED(pgm_exit2)	# only per or per+check ?
@@ -542,8 +531,6 @@ pgm_no_vtime2:
 	la	%r2,SP_PTREGS(%r15)	# address of register-save area
 	basr	%r14,%r7		# branch to interrupt-handler
 pgm_exit2:
-	TRACE_IRQS_ON
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
 	b	BASED(sysc_return)
 
 #
@@ -557,13 +544,11 @@ pgm_svcper:
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 	lh	%r7,0x8a		# get svc number from lowcore
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
-	TRACE_IRQS_OFF
 	l	%r8,__TI_task(%r9)
 	mvc	__THREAD_per+__PER_atmid(2,%r8),__LC_PER_ATMID
 	mvc	__THREAD_per+__PER_address(4,%r8),__LC_PER_ADDRESS
 	mvc	__THREAD_per+__PER_access_id(1,%r8),__LC_PER_ACCESS_ID
 	oi	__TI_flags+3(%r9),_TIF_SINGLE_STEP # set TIF_SINGLE_STEP
-	TRACE_IRQS_ON
 	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
 	lm	%r2,%r6,SP_R2(%r15)	# load svc arguments
 	b	BASED(sysc_do_svc)

arch/s390/kernel/entry.h

@@ -5,7 +5,7 @@
 #include <linux/signal.h>
 #include <asm/ptrace.h>
 
-typedef void pgm_check_handler_t(struct pt_regs *, long);
+typedef void pgm_check_handler_t(struct pt_regs *, long, unsigned long);
 extern pgm_check_handler_t *pgm_check_table[128];
 pgm_check_handler_t do_protection_exception;
 pgm_check_handler_t do_dat_exception;

arch/s390/kernel/entry64.S

@@ -79,25 +79,9 @@ _TIF_SYSCALL = (_TIF_SYSCALL_TRACE>>8 | _TIF_SYSCALL_AUDIT>>8 | \
 	basr	%r2,%r0
 	brasl	%r14,trace_hardirqs_off_caller
 	.endm
-
-	.macro TRACE_IRQS_CHECK_ON
-	tm	SP_PSW(%r15),0x03	# irqs enabled?
-	jz	0f
-	TRACE_IRQS_ON
-0:
-	.endm
-
-	.macro TRACE_IRQS_CHECK_OFF
-	tm	SP_PSW(%r15),0x03	# irqs enabled?
-	jz	0f
-	TRACE_IRQS_OFF
-0:
-	.endm
 #else
 #define TRACE_IRQS_ON
 #define TRACE_IRQS_OFF
-#define TRACE_IRQS_CHECK_ON
-#define TRACE_IRQS_CHECK_OFF
 #endif
 
 #ifdef CONFIG_LOCKDEP
@@ -207,6 +191,12 @@ _TIF_SYSCALL = (_TIF_SYSCALL_TRACE>>8 | _TIF_SYSCALL_AUDIT>>8 | \
 0:
 	.endm
 
+	.macro REENABLE_IRQS
+	mvc	__SF_EMPTY(1,%r15),SP_PSW(%r15)
+	ni	__SF_EMPTY(%r15),0xbf
+	ssm	__SF_EMPTY(%r15)
+	.endm
+
 /*
  * Scheduler resume function, called by switch_to
  *  gpr2 = (task_struct *) prev
@@ -443,14 +433,12 @@ kernel_execve:
 	br	%r14
 	# execve succeeded.
 0:	stnsm	__SF_EMPTY(%r15),0xfc	# disable interrupts
-#	TRACE_IRQS_OFF
 	lg	%r15,__LC_KERNEL_STACK	# load ksp
 	aghi	%r15,-SP_SIZE		# make room for registers & psw
 	lg	%r13,__LC_SVC_NEW_PSW+8
 	mvc	SP_PTREGS(__PT_SIZE,%r15),0(%r12)	# copy pt_regs
 	lg	%r12,__LC_THREAD_INFO
 	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
-#	TRACE_IRQS_ON
 	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
 	brasl	%r14,execve_tail
 	j	sysc_return
@@ -490,19 +478,18 @@ pgm_check_handler:
 	LAST_BREAK
 pgm_no_vtime:
 	HANDLE_SIE_INTERCEPT
-	TRACE_IRQS_CHECK_OFF
 	stg	%r11,SP_ARGS(%r15)
 	lgf	%r3,__LC_PGM_ILC	# load program interruption code
+	lg	%r4,__LC_TRANS_EXC_CODE
+	REENABLE_IRQS
 	lghi	%r8,0x7f
 	ngr	%r8,%r3
-pgm_do_call:
 	sll	%r8,3
 	larl	%r1,pgm_check_table
 	lg	%r1,0(%r8,%r1)		# load address of handler routine
 	la	%r2,SP_PTREGS(%r15)	# address of register-save area
 	basr	%r14,%r1		# branch to interrupt-handler
 pgm_exit:
-	TRACE_IRQS_CHECK_ON
 	j	sysc_return
 
 #
@@ -533,7 +520,6 @@ pgm_per_std:
 	LAST_BREAK
 pgm_no_vtime2:
 	HANDLE_SIE_INTERCEPT
-	TRACE_IRQS_CHECK_OFF
 	lg	%r1,__TI_task(%r12)
 	tm	SP_PSW+1(%r15),0x01	# kernel per event ?
 	jz	kernel_per
@@ -542,6 +528,8 @@ pgm_no_vtime2:
 	mvc	__THREAD_per+__PER_access_id(1,%r1),__LC_PER_ACCESS_ID
 	oi	__TI_flags+7(%r12),_TIF_SINGLE_STEP # set TIF_SINGLE_STEP
 	lgf	%r3,__LC_PGM_ILC	# load program interruption code
+	lg	%r4,__LC_TRANS_EXC_CODE
+	REENABLE_IRQS
 	lghi	%r8,0x7f
 	ngr	%r8,%r3			# clear per-event-bit and ilc
 	je	pgm_exit2
@@ -551,8 +539,6 @@ pgm_no_vtime2:
 	la	%r2,SP_PTREGS(%r15)	# address of register-save area
 	basr	%r14,%r1		# branch to interrupt-handler
 pgm_exit2:
-	TRACE_IRQS_ON
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
 	j	sysc_return
 
 #
@@ -568,13 +554,11 @@ pgm_svcper:
 	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 	LAST_BREAK
-	TRACE_IRQS_OFF
 	lg	%r8,__TI_task(%r12)
 	mvc	__THREAD_per+__PER_atmid(2,%r8),__LC_PER_ATMID
 	mvc	__THREAD_per+__PER_address(8,%r8),__LC_PER_ADDRESS
 	mvc	__THREAD_per+__PER_access_id(1,%r8),__LC_PER_ACCESS_ID
 	oi	__TI_flags+7(%r12),_TIF_SINGLE_STEP # set TIF_SINGLE_STEP
-	TRACE_IRQS_ON
 	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
 	lmg	%r2,%r6,SP_R2(%r15)	# load svc arguments
 	j	sysc_do_svc

arch/s390/mm/fault.c

@@ -333,12 +333,6 @@ static inline int do_exception(struct pt_regs *regs, int access,
 		goto out;
 
 	address = trans_exc_code & __FAIL_ADDR_MASK;
-	/*
-	 * When we get here, the fault happened in the current
-	 * task's user address space, so we can switch on the
-	 * interrupts again and then search the VMAs
-	 */
-	local_irq_enable();
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
 	down_read(&mm->mmap_sem);
 
@@ -397,20 +391,20 @@ static inline int do_exception(struct pt_regs *regs, int access,
 	return fault;
 }
 
-void __kprobes do_protection_exception(struct pt_regs *regs, long int_code)
+void __kprobes do_protection_exception(struct pt_regs *regs, long pgm_int_code,
+				       unsigned long trans_exc_code)
 {
-	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
 	int fault;
 
 	/* Protection exception is supressing, decrement psw address. */
-	regs->psw.addr -= (int_code >> 16);
+	regs->psw.addr -= (pgm_int_code >> 16);
 	/*
 	 * Check for low-address protection.  This needs to be treated
 	 * as a special case because the translation exception code
 	 * field is not guaranteed to contain valid data in this case.
 	 */
 	if (unlikely(!(trans_exc_code & 4))) {
-		do_low_address(regs, int_code, trans_exc_code);
+		do_low_address(regs, pgm_int_code, trans_exc_code);
 		return;
 	}
 	fault = do_exception(regs, VM_WRITE, trans_exc_code);
@@ -418,9 +412,9 @@ void __kprobes do_protection_exception(struct pt_regs *regs, long int_code)
 		do_fault_error(regs, 4, trans_exc_code, fault);
 }
 
-void __kprobes do_dat_exception(struct pt_regs *regs, long int_code)
+void __kprobes do_dat_exception(struct pt_regs *regs, long pgm_int_code,
+				unsigned long trans_exc_code)
 {
-	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
 	int access, fault;
 
 	access = VM_READ | VM_EXEC | VM_WRITE;
@@ -431,21 +425,19 @@ void __kprobes do_dat_exception(struct pt_regs *regs, long int_code)
 #endif
 	fault = do_exception(regs, access, trans_exc_code);
 	if (unlikely(fault))
-		do_fault_error(regs, int_code & 255, trans_exc_code, fault);
+		do_fault_error(regs, pgm_int_code & 255, trans_exc_code, fault);
 }
 
 #ifdef CONFIG_64BIT
-void __kprobes do_asce_exception(struct pt_regs *regs, long int_code)
+void __kprobes do_asce_exception(struct pt_regs *regs, long pgm_int_code,
+				 unsigned long trans_exc_code)
 {
-	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
 
 	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
 		goto no_context;
 
-	local_irq_enable();
-
 	down_read(&mm->mmap_sem);
 	vma = find_vma(mm, trans_exc_code & __FAIL_ADDR_MASK);
 	up_read(&mm->mmap_sem);
@@ -457,16 +449,16 @@ void __kprobes do_asce_exception(struct pt_regs *regs, long int_code)
 
 	/* User mode accesses just cause a SIGSEGV */
 	if (regs->psw.mask & PSW_MASK_PSTATE) {
-		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
+		do_sigsegv(regs, pgm_int_code, SEGV_MAPERR, trans_exc_code);
 		return;
 	}
 
 no_context:
-	do_no_context(regs, int_code, trans_exc_code);
+	do_no_context(regs, pgm_int_code, trans_exc_code);
 }
 #endif
 
-int __handle_fault(unsigned long uaddr, unsigned long int_code, int write_user)
+int __handle_fault(unsigned long uaddr, unsigned long pgm_int_code, int write)
 {
 	struct pt_regs regs;
 	int access, fault;
@@ -477,14 +469,14 @@ int __handle_fault(unsigned long uaddr, unsigned long int_code, int write_user)
 	regs.psw.addr = (unsigned long) __builtin_return_address(0);
 	regs.psw.addr |= PSW_ADDR_AMODE;
 	uaddr &= PAGE_MASK;
-	access = write_user ? VM_WRITE : VM_READ;
+	access = write ? VM_WRITE : VM_READ;
 	fault = do_exception(&regs, access, uaddr | 2);
 	if (unlikely(fault)) {
 		if (fault & VM_FAULT_OOM) {
 			pagefault_out_of_memory();
 			fault = 0;
 		} else if (fault & VM_FAULT_SIGBUS)
-			do_sigbus(&regs, int_code, uaddr);
+			do_sigbus(&regs, pgm_int_code, uaddr);
 	}
 	return fault ? -EFAULT : 0;
 }