s390: add stack for machine check handler

The previous code used the normal kernel stack for machine checks.
This is problematic when a machine check interrupts a system call
or interrupt handler right at the beginning where registers are set up.

Assume system_call is interrupted at the first instruction and a machine
check is triggered. The machine check handler is called, checks the PSW
to see whether it is coming from user space, notices that it is already
in kernel mode but %r15 still contains the user space stack. This would
lead to a kernel crash.

There are basically two ways of fixing that: Either using the 'critical
cleanup' approach which compares the address in the PSW to see whether
it is already at a point where the stack has been set up, or use an extra
stack for the machine check handler.

For simplicity, we will go with the second approach and allocate an extra
stack. This adds some memory overhead for large systems, but usually large
system have plenty of memory so this isn't really a concern. But it keeps
the mchk stack setup simple and less error prone.

Fixes: 0b0ed657fe ("s390: remove critical section cleanup from entry.S")
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Cc: <stable@kernel.org> # v5.8+
Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>

arch/s390/include/asm/lowcore.h

@@ -107,16 +107,15 @@ struct lowcore {
 	__u64	async_stack;			/* 0x0350 */
 	__u64	nodat_stack;			/* 0x0358 */
 	__u64	restart_stack;			/* 0x0360 */
-
+	__u64	mcck_stack;			/* 0x0368 */
 	/* Restart function and parameter. */
-	__u64	restart_fn;			/* 0x0368 */
-	__u64	restart_data;			/* 0x0370 */
-	__u64	restart_source;			/* 0x0378 */
+	__u64	restart_fn;			/* 0x0370 */
+	__u64	restart_data;			/* 0x0378 */
+	__u64	restart_source;			/* 0x0380 */
 
 	/* Address space pointer. */
-	__u64	kernel_asce;			/* 0x0380 */
-	__u64	user_asce;			/* 0x0388 */
-	__u8	pad_0x0390[0x0398-0x0390];	/* 0x0390 */
+	__u64	kernel_asce;			/* 0x0388 */
+	__u64	user_asce;			/* 0x0390 */
 
 	/*
 	 * The lpp and current_pid fields form a

arch/s390/kernel/asm-offsets.c

@@ -118,6 +118,7 @@ int main(void)
 	OFFSET(__LC_ASYNC_STACK, lowcore, async_stack);
 	OFFSET(__LC_NODAT_STACK, lowcore, nodat_stack);
 	OFFSET(__LC_RESTART_STACK, lowcore, restart_stack);
+	OFFSET(__LC_MCCK_STACK, lowcore, mcck_stack);
 	OFFSET(__LC_RESTART_FN, lowcore, restart_fn);
 	OFFSET(__LC_RESTART_DATA, lowcore, restart_data);
 	OFFSET(__LC_RESTART_SOURCE, lowcore, restart_source);

arch/s390/kernel/entry.S

@@ -70,6 +70,8 @@ _LPP_OFFSET	= __LC_LPP
 	je	\oklabel
 	clg	%r14,__LC_ASYNC_STACK
 	je	\oklabel
+	clg	%r14,__LC_MCCK_STACK
+	je	\oklabel
 	clg	%r14,__LC_NODAT_STACK
 	je	\oklabel
 	clg	%r14,__LC_RESTART_STACK
@@ -548,20 +550,16 @@ ENTRY(mcck_int_handler)
 	jhe	.Lmcck_stack
 	lghi	%r11,__LC_GPREGS_SAVE_AREA+64	# inside critical section, do cleanup
 	brasl	%r14,.Lcleanup_sie
-.Lmcck_stack:
 #endif
-	CHECK_STACK __LC_GPREGS_SAVE_AREA+64
-	lgr	%r11,%r15
-	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
-	stg	%r11,__SF_BACKCHAIN(%r15)
-	j	5f
+	j	.Lmcck_stack
 .Lmcck_user:
 	BPENTER __TI_flags(%r12),_TIF_ISOLATE_BP
-	lctlg	%c1,%c1,__LC_KERNEL_ASCE
-	lg	%r15,__LC_KERNEL_STACK
-	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
-5:	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+.Lmcck_stack:
+	lg	%r15,__LC_MCCK_STACK
 .Lmcck_skip:
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	lctlg	%c1,%c1,__LC_KERNEL_ASCE
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
 	lghi	%r14,__LC_GPREGS_SAVE_AREA+64
 	stmg	%r0,%r7,__PT_R0(%r11)
 	# clear user controlled registers to prevent speculative use
@@ -602,7 +600,6 @@ ENTRY(mcck_int_handler)
 
 .Lmcck_panic:
 	lg	%r15,__LC_NODAT_STACK
-	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	j	.Lmcck_skip
 ENDPROC(mcck_int_handler)
 

arch/s390/kernel/setup.c

@@ -338,7 +338,7 @@ int __init arch_early_irq_init(void)
 	return 0;
 }
 
-static int __init async_stack_realloc(void)
+static int __init stack_realloc(void)
 {
 	unsigned long old, new;
 
@@ -348,9 +348,16 @@ static int __init async_stack_realloc(void)
 		panic("Couldn't allocate async stack");
 	WRITE_ONCE(S390_lowcore.async_stack, new + STACK_INIT_OFFSET);
 	free_pages(old, THREAD_SIZE_ORDER);
+
+	old = S390_lowcore.mcck_stack - STACK_INIT_OFFSET;
+	new = stack_alloc();
+	if (!new)
+		panic("Couldn't allocate machine check stack");
+	WRITE_ONCE(S390_lowcore.mcck_stack, new + STACK_INIT_OFFSET);
+	memblock_free(old, THREAD_SIZE);
 	return 0;
 }
-early_initcall(async_stack_realloc);
+early_initcall(stack_realloc);
 
 void __init arch_call_rest_init(void)
 {
@@ -372,6 +379,7 @@ void __init arch_call_rest_init(void)
 static void __init setup_lowcore_dat_off(void)
 {
 	unsigned long int_psw_mask = PSW_KERNEL_BITS;
+	unsigned long mcck_stack;
 	struct lowcore *lc;
 
 	if (IS_ENABLED(CONFIG_KASAN))
@@ -439,6 +447,12 @@ static void __init setup_lowcore_dat_off(void)
 	lc->restart_data = 0;
 	lc->restart_source = -1UL;
 
+	mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
+	if (!mcck_stack)
+		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
+		      __func__, THREAD_SIZE, THREAD_SIZE);
+	lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
+
 	/* Setup absolute zero lowcore */
 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);

arch/s390/kernel/smp.c

@@ -189,7 +189,7 @@ static void pcpu_ec_call(struct pcpu *pcpu, int ec_bit)
 
 static int pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu)
 {
-	unsigned long async_stack, nodat_stack;
+	unsigned long async_stack, nodat_stack, mcck_stack;
 	struct lowcore *lc;
 
 	if (pcpu != &pcpu_devices[0]) {
@@ -202,13 +202,15 @@ static int pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu)
 		nodat_stack = pcpu->lowcore->nodat_stack - STACK_INIT_OFFSET;
 	}
 	async_stack = stack_alloc();
-	if (!async_stack)
-		goto out;
+	mcck_stack = stack_alloc();
+	if (!async_stack || !mcck_stack)
+		goto out_stack;
 	lc = pcpu->lowcore;
 	memcpy(lc, &S390_lowcore, 512);
 	memset((char *) lc + 512, 0, sizeof(*lc) - 512);
 	lc->async_stack = async_stack + STACK_INIT_OFFSET;
 	lc->nodat_stack = nodat_stack + STACK_INIT_OFFSET;
+	lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
 	lc->cpu_nr = cpu;
 	lc->spinlock_lockval = arch_spin_lockval(cpu);
 	lc->spinlock_index = 0;
@@ -216,12 +218,13 @@ static int pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu)
 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 	if (nmi_alloc_per_cpu(lc))
-		goto out_async;
+		goto out_stack;
 	lowcore_ptr[cpu] = lc;
 	pcpu_sigp_retry(pcpu, SIGP_SET_PREFIX, (u32)(unsigned long) lc);
 	return 0;
 
-out_async:
+out_stack:
+	stack_free(mcck_stack);
 	stack_free(async_stack);
 out:
 	if (pcpu != &pcpu_devices[0]) {
@@ -233,16 +236,18 @@ static int pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu)
 
 static void pcpu_free_lowcore(struct pcpu *pcpu)
 {
-	unsigned long async_stack, nodat_stack, lowcore;
+	unsigned long async_stack, nodat_stack, mcck_stack, lowcore;
 
 	nodat_stack = pcpu->lowcore->nodat_stack - STACK_INIT_OFFSET;
 	async_stack = pcpu->lowcore->async_stack - STACK_INIT_OFFSET;
+	mcck_stack = pcpu->lowcore->mcck_stack - STACK_INIT_OFFSET;
 	lowcore = (unsigned long) pcpu->lowcore;
 
 	pcpu_sigp_retry(pcpu, SIGP_SET_PREFIX, 0);
 	lowcore_ptr[pcpu - pcpu_devices] = NULL;
 	nmi_free_per_cpu(pcpu->lowcore);
 	stack_free(async_stack);
+	stack_free(mcck_stack);
 	if (pcpu == &pcpu_devices[0])
 		return;
 	free_pages(nodat_stack, THREAD_SIZE_ORDER);