s390/entry: add CIF_SIE flag and remove sie64a() address check

When a program check, interrupt or machine check is triggered, the
PSW address is compared to a certain range of the sie64a() function
to figure out whether SIE was interrupted and a cleanup of SIE is
needed.

This doesn't work with kprobes: If kprobes probes an instruction, it
copies the instruction to the kprobes instruction page and overwrites the
original instruction with an undefind instruction (Opcode 00). When this
instruction is hit later, kprobes single-steps the instruction on the
kprobes_instruction page.

However, if this instruction is a relative branch instruction it will now
point to a different location in memory due to being moved to the kprobes
instruction page. If the new branch target points into sie64a() the kernel
assumes it interrupted SIE when processing the breakpoint and will crash
trying to access the SIE control block.

Instead of comparing the address, introduce a new CIF_SIE flag which
indicates whether SIE was interrupted.

Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Suggested-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

2 files changed, 16 insertions, 3 deletions

diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h
index 7cf00cf8fb0b..db9982f0e8cd 100644
--- a/arch/s390/include/asm/processor.h
+++ b/arch/s390/include/asm/processor.h
@@ -14,11 +14,13 @@
 
 #include <linux/bits.h>
 
+#define CIF_SIE			0	/* CPU needs SIE exit cleanup */
 #define CIF_NOHZ_DELAY		2	/* delay HZ disable for a tick */
 #define CIF_ENABLED_WAIT	5	/* in enabled wait state */
 #define CIF_MCCK_GUEST		6	/* machine check happening in guest */
 #define CIF_DEDICATED_CPU	7	/* this CPU is dedicated */
 
+#define _CIF_SIE		BIT(CIF_SIE)
 #define _CIF_NOHZ_DELAY		BIT(CIF_NOHZ_DELAY)
 #define _CIF_ENABLED_WAIT	BIT(CIF_ENABLED_WAIT)
 #define _CIF_MCCK_GUEST		BIT(CIF_MCCK_GUEST)
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index fc5277eab554..4e0ff79ffee3 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -146,6 +146,7 @@ _LPP_OFFSET	= __LC_LPP
 	lg	%r9,__SF_SIE_CONTROL(%r15)	# get control block pointer
 	ni	__SIE_PROG0C+3(%r9),0xfe	# no longer in SIE
 	lctlg	%c1,%c1,__LC_KERNEL_ASCE	# load primary asce
+	ni	__LC_CPU_FLAGS+7,255-_CIF_SIE
 	larl	%r9,sie_exit			# skip forward to sie_exit
 	.endm
 #endif
@@ -214,6 +215,7 @@ SYM_FUNC_START(__sie64a)
 	lg	%r14,__LC_GMAP			# get gmap pointer
 	ltgr	%r14,%r14
 	jz	.Lsie_gmap
+	oi	__LC_CPU_FLAGS+7,_CIF_SIE
 	lctlg	%c1,%c1,__GMAP_ASCE(%r14)	# load primary asce
 .Lsie_gmap:
 	lg	%r14,__SF_SIE_CONTROL(%r15)	# get control block pointer
@@ -235,6 +237,7 @@ SYM_FUNC_START(__sie64a)
 	ni	__SIE_PROG0C+3(%r14),0xfe	# no longer in SIE
 	lctlg	%c1,%c1,__LC_KERNEL_ASCE	# load primary asce
 .Lsie_done:
+	ni	__LC_CPU_FLAGS+7,255-_CIF_SIE
 # some program checks are suppressing. C code (e.g. do_protection_exception)
 # will rewind the PSW by the ILC, which is often 4 bytes in case of SIE. There
 # are some corner cases (e.g. runtime instrumentation) where ILC is unpredictable.
@@ -346,7 +349,8 @@ SYM_CODE_START(pgm_check_handler)
 .Lpgm_skip_asce:
 #if IS_ENABLED(CONFIG_KVM)
 	# cleanup critical section for program checks in __sie64a
-	OUTSIDE	%r9,.Lsie_gmap,.Lsie_done,1f
+	TSTMSK	__LC_CPU_FLAGS,_CIF_SIE
+	jz	1f
 	BPENTER	__SF_SIE_FLAGS(%r15),_TIF_ISOLATE_BP_GUEST
 	SIEEXIT
 	lghi	%r10,_PIF_GUEST_FAULT
@@ -416,7 +420,8 @@ SYM_CODE_START(\name)
 	tmhh	%r8,0x0001			# interrupting from user ?
 	jnz	1f
 #if IS_ENABLED(CONFIG_KVM)
-	OUTSIDE	%r9,.Lsie_gmap,.Lsie_done,0f
+	TSTMSK	__LC_CPU_FLAGS,_CIF_SIE
+	jz	0f
 	BPENTER	__SF_SIE_FLAGS(%r15),_TIF_ISOLATE_BP_GUEST
 	SIEEXIT
 #endif
@@ -513,7 +518,13 @@ SYM_CODE_START(mcck_int_handler)
 	TSTMSK	__LC_MCCK_CODE,MCCK_CODE_PSW_IA_VALID
 	jno	.Lmcck_panic
 #if IS_ENABLED(CONFIG_KVM)
-	OUTSIDE	%r9,.Lsie_gmap,.Lsie_done,.Lmcck_user
+	TSTMSK	__LC_CPU_FLAGS,_CIF_SIE
+	jz	.Lmcck_user
+	# Need to compare the address instead of a CIF_SIE* flag.
+	# Otherwise there would be a race between setting the flag
+	# and entering SIE (or leaving and clearing the flag). This
+	# would cause machine checks targeted at the guest to be
+	# handled by the host.
 	OUTSIDE	%r9,.Lsie_entry,.Lsie_leave,4f
 	oi	__LC_CPU_FLAGS+7, _CIF_MCCK_GUEST
 4:	BPENTER	__SF_SIE_FLAGS(%r15),_TIF_ISOLATE_BP_GUEST