aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--Documentation/virtual/kvm/api.txt8
-rw-r--r--Documentation/virtual/kvm/devices/vm.txt26
-rw-r--r--Documentation/virtual/kvm/s390-diag.txt2
-rw-r--r--arch/s390/include/asm/ctl_reg.h14
-rw-r--r--arch/s390/include/asm/kvm_host.h149
-rw-r--r--arch/s390/include/asm/lowcore.h10
-rw-r--r--arch/s390/include/asm/mmu.h2
-rw-r--r--arch/s390/include/asm/mmu_context.h1
-rw-r--r--arch/s390/include/asm/pgalloc.h3
-rw-r--r--arch/s390/include/asm/pgtable.h169
-rw-r--r--arch/s390/include/asm/ptrace.h44
-rw-r--r--arch/s390/include/asm/sclp.h7
-rw-r--r--arch/s390/include/uapi/asm/kvm.h28
-rw-r--r--arch/s390/kernel/asm-offsets.c14
-rw-r--r--arch/s390/kernel/entry.S4
-rw-r--r--arch/s390/kernel/entry64.S4
-rw-r--r--arch/s390/kvm/Makefile4
-rw-r--r--arch/s390/kvm/diag.c15
-rw-r--r--arch/s390/kvm/gaccess.c645
-rw-r--r--arch/s390/kvm/gaccess.h373
-rw-r--r--arch/s390/kvm/guestdbg.c481
-rw-r--r--arch/s390/kvm/intercept.c110
-rw-r--r--arch/s390/kvm/interrupt.c339
-rw-r--r--arch/s390/kvm/kvm-s390.c373
-rw-r--r--arch/s390/kvm/kvm-s390.h59
-rw-r--r--arch/s390/kvm/priv.c272
-rw-r--r--arch/s390/kvm/sigp.c12
-rw-r--r--arch/s390/kvm/trace.h39
-rw-r--r--arch/s390/mm/pgtable.c93
-rw-r--r--drivers/s390/char/sclp_early.c22
-rw-r--r--include/linux/kvm_host.h7
-rw-r--r--include/uapi/linux/kvm.h1
-rw-r--r--virt/kvm/kvm_main.c2
33 files changed, 2831 insertions, 501 deletions
diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
index a9380ba54c8e..2014ff12b492 100644
--- a/Documentation/virtual/kvm/api.txt
+++ b/Documentation/virtual/kvm/api.txt
@@ -2314,8 +2314,8 @@ struct kvm_create_device {
2314 2314
23154.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR 23154.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
2316 2316
2317Capability: KVM_CAP_DEVICE_CTRL 2317Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device
2318Type: device ioctl 2318Type: device ioctl, vm ioctl
2319Parameters: struct kvm_device_attr 2319Parameters: struct kvm_device_attr
2320Returns: 0 on success, -1 on error 2320Returns: 0 on success, -1 on error
2321Errors: 2321Errors:
@@ -2340,8 +2340,8 @@ struct kvm_device_attr {
2340 2340
23414.81 KVM_HAS_DEVICE_ATTR 23414.81 KVM_HAS_DEVICE_ATTR
2342 2342
2343Capability: KVM_CAP_DEVICE_CTRL 2343Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device
2344Type: device ioctl 2344Type: device ioctl, vm ioctl
2345Parameters: struct kvm_device_attr 2345Parameters: struct kvm_device_attr
2346Returns: 0 on success, -1 on error 2346Returns: 0 on success, -1 on error
2347Errors: 2347Errors:
diff --git a/Documentation/virtual/kvm/devices/vm.txt b/Documentation/virtual/kvm/devices/vm.txt
new file mode 100644
index 000000000000..0d16f96c0eac
--- /dev/null
+++ b/Documentation/virtual/kvm/devices/vm.txt
@@ -0,0 +1,26 @@
1Generic vm interface
2====================================
3
4The virtual machine "device" also accepts the ioctls KVM_SET_DEVICE_ATTR,
5KVM_GET_DEVICE_ATTR, and KVM_HAS_DEVICE_ATTR. The interface uses the same
6struct kvm_device_attr as other devices, but targets VM-wide settings
7and controls.
8
9The groups and attributes per virtual machine, if any, are architecture
10specific.
11
121. GROUP: KVM_S390_VM_MEM_CTRL
13Architectures: s390
14
151.1. ATTRIBUTE: KVM_S390_VM_MEM_CTRL
16Parameters: none
17Returns: -EBUSY if already a vcpus is defined, otherwise 0
18
19Enables CMMA for the virtual machine
20
211.2. ATTRIBUTE: KVM_S390_VM_CLR_CMMA
22Parameteres: none
23Returns: 0
24
25Clear the CMMA status for all guest pages, so any pages the guest marked
26as unused are again used any may not be reclaimed by the host.
diff --git a/Documentation/virtual/kvm/s390-diag.txt b/Documentation/virtual/kvm/s390-diag.txt
index f1de4fbade15..48c4921794ed 100644
--- a/Documentation/virtual/kvm/s390-diag.txt
+++ b/Documentation/virtual/kvm/s390-diag.txt
@@ -78,3 +78,5 @@ DIAGNOSE function code 'X'501 - KVM breakpoint
78 78
79If the function code specifies 0x501, breakpoint functions may be performed. 79If the function code specifies 0x501, breakpoint functions may be performed.
80This function code is handled by userspace. 80This function code is handled by userspace.
81
82This diagnose function code has no subfunctions and uses no parameters.
diff --git a/arch/s390/include/asm/ctl_reg.h b/arch/s390/include/asm/ctl_reg.h
index 4e63f1a13600..31ab9f346d7e 100644
--- a/arch/s390/include/asm/ctl_reg.h
+++ b/arch/s390/include/asm/ctl_reg.h
@@ -57,6 +57,20 @@ static inline void __ctl_clear_bit(unsigned int cr, unsigned int bit)
57void smp_ctl_set_bit(int cr, int bit); 57void smp_ctl_set_bit(int cr, int bit);
58void smp_ctl_clear_bit(int cr, int bit); 58void smp_ctl_clear_bit(int cr, int bit);
59 59
60union ctlreg0 {
61 unsigned long val;
62 struct {
63#ifdef CONFIG_64BIT
64 unsigned long : 32;
65#endif
66 unsigned long : 3;
67 unsigned long lap : 1; /* Low-address-protection control */
68 unsigned long : 4;
69 unsigned long edat : 1; /* Enhanced-DAT-enablement control */
70 unsigned long : 23;
71 };
72};
73
60#ifdef CONFIG_SMP 74#ifdef CONFIG_SMP
61# define ctl_set_bit(cr, bit) smp_ctl_set_bit(cr, bit) 75# define ctl_set_bit(cr, bit) smp_ctl_set_bit(cr, bit)
62# define ctl_clear_bit(cr, bit) smp_ctl_clear_bit(cr, bit) 76# define ctl_clear_bit(cr, bit) smp_ctl_clear_bit(cr, bit)
diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h
index 154b60089be9..0d45f6fe734f 100644
--- a/arch/s390/include/asm/kvm_host.h
+++ b/arch/s390/include/asm/kvm_host.h
@@ -39,9 +39,17 @@ struct sca_entry {
39 __u64 reserved2[2]; 39 __u64 reserved2[2];
40} __attribute__((packed)); 40} __attribute__((packed));
41 41
42union ipte_control {
43 unsigned long val;
44 struct {
45 unsigned long k : 1;
46 unsigned long kh : 31;
47 unsigned long kg : 32;
48 };
49};
42 50
43struct sca_block { 51struct sca_block {
44 __u64 ipte_control; 52 union ipte_control ipte_control;
45 __u64 reserved[5]; 53 __u64 reserved[5];
46 __u64 mcn; 54 __u64 mcn;
47 __u64 reserved2; 55 __u64 reserved2;
@@ -85,12 +93,26 @@ struct kvm_s390_sie_block {
85 __u8 reserved40[4]; /* 0x0040 */ 93 __u8 reserved40[4]; /* 0x0040 */
86#define LCTL_CR0 0x8000 94#define LCTL_CR0 0x8000
87#define LCTL_CR6 0x0200 95#define LCTL_CR6 0x0200
96#define LCTL_CR9 0x0040
97#define LCTL_CR10 0x0020
98#define LCTL_CR11 0x0010
88#define LCTL_CR14 0x0002 99#define LCTL_CR14 0x0002
89 __u16 lctl; /* 0x0044 */ 100 __u16 lctl; /* 0x0044 */
90 __s16 icpua; /* 0x0046 */ 101 __s16 icpua; /* 0x0046 */
91#define ICTL_LPSW 0x00400000 102#define ICTL_PINT 0x20000000
103#define ICTL_LPSW 0x00400000
104#define ICTL_STCTL 0x00040000
105#define ICTL_ISKE 0x00004000
106#define ICTL_SSKE 0x00002000
107#define ICTL_RRBE 0x00001000
92 __u32 ictl; /* 0x0048 */ 108 __u32 ictl; /* 0x0048 */
93 __u32 eca; /* 0x004c */ 109 __u32 eca; /* 0x004c */
110#define ICPT_INST 0x04
111#define ICPT_PROGI 0x08
112#define ICPT_INSTPROGI 0x0C
113#define ICPT_OPEREXC 0x2C
114#define ICPT_PARTEXEC 0x38
115#define ICPT_IOINST 0x40
94 __u8 icptcode; /* 0x0050 */ 116 __u8 icptcode; /* 0x0050 */
95 __u8 reserved51; /* 0x0051 */ 117 __u8 reserved51; /* 0x0051 */
96 __u16 ihcpu; /* 0x0052 */ 118 __u16 ihcpu; /* 0x0052 */
@@ -109,9 +131,21 @@ struct kvm_s390_sie_block {
109 psw_t gpsw; /* 0x0090 */ 131 psw_t gpsw; /* 0x0090 */
110 __u64 gg14; /* 0x00a0 */ 132 __u64 gg14; /* 0x00a0 */
111 __u64 gg15; /* 0x00a8 */ 133 __u64 gg15; /* 0x00a8 */
112 __u8 reservedb0[30]; /* 0x00b0 */ 134 __u8 reservedb0[28]; /* 0x00b0 */
113 __u16 iprcc; /* 0x00ce */ 135 __u16 pgmilc; /* 0x00cc */
114 __u8 reservedd0[48]; /* 0x00d0 */ 136 __u16 iprcc; /* 0x00ce */
137 __u32 dxc; /* 0x00d0 */
138 __u16 mcn; /* 0x00d4 */
139 __u8 perc; /* 0x00d6 */
140 __u8 peratmid; /* 0x00d7 */
141 __u64 peraddr; /* 0x00d8 */
142 __u8 eai; /* 0x00e0 */
143 __u8 peraid; /* 0x00e1 */
144 __u8 oai; /* 0x00e2 */
145 __u8 armid; /* 0x00e3 */
146 __u8 reservede4[4]; /* 0x00e4 */
147 __u64 tecmc; /* 0x00e8 */
148 __u8 reservedf0[16]; /* 0x00f0 */
115 __u64 gcr[16]; /* 0x0100 */ 149 __u64 gcr[16]; /* 0x0100 */
116 __u64 gbea; /* 0x0180 */ 150 __u64 gbea; /* 0x0180 */
117 __u8 reserved188[24]; /* 0x0188 */ 151 __u8 reserved188[24]; /* 0x0188 */
@@ -146,6 +180,8 @@ struct kvm_vcpu_stat {
146 u32 exit_instruction; 180 u32 exit_instruction;
147 u32 instruction_lctl; 181 u32 instruction_lctl;
148 u32 instruction_lctlg; 182 u32 instruction_lctlg;
183 u32 instruction_stctl;
184 u32 instruction_stctg;
149 u32 exit_program_interruption; 185 u32 exit_program_interruption;
150 u32 exit_instr_and_program; 186 u32 exit_instr_and_program;
151 u32 deliver_external_call; 187 u32 deliver_external_call;
@@ -164,6 +200,7 @@ struct kvm_vcpu_stat {
164 u32 instruction_stpx; 200 u32 instruction_stpx;
165 u32 instruction_stap; 201 u32 instruction_stap;
166 u32 instruction_storage_key; 202 u32 instruction_storage_key;
203 u32 instruction_ipte_interlock;
167 u32 instruction_stsch; 204 u32 instruction_stsch;
168 u32 instruction_chsc; 205 u32 instruction_chsc;
169 u32 instruction_stsi; 206 u32 instruction_stsi;
@@ -183,13 +220,58 @@ struct kvm_vcpu_stat {
183 u32 diagnose_9c; 220 u32 diagnose_9c;
184}; 221};
185 222
186#define PGM_OPERATION 0x01 223#define PGM_OPERATION 0x01
187#define PGM_PRIVILEGED_OP 0x02 224#define PGM_PRIVILEGED_OP 0x02
188#define PGM_EXECUTE 0x03 225#define PGM_EXECUTE 0x03
189#define PGM_PROTECTION 0x04 226#define PGM_PROTECTION 0x04
190#define PGM_ADDRESSING 0x05 227#define PGM_ADDRESSING 0x05
191#define PGM_SPECIFICATION 0x06 228#define PGM_SPECIFICATION 0x06
192#define PGM_DATA 0x07 229#define PGM_DATA 0x07
230#define PGM_FIXED_POINT_OVERFLOW 0x08
231#define PGM_FIXED_POINT_DIVIDE 0x09
232#define PGM_DECIMAL_OVERFLOW 0x0a
233#define PGM_DECIMAL_DIVIDE 0x0b
234#define PGM_HFP_EXPONENT_OVERFLOW 0x0c
235#define PGM_HFP_EXPONENT_UNDERFLOW 0x0d
236#define PGM_HFP_SIGNIFICANCE 0x0e
237#define PGM_HFP_DIVIDE 0x0f
238#define PGM_SEGMENT_TRANSLATION 0x10
239#define PGM_PAGE_TRANSLATION 0x11
240#define PGM_TRANSLATION_SPEC 0x12
241#define PGM_SPECIAL_OPERATION 0x13
242#define PGM_OPERAND 0x15
243#define PGM_TRACE_TABEL 0x16
244#define PGM_SPACE_SWITCH 0x1c
245#define PGM_HFP_SQUARE_ROOT 0x1d
246#define PGM_PC_TRANSLATION_SPEC 0x1f
247#define PGM_AFX_TRANSLATION 0x20
248#define PGM_ASX_TRANSLATION 0x21
249#define PGM_LX_TRANSLATION 0x22
250#define PGM_EX_TRANSLATION 0x23
251#define PGM_PRIMARY_AUTHORITY 0x24
252#define PGM_SECONDARY_AUTHORITY 0x25
253#define PGM_LFX_TRANSLATION 0x26
254#define PGM_LSX_TRANSLATION 0x27
255#define PGM_ALET_SPECIFICATION 0x28
256#define PGM_ALEN_TRANSLATION 0x29
257#define PGM_ALE_SEQUENCE 0x2a
258#define PGM_ASTE_VALIDITY 0x2b
259#define PGM_ASTE_SEQUENCE 0x2c
260#define PGM_EXTENDED_AUTHORITY 0x2d
261#define PGM_LSTE_SEQUENCE 0x2e
262#define PGM_ASTE_INSTANCE 0x2f
263#define PGM_STACK_FULL 0x30
264#define PGM_STACK_EMPTY 0x31
265#define PGM_STACK_SPECIFICATION 0x32
266#define PGM_STACK_TYPE 0x33
267#define PGM_STACK_OPERATION 0x34
268#define PGM_ASCE_TYPE 0x38
269#define PGM_REGION_FIRST_TRANS 0x39
270#define PGM_REGION_SECOND_TRANS 0x3a
271#define PGM_REGION_THIRD_TRANS 0x3b
272#define PGM_MONITOR 0x40
273#define PGM_PER 0x80
274#define PGM_CRYPTO_OPERATION 0x119
193 275
194struct kvm_s390_interrupt_info { 276struct kvm_s390_interrupt_info {
195 struct list_head list; 277 struct list_head list;
@@ -229,6 +311,45 @@ struct kvm_s390_float_interrupt {
229 unsigned int irq_count; 311 unsigned int irq_count;
230}; 312};
231 313
314struct kvm_hw_wp_info_arch {
315 unsigned long addr;
316 unsigned long phys_addr;
317 int len;
318 char *old_data;
319};
320
321struct kvm_hw_bp_info_arch {
322 unsigned long addr;
323 int len;
324};
325
326/*
327 * Only the upper 16 bits of kvm_guest_debug->control are arch specific.
328 * Further KVM_GUESTDBG flags which an be used from userspace can be found in
329 * arch/s390/include/uapi/asm/kvm.h
330 */
331#define KVM_GUESTDBG_EXIT_PENDING 0x10000000
332
333#define guestdbg_enabled(vcpu) \
334 (vcpu->guest_debug & KVM_GUESTDBG_ENABLE)
335#define guestdbg_sstep_enabled(vcpu) \
336 (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
337#define guestdbg_hw_bp_enabled(vcpu) \
338 (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)
339#define guestdbg_exit_pending(vcpu) (guestdbg_enabled(vcpu) && \
340 (vcpu->guest_debug & KVM_GUESTDBG_EXIT_PENDING))
341
342struct kvm_guestdbg_info_arch {
343 unsigned long cr0;
344 unsigned long cr9;
345 unsigned long cr10;
346 unsigned long cr11;
347 struct kvm_hw_bp_info_arch *hw_bp_info;
348 struct kvm_hw_wp_info_arch *hw_wp_info;
349 int nr_hw_bp;
350 int nr_hw_wp;
351 unsigned long last_bp;
352};
232 353
233struct kvm_vcpu_arch { 354struct kvm_vcpu_arch {
234 struct kvm_s390_sie_block *sie_block; 355 struct kvm_s390_sie_block *sie_block;
@@ -238,11 +359,13 @@ struct kvm_vcpu_arch {
238 struct kvm_s390_local_interrupt local_int; 359 struct kvm_s390_local_interrupt local_int;
239 struct hrtimer ckc_timer; 360 struct hrtimer ckc_timer;
240 struct tasklet_struct tasklet; 361 struct tasklet_struct tasklet;
362 struct kvm_s390_pgm_info pgm;
241 union { 363 union {
242 struct cpuid cpu_id; 364 struct cpuid cpu_id;
243 u64 stidp_data; 365 u64 stidp_data;
244 }; 366 };
245 struct gmap *gmap; 367 struct gmap *gmap;
368 struct kvm_guestdbg_info_arch guestdbg;
246#define KVM_S390_PFAULT_TOKEN_INVALID (-1UL) 369#define KVM_S390_PFAULT_TOKEN_INVALID (-1UL)
247 unsigned long pfault_token; 370 unsigned long pfault_token;
248 unsigned long pfault_select; 371 unsigned long pfault_select;
@@ -285,7 +408,9 @@ struct kvm_arch{
285 struct gmap *gmap; 408 struct gmap *gmap;
286 int css_support; 409 int css_support;
287 int use_irqchip; 410 int use_irqchip;
411 int use_cmma;
288 struct s390_io_adapter *adapters[MAX_S390_IO_ADAPTERS]; 412 struct s390_io_adapter *adapters[MAX_S390_IO_ADAPTERS];
413 wait_queue_head_t ipte_wq;
289}; 414};
290 415
291#define KVM_HVA_ERR_BAD (-1UL) 416#define KVM_HVA_ERR_BAD (-1UL)
diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h
index bbf8141408cd..e88cb8c54130 100644
--- a/arch/s390/include/asm/lowcore.h
+++ b/arch/s390/include/asm/lowcore.h
@@ -56,13 +56,14 @@ struct _lowcore {
56 __u16 pgm_code; /* 0x008e */ 56 __u16 pgm_code; /* 0x008e */
57 __u32 trans_exc_code; /* 0x0090 */ 57 __u32 trans_exc_code; /* 0x0090 */
58 __u16 mon_class_num; /* 0x0094 */ 58 __u16 mon_class_num; /* 0x0094 */
59 __u16 per_perc_atmid; /* 0x0096 */ 59 __u8 per_code; /* 0x0096 */
60 __u8 per_atmid; /* 0x0097 */
60 __u32 per_address; /* 0x0098 */ 61 __u32 per_address; /* 0x0098 */
61 __u32 monitor_code; /* 0x009c */ 62 __u32 monitor_code; /* 0x009c */
62 __u8 exc_access_id; /* 0x00a0 */ 63 __u8 exc_access_id; /* 0x00a0 */
63 __u8 per_access_id; /* 0x00a1 */ 64 __u8 per_access_id; /* 0x00a1 */
64 __u8 op_access_id; /* 0x00a2 */ 65 __u8 op_access_id; /* 0x00a2 */
65 __u8 ar_access_id; /* 0x00a3 */ 66 __u8 ar_mode_id; /* 0x00a3 */
66 __u8 pad_0x00a4[0x00b8-0x00a4]; /* 0x00a4 */ 67 __u8 pad_0x00a4[0x00b8-0x00a4]; /* 0x00a4 */
67 __u16 subchannel_id; /* 0x00b8 */ 68 __u16 subchannel_id; /* 0x00b8 */
68 __u16 subchannel_nr; /* 0x00ba */ 69 __u16 subchannel_nr; /* 0x00ba */
@@ -196,12 +197,13 @@ struct _lowcore {
196 __u16 pgm_code; /* 0x008e */ 197 __u16 pgm_code; /* 0x008e */
197 __u32 data_exc_code; /* 0x0090 */ 198 __u32 data_exc_code; /* 0x0090 */
198 __u16 mon_class_num; /* 0x0094 */ 199 __u16 mon_class_num; /* 0x0094 */
199 __u16 per_perc_atmid; /* 0x0096 */ 200 __u8 per_code; /* 0x0096 */
201 __u8 per_atmid; /* 0x0097 */
200 __u64 per_address; /* 0x0098 */ 202 __u64 per_address; /* 0x0098 */
201 __u8 exc_access_id; /* 0x00a0 */ 203 __u8 exc_access_id; /* 0x00a0 */
202 __u8 per_access_id; /* 0x00a1 */ 204 __u8 per_access_id; /* 0x00a1 */
203 __u8 op_access_id; /* 0x00a2 */ 205 __u8 op_access_id; /* 0x00a2 */
204 __u8 ar_access_id; /* 0x00a3 */ 206 __u8 ar_mode_id; /* 0x00a3 */
205 __u8 pad_0x00a4[0x00a8-0x00a4]; /* 0x00a4 */ 207 __u8 pad_0x00a4[0x00a8-0x00a4]; /* 0x00a4 */
206 __u64 trans_exc_code; /* 0x00a8 */ 208 __u64 trans_exc_code; /* 0x00a8 */
207 __u64 monitor_code; /* 0x00b0 */ 209 __u64 monitor_code; /* 0x00b0 */
diff --git a/arch/s390/include/asm/mmu.h b/arch/s390/include/asm/mmu.h
index f77695a82f64..a5e656260a70 100644
--- a/arch/s390/include/asm/mmu.h
+++ b/arch/s390/include/asm/mmu.h
@@ -16,6 +16,8 @@ typedef struct {
16 unsigned long vdso_base; 16 unsigned long vdso_base;
17 /* The mmu context has extended page tables. */ 17 /* The mmu context has extended page tables. */
18 unsigned int has_pgste:1; 18 unsigned int has_pgste:1;
19 /* The mmu context uses storage keys. */
20 unsigned int use_skey:1;
19} mm_context_t; 21} mm_context_t;
20 22
21#define INIT_MM_CONTEXT(name) \ 23#define INIT_MM_CONTEXT(name) \
diff --git a/arch/s390/include/asm/mmu_context.h b/arch/s390/include/asm/mmu_context.h
index 71be346d0e3c..d42fb1b728d8 100644
--- a/arch/s390/include/asm/mmu_context.h
+++ b/arch/s390/include/asm/mmu_context.h
@@ -23,6 +23,7 @@ static inline int init_new_context(struct task_struct *tsk,
23 mm->context.asce_bits |= _ASCE_TYPE_REGION3; 23 mm->context.asce_bits |= _ASCE_TYPE_REGION3;
24#endif 24#endif
25 mm->context.has_pgste = 0; 25 mm->context.has_pgste = 0;
26 mm->context.use_skey = 0;
26 mm->context.asce_limit = STACK_TOP_MAX; 27 mm->context.asce_limit = STACK_TOP_MAX;
27 crst_table_init((unsigned long *) mm->pgd, pgd_entry_type(mm)); 28 crst_table_init((unsigned long *) mm->pgd, pgd_entry_type(mm));
28 return 0; 29 return 0;
diff --git a/arch/s390/include/asm/pgalloc.h b/arch/s390/include/asm/pgalloc.h
index 884017cbfa9f..9e18a61d3df3 100644
--- a/arch/s390/include/asm/pgalloc.h
+++ b/arch/s390/include/asm/pgalloc.h
@@ -22,7 +22,8 @@ unsigned long *page_table_alloc(struct mm_struct *, unsigned long);
22void page_table_free(struct mm_struct *, unsigned long *); 22void page_table_free(struct mm_struct *, unsigned long *);
23void page_table_free_rcu(struct mmu_gather *, unsigned long *); 23void page_table_free_rcu(struct mmu_gather *, unsigned long *);
24 24
25void page_table_reset_pgste(struct mm_struct *, unsigned long, unsigned long); 25void page_table_reset_pgste(struct mm_struct *, unsigned long, unsigned long,
26 bool init_skey);
26int set_guest_storage_key(struct mm_struct *mm, unsigned long addr, 27int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
27 unsigned long key, bool nq); 28 unsigned long key, bool nq);
28 29
diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h
index 12f75313e086..fcba5e03839f 100644
--- a/arch/s390/include/asm/pgtable.h
+++ b/arch/s390/include/asm/pgtable.h
@@ -309,7 +309,8 @@ extern unsigned long MODULES_END;
309#define PGSTE_HC_BIT 0x00200000UL 309#define PGSTE_HC_BIT 0x00200000UL
310#define PGSTE_GR_BIT 0x00040000UL 310#define PGSTE_GR_BIT 0x00040000UL
311#define PGSTE_GC_BIT 0x00020000UL 311#define PGSTE_GC_BIT 0x00020000UL
312#define PGSTE_IN_BIT 0x00008000UL /* IPTE notify bit */ 312#define PGSTE_UC_BIT 0x00008000UL /* user dirty (migration) */
313#define PGSTE_IN_BIT 0x00004000UL /* IPTE notify bit */
313 314
314#else /* CONFIG_64BIT */ 315#else /* CONFIG_64BIT */
315 316
@@ -391,7 +392,8 @@ extern unsigned long MODULES_END;
391#define PGSTE_HC_BIT 0x0020000000000000UL 392#define PGSTE_HC_BIT 0x0020000000000000UL
392#define PGSTE_GR_BIT 0x0004000000000000UL 393#define PGSTE_GR_BIT 0x0004000000000000UL
393#define PGSTE_GC_BIT 0x0002000000000000UL 394#define PGSTE_GC_BIT 0x0002000000000000UL
394#define PGSTE_IN_BIT 0x0000800000000000UL /* IPTE notify bit */ 395#define PGSTE_UC_BIT 0x0000800000000000UL /* user dirty (migration) */
396#define PGSTE_IN_BIT 0x0000400000000000UL /* IPTE notify bit */
395 397
396#endif /* CONFIG_64BIT */ 398#endif /* CONFIG_64BIT */
397 399
@@ -466,6 +468,16 @@ static inline int mm_has_pgste(struct mm_struct *mm)
466#endif 468#endif
467 return 0; 469 return 0;
468} 470}
471
472static inline int mm_use_skey(struct mm_struct *mm)
473{
474#ifdef CONFIG_PGSTE
475 if (mm->context.use_skey)
476 return 1;
477#endif
478 return 0;
479}
480
469/* 481/*
470 * pgd/pmd/pte query functions 482 * pgd/pmd/pte query functions
471 */ 483 */
@@ -699,26 +711,17 @@ static inline void pgste_set(pte_t *ptep, pgste_t pgste)
699#endif 711#endif
700} 712}
701 713
702static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste) 714static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste,
715 struct mm_struct *mm)
703{ 716{
704#ifdef CONFIG_PGSTE 717#ifdef CONFIG_PGSTE
705 unsigned long address, bits, skey; 718 unsigned long address, bits, skey;
706 719
707 if (pte_val(*ptep) & _PAGE_INVALID) 720 if (!mm_use_skey(mm) || pte_val(*ptep) & _PAGE_INVALID)
708 return pgste; 721 return pgste;
709 address = pte_val(*ptep) & PAGE_MASK; 722 address = pte_val(*ptep) & PAGE_MASK;
710 skey = (unsigned long) page_get_storage_key(address); 723 skey = (unsigned long) page_get_storage_key(address);
711 bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED); 724 bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
712 if (!(pgste_val(pgste) & PGSTE_HC_BIT) && (bits & _PAGE_CHANGED)) {
713 /* Transfer dirty + referenced bit to host bits in pgste */
714 pgste_val(pgste) |= bits << 52;
715 page_set_storage_key(address, skey ^ bits, 0);
716 } else if (!(pgste_val(pgste) & PGSTE_HR_BIT) &&
717 (bits & _PAGE_REFERENCED)) {
718 /* Transfer referenced bit to host bit in pgste */
719 pgste_val(pgste) |= PGSTE_HR_BIT;
720 page_reset_referenced(address);
721 }
722 /* Transfer page changed & referenced bit to guest bits in pgste */ 725 /* Transfer page changed & referenced bit to guest bits in pgste */
723 pgste_val(pgste) |= bits << 48; /* GR bit & GC bit */ 726 pgste_val(pgste) |= bits << 48; /* GR bit & GC bit */
724 /* Copy page access key and fetch protection bit to pgste */ 727 /* Copy page access key and fetch protection bit to pgste */
@@ -729,25 +732,14 @@ static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
729 732
730} 733}
731 734
732static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste) 735static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry,
733{ 736 struct mm_struct *mm)
734#ifdef CONFIG_PGSTE
735 if (pte_val(*ptep) & _PAGE_INVALID)
736 return pgste;
737 /* Get referenced bit from storage key */
738 if (page_reset_referenced(pte_val(*ptep) & PAGE_MASK))
739 pgste_val(pgste) |= PGSTE_HR_BIT | PGSTE_GR_BIT;
740#endif
741 return pgste;
742}
743
744static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry)
745{ 737{
746#ifdef CONFIG_PGSTE 738#ifdef CONFIG_PGSTE
747 unsigned long address; 739 unsigned long address;
748 unsigned long nkey; 740 unsigned long nkey;
749 741
750 if (pte_val(entry) & _PAGE_INVALID) 742 if (!mm_use_skey(mm) || pte_val(entry) & _PAGE_INVALID)
751 return; 743 return;
752 VM_BUG_ON(!(pte_val(*ptep) & _PAGE_INVALID)); 744 VM_BUG_ON(!(pte_val(*ptep) & _PAGE_INVALID));
753 address = pte_val(entry) & PAGE_MASK; 745 address = pte_val(entry) & PAGE_MASK;
@@ -757,23 +749,30 @@ static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry)
757 * key C/R to 0. 749 * key C/R to 0.
758 */ 750 */
759 nkey = (pgste_val(pgste) & (PGSTE_ACC_BITS | PGSTE_FP_BIT)) >> 56; 751 nkey = (pgste_val(pgste) & (PGSTE_ACC_BITS | PGSTE_FP_BIT)) >> 56;
752 nkey |= (pgste_val(pgste) & (PGSTE_GR_BIT | PGSTE_GC_BIT)) >> 48;
760 page_set_storage_key(address, nkey, 0); 753 page_set_storage_key(address, nkey, 0);
761#endif 754#endif
762} 755}
763 756
764static inline void pgste_set_pte(pte_t *ptep, pte_t entry) 757static inline pgste_t pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry)
765{ 758{
766 if (!MACHINE_HAS_ESOP && 759 if ((pte_val(entry) & _PAGE_PRESENT) &&
767 (pte_val(entry) & _PAGE_PRESENT) && 760 (pte_val(entry) & _PAGE_WRITE) &&
768 (pte_val(entry) & _PAGE_WRITE)) { 761 !(pte_val(entry) & _PAGE_INVALID)) {
769 /* 762 if (!MACHINE_HAS_ESOP) {
770 * Without enhanced suppression-on-protection force 763 /*
771 * the dirty bit on for all writable ptes. 764 * Without enhanced suppression-on-protection force
772 */ 765 * the dirty bit on for all writable ptes.
773 pte_val(entry) |= _PAGE_DIRTY; 766 */
774 pte_val(entry) &= ~_PAGE_PROTECT; 767 pte_val(entry) |= _PAGE_DIRTY;
768 pte_val(entry) &= ~_PAGE_PROTECT;
769 }
770 if (!(pte_val(entry) & _PAGE_PROTECT))
771 /* This pte allows write access, set user-dirty */
772 pgste_val(pgste) |= PGSTE_UC_BIT;
775 } 773 }
776 *ptep = entry; 774 *ptep = entry;
775 return pgste;
777} 776}
778 777
779/** 778/**
@@ -839,6 +838,8 @@ unsigned long __gmap_fault(unsigned long address, struct gmap *);
839unsigned long gmap_fault(unsigned long address, struct gmap *); 838unsigned long gmap_fault(unsigned long address, struct gmap *);
840void gmap_discard(unsigned long from, unsigned long to, struct gmap *); 839void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
841void __gmap_zap(unsigned long address, struct gmap *); 840void __gmap_zap(unsigned long address, struct gmap *);
841bool gmap_test_and_clear_dirty(unsigned long address, struct gmap *);
842
842 843
843void gmap_register_ipte_notifier(struct gmap_notifier *); 844void gmap_register_ipte_notifier(struct gmap_notifier *);
844void gmap_unregister_ipte_notifier(struct gmap_notifier *); 845void gmap_unregister_ipte_notifier(struct gmap_notifier *);
@@ -870,8 +871,8 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
870 if (mm_has_pgste(mm)) { 871 if (mm_has_pgste(mm)) {
871 pgste = pgste_get_lock(ptep); 872 pgste = pgste_get_lock(ptep);
872 pgste_val(pgste) &= ~_PGSTE_GPS_ZERO; 873 pgste_val(pgste) &= ~_PGSTE_GPS_ZERO;
873 pgste_set_key(ptep, pgste, entry); 874 pgste_set_key(ptep, pgste, entry, mm);
874 pgste_set_pte(ptep, entry); 875 pgste = pgste_set_pte(ptep, pgste, entry);
875 pgste_set_unlock(ptep, pgste); 876 pgste_set_unlock(ptep, pgste);
876 } else { 877 } else {
877 if (!(pte_val(entry) & _PAGE_INVALID) && MACHINE_HAS_EDAT1) 878 if (!(pte_val(entry) & _PAGE_INVALID) && MACHINE_HAS_EDAT1)
@@ -1017,45 +1018,6 @@ static inline pte_t pte_mkhuge(pte_t pte)
1017} 1018}
1018#endif 1019#endif
1019 1020
1020/*
1021 * Get (and clear) the user dirty bit for a pte.
1022 */
1023static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
1024 pte_t *ptep)
1025{
1026 pgste_t pgste;
1027 int dirty = 0;
1028
1029 if (mm_has_pgste(mm)) {
1030 pgste = pgste_get_lock(ptep);
1031 pgste = pgste_update_all(ptep, pgste);
1032 dirty = !!(pgste_val(pgste) & PGSTE_HC_BIT);
1033 pgste_val(pgste) &= ~PGSTE_HC_BIT;
1034 pgste_set_unlock(ptep, pgste);
1035 return dirty;
1036 }
1037 return dirty;
1038}
1039
1040/*
1041 * Get (and clear) the user referenced bit for a pte.
1042 */
1043static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
1044 pte_t *ptep)
1045{
1046 pgste_t pgste;
1047 int young = 0;
1048
1049 if (mm_has_pgste(mm)) {
1050 pgste = pgste_get_lock(ptep);
1051 pgste = pgste_update_young(ptep, pgste);
1052 young = !!(pgste_val(pgste) & PGSTE_HR_BIT);
1053 pgste_val(pgste) &= ~PGSTE_HR_BIT;
1054 pgste_set_unlock(ptep, pgste);
1055 }
1056 return young;
1057}
1058
1059static inline void __ptep_ipte(unsigned long address, pte_t *ptep) 1021static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
1060{ 1022{
1061 unsigned long pto = (unsigned long) ptep; 1023 unsigned long pto = (unsigned long) ptep;
@@ -1118,6 +1080,36 @@ static inline void ptep_flush_lazy(struct mm_struct *mm,
1118 atomic_sub(0x10000, &mm->context.attach_count); 1080 atomic_sub(0x10000, &mm->context.attach_count);
1119} 1081}
1120 1082
1083/*
1084 * Get (and clear) the user dirty bit for a pte.
1085 */
1086static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
1087 unsigned long addr,
1088 pte_t *ptep)
1089{
1090 pgste_t pgste;
1091 pte_t pte;
1092 int dirty;
1093
1094 if (!mm_has_pgste(mm))
1095 return 0;
1096 pgste = pgste_get_lock(ptep);
1097 dirty = !!(pgste_val(pgste) & PGSTE_UC_BIT);
1098 pgste_val(pgste) &= ~PGSTE_UC_BIT;
1099 pte = *ptep;
1100 if (dirty && (pte_val(pte) & _PAGE_PRESENT)) {
1101 pgste = pgste_ipte_notify(mm, ptep, pgste);
1102 __ptep_ipte(addr, ptep);
1103 if (MACHINE_HAS_ESOP || !(pte_val(pte) & _PAGE_WRITE))
1104 pte_val(pte) |= _PAGE_PROTECT;
1105 else
1106 pte_val(pte) |= _PAGE_INVALID;
1107 *ptep = pte;
1108 }
1109 pgste_set_unlock(ptep, pgste);
1110 return dirty;
1111}
1112
1121#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG 1113#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1122static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, 1114static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1123 unsigned long addr, pte_t *ptep) 1115 unsigned long addr, pte_t *ptep)
@@ -1137,7 +1129,7 @@ static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1137 pte = pte_mkold(pte); 1129 pte = pte_mkold(pte);
1138 1130
1139 if (mm_has_pgste(vma->vm_mm)) { 1131 if (mm_has_pgste(vma->vm_mm)) {
1140 pgste_set_pte(ptep, pte); 1132 pgste = pgste_set_pte(ptep, pgste, pte);
1141 pgste_set_unlock(ptep, pgste); 1133 pgste_set_unlock(ptep, pgste);
1142 } else 1134 } else
1143 *ptep = pte; 1135 *ptep = pte;
@@ -1182,7 +1174,7 @@ static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1182 pte_val(*ptep) = _PAGE_INVALID; 1174 pte_val(*ptep) = _PAGE_INVALID;
1183 1175
1184 if (mm_has_pgste(mm)) { 1176 if (mm_has_pgste(mm)) {
1185 pgste = pgste_update_all(&pte, pgste); 1177 pgste = pgste_update_all(&pte, pgste, mm);
1186 pgste_set_unlock(ptep, pgste); 1178 pgste_set_unlock(ptep, pgste);
1187 } 1179 }
1188 return pte; 1180 return pte;
@@ -1205,7 +1197,7 @@ static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1205 ptep_flush_lazy(mm, address, ptep); 1197 ptep_flush_lazy(mm, address, ptep);
1206 1198
1207 if (mm_has_pgste(mm)) { 1199 if (mm_has_pgste(mm)) {
1208 pgste = pgste_update_all(&pte, pgste); 1200 pgste = pgste_update_all(&pte, pgste, mm);
1209 pgste_set(ptep, pgste); 1201 pgste_set(ptep, pgste);
1210 } 1202 }
1211 return pte; 1203 return pte;
@@ -1219,8 +1211,8 @@ static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1219 1211
1220 if (mm_has_pgste(mm)) { 1212 if (mm_has_pgste(mm)) {
1221 pgste = pgste_get(ptep); 1213 pgste = pgste_get(ptep);
1222 pgste_set_key(ptep, pgste, pte); 1214 pgste_set_key(ptep, pgste, pte, mm);
1223 pgste_set_pte(ptep, pte); 1215 pgste = pgste_set_pte(ptep, pgste, pte);
1224 pgste_set_unlock(ptep, pgste); 1216 pgste_set_unlock(ptep, pgste);
1225 } else 1217 } else
1226 *ptep = pte; 1218 *ptep = pte;
@@ -1246,7 +1238,7 @@ static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1246 if ((pgste_val(pgste) & _PGSTE_GPS_USAGE_MASK) == 1238 if ((pgste_val(pgste) & _PGSTE_GPS_USAGE_MASK) ==
1247 _PGSTE_GPS_USAGE_UNUSED) 1239 _PGSTE_GPS_USAGE_UNUSED)
1248 pte_val(pte) |= _PAGE_UNUSED; 1240 pte_val(pte) |= _PAGE_UNUSED;
1249 pgste = pgste_update_all(&pte, pgste); 1241 pgste = pgste_update_all(&pte, pgste, vma->vm_mm);
1250 pgste_set_unlock(ptep, pgste); 1242 pgste_set_unlock(ptep, pgste);
1251 } 1243 }
1252 return pte; 1244 return pte;
@@ -1278,7 +1270,7 @@ static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1278 pte_val(*ptep) = _PAGE_INVALID; 1270 pte_val(*ptep) = _PAGE_INVALID;
1279 1271
1280 if (!full && mm_has_pgste(mm)) { 1272 if (!full && mm_has_pgste(mm)) {
1281 pgste = pgste_update_all(&pte, pgste); 1273 pgste = pgste_update_all(&pte, pgste, mm);
1282 pgste_set_unlock(ptep, pgste); 1274 pgste_set_unlock(ptep, pgste);
1283 } 1275 }
1284 return pte; 1276 return pte;
@@ -1301,7 +1293,7 @@ static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1301 pte = pte_wrprotect(pte); 1293 pte = pte_wrprotect(pte);
1302 1294
1303 if (mm_has_pgste(mm)) { 1295 if (mm_has_pgste(mm)) {
1304 pgste_set_pte(ptep, pte); 1296 pgste = pgste_set_pte(ptep, pgste, pte);
1305 pgste_set_unlock(ptep, pgste); 1297 pgste_set_unlock(ptep, pgste);
1306 } else 1298 } else
1307 *ptep = pte; 1299 *ptep = pte;
@@ -1326,7 +1318,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1326 ptep_flush_direct(vma->vm_mm, address, ptep); 1318 ptep_flush_direct(vma->vm_mm, address, ptep);
1327 1319
1328 if (mm_has_pgste(vma->vm_mm)) { 1320 if (mm_has_pgste(vma->vm_mm)) {
1329 pgste_set_pte(ptep, entry); 1321 pgste = pgste_set_pte(ptep, pgste, entry);
1330 pgste_set_unlock(ptep, pgste); 1322 pgste_set_unlock(ptep, pgste);
1331 } else 1323 } else
1332 *ptep = entry; 1324 *ptep = entry;
@@ -1734,6 +1726,7 @@ static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1734extern int vmem_add_mapping(unsigned long start, unsigned long size); 1726extern int vmem_add_mapping(unsigned long start, unsigned long size);
1735extern int vmem_remove_mapping(unsigned long start, unsigned long size); 1727extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1736extern int s390_enable_sie(void); 1728extern int s390_enable_sie(void);
1729extern void s390_enable_skey(void);
1737 1730
1738/* 1731/*
1739 * No page table caches to initialise 1732 * No page table caches to initialise
diff --git a/arch/s390/include/asm/ptrace.h b/arch/s390/include/asm/ptrace.h
index f4783c0b7b43..6e7708f3d866 100644
--- a/arch/s390/include/asm/ptrace.h
+++ b/arch/s390/include/asm/ptrace.h
@@ -16,6 +16,50 @@
16 PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK | \ 16 PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK | \
17 PSW_MASK_PSTATE | PSW_ASC_PRIMARY) 17 PSW_MASK_PSTATE | PSW_ASC_PRIMARY)
18 18
19struct psw_bits {
20 unsigned long long : 1;
21 unsigned long long r : 1; /* PER-Mask */
22 unsigned long long : 3;
23 unsigned long long t : 1; /* DAT Mode */
24 unsigned long long i : 1; /* Input/Output Mask */
25 unsigned long long e : 1; /* External Mask */
26 unsigned long long key : 4; /* PSW Key */
27 unsigned long long : 1;
28 unsigned long long m : 1; /* Machine-Check Mask */
29 unsigned long long w : 1; /* Wait State */
30 unsigned long long p : 1; /* Problem State */
31 unsigned long long as : 2; /* Address Space Control */
32 unsigned long long cc : 2; /* Condition Code */
33 unsigned long long pm : 4; /* Program Mask */
34 unsigned long long ri : 1; /* Runtime Instrumentation */
35 unsigned long long : 6;
36 unsigned long long eaba : 2; /* Addressing Mode */
37#ifdef CONFIG_64BIT
38 unsigned long long : 31;
39 unsigned long long ia : 64;/* Instruction Address */
40#else
41 unsigned long long ia : 31;/* Instruction Address */
42#endif
43};
44
45enum {
46 PSW_AMODE_24BIT = 0,
47 PSW_AMODE_31BIT = 1,
48 PSW_AMODE_64BIT = 3
49};
50
51enum {
52 PSW_AS_PRIMARY = 0,
53 PSW_AS_ACCREG = 1,
54 PSW_AS_SECONDARY = 2,
55 PSW_AS_HOME = 3
56};
57
58#define psw_bits(__psw) (*({ \
59 typecheck(psw_t, __psw); \
60 &(*(struct psw_bits *)(&(__psw))); \
61}))
62
19/* 63/*
20 * The pt_regs struct defines the way the registers are stored on 64 * The pt_regs struct defines the way the registers are stored on
21 * the stack during a system call. 65 * the stack during a system call.
diff --git a/arch/s390/include/asm/sclp.h b/arch/s390/include/asm/sclp.h
index 2f5e9932b4de..943d43451116 100644
--- a/arch/s390/include/asm/sclp.h
+++ b/arch/s390/include/asm/sclp.h
@@ -28,7 +28,11 @@ struct sclp_ipl_info {
28 28
29struct sclp_cpu_entry { 29struct sclp_cpu_entry {
30 u8 address; 30 u8 address;
31 u8 reserved0[13]; 31 u8 reserved0[2];
32 u8 : 3;
33 u8 siif : 1;
34 u8 : 4;
35 u8 reserved2[10];
32 u8 type; 36 u8 type;
33 u8 reserved1; 37 u8 reserved1;
34} __attribute__((packed)); 38} __attribute__((packed));
@@ -61,5 +65,6 @@ int sclp_pci_deconfigure(u32 fid);
61int memcpy_hsa(void *dest, unsigned long src, size_t count, int mode); 65int memcpy_hsa(void *dest, unsigned long src, size_t count, int mode);
62unsigned long sclp_get_hsa_size(void); 66unsigned long sclp_get_hsa_size(void);
63void sclp_early_detect(void); 67void sclp_early_detect(void);
68int sclp_has_siif(void);
64 69
65#endif /* _ASM_S390_SCLP_H */ 70#endif /* _ASM_S390_SCLP_H */
diff --git a/arch/s390/include/uapi/asm/kvm.h b/arch/s390/include/uapi/asm/kvm.h
index c003c6a73b1e..0fc26430a1e5 100644
--- a/arch/s390/include/uapi/asm/kvm.h
+++ b/arch/s390/include/uapi/asm/kvm.h
@@ -15,6 +15,7 @@
15#include <linux/types.h> 15#include <linux/types.h>
16 16
17#define __KVM_S390 17#define __KVM_S390
18#define __KVM_HAVE_GUEST_DEBUG
18 19
19/* Device control API: s390-specific devices */ 20/* Device control API: s390-specific devices */
20#define KVM_DEV_FLIC_GET_ALL_IRQS 1 21#define KVM_DEV_FLIC_GET_ALL_IRQS 1
@@ -54,6 +55,13 @@ struct kvm_s390_io_adapter_req {
54 __u64 addr; 55 __u64 addr;
55}; 56};
56 57
58/* kvm attr_group on vm fd */
59#define KVM_S390_VM_MEM_CTRL 0
60
61/* kvm attributes for mem_ctrl */
62#define KVM_S390_VM_MEM_ENABLE_CMMA 0
63#define KVM_S390_VM_MEM_CLR_CMMA 1
64
57/* for KVM_GET_REGS and KVM_SET_REGS */ 65/* for KVM_GET_REGS and KVM_SET_REGS */
58struct kvm_regs { 66struct kvm_regs {
59 /* general purpose regs for s390 */ 67 /* general purpose regs for s390 */
@@ -72,11 +80,31 @@ struct kvm_fpu {
72 __u64 fprs[16]; 80 __u64 fprs[16];
73}; 81};
74 82
83#define KVM_GUESTDBG_USE_HW_BP 0x00010000
84
85#define KVM_HW_BP 1
86#define KVM_HW_WP_WRITE 2
87#define KVM_SINGLESTEP 4
88
75struct kvm_debug_exit_arch { 89struct kvm_debug_exit_arch {
90 __u64 addr;
91 __u8 type;
92 __u8 pad[7]; /* Should be set to 0 */
93};
94
95struct kvm_hw_breakpoint {
96 __u64 addr;
97 __u64 phys_addr;
98 __u64 len;
99 __u8 type;
100 __u8 pad[7]; /* Should be set to 0 */
76}; 101};
77 102
78/* for KVM_SET_GUEST_DEBUG */ 103/* for KVM_SET_GUEST_DEBUG */
79struct kvm_guest_debug_arch { 104struct kvm_guest_debug_arch {
105 __u32 nr_hw_bp;
106 __u32 pad; /* Should be set to 0 */
107 struct kvm_hw_breakpoint __user *hw_bp;
80}; 108};
81 109
82#define KVM_SYNC_PREFIX (1UL << 0) 110#define KVM_SYNC_PREFIX (1UL << 0)
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index cc10cdd4d6a2..859a7ed36c4b 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -89,16 +89,22 @@ int main(void)
89 DEFINE(__LC_PGM_ILC, offsetof(struct _lowcore, pgm_ilc)); 89 DEFINE(__LC_PGM_ILC, offsetof(struct _lowcore, pgm_ilc));
90 DEFINE(__LC_PGM_INT_CODE, offsetof(struct _lowcore, pgm_code)); 90 DEFINE(__LC_PGM_INT_CODE, offsetof(struct _lowcore, pgm_code));
91 DEFINE(__LC_TRANS_EXC_CODE, offsetof(struct _lowcore, trans_exc_code)); 91 DEFINE(__LC_TRANS_EXC_CODE, offsetof(struct _lowcore, trans_exc_code));
92 DEFINE(__LC_PER_CAUSE, offsetof(struct _lowcore, per_perc_atmid)); 92 DEFINE(__LC_MON_CLASS_NR, offsetof(struct _lowcore, mon_class_num));
93 DEFINE(__LC_PER_CODE, offsetof(struct _lowcore, per_code));
94 DEFINE(__LC_PER_ATMID, offsetof(struct _lowcore, per_atmid));
93 DEFINE(__LC_PER_ADDRESS, offsetof(struct _lowcore, per_address)); 95 DEFINE(__LC_PER_ADDRESS, offsetof(struct _lowcore, per_address));
94 DEFINE(__LC_PER_PAID, offsetof(struct _lowcore, per_access_id)); 96 DEFINE(__LC_EXC_ACCESS_ID, offsetof(struct _lowcore, exc_access_id));
95 DEFINE(__LC_AR_MODE_ID, offsetof(struct _lowcore, ar_access_id)); 97 DEFINE(__LC_PER_ACCESS_ID, offsetof(struct _lowcore, per_access_id));
98 DEFINE(__LC_OP_ACCESS_ID, offsetof(struct _lowcore, op_access_id));
99 DEFINE(__LC_AR_MODE_ID, offsetof(struct _lowcore, ar_mode_id));
100 DEFINE(__LC_MON_CODE, offsetof(struct _lowcore, monitor_code));
96 DEFINE(__LC_SUBCHANNEL_ID, offsetof(struct _lowcore, subchannel_id)); 101 DEFINE(__LC_SUBCHANNEL_ID, offsetof(struct _lowcore, subchannel_id));
97 DEFINE(__LC_SUBCHANNEL_NR, offsetof(struct _lowcore, subchannel_nr)); 102 DEFINE(__LC_SUBCHANNEL_NR, offsetof(struct _lowcore, subchannel_nr));
98 DEFINE(__LC_IO_INT_PARM, offsetof(struct _lowcore, io_int_parm)); 103 DEFINE(__LC_IO_INT_PARM, offsetof(struct _lowcore, io_int_parm));
99 DEFINE(__LC_IO_INT_WORD, offsetof(struct _lowcore, io_int_word)); 104 DEFINE(__LC_IO_INT_WORD, offsetof(struct _lowcore, io_int_word));
100 DEFINE(__LC_STFL_FAC_LIST, offsetof(struct _lowcore, stfl_fac_list)); 105 DEFINE(__LC_STFL_FAC_LIST, offsetof(struct _lowcore, stfl_fac_list));
101 DEFINE(__LC_MCCK_CODE, offsetof(struct _lowcore, mcck_interruption_code)); 106 DEFINE(__LC_MCCK_CODE, offsetof(struct _lowcore, mcck_interruption_code));
107 DEFINE(__LC_MCCK_EXT_DAM_CODE, offsetof(struct _lowcore, external_damage_code));
102 DEFINE(__LC_RST_OLD_PSW, offsetof(struct _lowcore, restart_old_psw)); 108 DEFINE(__LC_RST_OLD_PSW, offsetof(struct _lowcore, restart_old_psw));
103 DEFINE(__LC_EXT_OLD_PSW, offsetof(struct _lowcore, external_old_psw)); 109 DEFINE(__LC_EXT_OLD_PSW, offsetof(struct _lowcore, external_old_psw));
104 DEFINE(__LC_SVC_OLD_PSW, offsetof(struct _lowcore, svc_old_psw)); 110 DEFINE(__LC_SVC_OLD_PSW, offsetof(struct _lowcore, svc_old_psw));
@@ -156,6 +162,8 @@ int main(void)
156#ifdef CONFIG_32BIT 162#ifdef CONFIG_32BIT
157 DEFINE(SAVE_AREA_BASE, offsetof(struct _lowcore, extended_save_area_addr)); 163 DEFINE(SAVE_AREA_BASE, offsetof(struct _lowcore, extended_save_area_addr));
158#else /* CONFIG_32BIT */ 164#else /* CONFIG_32BIT */
165 DEFINE(__LC_DATA_EXC_CODE, offsetof(struct _lowcore, data_exc_code));
166 DEFINE(__LC_MCCK_FAIL_STOR_ADDR, offsetof(struct _lowcore, failing_storage_address));
159 DEFINE(__LC_EXT_PARAMS2, offsetof(struct _lowcore, ext_params2)); 167 DEFINE(__LC_EXT_PARAMS2, offsetof(struct _lowcore, ext_params2));
160 DEFINE(SAVE_AREA_BASE, offsetof(struct _lowcore, floating_pt_save_area)); 168 DEFINE(SAVE_AREA_BASE, offsetof(struct _lowcore, floating_pt_save_area));
161 DEFINE(__LC_PASTE, offsetof(struct _lowcore, paste)); 169 DEFINE(__LC_PASTE, offsetof(struct _lowcore, paste));
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index 1662038516c0..e66f046b9c43 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -391,8 +391,8 @@ ENTRY(pgm_check_handler)
391 jz pgm_kprobe 391 jz pgm_kprobe
392 oi __TI_flags+3(%r12),_TIF_PER_TRAP 392 oi __TI_flags+3(%r12),_TIF_PER_TRAP
393 mvc __THREAD_per_address(4,%r1),__LC_PER_ADDRESS 393 mvc __THREAD_per_address(4,%r1),__LC_PER_ADDRESS
394 mvc __THREAD_per_cause(2,%r1),__LC_PER_CAUSE 394 mvc __THREAD_per_cause(2,%r1),__LC_PER_CODE
395 mvc __THREAD_per_paid(1,%r1),__LC_PER_PAID 395 mvc __THREAD_per_paid(1,%r1),__LC_PER_ACCESS_ID
3960: REENABLE_IRQS 3960: REENABLE_IRQS
397 xc __SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15) 397 xc __SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
398 l %r1,BASED(.Ljump_table) 398 l %r1,BASED(.Ljump_table)
diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S
index 5963e43618bb..3c34753de6ad 100644
--- a/arch/s390/kernel/entry64.S
+++ b/arch/s390/kernel/entry64.S
@@ -423,8 +423,8 @@ ENTRY(pgm_check_handler)
423 jz pgm_kprobe 423 jz pgm_kprobe
424 oi __TI_flags+7(%r12),_TIF_PER_TRAP 424 oi __TI_flags+7(%r12),_TIF_PER_TRAP
425 mvc __THREAD_per_address(8,%r14),__LC_PER_ADDRESS 425 mvc __THREAD_per_address(8,%r14),__LC_PER_ADDRESS
426 mvc __THREAD_per_cause(2,%r14),__LC_PER_CAUSE 426 mvc __THREAD_per_cause(2,%r14),__LC_PER_CODE
427 mvc __THREAD_per_paid(1,%r14),__LC_PER_PAID 427 mvc __THREAD_per_paid(1,%r14),__LC_PER_ACCESS_ID
4280: REENABLE_IRQS 4280: REENABLE_IRQS
429 xc __SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15) 429 xc __SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
430 larl %r1,pgm_check_table 430 larl %r1,pgm_check_table
diff --git a/arch/s390/kvm/Makefile b/arch/s390/kvm/Makefile
index d3adb37e93a4..b3b553469650 100644
--- a/arch/s390/kvm/Makefile
+++ b/arch/s390/kvm/Makefile
@@ -11,5 +11,7 @@ common-objs = $(KVM)/kvm_main.o $(KVM)/eventfd.o $(KVM)/async_pf.o $(KVM)/irqch
11 11
12ccflags-y := -Ivirt/kvm -Iarch/s390/kvm 12ccflags-y := -Ivirt/kvm -Iarch/s390/kvm
13 13
14kvm-objs := $(common-objs) kvm-s390.o intercept.o interrupt.o priv.o sigp.o diag.o 14kvm-objs := $(common-objs) kvm-s390.o intercept.o interrupt.o priv.o sigp.o
15kvm-objs += diag.o gaccess.o guestdbg.o
16
15obj-$(CONFIG_KVM) += kvm.o 17obj-$(CONFIG_KVM) += kvm.o
diff --git a/arch/s390/kvm/diag.c b/arch/s390/kvm/diag.c
index 08dfc839a6cf..5521ace8b60d 100644
--- a/arch/s390/kvm/diag.c
+++ b/arch/s390/kvm/diag.c
@@ -64,12 +64,12 @@ static int __diag_page_ref_service(struct kvm_vcpu *vcpu)
64 int rc; 64 int rc;
65 u16 rx = (vcpu->arch.sie_block->ipa & 0xf0) >> 4; 65 u16 rx = (vcpu->arch.sie_block->ipa & 0xf0) >> 4;
66 u16 ry = (vcpu->arch.sie_block->ipa & 0x0f); 66 u16 ry = (vcpu->arch.sie_block->ipa & 0x0f);
67 unsigned long hva_token = KVM_HVA_ERR_BAD;
68 67
69 if (vcpu->run->s.regs.gprs[rx] & 7) 68 if (vcpu->run->s.regs.gprs[rx] & 7)
70 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 69 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
71 if (copy_from_guest(vcpu, &parm, vcpu->run->s.regs.gprs[rx], sizeof(parm))) 70 rc = read_guest(vcpu, vcpu->run->s.regs.gprs[rx], &parm, sizeof(parm));
72 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 71 if (rc)
72 return kvm_s390_inject_prog_cond(vcpu, rc);
73 if (parm.parm_version != 2 || parm.parm_len < 5 || parm.code != 0x258) 73 if (parm.parm_version != 2 || parm.parm_len < 5 || parm.code != 0x258)
74 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 74 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
75 75
@@ -89,8 +89,7 @@ static int __diag_page_ref_service(struct kvm_vcpu *vcpu)
89 parm.token_addr & 7 || parm.zarch != 0x8000000000000000ULL) 89 parm.token_addr & 7 || parm.zarch != 0x8000000000000000ULL)
90 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 90 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
91 91
92 hva_token = gfn_to_hva(vcpu->kvm, gpa_to_gfn(parm.token_addr)); 92 if (kvm_is_error_gpa(vcpu->kvm, parm.token_addr))
93 if (kvm_is_error_hva(hva_token))
94 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 93 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
95 94
96 vcpu->arch.pfault_token = parm.token_addr; 95 vcpu->arch.pfault_token = parm.token_addr;
@@ -167,17 +166,11 @@ static int __diag_ipl_functions(struct kvm_vcpu *vcpu)
167 166
168 VCPU_EVENT(vcpu, 5, "diag ipl functions, subcode %lx", subcode); 167 VCPU_EVENT(vcpu, 5, "diag ipl functions, subcode %lx", subcode);
169 switch (subcode) { 168 switch (subcode) {
170 case 0:
171 case 1:
172 page_table_reset_pgste(current->mm, 0, TASK_SIZE);
173 return -EOPNOTSUPP;
174 case 3: 169 case 3:
175 vcpu->run->s390_reset_flags = KVM_S390_RESET_CLEAR; 170 vcpu->run->s390_reset_flags = KVM_S390_RESET_CLEAR;
176 page_table_reset_pgste(current->mm, 0, TASK_SIZE);
177 break; 171 break;
178 case 4: 172 case 4:
179 vcpu->run->s390_reset_flags = 0; 173 vcpu->run->s390_reset_flags = 0;
180 page_table_reset_pgste(current->mm, 0, TASK_SIZE);
181 break; 174 break;
182 default: 175 default:
183 return -EOPNOTSUPP; 176 return -EOPNOTSUPP;
diff --git a/arch/s390/kvm/gaccess.c b/arch/s390/kvm/gaccess.c
new file mode 100644
index 000000000000..691fdb776c90
--- /dev/null
+++ b/arch/s390/kvm/gaccess.c
@@ -0,0 +1,645 @@
1/*
2 * guest access functions
3 *
4 * Copyright IBM Corp. 2014
5 *
6 */
7
8#include <linux/vmalloc.h>
9#include <linux/err.h>
10#include <asm/pgtable.h>
11#include "kvm-s390.h"
12#include "gaccess.h"
13
14union asce {
15 unsigned long val;
16 struct {
17 unsigned long origin : 52; /* Region- or Segment-Table Origin */
18 unsigned long : 2;
19 unsigned long g : 1; /* Subspace Group Control */
20 unsigned long p : 1; /* Private Space Control */
21 unsigned long s : 1; /* Storage-Alteration-Event Control */
22 unsigned long x : 1; /* Space-Switch-Event Control */
23 unsigned long r : 1; /* Real-Space Control */
24 unsigned long : 1;
25 unsigned long dt : 2; /* Designation-Type Control */
26 unsigned long tl : 2; /* Region- or Segment-Table Length */
27 };
28};
29
30enum {
31 ASCE_TYPE_SEGMENT = 0,
32 ASCE_TYPE_REGION3 = 1,
33 ASCE_TYPE_REGION2 = 2,
34 ASCE_TYPE_REGION1 = 3
35};
36
37union region1_table_entry {
38 unsigned long val;
39 struct {
40 unsigned long rto: 52;/* Region-Table Origin */
41 unsigned long : 2;
42 unsigned long p : 1; /* DAT-Protection Bit */
43 unsigned long : 1;
44 unsigned long tf : 2; /* Region-Second-Table Offset */
45 unsigned long i : 1; /* Region-Invalid Bit */
46 unsigned long : 1;
47 unsigned long tt : 2; /* Table-Type Bits */
48 unsigned long tl : 2; /* Region-Second-Table Length */
49 };
50};
51
52union region2_table_entry {
53 unsigned long val;
54 struct {
55 unsigned long rto: 52;/* Region-Table Origin */
56 unsigned long : 2;
57 unsigned long p : 1; /* DAT-Protection Bit */
58 unsigned long : 1;
59 unsigned long tf : 2; /* Region-Third-Table Offset */
60 unsigned long i : 1; /* Region-Invalid Bit */
61 unsigned long : 1;
62 unsigned long tt : 2; /* Table-Type Bits */
63 unsigned long tl : 2; /* Region-Third-Table Length */
64 };
65};
66
67struct region3_table_entry_fc0 {
68 unsigned long sto: 52;/* Segment-Table Origin */
69 unsigned long : 1;
70 unsigned long fc : 1; /* Format-Control */
71 unsigned long p : 1; /* DAT-Protection Bit */
72 unsigned long : 1;
73 unsigned long tf : 2; /* Segment-Table Offset */
74 unsigned long i : 1; /* Region-Invalid Bit */
75 unsigned long cr : 1; /* Common-Region Bit */
76 unsigned long tt : 2; /* Table-Type Bits */
77 unsigned long tl : 2; /* Segment-Table Length */
78};
79
80struct region3_table_entry_fc1 {
81 unsigned long rfaa : 33; /* Region-Frame Absolute Address */
82 unsigned long : 14;
83 unsigned long av : 1; /* ACCF-Validity Control */
84 unsigned long acc: 4; /* Access-Control Bits */
85 unsigned long f : 1; /* Fetch-Protection Bit */
86 unsigned long fc : 1; /* Format-Control */
87 unsigned long p : 1; /* DAT-Protection Bit */
88 unsigned long co : 1; /* Change-Recording Override */
89 unsigned long : 2;
90 unsigned long i : 1; /* Region-Invalid Bit */
91 unsigned long cr : 1; /* Common-Region Bit */
92 unsigned long tt : 2; /* Table-Type Bits */
93 unsigned long : 2;
94};
95
96union region3_table_entry {
97 unsigned long val;
98 struct region3_table_entry_fc0 fc0;
99 struct region3_table_entry_fc1 fc1;
100 struct {
101 unsigned long : 53;
102 unsigned long fc : 1; /* Format-Control */
103 unsigned long : 4;
104 unsigned long i : 1; /* Region-Invalid Bit */
105 unsigned long cr : 1; /* Common-Region Bit */
106 unsigned long tt : 2; /* Table-Type Bits */
107 unsigned long : 2;
108 };
109};
110
111struct segment_entry_fc0 {
112 unsigned long pto: 53;/* Page-Table Origin */
113 unsigned long fc : 1; /* Format-Control */
114 unsigned long p : 1; /* DAT-Protection Bit */
115 unsigned long : 3;
116 unsigned long i : 1; /* Segment-Invalid Bit */
117 unsigned long cs : 1; /* Common-Segment Bit */
118 unsigned long tt : 2; /* Table-Type Bits */
119 unsigned long : 2;
120};
121
122struct segment_entry_fc1 {
123 unsigned long sfaa : 44; /* Segment-Frame Absolute Address */
124 unsigned long : 3;
125 unsigned long av : 1; /* ACCF-Validity Control */
126 unsigned long acc: 4; /* Access-Control Bits */
127 unsigned long f : 1; /* Fetch-Protection Bit */
128 unsigned long fc : 1; /* Format-Control */
129 unsigned long p : 1; /* DAT-Protection Bit */
130 unsigned long co : 1; /* Change-Recording Override */
131 unsigned long : 2;
132 unsigned long i : 1; /* Segment-Invalid Bit */
133 unsigned long cs : 1; /* Common-Segment Bit */
134 unsigned long tt : 2; /* Table-Type Bits */
135 unsigned long : 2;
136};
137
138union segment_table_entry {
139 unsigned long val;
140 struct segment_entry_fc0 fc0;
141 struct segment_entry_fc1 fc1;
142 struct {
143 unsigned long : 53;
144 unsigned long fc : 1; /* Format-Control */
145 unsigned long : 4;
146 unsigned long i : 1; /* Segment-Invalid Bit */
147 unsigned long cs : 1; /* Common-Segment Bit */
148 unsigned long tt : 2; /* Table-Type Bits */
149 unsigned long : 2;
150 };
151};
152
153enum {
154 TABLE_TYPE_SEGMENT = 0,
155 TABLE_TYPE_REGION3 = 1,
156 TABLE_TYPE_REGION2 = 2,
157 TABLE_TYPE_REGION1 = 3
158};
159
160union page_table_entry {
161 unsigned long val;
162 struct {
163 unsigned long pfra : 52; /* Page-Frame Real Address */
164 unsigned long z : 1; /* Zero Bit */
165 unsigned long i : 1; /* Page-Invalid Bit */
166 unsigned long p : 1; /* DAT-Protection Bit */
167 unsigned long co : 1; /* Change-Recording Override */
168 unsigned long : 8;
169 };
170};
171
172/*
173 * vaddress union in order to easily decode a virtual address into its
174 * region first index, region second index etc. parts.
175 */
176union vaddress {
177 unsigned long addr;
178 struct {
179 unsigned long rfx : 11;
180 unsigned long rsx : 11;
181 unsigned long rtx : 11;
182 unsigned long sx : 11;
183 unsigned long px : 8;
184 unsigned long bx : 12;
185 };
186 struct {
187 unsigned long rfx01 : 2;
188 unsigned long : 9;
189 unsigned long rsx01 : 2;
190 unsigned long : 9;
191 unsigned long rtx01 : 2;
192 unsigned long : 9;
193 unsigned long sx01 : 2;
194 unsigned long : 29;
195 };
196};
197
198/*
199 * raddress union which will contain the result (real or absolute address)
200 * after a page table walk. The rfaa, sfaa and pfra members are used to
201 * simply assign them the value of a region, segment or page table entry.
202 */
203union raddress {
204 unsigned long addr;
205 unsigned long rfaa : 33; /* Region-Frame Absolute Address */
206 unsigned long sfaa : 44; /* Segment-Frame Absolute Address */
207 unsigned long pfra : 52; /* Page-Frame Real Address */
208};
209
210static int ipte_lock_count;
211static DEFINE_MUTEX(ipte_mutex);
212
213int ipte_lock_held(struct kvm_vcpu *vcpu)
214{
215 union ipte_control *ic = &vcpu->kvm->arch.sca->ipte_control;
216
217 if (vcpu->arch.sie_block->eca & 1)
218 return ic->kh != 0;
219 return ipte_lock_count != 0;
220}
221
222static void ipte_lock_simple(struct kvm_vcpu *vcpu)
223{
224 union ipte_control old, new, *ic;
225
226 mutex_lock(&ipte_mutex);
227 ipte_lock_count++;
228 if (ipte_lock_count > 1)
229 goto out;
230 ic = &vcpu->kvm->arch.sca->ipte_control;
231 do {
232 old = ACCESS_ONCE(*ic);
233 while (old.k) {
234 cond_resched();
235 old = ACCESS_ONCE(*ic);
236 }
237 new = old;
238 new.k = 1;
239 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
240out:
241 mutex_unlock(&ipte_mutex);
242}
243
244static void ipte_unlock_simple(struct kvm_vcpu *vcpu)
245{
246 union ipte_control old, new, *ic;
247
248 mutex_lock(&ipte_mutex);
249 ipte_lock_count--;
250 if (ipte_lock_count)
251 goto out;
252 ic = &vcpu->kvm->arch.sca->ipte_control;
253 do {
254 new = old = ACCESS_ONCE(*ic);
255 new.k = 0;
256 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
257 if (!ipte_lock_count)
258 wake_up(&vcpu->kvm->arch.ipte_wq);
259out:
260 mutex_unlock(&ipte_mutex);
261}
262
263static void ipte_lock_siif(struct kvm_vcpu *vcpu)
264{
265 union ipte_control old, new, *ic;
266
267 ic = &vcpu->kvm->arch.sca->ipte_control;
268 do {
269 old = ACCESS_ONCE(*ic);
270 while (old.kg) {
271 cond_resched();
272 old = ACCESS_ONCE(*ic);
273 }
274 new = old;
275 new.k = 1;
276 new.kh++;
277 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
278}
279
280static void ipte_unlock_siif(struct kvm_vcpu *vcpu)
281{
282 union ipte_control old, new, *ic;
283
284 ic = &vcpu->kvm->arch.sca->ipte_control;
285 do {
286 new = old = ACCESS_ONCE(*ic);
287 new.kh--;
288 if (!new.kh)
289 new.k = 0;
290 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
291 if (!new.kh)
292 wake_up(&vcpu->kvm->arch.ipte_wq);
293}
294
295static void ipte_lock(struct kvm_vcpu *vcpu)
296{
297 if (vcpu->arch.sie_block->eca & 1)
298 ipte_lock_siif(vcpu);
299 else
300 ipte_lock_simple(vcpu);
301}
302
303static void ipte_unlock(struct kvm_vcpu *vcpu)
304{
305 if (vcpu->arch.sie_block->eca & 1)
306 ipte_unlock_siif(vcpu);
307 else
308 ipte_unlock_simple(vcpu);
309}
310
311static unsigned long get_vcpu_asce(struct kvm_vcpu *vcpu)
312{
313 switch (psw_bits(vcpu->arch.sie_block->gpsw).as) {
314 case PSW_AS_PRIMARY:
315 return vcpu->arch.sie_block->gcr[1];
316 case PSW_AS_SECONDARY:
317 return vcpu->arch.sie_block->gcr[7];
318 case PSW_AS_HOME:
319 return vcpu->arch.sie_block->gcr[13];
320 }
321 return 0;
322}
323
324static int deref_table(struct kvm *kvm, unsigned long gpa, unsigned long *val)
325{
326 return kvm_read_guest(kvm, gpa, val, sizeof(*val));
327}
328
329/**
330 * guest_translate - translate a guest virtual into a guest absolute address
331 * @vcpu: virtual cpu
332 * @gva: guest virtual address
333 * @gpa: points to where guest physical (absolute) address should be stored
334 * @write: indicates if access is a write access
335 *
336 * Translate a guest virtual address into a guest absolute address by means
337 * of dynamic address translation as specified by the architecuture.
338 * If the resulting absolute address is not available in the configuration
339 * an addressing exception is indicated and @gpa will not be changed.
340 *
341 * Returns: - zero on success; @gpa contains the resulting absolute address
342 * - a negative value if guest access failed due to e.g. broken
343 * guest mapping
344 * - a positve value if an access exception happened. In this case
345 * the returned value is the program interruption code as defined
346 * by the architecture
347 */
348static unsigned long guest_translate(struct kvm_vcpu *vcpu, unsigned long gva,
349 unsigned long *gpa, int write)
350{
351 union vaddress vaddr = {.addr = gva};
352 union raddress raddr = {.addr = gva};
353 union page_table_entry pte;
354 int dat_protection = 0;
355 union ctlreg0 ctlreg0;
356 unsigned long ptr;
357 int edat1, edat2;
358 union asce asce;
359
360 ctlreg0.val = vcpu->arch.sie_block->gcr[0];
361 edat1 = ctlreg0.edat && test_vfacility(8);
362 edat2 = edat1 && test_vfacility(78);
363 asce.val = get_vcpu_asce(vcpu);
364 if (asce.r)
365 goto real_address;
366 ptr = asce.origin * 4096;
367 switch (asce.dt) {
368 case ASCE_TYPE_REGION1:
369 if (vaddr.rfx01 > asce.tl)
370 return PGM_REGION_FIRST_TRANS;
371 ptr += vaddr.rfx * 8;
372 break;
373 case ASCE_TYPE_REGION2:
374 if (vaddr.rfx)
375 return PGM_ASCE_TYPE;
376 if (vaddr.rsx01 > asce.tl)
377 return PGM_REGION_SECOND_TRANS;
378 ptr += vaddr.rsx * 8;
379 break;
380 case ASCE_TYPE_REGION3:
381 if (vaddr.rfx || vaddr.rsx)
382 return PGM_ASCE_TYPE;
383 if (vaddr.rtx01 > asce.tl)
384 return PGM_REGION_THIRD_TRANS;
385 ptr += vaddr.rtx * 8;
386 break;
387 case ASCE_TYPE_SEGMENT:
388 if (vaddr.rfx || vaddr.rsx || vaddr.rtx)
389 return PGM_ASCE_TYPE;
390 if (vaddr.sx01 > asce.tl)
391 return PGM_SEGMENT_TRANSLATION;
392 ptr += vaddr.sx * 8;
393 break;
394 }
395 switch (asce.dt) {
396 case ASCE_TYPE_REGION1: {
397 union region1_table_entry rfte;
398
399 if (kvm_is_error_gpa(vcpu->kvm, ptr))
400 return PGM_ADDRESSING;
401 if (deref_table(vcpu->kvm, ptr, &rfte.val))
402 return -EFAULT;
403 if (rfte.i)
404 return PGM_REGION_FIRST_TRANS;
405 if (rfte.tt != TABLE_TYPE_REGION1)
406 return PGM_TRANSLATION_SPEC;
407 if (vaddr.rsx01 < rfte.tf || vaddr.rsx01 > rfte.tl)
408 return PGM_REGION_SECOND_TRANS;
409 if (edat1)
410 dat_protection |= rfte.p;
411 ptr = rfte.rto * 4096 + vaddr.rsx * 8;
412 }
413 /* fallthrough */
414 case ASCE_TYPE_REGION2: {
415 union region2_table_entry rste;
416
417 if (kvm_is_error_gpa(vcpu->kvm, ptr))
418 return PGM_ADDRESSING;
419 if (deref_table(vcpu->kvm, ptr, &rste.val))
420 return -EFAULT;
421 if (rste.i)
422 return PGM_REGION_SECOND_TRANS;
423 if (rste.tt != TABLE_TYPE_REGION2)
424 return PGM_TRANSLATION_SPEC;
425 if (vaddr.rtx01 < rste.tf || vaddr.rtx01 > rste.tl)
426 return PGM_REGION_THIRD_TRANS;
427 if (edat1)
428 dat_protection |= rste.p;
429 ptr = rste.rto * 4096 + vaddr.rtx * 8;
430 }
431 /* fallthrough */
432 case ASCE_TYPE_REGION3: {
433 union region3_table_entry rtte;
434
435 if (kvm_is_error_gpa(vcpu->kvm, ptr))
436 return PGM_ADDRESSING;
437 if (deref_table(vcpu->kvm, ptr, &rtte.val))
438 return -EFAULT;
439 if (rtte.i)
440 return PGM_REGION_THIRD_TRANS;
441 if (rtte.tt != TABLE_TYPE_REGION3)
442 return PGM_TRANSLATION_SPEC;
443 if (rtte.cr && asce.p && edat2)
444 return PGM_TRANSLATION_SPEC;
445 if (rtte.fc && edat2) {
446 dat_protection |= rtte.fc1.p;
447 raddr.rfaa = rtte.fc1.rfaa;
448 goto absolute_address;
449 }
450 if (vaddr.sx01 < rtte.fc0.tf)
451 return PGM_SEGMENT_TRANSLATION;
452 if (vaddr.sx01 > rtte.fc0.tl)
453 return PGM_SEGMENT_TRANSLATION;
454 if (edat1)
455 dat_protection |= rtte.fc0.p;
456 ptr = rtte.fc0.sto * 4096 + vaddr.sx * 8;
457 }
458 /* fallthrough */
459 case ASCE_TYPE_SEGMENT: {
460 union segment_table_entry ste;
461
462 if (kvm_is_error_gpa(vcpu->kvm, ptr))
463 return PGM_ADDRESSING;
464 if (deref_table(vcpu->kvm, ptr, &ste.val))
465 return -EFAULT;
466 if (ste.i)
467 return PGM_SEGMENT_TRANSLATION;
468 if (ste.tt != TABLE_TYPE_SEGMENT)
469 return PGM_TRANSLATION_SPEC;
470 if (ste.cs && asce.p)
471 return PGM_TRANSLATION_SPEC;
472 if (ste.fc && edat1) {
473 dat_protection |= ste.fc1.p;
474 raddr.sfaa = ste.fc1.sfaa;
475 goto absolute_address;
476 }
477 dat_protection |= ste.fc0.p;
478 ptr = ste.fc0.pto * 2048 + vaddr.px * 8;
479 }
480 }
481 if (kvm_is_error_gpa(vcpu->kvm, ptr))
482 return PGM_ADDRESSING;
483 if (deref_table(vcpu->kvm, ptr, &pte.val))
484 return -EFAULT;
485 if (pte.i)
486 return PGM_PAGE_TRANSLATION;
487 if (pte.z)
488 return PGM_TRANSLATION_SPEC;
489 if (pte.co && !edat1)
490 return PGM_TRANSLATION_SPEC;
491 dat_protection |= pte.p;
492 raddr.pfra = pte.pfra;
493real_address:
494 raddr.addr = kvm_s390_real_to_abs(vcpu, raddr.addr);
495absolute_address:
496 if (write && dat_protection)
497 return PGM_PROTECTION;
498 if (kvm_is_error_gpa(vcpu->kvm, raddr.addr))
499 return PGM_ADDRESSING;
500 *gpa = raddr.addr;
501 return 0;
502}
503
504static inline int is_low_address(unsigned long ga)
505{
506 /* Check for address ranges 0..511 and 4096..4607 */
507 return (ga & ~0x11fful) == 0;
508}
509
510static int low_address_protection_enabled(struct kvm_vcpu *vcpu)
511{
512 union ctlreg0 ctlreg0 = {.val = vcpu->arch.sie_block->gcr[0]};
513 psw_t *psw = &vcpu->arch.sie_block->gpsw;
514 union asce asce;
515
516 if (!ctlreg0.lap)
517 return 0;
518 asce.val = get_vcpu_asce(vcpu);
519 if (psw_bits(*psw).t && asce.p)
520 return 0;
521 return 1;
522}
523
524struct trans_exc_code_bits {
525 unsigned long addr : 52; /* Translation-exception Address */
526 unsigned long fsi : 2; /* Access Exception Fetch/Store Indication */
527 unsigned long : 7;
528 unsigned long b61 : 1;
529 unsigned long as : 2; /* ASCE Identifier */
530};
531
532enum {
533 FSI_UNKNOWN = 0, /* Unknown wether fetch or store */
534 FSI_STORE = 1, /* Exception was due to store operation */
535 FSI_FETCH = 2 /* Exception was due to fetch operation */
536};
537
538static int guest_page_range(struct kvm_vcpu *vcpu, unsigned long ga,
539 unsigned long *pages, unsigned long nr_pages,
540 int write)
541{
542 struct kvm_s390_pgm_info *pgm = &vcpu->arch.pgm;
543 psw_t *psw = &vcpu->arch.sie_block->gpsw;
544 struct trans_exc_code_bits *tec_bits;
545 int lap_enabled, rc;
546
547 memset(pgm, 0, sizeof(*pgm));
548 tec_bits = (struct trans_exc_code_bits *)&pgm->trans_exc_code;
549 tec_bits->fsi = write ? FSI_STORE : FSI_FETCH;
550 tec_bits->as = psw_bits(*psw).as;
551 lap_enabled = low_address_protection_enabled(vcpu);
552 while (nr_pages) {
553 ga = kvm_s390_logical_to_effective(vcpu, ga);
554 tec_bits->addr = ga >> PAGE_SHIFT;
555 if (write && lap_enabled && is_low_address(ga)) {
556 pgm->code = PGM_PROTECTION;
557 return pgm->code;
558 }
559 ga &= PAGE_MASK;
560 if (psw_bits(*psw).t) {
561 rc = guest_translate(vcpu, ga, pages, write);
562 if (rc < 0)
563 return rc;
564 if (rc == PGM_PROTECTION)
565 tec_bits->b61 = 1;
566 if (rc)
567 pgm->code = rc;
568 } else {
569 *pages = kvm_s390_real_to_abs(vcpu, ga);
570 if (kvm_is_error_gpa(vcpu->kvm, *pages))
571 pgm->code = PGM_ADDRESSING;
572 }
573 if (pgm->code)
574 return pgm->code;
575 ga += PAGE_SIZE;
576 pages++;
577 nr_pages--;
578 }
579 return 0;
580}
581
582int access_guest(struct kvm_vcpu *vcpu, unsigned long ga, void *data,
583 unsigned long len, int write)
584{
585 psw_t *psw = &vcpu->arch.sie_block->gpsw;
586 unsigned long _len, nr_pages, gpa, idx;
587 unsigned long pages_array[2];
588 unsigned long *pages;
589 int need_ipte_lock;
590 union asce asce;
591 int rc;
592
593 if (!len)
594 return 0;
595 /* Access register mode is not supported yet. */
596 if (psw_bits(*psw).t && psw_bits(*psw).as == PSW_AS_ACCREG)
597 return -EOPNOTSUPP;
598 nr_pages = (((ga & ~PAGE_MASK) + len - 1) >> PAGE_SHIFT) + 1;
599 pages = pages_array;
600 if (nr_pages > ARRAY_SIZE(pages_array))
601 pages = vmalloc(nr_pages * sizeof(unsigned long));
602 if (!pages)
603 return -ENOMEM;
604 asce.val = get_vcpu_asce(vcpu);
605 need_ipte_lock = psw_bits(*psw).t && !asce.r;
606 if (need_ipte_lock)
607 ipte_lock(vcpu);
608 rc = guest_page_range(vcpu, ga, pages, nr_pages, write);
609 for (idx = 0; idx < nr_pages && !rc; idx++) {
610 gpa = *(pages + idx) + (ga & ~PAGE_MASK);
611 _len = min(PAGE_SIZE - (gpa & ~PAGE_MASK), len);
612 if (write)
613 rc = kvm_write_guest(vcpu->kvm, gpa, data, _len);
614 else
615 rc = kvm_read_guest(vcpu->kvm, gpa, data, _len);
616 len -= _len;
617 ga += _len;
618 data += _len;
619 }
620 if (need_ipte_lock)
621 ipte_unlock(vcpu);
622 if (nr_pages > ARRAY_SIZE(pages_array))
623 vfree(pages);
624 return rc;
625}
626
627int access_guest_real(struct kvm_vcpu *vcpu, unsigned long gra,
628 void *data, unsigned long len, int write)
629{
630 unsigned long _len, gpa;
631 int rc = 0;
632
633 while (len && !rc) {
634 gpa = kvm_s390_real_to_abs(vcpu, gra);
635 _len = min(PAGE_SIZE - (gpa & ~PAGE_MASK), len);
636 if (write)
637 rc = write_guest_abs(vcpu, gpa, data, _len);
638 else
639 rc = read_guest_abs(vcpu, gpa, data, _len);
640 len -= _len;
641 gra += _len;
642 data += _len;
643 }
644 return rc;
645}
diff --git a/arch/s390/kvm/gaccess.h b/arch/s390/kvm/gaccess.h
index 374a439ccc60..1079c8fc6d0d 100644
--- a/arch/s390/kvm/gaccess.h
+++ b/arch/s390/kvm/gaccess.h
@@ -1,7 +1,7 @@
1/* 1/*
2 * access guest memory 2 * access guest memory
3 * 3 *
4 * Copyright IBM Corp. 2008, 2009 4 * Copyright IBM Corp. 2008, 2014
5 * 5 *
6 * This program is free software; you can redistribute it and/or modify 6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License (version 2 only) 7 * it under the terms of the GNU General Public License (version 2 only)
@@ -15,100 +15,315 @@
15 15
16#include <linux/compiler.h> 16#include <linux/compiler.h>
17#include <linux/kvm_host.h> 17#include <linux/kvm_host.h>
18#include <asm/uaccess.h> 18#include <linux/uaccess.h>
19#include <linux/ptrace.h>
19#include "kvm-s390.h" 20#include "kvm-s390.h"
20 21
21/* Convert real to absolute address by applying the prefix of the CPU */ 22/**
23 * kvm_s390_real_to_abs - convert guest real address to guest absolute address
24 * @vcpu - guest virtual cpu
25 * @gra - guest real address
26 *
27 * Returns the guest absolute address that corresponds to the passed guest real
28 * address @gra of a virtual guest cpu by applying its prefix.
29 */
22static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu, 30static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
23 unsigned long gaddr) 31 unsigned long gra)
24{ 32{
25 unsigned long prefix = vcpu->arch.sie_block->prefix; 33 unsigned long prefix = vcpu->arch.sie_block->prefix;
26 if (gaddr < 2 * PAGE_SIZE) 34
27 gaddr += prefix; 35 if (gra < 2 * PAGE_SIZE)
28 else if (gaddr >= prefix && gaddr < prefix + 2 * PAGE_SIZE) 36 gra += prefix;
29 gaddr -= prefix; 37 else if (gra >= prefix && gra < prefix + 2 * PAGE_SIZE)
30 return gaddr; 38 gra -= prefix;
39 return gra;
31} 40}
32 41
33static inline void __user *__gptr_to_uptr(struct kvm_vcpu *vcpu, 42/**
34 void __user *gptr, 43 * kvm_s390_logical_to_effective - convert guest logical to effective address
35 int prefixing) 44 * @vcpu: guest virtual cpu
45 * @ga: guest logical address
46 *
47 * Convert a guest vcpu logical address to a guest vcpu effective address by
48 * applying the rules of the vcpu's addressing mode defined by PSW bits 31
49 * and 32 (extendended/basic addressing mode).
50 *
51 * Depending on the vcpu's addressing mode the upper 40 bits (24 bit addressing
52 * mode), 33 bits (31 bit addressing mode) or no bits (64 bit addressing mode)
53 * of @ga will be zeroed and the remaining bits will be returned.
54 */
55static inline unsigned long kvm_s390_logical_to_effective(struct kvm_vcpu *vcpu,
56 unsigned long ga)
36{ 57{
37 unsigned long gaddr = (unsigned long) gptr; 58 psw_t *psw = &vcpu->arch.sie_block->gpsw;
38 unsigned long uaddr; 59
39 60 if (psw_bits(*psw).eaba == PSW_AMODE_64BIT)
40 if (prefixing) 61 return ga;
41 gaddr = kvm_s390_real_to_abs(vcpu, gaddr); 62 if (psw_bits(*psw).eaba == PSW_AMODE_31BIT)
42 uaddr = gmap_fault(gaddr, vcpu->arch.gmap); 63 return ga & ((1UL << 31) - 1);
43 if (IS_ERR_VALUE(uaddr)) 64 return ga & ((1UL << 24) - 1);
44 uaddr = -EFAULT;
45 return (void __user *)uaddr;
46} 65}
47 66
48#define get_guest(vcpu, x, gptr) \ 67/*
49({ \ 68 * put_guest_lc, read_guest_lc and write_guest_lc are guest access functions
50 __typeof__(gptr) __uptr = __gptr_to_uptr(vcpu, gptr, 1);\ 69 * which shall only be used to access the lowcore of a vcpu.
51 int __mask = sizeof(__typeof__(*(gptr))) - 1; \ 70 * These functions should be used for e.g. interrupt handlers where no
52 int __ret; \ 71 * guest memory access protection facilities, like key or low address
53 \ 72 * protection, are applicable.
54 if (IS_ERR((void __force *)__uptr)) { \ 73 * At a later point guest vcpu lowcore access should happen via pinned
55 __ret = PTR_ERR((void __force *)__uptr); \ 74 * prefix pages, so that these pages can be accessed directly via the
56 } else { \ 75 * kernel mapping. All of these *_lc functions can be removed then.
57 BUG_ON((unsigned long)__uptr & __mask); \ 76 */
58 __ret = get_user(x, __uptr); \
59 } \
60 __ret; \
61})
62 77
63#define put_guest(vcpu, x, gptr) \ 78/**
79 * put_guest_lc - write a simple variable to a guest vcpu's lowcore
80 * @vcpu: virtual cpu
81 * @x: value to copy to guest
82 * @gra: vcpu's destination guest real address
83 *
84 * Copies a simple value from kernel space to a guest vcpu's lowcore.
85 * The size of the variable may be 1, 2, 4 or 8 bytes. The destination
86 * must be located in the vcpu's lowcore. Otherwise the result is undefined.
87 *
88 * Returns zero on success or -EFAULT on error.
89 *
90 * Note: an error indicates that either the kernel is out of memory or
91 * the guest memory mapping is broken. In any case the best solution
92 * would be to terminate the guest.
93 * It is wrong to inject a guest exception.
94 */
95#define put_guest_lc(vcpu, x, gra) \
64({ \ 96({ \
65 __typeof__(gptr) __uptr = __gptr_to_uptr(vcpu, gptr, 1);\ 97 struct kvm_vcpu *__vcpu = (vcpu); \
66 int __mask = sizeof(__typeof__(*(gptr))) - 1; \ 98 __typeof__(*(gra)) __x = (x); \
67 int __ret; \ 99 unsigned long __gpa; \
68 \ 100 \
69 if (IS_ERR((void __force *)__uptr)) { \ 101 __gpa = (unsigned long)(gra); \
70 __ret = PTR_ERR((void __force *)__uptr); \ 102 __gpa += __vcpu->arch.sie_block->prefix; \
71 } else { \ 103 kvm_write_guest(__vcpu->kvm, __gpa, &__x, sizeof(__x)); \
72 BUG_ON((unsigned long)__uptr & __mask); \
73 __ret = put_user(x, __uptr); \
74 } \
75 __ret; \
76}) 104})
77 105
78static inline int __copy_guest(struct kvm_vcpu *vcpu, unsigned long to, 106/**
79 unsigned long from, unsigned long len, 107 * write_guest_lc - copy data from kernel space to guest vcpu's lowcore
80 int to_guest, int prefixing) 108 * @vcpu: virtual cpu
109 * @gra: vcpu's source guest real address
110 * @data: source address in kernel space
111 * @len: number of bytes to copy
112 *
113 * Copy data from kernel space to guest vcpu's lowcore. The entire range must
114 * be located within the vcpu's lowcore, otherwise the result is undefined.
115 *
116 * Returns zero on success or -EFAULT on error.
117 *
118 * Note: an error indicates that either the kernel is out of memory or
119 * the guest memory mapping is broken. In any case the best solution
120 * would be to terminate the guest.
121 * It is wrong to inject a guest exception.
122 */
123static inline __must_check
124int write_guest_lc(struct kvm_vcpu *vcpu, unsigned long gra, void *data,
125 unsigned long len)
126{
127 unsigned long gpa = gra + vcpu->arch.sie_block->prefix;
128
129 return kvm_write_guest(vcpu->kvm, gpa, data, len);
130}
131
132/**
133 * read_guest_lc - copy data from guest vcpu's lowcore to kernel space
134 * @vcpu: virtual cpu
135 * @gra: vcpu's source guest real address
136 * @data: destination address in kernel space
137 * @len: number of bytes to copy
138 *
139 * Copy data from guest vcpu's lowcore to kernel space. The entire range must
140 * be located within the vcpu's lowcore, otherwise the result is undefined.
141 *
142 * Returns zero on success or -EFAULT on error.
143 *
144 * Note: an error indicates that either the kernel is out of memory or
145 * the guest memory mapping is broken. In any case the best solution
146 * would be to terminate the guest.
147 * It is wrong to inject a guest exception.
148 */
149static inline __must_check
150int read_guest_lc(struct kvm_vcpu *vcpu, unsigned long gra, void *data,
151 unsigned long len)
152{
153 unsigned long gpa = gra + vcpu->arch.sie_block->prefix;
154
155 return kvm_read_guest(vcpu->kvm, gpa, data, len);
156}
157
158int access_guest(struct kvm_vcpu *vcpu, unsigned long ga, void *data,
159 unsigned long len, int write);
160
161int access_guest_real(struct kvm_vcpu *vcpu, unsigned long gra,
162 void *data, unsigned long len, int write);
163
164/**
165 * write_guest - copy data from kernel space to guest space
166 * @vcpu: virtual cpu
167 * @ga: guest address
168 * @data: source address in kernel space
169 * @len: number of bytes to copy
170 *
171 * Copy @len bytes from @data (kernel space) to @ga (guest address).
172 * In order to copy data to guest space the PSW of the vcpu is inspected:
173 * If DAT is off data will be copied to guest real or absolute memory.
174 * If DAT is on data will be copied to the address space as specified by
175 * the address space bits of the PSW:
176 * Primary, secondory or home space (access register mode is currently not
177 * implemented).
178 * The addressing mode of the PSW is also inspected, so that address wrap
179 * around is taken into account for 24-, 31- and 64-bit addressing mode,
180 * if the to be copied data crosses page boundaries in guest address space.
181 * In addition also low address and DAT protection are inspected before
182 * copying any data (key protection is currently not implemented).
183 *
184 * This function modifies the 'struct kvm_s390_pgm_info pgm' member of @vcpu.
185 * In case of an access exception (e.g. protection exception) pgm will contain
186 * all data necessary so that a subsequent call to 'kvm_s390_inject_prog_vcpu()'
187 * will inject a correct exception into the guest.
188 * If no access exception happened, the contents of pgm are undefined when
189 * this function returns.
190 *
191 * Returns: - zero on success
192 * - a negative value if e.g. the guest mapping is broken or in
193 * case of out-of-memory. In this case the contents of pgm are
194 * undefined. Also parts of @data may have been copied to guest
195 * space.
196 * - a positive value if an access exception happened. In this case
197 * the returned value is the program interruption code and the
198 * contents of pgm may be used to inject an exception into the
199 * guest. No data has been copied to guest space.
200 *
201 * Note: in case an access exception is recognized no data has been copied to
202 * guest space (this is also true, if the to be copied data would cross
203 * one or more page boundaries in guest space).
204 * Therefore this function may be used for nullifying and suppressing
205 * instruction emulation.
206 * It may also be used for terminating instructions, if it is undefined
207 * if data has been changed in guest space in case of an exception.
208 */
209static inline __must_check
210int write_guest(struct kvm_vcpu *vcpu, unsigned long ga, void *data,
211 unsigned long len)
212{
213 return access_guest(vcpu, ga, data, len, 1);
214}
215
216/**
217 * read_guest - copy data from guest space to kernel space
218 * @vcpu: virtual cpu
219 * @ga: guest address
220 * @data: destination address in kernel space
221 * @len: number of bytes to copy
222 *
223 * Copy @len bytes from @ga (guest address) to @data (kernel space).
224 *
225 * The behaviour of read_guest is identical to write_guest, except that
226 * data will be copied from guest space to kernel space.
227 */
228static inline __must_check
229int read_guest(struct kvm_vcpu *vcpu, unsigned long ga, void *data,
230 unsigned long len)
231{
232 return access_guest(vcpu, ga, data, len, 0);
233}
234
235/**
236 * write_guest_abs - copy data from kernel space to guest space absolute
237 * @vcpu: virtual cpu
238 * @gpa: guest physical (absolute) address
239 * @data: source address in kernel space
240 * @len: number of bytes to copy
241 *
242 * Copy @len bytes from @data (kernel space) to @gpa (guest absolute address).
243 * It is up to the caller to ensure that the entire guest memory range is
244 * valid memory before calling this function.
245 * Guest low address and key protection are not checked.
246 *
247 * Returns zero on success or -EFAULT on error.
248 *
249 * If an error occurs data may have been copied partially to guest memory.
250 */
251static inline __must_check
252int write_guest_abs(struct kvm_vcpu *vcpu, unsigned long gpa, void *data,
253 unsigned long len)
254{
255 return kvm_write_guest(vcpu->kvm, gpa, data, len);
256}
257
258/**
259 * read_guest_abs - copy data from guest space absolute to kernel space
260 * @vcpu: virtual cpu
261 * @gpa: guest physical (absolute) address
262 * @data: destination address in kernel space
263 * @len: number of bytes to copy
264 *
265 * Copy @len bytes from @gpa (guest absolute address) to @data (kernel space).
266 * It is up to the caller to ensure that the entire guest memory range is
267 * valid memory before calling this function.
268 * Guest key protection is not checked.
269 *
270 * Returns zero on success or -EFAULT on error.
271 *
272 * If an error occurs data may have been copied partially to kernel space.
273 */
274static inline __must_check
275int read_guest_abs(struct kvm_vcpu *vcpu, unsigned long gpa, void *data,
276 unsigned long len)
277{
278 return kvm_read_guest(vcpu->kvm, gpa, data, len);
279}
280
281/**
282 * write_guest_real - copy data from kernel space to guest space real
283 * @vcpu: virtual cpu
284 * @gra: guest real address
285 * @data: source address in kernel space
286 * @len: number of bytes to copy
287 *
288 * Copy @len bytes from @data (kernel space) to @gra (guest real address).
289 * It is up to the caller to ensure that the entire guest memory range is
290 * valid memory before calling this function.
291 * Guest low address and key protection are not checked.
292 *
293 * Returns zero on success or -EFAULT on error.
294 *
295 * If an error occurs data may have been copied partially to guest memory.
296 */
297static inline __must_check
298int write_guest_real(struct kvm_vcpu *vcpu, unsigned long gra, void *data,
299 unsigned long len)
300{
301 return access_guest_real(vcpu, gra, data, len, 1);
302}
303
304/**
305 * read_guest_real - copy data from guest space real to kernel space
306 * @vcpu: virtual cpu
307 * @gra: guest real address
308 * @data: destination address in kernel space
309 * @len: number of bytes to copy
310 *
311 * Copy @len bytes from @gra (guest real address) to @data (kernel space).
312 * It is up to the caller to ensure that the entire guest memory range is
313 * valid memory before calling this function.
314 * Guest key protection is not checked.
315 *
316 * Returns zero on success or -EFAULT on error.
317 *
318 * If an error occurs data may have been copied partially to kernel space.
319 */
320static inline __must_check
321int read_guest_real(struct kvm_vcpu *vcpu, unsigned long gra, void *data,
322 unsigned long len)
81{ 323{
82 unsigned long _len, rc; 324 return access_guest_real(vcpu, gra, data, len, 0);
83 void __user *uptr;
84
85 while (len) {
86 uptr = to_guest ? (void __user *)to : (void __user *)from;
87 uptr = __gptr_to_uptr(vcpu, uptr, prefixing);
88 if (IS_ERR((void __force *)uptr))
89 return -EFAULT;
90 _len = PAGE_SIZE - ((unsigned long)uptr & (PAGE_SIZE - 1));
91 _len = min(_len, len);
92 if (to_guest)
93 rc = copy_to_user((void __user *) uptr, (void *)from, _len);
94 else
95 rc = copy_from_user((void *)to, (void __user *)uptr, _len);
96 if (rc)
97 return -EFAULT;
98 len -= _len;
99 from += _len;
100 to += _len;
101 }
102 return 0;
103} 325}
104 326
105#define copy_to_guest(vcpu, to, from, size) \ 327int ipte_lock_held(struct kvm_vcpu *vcpu);
106 __copy_guest(vcpu, to, (unsigned long)from, size, 1, 1)
107#define copy_from_guest(vcpu, to, from, size) \
108 __copy_guest(vcpu, (unsigned long)to, from, size, 0, 1)
109#define copy_to_guest_absolute(vcpu, to, from, size) \
110 __copy_guest(vcpu, to, (unsigned long)from, size, 1, 0)
111#define copy_from_guest_absolute(vcpu, to, from, size) \
112 __copy_guest(vcpu, (unsigned long)to, from, size, 0, 0)
113 328
114#endif /* __KVM_S390_GACCESS_H */ 329#endif /* __KVM_S390_GACCESS_H */
diff --git a/arch/s390/kvm/guestdbg.c b/arch/s390/kvm/guestdbg.c
new file mode 100644
index 000000000000..757ccef62fd5
--- /dev/null
+++ b/arch/s390/kvm/guestdbg.c
@@ -0,0 +1,481 @@
1/*
2 * kvm guest debug support
3 *
4 * Copyright IBM Corp. 2014
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License (version 2 only)
8 * as published by the Free Software Foundation.
9 *
10 * Author(s): David Hildenbrand <dahi@linux.vnet.ibm.com>
11 */
12#include <linux/kvm_host.h>
13#include <linux/errno.h>
14#include "kvm-s390.h"
15#include "gaccess.h"
16
17/*
18 * Extends the address range given by *start and *stop to include the address
19 * range starting with estart and the length len. Takes care of overflowing
20 * intervals and tries to minimize the overall intervall size.
21 */
22static void extend_address_range(u64 *start, u64 *stop, u64 estart, int len)
23{
24 u64 estop;
25
26 if (len > 0)
27 len--;
28 else
29 len = 0;
30
31 estop = estart + len;
32
33 /* 0-0 range represents "not set" */
34 if ((*start == 0) && (*stop == 0)) {
35 *start = estart;
36 *stop = estop;
37 } else if (*start <= *stop) {
38 /* increase the existing range */
39 if (estart < *start)
40 *start = estart;
41 if (estop > *stop)
42 *stop = estop;
43 } else {
44 /* "overflowing" interval, whereby *stop > *start */
45 if (estart <= *stop) {
46 if (estop > *stop)
47 *stop = estop;
48 } else if (estop > *start) {
49 if (estart < *start)
50 *start = estart;
51 }
52 /* minimize the range */
53 else if ((estop - *stop) < (*start - estart))
54 *stop = estop;
55 else
56 *start = estart;
57 }
58}
59
60#define MAX_INST_SIZE 6
61
62static void enable_all_hw_bp(struct kvm_vcpu *vcpu)
63{
64 unsigned long start, len;
65 u64 *cr9 = &vcpu->arch.sie_block->gcr[9];
66 u64 *cr10 = &vcpu->arch.sie_block->gcr[10];
67 u64 *cr11 = &vcpu->arch.sie_block->gcr[11];
68 int i;
69
70 if (vcpu->arch.guestdbg.nr_hw_bp <= 0 ||
71 vcpu->arch.guestdbg.hw_bp_info == NULL)
72 return;
73
74 /*
75 * If the guest is not interrested in branching events, we can savely
76 * limit them to the PER address range.
77 */
78 if (!(*cr9 & PER_EVENT_BRANCH))
79 *cr9 |= PER_CONTROL_BRANCH_ADDRESS;
80 *cr9 |= PER_EVENT_IFETCH | PER_EVENT_BRANCH;
81
82 for (i = 0; i < vcpu->arch.guestdbg.nr_hw_bp; i++) {
83 start = vcpu->arch.guestdbg.hw_bp_info[i].addr;
84 len = vcpu->arch.guestdbg.hw_bp_info[i].len;
85
86 /*
87 * The instruction in front of the desired bp has to
88 * report instruction-fetching events
89 */
90 if (start < MAX_INST_SIZE) {
91 len += start;
92 start = 0;
93 } else {
94 start -= MAX_INST_SIZE;
95 len += MAX_INST_SIZE;
96 }
97
98 extend_address_range(cr10, cr11, start, len);
99 }
100}
101
102static void enable_all_hw_wp(struct kvm_vcpu *vcpu)
103{
104 unsigned long start, len;
105 u64 *cr9 = &vcpu->arch.sie_block->gcr[9];
106 u64 *cr10 = &vcpu->arch.sie_block->gcr[10];
107 u64 *cr11 = &vcpu->arch.sie_block->gcr[11];
108 int i;
109
110 if (vcpu->arch.guestdbg.nr_hw_wp <= 0 ||
111 vcpu->arch.guestdbg.hw_wp_info == NULL)
112 return;
113
114 /* if host uses storage alternation for special address
115 * spaces, enable all events and give all to the guest */
116 if (*cr9 & PER_EVENT_STORE && *cr9 & PER_CONTROL_ALTERATION) {
117 *cr9 &= ~PER_CONTROL_ALTERATION;
118 *cr10 = 0;
119 *cr11 = PSW_ADDR_INSN;
120 } else {
121 *cr9 &= ~PER_CONTROL_ALTERATION;
122 *cr9 |= PER_EVENT_STORE;
123
124 for (i = 0; i < vcpu->arch.guestdbg.nr_hw_wp; i++) {
125 start = vcpu->arch.guestdbg.hw_wp_info[i].addr;
126 len = vcpu->arch.guestdbg.hw_wp_info[i].len;
127
128 extend_address_range(cr10, cr11, start, len);
129 }
130 }
131}
132
133void kvm_s390_backup_guest_per_regs(struct kvm_vcpu *vcpu)
134{
135 vcpu->arch.guestdbg.cr0 = vcpu->arch.sie_block->gcr[0];
136 vcpu->arch.guestdbg.cr9 = vcpu->arch.sie_block->gcr[9];
137 vcpu->arch.guestdbg.cr10 = vcpu->arch.sie_block->gcr[10];
138 vcpu->arch.guestdbg.cr11 = vcpu->arch.sie_block->gcr[11];
139}
140
141void kvm_s390_restore_guest_per_regs(struct kvm_vcpu *vcpu)
142{
143 vcpu->arch.sie_block->gcr[0] = vcpu->arch.guestdbg.cr0;
144 vcpu->arch.sie_block->gcr[9] = vcpu->arch.guestdbg.cr9;
145 vcpu->arch.sie_block->gcr[10] = vcpu->arch.guestdbg.cr10;
146 vcpu->arch.sie_block->gcr[11] = vcpu->arch.guestdbg.cr11;
147}
148
149void kvm_s390_patch_guest_per_regs(struct kvm_vcpu *vcpu)
150{
151 /*
152 * TODO: if guest psw has per enabled, otherwise 0s!
153 * This reduces the amount of reported events.
154 * Need to intercept all psw changes!
155 */
156
157 if (guestdbg_sstep_enabled(vcpu)) {
158 /* disable timer (clock-comparator) interrupts */
159 vcpu->arch.sie_block->gcr[0] &= ~0x800ul;
160 vcpu->arch.sie_block->gcr[9] |= PER_EVENT_IFETCH;
161 vcpu->arch.sie_block->gcr[10] = 0;
162 vcpu->arch.sie_block->gcr[11] = PSW_ADDR_INSN;
163 }
164
165 if (guestdbg_hw_bp_enabled(vcpu)) {
166 enable_all_hw_bp(vcpu);
167 enable_all_hw_wp(vcpu);
168 }
169
170 /* TODO: Instruction-fetching-nullification not allowed for now */
171 if (vcpu->arch.sie_block->gcr[9] & PER_EVENT_NULLIFICATION)
172 vcpu->arch.sie_block->gcr[9] &= ~PER_EVENT_NULLIFICATION;
173}
174
175#define MAX_WP_SIZE 100
176
177static int __import_wp_info(struct kvm_vcpu *vcpu,
178 struct kvm_hw_breakpoint *bp_data,
179 struct kvm_hw_wp_info_arch *wp_info)
180{
181 int ret = 0;
182 wp_info->len = bp_data->len;
183 wp_info->addr = bp_data->addr;
184 wp_info->phys_addr = bp_data->phys_addr;
185 wp_info->old_data = NULL;
186
187 if (wp_info->len < 0 || wp_info->len > MAX_WP_SIZE)
188 return -EINVAL;
189
190 wp_info->old_data = kmalloc(bp_data->len, GFP_KERNEL);
191 if (!wp_info->old_data)
192 return -ENOMEM;
193 /* try to backup the original value */
194 ret = read_guest(vcpu, wp_info->phys_addr, wp_info->old_data,
195 wp_info->len);
196 if (ret) {
197 kfree(wp_info->old_data);
198 wp_info->old_data = NULL;
199 }
200
201 return ret;
202}
203
204#define MAX_BP_COUNT 50
205
206int kvm_s390_import_bp_data(struct kvm_vcpu *vcpu,
207 struct kvm_guest_debug *dbg)
208{
209 int ret = 0, nr_wp = 0, nr_bp = 0, i, size;
210 struct kvm_hw_breakpoint *bp_data = NULL;
211 struct kvm_hw_wp_info_arch *wp_info = NULL;
212 struct kvm_hw_bp_info_arch *bp_info = NULL;
213
214 if (dbg->arch.nr_hw_bp <= 0 || !dbg->arch.hw_bp)
215 return 0;
216 else if (dbg->arch.nr_hw_bp > MAX_BP_COUNT)
217 return -EINVAL;
218
219 size = dbg->arch.nr_hw_bp * sizeof(struct kvm_hw_breakpoint);
220 bp_data = kmalloc(size, GFP_KERNEL);
221 if (!bp_data) {
222 ret = -ENOMEM;
223 goto error;
224 }
225
226 ret = copy_from_user(bp_data, dbg->arch.hw_bp, size);
227 if (ret)
228 goto error;
229
230 for (i = 0; i < dbg->arch.nr_hw_bp; i++) {
231 switch (bp_data[i].type) {
232 case KVM_HW_WP_WRITE:
233 nr_wp++;
234 break;
235 case KVM_HW_BP:
236 nr_bp++;
237 break;
238 default:
239 break;
240 }
241 }
242
243 size = nr_wp * sizeof(struct kvm_hw_wp_info_arch);
244 if (size > 0) {
245 wp_info = kmalloc(size, GFP_KERNEL);
246 if (!wp_info) {
247 ret = -ENOMEM;
248 goto error;
249 }
250 }
251 size = nr_bp * sizeof(struct kvm_hw_bp_info_arch);
252 if (size > 0) {
253 bp_info = kmalloc(size, GFP_KERNEL);
254 if (!bp_info) {
255 ret = -ENOMEM;
256 goto error;
257 }
258 }
259
260 for (nr_wp = 0, nr_bp = 0, i = 0; i < dbg->arch.nr_hw_bp; i++) {
261 switch (bp_data[i].type) {
262 case KVM_HW_WP_WRITE:
263 ret = __import_wp_info(vcpu, &bp_data[i],
264 &wp_info[nr_wp]);
265 if (ret)
266 goto error;
267 nr_wp++;
268 break;
269 case KVM_HW_BP:
270 bp_info[nr_bp].len = bp_data[i].len;
271 bp_info[nr_bp].addr = bp_data[i].addr;
272 nr_bp++;
273 break;
274 }
275 }
276
277 vcpu->arch.guestdbg.nr_hw_bp = nr_bp;
278 vcpu->arch.guestdbg.hw_bp_info = bp_info;
279 vcpu->arch.guestdbg.nr_hw_wp = nr_wp;
280 vcpu->arch.guestdbg.hw_wp_info = wp_info;
281 return 0;
282error:
283 kfree(bp_data);
284 kfree(wp_info);
285 kfree(bp_info);
286 return ret;
287}
288
289void kvm_s390_clear_bp_data(struct kvm_vcpu *vcpu)
290{
291 int i;
292 struct kvm_hw_wp_info_arch *hw_wp_info = NULL;
293
294 for (i = 0; i < vcpu->arch.guestdbg.nr_hw_wp; i++) {
295 hw_wp_info = &vcpu->arch.guestdbg.hw_wp_info[i];
296 kfree(hw_wp_info->old_data);
297 hw_wp_info->old_data = NULL;
298 }
299 kfree(vcpu->arch.guestdbg.hw_wp_info);
300 vcpu->arch.guestdbg.hw_wp_info = NULL;
301
302 kfree(vcpu->arch.guestdbg.hw_bp_info);
303 vcpu->arch.guestdbg.hw_bp_info = NULL;
304
305 vcpu->arch.guestdbg.nr_hw_wp = 0;
306 vcpu->arch.guestdbg.nr_hw_bp = 0;
307}
308
309static inline int in_addr_range(u64 addr, u64 a, u64 b)
310{
311 if (a <= b)
312 return (addr >= a) && (addr <= b);
313 else
314 /* "overflowing" interval */
315 return (addr <= a) && (addr >= b);
316}
317
318#define end_of_range(bp_info) (bp_info->addr + bp_info->len - 1)
319
320static struct kvm_hw_bp_info_arch *find_hw_bp(struct kvm_vcpu *vcpu,
321 unsigned long addr)
322{
323 struct kvm_hw_bp_info_arch *bp_info = vcpu->arch.guestdbg.hw_bp_info;
324 int i;
325
326 if (vcpu->arch.guestdbg.nr_hw_bp == 0)
327 return NULL;
328
329 for (i = 0; i < vcpu->arch.guestdbg.nr_hw_bp; i++) {
330 /* addr is directly the start or in the range of a bp */
331 if (addr == bp_info->addr)
332 goto found;
333 if (bp_info->len > 0 &&
334 in_addr_range(addr, bp_info->addr, end_of_range(bp_info)))
335 goto found;
336
337 bp_info++;
338 }
339
340 return NULL;
341found:
342 return bp_info;
343}
344
345static struct kvm_hw_wp_info_arch *any_wp_changed(struct kvm_vcpu *vcpu)
346{
347 int i;
348 struct kvm_hw_wp_info_arch *wp_info = NULL;
349 void *temp = NULL;
350
351 if (vcpu->arch.guestdbg.nr_hw_wp == 0)
352 return NULL;
353
354 for (i = 0; i < vcpu->arch.guestdbg.nr_hw_wp; i++) {
355 wp_info = &vcpu->arch.guestdbg.hw_wp_info[i];
356 if (!wp_info || !wp_info->old_data || wp_info->len <= 0)
357 continue;
358
359 temp = kmalloc(wp_info->len, GFP_KERNEL);
360 if (!temp)
361 continue;
362
363 /* refetch the wp data and compare it to the old value */
364 if (!read_guest(vcpu, wp_info->phys_addr, temp,
365 wp_info->len)) {
366 if (memcmp(temp, wp_info->old_data, wp_info->len)) {
367 kfree(temp);
368 return wp_info;
369 }
370 }
371 kfree(temp);
372 temp = NULL;
373 }
374
375 return NULL;
376}
377
378void kvm_s390_prepare_debug_exit(struct kvm_vcpu *vcpu)
379{
380 vcpu->run->exit_reason = KVM_EXIT_DEBUG;
381 vcpu->guest_debug &= ~KVM_GUESTDBG_EXIT_PENDING;
382}
383
384#define per_bp_event(code) \
385 (code & (PER_EVENT_IFETCH | PER_EVENT_BRANCH))
386#define per_write_wp_event(code) \
387 (code & (PER_EVENT_STORE | PER_EVENT_STORE_REAL))
388
389static int debug_exit_required(struct kvm_vcpu *vcpu)
390{
391 u32 perc = (vcpu->arch.sie_block->perc << 24);
392 struct kvm_debug_exit_arch *debug_exit = &vcpu->run->debug.arch;
393 struct kvm_hw_wp_info_arch *wp_info = NULL;
394 struct kvm_hw_bp_info_arch *bp_info = NULL;
395 unsigned long addr = vcpu->arch.sie_block->gpsw.addr;
396 unsigned long peraddr = vcpu->arch.sie_block->peraddr;
397
398 if (guestdbg_hw_bp_enabled(vcpu)) {
399 if (per_write_wp_event(perc) &&
400 vcpu->arch.guestdbg.nr_hw_wp > 0) {
401 wp_info = any_wp_changed(vcpu);
402 if (wp_info) {
403 debug_exit->addr = wp_info->addr;
404 debug_exit->type = KVM_HW_WP_WRITE;
405 goto exit_required;
406 }
407 }
408 if (per_bp_event(perc) &&
409 vcpu->arch.guestdbg.nr_hw_bp > 0) {
410 bp_info = find_hw_bp(vcpu, addr);
411 /* remove duplicate events if PC==PER address */
412 if (bp_info && (addr != peraddr)) {
413 debug_exit->addr = addr;
414 debug_exit->type = KVM_HW_BP;
415 vcpu->arch.guestdbg.last_bp = addr;
416 goto exit_required;
417 }
418 /* breakpoint missed */
419 bp_info = find_hw_bp(vcpu, peraddr);
420 if (bp_info && vcpu->arch.guestdbg.last_bp != peraddr) {
421 debug_exit->addr = peraddr;
422 debug_exit->type = KVM_HW_BP;
423 goto exit_required;
424 }
425 }
426 }
427 if (guestdbg_sstep_enabled(vcpu) && per_bp_event(perc)) {
428 debug_exit->addr = addr;
429 debug_exit->type = KVM_SINGLESTEP;
430 goto exit_required;
431 }
432
433 return 0;
434exit_required:
435 return 1;
436}
437
438#define guest_per_enabled(vcpu) \
439 (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PER)
440
441static void filter_guest_per_event(struct kvm_vcpu *vcpu)
442{
443 u32 perc = vcpu->arch.sie_block->perc << 24;
444 u64 peraddr = vcpu->arch.sie_block->peraddr;
445 u64 addr = vcpu->arch.sie_block->gpsw.addr;
446 u64 cr9 = vcpu->arch.sie_block->gcr[9];
447 u64 cr10 = vcpu->arch.sie_block->gcr[10];
448 u64 cr11 = vcpu->arch.sie_block->gcr[11];
449 /* filter all events, demanded by the guest */
450 u32 guest_perc = perc & cr9 & PER_EVENT_MASK;
451
452 if (!guest_per_enabled(vcpu))
453 guest_perc = 0;
454
455 /* filter "successful-branching" events */
456 if (guest_perc & PER_EVENT_BRANCH &&
457 cr9 & PER_CONTROL_BRANCH_ADDRESS &&
458 !in_addr_range(addr, cr10, cr11))
459 guest_perc &= ~PER_EVENT_BRANCH;
460
461 /* filter "instruction-fetching" events */
462 if (guest_perc & PER_EVENT_IFETCH &&
463 !in_addr_range(peraddr, cr10, cr11))
464 guest_perc &= ~PER_EVENT_IFETCH;
465
466 /* All other PER events will be given to the guest */
467 /* TODO: Check alterated address/address space */
468
469 vcpu->arch.sie_block->perc = guest_perc >> 24;
470
471 if (!guest_perc)
472 vcpu->arch.sie_block->iprcc &= ~PGM_PER;
473}
474
475void kvm_s390_handle_per_event(struct kvm_vcpu *vcpu)
476{
477 if (debug_exit_required(vcpu))
478 vcpu->guest_debug |= KVM_GUESTDBG_EXIT_PENDING;
479
480 filter_guest_per_event(vcpu);
481}
diff --git a/arch/s390/kvm/intercept.c b/arch/s390/kvm/intercept.c
index eeb1ac7d8fa4..30e1c5eb726a 100644
--- a/arch/s390/kvm/intercept.c
+++ b/arch/s390/kvm/intercept.c
@@ -16,6 +16,7 @@
16#include <linux/pagemap.h> 16#include <linux/pagemap.h>
17 17
18#include <asm/kvm_host.h> 18#include <asm/kvm_host.h>
19#include <asm/asm-offsets.h>
19 20
20#include "kvm-s390.h" 21#include "kvm-s390.h"
21#include "gaccess.h" 22#include "gaccess.h"
@@ -29,6 +30,7 @@ static const intercept_handler_t instruction_handlers[256] = {
29 [0x83] = kvm_s390_handle_diag, 30 [0x83] = kvm_s390_handle_diag,
30 [0xae] = kvm_s390_handle_sigp, 31 [0xae] = kvm_s390_handle_sigp,
31 [0xb2] = kvm_s390_handle_b2, 32 [0xb2] = kvm_s390_handle_b2,
33 [0xb6] = kvm_s390_handle_stctl,
32 [0xb7] = kvm_s390_handle_lctl, 34 [0xb7] = kvm_s390_handle_lctl,
33 [0xb9] = kvm_s390_handle_b9, 35 [0xb9] = kvm_s390_handle_b9,
34 [0xe5] = kvm_s390_handle_e5, 36 [0xe5] = kvm_s390_handle_e5,
@@ -109,22 +111,112 @@ static int handle_instruction(struct kvm_vcpu *vcpu)
109 return -EOPNOTSUPP; 111 return -EOPNOTSUPP;
110} 112}
111 113
114static void __extract_prog_irq(struct kvm_vcpu *vcpu,
115 struct kvm_s390_pgm_info *pgm_info)
116{
117 memset(pgm_info, 0, sizeof(struct kvm_s390_pgm_info));
118 pgm_info->code = vcpu->arch.sie_block->iprcc;
119
120 switch (vcpu->arch.sie_block->iprcc & ~PGM_PER) {
121 case PGM_AFX_TRANSLATION:
122 case PGM_ASX_TRANSLATION:
123 case PGM_EX_TRANSLATION:
124 case PGM_LFX_TRANSLATION:
125 case PGM_LSTE_SEQUENCE:
126 case PGM_LSX_TRANSLATION:
127 case PGM_LX_TRANSLATION:
128 case PGM_PRIMARY_AUTHORITY:
129 case PGM_SECONDARY_AUTHORITY:
130 case PGM_SPACE_SWITCH:
131 pgm_info->trans_exc_code = vcpu->arch.sie_block->tecmc;
132 break;
133 case PGM_ALEN_TRANSLATION:
134 case PGM_ALE_SEQUENCE:
135 case PGM_ASTE_INSTANCE:
136 case PGM_ASTE_SEQUENCE:
137 case PGM_ASTE_VALIDITY:
138 case PGM_EXTENDED_AUTHORITY:
139 pgm_info->exc_access_id = vcpu->arch.sie_block->eai;
140 break;
141 case PGM_ASCE_TYPE:
142 case PGM_PAGE_TRANSLATION:
143 case PGM_REGION_FIRST_TRANS:
144 case PGM_REGION_SECOND_TRANS:
145 case PGM_REGION_THIRD_TRANS:
146 case PGM_SEGMENT_TRANSLATION:
147 pgm_info->trans_exc_code = vcpu->arch.sie_block->tecmc;
148 pgm_info->exc_access_id = vcpu->arch.sie_block->eai;
149 pgm_info->op_access_id = vcpu->arch.sie_block->oai;
150 break;
151 case PGM_MONITOR:
152 pgm_info->mon_class_nr = vcpu->arch.sie_block->mcn;
153 pgm_info->mon_code = vcpu->arch.sie_block->tecmc;
154 break;
155 case PGM_DATA:
156 pgm_info->data_exc_code = vcpu->arch.sie_block->dxc;
157 break;
158 case PGM_PROTECTION:
159 pgm_info->trans_exc_code = vcpu->arch.sie_block->tecmc;
160 pgm_info->exc_access_id = vcpu->arch.sie_block->eai;
161 break;
162 default:
163 break;
164 }
165
166 if (vcpu->arch.sie_block->iprcc & PGM_PER) {
167 pgm_info->per_code = vcpu->arch.sie_block->perc;
168 pgm_info->per_atmid = vcpu->arch.sie_block->peratmid;
169 pgm_info->per_address = vcpu->arch.sie_block->peraddr;
170 pgm_info->per_access_id = vcpu->arch.sie_block->peraid;
171 }
172}
173
174/*
175 * restore ITDB to program-interruption TDB in guest lowcore
176 * and set TX abort indication if required
177*/
178static int handle_itdb(struct kvm_vcpu *vcpu)
179{
180 struct kvm_s390_itdb *itdb;
181 int rc;
182
183 if (!IS_TE_ENABLED(vcpu) || !IS_ITDB_VALID(vcpu))
184 return 0;
185 if (current->thread.per_flags & PER_FLAG_NO_TE)
186 return 0;
187 itdb = (struct kvm_s390_itdb *)vcpu->arch.sie_block->itdba;
188 rc = write_guest_lc(vcpu, __LC_PGM_TDB, itdb, sizeof(*itdb));
189 if (rc)
190 return rc;
191 memset(itdb, 0, sizeof(*itdb));
192
193 return 0;
194}
195
196#define per_event(vcpu) (vcpu->arch.sie_block->iprcc & PGM_PER)
197
112static int handle_prog(struct kvm_vcpu *vcpu) 198static int handle_prog(struct kvm_vcpu *vcpu)
113{ 199{
200 struct kvm_s390_pgm_info pgm_info;
201 int rc;
202
114 vcpu->stat.exit_program_interruption++; 203 vcpu->stat.exit_program_interruption++;
115 204
116 /* Restore ITDB to Program-Interruption TDB in guest memory */ 205 if (guestdbg_enabled(vcpu) && per_event(vcpu)) {
117 if (IS_TE_ENABLED(vcpu) && 206 kvm_s390_handle_per_event(vcpu);
118 !(current->thread.per_flags & PER_FLAG_NO_TE) && 207 /* the interrupt might have been filtered out completely */
119 IS_ITDB_VALID(vcpu)) { 208 if (vcpu->arch.sie_block->iprcc == 0)
120 copy_to_guest(vcpu, TDB_ADDR, vcpu->arch.sie_block->itdba, 209 return 0;
121 sizeof(struct kvm_s390_itdb));
122 memset((void *) vcpu->arch.sie_block->itdba, 0,
123 sizeof(struct kvm_s390_itdb));
124 } 210 }
125 211
126 trace_kvm_s390_intercept_prog(vcpu, vcpu->arch.sie_block->iprcc); 212 trace_kvm_s390_intercept_prog(vcpu, vcpu->arch.sie_block->iprcc);
127 return kvm_s390_inject_program_int(vcpu, vcpu->arch.sie_block->iprcc); 213
214 rc = handle_itdb(vcpu);
215 if (rc)
216 return rc;
217
218 __extract_prog_irq(vcpu, &pgm_info);
219 return kvm_s390_inject_prog_irq(vcpu, &pgm_info);
128} 220}
129 221
130static int handle_instruction_and_prog(struct kvm_vcpu *vcpu) 222static int handle_instruction_and_prog(struct kvm_vcpu *vcpu)
diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c
index 200a8f9390b6..077e4738ebdc 100644
--- a/arch/s390/kvm/interrupt.c
+++ b/arch/s390/kvm/interrupt.c
@@ -56,6 +56,17 @@ static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
56 return 1; 56 return 1;
57} 57}
58 58
59static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
60{
61 if (psw_extint_disabled(vcpu) ||
62 !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
63 return 0;
64 if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
65 /* No timer interrupts when single stepping */
66 return 0;
67 return 1;
68}
69
59static u64 int_word_to_isc_bits(u32 int_word) 70static u64 int_word_to_isc_bits(u32 int_word)
60{ 71{
61 u8 isc = (int_word & 0x38000000) >> 27; 72 u8 isc = (int_word & 0x38000000) >> 27;
@@ -131,7 +142,13 @@ static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
131 CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT, 142 CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
132 &vcpu->arch.sie_block->cpuflags); 143 &vcpu->arch.sie_block->cpuflags);
133 vcpu->arch.sie_block->lctl = 0x0000; 144 vcpu->arch.sie_block->lctl = 0x0000;
134 vcpu->arch.sie_block->ictl &= ~ICTL_LPSW; 145 vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
146
147 if (guestdbg_enabled(vcpu)) {
148 vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
149 LCTL_CR10 | LCTL_CR11);
150 vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
151 }
135} 152}
136 153
137static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag) 154static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
@@ -174,6 +191,106 @@ static void __set_intercept_indicator(struct kvm_vcpu *vcpu,
174 } 191 }
175} 192}
176 193
194static int __deliver_prog_irq(struct kvm_vcpu *vcpu,
195 struct kvm_s390_pgm_info *pgm_info)
196{
197 const unsigned short table[] = { 2, 4, 4, 6 };
198 int rc = 0;
199
200 switch (pgm_info->code & ~PGM_PER) {
201 case PGM_AFX_TRANSLATION:
202 case PGM_ASX_TRANSLATION:
203 case PGM_EX_TRANSLATION:
204 case PGM_LFX_TRANSLATION:
205 case PGM_LSTE_SEQUENCE:
206 case PGM_LSX_TRANSLATION:
207 case PGM_LX_TRANSLATION:
208 case PGM_PRIMARY_AUTHORITY:
209 case PGM_SECONDARY_AUTHORITY:
210 case PGM_SPACE_SWITCH:
211 rc = put_guest_lc(vcpu, pgm_info->trans_exc_code,
212 (u64 *)__LC_TRANS_EXC_CODE);
213 break;
214 case PGM_ALEN_TRANSLATION:
215 case PGM_ALE_SEQUENCE:
216 case PGM_ASTE_INSTANCE:
217 case PGM_ASTE_SEQUENCE:
218 case PGM_ASTE_VALIDITY:
219 case PGM_EXTENDED_AUTHORITY:
220 rc = put_guest_lc(vcpu, pgm_info->exc_access_id,
221 (u8 *)__LC_EXC_ACCESS_ID);
222 break;
223 case PGM_ASCE_TYPE:
224 case PGM_PAGE_TRANSLATION:
225 case PGM_REGION_FIRST_TRANS:
226 case PGM_REGION_SECOND_TRANS:
227 case PGM_REGION_THIRD_TRANS:
228 case PGM_SEGMENT_TRANSLATION:
229 rc = put_guest_lc(vcpu, pgm_info->trans_exc_code,
230 (u64 *)__LC_TRANS_EXC_CODE);
231 rc |= put_guest_lc(vcpu, pgm_info->exc_access_id,
232 (u8 *)__LC_EXC_ACCESS_ID);
233 rc |= put_guest_lc(vcpu, pgm_info->op_access_id,
234 (u8 *)__LC_OP_ACCESS_ID);
235 break;
236 case PGM_MONITOR:
237 rc = put_guest_lc(vcpu, pgm_info->mon_class_nr,
238 (u64 *)__LC_MON_CLASS_NR);
239 rc |= put_guest_lc(vcpu, pgm_info->mon_code,
240 (u64 *)__LC_MON_CODE);
241 break;
242 case PGM_DATA:
243 rc = put_guest_lc(vcpu, pgm_info->data_exc_code,
244 (u32 *)__LC_DATA_EXC_CODE);
245 break;
246 case PGM_PROTECTION:
247 rc = put_guest_lc(vcpu, pgm_info->trans_exc_code,
248 (u64 *)__LC_TRANS_EXC_CODE);
249 rc |= put_guest_lc(vcpu, pgm_info->exc_access_id,
250 (u8 *)__LC_EXC_ACCESS_ID);
251 break;
252 }
253
254 if (pgm_info->code & PGM_PER) {
255 rc |= put_guest_lc(vcpu, pgm_info->per_code,
256 (u8 *) __LC_PER_CODE);
257 rc |= put_guest_lc(vcpu, pgm_info->per_atmid,
258 (u8 *)__LC_PER_ATMID);
259 rc |= put_guest_lc(vcpu, pgm_info->per_address,
260 (u64 *) __LC_PER_ADDRESS);
261 rc |= put_guest_lc(vcpu, pgm_info->per_access_id,
262 (u8 *) __LC_PER_ACCESS_ID);
263 }
264
265 switch (vcpu->arch.sie_block->icptcode) {
266 case ICPT_INST:
267 case ICPT_INSTPROGI:
268 case ICPT_OPEREXC:
269 case ICPT_PARTEXEC:
270 case ICPT_IOINST:
271 /* last instruction only stored for these icptcodes */
272 rc |= put_guest_lc(vcpu, table[vcpu->arch.sie_block->ipa >> 14],
273 (u16 *) __LC_PGM_ILC);
274 break;
275 case ICPT_PROGI:
276 rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->pgmilc,
277 (u16 *) __LC_PGM_ILC);
278 break;
279 default:
280 rc |= put_guest_lc(vcpu, 0,
281 (u16 *) __LC_PGM_ILC);
282 }
283
284 rc |= put_guest_lc(vcpu, pgm_info->code,
285 (u16 *)__LC_PGM_INT_CODE);
286 rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
287 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
288 rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
289 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
290
291 return rc;
292}
293
177static void __do_deliver_interrupt(struct kvm_vcpu *vcpu, 294static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
178 struct kvm_s390_interrupt_info *inti) 295 struct kvm_s390_interrupt_info *inti)
179{ 296{
@@ -186,26 +303,28 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
186 vcpu->stat.deliver_emergency_signal++; 303 vcpu->stat.deliver_emergency_signal++;
187 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 304 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
188 inti->emerg.code, 0); 305 inti->emerg.code, 0);
189 rc = put_guest(vcpu, 0x1201, (u16 __user *)__LC_EXT_INT_CODE); 306 rc = put_guest_lc(vcpu, 0x1201, (u16 *)__LC_EXT_INT_CODE);
190 rc |= put_guest(vcpu, inti->emerg.code, 307 rc |= put_guest_lc(vcpu, inti->emerg.code,
191 (u16 __user *)__LC_EXT_CPU_ADDR); 308 (u16 *)__LC_EXT_CPU_ADDR);
192 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 309 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
310 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
311 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
193 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 312 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
194 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw,
195 __LC_EXT_NEW_PSW, sizeof(psw_t));
196 break; 313 break;
197 case KVM_S390_INT_EXTERNAL_CALL: 314 case KVM_S390_INT_EXTERNAL_CALL:
198 VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp ext call"); 315 VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp ext call");
199 vcpu->stat.deliver_external_call++; 316 vcpu->stat.deliver_external_call++;
200 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 317 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
201 inti->extcall.code, 0); 318 inti->extcall.code, 0);
202 rc = put_guest(vcpu, 0x1202, (u16 __user *)__LC_EXT_INT_CODE); 319 rc = put_guest_lc(vcpu, 0x1202, (u16 *)__LC_EXT_INT_CODE);
203 rc |= put_guest(vcpu, inti->extcall.code, 320 rc |= put_guest_lc(vcpu, inti->extcall.code,
204 (u16 __user *)__LC_EXT_CPU_ADDR); 321 (u16 *)__LC_EXT_CPU_ADDR);
205 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 322 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
206 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 323 &vcpu->arch.sie_block->gpsw,
207 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 324 sizeof(psw_t));
208 __LC_EXT_NEW_PSW, sizeof(psw_t)); 325 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
326 &vcpu->arch.sie_block->gpsw,
327 sizeof(psw_t));
209 break; 328 break;
210 case KVM_S390_INT_SERVICE: 329 case KVM_S390_INT_SERVICE:
211 VCPU_EVENT(vcpu, 4, "interrupt: sclp parm:%x", 330 VCPU_EVENT(vcpu, 4, "interrupt: sclp parm:%x",
@@ -213,37 +332,39 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
213 vcpu->stat.deliver_service_signal++; 332 vcpu->stat.deliver_service_signal++;
214 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 333 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
215 inti->ext.ext_params, 0); 334 inti->ext.ext_params, 0);
216 rc = put_guest(vcpu, 0x2401, (u16 __user *)__LC_EXT_INT_CODE); 335 rc = put_guest_lc(vcpu, 0x2401, (u16 *)__LC_EXT_INT_CODE);
217 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 336 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
337 &vcpu->arch.sie_block->gpsw,
338 sizeof(psw_t));
339 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
218 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 340 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
219 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 341 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
220 __LC_EXT_NEW_PSW, sizeof(psw_t)); 342 (u32 *)__LC_EXT_PARAMS);
221 rc |= put_guest(vcpu, inti->ext.ext_params,
222 (u32 __user *)__LC_EXT_PARAMS);
223 break; 343 break;
224 case KVM_S390_INT_PFAULT_INIT: 344 case KVM_S390_INT_PFAULT_INIT:
225 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 0, 345 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 0,
226 inti->ext.ext_params2); 346 inti->ext.ext_params2);
227 rc = put_guest(vcpu, 0x2603, (u16 __user *) __LC_EXT_INT_CODE); 347 rc = put_guest_lc(vcpu, 0x2603, (u16 *) __LC_EXT_INT_CODE);
228 rc |= put_guest(vcpu, 0x0600, (u16 __user *) __LC_EXT_CPU_ADDR); 348 rc |= put_guest_lc(vcpu, 0x0600, (u16 *) __LC_EXT_CPU_ADDR);
229 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 349 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
350 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
351 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
230 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 352 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
231 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 353 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
232 __LC_EXT_NEW_PSW, sizeof(psw_t)); 354 (u64 *) __LC_EXT_PARAMS2);
233 rc |= put_guest(vcpu, inti->ext.ext_params2,
234 (u64 __user *) __LC_EXT_PARAMS2);
235 break; 355 break;
236 case KVM_S390_INT_PFAULT_DONE: 356 case KVM_S390_INT_PFAULT_DONE:
237 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 0, 357 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 0,
238 inti->ext.ext_params2); 358 inti->ext.ext_params2);
239 rc = put_guest(vcpu, 0x2603, (u16 __user *) __LC_EXT_INT_CODE); 359 rc = put_guest_lc(vcpu, 0x2603, (u16 *)__LC_EXT_INT_CODE);
240 rc |= put_guest(vcpu, 0x0680, (u16 __user *) __LC_EXT_CPU_ADDR); 360 rc |= put_guest_lc(vcpu, 0x0680, (u16 *)__LC_EXT_CPU_ADDR);
241 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 361 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
362 &vcpu->arch.sie_block->gpsw,
363 sizeof(psw_t));
364 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
242 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 365 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
243 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 366 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
244 __LC_EXT_NEW_PSW, sizeof(psw_t)); 367 (u64 *)__LC_EXT_PARAMS2);
245 rc |= put_guest(vcpu, inti->ext.ext_params2,
246 (u64 __user *) __LC_EXT_PARAMS2);
247 break; 368 break;
248 case KVM_S390_INT_VIRTIO: 369 case KVM_S390_INT_VIRTIO:
249 VCPU_EVENT(vcpu, 4, "interrupt: virtio parm:%x,parm64:%llx", 370 VCPU_EVENT(vcpu, 4, "interrupt: virtio parm:%x,parm64:%llx",
@@ -252,16 +373,17 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
252 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 373 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
253 inti->ext.ext_params, 374 inti->ext.ext_params,
254 inti->ext.ext_params2); 375 inti->ext.ext_params2);
255 rc = put_guest(vcpu, 0x2603, (u16 __user *)__LC_EXT_INT_CODE); 376 rc = put_guest_lc(vcpu, 0x2603, (u16 *)__LC_EXT_INT_CODE);
256 rc |= put_guest(vcpu, 0x0d00, (u16 __user *)__LC_EXT_CPU_ADDR); 377 rc |= put_guest_lc(vcpu, 0x0d00, (u16 *)__LC_EXT_CPU_ADDR);
257 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 378 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
379 &vcpu->arch.sie_block->gpsw,
380 sizeof(psw_t));
381 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
258 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 382 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
259 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 383 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
260 __LC_EXT_NEW_PSW, sizeof(psw_t)); 384 (u32 *)__LC_EXT_PARAMS);
261 rc |= put_guest(vcpu, inti->ext.ext_params, 385 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
262 (u32 __user *)__LC_EXT_PARAMS); 386 (u64 *)__LC_EXT_PARAMS2);
263 rc |= put_guest(vcpu, inti->ext.ext_params2,
264 (u64 __user *)__LC_EXT_PARAMS2);
265 break; 387 break;
266 case KVM_S390_SIGP_STOP: 388 case KVM_S390_SIGP_STOP:
267 VCPU_EVENT(vcpu, 4, "%s", "interrupt: cpu stop"); 389 VCPU_EVENT(vcpu, 4, "%s", "interrupt: cpu stop");
@@ -285,12 +407,12 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
285 vcpu->stat.deliver_restart_signal++; 407 vcpu->stat.deliver_restart_signal++;
286 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 408 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
287 0, 0); 409 0, 0);
288 rc = copy_to_guest(vcpu, 410 rc = write_guest_lc(vcpu,
289 offsetof(struct _lowcore, restart_old_psw), 411 offsetof(struct _lowcore, restart_old_psw),
290 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 412 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
291 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 413 rc |= read_guest_lc(vcpu, offsetof(struct _lowcore, restart_psw),
292 offsetof(struct _lowcore, restart_psw), 414 &vcpu->arch.sie_block->gpsw,
293 sizeof(psw_t)); 415 sizeof(psw_t));
294 atomic_clear_mask(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags); 416 atomic_clear_mask(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
295 break; 417 break;
296 case KVM_S390_PROGRAM_INT: 418 case KVM_S390_PROGRAM_INT:
@@ -300,13 +422,7 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
300 vcpu->stat.deliver_program_int++; 422 vcpu->stat.deliver_program_int++;
301 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 423 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
302 inti->pgm.code, 0); 424 inti->pgm.code, 0);
303 rc = put_guest(vcpu, inti->pgm.code, (u16 __user *)__LC_PGM_INT_CODE); 425 rc = __deliver_prog_irq(vcpu, &inti->pgm);
304 rc |= put_guest(vcpu, table[vcpu->arch.sie_block->ipa >> 14],
305 (u16 __user *)__LC_PGM_ILC);
306 rc |= copy_to_guest(vcpu, __LC_PGM_OLD_PSW,
307 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
308 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw,
309 __LC_PGM_NEW_PSW, sizeof(psw_t));
310 break; 426 break;
311 427
312 case KVM_S390_MCHK: 428 case KVM_S390_MCHK:
@@ -317,11 +433,12 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
317 inti->mchk.mcic); 433 inti->mchk.mcic);
318 rc = kvm_s390_vcpu_store_status(vcpu, 434 rc = kvm_s390_vcpu_store_status(vcpu,
319 KVM_S390_STORE_STATUS_PREFIXED); 435 KVM_S390_STORE_STATUS_PREFIXED);
320 rc |= put_guest(vcpu, inti->mchk.mcic, (u64 __user *) __LC_MCCK_CODE); 436 rc |= put_guest_lc(vcpu, inti->mchk.mcic, (u64 *)__LC_MCCK_CODE);
321 rc |= copy_to_guest(vcpu, __LC_MCK_OLD_PSW, 437 rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
438 &vcpu->arch.sie_block->gpsw,
439 sizeof(psw_t));
440 rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
322 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 441 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
323 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw,
324 __LC_MCK_NEW_PSW, sizeof(psw_t));
325 break; 442 break;
326 443
327 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX: 444 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
@@ -334,18 +451,20 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
334 vcpu->stat.deliver_io_int++; 451 vcpu->stat.deliver_io_int++;
335 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type, 452 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
336 param0, param1); 453 param0, param1);
337 rc = put_guest(vcpu, inti->io.subchannel_id, 454 rc = put_guest_lc(vcpu, inti->io.subchannel_id,
338 (u16 __user *) __LC_SUBCHANNEL_ID); 455 (u16 *)__LC_SUBCHANNEL_ID);
339 rc |= put_guest(vcpu, inti->io.subchannel_nr, 456 rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
340 (u16 __user *) __LC_SUBCHANNEL_NR); 457 (u16 *)__LC_SUBCHANNEL_NR);
341 rc |= put_guest(vcpu, inti->io.io_int_parm, 458 rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
342 (u32 __user *) __LC_IO_INT_PARM); 459 (u32 *)__LC_IO_INT_PARM);
343 rc |= put_guest(vcpu, inti->io.io_int_word, 460 rc |= put_guest_lc(vcpu, inti->io.io_int_word,
344 (u32 __user *) __LC_IO_INT_WORD); 461 (u32 *)__LC_IO_INT_WORD);
345 rc |= copy_to_guest(vcpu, __LC_IO_OLD_PSW, 462 rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
346 &vcpu->arch.sie_block->gpsw, sizeof(psw_t)); 463 &vcpu->arch.sie_block->gpsw,
347 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw, 464 sizeof(psw_t));
348 __LC_IO_NEW_PSW, sizeof(psw_t)); 465 rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
466 &vcpu->arch.sie_block->gpsw,
467 sizeof(psw_t));
349 break; 468 break;
350 } 469 }
351 default: 470 default:
@@ -358,25 +477,21 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
358 } 477 }
359} 478}
360 479
361static int __try_deliver_ckc_interrupt(struct kvm_vcpu *vcpu) 480static void deliver_ckc_interrupt(struct kvm_vcpu *vcpu)
362{ 481{
363 int rc; 482 int rc;
364 483
365 if (psw_extint_disabled(vcpu)) 484 rc = put_guest_lc(vcpu, 0x1004, (u16 __user *)__LC_EXT_INT_CODE);
366 return 0; 485 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
367 if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul)) 486 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
368 return 0; 487 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
369 rc = put_guest(vcpu, 0x1004, (u16 __user *)__LC_EXT_INT_CODE); 488 &vcpu->arch.sie_block->gpsw,
370 rc |= copy_to_guest(vcpu, __LC_EXT_OLD_PSW, 489 sizeof(psw_t));
371 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
372 rc |= copy_from_guest(vcpu, &vcpu->arch.sie_block->gpsw,
373 __LC_EXT_NEW_PSW, sizeof(psw_t));
374 if (rc) { 490 if (rc) {
375 printk("kvm: The guest lowcore is not mapped during interrupt " 491 printk("kvm: The guest lowcore is not mapped during interrupt "
376 "delivery, killing userspace\n"); 492 "delivery, killing userspace\n");
377 do_exit(SIGKILL); 493 do_exit(SIGKILL);
378 } 494 }
379 return 1;
380} 495}
381 496
382int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu) 497int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu)
@@ -406,19 +521,20 @@ int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu)
406 spin_unlock(&fi->lock); 521 spin_unlock(&fi->lock);
407 } 522 }
408 523
409 if ((!rc) && (vcpu->arch.sie_block->ckc < 524 if (!rc && kvm_cpu_has_pending_timer(vcpu))
410 get_tod_clock_fast() + vcpu->arch.sie_block->epoch)) { 525 rc = 1;
411 if ((!psw_extint_disabled(vcpu)) &&
412 (vcpu->arch.sie_block->gcr[0] & 0x800ul))
413 rc = 1;
414 }
415 526
416 return rc; 527 return rc;
417} 528}
418 529
419int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu) 530int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
420{ 531{
421 return 0; 532 if (!(vcpu->arch.sie_block->ckc <
533 get_tod_clock_fast() + vcpu->arch.sie_block->epoch))
534 return 0;
535 if (!ckc_interrupts_enabled(vcpu))
536 return 0;
537 return 1;
422} 538}
423 539
424int kvm_s390_handle_wait(struct kvm_vcpu *vcpu) 540int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
@@ -441,8 +557,7 @@ int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
441 return -EOPNOTSUPP; /* disabled wait */ 557 return -EOPNOTSUPP; /* disabled wait */
442 } 558 }
443 559
444 if (psw_extint_disabled(vcpu) || 560 if (!ckc_interrupts_enabled(vcpu)) {
445 (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))) {
446 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer"); 561 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
447 goto no_timer; 562 goto no_timer;
448 } 563 }
@@ -554,9 +669,8 @@ void kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
554 } while (deliver); 669 } while (deliver);
555 } 670 }
556 671
557 if ((vcpu->arch.sie_block->ckc < 672 if (kvm_cpu_has_pending_timer(vcpu))
558 get_tod_clock_fast() + vcpu->arch.sie_block->epoch)) 673 deliver_ckc_interrupt(vcpu);
559 __try_deliver_ckc_interrupt(vcpu);
560 674
561 if (atomic_read(&fi->active)) { 675 if (atomic_read(&fi->active)) {
562 do { 676 do {
@@ -660,6 +774,31 @@ int kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code)
660 return 0; 774 return 0;
661} 775}
662 776
777int kvm_s390_inject_prog_irq(struct kvm_vcpu *vcpu,
778 struct kvm_s390_pgm_info *pgm_info)
779{
780 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
781 struct kvm_s390_interrupt_info *inti;
782
783 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
784 if (!inti)
785 return -ENOMEM;
786
787 VCPU_EVENT(vcpu, 3, "inject: prog irq %d (from kernel)",
788 pgm_info->code);
789 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
790 pgm_info->code, 0, 1);
791
792 inti->type = KVM_S390_PROGRAM_INT;
793 memcpy(&inti->pgm, pgm_info, sizeof(inti->pgm));
794 spin_lock_bh(&li->lock);
795 list_add(&inti->list, &li->list);
796 atomic_set(&li->active, 1);
797 BUG_ON(waitqueue_active(li->wq));
798 spin_unlock_bh(&li->lock);
799 return 0;
800}
801
663struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm, 802struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
664 u64 cr6, u64 schid) 803 u64 cr6, u64 schid)
665{ 804{
@@ -810,6 +949,12 @@ int kvm_s390_inject_vm(struct kvm *kvm,
810 return __inject_vm(kvm, inti); 949 return __inject_vm(kvm, inti);
811} 950}
812 951
952void kvm_s390_reinject_io_int(struct kvm *kvm,
953 struct kvm_s390_interrupt_info *inti)
954{
955 __inject_vm(kvm, inti);
956}
957
813int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, 958int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
814 struct kvm_s390_interrupt *s390int) 959 struct kvm_s390_interrupt *s390int)
815{ 960{
@@ -900,7 +1045,7 @@ int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
900 return 0; 1045 return 0;
901} 1046}
902 1047
903static void clear_floating_interrupts(struct kvm *kvm) 1048void kvm_s390_clear_float_irqs(struct kvm *kvm)
904{ 1049{
905 struct kvm_s390_float_interrupt *fi; 1050 struct kvm_s390_float_interrupt *fi;
906 struct kvm_s390_interrupt_info *n, *inti = NULL; 1051 struct kvm_s390_interrupt_info *n, *inti = NULL;
@@ -1246,7 +1391,7 @@ static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1246 break; 1391 break;
1247 case KVM_DEV_FLIC_CLEAR_IRQS: 1392 case KVM_DEV_FLIC_CLEAR_IRQS:
1248 r = 0; 1393 r = 0;
1249 clear_floating_interrupts(dev->kvm); 1394 kvm_s390_clear_float_irqs(dev->kvm);
1250 break; 1395 break;
1251 case KVM_DEV_FLIC_APF_ENABLE: 1396 case KVM_DEV_FLIC_APF_ENABLE:
1252 dev->kvm->arch.gmap->pfault_enabled = 1; 1397 dev->kvm->arch.gmap->pfault_enabled = 1;
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index b3ecb8f5b6ce..b32c42cbc706 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -11,6 +11,7 @@
11 * Christian Borntraeger <borntraeger@de.ibm.com> 11 * Christian Borntraeger <borntraeger@de.ibm.com>
12 * Heiko Carstens <heiko.carstens@de.ibm.com> 12 * Heiko Carstens <heiko.carstens@de.ibm.com>
13 * Christian Ehrhardt <ehrhardt@de.ibm.com> 13 * Christian Ehrhardt <ehrhardt@de.ibm.com>
14 * Jason J. Herne <jjherne@us.ibm.com>
14 */ 15 */
15 16
16#include <linux/compiler.h> 17#include <linux/compiler.h>
@@ -51,6 +52,8 @@ struct kvm_stats_debugfs_item debugfs_entries[] = {
51 { "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program) }, 52 { "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program) },
52 { "instruction_lctlg", VCPU_STAT(instruction_lctlg) }, 53 { "instruction_lctlg", VCPU_STAT(instruction_lctlg) },
53 { "instruction_lctl", VCPU_STAT(instruction_lctl) }, 54 { "instruction_lctl", VCPU_STAT(instruction_lctl) },
55 { "instruction_stctl", VCPU_STAT(instruction_stctl) },
56 { "instruction_stctg", VCPU_STAT(instruction_stctg) },
54 { "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal) }, 57 { "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal) },
55 { "deliver_external_call", VCPU_STAT(deliver_external_call) }, 58 { "deliver_external_call", VCPU_STAT(deliver_external_call) },
56 { "deliver_service_signal", VCPU_STAT(deliver_service_signal) }, 59 { "deliver_service_signal", VCPU_STAT(deliver_service_signal) },
@@ -66,6 +69,7 @@ struct kvm_stats_debugfs_item debugfs_entries[] = {
66 { "instruction_stpx", VCPU_STAT(instruction_stpx) }, 69 { "instruction_stpx", VCPU_STAT(instruction_stpx) },
67 { "instruction_stap", VCPU_STAT(instruction_stap) }, 70 { "instruction_stap", VCPU_STAT(instruction_stap) },
68 { "instruction_storage_key", VCPU_STAT(instruction_storage_key) }, 71 { "instruction_storage_key", VCPU_STAT(instruction_storage_key) },
72 { "instruction_ipte_interlock", VCPU_STAT(instruction_ipte_interlock) },
69 { "instruction_stsch", VCPU_STAT(instruction_stsch) }, 73 { "instruction_stsch", VCPU_STAT(instruction_stsch) },
70 { "instruction_chsc", VCPU_STAT(instruction_chsc) }, 74 { "instruction_chsc", VCPU_STAT(instruction_chsc) },
71 { "instruction_essa", VCPU_STAT(instruction_essa) }, 75 { "instruction_essa", VCPU_STAT(instruction_essa) },
@@ -90,7 +94,7 @@ unsigned long *vfacilities;
90static struct gmap_notifier gmap_notifier; 94static struct gmap_notifier gmap_notifier;
91 95
92/* test availability of vfacility */ 96/* test availability of vfacility */
93static inline int test_vfacility(unsigned long nr) 97int test_vfacility(unsigned long nr)
94{ 98{
95 return __test_facility(nr, (void *) vfacilities); 99 return __test_facility(nr, (void *) vfacilities);
96} 100}
@@ -161,6 +165,7 @@ int kvm_dev_ioctl_check_extension(long ext)
161 case KVM_CAP_IOEVENTFD: 165 case KVM_CAP_IOEVENTFD:
162 case KVM_CAP_DEVICE_CTRL: 166 case KVM_CAP_DEVICE_CTRL:
163 case KVM_CAP_ENABLE_CAP_VM: 167 case KVM_CAP_ENABLE_CAP_VM:
168 case KVM_CAP_VM_ATTRIBUTES:
164 r = 1; 169 r = 1;
165 break; 170 break;
166 case KVM_CAP_NR_VCPUS: 171 case KVM_CAP_NR_VCPUS:
@@ -179,6 +184,25 @@ int kvm_dev_ioctl_check_extension(long ext)
179 return r; 184 return r;
180} 185}
181 186
187static void kvm_s390_sync_dirty_log(struct kvm *kvm,
188 struct kvm_memory_slot *memslot)
189{
190 gfn_t cur_gfn, last_gfn;
191 unsigned long address;
192 struct gmap *gmap = kvm->arch.gmap;
193
194 down_read(&gmap->mm->mmap_sem);
195 /* Loop over all guest pages */
196 last_gfn = memslot->base_gfn + memslot->npages;
197 for (cur_gfn = memslot->base_gfn; cur_gfn <= last_gfn; cur_gfn++) {
198 address = gfn_to_hva_memslot(memslot, cur_gfn);
199
200 if (gmap_test_and_clear_dirty(address, gmap))
201 mark_page_dirty(kvm, cur_gfn);
202 }
203 up_read(&gmap->mm->mmap_sem);
204}
205
182/* Section: vm related */ 206/* Section: vm related */
183/* 207/*
184 * Get (and clear) the dirty memory log for a memory slot. 208 * Get (and clear) the dirty memory log for a memory slot.
@@ -186,7 +210,36 @@ int kvm_dev_ioctl_check_extension(long ext)
186int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, 210int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
187 struct kvm_dirty_log *log) 211 struct kvm_dirty_log *log)
188{ 212{
189 return 0; 213 int r;
214 unsigned long n;
215 struct kvm_memory_slot *memslot;
216 int is_dirty = 0;
217
218 mutex_lock(&kvm->slots_lock);
219
220 r = -EINVAL;
221 if (log->slot >= KVM_USER_MEM_SLOTS)
222 goto out;
223
224 memslot = id_to_memslot(kvm->memslots, log->slot);
225 r = -ENOENT;
226 if (!memslot->dirty_bitmap)
227 goto out;
228
229 kvm_s390_sync_dirty_log(kvm, memslot);
230 r = kvm_get_dirty_log(kvm, log, &is_dirty);
231 if (r)
232 goto out;
233
234 /* Clear the dirty log */
235 if (is_dirty) {
236 n = kvm_dirty_bitmap_bytes(memslot);
237 memset(memslot->dirty_bitmap, 0, n);
238 }
239 r = 0;
240out:
241 mutex_unlock(&kvm->slots_lock);
242 return r;
190} 243}
191 244
192static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) 245static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
@@ -208,11 +261,86 @@ static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
208 return r; 261 return r;
209} 262}
210 263
264static int kvm_s390_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
265{
266 int ret;
267 unsigned int idx;
268 switch (attr->attr) {
269 case KVM_S390_VM_MEM_ENABLE_CMMA:
270 ret = -EBUSY;
271 mutex_lock(&kvm->lock);
272 if (atomic_read(&kvm->online_vcpus) == 0) {
273 kvm->arch.use_cmma = 1;
274 ret = 0;
275 }
276 mutex_unlock(&kvm->lock);
277 break;
278 case KVM_S390_VM_MEM_CLR_CMMA:
279 mutex_lock(&kvm->lock);
280 idx = srcu_read_lock(&kvm->srcu);
281 page_table_reset_pgste(kvm->arch.gmap->mm, 0, TASK_SIZE, false);
282 srcu_read_unlock(&kvm->srcu, idx);
283 mutex_unlock(&kvm->lock);
284 ret = 0;
285 break;
286 default:
287 ret = -ENXIO;
288 break;
289 }
290 return ret;
291}
292
293static int kvm_s390_vm_set_attr(struct kvm *kvm, struct kvm_device_attr *attr)
294{
295 int ret;
296
297 switch (attr->group) {
298 case KVM_S390_VM_MEM_CTRL:
299 ret = kvm_s390_mem_control(kvm, attr);
300 break;
301 default:
302 ret = -ENXIO;
303 break;
304 }
305
306 return ret;
307}
308
309static int kvm_s390_vm_get_attr(struct kvm *kvm, struct kvm_device_attr *attr)
310{
311 return -ENXIO;
312}
313
314static int kvm_s390_vm_has_attr(struct kvm *kvm, struct kvm_device_attr *attr)
315{
316 int ret;
317
318 switch (attr->group) {
319 case KVM_S390_VM_MEM_CTRL:
320 switch (attr->attr) {
321 case KVM_S390_VM_MEM_ENABLE_CMMA:
322 case KVM_S390_VM_MEM_CLR_CMMA:
323 ret = 0;
324 break;
325 default:
326 ret = -ENXIO;
327 break;
328 }
329 break;
330 default:
331 ret = -ENXIO;
332 break;
333 }
334
335 return ret;
336}
337
211long kvm_arch_vm_ioctl(struct file *filp, 338long kvm_arch_vm_ioctl(struct file *filp,
212 unsigned int ioctl, unsigned long arg) 339 unsigned int ioctl, unsigned long arg)
213{ 340{
214 struct kvm *kvm = filp->private_data; 341 struct kvm *kvm = filp->private_data;
215 void __user *argp = (void __user *)arg; 342 void __user *argp = (void __user *)arg;
343 struct kvm_device_attr attr;
216 int r; 344 int r;
217 345
218 switch (ioctl) { 346 switch (ioctl) {
@@ -245,6 +373,27 @@ long kvm_arch_vm_ioctl(struct file *filp,
245 } 373 }
246 break; 374 break;
247 } 375 }
376 case KVM_SET_DEVICE_ATTR: {
377 r = -EFAULT;
378 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
379 break;
380 r = kvm_s390_vm_set_attr(kvm, &attr);
381 break;
382 }
383 case KVM_GET_DEVICE_ATTR: {
384 r = -EFAULT;
385 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
386 break;
387 r = kvm_s390_vm_get_attr(kvm, &attr);
388 break;
389 }
390 case KVM_HAS_DEVICE_ATTR: {
391 r = -EFAULT;
392 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
393 break;
394 r = kvm_s390_vm_has_attr(kvm, &attr);
395 break;
396 }
248 default: 397 default:
249 r = -ENOTTY; 398 r = -ENOTTY;
250 } 399 }
@@ -291,6 +440,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
291 440
292 spin_lock_init(&kvm->arch.float_int.lock); 441 spin_lock_init(&kvm->arch.float_int.lock);
293 INIT_LIST_HEAD(&kvm->arch.float_int.list); 442 INIT_LIST_HEAD(&kvm->arch.float_int.list);
443 init_waitqueue_head(&kvm->arch.ipte_wq);
294 444
295 debug_register_view(kvm->arch.dbf, &debug_sprintf_view); 445 debug_register_view(kvm->arch.dbf, &debug_sprintf_view);
296 VM_EVENT(kvm, 3, "%s", "vm created"); 446 VM_EVENT(kvm, 3, "%s", "vm created");
@@ -321,6 +471,7 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
321{ 471{
322 VCPU_EVENT(vcpu, 3, "%s", "free cpu"); 472 VCPU_EVENT(vcpu, 3, "%s", "free cpu");
323 trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id); 473 trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id);
474 kvm_s390_clear_local_irqs(vcpu);
324 kvm_clear_async_pf_completion_queue(vcpu); 475 kvm_clear_async_pf_completion_queue(vcpu);
325 if (!kvm_is_ucontrol(vcpu->kvm)) { 476 if (!kvm_is_ucontrol(vcpu->kvm)) {
326 clear_bit(63 - vcpu->vcpu_id, 477 clear_bit(63 - vcpu->vcpu_id,
@@ -334,9 +485,8 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
334 if (kvm_is_ucontrol(vcpu->kvm)) 485 if (kvm_is_ucontrol(vcpu->kvm))
335 gmap_free(vcpu->arch.gmap); 486 gmap_free(vcpu->arch.gmap);
336 487
337 if (vcpu->arch.sie_block->cbrlo) 488 if (kvm_s390_cmma_enabled(vcpu->kvm))
338 __free_page(__pfn_to_page( 489 kvm_s390_vcpu_unsetup_cmma(vcpu);
339 vcpu->arch.sie_block->cbrlo >> PAGE_SHIFT));
340 free_page((unsigned long)(vcpu->arch.sie_block)); 490 free_page((unsigned long)(vcpu->arch.sie_block));
341 491
342 kvm_vcpu_uninit(vcpu); 492 kvm_vcpu_uninit(vcpu);
@@ -371,6 +521,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
371 if (!kvm_is_ucontrol(kvm)) 521 if (!kvm_is_ucontrol(kvm))
372 gmap_free(kvm->arch.gmap); 522 gmap_free(kvm->arch.gmap);
373 kvm_s390_destroy_adapters(kvm); 523 kvm_s390_destroy_adapters(kvm);
524 kvm_s390_clear_float_irqs(kvm);
374} 525}
375 526
376/* Section: vcpu related */ 527/* Section: vcpu related */
@@ -450,9 +601,26 @@ int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
450 return 0; 601 return 0;
451} 602}
452 603
604void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu *vcpu)
605{
606 free_page(vcpu->arch.sie_block->cbrlo);
607 vcpu->arch.sie_block->cbrlo = 0;
608}
609
610int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu *vcpu)
611{
612 vcpu->arch.sie_block->cbrlo = get_zeroed_page(GFP_KERNEL);
613 if (!vcpu->arch.sie_block->cbrlo)
614 return -ENOMEM;
615
616 vcpu->arch.sie_block->ecb2 |= 0x80;
617 vcpu->arch.sie_block->ecb2 &= ~0x08;
618 return 0;
619}
620
453int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu) 621int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
454{ 622{
455 struct page *cbrl; 623 int rc = 0;
456 624
457 atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH | 625 atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH |
458 CPUSTAT_SM | 626 CPUSTAT_SM |
@@ -463,15 +631,15 @@ int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
463 vcpu->arch.sie_block->ecb |= 0x10; 631 vcpu->arch.sie_block->ecb |= 0x10;
464 632
465 vcpu->arch.sie_block->ecb2 = 8; 633 vcpu->arch.sie_block->ecb2 = 8;
466 vcpu->arch.sie_block->eca = 0xC1002001U; 634 vcpu->arch.sie_block->eca = 0xC1002000U;
635 if (sclp_has_siif())
636 vcpu->arch.sie_block->eca |= 1;
467 vcpu->arch.sie_block->fac = (int) (long) vfacilities; 637 vcpu->arch.sie_block->fac = (int) (long) vfacilities;
468 if (kvm_enabled_cmma()) { 638 vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
469 cbrl = alloc_page(GFP_KERNEL | __GFP_ZERO); 639 if (kvm_s390_cmma_enabled(vcpu->kvm)) {
470 if (cbrl) { 640 rc = kvm_s390_vcpu_setup_cmma(vcpu);
471 vcpu->arch.sie_block->ecb2 |= 0x80; 641 if (rc)
472 vcpu->arch.sie_block->ecb2 &= ~0x08; 642 return rc;
473 vcpu->arch.sie_block->cbrlo = page_to_phys(cbrl);
474 }
475 } 643 }
476 hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS); 644 hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
477 tasklet_init(&vcpu->arch.tasklet, kvm_s390_tasklet, 645 tasklet_init(&vcpu->arch.tasklet, kvm_s390_tasklet,
@@ -479,7 +647,7 @@ int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
479 vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup; 647 vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup;
480 get_cpu_id(&vcpu->arch.cpu_id); 648 get_cpu_id(&vcpu->arch.cpu_id);
481 vcpu->arch.cpu_id.version = 0xff; 649 vcpu->arch.cpu_id.version = 0xff;
482 return 0; 650 return rc;
483} 651}
484 652
485struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, 653struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm,
@@ -768,10 +936,40 @@ int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
768 return -EINVAL; /* not implemented yet */ 936 return -EINVAL; /* not implemented yet */
769} 937}
770 938
939#define VALID_GUESTDBG_FLAGS (KVM_GUESTDBG_SINGLESTEP | \
940 KVM_GUESTDBG_USE_HW_BP | \
941 KVM_GUESTDBG_ENABLE)
942
771int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, 943int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
772 struct kvm_guest_debug *dbg) 944 struct kvm_guest_debug *dbg)
773{ 945{
774 return -EINVAL; /* not implemented yet */ 946 int rc = 0;
947
948 vcpu->guest_debug = 0;
949 kvm_s390_clear_bp_data(vcpu);
950
951 if (vcpu->guest_debug & ~VALID_GUESTDBG_FLAGS)
952 return -EINVAL;
953
954 if (dbg->control & KVM_GUESTDBG_ENABLE) {
955 vcpu->guest_debug = dbg->control;
956 /* enforce guest PER */
957 atomic_set_mask(CPUSTAT_P, &vcpu->arch.sie_block->cpuflags);
958
959 if (dbg->control & KVM_GUESTDBG_USE_HW_BP)
960 rc = kvm_s390_import_bp_data(vcpu, dbg);
961 } else {
962 atomic_clear_mask(CPUSTAT_P, &vcpu->arch.sie_block->cpuflags);
963 vcpu->arch.guestdbg.last_bp = 0;
964 }
965
966 if (rc) {
967 vcpu->guest_debug = 0;
968 kvm_s390_clear_bp_data(vcpu);
969 atomic_clear_mask(CPUSTAT_P, &vcpu->arch.sie_block->cpuflags);
970 }
971
972 return rc;
775} 973}
776 974
777int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, 975int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
@@ -786,6 +984,18 @@ int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
786 return -EINVAL; /* not implemented yet */ 984 return -EINVAL; /* not implemented yet */
787} 985}
788 986
987bool kvm_s390_cmma_enabled(struct kvm *kvm)
988{
989 if (!MACHINE_IS_LPAR)
990 return false;
991 /* only enable for z10 and later */
992 if (!MACHINE_HAS_EDAT1)
993 return false;
994 if (!kvm->arch.use_cmma)
995 return false;
996 return true;
997}
998
789static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu) 999static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
790{ 1000{
791 /* 1001 /*
@@ -882,8 +1092,9 @@ static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu)
882 if (!vcpu->arch.gmap->pfault_enabled) 1092 if (!vcpu->arch.gmap->pfault_enabled)
883 return 0; 1093 return 0;
884 1094
885 hva = gmap_fault(current->thread.gmap_addr, vcpu->arch.gmap); 1095 hva = gfn_to_hva(vcpu->kvm, gpa_to_gfn(current->thread.gmap_addr));
886 if (copy_from_guest(vcpu, &arch.pfault_token, vcpu->arch.pfault_token, 8)) 1096 hva += current->thread.gmap_addr & ~PAGE_MASK;
1097 if (read_guest_real(vcpu, vcpu->arch.pfault_token, &arch.pfault_token, 8))
887 return 0; 1098 return 0;
888 1099
889 rc = kvm_setup_async_pf(vcpu, current->thread.gmap_addr, hva, &arch); 1100 rc = kvm_setup_async_pf(vcpu, current->thread.gmap_addr, hva, &arch);
@@ -916,6 +1127,11 @@ static int vcpu_pre_run(struct kvm_vcpu *vcpu)
916 if (rc) 1127 if (rc)
917 return rc; 1128 return rc;
918 1129
1130 if (guestdbg_enabled(vcpu)) {
1131 kvm_s390_backup_guest_per_regs(vcpu);
1132 kvm_s390_patch_guest_per_regs(vcpu);
1133 }
1134
919 vcpu->arch.sie_block->icptcode = 0; 1135 vcpu->arch.sie_block->icptcode = 0;
920 cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags); 1136 cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags);
921 VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags); 1137 VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags);
@@ -932,6 +1148,9 @@ static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason)
932 vcpu->arch.sie_block->icptcode); 1148 vcpu->arch.sie_block->icptcode);
933 trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode); 1149 trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode);
934 1150
1151 if (guestdbg_enabled(vcpu))
1152 kvm_s390_restore_guest_per_regs(vcpu);
1153
935 if (exit_reason >= 0) { 1154 if (exit_reason >= 0) {
936 rc = 0; 1155 rc = 0;
937 } else if (kvm_is_ucontrol(vcpu->kvm)) { 1156 } else if (kvm_is_ucontrol(vcpu->kvm)) {
@@ -968,16 +1187,6 @@ static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason)
968 return rc; 1187 return rc;
969} 1188}
970 1189
971bool kvm_enabled_cmma(void)
972{
973 if (!MACHINE_IS_LPAR)
974 return false;
975 /* only enable for z10 and later */
976 if (!MACHINE_HAS_EDAT1)
977 return false;
978 return true;
979}
980
981static int __vcpu_run(struct kvm_vcpu *vcpu) 1190static int __vcpu_run(struct kvm_vcpu *vcpu)
982{ 1191{
983 int rc, exit_reason; 1192 int rc, exit_reason;
@@ -1007,7 +1216,7 @@ static int __vcpu_run(struct kvm_vcpu *vcpu)
1007 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); 1216 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1008 1217
1009 rc = vcpu_post_run(vcpu, exit_reason); 1218 rc = vcpu_post_run(vcpu, exit_reason);
1010 } while (!signal_pending(current) && !rc); 1219 } while (!signal_pending(current) && !guestdbg_exit_pending(vcpu) && !rc);
1011 1220
1012 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); 1221 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
1013 return rc; 1222 return rc;
@@ -1018,6 +1227,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
1018 int rc; 1227 int rc;
1019 sigset_t sigsaved; 1228 sigset_t sigsaved;
1020 1229
1230 if (guestdbg_exit_pending(vcpu)) {
1231 kvm_s390_prepare_debug_exit(vcpu);
1232 return 0;
1233 }
1234
1021 if (vcpu->sigset_active) 1235 if (vcpu->sigset_active)
1022 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); 1236 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
1023 1237
@@ -1030,6 +1244,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
1030 case KVM_EXIT_S390_RESET: 1244 case KVM_EXIT_S390_RESET:
1031 case KVM_EXIT_S390_UCONTROL: 1245 case KVM_EXIT_S390_UCONTROL:
1032 case KVM_EXIT_S390_TSCH: 1246 case KVM_EXIT_S390_TSCH:
1247 case KVM_EXIT_DEBUG:
1033 break; 1248 break;
1034 default: 1249 default:
1035 BUG(); 1250 BUG();
@@ -1055,6 +1270,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
1055 rc = -EINTR; 1270 rc = -EINTR;
1056 } 1271 }
1057 1272
1273 if (guestdbg_exit_pending(vcpu) && !rc) {
1274 kvm_s390_prepare_debug_exit(vcpu);
1275 rc = 0;
1276 }
1277
1058 if (rc == -EOPNOTSUPP) { 1278 if (rc == -EOPNOTSUPP) {
1059 /* intercept cannot be handled in-kernel, prepare kvm-run */ 1279 /* intercept cannot be handled in-kernel, prepare kvm-run */
1060 kvm_run->exit_reason = KVM_EXIT_S390_SIEIC; 1280 kvm_run->exit_reason = KVM_EXIT_S390_SIEIC;
@@ -1082,83 +1302,50 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
1082 return rc; 1302 return rc;
1083} 1303}
1084 1304
1085static int __guestcopy(struct kvm_vcpu *vcpu, u64 guestdest, void *from,
1086 unsigned long n, int prefix)
1087{
1088 if (prefix)
1089 return copy_to_guest(vcpu, guestdest, from, n);
1090 else
1091 return copy_to_guest_absolute(vcpu, guestdest, from, n);
1092}
1093
1094/* 1305/*
1095 * store status at address 1306 * store status at address
1096 * we use have two special cases: 1307 * we use have two special cases:
1097 * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit 1308 * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
1098 * KVM_S390_STORE_STATUS_PREFIXED: -> prefix 1309 * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
1099 */ 1310 */
1100int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long addr) 1311int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long gpa)
1101{ 1312{
1102 unsigned char archmode = 1; 1313 unsigned char archmode = 1;
1103 int prefix;
1104 u64 clkcomp; 1314 u64 clkcomp;
1315 int rc;
1105 1316
1106 if (addr == KVM_S390_STORE_STATUS_NOADDR) { 1317 if (gpa == KVM_S390_STORE_STATUS_NOADDR) {
1107 if (copy_to_guest_absolute(vcpu, 163ul, &archmode, 1)) 1318 if (write_guest_abs(vcpu, 163, &archmode, 1))
1108 return -EFAULT; 1319 return -EFAULT;
1109 addr = SAVE_AREA_BASE; 1320 gpa = SAVE_AREA_BASE;
1110 prefix = 0; 1321 } else if (gpa == KVM_S390_STORE_STATUS_PREFIXED) {
1111 } else if (addr == KVM_S390_STORE_STATUS_PREFIXED) { 1322 if (write_guest_real(vcpu, 163, &archmode, 1))
1112 if (copy_to_guest(vcpu, 163ul, &archmode, 1))
1113 return -EFAULT; 1323 return -EFAULT;
1114 addr = SAVE_AREA_BASE; 1324 gpa = kvm_s390_real_to_abs(vcpu, SAVE_AREA_BASE);
1115 prefix = 1; 1325 }
1116 } else 1326 rc = write_guest_abs(vcpu, gpa + offsetof(struct save_area, fp_regs),
1117 prefix = 0; 1327 vcpu->arch.guest_fpregs.fprs, 128);
1118 1328 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, gp_regs),
1119 if (__guestcopy(vcpu, addr + offsetof(struct save_area, fp_regs), 1329 vcpu->run->s.regs.gprs, 128);
1120 vcpu->arch.guest_fpregs.fprs, 128, prefix)) 1330 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, psw),
1121 return -EFAULT; 1331 &vcpu->arch.sie_block->gpsw, 16);
1122 1332 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, pref_reg),
1123 if (__guestcopy(vcpu, addr + offsetof(struct save_area, gp_regs), 1333 &vcpu->arch.sie_block->prefix, 4);
1124 vcpu->run->s.regs.gprs, 128, prefix)) 1334 rc |= write_guest_abs(vcpu,
1125 return -EFAULT; 1335 gpa + offsetof(struct save_area, fp_ctrl_reg),
1126 1336 &vcpu->arch.guest_fpregs.fpc, 4);
1127 if (__guestcopy(vcpu, addr + offsetof(struct save_area, psw), 1337 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, tod_reg),
1128 &vcpu->arch.sie_block->gpsw, 16, prefix)) 1338 &vcpu->arch.sie_block->todpr, 4);
1129 return -EFAULT; 1339 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, timer),
1130 1340 &vcpu->arch.sie_block->cputm, 8);
1131 if (__guestcopy(vcpu, addr + offsetof(struct save_area, pref_reg),
1132 &vcpu->arch.sie_block->prefix, 4, prefix))
1133 return -EFAULT;
1134
1135 if (__guestcopy(vcpu,
1136 addr + offsetof(struct save_area, fp_ctrl_reg),
1137 &vcpu->arch.guest_fpregs.fpc, 4, prefix))
1138 return -EFAULT;
1139
1140 if (__guestcopy(vcpu, addr + offsetof(struct save_area, tod_reg),
1141 &vcpu->arch.sie_block->todpr, 4, prefix))
1142 return -EFAULT;
1143
1144 if (__guestcopy(vcpu, addr + offsetof(struct save_area, timer),
1145 &vcpu->arch.sie_block->cputm, 8, prefix))
1146 return -EFAULT;
1147
1148 clkcomp = vcpu->arch.sie_block->ckc >> 8; 1341 clkcomp = vcpu->arch.sie_block->ckc >> 8;
1149 if (__guestcopy(vcpu, addr + offsetof(struct save_area, clk_cmp), 1342 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, clk_cmp),
1150 &clkcomp, 8, prefix)) 1343 &clkcomp, 8);
1151 return -EFAULT; 1344 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, acc_regs),
1152 1345 &vcpu->run->s.regs.acrs, 64);
1153 if (__guestcopy(vcpu, addr + offsetof(struct save_area, acc_regs), 1346 rc |= write_guest_abs(vcpu, gpa + offsetof(struct save_area, ctrl_regs),
1154 &vcpu->run->s.regs.acrs, 64, prefix)) 1347 &vcpu->arch.sie_block->gcr, 128);
1155 return -EFAULT; 1348 return rc ? -EFAULT : 0;
1156
1157 if (__guestcopy(vcpu,
1158 addr + offsetof(struct save_area, ctrl_regs),
1159 &vcpu->arch.sie_block->gcr, 128, prefix))
1160 return -EFAULT;
1161 return 0;
1162} 1349}
1163 1350
1164int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr) 1351int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
diff --git a/arch/s390/kvm/kvm-s390.h b/arch/s390/kvm/kvm-s390.h
index 3c1e2274d9ea..9b5680d1f6cc 100644
--- a/arch/s390/kvm/kvm-s390.h
+++ b/arch/s390/kvm/kvm-s390.h
@@ -28,7 +28,6 @@ int kvm_handle_sie_intercept(struct kvm_vcpu *vcpu);
28 28
29/* Transactional Memory Execution related macros */ 29/* Transactional Memory Execution related macros */
30#define IS_TE_ENABLED(vcpu) ((vcpu->arch.sie_block->ecb & 0x10)) 30#define IS_TE_ENABLED(vcpu) ((vcpu->arch.sie_block->ecb & 0x10))
31#define TDB_ADDR 0x1800UL
32#define TDB_FORMAT1 1 31#define TDB_FORMAT1 1
33#define IS_ITDB_VALID(vcpu) ((*(char *)vcpu->arch.sie_block->itdba == TDB_FORMAT1)) 32#define IS_ITDB_VALID(vcpu) ((*(char *)vcpu->arch.sie_block->itdba == TDB_FORMAT1))
34 33
@@ -130,6 +129,7 @@ void kvm_s390_tasklet(unsigned long parm);
130void kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu); 129void kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu);
131void kvm_s390_deliver_pending_machine_checks(struct kvm_vcpu *vcpu); 130void kvm_s390_deliver_pending_machine_checks(struct kvm_vcpu *vcpu);
132void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu); 131void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu);
132void kvm_s390_clear_float_irqs(struct kvm *kvm);
133int __must_check kvm_s390_inject_vm(struct kvm *kvm, 133int __must_check kvm_s390_inject_vm(struct kvm *kvm,
134 struct kvm_s390_interrupt *s390int); 134 struct kvm_s390_interrupt *s390int);
135int __must_check kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, 135int __must_check kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
@@ -137,6 +137,8 @@ int __must_check kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
137int __must_check kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code); 137int __must_check kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code);
138struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm, 138struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
139 u64 cr6, u64 schid); 139 u64 cr6, u64 schid);
140void kvm_s390_reinject_io_int(struct kvm *kvm,
141 struct kvm_s390_interrupt_info *inti);
140int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked); 142int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked);
141 143
142/* implemented in priv.c */ 144/* implemented in priv.c */
@@ -145,6 +147,7 @@ int kvm_s390_handle_e5(struct kvm_vcpu *vcpu);
145int kvm_s390_handle_01(struct kvm_vcpu *vcpu); 147int kvm_s390_handle_01(struct kvm_vcpu *vcpu);
146int kvm_s390_handle_b9(struct kvm_vcpu *vcpu); 148int kvm_s390_handle_b9(struct kvm_vcpu *vcpu);
147int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu); 149int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu);
150int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu);
148int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu); 151int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu);
149int kvm_s390_handle_eb(struct kvm_vcpu *vcpu); 152int kvm_s390_handle_eb(struct kvm_vcpu *vcpu);
150 153
@@ -158,14 +161,64 @@ void s390_vcpu_block(struct kvm_vcpu *vcpu);
158void s390_vcpu_unblock(struct kvm_vcpu *vcpu); 161void s390_vcpu_unblock(struct kvm_vcpu *vcpu);
159void exit_sie(struct kvm_vcpu *vcpu); 162void exit_sie(struct kvm_vcpu *vcpu);
160void exit_sie_sync(struct kvm_vcpu *vcpu); 163void exit_sie_sync(struct kvm_vcpu *vcpu);
161/* are we going to support cmma? */ 164int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu *vcpu);
162bool kvm_enabled_cmma(void); 165void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu *vcpu);
166/* is cmma enabled */
167bool kvm_s390_cmma_enabled(struct kvm *kvm);
168int test_vfacility(unsigned long nr);
169
163/* implemented in diag.c */ 170/* implemented in diag.c */
164int kvm_s390_handle_diag(struct kvm_vcpu *vcpu); 171int kvm_s390_handle_diag(struct kvm_vcpu *vcpu);
172/* implemented in interrupt.c */
173int kvm_s390_inject_prog_irq(struct kvm_vcpu *vcpu,
174 struct kvm_s390_pgm_info *pgm_info);
175
176/**
177 * kvm_s390_inject_prog_cond - conditionally inject a program check
178 * @vcpu: virtual cpu
179 * @rc: original return/error code
180 *
181 * This function is supposed to be used after regular guest access functions
182 * failed, to conditionally inject a program check to a vcpu. The typical
183 * pattern would look like
184 *
185 * rc = write_guest(vcpu, addr, data, len);
186 * if (rc)
187 * return kvm_s390_inject_prog_cond(vcpu, rc);
188 *
189 * A negative return code from guest access functions implies an internal error
190 * like e.g. out of memory. In these cases no program check should be injected
191 * to the guest.
192 * A positive value implies that an exception happened while accessing a guest's
193 * memory. In this case all data belonging to the corresponding program check
194 * has been stored in vcpu->arch.pgm and can be injected with
195 * kvm_s390_inject_prog_irq().
196 *
197 * Returns: - the original @rc value if @rc was negative (internal error)
198 * - zero if @rc was already zero
199 * - zero or error code from injecting if @rc was positive
200 * (program check injected to @vcpu)
201 */
202static inline int kvm_s390_inject_prog_cond(struct kvm_vcpu *vcpu, int rc)
203{
204 if (rc <= 0)
205 return rc;
206 return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
207}
165 208
166/* implemented in interrupt.c */ 209/* implemented in interrupt.c */
167int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu); 210int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
168int psw_extint_disabled(struct kvm_vcpu *vcpu); 211int psw_extint_disabled(struct kvm_vcpu *vcpu);
169void kvm_s390_destroy_adapters(struct kvm *kvm); 212void kvm_s390_destroy_adapters(struct kvm *kvm);
170 213
214/* implemented in guestdbg.c */
215void kvm_s390_backup_guest_per_regs(struct kvm_vcpu *vcpu);
216void kvm_s390_restore_guest_per_regs(struct kvm_vcpu *vcpu);
217void kvm_s390_patch_guest_per_regs(struct kvm_vcpu *vcpu);
218int kvm_s390_import_bp_data(struct kvm_vcpu *vcpu,
219 struct kvm_guest_debug *dbg);
220void kvm_s390_clear_bp_data(struct kvm_vcpu *vcpu);
221void kvm_s390_prepare_debug_exit(struct kvm_vcpu *vcpu);
222void kvm_s390_handle_per_event(struct kvm_vcpu *vcpu);
223
171#endif 224#endif
diff --git a/arch/s390/kvm/priv.c b/arch/s390/kvm/priv.c
index 476e9e218f43..27f9051a78f8 100644
--- a/arch/s390/kvm/priv.c
+++ b/arch/s390/kvm/priv.c
@@ -35,8 +35,8 @@ static int handle_set_clock(struct kvm_vcpu *vcpu)
35{ 35{
36 struct kvm_vcpu *cpup; 36 struct kvm_vcpu *cpup;
37 s64 hostclk, val; 37 s64 hostclk, val;
38 int i, rc;
38 u64 op2; 39 u64 op2;
39 int i;
40 40
41 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 41 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
42 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 42 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
@@ -44,8 +44,9 @@ static int handle_set_clock(struct kvm_vcpu *vcpu)
44 op2 = kvm_s390_get_base_disp_s(vcpu); 44 op2 = kvm_s390_get_base_disp_s(vcpu);
45 if (op2 & 7) /* Operand must be on a doubleword boundary */ 45 if (op2 & 7) /* Operand must be on a doubleword boundary */
46 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 46 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
47 if (get_guest(vcpu, val, (u64 __user *) op2)) 47 rc = read_guest(vcpu, op2, &val, sizeof(val));
48 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 48 if (rc)
49 return kvm_s390_inject_prog_cond(vcpu, rc);
49 50
50 if (store_tod_clock(&hostclk)) { 51 if (store_tod_clock(&hostclk)) {
51 kvm_s390_set_psw_cc(vcpu, 3); 52 kvm_s390_set_psw_cc(vcpu, 3);
@@ -65,8 +66,8 @@ static int handle_set_clock(struct kvm_vcpu *vcpu)
65static int handle_set_prefix(struct kvm_vcpu *vcpu) 66static int handle_set_prefix(struct kvm_vcpu *vcpu)
66{ 67{
67 u64 operand2; 68 u64 operand2;
68 u32 address = 0; 69 u32 address;
69 u8 tmp; 70 int rc;
70 71
71 vcpu->stat.instruction_spx++; 72 vcpu->stat.instruction_spx++;
72 73
@@ -80,14 +81,18 @@ static int handle_set_prefix(struct kvm_vcpu *vcpu)
80 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 81 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
81 82
82 /* get the value */ 83 /* get the value */
83 if (get_guest(vcpu, address, (u32 __user *) operand2)) 84 rc = read_guest(vcpu, operand2, &address, sizeof(address));
84 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 85 if (rc)
86 return kvm_s390_inject_prog_cond(vcpu, rc);
85 87
86 address = address & 0x7fffe000u; 88 address &= 0x7fffe000u;
87 89
88 /* make sure that the new value is valid memory */ 90 /*
89 if (copy_from_guest_absolute(vcpu, &tmp, address, 1) || 91 * Make sure the new value is valid memory. We only need to check the
90 (copy_from_guest_absolute(vcpu, &tmp, address + PAGE_SIZE, 1))) 92 * first page, since address is 8k aligned and memory pieces are always
93 * at least 1MB aligned and have at least a size of 1MB.
94 */
95 if (kvm_is_error_gpa(vcpu->kvm, address))
91 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 96 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
92 97
93 kvm_s390_set_prefix(vcpu, address); 98 kvm_s390_set_prefix(vcpu, address);
@@ -101,6 +106,7 @@ static int handle_store_prefix(struct kvm_vcpu *vcpu)
101{ 106{
102 u64 operand2; 107 u64 operand2;
103 u32 address; 108 u32 address;
109 int rc;
104 110
105 vcpu->stat.instruction_stpx++; 111 vcpu->stat.instruction_stpx++;
106 112
@@ -117,8 +123,9 @@ static int handle_store_prefix(struct kvm_vcpu *vcpu)
117 address = address & 0x7fffe000u; 123 address = address & 0x7fffe000u;
118 124
119 /* get the value */ 125 /* get the value */
120 if (put_guest(vcpu, address, (u32 __user *)operand2)) 126 rc = write_guest(vcpu, operand2, &address, sizeof(address));
121 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 127 if (rc)
128 return kvm_s390_inject_prog_cond(vcpu, rc);
122 129
123 VCPU_EVENT(vcpu, 5, "storing prefix to %x", address); 130 VCPU_EVENT(vcpu, 5, "storing prefix to %x", address);
124 trace_kvm_s390_handle_prefix(vcpu, 0, address); 131 trace_kvm_s390_handle_prefix(vcpu, 0, address);
@@ -127,28 +134,44 @@ static int handle_store_prefix(struct kvm_vcpu *vcpu)
127 134
128static int handle_store_cpu_address(struct kvm_vcpu *vcpu) 135static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
129{ 136{
130 u64 useraddr; 137 u16 vcpu_id = vcpu->vcpu_id;
138 u64 ga;
139 int rc;
131 140
132 vcpu->stat.instruction_stap++; 141 vcpu->stat.instruction_stap++;
133 142
134 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 143 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
135 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 144 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
136 145
137 useraddr = kvm_s390_get_base_disp_s(vcpu); 146 ga = kvm_s390_get_base_disp_s(vcpu);
138 147
139 if (useraddr & 1) 148 if (ga & 1)
140 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 149 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
141 150
142 if (put_guest(vcpu, vcpu->vcpu_id, (u16 __user *)useraddr)) 151 rc = write_guest(vcpu, ga, &vcpu_id, sizeof(vcpu_id));
143 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 152 if (rc)
153 return kvm_s390_inject_prog_cond(vcpu, rc);
144 154
145 VCPU_EVENT(vcpu, 5, "storing cpu address to %llx", useraddr); 155 VCPU_EVENT(vcpu, 5, "storing cpu address to %llx", ga);
146 trace_kvm_s390_handle_stap(vcpu, useraddr); 156 trace_kvm_s390_handle_stap(vcpu, ga);
147 return 0; 157 return 0;
148} 158}
149 159
160static void __skey_check_enable(struct kvm_vcpu *vcpu)
161{
162 if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
163 return;
164
165 s390_enable_skey();
166 trace_kvm_s390_skey_related_inst(vcpu);
167 vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
168}
169
170
150static int handle_skey(struct kvm_vcpu *vcpu) 171static int handle_skey(struct kvm_vcpu *vcpu)
151{ 172{
173 __skey_check_enable(vcpu);
174
152 vcpu->stat.instruction_storage_key++; 175 vcpu->stat.instruction_storage_key++;
153 176
154 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 177 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
@@ -160,9 +183,21 @@ static int handle_skey(struct kvm_vcpu *vcpu)
160 return 0; 183 return 0;
161} 184}
162 185
186static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
187{
188 psw_t *psw = &vcpu->arch.sie_block->gpsw;
189
190 vcpu->stat.instruction_ipte_interlock++;
191 if (psw_bits(*psw).p)
192 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
193 wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
194 psw->addr = __rewind_psw(*psw, 4);
195 VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
196 return 0;
197}
198
163static int handle_test_block(struct kvm_vcpu *vcpu) 199static int handle_test_block(struct kvm_vcpu *vcpu)
164{ 200{
165 unsigned long hva;
166 gpa_t addr; 201 gpa_t addr;
167 int reg2; 202 int reg2;
168 203
@@ -173,14 +208,13 @@ static int handle_test_block(struct kvm_vcpu *vcpu)
173 addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK; 208 addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
174 addr = kvm_s390_real_to_abs(vcpu, addr); 209 addr = kvm_s390_real_to_abs(vcpu, addr);
175 210
176 hva = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr)); 211 if (kvm_is_error_gpa(vcpu->kvm, addr))
177 if (kvm_is_error_hva(hva))
178 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 212 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
179 /* 213 /*
180 * We don't expect errors on modern systems, and do not care 214 * We don't expect errors on modern systems, and do not care
181 * about storage keys (yet), so let's just clear the page. 215 * about storage keys (yet), so let's just clear the page.
182 */ 216 */
183 if (clear_user((void __user *)hva, PAGE_SIZE) != 0) 217 if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
184 return -EFAULT; 218 return -EFAULT;
185 kvm_s390_set_psw_cc(vcpu, 0); 219 kvm_s390_set_psw_cc(vcpu, 0);
186 vcpu->run->s.regs.gprs[0] = 0; 220 vcpu->run->s.regs.gprs[0] = 0;
@@ -190,9 +224,12 @@ static int handle_test_block(struct kvm_vcpu *vcpu)
190static int handle_tpi(struct kvm_vcpu *vcpu) 224static int handle_tpi(struct kvm_vcpu *vcpu)
191{ 225{
192 struct kvm_s390_interrupt_info *inti; 226 struct kvm_s390_interrupt_info *inti;
227 unsigned long len;
228 u32 tpi_data[3];
229 int cc, rc;
193 u64 addr; 230 u64 addr;
194 int cc;
195 231
232 rc = 0;
196 addr = kvm_s390_get_base_disp_s(vcpu); 233 addr = kvm_s390_get_base_disp_s(vcpu);
197 if (addr & 3) 234 if (addr & 3)
198 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 235 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
@@ -201,30 +238,41 @@ static int handle_tpi(struct kvm_vcpu *vcpu)
201 if (!inti) 238 if (!inti)
202 goto no_interrupt; 239 goto no_interrupt;
203 cc = 1; 240 cc = 1;
241 tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
242 tpi_data[1] = inti->io.io_int_parm;
243 tpi_data[2] = inti->io.io_int_word;
204 if (addr) { 244 if (addr) {
205 /* 245 /*
206 * Store the two-word I/O interruption code into the 246 * Store the two-word I/O interruption code into the
207 * provided area. 247 * provided area.
208 */ 248 */
209 if (put_guest(vcpu, inti->io.subchannel_id, (u16 __user *)addr) 249 len = sizeof(tpi_data) - 4;
210 || put_guest(vcpu, inti->io.subchannel_nr, (u16 __user *)(addr + 2)) 250 rc = write_guest(vcpu, addr, &tpi_data, len);
211 || put_guest(vcpu, inti->io.io_int_parm, (u32 __user *)(addr + 4))) 251 if (rc)
212 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 252 return kvm_s390_inject_prog_cond(vcpu, rc);
213 } else { 253 } else {
214 /* 254 /*
215 * Store the three-word I/O interruption code into 255 * Store the three-word I/O interruption code into
216 * the appropriate lowcore area. 256 * the appropriate lowcore area.
217 */ 257 */
218 put_guest(vcpu, inti->io.subchannel_id, (u16 __user *) __LC_SUBCHANNEL_ID); 258 len = sizeof(tpi_data);
219 put_guest(vcpu, inti->io.subchannel_nr, (u16 __user *) __LC_SUBCHANNEL_NR); 259 if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len))
220 put_guest(vcpu, inti->io.io_int_parm, (u32 __user *) __LC_IO_INT_PARM); 260 rc = -EFAULT;
221 put_guest(vcpu, inti->io.io_int_word, (u32 __user *) __LC_IO_INT_WORD);
222 } 261 }
223 kfree(inti); 262 /*
263 * If we encounter a problem storing the interruption code, the
264 * instruction is suppressed from the guest's view: reinject the
265 * interrupt.
266 */
267 if (!rc)
268 kfree(inti);
269 else
270 kvm_s390_reinject_io_int(vcpu->kvm, inti);
224no_interrupt: 271no_interrupt:
225 /* Set condition code and we're done. */ 272 /* Set condition code and we're done. */
226 kvm_s390_set_psw_cc(vcpu, cc); 273 if (!rc)
227 return 0; 274 kvm_s390_set_psw_cc(vcpu, cc);
275 return rc ? -EFAULT : 0;
228} 276}
229 277
230static int handle_tsch(struct kvm_vcpu *vcpu) 278static int handle_tsch(struct kvm_vcpu *vcpu)
@@ -292,10 +340,10 @@ static int handle_stfl(struct kvm_vcpu *vcpu)
292 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 340 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
293 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 341 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
294 342
295 rc = copy_to_guest(vcpu, offsetof(struct _lowcore, stfl_fac_list), 343 rc = write_guest_lc(vcpu, offsetof(struct _lowcore, stfl_fac_list),
296 vfacilities, 4); 344 vfacilities, 4);
297 if (rc) 345 if (rc)
298 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 346 return rc;
299 VCPU_EVENT(vcpu, 5, "store facility list value %x", 347 VCPU_EVENT(vcpu, 5, "store facility list value %x",
300 *(unsigned int *) vfacilities); 348 *(unsigned int *) vfacilities);
301 trace_kvm_s390_handle_stfl(vcpu, *(unsigned int *) vfacilities); 349 trace_kvm_s390_handle_stfl(vcpu, *(unsigned int *) vfacilities);
@@ -333,6 +381,7 @@ int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
333 psw_t *gpsw = &vcpu->arch.sie_block->gpsw; 381 psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
334 psw_compat_t new_psw; 382 psw_compat_t new_psw;
335 u64 addr; 383 u64 addr;
384 int rc;
336 385
337 if (gpsw->mask & PSW_MASK_PSTATE) 386 if (gpsw->mask & PSW_MASK_PSTATE)
338 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 387 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
@@ -340,8 +389,10 @@ int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
340 addr = kvm_s390_get_base_disp_s(vcpu); 389 addr = kvm_s390_get_base_disp_s(vcpu);
341 if (addr & 7) 390 if (addr & 7)
342 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 391 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
343 if (copy_from_guest(vcpu, &new_psw, addr, sizeof(new_psw))) 392
344 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 393 rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
394 if (rc)
395 return kvm_s390_inject_prog_cond(vcpu, rc);
345 if (!(new_psw.mask & PSW32_MASK_BASE)) 396 if (!(new_psw.mask & PSW32_MASK_BASE))
346 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 397 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
347 gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32; 398 gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
@@ -357,6 +408,7 @@ static int handle_lpswe(struct kvm_vcpu *vcpu)
357{ 408{
358 psw_t new_psw; 409 psw_t new_psw;
359 u64 addr; 410 u64 addr;
411 int rc;
360 412
361 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 413 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
362 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 414 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
@@ -364,8 +416,9 @@ static int handle_lpswe(struct kvm_vcpu *vcpu)
364 addr = kvm_s390_get_base_disp_s(vcpu); 416 addr = kvm_s390_get_base_disp_s(vcpu);
365 if (addr & 7) 417 if (addr & 7)
366 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 418 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
367 if (copy_from_guest(vcpu, &new_psw, addr, sizeof(new_psw))) 419 rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
368 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 420 if (rc)
421 return kvm_s390_inject_prog_cond(vcpu, rc);
369 vcpu->arch.sie_block->gpsw = new_psw; 422 vcpu->arch.sie_block->gpsw = new_psw;
370 if (!is_valid_psw(&vcpu->arch.sie_block->gpsw)) 423 if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
371 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 424 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
@@ -375,7 +428,9 @@ static int handle_lpswe(struct kvm_vcpu *vcpu)
375 428
376static int handle_stidp(struct kvm_vcpu *vcpu) 429static int handle_stidp(struct kvm_vcpu *vcpu)
377{ 430{
431 u64 stidp_data = vcpu->arch.stidp_data;
378 u64 operand2; 432 u64 operand2;
433 int rc;
379 434
380 vcpu->stat.instruction_stidp++; 435 vcpu->stat.instruction_stidp++;
381 436
@@ -387,8 +442,9 @@ static int handle_stidp(struct kvm_vcpu *vcpu)
387 if (operand2 & 7) 442 if (operand2 & 7)
388 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 443 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
389 444
390 if (put_guest(vcpu, vcpu->arch.stidp_data, (u64 __user *)operand2)) 445 rc = write_guest(vcpu, operand2, &stidp_data, sizeof(stidp_data));
391 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 446 if (rc)
447 return kvm_s390_inject_prog_cond(vcpu, rc);
392 448
393 VCPU_EVENT(vcpu, 5, "%s", "store cpu id"); 449 VCPU_EVENT(vcpu, 5, "%s", "store cpu id");
394 return 0; 450 return 0;
@@ -474,9 +530,10 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
474 break; 530 break;
475 } 531 }
476 532
477 if (copy_to_guest_absolute(vcpu, operand2, (void *) mem, PAGE_SIZE)) { 533 rc = write_guest(vcpu, operand2, (void *)mem, PAGE_SIZE);
478 rc = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 534 if (rc) {
479 goto out_exception; 535 rc = kvm_s390_inject_prog_cond(vcpu, rc);
536 goto out;
480 } 537 }
481 trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2); 538 trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
482 free_page(mem); 539 free_page(mem);
@@ -485,7 +542,7 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
485 return 0; 542 return 0;
486out_no_data: 543out_no_data:
487 kvm_s390_set_psw_cc(vcpu, 3); 544 kvm_s390_set_psw_cc(vcpu, 3);
488out_exception: 545out:
489 free_page(mem); 546 free_page(mem);
490 return rc; 547 return rc;
491} 548}
@@ -496,6 +553,7 @@ static const intercept_handler_t b2_handlers[256] = {
496 [0x10] = handle_set_prefix, 553 [0x10] = handle_set_prefix,
497 [0x11] = handle_store_prefix, 554 [0x11] = handle_store_prefix,
498 [0x12] = handle_store_cpu_address, 555 [0x12] = handle_store_cpu_address,
556 [0x21] = handle_ipte_interlock,
499 [0x29] = handle_skey, 557 [0x29] = handle_skey,
500 [0x2a] = handle_skey, 558 [0x2a] = handle_skey,
501 [0x2b] = handle_skey, 559 [0x2b] = handle_skey,
@@ -513,6 +571,7 @@ static const intercept_handler_t b2_handlers[256] = {
513 [0x3a] = handle_io_inst, 571 [0x3a] = handle_io_inst,
514 [0x3b] = handle_io_inst, 572 [0x3b] = handle_io_inst,
515 [0x3c] = handle_io_inst, 573 [0x3c] = handle_io_inst,
574 [0x50] = handle_ipte_interlock,
516 [0x5f] = handle_io_inst, 575 [0x5f] = handle_io_inst,
517 [0x74] = handle_io_inst, 576 [0x74] = handle_io_inst,
518 [0x76] = handle_io_inst, 577 [0x76] = handle_io_inst,
@@ -618,6 +677,7 @@ static int handle_pfmf(struct kvm_vcpu *vcpu)
618 } 677 }
619 678
620 if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) { 679 if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
680 __skey_check_enable(vcpu);
621 if (set_guest_storage_key(current->mm, useraddr, 681 if (set_guest_storage_key(current->mm, useraddr,
622 vcpu->run->s.regs.gprs[reg1] & PFMF_KEY, 682 vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
623 vcpu->run->s.regs.gprs[reg1] & PFMF_NQ)) 683 vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
@@ -642,7 +702,7 @@ static int handle_essa(struct kvm_vcpu *vcpu)
642 VCPU_EVENT(vcpu, 5, "cmma release %d pages", entries); 702 VCPU_EVENT(vcpu, 5, "cmma release %d pages", entries);
643 gmap = vcpu->arch.gmap; 703 gmap = vcpu->arch.gmap;
644 vcpu->stat.instruction_essa++; 704 vcpu->stat.instruction_essa++;
645 if (!kvm_enabled_cmma() || !vcpu->arch.sie_block->cbrlo) 705 if (!kvm_s390_cmma_enabled(vcpu->kvm))
646 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION); 706 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
647 707
648 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 708 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
@@ -672,7 +732,10 @@ static int handle_essa(struct kvm_vcpu *vcpu)
672} 732}
673 733
674static const intercept_handler_t b9_handlers[256] = { 734static const intercept_handler_t b9_handlers[256] = {
735 [0x8a] = handle_ipte_interlock,
675 [0x8d] = handle_epsw, 736 [0x8d] = handle_epsw,
737 [0x8e] = handle_ipte_interlock,
738 [0x8f] = handle_ipte_interlock,
676 [0xab] = handle_essa, 739 [0xab] = handle_essa,
677 [0xaf] = handle_pfmf, 740 [0xaf] = handle_pfmf,
678}; 741};
@@ -693,32 +756,67 @@ int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
693{ 756{
694 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4; 757 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
695 int reg3 = vcpu->arch.sie_block->ipa & 0x000f; 758 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
696 u64 useraddr;
697 u32 val = 0; 759 u32 val = 0;
698 int reg, rc; 760 int reg, rc;
761 u64 ga;
699 762
700 vcpu->stat.instruction_lctl++; 763 vcpu->stat.instruction_lctl++;
701 764
702 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 765 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
703 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 766 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
704 767
705 useraddr = kvm_s390_get_base_disp_rs(vcpu); 768 ga = kvm_s390_get_base_disp_rs(vcpu);
706 769
707 if (useraddr & 3) 770 if (ga & 3)
708 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 771 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
709 772
710 VCPU_EVENT(vcpu, 5, "lctl r1:%x, r3:%x, addr:%llx", reg1, reg3, 773 VCPU_EVENT(vcpu, 5, "lctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
711 useraddr); 774 trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
712 trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, useraddr);
713 775
714 reg = reg1; 776 reg = reg1;
715 do { 777 do {
716 rc = get_guest(vcpu, val, (u32 __user *) useraddr); 778 rc = read_guest(vcpu, ga, &val, sizeof(val));
717 if (rc) 779 if (rc)
718 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 780 return kvm_s390_inject_prog_cond(vcpu, rc);
719 vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul; 781 vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
720 vcpu->arch.sie_block->gcr[reg] |= val; 782 vcpu->arch.sie_block->gcr[reg] |= val;
721 useraddr += 4; 783 ga += 4;
784 if (reg == reg3)
785 break;
786 reg = (reg + 1) % 16;
787 } while (1);
788
789 return 0;
790}
791
792int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
793{
794 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
795 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
796 u64 ga;
797 u32 val;
798 int reg, rc;
799
800 vcpu->stat.instruction_stctl++;
801
802 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
803 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
804
805 ga = kvm_s390_get_base_disp_rs(vcpu);
806
807 if (ga & 3)
808 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
809
810 VCPU_EVENT(vcpu, 5, "stctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
811 trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
812
813 reg = reg1;
814 do {
815 val = vcpu->arch.sie_block->gcr[reg] & 0x00000000fffffffful;
816 rc = write_guest(vcpu, ga, &val, sizeof(val));
817 if (rc)
818 return kvm_s390_inject_prog_cond(vcpu, rc);
819 ga += 4;
722 if (reg == reg3) 820 if (reg == reg3)
723 break; 821 break;
724 reg = (reg + 1) % 16; 822 reg = (reg + 1) % 16;
@@ -731,7 +829,7 @@ static int handle_lctlg(struct kvm_vcpu *vcpu)
731{ 829{
732 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4; 830 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
733 int reg3 = vcpu->arch.sie_block->ipa & 0x000f; 831 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
734 u64 useraddr; 832 u64 ga, val;
735 int reg, rc; 833 int reg, rc;
736 834
737 vcpu->stat.instruction_lctlg++; 835 vcpu->stat.instruction_lctlg++;
@@ -739,23 +837,58 @@ static int handle_lctlg(struct kvm_vcpu *vcpu)
739 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE) 837 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
740 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP); 838 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
741 839
742 useraddr = kvm_s390_get_base_disp_rsy(vcpu); 840 ga = kvm_s390_get_base_disp_rsy(vcpu);
743 841
744 if (useraddr & 7) 842 if (ga & 7)
745 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION); 843 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
746 844
747 reg = reg1; 845 reg = reg1;
748 846
749 VCPU_EVENT(vcpu, 5, "lctlg r1:%x, r3:%x, addr:%llx", reg1, reg3, 847 VCPU_EVENT(vcpu, 5, "lctlg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
750 useraddr); 848 trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
751 trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, useraddr);
752 849
753 do { 850 do {
754 rc = get_guest(vcpu, vcpu->arch.sie_block->gcr[reg], 851 rc = read_guest(vcpu, ga, &val, sizeof(val));
755 (u64 __user *) useraddr);
756 if (rc) 852 if (rc)
757 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); 853 return kvm_s390_inject_prog_cond(vcpu, rc);
758 useraddr += 8; 854 vcpu->arch.sie_block->gcr[reg] = val;
855 ga += 8;
856 if (reg == reg3)
857 break;
858 reg = (reg + 1) % 16;
859 } while (1);
860
861 return 0;
862}
863
864static int handle_stctg(struct kvm_vcpu *vcpu)
865{
866 int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
867 int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
868 u64 ga, val;
869 int reg, rc;
870
871 vcpu->stat.instruction_stctg++;
872
873 if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
874 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
875
876 ga = kvm_s390_get_base_disp_rsy(vcpu);
877
878 if (ga & 7)
879 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
880
881 reg = reg1;
882
883 VCPU_EVENT(vcpu, 5, "stctg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
884 trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
885
886 do {
887 val = vcpu->arch.sie_block->gcr[reg];
888 rc = write_guest(vcpu, ga, &val, sizeof(val));
889 if (rc)
890 return kvm_s390_inject_prog_cond(vcpu, rc);
891 ga += 8;
759 if (reg == reg3) 892 if (reg == reg3)
760 break; 893 break;
761 reg = (reg + 1) % 16; 894 reg = (reg + 1) % 16;
@@ -766,6 +899,7 @@ static int handle_lctlg(struct kvm_vcpu *vcpu)
766 899
767static const intercept_handler_t eb_handlers[256] = { 900static const intercept_handler_t eb_handlers[256] = {
768 [0x2f] = handle_lctlg, 901 [0x2f] = handle_lctlg,
902 [0x25] = handle_stctg,
769}; 903};
770 904
771int kvm_s390_handle_eb(struct kvm_vcpu *vcpu) 905int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
diff --git a/arch/s390/kvm/sigp.c b/arch/s390/kvm/sigp.c
index 26caeb530a78..c0b99e0f6b63 100644
--- a/arch/s390/kvm/sigp.c
+++ b/arch/s390/kvm/sigp.c
@@ -235,7 +235,6 @@ static int __sigp_set_prefix(struct kvm_vcpu *vcpu, u16 cpu_addr, u32 address,
235 struct kvm_vcpu *dst_vcpu = NULL; 235 struct kvm_vcpu *dst_vcpu = NULL;
236 struct kvm_s390_interrupt_info *inti; 236 struct kvm_s390_interrupt_info *inti;
237 int rc; 237 int rc;
238 u8 tmp;
239 238
240 if (cpu_addr < KVM_MAX_VCPUS) 239 if (cpu_addr < KVM_MAX_VCPUS)
241 dst_vcpu = kvm_get_vcpu(vcpu->kvm, cpu_addr); 240 dst_vcpu = kvm_get_vcpu(vcpu->kvm, cpu_addr);
@@ -243,10 +242,13 @@ static int __sigp_set_prefix(struct kvm_vcpu *vcpu, u16 cpu_addr, u32 address,
243 return SIGP_CC_NOT_OPERATIONAL; 242 return SIGP_CC_NOT_OPERATIONAL;
244 li = &dst_vcpu->arch.local_int; 243 li = &dst_vcpu->arch.local_int;
245 244
246 /* make sure that the new value is valid memory */ 245 /*
247 address = address & 0x7fffe000u; 246 * Make sure the new value is valid memory. We only need to check the
248 if (copy_from_guest_absolute(vcpu, &tmp, address, 1) || 247 * first page, since address is 8k aligned and memory pieces are always
249 copy_from_guest_absolute(vcpu, &tmp, address + PAGE_SIZE, 1)) { 248 * at least 1MB aligned and have at least a size of 1MB.
249 */
250 address &= 0x7fffe000u;
251 if (kvm_is_error_gpa(vcpu->kvm, address)) {
250 *reg &= 0xffffffff00000000UL; 252 *reg &= 0xffffffff00000000UL;
251 *reg |= SIGP_STATUS_INVALID_PARAMETER; 253 *reg |= SIGP_STATUS_INVALID_PARAMETER;
252 return SIGP_CC_STATUS_STORED; 254 return SIGP_CC_STATUS_STORED;
diff --git a/arch/s390/kvm/trace.h b/arch/s390/kvm/trace.h
index e8e7213d4cc5..abf6ba52769e 100644
--- a/arch/s390/kvm/trace.h
+++ b/arch/s390/kvm/trace.h
@@ -30,6 +30,20 @@
30 TP_printk("%02d[%016lx-%016lx]: " p_str, __entry->id, \ 30 TP_printk("%02d[%016lx-%016lx]: " p_str, __entry->id, \
31 __entry->pswmask, __entry->pswaddr, p_args) 31 __entry->pswmask, __entry->pswaddr, p_args)
32 32
33TRACE_EVENT(kvm_s390_skey_related_inst,
34 TP_PROTO(VCPU_PROTO_COMMON),
35 TP_ARGS(VCPU_ARGS_COMMON),
36
37 TP_STRUCT__entry(
38 VCPU_FIELD_COMMON
39 ),
40
41 TP_fast_assign(
42 VCPU_ASSIGN_COMMON
43 ),
44 VCPU_TP_PRINTK("%s", "first instruction related to skeys on vcpu")
45 );
46
33TRACE_EVENT(kvm_s390_major_guest_pfault, 47TRACE_EVENT(kvm_s390_major_guest_pfault,
34 TP_PROTO(VCPU_PROTO_COMMON), 48 TP_PROTO(VCPU_PROTO_COMMON),
35 TP_ARGS(VCPU_ARGS_COMMON), 49 TP_ARGS(VCPU_ARGS_COMMON),
@@ -301,6 +315,31 @@ TRACE_EVENT(kvm_s390_handle_lctl,
301 __entry->reg1, __entry->reg3, __entry->addr) 315 __entry->reg1, __entry->reg3, __entry->addr)
302 ); 316 );
303 317
318TRACE_EVENT(kvm_s390_handle_stctl,
319 TP_PROTO(VCPU_PROTO_COMMON, int g, int reg1, int reg3, u64 addr),
320 TP_ARGS(VCPU_ARGS_COMMON, g, reg1, reg3, addr),
321
322 TP_STRUCT__entry(
323 VCPU_FIELD_COMMON
324 __field(int, g)
325 __field(int, reg1)
326 __field(int, reg3)
327 __field(u64, addr)
328 ),
329
330 TP_fast_assign(
331 VCPU_ASSIGN_COMMON
332 __entry->g = g;
333 __entry->reg1 = reg1;
334 __entry->reg3 = reg3;
335 __entry->addr = addr;
336 ),
337
338 VCPU_TP_PRINTK("%s: storing cr %x-%x to %016llx",
339 __entry->g ? "stctg" : "stctl",
340 __entry->reg1, __entry->reg3, __entry->addr)
341 );
342
304TRACE_EVENT(kvm_s390_handle_prefix, 343TRACE_EVENT(kvm_s390_handle_prefix,
305 TP_PROTO(VCPU_PROTO_COMMON, int set, u32 address), 344 TP_PROTO(VCPU_PROTO_COMMON, int set, u32 address),
306 TP_ARGS(VCPU_ARGS_COMMON, set, address), 345 TP_ARGS(VCPU_ARGS_COMMON, set, address),
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index d7cfd57815fb..ea4a31b95990 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -832,6 +832,7 @@ void gmap_do_ipte_notify(struct mm_struct *mm, pte_t *pte)
832 } 832 }
833 spin_unlock(&gmap_notifier_lock); 833 spin_unlock(&gmap_notifier_lock);
834} 834}
835EXPORT_SYMBOL_GPL(gmap_do_ipte_notify);
835 836
836static inline int page_table_with_pgste(struct page *page) 837static inline int page_table_with_pgste(struct page *page)
837{ 838{
@@ -864,8 +865,7 @@ static inline unsigned long *page_table_alloc_pgste(struct mm_struct *mm,
864 atomic_set(&page->_mapcount, 0); 865 atomic_set(&page->_mapcount, 0);
865 table = (unsigned long *) page_to_phys(page); 866 table = (unsigned long *) page_to_phys(page);
866 clear_table(table, _PAGE_INVALID, PAGE_SIZE/2); 867 clear_table(table, _PAGE_INVALID, PAGE_SIZE/2);
867 clear_table(table + PTRS_PER_PTE, PGSTE_HR_BIT | PGSTE_HC_BIT, 868 clear_table(table + PTRS_PER_PTE, 0, PAGE_SIZE/2);
868 PAGE_SIZE/2);
869 return table; 869 return table;
870} 870}
871 871
@@ -883,8 +883,8 @@ static inline void page_table_free_pgste(unsigned long *table)
883 __free_page(page); 883 __free_page(page);
884} 884}
885 885
886static inline unsigned long page_table_reset_pte(struct mm_struct *mm, 886static inline unsigned long page_table_reset_pte(struct mm_struct *mm, pmd_t *pmd,
887 pmd_t *pmd, unsigned long addr, unsigned long end) 887 unsigned long addr, unsigned long end, bool init_skey)
888{ 888{
889 pte_t *start_pte, *pte; 889 pte_t *start_pte, *pte;
890 spinlock_t *ptl; 890 spinlock_t *ptl;
@@ -895,6 +895,22 @@ static inline unsigned long page_table_reset_pte(struct mm_struct *mm,
895 do { 895 do {
896 pgste = pgste_get_lock(pte); 896 pgste = pgste_get_lock(pte);
897 pgste_val(pgste) &= ~_PGSTE_GPS_USAGE_MASK; 897 pgste_val(pgste) &= ~_PGSTE_GPS_USAGE_MASK;
898 if (init_skey) {
899 unsigned long address;
900
901 pgste_val(pgste) &= ~(PGSTE_ACC_BITS | PGSTE_FP_BIT |
902 PGSTE_GR_BIT | PGSTE_GC_BIT);
903
904 /* skip invalid and not writable pages */
905 if (pte_val(*pte) & _PAGE_INVALID ||
906 !(pte_val(*pte) & _PAGE_WRITE)) {
907 pgste_set_unlock(pte, pgste);
908 continue;
909 }
910
911 address = pte_val(*pte) & PAGE_MASK;
912 page_set_storage_key(address, PAGE_DEFAULT_KEY, 1);
913 }
898 pgste_set_unlock(pte, pgste); 914 pgste_set_unlock(pte, pgste);
899 } while (pte++, addr += PAGE_SIZE, addr != end); 915 } while (pte++, addr += PAGE_SIZE, addr != end);
900 pte_unmap_unlock(start_pte, ptl); 916 pte_unmap_unlock(start_pte, ptl);
@@ -902,8 +918,8 @@ static inline unsigned long page_table_reset_pte(struct mm_struct *mm,
902 return addr; 918 return addr;
903} 919}
904 920
905static inline unsigned long page_table_reset_pmd(struct mm_struct *mm, 921static inline unsigned long page_table_reset_pmd(struct mm_struct *mm, pud_t *pud,
906 pud_t *pud, unsigned long addr, unsigned long end) 922 unsigned long addr, unsigned long end, bool init_skey)
907{ 923{
908 unsigned long next; 924 unsigned long next;
909 pmd_t *pmd; 925 pmd_t *pmd;
@@ -913,14 +929,14 @@ static inline unsigned long page_table_reset_pmd(struct mm_struct *mm,
913 next = pmd_addr_end(addr, end); 929 next = pmd_addr_end(addr, end);
914 if (pmd_none_or_clear_bad(pmd)) 930 if (pmd_none_or_clear_bad(pmd))
915 continue; 931 continue;
916 next = page_table_reset_pte(mm, pmd, addr, next); 932 next = page_table_reset_pte(mm, pmd, addr, next, init_skey);
917 } while (pmd++, addr = next, addr != end); 933 } while (pmd++, addr = next, addr != end);
918 934
919 return addr; 935 return addr;
920} 936}
921 937
922static inline unsigned long page_table_reset_pud(struct mm_struct *mm, 938static inline unsigned long page_table_reset_pud(struct mm_struct *mm, pgd_t *pgd,
923 pgd_t *pgd, unsigned long addr, unsigned long end) 939 unsigned long addr, unsigned long end, bool init_skey)
924{ 940{
925 unsigned long next; 941 unsigned long next;
926 pud_t *pud; 942 pud_t *pud;
@@ -930,14 +946,14 @@ static inline unsigned long page_table_reset_pud(struct mm_struct *mm,
930 next = pud_addr_end(addr, end); 946 next = pud_addr_end(addr, end);
931 if (pud_none_or_clear_bad(pud)) 947 if (pud_none_or_clear_bad(pud))
932 continue; 948 continue;
933 next = page_table_reset_pmd(mm, pud, addr, next); 949 next = page_table_reset_pmd(mm, pud, addr, next, init_skey);
934 } while (pud++, addr = next, addr != end); 950 } while (pud++, addr = next, addr != end);
935 951
936 return addr; 952 return addr;
937} 953}
938 954
939void page_table_reset_pgste(struct mm_struct *mm, 955void page_table_reset_pgste(struct mm_struct *mm, unsigned long start,
940 unsigned long start, unsigned long end) 956 unsigned long end, bool init_skey)
941{ 957{
942 unsigned long addr, next; 958 unsigned long addr, next;
943 pgd_t *pgd; 959 pgd_t *pgd;
@@ -949,7 +965,7 @@ void page_table_reset_pgste(struct mm_struct *mm,
949 next = pgd_addr_end(addr, end); 965 next = pgd_addr_end(addr, end);
950 if (pgd_none_or_clear_bad(pgd)) 966 if (pgd_none_or_clear_bad(pgd))
951 continue; 967 continue;
952 next = page_table_reset_pud(mm, pgd, addr, next); 968 next = page_table_reset_pud(mm, pgd, addr, next, init_skey);
953 } while (pgd++, addr = next, addr != end); 969 } while (pgd++, addr = next, addr != end);
954 up_read(&mm->mmap_sem); 970 up_read(&mm->mmap_sem);
955} 971}
@@ -989,7 +1005,7 @@ int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
989 /* changing the guest storage key is considered a change of the page */ 1005 /* changing the guest storage key is considered a change of the page */
990 if ((pgste_val(new) ^ pgste_val(old)) & 1006 if ((pgste_val(new) ^ pgste_val(old)) &
991 (PGSTE_ACC_BITS | PGSTE_FP_BIT | PGSTE_GR_BIT | PGSTE_GC_BIT)) 1007 (PGSTE_ACC_BITS | PGSTE_FP_BIT | PGSTE_GR_BIT | PGSTE_GC_BIT))
992 pgste_val(new) |= PGSTE_HC_BIT; 1008 pgste_val(new) |= PGSTE_UC_BIT;
993 1009
994 pgste_set_unlock(ptep, new); 1010 pgste_set_unlock(ptep, new);
995 pte_unmap_unlock(*ptep, ptl); 1011 pte_unmap_unlock(*ptep, ptl);
@@ -1011,6 +1027,11 @@ static inline unsigned long *page_table_alloc_pgste(struct mm_struct *mm,
1011 return NULL; 1027 return NULL;
1012} 1028}
1013 1029
1030void page_table_reset_pgste(struct mm_struct *mm, unsigned long start,
1031 unsigned long end, bool init_skey)
1032{
1033}
1034
1014static inline void page_table_free_pgste(unsigned long *table) 1035static inline void page_table_free_pgste(unsigned long *table)
1015{ 1036{
1016} 1037}
@@ -1357,6 +1378,50 @@ int s390_enable_sie(void)
1357} 1378}
1358EXPORT_SYMBOL_GPL(s390_enable_sie); 1379EXPORT_SYMBOL_GPL(s390_enable_sie);
1359 1380
1381/*
1382 * Enable storage key handling from now on and initialize the storage
1383 * keys with the default key.
1384 */
1385void s390_enable_skey(void)
1386{
1387 /*
1388 * To avoid races between multiple vcpus, ending in calling
1389 * page_table_reset twice or more,
1390 * the page_table_lock is taken for serialization.
1391 */
1392 spin_lock(&current->mm->page_table_lock);
1393 if (mm_use_skey(current->mm)) {
1394 spin_unlock(&current->mm->page_table_lock);
1395 return;
1396 }
1397
1398 current->mm->context.use_skey = 1;
1399 spin_unlock(&current->mm->page_table_lock);
1400 page_table_reset_pgste(current->mm, 0, TASK_SIZE, true);
1401}
1402EXPORT_SYMBOL_GPL(s390_enable_skey);
1403
1404/*
1405 * Test and reset if a guest page is dirty
1406 */
1407bool gmap_test_and_clear_dirty(unsigned long address, struct gmap *gmap)
1408{
1409 pte_t *pte;
1410 spinlock_t *ptl;
1411 bool dirty = false;
1412
1413 pte = get_locked_pte(gmap->mm, address, &ptl);
1414 if (unlikely(!pte))
1415 return false;
1416
1417 if (ptep_test_and_clear_user_dirty(gmap->mm, address, pte))
1418 dirty = true;
1419
1420 spin_unlock(ptl);
1421 return dirty;
1422}
1423EXPORT_SYMBOL_GPL(gmap_test_and_clear_dirty);
1424
1360#ifdef CONFIG_TRANSPARENT_HUGEPAGE 1425#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1361int pmdp_clear_flush_young(struct vm_area_struct *vma, unsigned long address, 1426int pmdp_clear_flush_young(struct vm_area_struct *vma, unsigned long address,
1362 pmd_t *pmdp) 1427 pmd_t *pmdp)
diff --git a/drivers/s390/char/sclp_early.c b/drivers/s390/char/sclp_early.c
index 14196ea0fdf3..b57fe0efb422 100644
--- a/drivers/s390/char/sclp_early.c
+++ b/drivers/s390/char/sclp_early.c
@@ -22,7 +22,8 @@ struct read_info_sccb {
22 u8 rnsize; /* 10 */ 22 u8 rnsize; /* 10 */
23 u8 _reserved0[16 - 11]; /* 11-15 */ 23 u8 _reserved0[16 - 11]; /* 11-15 */
24 u16 ncpurl; /* 16-17 */ 24 u16 ncpurl; /* 16-17 */
25 u8 _reserved7[24 - 18]; /* 18-23 */ 25 u16 cpuoff; /* 18-19 */
26 u8 _reserved7[24 - 20]; /* 20-23 */
26 u8 loadparm[8]; /* 24-31 */ 27 u8 loadparm[8]; /* 24-31 */
27 u8 _reserved1[48 - 32]; /* 32-47 */ 28 u8 _reserved1[48 - 32]; /* 32-47 */
28 u64 facilities; /* 48-55 */ 29 u64 facilities; /* 48-55 */
@@ -45,6 +46,7 @@ static unsigned int sclp_con_has_linemode __initdata;
45static unsigned long sclp_hsa_size; 46static unsigned long sclp_hsa_size;
46static unsigned int sclp_max_cpu; 47static unsigned int sclp_max_cpu;
47static struct sclp_ipl_info sclp_ipl_info; 48static struct sclp_ipl_info sclp_ipl_info;
49static unsigned char sclp_siif;
48 50
49u64 sclp_facilities; 51u64 sclp_facilities;
50u8 sclp_fac84; 52u8 sclp_fac84;
@@ -96,6 +98,9 @@ static int __init sclp_read_info_early(struct read_info_sccb *sccb)
96 98
97static void __init sclp_facilities_detect(struct read_info_sccb *sccb) 99static void __init sclp_facilities_detect(struct read_info_sccb *sccb)
98{ 100{
101 struct sclp_cpu_entry *cpue;
102 u16 boot_cpu_address, cpu;
103
99 if (sclp_read_info_early(sccb)) 104 if (sclp_read_info_early(sccb))
100 return; 105 return;
101 106
@@ -116,6 +121,15 @@ static void __init sclp_facilities_detect(struct read_info_sccb *sccb)
116 sclp_max_cpu = sccb->hcpua + 1; 121 sclp_max_cpu = sccb->hcpua + 1;
117 } 122 }
118 123
124 boot_cpu_address = stap();
125 cpue = (void *)sccb + sccb->cpuoff;
126 for (cpu = 0; cpu < sccb->ncpurl; cpue++, cpu++) {
127 if (boot_cpu_address != cpue->address)
128 continue;
129 sclp_siif = cpue->siif;
130 break;
131 }
132
119 /* Save IPL information */ 133 /* Save IPL information */
120 sclp_ipl_info.is_valid = 1; 134 sclp_ipl_info.is_valid = 1;
121 if (sccb->flags & 0x2) 135 if (sccb->flags & 0x2)
@@ -148,6 +162,12 @@ unsigned int sclp_get_max_cpu(void)
148 return sclp_max_cpu; 162 return sclp_max_cpu;
149} 163}
150 164
165int sclp_has_siif(void)
166{
167 return sclp_siif;
168}
169EXPORT_SYMBOL(sclp_has_siif);
170
151/* 171/*
152 * This function will be called after sclp_facilities_detect(), which gets 172 * This function will be called after sclp_facilities_detect(), which gets
153 * called from early.c code. The sclp_facilities_detect() function retrieves 173 * called from early.c code. The sclp_facilities_detect() function retrieves
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 6c3c2eb96d06..32d263f683dc 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -880,6 +880,13 @@ static inline hpa_t pfn_to_hpa(pfn_t pfn)
880 return (hpa_t)pfn << PAGE_SHIFT; 880 return (hpa_t)pfn << PAGE_SHIFT;
881} 881}
882 882
883static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
884{
885 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
886
887 return kvm_is_error_hva(hva);
888}
889
883static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu) 890static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
884{ 891{
885 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests); 892 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index d8a6ce4c2a83..836e15b7abc8 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -745,6 +745,7 @@ struct kvm_ppc_smmu_info {
745#define KVM_CAP_ENABLE_CAP_VM 98 745#define KVM_CAP_ENABLE_CAP_VM 98
746#define KVM_CAP_S390_IRQCHIP 99 746#define KVM_CAP_S390_IRQCHIP 99
747#define KVM_CAP_IOEVENTFD_NO_LENGTH 100 747#define KVM_CAP_IOEVENTFD_NO_LENGTH 100
748#define KVM_CAP_VM_ATTRIBUTES 101
748 749
749#ifdef KVM_CAP_IRQ_ROUTING 750#ifdef KVM_CAP_IRQ_ROUTING
750 751
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 96456ac888ba..ea46d64c8e75 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -637,14 +637,12 @@ static int kvm_vm_release(struct inode *inode, struct file *filp)
637 */ 637 */
638static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot) 638static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
639{ 639{
640#ifndef CONFIG_S390
641 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot); 640 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
642 641
643 memslot->dirty_bitmap = kvm_kvzalloc(dirty_bytes); 642 memslot->dirty_bitmap = kvm_kvzalloc(dirty_bytes);
644 if (!memslot->dirty_bitmap) 643 if (!memslot->dirty_bitmap)
645 return -ENOMEM; 644 return -ENOMEM;
646 645
647#endif /* !CONFIG_S390 */
648 return 0; 646 return 0;
649} 647}
650 648
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-19 22:58:53 -0500