Merge tag 'nmiforkvm' of git://git.kernel.org/pub/scm/linux/kernel/git/kvms390/linux into features

Pull kvm patches from Christian Borntraeger:
"s390,kvm: provide plumbing for machines checks when running guests"

This provides the basic plumbing for handling machine checks when
running guests

7 files changed, 115 insertions, 12 deletions

diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h
index 426614a882a9..c6e1d5fa1ad1 100644
--- a/arch/s390/include/asm/kvm_host.h
+++ b/arch/s390/include/asm/kvm_host.h
@@ -107,6 +107,20 @@ struct esca_block {
         struct esca_entry cpu[KVM_S390_ESCA_CPU_SLOTS];
 } __packed;
 
+/*
+ * This struct is used to store some machine check info from lowcore
+ * for machine checks that happen while the guest is running.
+ * This info in host's lowcore might be overwritten by a second machine
+ * check from host when host is in the machine check's high-level handling.
+ * The size is 24 bytes.
+ */
+struct mcck_volatile_info {
+        __u64 mcic;
+        __u64 failing_storage_address;
+        __u32 ext_damage_code;
+        __u32 reserved;
+};
+
 #define CPUSTAT_STOPPED    0x80000000
 #define CPUSTAT_WAIT       0x10000000
 #define CPUSTAT_ECALL_PEND 0x08000000
@@ -264,7 +278,8 @@ struct kvm_s390_itdb {
 
 struct sie_page {
         struct kvm_s390_sie_block sie_block;
-        __u8 reserved200[1024];         /* 0x0200 */
+        struct mcck_volatile_info mcck_info;    /* 0x0200 */
+        __u8 reserved218[1000];         /* 0x0218 */
         struct kvm_s390_itdb itdb;      /* 0x0600 */
         __u8 reserved700[2304];         /* 0x0700 */
 } __packed;
diff --git a/arch/s390/include/asm/nmi.h b/arch/s390/include/asm/nmi.h
index e3e8895f5d3e..13623b9991d4 100644
--- a/arch/s390/include/asm/nmi.h
+++ b/arch/s390/include/asm/nmi.h
@@ -14,7 +14,14 @@
 #include <linux/const.h>
 #include <linux/types.h>
 
+#define MCIC_SUBCLASS_MASK      (1ULL<<63 | 1ULL<<62 | 1ULL<<61 | \
+                                1ULL<<59 | 1ULL<<58 | 1ULL<<56 | \
+                                1ULL<<55 | 1ULL<<54 | 1ULL<<53 | \
+                                1ULL<<52 | 1ULL<<47 | 1ULL<<46 | \
+                                1ULL<<45 | 1ULL<<44)
 #define MCCK_CODE_SYSTEM_DAMAGE         _BITUL(63)
+#define MCCK_CODE_EXT_DAMAGE            _BITUL(63 - 5)
+#define MCCK_CODE_CP                    _BITUL(63 - 9)
 #define MCCK_CODE_CPU_TIMER_VALID       _BITUL(63 - 46)
 #define MCCK_CODE_PSW_MWP_VALID         _BITUL(63 - 20)
 #define MCCK_CODE_PSW_IA_VALID          _BITUL(63 - 23)
diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h
index f57c017a5c03..72c7b88f8d2c 100644
--- a/arch/s390/include/asm/processor.h
+++ b/arch/s390/include/asm/processor.h
@@ -20,6 +20,7 @@
 #define CIF_FPU                 4       /* restore FPU registers */
 #define CIF_IGNORE_IRQ          5       /* ignore interrupt (for udelay) */
 #define CIF_ENABLED_WAIT        6       /* in enabled wait state */
+#define CIF_MCCK_GUEST          7       /* machine check happening in guest */
 
 #define _CIF_MCCK_PENDING       _BITUL(CIF_MCCK_PENDING)
 #define _CIF_ASCE_PRIMARY       _BITUL(CIF_ASCE_PRIMARY)
@@ -28,6 +29,7 @@
 #define _CIF_FPU                _BITUL(CIF_FPU)
 #define _CIF_IGNORE_IRQ         _BITUL(CIF_IGNORE_IRQ)
 #define _CIF_ENABLED_WAIT       _BITUL(CIF_ENABLED_WAIT)
+#define _CIF_MCCK_GUEST         _BITUL(CIF_MCCK_GUEST)
 
 #ifndef __ASSEMBLY__
 
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index 6bb29633e1f1..b65c414b6c0e 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -58,6 +58,9 @@ int main(void)
         OFFSET(__SF_BACKCHAIN, stack_frame, back_chain);
         OFFSET(__SF_GPRS, stack_frame, gprs);
         OFFSET(__SF_EMPTY, stack_frame, empty1);
+        OFFSET(__SF_SIE_CONTROL, stack_frame, empty1[0]);
+        OFFSET(__SF_SIE_SAVEAREA, stack_frame, empty1[1]);
+        OFFSET(__SF_SIE_REASON, stack_frame, empty1[2]);
         BLANK();
         /* timeval/timezone offsets for use by vdso */
         OFFSET(__VDSO_UPD_COUNT, vdso_data, tb_update_count);
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index e83aff630bcf..21900e1cee9c 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -225,6 +225,7 @@ ENTRY(sie64a)
         jnz     .Lsie_skip
         TSTMSK  __LC_CPU_FLAGS,_CIF_FPU
         jo      .Lsie_skip                      # exit if fp/vx regs changed
+.Lsie_entry:
         sie     0(%r14)
 .Lsie_skip:
         ni      __SIE_PROG0C+3(%r14),0xfe       # no longer in SIE
@@ -1122,7 +1123,13 @@ cleanup_critical:
         .quad   .Lsie_done
 
 .Lcleanup_sie:
-        lg      %r9,__SF_EMPTY(%r15)            # get control block pointer
+        cghi    %r11,__LC_SAVE_AREA_ASYNC       #Is this in normal interrupt?
+        je      1f
+        slg     %r9,BASED(.Lsie_crit_mcck_start)
+        clg     %r9,BASED(.Lsie_crit_mcck_length)
+        jh      1f
+        oi      __LC_CPU_FLAGS+7, _CIF_MCCK_GUEST
+1:      lg      %r9,__SF_EMPTY(%r15)            # get control block pointer
         ni      __SIE_PROG0C+3(%r9),0xfe        # no longer in SIE
         lctlg   %c1,%c1,__LC_USER_ASCE          # load primary asce
         larl    %r9,sie_exit                    # skip forward to sie_exit
@@ -1307,6 +1314,10 @@ cleanup_critical:
         .quad   .Lsie_gmap
 .Lsie_critical_length:
         .quad   .Lsie_done - .Lsie_gmap
+.Lsie_crit_mcck_start:
+        .quad   .Lsie_entry
+.Lsie_crit_mcck_length:
+        .quad   .Lsie_skip - .Lsie_entry
 #endif
 
         .section .rodata, "a"
diff --git a/arch/s390/kernel/nmi.c b/arch/s390/kernel/nmi.c
index 985589523970..31d03a84126c 100644
--- a/arch/s390/kernel/nmi.c
+++ b/arch/s390/kernel/nmi.c
@@ -25,6 +25,8 @@
 #include <asm/crw.h>
 #include <asm/switch_to.h>
 #include <asm/ctl_reg.h>
+#include <asm/asm-offsets.h>
+#include <linux/kvm_host.h>
 
 struct mcck_struct {
         unsigned int kill_task : 1;
@@ -274,12 +276,39 @@ static int notrace s390_validate_registers(union mci mci, int umode)
         return kill_task;
 }
 
+/*
+ * Backup the guest's machine check info to its description block
+ */
+static void notrace s390_backup_mcck_info(struct pt_regs *regs)
+{
+        struct mcck_volatile_info *mcck_backup;
+        struct sie_page *sie_page;
+
+        /* r14 contains the sie block, which was set in sie64a */
+        struct kvm_s390_sie_block *sie_block =
+                        (struct kvm_s390_sie_block *) regs->gprs[14];
+
+        if (sie_block == NULL)
+                /* Something's seriously wrong, stop system. */
+                s390_handle_damage();
+
+        sie_page = container_of(sie_block, struct sie_page, sie_block);
+        mcck_backup = &sie_page->mcck_info;
+        mcck_backup->mcic = S390_lowcore.mcck_interruption_code &
+                                ~(MCCK_CODE_CP | MCCK_CODE_EXT_DAMAGE);
+        mcck_backup->ext_damage_code = S390_lowcore.external_damage_code;
+        mcck_backup->failing_storage_address
+                        = S390_lowcore.failing_storage_address;
+}
+
 #define MAX_IPD_COUNT   29
 #define MAX_IPD_TIME    (5 * 60 * USEC_PER_SEC) /* 5 minutes */
 
 #define ED_STP_ISLAND   6       /* External damage STP island check */
 #define ED_STP_SYNC     7       /* External damage STP sync check */
 
+#define MCCK_CODE_NO_GUEST      (MCCK_CODE_CP | MCCK_CODE_EXT_DAMAGE)
+
 /*
  * machine check handler.
  */
@@ -291,6 +320,7 @@ void notrace s390_do_machine_check(struct pt_regs *regs)
         struct mcck_struct *mcck;
         unsigned long long tmp;
         union mci mci;
+        unsigned long mcck_dam_code;
 
         nmi_enter();
         inc_irq_stat(NMI_NMI);
@@ -301,7 +331,13 @@ void notrace s390_do_machine_check(struct pt_regs *regs)
                 /* System damage -> stopping machine */
                 s390_handle_damage();
         }
-        if (mci.pd) {
+
+        /*
+         * Reinject the instruction processing damages' machine checks
+         * including Delayed Access Exception into the guest
+         * instead of damaging the host if they happen in the guest.
+         */
+        if (mci.pd && !test_cpu_flag(CIF_MCCK_GUEST)) {
                 if (mci.b) {
                         /* Processing backup -> verify if we can survive this */
                         u64 z_mcic, o_mcic, t_mcic;
@@ -345,6 +381,14 @@ void notrace s390_do_machine_check(struct pt_regs *regs)
                 mcck->mcck_code = mci.val;
                 set_cpu_flag(CIF_MCCK_PENDING);
         }
+
+        /*
+         * Backup the machine check's info if it happens when the guest
+         * is running.
+         */
+        if (test_cpu_flag(CIF_MCCK_GUEST))
+                s390_backup_mcck_info(regs);
+
         if (mci.cd) {
                 /* Timing facility damage */
                 s390_handle_damage();
@@ -358,15 +402,22 @@ void notrace s390_do_machine_check(struct pt_regs *regs)
                 if (mcck->stp_queue)
                         set_cpu_flag(CIF_MCCK_PENDING);
         }
-        if (mci.se)
-                /* Storage error uncorrected */
-                s390_handle_damage();
-        if (mci.ke)
-                /* Storage key-error uncorrected */
-                s390_handle_damage();
-        if (mci.ds && mci.fa)
-                /* Storage degradation */
-                s390_handle_damage();
+
+        /*
+         * Reinject storage related machine checks into the guest if they
+         * happen when the guest is running.
+         */
+        if (!test_cpu_flag(CIF_MCCK_GUEST)) {
+                if (mci.se)
+                        /* Storage error uncorrected */
+                        s390_handle_damage();
+                if (mci.ke)
+                        /* Storage key-error uncorrected */
+                        s390_handle_damage();
+                if (mci.ds && mci.fa)
+                        /* Storage degradation */
+                        s390_handle_damage();
+        }
         if (mci.cp) {
                 /* Channel report word pending */
                 mcck->channel_report = 1;
@@ -377,6 +428,19 @@ void notrace s390_do_machine_check(struct pt_regs *regs)
                 mcck->warning = 1;
                 set_cpu_flag(CIF_MCCK_PENDING);
         }
+
+        /*
+         * If there are only Channel Report Pending and External Damage
+         * machine checks, they will not be reinjected into the guest
+         * because they refer to host conditions only.
+         */
+        mcck_dam_code = (mci.val & MCIC_SUBCLASS_MASK);
+        if (test_cpu_flag(CIF_MCCK_GUEST) &&
+        (mcck_dam_code & MCCK_CODE_NO_GUEST) != mcck_dam_code) {
+                /* Set exit reason code for host's later handling */
+                *((long *)(regs->gprs[15] + __SF_SIE_REASON)) = -EINTR;
+        }
+        clear_cpu_flag(CIF_MCCK_GUEST);
         nmi_exit();
 }
 
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index 689ac48361c6..0457e03199c5 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -2069,6 +2069,7 @@ struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm,
         if (!vcpu)
                 goto out;
 
+        BUILD_BUG_ON(sizeof(struct sie_page) != 4096);
         sie_page = (struct sie_page *) get_zeroed_page(GFP_KERNEL);
         if (!sie_page)
                 goto out_free_cpu;