[Qemu-devel] [PULL v3 00/17] KVM and cpu-exec patches for 2.9 soft freeze

[Qemu-devel] [PULL v3 00/17] KVM and cpu-exec patches for 2.9 soft freeze

[Paolo Bonzini]

The following changes since commit d992f2f1368ceb92e6bfd8efece174110f4236ff:

  Merge remote-tracking branch 'remotes/artyom/tags/pull-sun4v-20170226' into staging (2017-02-26 22:40:23 +0000)

are available in the git repository at:

  git://github.com/bonzini/qemu.git tags/for-upstream

for you to fetch changes up to 0819248d30877a9e0f063e7246e522a67548fc3e:

  qmp-events: fix GUEST_PANICKED description formatting (2017-02-27 16:31:08 +0100)

v2->v3: replace SIGPOLL with SIGIO

        another missing #ifdef KVM_HAVE_MCE_INJECTION

v1->v2: add missing #ifdef KVM_HAVE_MCE_INJECTION

        rebase over MTTCG pull request.  -icount is currently broken but
        I'd like to preserve bisectability, so this pull request brings
        in the icount patches through a merge commit.

----------------------------------------------------------------
* kernel header update (requested by David and Vijay)
* GuestPanicInformation fixups (Anton)
* record/replay icount fixes (Pavel)
* cpu-exec cleanup, unification of icount_decr with tcg_exit_req (me)
* KVM_CAP_IMMEDIATE_EXIT support (me)
* vmxcap update (me)

----------------------------------------------------------------
Anton Nefedov (2):
      qapi: flatten GuestPanicInformation union
      qmp-events: fix GUEST_PANICKED description formatting

Paolo Bonzini (14):
      cpu-exec: unify icount_decr and tcg_exit_req
      cpu-exec: remove unnecessary check of cpu->exit_request
      update-linux-headers: update for 4.11
      update Linux headers to 4.11
      cpus: remove ugly cast on sigbus_handler
      KVM: x86: cleanup SIGBUS handlers
      cpus: reorganize signal handling code
      KVM: remove kvm_arch_on_sigbus
      KVM: do not use sigtimedwait to catch SIGBUS
      KVM: move SIG_IPI handling to kvm-all.c
      kvm: use atomic_read/atomic_set to access cpu->exit_request
      KVM: use KVM_CAP_IMMEDIATE_EXIT
      vmxcap: port to Python 3
      vmxcap: update for September 2016 SDM

Pavel Dovgalyuk (1):
      replay: check icount in cpu exec loop

 cpu-exec.c                                         |  93 ++---
 cpus.c                                             | 102 +----
 include/exec/gen-icount.h                          |  53 ++-
 include/qemu/compatfd.h                            |  42 ---
 include/qemu/osdep.h                               |  37 ++
 include/qom/cpu.h                                  |  15 +-
 include/standard-headers/asm-x86/hyperv.h          |   8 +
 include/standard-headers/linux/input-event-codes.h |   2 +-
 include/standard-headers/linux/pci_regs.h          |  25 ++
 include/standard-headers/linux/virtio_ids.h        |   1 +
 include/sysemu/kvm.h                               |  11 +-
 kvm-all.c                                          | 152 +++++++-
 kvm-stub.c                                         |  12 +-
 linux-headers/asm-arm/kvm.h                        |  15 +
 linux-headers/asm-arm/unistd-common.h              | 357 ++++++++++++++++++
 linux-headers/asm-arm/unistd-eabi.h                |   5 +
 linux-headers/asm-arm/unistd-oabi.h                |  17 +
 linux-headers/asm-arm/unistd.h                     | 419 +--------------------
 linux-headers/asm-arm64/kvm.h                      |  13 +
 linux-headers/asm-powerpc/kvm.h                    |  27 ++
 linux-headers/asm-powerpc/unistd.h                 |   1 +
 linux-headers/asm-x86/kvm_para.h                   |  13 +-
 linux-headers/linux/kvm.h                          |  24 +-
 linux-headers/linux/kvm_para.h                     |   2 +
 linux-headers/linux/userfaultfd.h                  |  67 +++-
 linux-headers/linux/vfio.h                         |  10 +
 qapi-schema.json                                   |  12 +
 qapi/event.json                                    |   4 +-
 qom/cpu.c                                          |   2 +-
 scripts/kvm/vmxcap                                 |  23 +-
 scripts/update-linux-headers.sh                    |  13 +-
 target/arm/kvm.c                                   |  10 -
 target/i386/cpu.c                                  |  15 +-
 target/i386/kvm.c                                  |  81 ++--
 target/mips/kvm.c                                  |  12 -
 target/ppc/kvm.c                                   |  10 -
 target/s390x/kvm.c                                 |  10 -
 tcg/tcg.h                                          |   1 -
 translate-all.c                                    |   2 +-
 translate-common.c                                 |  13 +-
 util/compatfd.c                                    |   1 -
 util/main-loop.c                                   |   5 +-
 util/oslib-posix.c                                 |  33 ++
 vl.c                                               |  12 +-
 44 files changed, 970 insertions(+), 812 deletions(-)
 delete mode 100644 include/qemu/compatfd.h
 create mode 100644 linux-headers/asm-arm/unistd-common.h
 create mode 100644 linux-headers/asm-arm/unistd-eabi.h
 create mode 100644 linux-headers/asm-arm/unistd-oabi.h

[Qemu-devel] [PULL 06/17] cpus: remove ugly cast on sigbus_handler

[Paolo Bonzini]

The cast is there because sigbus_handler is invoked via sigfd_handler.
But it feels just wrong to use struct qemu_signalfd_siginfo in the
prototype of a function that is passed to sigaction.

Instead, do a simple-minded conversion of qemu_signalfd_siginfo to
siginfo_t.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 cpus.c                  | 12 +++---------
 include/qemu/compatfd.h | 42 ------------------------------------------
 include/qemu/osdep.h    | 28 ++++++++++++++++++++++++++++
 util/compatfd.c         |  1 -
 util/main-loop.c        |  5 +----
 util/oslib-posix.c      | 33 +++++++++++++++++++++++++++++++++
 6 files changed, 65 insertions(+), 56 deletions(-)
 delete mode 100644 include/qemu/compatfd.h

diff --git a/cpus.c b/cpus.c
index 8200ac6..a628cde 100644
--- a/cpus.c
+++ b/cpus.c
@@ -51,10 +51,6 @@
 #include "hw/nmi.h"
 #include "sysemu/replay.h"
 
-#ifndef _WIN32
-#include "qemu/compatfd.h"
-#endif
-
 #ifdef CONFIG_LINUX
 
 #include <sys/prctl.h>
@@ -924,11 +920,9 @@ static void sigbus_reraise(void)
     abort();
 }
 
-static void sigbus_handler(int n, struct qemu_signalfd_siginfo *siginfo,
-                           void *ctx)
+static void sigbus_handler(int n, siginfo_t *siginfo, void *ctx)
 {
-    if (kvm_on_sigbus(siginfo->ssi_code,
-                      (void *)(intptr_t)siginfo->ssi_addr)) {
+    if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
         sigbus_reraise();
     }
 }
@@ -939,7 +933,7 @@ static void qemu_init_sigbus(void)
 
     memset(&action, 0, sizeof(action));
     action.sa_flags = SA_SIGINFO;
-    action.sa_sigaction = (void (*)(int, siginfo_t*, void*))sigbus_handler;
+    action.sa_sigaction = sigbus_handler;
     sigaction(SIGBUS, &action, NULL);
 
     prctl(PR_MCE_KILL, PR_MCE_KILL_SET, PR_MCE_KILL_EARLY, 0, 0);
diff --git a/include/qemu/compatfd.h b/include/qemu/compatfd.h
deleted file mode 100644
index aa12ee9..0000000
--- a/include/qemu/compatfd.h
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * signalfd/eventfd compatibility
- *
- * Copyright IBM, Corp. 2008
- *
- * Authors:
- *  Anthony Liguori   <aliguori@us.ibm.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2.  See
- * the COPYING file in the top-level directory.
- *
- */
-
-#ifndef QEMU_COMPATFD_H
-#define QEMU_COMPATFD_H
-
-
-struct qemu_signalfd_siginfo {
-    uint32_t ssi_signo;   /* Signal number */
-    int32_t  ssi_errno;   /* Error number (unused) */
-    int32_t  ssi_code;    /* Signal code */
-    uint32_t ssi_pid;     /* PID of sender */
-    uint32_t ssi_uid;     /* Real UID of sender */
-    int32_t  ssi_fd;      /* File descriptor (SIGIO) */
-    uint32_t ssi_tid;     /* Kernel timer ID (POSIX timers) */
-    uint32_t ssi_band;    /* Band event (SIGIO) */
-    uint32_t ssi_overrun; /* POSIX timer overrun count */
-    uint32_t ssi_trapno;  /* Trap number that caused signal */
-    int32_t  ssi_status;  /* Exit status or signal (SIGCHLD) */
-    int32_t  ssi_int;     /* Integer sent by sigqueue(2) */
-    uint64_t ssi_ptr;     /* Pointer sent by sigqueue(2) */
-    uint64_t ssi_utime;   /* User CPU time consumed (SIGCHLD) */
-    uint64_t ssi_stime;   /* System CPU time consumed (SIGCHLD) */
-    uint64_t ssi_addr;    /* Address that generated signal
-                             (for hardware-generated signals) */
-    uint8_t  pad[48];     /* Pad size to 128 bytes (allow for
-                             additional fields in the future) */
-};
-
-int qemu_signalfd(const sigset_t *mask);
-
-#endif
diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h
index 56c9e22..6932709 100644
--- a/include/qemu/osdep.h
+++ b/include/qemu/osdep.h
@@ -297,6 +297,34 @@ void qemu_anon_ram_free(void *ptr, size_t size);
 #  define QEMU_VMALLOC_ALIGN getpagesize()
 #endif
 
+#ifdef CONFIG_POSIX
+struct qemu_signalfd_siginfo {
+    uint32_t ssi_signo;   /* Signal number */
+    int32_t  ssi_errno;   /* Error number (unused) */
+    int32_t  ssi_code;    /* Signal code */
+    uint32_t ssi_pid;     /* PID of sender */
+    uint32_t ssi_uid;     /* Real UID of sender */
+    int32_t  ssi_fd;      /* File descriptor (SIGIO) */
+    uint32_t ssi_tid;     /* Kernel timer ID (POSIX timers) */
+    uint32_t ssi_band;    /* Band event (SIGIO) */
+    uint32_t ssi_overrun; /* POSIX timer overrun count */
+    uint32_t ssi_trapno;  /* Trap number that caused signal */
+    int32_t  ssi_status;  /* Exit status or signal (SIGCHLD) */
+    int32_t  ssi_int;     /* Integer sent by sigqueue(2) */
+    uint64_t ssi_ptr;     /* Pointer sent by sigqueue(2) */
+    uint64_t ssi_utime;   /* User CPU time consumed (SIGCHLD) */
+    uint64_t ssi_stime;   /* System CPU time consumed (SIGCHLD) */
+    uint64_t ssi_addr;    /* Address that generated signal
+                             (for hardware-generated signals) */
+    uint8_t  pad[48];     /* Pad size to 128 bytes (allow for
+                             additional fields in the future) */
+};
+
+int qemu_signalfd(const sigset_t *mask);
+void sigaction_invoke(struct sigaction *action,
+                      struct qemu_signalfd_siginfo *info);
+#endif
+
 int qemu_madvise(void *addr, size_t len, int advice);
 
 int qemu_open(const char *name, int flags, ...);
diff --git a/util/compatfd.c b/util/compatfd.c
index 9a43042..980bd33 100644
--- a/util/compatfd.c
+++ b/util/compatfd.c
@@ -15,7 +15,6 @@
 
 #include "qemu/osdep.h"
 #include "qemu-common.h"
-#include "qemu/compatfd.h"
 #include "qemu/thread.h"
 
 #include <sys/syscall.h>
diff --git a/util/main-loop.c b/util/main-loop.c
index ad10bca..ca7bb07 100644
--- a/util/main-loop.c
+++ b/util/main-loop.c
@@ -34,8 +34,6 @@
 
 #ifndef _WIN32
 
-#include "qemu/compatfd.h"
-
 /* If we have signalfd, we mask out the signals we want to handle and then
  * use signalfd to listen for them.  We rely on whatever the current signal
  * handler is to dispatch the signals when we receive them.
@@ -63,8 +61,7 @@ static void sigfd_handler(void *opaque)
 
         sigaction(info.ssi_signo, NULL, &action);
         if ((action.sa_flags & SA_SIGINFO) && action.sa_sigaction) {
-            action.sa_sigaction(info.ssi_signo,
-                                (siginfo_t *)&info, NULL);
+            sigaction_invoke(&action, &info);
         } else if (action.sa_handler) {
             action.sa_handler(info.ssi_signo);
         }
diff --git a/util/oslib-posix.c b/util/oslib-posix.c
index f631464..cd686aa 100644
--- a/util/oslib-posix.c
+++ b/util/oslib-posix.c
@@ -603,3 +603,36 @@ void qemu_free_stack(void *stack, size_t sz)
 
     munmap(stack, sz);
 }
+
+void sigaction_invoke(struct sigaction *action,
+                      struct qemu_signalfd_siginfo *info)
+{
+    siginfo_t si = { 0 };
+    si.si_signo = info->ssi_signo;
+    si.si_errno = info->ssi_errno;
+    si.si_code = info->ssi_code;
+
+    /* Convert the minimal set of fields defined by POSIX.
+     * Positive si_code values are reserved for kernel-generated
+     * signals, where the valid siginfo fields are determined by
+     * the signal number.  But according to POSIX, it is unspecified
+     * whether SI_USER and SI_QUEUE have values less than or equal to
+     * zero.
+     */
+    if (info->ssi_code == SI_USER || info->ssi_code == SI_QUEUE ||
+        info->ssi_code <= 0) {
+        /* SIGTERM, etc.  */
+        si.si_pid = info->ssi_pid;
+        si.si_uid = info->ssi_uid;
+    } else if (info->ssi_signo == SIGILL || info->ssi_signo == SIGFPE ||
+               info->ssi_signo == SIGSEGV || info->ssi_signo == SIGBUS) {
+        si.si_addr = (void *)(uintptr_t)info->ssi_addr;
+    } else if (info->ssi_signo == SIGCHLD) {
+        si.si_pid = info->ssi_pid;
+        si.si_status = info->ssi_status;
+        si.si_uid = info->ssi_uid;
+    } else if (info->ssi_signo == SIGIO) {
+        si.si_band = info->ssi_band;
+    }
+    action->sa_sigaction(info->ssi_signo, &si, NULL);
+}
-- 
2.9.3

[Qemu-devel] [PULL 10/17] KVM: do not use sigtimedwait to catch SIGBUS

[Paolo Bonzini]

Call kvm_on_sigbus_vcpu asynchronously from the VCPU thread.
Information for the SIGBUS can be stored in thread-local variables
and processed later in kvm_cpu_exec.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 cpus.c               | 31 +++++++++++++------------------
 include/sysemu/kvm.h |  5 ++++-
 kvm-all.c            | 37 ++++++++++++++++++++++++++++++++++++-
 target/arm/kvm.c     |  5 -----
 target/i386/kvm.c    |  5 ++---
 target/mips/kvm.c    |  6 ------
 target/ppc/kvm.c     |  5 -----
 target/s390x/kvm.c   |  5 -----
 8 files changed, 55 insertions(+), 44 deletions(-)

diff --git a/cpus.c b/cpus.c
index 399e271..56b1338 100644
--- a/cpus.c
+++ b/cpus.c
@@ -926,8 +926,16 @@ static void sigbus_handler(int n, siginfo_t *siginfo, void *ctx)
         sigbus_reraise();
     }
 
-    if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
-        sigbus_reraise();
+    if (current_cpu) {
+        /* Called asynchronously in VCPU thread.  */
+        if (kvm_on_sigbus_vcpu(current_cpu, siginfo->si_code, siginfo->si_addr)) {
+            sigbus_reraise();
+        }
+    } else {
+        /* Called synchronously (via signalfd) in main thread.  */
+        if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
+            sigbus_reraise();
+        }
     }
 }
 
@@ -958,8 +966,9 @@ static void qemu_kvm_init_cpu_signals(CPUState *cpu)
     sigaction(SIG_IPI, &sigact, NULL);
 
     pthread_sigmask(SIG_BLOCK, NULL, &set);
-    sigdelset(&set, SIG_IPI);
     sigdelset(&set, SIGBUS);
+    pthread_sigmask(SIG_SETMASK, &set, NULL);
+    sigdelset(&set, SIG_IPI);
     r = kvm_set_signal_mask(cpu, &set);
     if (r) {
         fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r));
@@ -977,7 +986,6 @@ static void qemu_kvm_eat_signals(CPUState *cpu)
 
     sigemptyset(&waitset);
     sigaddset(&waitset, SIG_IPI);
-    sigaddset(&waitset, SIGBUS);
 
     do {
         r = sigtimedwait(&waitset, &siginfo, &ts);
@@ -986,25 +994,12 @@ static void qemu_kvm_eat_signals(CPUState *cpu)
             exit(1);
         }
 
-        switch (r) {
-        case SIGBUS:
-            if (siginfo.si_code != BUS_MCEERR_AO && siginfo.si_code != BUS_MCEERR_AR) {
-                sigbus_reraise();
-            }
-            if (kvm_on_sigbus_vcpu(cpu, siginfo.si_code, siginfo.si_addr)) {
-                sigbus_reraise();
-            }
-            break;
-        default:
-            break;
-        }
-
         r = sigpending(&chkset);
         if (r == -1) {
             perror("sigpending");
             exit(1);
         }
-    } while (sigismember(&chkset, SIG_IPI) || sigismember(&chkset, SIGBUS));
+    } while (sigismember(&chkset, SIG_IPI));
 }
 #else /* !CONFIG_LINUX */
 static void qemu_init_sigbus(void)
diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
index 6ecb61c..a1b019d 100644
--- a/include/sysemu/kvm.h
+++ b/include/sysemu/kvm.h
@@ -357,7 +357,10 @@ bool kvm_vcpu_id_is_valid(int vcpu_id);
 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
+#ifdef TARGET_I386
+#define KVM_HAVE_MCE_INJECTION 1
+void kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
+#endif
 
 void kvm_arch_init_irq_routing(KVMState *s);
 
diff --git a/kvm-all.c b/kvm-all.c
index a433ad3..9e7c09f 100644
--- a/kvm-all.c
+++ b/kvm-all.c
@@ -1893,6 +1893,12 @@ void kvm_cpu_synchronize_post_init(CPUState *cpu)
     run_on_cpu(cpu, do_kvm_cpu_synchronize_post_init, RUN_ON_CPU_NULL);
 }
 
+#ifdef KVM_HAVE_MCE_INJECTION
+static __thread void *pending_sigbus_addr;
+static __thread int pending_sigbus_code;
+static __thread bool have_sigbus_pending;
+#endif
+
 int kvm_cpu_exec(CPUState *cpu)
 {
     struct kvm_run *run = cpu->kvm_run;
@@ -1930,6 +1936,18 @@ int kvm_cpu_exec(CPUState *cpu)
 
         attrs = kvm_arch_post_run(cpu, run);
 
+        if (unlikely(have_sigbus_pending)) {
+#ifdef KVM_HAVE_MCE_INJECTION
+            qemu_mutex_lock_iothread();
+            kvm_arch_on_sigbus_vcpu(cpu, pending_sigbus_code,
+                                    pending_sigbus_addr);
+            have_sigbus_pending = false;
+            qemu_mutex_unlock_iothread();
+#else
+            abort();
+#endif
+        }
+
         if (run_ret < 0) {
             if (run_ret == -EINTR || run_ret == -EAGAIN) {
                 DPRINTF("io window exit\n");
@@ -2392,13 +2410,27 @@ int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset)
     return r;
 }
 
+/* Called asynchronously in VCPU thread.  */
 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
 {
-    return kvm_arch_on_sigbus_vcpu(cpu, code, addr);
+#ifdef KVM_HAVE_MCE_INJECTION
+    if (have_sigbus_pending) {
+        return 1;
+    }
+    have_sigbus_pending = true;
+    pending_sigbus_addr = addr;
+    pending_sigbus_code = code;
+    atomic_set(&cpu->exit_request, 1);
+    return 0;
+#else
+    return 1;
+#endif
 }
 
+/* Called synchronously (via signalfd) in main thread.  */
 int kvm_on_sigbus(int code, void *addr)
 {
+#ifdef KVM_HAVE_MCE_INJECTION
     /* Action required MCE kills the process if SIGBUS is blocked.  Because
      * that's what happens in the I/O thread, where we handle MCE via signalfd,
      * we can only get action optional here.
@@ -2406,6 +2438,9 @@ int kvm_on_sigbus(int code, void *addr)
     assert(code != BUS_MCEERR_AR);
     kvm_arch_on_sigbus_vcpu(first_cpu, code, addr);
     return 0;
+#else
+    return 1;
+#endif
 }
 
 int kvm_create_device(KVMState *s, uint64_t type, bool test)
diff --git a/target/arm/kvm.c b/target/arm/kvm.c
index e5218f6..4555468 100644
--- a/target/arm/kvm.c
+++ b/target/arm/kvm.c
@@ -560,11 +560,6 @@ int kvm_arch_process_async_events(CPUState *cs)
     return 0;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cs, int code, void *addr)
-{
-    return 1;
-}
-
 /* The #ifdef protections are until 32bit headers are imported and can
  * be removed once both 32 and 64 bit reach feature parity.
  */
diff --git a/target/i386/kvm.c b/target/i386/kvm.c
index 2adf992..7698421 100644
--- a/target/i386/kvm.c
+++ b/target/i386/kvm.c
@@ -455,7 +455,7 @@ static void hardware_memory_error(void)
     exit(1);
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
+void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
 {
     X86CPU *cpu = X86_CPU(c);
     CPUX86State *env = &cpu->env;
@@ -475,7 +475,7 @@ int kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
             kvm_physical_memory_addr_from_host(c->kvm_state, addr, &paddr)) {
             kvm_hwpoison_page_add(ram_addr);
             kvm_mce_inject(cpu, paddr, code);
-            return 0;
+            return;
         }
 
         fprintf(stderr, "Hardware memory error for memory used by "
@@ -487,7 +487,6 @@ int kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
     }
 
     /* Hope we are lucky for AO MCE */
-    return 0;
 }
 
 static int kvm_inject_mce_oldstyle(X86CPU *cpu)
diff --git a/target/mips/kvm.c b/target/mips/kvm.c
index 3e686e7..0982e87 100644
--- a/target/mips/kvm.c
+++ b/target/mips/kvm.c
@@ -180,12 +180,6 @@ bool kvm_arch_stop_on_emulation_error(CPUState *cs)
     return true;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cs, int code, void *addr)
-{
-    DPRINTF("%s\n", __func__);
-    return 1;
-}
-
 void kvm_arch_init_irq_routing(KVMState *s)
 {
 }
diff --git a/target/ppc/kvm.c b/target/ppc/kvm.c
index bc011c6..f22ecab 100644
--- a/target/ppc/kvm.c
+++ b/target/ppc/kvm.c
@@ -2582,11 +2582,6 @@ bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
     return true;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
-{
-    return 1;
-}
-
 void kvm_arch_init_irq_routing(KVMState *s)
 {
 }
diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
index e7eea6d..ac47154 100644
--- a/target/s390x/kvm.c
+++ b/target/s390x/kvm.c
@@ -2140,11 +2140,6 @@ bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
     return true;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
-{
-    return 1;
-}
-
 void kvm_s390_io_interrupt(uint16_t subchannel_id,
                            uint16_t subchannel_nr, uint32_t io_int_parm,
                            uint32_t io_int_word)
-- 
2.9.3

Re: [Qemu-devel] [PULL v3 00/17] KVM and cpu-exec patches for 2.9 soft freeze

[Peter Maydell]

On 27 February 2017 at 16:49, Paolo Bonzini <pbonzini@redhat.com> wrote:
> The following changes since commit d992f2f1368ceb92e6bfd8efece174110f4236ff:
>
>   Merge remote-tracking branch 'remotes/artyom/tags/pull-sun4v-20170226' into staging (2017-02-26 22:40:23 +0000)
>
> are available in the git repository at:
>
>   git://github.com/bonzini/qemu.git tags/for-upstream
>
> for you to fetch changes up to 0819248d30877a9e0f063e7246e522a67548fc3e:
>
>   qmp-events: fix GUEST_PANICKED description formatting (2017-02-27 16:31:08 +0100)
>
> v2->v3: replace SIGPOLL with SIGIO
>
>         another missing #ifdef KVM_HAVE_MCE_INJECTION
>
> v1->v2: add missing #ifdef KVM_HAVE_MCE_INJECTION
>
>         rebase over MTTCG pull request.  -icount is currently broken but
>         I'd like to preserve bisectability, so this pull request brings
>         in the icount patches through a merge commit.
>
> ----------------------------------------------------------------
> * kernel header update (requested by David and Vijay)
> * GuestPanicInformation fixups (Anton)
> * record/replay icount fixes (Pavel)
> * cpu-exec cleanup, unification of icount_decr with tcg_exit_req (me)
> * KVM_CAP_IMMEDIATE_EXIT support (me)
> * vmxcap update (me)

Fails to build, OSX:

/home/petmay01/qemu/kvm-all.c: In function 'kvm_cpu_exec':
/home/petmay01/qemu/kvm-all.c:1995:33: error: 'have_sigbus_pending'
undeclared (first use in this function)
         if (unlikely(have_sigbus_pending)) {
                                 ^
/home/petmay01/qemu/kvm-all.c:1995:33: note: each undeclared
identifier is reported only once for each function it appears in

thanks
-- PMM

Re: [Qemu-devel] [PULL 06/17] cpus: remove ugly cast on sigbus_handler

[Peter Maydell]

On 27 February 2017 at 16:49, Paolo Bonzini <pbonzini@redhat.com> wrote:
> The cast is there because sigbus_handler is invoked via sigfd_handler.
> But it feels just wrong to use struct qemu_signalfd_siginfo in the
> prototype of a function that is passed to sigaction.
>
> Instead, do a simple-minded conversion of qemu_signalfd_siginfo to
> siginfo_t.
>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> +    } else if (info->ssi_signo == SIGIO) {
> +        si.si_band = info->ssi_band;
> +    }

This doesn't build on OpenBSD:

In file included from /usr/include/sys/signal.h:107:0,
                 from /usr/include/signal.h:38,
                 from /root/qemu/include/qemu/osdep.h:86,
                 from /root/qemu/util/oslib-posix.c:29:
/root/qemu/util/oslib-posix.c: In function 'sigaction_invoke':
/root/qemu/util/oslib-posix.c:713:12: error: 'union <anonymous>' has
no member named '_file'
         si.si_band = info->ssi_band;
            ^

I dunno how much we care.

thanks
-- PMM

Re: [Qemu-devel] [PULL 06/17] cpus: remove ugly cast on sigbus_handler

[Paolo Bonzini]

On 17/03/2017 16:08, Peter Maydell wrote:
> On 27 February 2017 at 16:49, Paolo Bonzini <pbonzini@redhat.com> wrote:
>> The cast is there because sigbus_handler is invoked via sigfd_handler.
>> But it feels just wrong to use struct qemu_signalfd_siginfo in the
>> prototype of a function that is passed to sigaction.
>>
>> Instead, do a simple-minded conversion of qemu_signalfd_siginfo to
>> siginfo_t.
>>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>> +    } else if (info->ssi_signo == SIGIO) {
>> +        si.si_band = info->ssi_band;
>> +    }
> 
> This doesn't build on OpenBSD:
> 
> In file included from /usr/include/sys/signal.h:107:0,
>                  from /usr/include/signal.h:38,
>                  from /root/qemu/include/qemu/osdep.h:86,
>                  from /root/qemu/util/oslib-posix.c:29:
> /root/qemu/util/oslib-posix.c: In function 'sigaction_invoke':
> /root/qemu/util/oslib-posix.c:713:12: error: 'union <anonymous>' has
> no member named '_file'
>          si.si_band = info->ssi_band;
>             ^
> 
> I dunno how much we care.

It's marked as obsolescent in POSIX, so it is sane to remove it.  I'll
send a patch.

Paolo

[Qemu-devel] [PULL 10/17] KVM: do not use sigtimedwait to catch SIGBUS

[Paolo Bonzini]

Call kvm_on_sigbus_vcpu asynchronously from the VCPU thread.
Information for the SIGBUS can be stored in thread-local variables
and processed later in kvm_cpu_exec.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 cpus.c               | 31 +++++++++++++------------------
 include/sysemu/kvm.h |  5 ++++-
 kvm-all.c            | 31 ++++++++++++++++++++++++++++++-
 target/arm/kvm.c     |  5 -----
 target/i386/kvm.c    |  5 ++---
 target/mips/kvm.c    |  6 ------
 target/ppc/kvm.c     |  5 -----
 target/s390x/kvm.c   |  5 -----
 8 files changed, 49 insertions(+), 44 deletions(-)

diff --git a/cpus.c b/cpus.c
index e50284f..a2bc916 100644
--- a/cpus.c
+++ b/cpus.c
@@ -796,8 +796,16 @@ static void sigbus_handler(int n, siginfo_t *siginfo, void *ctx)
         sigbus_reraise();
     }
 
-    if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
-        sigbus_reraise();
+    if (current_cpu) {
+        /* Called asynchronously in VCPU thread.  */
+        if (kvm_on_sigbus_vcpu(current_cpu, siginfo->si_code, siginfo->si_addr)) {
+            sigbus_reraise();
+        }
+    } else {
+        /* Called synchronously (via signalfd) in main thread.  */
+        if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
+            sigbus_reraise();
+        }
     }
 }
 
@@ -828,8 +836,9 @@ static void qemu_kvm_init_cpu_signals(CPUState *cpu)
     sigaction(SIG_IPI, &sigact, NULL);
 
     pthread_sigmask(SIG_BLOCK, NULL, &set);
-    sigdelset(&set, SIG_IPI);
     sigdelset(&set, SIGBUS);
+    pthread_sigmask(SIG_SETMASK, &set, NULL);
+    sigdelset(&set, SIG_IPI);
     r = kvm_set_signal_mask(cpu, &set);
     if (r) {
         fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r));
@@ -847,7 +856,6 @@ static void qemu_kvm_eat_signals(CPUState *cpu)
 
     sigemptyset(&waitset);
     sigaddset(&waitset, SIG_IPI);
-    sigaddset(&waitset, SIGBUS);
 
     do {
         r = sigtimedwait(&waitset, &siginfo, &ts);
@@ -856,25 +864,12 @@ static void qemu_kvm_eat_signals(CPUState *cpu)
             exit(1);
         }
 
-        switch (r) {
-        case SIGBUS:
-            if (siginfo.si_code != BUS_MCEERR_AO && siginfo.si_code != BUS_MCEERR_AR) {
-                sigbus_reraise();
-            }
-            if (kvm_on_sigbus_vcpu(cpu, siginfo.si_code, siginfo.si_addr)) {
-                sigbus_reraise();
-            }
-            break;
-        default:
-            break;
-        }
-
         r = sigpending(&chkset);
         if (r == -1) {
             perror("sigpending");
             exit(1);
         }
-    } while (sigismember(&chkset, SIG_IPI) || sigismember(&chkset, SIGBUS));
+    } while (sigismember(&chkset, SIG_IPI));
 }
 #else /* !CONFIG_LINUX */
 static void qemu_init_sigbus(void)
diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
index 6ecb61c..a1b019d 100644
--- a/include/sysemu/kvm.h
+++ b/include/sysemu/kvm.h
@@ -357,7 +357,10 @@ bool kvm_vcpu_id_is_valid(int vcpu_id);
 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
+#ifdef TARGET_I386
+#define KVM_HAVE_MCE_INJECTION 1
+void kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
+#endif
 
 void kvm_arch_init_irq_routing(KVMState *s);
 
diff --git a/kvm-all.c b/kvm-all.c
index a433ad3..d3b0487 100644
--- a/kvm-all.c
+++ b/kvm-all.c
@@ -1893,6 +1893,10 @@ void kvm_cpu_synchronize_post_init(CPUState *cpu)
     run_on_cpu(cpu, do_kvm_cpu_synchronize_post_init, RUN_ON_CPU_NULL);
 }
 
+static __thread void *pending_sigbus_addr;
+static __thread int pending_sigbus_code;
+static __thread bool have_sigbus_pending;
+
 int kvm_cpu_exec(CPUState *cpu)
 {
     struct kvm_run *run = cpu->kvm_run;
@@ -1930,6 +1934,14 @@ int kvm_cpu_exec(CPUState *cpu)
 
         attrs = kvm_arch_post_run(cpu, run);
 
+        if (unlikely(have_sigbus_pending)) {
+            qemu_mutex_lock_iothread();
+            kvm_arch_on_sigbus_vcpu(cpu, pending_sigbus_code,
+                                    pending_sigbus_addr);
+            have_sigbus_pending = false;
+            qemu_mutex_unlock_iothread();
+        }
+
         if (run_ret < 0) {
             if (run_ret == -EINTR || run_ret == -EAGAIN) {
                 DPRINTF("io window exit\n");
@@ -2392,13 +2404,27 @@ int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset)
     return r;
 }
 
+/* Called asynchronously in VCPU thread.  */
 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
 {
-    return kvm_arch_on_sigbus_vcpu(cpu, code, addr);
+#ifdef KVM_HAVE_MCE_INJECTION
+    if (have_sigbus_pending) {
+        return 1;
+    }
+    have_sigbus_pending = true;
+    pending_sigbus_addr = addr;
+    pending_sigbus_code = code;
+    atomic_set(&cpu->exit_request, 1);
+    return 0;
+#else
+    return 1;
+#endif
 }
 
+/* Called synchronously (via signalfd) in main thread.  */
 int kvm_on_sigbus(int code, void *addr)
 {
+#ifdef KVM_HAVE_MCE_INJECTION
     /* Action required MCE kills the process if SIGBUS is blocked.  Because
      * that's what happens in the I/O thread, where we handle MCE via signalfd,
      * we can only get action optional here.
@@ -2406,6 +2432,9 @@ int kvm_on_sigbus(int code, void *addr)
     assert(code != BUS_MCEERR_AR);
     kvm_arch_on_sigbus_vcpu(first_cpu, code, addr);
     return 0;
+#else
+    return 1;
+#endif
 }
 
 int kvm_create_device(KVMState *s, uint64_t type, bool test)
diff --git a/target/arm/kvm.c b/target/arm/kvm.c
index 6f2d590..c6d0b3a 100644
--- a/target/arm/kvm.c
+++ b/target/arm/kvm.c
@@ -559,11 +559,6 @@ int kvm_arch_process_async_events(CPUState *cs)
     return 0;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cs, int code, void *addr)
-{
-    return 1;
-}
-
 /* The #ifdef protections are until 32bit headers are imported and can
  * be removed once both 32 and 64 bit reach feature parity.
  */
diff --git a/target/i386/kvm.c b/target/i386/kvm.c
index 2adf992..7698421 100644
--- a/target/i386/kvm.c
+++ b/target/i386/kvm.c
@@ -455,7 +455,7 @@ static void hardware_memory_error(void)
     exit(1);
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
+void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
 {
     X86CPU *cpu = X86_CPU(c);
     CPUX86State *env = &cpu->env;
@@ -475,7 +475,7 @@ int kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
             kvm_physical_memory_addr_from_host(c->kvm_state, addr, &paddr)) {
             kvm_hwpoison_page_add(ram_addr);
             kvm_mce_inject(cpu, paddr, code);
-            return 0;
+            return;
         }
 
         fprintf(stderr, "Hardware memory error for memory used by "
@@ -487,7 +487,6 @@ int kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
     }
 
     /* Hope we are lucky for AO MCE */
-    return 0;
 }
 
 static int kvm_inject_mce_oldstyle(X86CPU *cpu)
diff --git a/target/mips/kvm.c b/target/mips/kvm.c
index 3e686e7..0982e87 100644
--- a/target/mips/kvm.c
+++ b/target/mips/kvm.c
@@ -180,12 +180,6 @@ bool kvm_arch_stop_on_emulation_error(CPUState *cs)
     return true;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cs, int code, void *addr)
-{
-    DPRINTF("%s\n", __func__);
-    return 1;
-}
-
 void kvm_arch_init_irq_routing(KVMState *s)
 {
 }
diff --git a/target/ppc/kvm.c b/target/ppc/kvm.c
index 80f2728..3125255 100644
--- a/target/ppc/kvm.c
+++ b/target/ppc/kvm.c
@@ -2558,11 +2558,6 @@ bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
     return true;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
-{
-    return 1;
-}
-
 void kvm_arch_init_irq_routing(KVMState *s)
 {
 }
diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
index 92cf6d4..8b9b211 100644
--- a/target/s390x/kvm.c
+++ b/target/s390x/kvm.c
@@ -2103,11 +2103,6 @@ bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
     return true;
 }
 
-int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
-{
-    return 1;
-}
-
 void kvm_s390_io_interrupt(uint16_t subchannel_id,
                            uint16_t subchannel_nr, uint32_t io_int_parm,
                            uint32_t io_int_word)
-- 
1.8.3.1