[PATCH 00/10] qemu-kvm: Cleanup and switch to upstream - Part III

[PATCH 00/10] qemu-kvm: Cleanup and switch to upstream - Part III

[Jan Kiszka]

This is a rather short round as the next and final one cannot be split
up very well.

We start with three code cleanup patches, then work towards using
upstream kvm_cpu_exec, and finally rework the core's PIO access
management used for device assignment.

Please review/merge.

CC: Alexander Graf <agraf@suse.de>

Jan Kiszka (10):
  qemu-kvm: Remove obsolete inclusions
  qemu-kvm: Remove remaining s390 fragments
  qemu-kvm: ppc: Drop diff to upstream
  qemu-kvm: Refactor vm exit in kvm_run
  qemu-kvm: Use thread_kicked instead of KVMCPUState::signalled
  qemu-kvm: Activate iothread version of qemu_cpu_kick_self
  qemu-kvm: Refactor exit_request processing in kvm_run
  qemu-kvm: Use upstream kvm_arch_process_async_events
  qemu-kvm: Use upstream kvm_cpu_exec
  qemu-kvm: Rework ioport access management

 cpu-defs.h             |    1 -
 cpu-exec.c             |    2 -
 cpus.c                 |    6 +
 exec.c                 |    1 -
 gdbstub.c              |    1 -
 hw/acpi.c              |    3 -
 hw/device-assignment.c |   20 +--
 hw/ppc440.c            |    1 -
 hw/ppc440_bamboo.c     |    1 -
 hw/ppce500_mpc8544ds.c |    1 -
 hw/virtio-balloon.c    |    1 -
 kvm-all.c              |    3 -
 qemu-kvm-x86.c         |  158 +--------------------
 qemu-kvm.c             |  359 ++++++++++++-----------------------------------
 qemu-kvm.h             |   77 +----------
 target-i386/helper.c   |    1 -
 target-i386/kvm.c      |    4 -
 target-ppc/cpu.h       |   10 +--
 target-ppc/fake-exec.c |  104 --------------
 target-ppc/helper.c    |    1 -
 target-ppc/machine.c   |    1 -
 21 files changed, 121 insertions(+), 635 deletions(-)
 delete mode 100644 target-ppc/fake-exec.c

[PATCH 01/10] qemu-kvm: Remove obsolete inclusions

[Jan Kiszka]

Reduces the diff to upstream.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 cpu-exec.c             |    2 --
 exec.c                 |    1 -
 gdbstub.c              |    1 -
 hw/acpi.c              |    3 ---
 hw/ppc440.c            |    1 -
 hw/ppc440_bamboo.c     |    1 -
 hw/ppce500_mpc8544ds.c |    1 -
 hw/virtio-balloon.c    |    1 -
 target-i386/helper.c   |    1 -
 9 files changed, 0 insertions(+), 12 deletions(-)

diff --git a/cpu-exec.c b/cpu-exec.c
index 583d7d6..6ec519d 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -41,8 +41,6 @@
 #endif
 #endif
 
-#include "qemu-kvm.h"
-
 #if defined(__sparc__) && !defined(CONFIG_SOLARIS)
 // Work around ugly bugs in glibc that mangle global register contents
 #undef env
diff --git a/exec.c b/exec.c
index 1dc970f..9a493d0 100644
--- a/exec.c
+++ b/exec.c
@@ -32,7 +32,6 @@
 #if !defined(TARGET_IA64)
 #include "tcg.h"
 #endif
-#include "qemu-kvm.h"
 
 #include "hw/hw.h"
 #include "hw/qdev.h"
diff --git a/gdbstub.c b/gdbstub.c
index df93691..0838948 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -34,7 +34,6 @@
 #include "sysemu.h"
 #include "gdbstub.h"
 #endif
-#include "qemu-kvm.h"
 
 #define MAX_PACKET_LENGTH 4096
 
diff --git a/hw/acpi.c b/hw/acpi.c
index 3d78d5f..ad40fb4 100644
--- a/hw/acpi.c
+++ b/hw/acpi.c
@@ -19,9 +19,6 @@
 #include "hw.h"
 #include "pc.h"
 #include "acpi.h"
-#include "kvm.h"
-#include "qemu-kvm.h"
-#include "string.h"
 
 struct acpi_table_header
 {
diff --git a/hw/ppc440.c b/hw/ppc440.c
index 7d651e9..1ed001a 100644
--- a/hw/ppc440.c
+++ b/hw/ppc440.c
@@ -20,7 +20,6 @@
 #include "ppc405.h"
 #include "sysemu.h"
 #include "kvm.h"
-#include "qemu-kvm.h"
 
 #define PPC440EP_PCI_CONFIG     0xeec00000
 #define PPC440EP_PCI_INTACK     0xeed00000
diff --git a/hw/ppc440_bamboo.c b/hw/ppc440_bamboo.c
index 6627cd8..20b8629 100644
--- a/hw/ppc440_bamboo.c
+++ b/hw/ppc440_bamboo.c
@@ -23,7 +23,6 @@
 #include "device_tree.h"
 #include "loader.h"
 #include "elf.h"
-#include "qemu-kvm.h"
 
 #define BINARY_DEVICE_TREE_FILE "bamboo.dtb"
 
diff --git a/hw/ppce500_mpc8544ds.c b/hw/ppce500_mpc8544ds.c
index 13153dd..17b0165 100644
--- a/hw/ppce500_mpc8544ds.c
+++ b/hw/ppce500_mpc8544ds.c
@@ -32,7 +32,6 @@
 #include "loader.h"
 #include "elf.h"
 #include "sysbus.h"
-#include "qemu-kvm.h"
 
 #define BINARY_DEVICE_TREE_FILE    "mpc8544ds.dtb"
 #define UIMAGE_LOAD_BASE           0
diff --git a/hw/virtio-balloon.c b/hw/virtio-balloon.c
index 447b337..70a8710 100644
--- a/hw/virtio-balloon.c
+++ b/hw/virtio-balloon.c
@@ -20,7 +20,6 @@
 #include "balloon.h"
 #include "virtio-balloon.h"
 #include "kvm.h"
-#include "qemu-kvm.h"
 #include "qlist.h"
 #include "qint.h"
 #include "qstring.h"
diff --git a/target-i386/helper.c b/target-i386/helper.c
index 4ab5241..2315c84 100644
--- a/target-i386/helper.c
+++ b/target-i386/helper.c
@@ -32,7 +32,6 @@
 #include "monitor.h"
 #endif
 
-#include "qemu-kvm.h"
 #ifndef OBSOLETE_KVM_IMPL
 #define run_on_cpu on_vcpu
 #endif /* !OBSOLETE_KVM_IMPL */
-- 
1.7.1

[PATCH 02/10] qemu-kvm: Remove remaining s390 fragments

[Jan Kiszka]

qemu-kvm is not used with s390, only upstream works theses days. So
remove any conditionals related to this arch.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 qemu-kvm.c |   12 ------------
 qemu-kvm.h |   10 +---------
 2 files changed, 1 insertions(+), 21 deletions(-)

diff --git a/qemu-kvm.c b/qemu-kvm.c
index c9ff5fe..19ede6e 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -307,11 +307,9 @@ static int kvm_run(CPUState *env)
         env->kvm_vcpu_dirty = 0;
     }
     push_nmi();
-#if !defined(__s390__)
     if (!kvm_state->irqchip_in_kernel) {
         run->request_interrupt_window = kvm_arch_try_push_interrupts(env);
     }
-#endif
 
     r = pre_kvm_run(env);
     if (r) {
@@ -334,11 +332,9 @@ static int kvm_run(CPUState *env)
 
     kvm_flush_coalesced_mmio_buffer();
 
-#if !defined(__s390__)
     if (r == -1) {
         return 1;
     }
-#endif
     if (1) {
         switch (run->exit_reason) {
         case KVM_EXIT_UNKNOWN:
@@ -372,14 +368,6 @@ static int kvm_run(CPUState *env)
         case KVM_EXIT_SHUTDOWN:
             r = handle_shutdown(env);
             break;
-#if defined(__s390__)
-        case KVM_EXIT_S390_SIEIC:
-            r = kvm_s390_handle_intercept(kvm, env, run);
-            break;
-        case KVM_EXIT_S390_RESET:
-            r = kvm_s390_handle_reset(kvm, env, run);
-            break;
-#endif
 	case KVM_EXIT_INTERNAL_ERROR:
             r = kvm_handle_internal_error(env, run);
 	    break;
diff --git a/qemu-kvm.h b/qemu-kvm.h
index b552e1e..8ff773a 100644
--- a/qemu-kvm.h
+++ b/qemu-kvm.h
@@ -15,10 +15,6 @@
 
 #ifdef CONFIG_KVM
 
-#if defined(__s390__)
-#include <asm/ptrace.h>
-#endif
-
 #include <stdint.h>
 
 #ifndef __user
@@ -31,11 +27,7 @@
 
 /* FIXME: share this number with kvm */
 /* FIXME: or dynamically alloc/realloc regions */
-#ifdef __s390__
-#define KVM_MAX_NUM_MEM_REGIONS 1u
-#define MAX_VCPUS 64
-#define LIBKVM_S390_ORIGIN (0UL)
-#elif defined(__ia64__)
+#if defined(__ia64__)
 #define KVM_MAX_NUM_MEM_REGIONS 32u
 #define MAX_VCPUS 256
 #else
-- 
1.7.1

[PATCH 03/10] qemu-kvm: ppc: Drop diff to upstream

[Jan Kiszka]

qemu-kvm is x86-only today. Remove the remaining delta of the former ppc
support to avoid needless conflicts when sync'ing with upstream.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
CC: Alexander Graf <agraf@suse.de>
---
 target-ppc/cpu.h       |   10 +----
 target-ppc/fake-exec.c |  104 ------------------------------------------------
 target-ppc/helper.c    |    1 -
 target-ppc/machine.c   |    1 -
 4 files changed, 1 insertions(+), 115 deletions(-)
 delete mode 100644 target-ppc/fake-exec.c

diff --git a/target-ppc/cpu.h b/target-ppc/cpu.h
index 133df95..7a6a7df 100644
--- a/target-ppc/cpu.h
+++ b/target-ppc/cpu.h
@@ -52,10 +52,9 @@
 #if defined(TARGET_PPCEMB)
 /* Specific definitions for PowerPC embedded */
 /* BookE have 36 bits physical address space */
-#if defined(CONFIG_USER_ONLY) || defined(USE_KVM)
+#if defined(CONFIG_USER_ONLY)
 /* It looks like a lot of Linux programs assume page size
  * is 4kB long. This is evil, but we have to deal with it...
- * Also kvm for embedded powerpc needs (atm) 4kB aligned pages
  */
 #define TARGET_PAGE_BITS 12
 #else /* defined(CONFIG_USER_ONLY) */
@@ -1927,13 +1926,6 @@ static inline void cpu_set_tls(CPUState *env, target_ulong newtls)
 #endif
 }
 
-/* hidden flags (hflags) - used internally by qemu to represent additional
- * cpu states.
- */
-#define HF_HALTED_SHIFT 1
-
-#define HF_HALTED_MASK 1<<HF_HALTED_SHIFT
-
 #if !defined(CONFIG_USER_ONLY)
 static inline int booke206_tlbe_id(CPUState *env, ppcemb_tlb_t *tlbe)
 {
diff --git a/target-ppc/fake-exec.c b/target-ppc/fake-exec.c
deleted file mode 100644
index 259e06d..0000000
--- a/target-ppc/fake-exec.c
+++ /dev/null
@@ -1,104 +0,0 @@
-/*
- * fake-exec.c
- *
- * This is a file for stub functions so that compilation is possible
- * when TCG CPU emulation is disabled during compilation.
- *
- * Copyright 2007 IBM Corporation.
- * Added by & Authors:
- * 	Jerone Young <jyoung5@us.ibm.com>
- * This work is licensed under the GNU GPL licence version 2 or later.
- *
- */
-
-#include <stdarg.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <inttypes.h>
-
-#include "cpu.h"
-#include "exec-all.h"
-
-
-struct ppc_def_t {
-    const unsigned char *name;
-    uint32_t pvr;
-    uint32_t svr;
-    uint64_t insns_flags;
-    uint64_t msr_mask;
-    powerpc_mmu_t   mmu_model;
-    powerpc_excp_t  excp_model;
-    powerpc_input_t bus_model;
-    uint32_t flags;
-    int bfd_mach;
-    void (*init_proc)(CPUPPCState *env);
-    int  (*check_pow)(CPUPPCState *env);
-};
-
-int code_copy_enabled = 0;
-
-void cpu_dump_state (CPUState *env, FILE *f,
-                     int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
-                     int flags)
-{
-}
-
-void ppc_cpu_list (FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...))
-{
-}
-
-void cpu_dump_statistics (CPUState *env, FILE*f,
-                          int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
-                          int flags)
-{
-}
-
-unsigned long code_gen_max_block_size(void)
-{
-    return 32;
-}
-
-void cpu_gen_init(void)
-{
-}
-
-int cpu_restore_state(TranslationBlock *tb,
-                      CPUState *env, unsigned long searched_pc,
-                      void *puc)
-
-{
-    return 0;
-}
-
-int cpu_ppc_gen_code(CPUState *env, TranslationBlock *tb, int *gen_code_size_ptr)
-{
-    return 0;
-}
-
-void init_proc_ppc440ep_kvm(CPUPPCState *env)
-{
-    ppc40x_irq_init(env);
-}
-
-static ppc_def_t ppc440ep_kvm = {
-    .name = "440EP KVM",
-    .mmu_model = POWERPC_MMU_SOFT_4xx, /*XXX needed for GDB stub */
-    .init_proc = init_proc_ppc440ep_kvm,
-};
-
-const ppc_def_t *cpu_ppc_find_by_name (const unsigned char *name)
-{
-    return &ppc440ep_kvm;
-}
-
-int cpu_ppc_register_internal (CPUPPCState *env, const ppc_def_t *def)
-{
-    env->mmu_model = def->mmu_model;
-    (*def->init_proc)(env);
-    return 0;
-}
-
-void flush_icache_range(unsigned long start, unsigned long stop)
-{
-}
diff --git a/target-ppc/helper.c b/target-ppc/helper.c
index 95f81ec..4238be6 100644
--- a/target-ppc/helper.c
+++ b/target-ppc/helper.c
@@ -28,7 +28,6 @@
 #include "helper_regs.h"
 #include "qemu-common.h"
 #include "kvm.h"
-#include "qemu-kvm.h"
 
 //#define DEBUG_MMU
 //#define DEBUG_BATS
diff --git a/target-ppc/machine.c b/target-ppc/machine.c
index 4b98113..0c1986e 100644
--- a/target-ppc/machine.c
+++ b/target-ppc/machine.c
@@ -1,7 +1,6 @@
 #include "hw/hw.h"
 #include "hw/boards.h"
 #include "kvm.h"
-#include "qemu-kvm.h"
 
 void cpu_save(QEMUFile *f, void *opaque)
 {
-- 
1.7.1

[PATCH 04/10] qemu-kvm: Refactor vm exit in kvm_run

[Jan Kiszka]

Use kvm_vcpu_ioctl to simplify the vm exit code a bit. This increases
the similarity to upstream.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 qemu-kvm.c |   17 +++++++----------
 1 files changed, 7 insertions(+), 10 deletions(-)

diff --git a/qemu-kvm.c b/qemu-kvm.c
index 19ede6e..e5b8e42 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -299,7 +299,6 @@ static int kvm_run(CPUState *env)
 {
     int r;
     struct kvm_run *run = env->kvm_run;
-    int fd = env->kvm_fd;
 
   again:
     if (env->kvm_vcpu_dirty) {
@@ -319,22 +318,20 @@ static int kvm_run(CPUState *env)
         env->exit_request = 0;
         pthread_kill(env->thread->thread, SIG_IPI);
     }
-    r = ioctl(fd, KVM_RUN, 0);
-
-    if (r == -1 && errno != EINTR && errno != EAGAIN) {
-        r = -errno;
-        post_kvm_run(env);
-        fprintf(stderr, "kvm_run: %s\n", strerror(-r));
-        return r;
-    }
+    r = kvm_vcpu_ioctl(env, KVM_RUN, 0);
 
     post_kvm_run(env);
 
     kvm_flush_coalesced_mmio_buffer();
 
-    if (r == -1) {
+    if (r == -EINTR || r == -EAGAIN) {
         return 1;
     }
+    if (r < 0) {
+        fprintf(stderr, "kvm_run: %s\n", strerror(-r));
+        return r;
+    }
+
     if (1) {
         switch (run->exit_reason) {
         case KVM_EXIT_UNKNOWN:
-- 
1.7.1

[PATCH 05/10] qemu-kvm: Use thread_kicked instead of KVMCPUState::signalled

[Jan Kiszka]

Another step forward to use signaling services from cpus.c.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 cpu-defs.h |    1 -
 qemu-kvm.c |    6 +++---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/cpu-defs.h b/cpu-defs.h
index 624fd1a..5a0f11d 100644
--- a/cpu-defs.h
+++ b/cpu-defs.h
@@ -158,7 +158,6 @@ typedef struct CPUWatchpoint {
 struct qemu_work_item;
 
 struct KVMCPUState {
-    int signalled;
     struct qemu_work_item *queued_work_first, *queued_work_last;
 };
 
diff --git a/qemu-kvm.c b/qemu-kvm.c
index e5b8e42..843c973 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -840,12 +840,12 @@ void kvm_update_interrupt_request(CPUState *env)
          * Testing for created here is really redundant
          */
         if (current_env && current_env->created &&
-            env != current_env && !env->kvm_cpu_state.signalled) {
+            env != current_env && !env->thread_kicked) {
             signal = 1;
         }
 
         if (signal) {
-            env->kvm_cpu_state.signalled = 1;
+            env->thread_kicked = true;
             if (env->thread) {
                 pthread_kill(env->thread->thread, SIG_IPI);
             }
@@ -941,7 +941,7 @@ static void kvm_main_loop_wait(CPUState *env, int timeout)
         pthread_cond_signal(&qemu_pause_cond);
     }
 
-    env->kvm_cpu_state.signalled = 0;
+    env->thread_kicked = false;
 }
 
 static int all_threads_paused(void)
-- 
1.7.1

[PATCH 06/10] qemu-kvm: Activate iothread version of qemu_cpu_kick_self

[Jan Kiszka]

This is a temporary hack: We will soon use this service for kvm which is
running in iothread mode - but with CONFIG_IOTHREAD disabled.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 cpus.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/cpus.c b/cpus.c
index 5d1c396..b64c29f 100644
--- a/cpus.c
+++ b/cpus.c
@@ -590,6 +590,7 @@ void qemu_cpu_kick(void *env)
 {
 }
 
+#ifdef UNUSED_IMPL
 void qemu_cpu_kick_self(void)
 {
 #ifndef _WIN32
@@ -600,6 +601,7 @@ void qemu_cpu_kick_self(void)
     abort();
 #endif
 }
+#endif
 
 void qemu_notify_event(void)
 {
@@ -848,6 +850,7 @@ static void *qemu_tcg_cpu_thread_fn(void *arg)
     return NULL;
 }
 
+#endif
 static void qemu_cpu_kick_thread(CPUState *env)
 {
 #ifndef _WIN32
@@ -866,6 +869,7 @@ static void qemu_cpu_kick_thread(CPUState *env)
     }
 #endif
 }
+#ifdef CONFIG_IOTHREAD
 
 void qemu_cpu_kick(void *_env)
 {
@@ -878,6 +882,7 @@ void qemu_cpu_kick(void *_env)
     }
 }
 
+#endif
 void qemu_cpu_kick_self(void)
 {
 #ifndef _WIN32
@@ -891,6 +896,7 @@ void qemu_cpu_kick_self(void)
     abort();
 #endif
 }
+#ifdef CONFIG_IOTHREAD
 
 int qemu_cpu_is_self(void *_env)
 {
-- 
1.7.1

[PATCH 07/10] qemu-kvm: Refactor exit_request processing in kvm_run

[Jan Kiszka]

Move closer to upstream by reusing its self-signaling service in
kvm_run, maintain cpu_single_env the same way, push this under
qemu_mutex, and clear exit_request unconditionally when leaving kvm_run.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 qemu-kvm.c |   25 ++++++++++++++-----------
 1 files changed, 14 insertions(+), 11 deletions(-)

diff --git a/qemu-kvm.c b/qemu-kvm.c
index 843c973..24c667c 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -278,16 +278,21 @@ static inline void push_nmi(void)
 static void post_kvm_run(CPUState *env)
 {
     pthread_mutex_lock(&qemu_mutex);
-    kvm_arch_post_run(env, env->kvm_run);
     cpu_single_env = env;
+
+    kvm_arch_post_run(env, env->kvm_run);
 }
 
-static int pre_kvm_run(CPUState *env)
+static void pre_kvm_run(CPUState *env)
 {
     kvm_arch_pre_run(env, env->kvm_run);
 
+    if (env->exit_request) {
+        qemu_cpu_kick_self();
+    }
+
+    cpu_single_env = NULL;
     pthread_mutex_unlock(&qemu_mutex);
-    return 0;
 }
 
 int kvm_is_ready_for_interrupt_injection(CPUState *env)
@@ -300,6 +305,8 @@ static int kvm_run(CPUState *env)
     int r;
     struct kvm_run *run = env->kvm_run;
 
+    cpu_single_env = env;
+
   again:
     if (env->kvm_vcpu_dirty) {
         kvm_arch_put_registers(env, KVM_PUT_RUNTIME_STATE);
@@ -310,14 +317,8 @@ static int kvm_run(CPUState *env)
         run->request_interrupt_window = kvm_arch_try_push_interrupts(env);
     }
 
-    r = pre_kvm_run(env);
-    if (r) {
-        return r;
-    }
-    if (env->exit_request) {
-        env->exit_request = 0;
-        pthread_kill(env->thread->thread, SIG_IPI);
-    }
+    pre_kvm_run(env);
+
     r = kvm_vcpu_ioctl(env, KVM_RUN, 0);
 
     post_kvm_run(env);
@@ -325,6 +326,7 @@ static int kvm_run(CPUState *env)
     kvm_flush_coalesced_mmio_buffer();
 
     if (r == -EINTR || r == -EAGAIN) {
+        env->exit_request = 0;
         return 1;
     }
     if (r < 0) {
@@ -384,6 +386,7 @@ static int kvm_run(CPUState *env)
     if (!r) {
         goto again;
     }
+    env->exit_request = 0;
     return r;
 }
 
-- 
1.7.1

[PATCH 08/10] qemu-kvm: Use upstream kvm_arch_process_async_events

[Jan Kiszka]

Upstream's kvm_arch_process_async_events is now identical to qemu-kvm's
combination of kvm_arch_process_async_events and process_irqchip_events.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 qemu-kvm-x86.c    |   49 -------------------------------------------------
 qemu-kvm.c        |   13 +------------
 qemu-kvm.h        |    2 --
 target-i386/kvm.c |    2 --
 4 files changed, 1 insertions(+), 65 deletions(-)

diff --git a/qemu-kvm-x86.c b/qemu-kvm-x86.c
index b009251..18f7c3a 100644
--- a/qemu-kvm-x86.c
+++ b/qemu-kvm-x86.c
@@ -225,16 +225,6 @@ void kvm_arch_pre_run(CPUState *env, struct kvm_run *run)
     }
 }
 
-int kvm_arch_has_work(CPUState *env)
-{
-    if (((env->interrupt_request & CPU_INTERRUPT_HARD) &&
-         (env->eflags & IF_MASK)) ||
-        (env->interrupt_request & CPU_INTERRUPT_NMI)) {
-        return 1;
-    }
-    return 0;
-}
-
 int kvm_arch_try_push_interrupts(void *opaque)
 {
     CPUState *env = cpu_single_env;
@@ -331,42 +321,3 @@ int kvm_arch_init_irq_routing(void)
 
     return 0;
 }
-
-void kvm_arch_process_irqchip_events(CPUState *env)
-{
-    if (env->interrupt_request & CPU_INTERRUPT_INIT) {
-        kvm_cpu_synchronize_state(env);
-        do_cpu_init(env);
-    }
-    if (env->interrupt_request & CPU_INTERRUPT_SIPI) {
-        kvm_cpu_synchronize_state(env);
-        do_cpu_sipi(env);
-    }
-}
-
-int kvm_arch_process_async_events(CPUState *env)
-{
-    if (env->interrupt_request & CPU_INTERRUPT_MCE) {
-        /* We must not raise CPU_INTERRUPT_MCE if it's not supported. */
-        assert(env->mcg_cap);
-
-        env->interrupt_request &= ~CPU_INTERRUPT_MCE;
-
-        kvm_cpu_synchronize_state(env);
-
-        if (env->exception_injected == EXCP08_DBLE) {
-            /* this means triple fault */
-            qemu_system_reset_request();
-            env->exit_request = 1;
-            return 0;
-        }
-        env->exception_injected = EXCP12_MCHK;
-        env->has_error_code = 0;
-
-        env->halted = 0;
-        if (kvm_irqchip_in_kernel() && env->mp_state == KVM_MP_STATE_HALTED) {
-            env->mp_state = KVM_MP_STATE_RUNNABLE;
-        }
-    }
-    return 0;
-}
diff --git a/qemu-kvm.c b/qemu-kvm.c
index 24c667c..41c4219 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -1032,23 +1032,12 @@ static void qemu_kvm_system_reset(void)
     resume_all_threads();
 }
 
-static void process_irqchip_events(CPUState *env)
-{
-    kvm_arch_process_irqchip_events(env);
-    if (kvm_arch_has_work(env))
-        env->halted = 0;
-}
-
 static int kvm_main_loop_cpu(CPUState *env)
 {
     while (1) {
         int run_cpu = !kvm_cpu_is_stopped(env);
         if (run_cpu) {
-            kvm_arch_process_async_events(env);
-            if (!kvm_irqchip_in_kernel()) {
-                process_irqchip_events(env);
-                run_cpu = !env->halted;
-            }
+            run_cpu = !kvm_arch_process_async_events(env);
         }
         if (run_cpu) {
             kvm_cpu_exec(env);
diff --git a/qemu-kvm.h b/qemu-kvm.h
index 8ff773a..5621cfa 100644
--- a/qemu-kvm.h
+++ b/qemu-kvm.h
@@ -314,8 +314,6 @@ void kvm_hpet_disable_kpit(void);
 void on_vcpu(CPUState *env, void (*func)(void *data), void *data);
 void kvm_update_interrupt_request(CPUState *env);
 
-int kvm_arch_has_work(CPUState *env);
-void kvm_arch_process_irqchip_events(CPUState *env);
 int kvm_arch_try_push_interrupts(void *opaque);
 void kvm_arch_push_nmi(void);
 int kvm_set_boot_cpu_id(KVMState *s, uint32_t id);
diff --git a/target-i386/kvm.c b/target-i386/kvm.c
index a4365bf..e98eccf 100644
--- a/target-i386/kvm.c
+++ b/target-i386/kvm.c
@@ -1636,7 +1636,6 @@ void kvm_arch_post_run(CPUState *env, struct kvm_run *run)
     cpu_set_apic_base(env->apic_state, run->apic_base);
 }
 
-#ifdef OBSOLETE_KVM_IMPL
 int kvm_arch_process_async_events(CPUState *env)
 {
     if (env->interrupt_request & CPU_INTERRUPT_MCE) {
@@ -1682,7 +1681,6 @@ int kvm_arch_process_async_events(CPUState *env)
 
     return env->halted;
 }
-#endif
 
 static int kvm_handle_halt(CPUState *env)
 {
-- 
1.7.1

[PATCH 09/10] qemu-kvm: Use upstream kvm_cpu_exec

[Jan Kiszka]

Upstream's and qemu-kvm's kvm_cpu_exec are not logically equivalent so
that we can safely switch the implementations.

A bit refactoring of kvm_main_loop_cpu is required as upstream's cpu
loop already contains the asynchronous event processing which ran
outside so far.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 kvm-all.c         |    3 -
 qemu-kvm-x86.c    |  101 +-----------------------
 qemu-kvm.c        |  225 ++++-------------------------------------------------
 qemu-kvm.h        |   44 +----------
 target-i386/kvm.c |    2 -
 5 files changed, 17 insertions(+), 358 deletions(-)

diff --git a/kvm-all.c b/kvm-all.c
index 5ac177f..88d0785 100644
--- a/kvm-all.c
+++ b/kvm-all.c
@@ -958,8 +958,6 @@ void kvm_cpu_synchronize_post_init(CPUState *env)
     env->kvm_vcpu_dirty = 0;
 }
 
-#ifdef OBSOLETE_KVM_IMPL
-
 int kvm_cpu_exec(CPUState *env)
 {
     struct kvm_run *run = env->kvm_run;
@@ -1065,7 +1063,6 @@ int kvm_cpu_exec(CPUState *env)
     return ret;
 }
 
-#endif
 int kvm_ioctl(KVMState *s, int type, ...)
 {
     int ret;
diff --git a/qemu-kvm-x86.c b/qemu-kvm-x86.c
index 18f7c3a..89bb692 100644
--- a/qemu-kvm-x86.c
+++ b/qemu-kvm-x86.c
@@ -49,13 +49,13 @@ static int kvm_create_pit(KVMState *s)
 
 #ifdef KVM_EXIT_TPR_ACCESS
 
-static int kvm_handle_tpr_access(CPUState *env)
+int kvm_handle_tpr_access(CPUState *env)
 {
     struct kvm_run *run = env->kvm_run;
     kvm_tpr_access_report(env,
                           run->tpr_access.rip,
                           run->tpr_access.is_write);
-    return 0;
+    return 1;
 }
 
 
@@ -70,41 +70,6 @@ int kvm_enable_vapic(CPUState *env, uint64_t vapic)
 
 #endif
 
-extern CPUState *kvm_debug_cpu_requested;
-
-int kvm_arch_run(CPUState *env)
-{
-    int r = 0;
-    struct kvm_run *run = env->kvm_run;
-
-    switch (run->exit_reason) {
-#ifdef KVM_EXIT_SET_TPR
-    case KVM_EXIT_SET_TPR:
-        break;
-#endif
-#ifdef KVM_EXIT_TPR_ACCESS
-    case KVM_EXIT_TPR_ACCESS:
-        r = kvm_handle_tpr_access(env);
-        break;
-#endif
-#ifdef KVM_CAP_SET_GUEST_DEBUG
-    case KVM_EXIT_DEBUG:
-        DPRINTF("kvm_exit_debug\n");
-        r = kvm_handle_debug(&run->debug.arch);
-        if (r == EXCP_DEBUG) {
-            kvm_debug_cpu_requested = env;
-            env->stopped = 1;
-        }
-        break;
-#endif /* KVM_CAP_SET_GUEST_DEBUG */
-    default:
-        r = -1;
-        break;
-    }
-
-    return r;
-}
-
 #ifdef KVM_CAP_IRQCHIP
 
 int kvm_get_lapic(CPUState *env, struct kvm_lapic_state *s)
@@ -178,11 +143,6 @@ int kvm_set_pit2(KVMState *s, struct kvm_pit_state2 *ps2)
 #endif
 #endif
 
-static void kvm_set_cr8(CPUState *env, uint64_t cr8)
-{
-    env->kvm_run->cr8 = cr8;
-}
-
 #ifdef KVM_CAP_VAPIC
 static int kvm_enable_tpr_access_reporting(CPUState *env)
 {
@@ -207,63 +167,6 @@ static int _kvm_arch_init_vcpu(CPUState *env)
     return 0;
 }
 
-int kvm_arch_halt(CPUState *env)
-{
-
-    if (!((env->interrupt_request & CPU_INTERRUPT_HARD) &&
-          (env->eflags & IF_MASK)) &&
-        !(env->interrupt_request & CPU_INTERRUPT_NMI)) {
-        env->halted = 1;
-    }
-    return 1;
-}
-
-void kvm_arch_pre_run(CPUState *env, struct kvm_run *run)
-{
-    if (!kvm_irqchip_in_kernel()) {
-        kvm_set_cr8(env, cpu_get_apic_tpr(env->apic_state));
-    }
-}
-
-int kvm_arch_try_push_interrupts(void *opaque)
-{
-    CPUState *env = cpu_single_env;
-    int r, irq;
-
-    if (kvm_is_ready_for_interrupt_injection(env) &&
-        (env->interrupt_request & CPU_INTERRUPT_HARD) &&
-        (env->eflags & IF_MASK)) {
-        env->interrupt_request &= ~CPU_INTERRUPT_HARD;
-        irq = cpu_get_pic_interrupt(env);
-        if (irq >= 0) {
-            r = kvm_inject_irq(env, irq);
-            if (r < 0) {
-                printf("cpu %d fail inject %x\n", env->cpu_index, irq);
-            }
-        }
-    }
-
-    return (env->interrupt_request & CPU_INTERRUPT_HARD) != 0;
-}
-
-#ifdef KVM_CAP_USER_NMI
-void kvm_arch_push_nmi(void)
-{
-    CPUState *env = cpu_single_env;
-    int r;
-
-    if (likely(!(env->interrupt_request & CPU_INTERRUPT_NMI))) {
-        return;
-    }
-
-    env->interrupt_request &= ~CPU_INTERRUPT_NMI;
-    r = kvm_inject_nmi(env);
-    if (r < 0) {
-        printf("cpu %d fail inject NMI\n", env->cpu_index);
-    }
-}
-#endif /* KVM_CAP_USER_NMI */
-
 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 void kvm_arch_do_ioperm(void *_data)
 {
diff --git a/qemu-kvm.c b/qemu-kvm.c
index 41c4219..b2387df 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -73,31 +73,6 @@ static QLIST_HEAD(, ioperm_data) ioperm_head;
 
 #define ALIGN(x, y) (((x)+(y)-1) & ~((y)-1))
 
-static int handle_unhandled(uint64_t reason)
-{
-    fprintf(stderr, "kvm: unhandled exit %" PRIx64 "\n", reason);
-    return -EINVAL;
-}
-
-#define VMX_INVALID_GUEST_STATE 0x80000021
-
-static int handle_failed_vmentry(uint64_t reason)
-{
-    fprintf(stderr, "kvm: vm entry failed with error 0x%" PRIx64 "\n\n", reason);
-
-    /* Perhaps we will need to check if this machine is intel since exit reason 0x21
-       has a different interpretation on SVM */
-    if (reason == VMX_INVALID_GUEST_STATE) {
-        fprintf(stderr, "If you're runnning a guest on an Intel machine without\n");
-        fprintf(stderr, "unrestricted mode support, the failure can be most likely\n");
-        fprintf(stderr, "due to the guest entering an invalid state for Intel VT.\n");
-        fprintf(stderr, "For example, the guest maybe running in big real mode\n");
-        fprintf(stderr, "which is not supported on less recent Intel processors.\n\n");
-    }
-
-    return -EINVAL;
-}
-
 static inline void set_gsi(KVMState *s, unsigned int gsi)
 {
     uint32_t *bitmap = s->used_gsi_bitmap;
@@ -244,169 +219,6 @@ int kvm_set_irqchip(KVMState *s, struct kvm_irqchip *chip)
 
 #endif
 
-static int handle_mmio(CPUState *env)
-{
-    unsigned long addr = env->kvm_run->mmio.phys_addr;
-    struct kvm_run *kvm_run = env->kvm_run;
-    void *data = kvm_run->mmio.data;
-
-    /* hack: Red Hat 7.1 generates these weird accesses. */
-    if ((addr > 0xa0000 - 4 && addr <= 0xa0000) && kvm_run->mmio.len == 3) {
-        return 0;
-    }
-
-    cpu_physical_memory_rw(addr, data, kvm_run->mmio.len, kvm_run->mmio.is_write);
-    return 0;
-}
-
-static int handle_shutdown(CPUState *env)
-{
-    /* stop the current vcpu from going back to guest mode */
-    env->stopped = 1;
-
-    qemu_system_reset_request();
-    return 1;
-}
-
-static inline void push_nmi(void)
-{
-#ifdef KVM_CAP_USER_NMI
-    kvm_arch_push_nmi();
-#endif                          /* KVM_CAP_USER_NMI */
-}
-
-static void post_kvm_run(CPUState *env)
-{
-    pthread_mutex_lock(&qemu_mutex);
-    cpu_single_env = env;
-
-    kvm_arch_post_run(env, env->kvm_run);
-}
-
-static void pre_kvm_run(CPUState *env)
-{
-    kvm_arch_pre_run(env, env->kvm_run);
-
-    if (env->exit_request) {
-        qemu_cpu_kick_self();
-    }
-
-    cpu_single_env = NULL;
-    pthread_mutex_unlock(&qemu_mutex);
-}
-
-int kvm_is_ready_for_interrupt_injection(CPUState *env)
-{
-    return env->kvm_run->ready_for_interrupt_injection;
-}
-
-static int kvm_run(CPUState *env)
-{
-    int r;
-    struct kvm_run *run = env->kvm_run;
-
-    cpu_single_env = env;
-
-  again:
-    if (env->kvm_vcpu_dirty) {
-        kvm_arch_put_registers(env, KVM_PUT_RUNTIME_STATE);
-        env->kvm_vcpu_dirty = 0;
-    }
-    push_nmi();
-    if (!kvm_state->irqchip_in_kernel) {
-        run->request_interrupt_window = kvm_arch_try_push_interrupts(env);
-    }
-
-    pre_kvm_run(env);
-
-    r = kvm_vcpu_ioctl(env, KVM_RUN, 0);
-
-    post_kvm_run(env);
-
-    kvm_flush_coalesced_mmio_buffer();
-
-    if (r == -EINTR || r == -EAGAIN) {
-        env->exit_request = 0;
-        return 1;
-    }
-    if (r < 0) {
-        fprintf(stderr, "kvm_run: %s\n", strerror(-r));
-        return r;
-    }
-
-    if (1) {
-        switch (run->exit_reason) {
-        case KVM_EXIT_UNKNOWN:
-            r = handle_unhandled(run->hw.hardware_exit_reason);
-            break;
-        case KVM_EXIT_FAIL_ENTRY:
-            r = handle_failed_vmentry(run->fail_entry.hardware_entry_failure_reason);
-            break;
-        case KVM_EXIT_EXCEPTION:
-            fprintf(stderr, "exception %d (%x)\n", run->ex.exception,
-                    run->ex.error_code);
-            cpu_dump_state(env, stderr, fprintf, CPU_DUMP_CODE);
-            abort();
-            break;
-        case KVM_EXIT_IO:
-            kvm_handle_io(run->io.port,
-                                (uint8_t *)run + run->io.data_offset,
-                                run->io.direction,
-                                run->io.size,
-                                run->io.count);
-            r = 0;
-            break;
-        case KVM_EXIT_MMIO:
-            r = handle_mmio(env);
-            break;
-        case KVM_EXIT_HLT:
-            r = kvm_arch_halt(env);
-            break;
-        case KVM_EXIT_IRQ_WINDOW_OPEN:
-            break;
-        case KVM_EXIT_SHUTDOWN:
-            r = handle_shutdown(env);
-            break;
-	case KVM_EXIT_INTERNAL_ERROR:
-            r = kvm_handle_internal_error(env, run);
-	    break;
-        default:
-            r = kvm_arch_run(env);
-            if (r < 0) {
-                fprintf(stderr, "unhandled vm exit: 0x%x\n", run->exit_reason);
-                cpu_dump_state(env, stderr, fprintf, CPU_DUMP_CODE);
-                abort();
-            }
-            if (r > 0) {
-                return r;
-            }
-            break;
-        }
-    }
-    if (!r) {
-        goto again;
-    }
-    env->exit_request = 0;
-    return r;
-}
-
-int kvm_inject_irq(CPUState *env, unsigned irq)
-{
-    struct kvm_interrupt intr;
-
-    intr.irq = irq;
-    return kvm_vcpu_ioctl(env, KVM_INTERRUPT, &intr);
-}
-
-int kvm_inject_nmi(CPUState *env)
-{
-#ifdef KVM_CAP_USER_NMI
-    return kvm_vcpu_ioctl(env, KVM_NMI);
-#else
-    return -ENOSYS;
-#endif
-}
-
 #ifdef KVM_CAP_DEVICE_ASSIGNMENT
 int kvm_assign_pci_device(KVMState *s,
                           struct kvm_assigned_pci_dev *assigned_dev)
@@ -856,20 +668,6 @@ void kvm_update_interrupt_request(CPUState *env)
     }
 }
 
-int kvm_cpu_exec(CPUState *env)
-{
-    int r;
-
-    r = kvm_run(env);
-    if (r < 0) {
-        printf("kvm_run returned %d\n", r);
-        cpu_dump_state(env, stderr, fprintf, CPU_DUMP_CODE);
-        vm_stop(VMSTOP_PANIC);
-    }
-
-    return 0;
-}
-
 static int kvm_cpu_is_stopped(CPUState *env)
 {
     return !vm_running || env->stopped;
@@ -1035,16 +833,21 @@ static void qemu_kvm_system_reset(void)
 static int kvm_main_loop_cpu(CPUState *env)
 {
     while (1) {
-        int run_cpu = !kvm_cpu_is_stopped(env);
-        if (run_cpu) {
-            run_cpu = !kvm_arch_process_async_events(env);
-        }
-        if (run_cpu) {
-            kvm_cpu_exec(env);
-            kvm_main_loop_wait(env, 0);
-        } else {
-            kvm_main_loop_wait(env, 1000);
+        int timeout = 1000;
+        if (!kvm_cpu_is_stopped(env)) {
+            switch (kvm_cpu_exec(env)) {
+            case EXCP_HLT:
+                break;
+            case EXCP_DEBUG:
+                kvm_debug_cpu_requested = env;
+                env->stopped = 1;
+                break;
+            default:
+                timeout = 0;
+                break;
+            }
         }
+        kvm_main_loop_wait(env, timeout);
     }
     pthread_mutex_unlock(&qemu_mutex);
     return 0;
diff --git a/qemu-kvm.h b/qemu-kvm.h
index 5621cfa..57dc6aa 100644
--- a/qemu-kvm.h
+++ b/qemu-kvm.h
@@ -37,35 +37,8 @@
 
 #include "kvm.h"
 
-int kvm_arch_run(CPUState *env);
-
 int kvm_create_irqchip(KVMState *s);
 
-/*!
- * \brief Check if a vcpu is ready for interrupt injection
- *
- * This checks if vcpu interrupts are not masked by mov ss or sti.
- *
- * \param kvm Pointer to the current kvm_context
- * \param vcpu Which virtual CPU should get dumped
- * \return boolean indicating interrupt injection readiness
- */
-int kvm_is_ready_for_interrupt_injection(CPUState *env);
-
-#if defined(__i386__) || defined(__x86_64__)
-/*!
- * \brief Simulate an external vectored interrupt
- *
- * This allows you to simulate an external vectored interrupt.
- *
- * \param kvm Pointer to the current kvm_context
- * \param vcpu Which virtual CPU should get dumped
- * \param irq Vector number
- * \return 0 on success
- */
-int kvm_inject_irq(CPUState *env, unsigned irq);
-#endif
-
 #ifdef KVM_CAP_IRQCHIP
 /*!
  * \brief Dump in kernel IRQCHIP contents
@@ -115,17 +88,6 @@ int kvm_set_lapic(CPUState *env, struct kvm_lapic_state *s);
 
 #endif
 
-/*!
- * \brief Simulate an NMI
- *
- * This allows you to simulate a non-maskable interrupt.
- *
- * \param kvm Pointer to the current kvm_context
- * \param vcpu Which virtual CPU should get dumped
- * \return 0 on success
- */
-int kvm_inject_nmi(CPUState *env);
-
 #endif
 
 #ifdef KVM_CAP_PIT
@@ -314,8 +276,6 @@ void kvm_hpet_disable_kpit(void);
 void on_vcpu(CPUState *env, void (*func)(void *data), void *data);
 void kvm_update_interrupt_request(CPUState *env);
 
-int kvm_arch_try_push_interrupts(void *opaque);
-void kvm_arch_push_nmi(void);
 int kvm_set_boot_cpu_id(KVMState *s, uint32_t id);
 
 void kvm_tpr_access_report(CPUState *env, uint64_t rip, int is_write);
@@ -347,9 +307,7 @@ struct ioperm_data {
     QLIST_ENTRY(ioperm_data) entries;
 };
 
-int kvm_arch_halt(CPUState *env);
-int handle_tpr_access(void *opaque, CPUState *env, uint64_t rip,
-                      int is_write);
+int kvm_handle_tpr_access(CPUState *env);
 
 #else
 #define kvm_nested 0
diff --git a/target-i386/kvm.c b/target-i386/kvm.c
index e98eccf..bccd87c 100644
--- a/target-i386/kvm.c
+++ b/target-i386/kvm.c
@@ -1565,7 +1565,6 @@ int kvm_arch_get_registers(CPUState *env)
     return 0;
 }
 
-#ifdef OBSOLETE_KVM_IMPL
 void kvm_arch_pre_run(CPUState *env, struct kvm_run *run)
 {
     int ret;
@@ -1623,7 +1622,6 @@ void kvm_arch_pre_run(CPUState *env, struct kvm_run *run)
         run->cr8 = cpu_get_apic_tpr(env->apic_state);
     }
 }
-#endif
 
 void kvm_arch_post_run(CPUState *env, struct kvm_run *run)
 {
-- 
1.7.1

[PATCH 10/10] qemu-kvm: Rework ioport access management

[Jan Kiszka]

Clean up the interface for enabling/disabling direct ioport access for
assigned devices. There is now only a register and a deregister service.
Both are automatically updating the access on all vcpus. Besides that,
there is an update service for newly created VCPUs that applies all
currently registered regions.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 hw/device-assignment.c |   20 +++------
 qemu-kvm-x86.c         |   12 ++++--
 qemu-kvm.c             |  107 ++++++++++++++++++++++++++++++++++--------------
 qemu-kvm.h             |   21 ++-------
 4 files changed, 96 insertions(+), 64 deletions(-)

diff --git a/hw/device-assignment.c b/hw/device-assignment.c
index abc38f8..57d8dc0 100644
--- a/hw/device-assignment.c
+++ b/hw/device-assignment.c
@@ -298,7 +298,7 @@ static void assigned_dev_ioport_map(PCIDevice *pci_dev, int region_num,
     AssignedDevice *r_dev = DO_UPCAST(AssignedDevice, dev, pci_dev);
     AssignedDevRegion *region = &r_dev->v_addrs[region_num];
     int first_map = (region->e_size == 0);
-    CPUState *env;
+    int r;
 
     region->e_physbase = addr;
     region->e_size = size;
@@ -307,17 +307,11 @@ static void assigned_dev_ioport_map(PCIDevice *pci_dev, int region_num,
           addr, region->u.r_baseport, type, size, region_num);
 
     if (first_map && region->region->resource_fd < 0) {
-	struct ioperm_data *data;
-
-	data = qemu_mallocz(sizeof(struct ioperm_data));
-	data->start_port = region->u.r_baseport;
-	data->num = region->r_size;
-	data->turn_on = 1;
-
-	kvm_add_ioperm_data(data);
-
-	for (env = first_cpu; env; env = env->next_cpu)
-	    kvm_ioperm(env, data);
+        r = kvm_add_ioport_region(region->u.r_baseport, region->r_size);
+        if (r < 0) {
+            fprintf(stderr, "%s: failed to enable ioport access (%m)\n",
+                    __func__);
+        }
     }
 
     register_ioport_read(addr, size, 1, assigned_dev_ioport_readb,
@@ -832,7 +826,7 @@ static void free_assigned_device(AssignedDevice *dev)
         }
         if (pci_region->type & IORESOURCE_IO) {
             if (pci_region->resource_fd < 0) {
-                kvm_remove_ioperm_data(region->u.r_baseport, region->r_size);
+                kvm_remove_ioport_region(region->u.r_baseport, region->r_size);
             }
         } else if (pci_region->type & IORESOURCE_MEM) {
             if (region->u.r_virtbase) {
diff --git a/qemu-kvm-x86.c b/qemu-kvm-x86.c
index 89bb692..2a01ccc 100644
--- a/qemu-kvm-x86.c
+++ b/qemu-kvm-x86.c
@@ -164,14 +164,18 @@ static int _kvm_arch_init_vcpu(CPUState *env)
 #ifdef KVM_EXIT_TPR_ACCESS
     kvm_enable_tpr_access_reporting(env);
 #endif
-    return 0;
+
+    return kvm_update_ioport_access(env);
 }
 
 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
-void kvm_arch_do_ioperm(void *_data)
+int kvm_arch_set_ioport_access(unsigned long start, unsigned long size,
+                               bool enable)
 {
-    struct ioperm_data *data = _data;
-    ioperm(data->start_port, data->num, data->turn_on);
+    if (ioperm(start, size, enable) < 0) {
+        return -errno;
+    }
+    return 0;
 }
 #endif
 
diff --git a/qemu-kvm.c b/qemu-kvm.c
index b2387df..94e12f3 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -66,11 +66,6 @@ static int qemu_system_ready;
 
 CPUState *kvm_debug_cpu_requested;
 
-#ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
-/* The list of ioperm_data */
-static QLIST_HEAD(, ioperm_data) ioperm_head;
-#endif
-
 #define ALIGN(x, y) (((x)+(y)-1) & ~((y)-1))
 
 static inline void set_gsi(KVMState *s, unsigned int gsi)
@@ -856,19 +851,10 @@ static int kvm_main_loop_cpu(CPUState *env)
 static void *ap_main_loop(void *_env)
 {
     CPUState *env = _env;
-#ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
-    struct ioperm_data *data = NULL;
-#endif
 
     current_env = env;
     env->thread_id = kvm_get_thread_id();
 
-#ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
-    /* do ioperm for io ports of assigned devices */
-    QLIST_FOREACH(data, &ioperm_head, entries)
-        on_vcpu(env, kvm_arch_do_ioperm, data);
-#endif
-
     pthread_mutex_lock(&qemu_mutex);
     cpu_single_env = env;
 
@@ -1069,36 +1055,95 @@ void qemu_mutex_lock_iothread(void)
 }
 
 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
-void kvm_add_ioperm_data(struct ioperm_data *data)
+typedef struct KVMIOPortRegion {
+    unsigned long start;
+    unsigned long size;
+    int status;
+    QLIST_ENTRY(KVMIOPortRegion) entry;
+} KVMIOPortRegion;
+
+static QLIST_HEAD(, KVMIOPortRegion) ioport_regions;
+
+static void do_set_ioport_access(void *data)
 {
-    QLIST_INSERT_HEAD(&ioperm_head, data, entries);
+    KVMIOPortRegion *region = data;
+    bool enable = region->status > 0;
+    int r;
+
+    r = kvm_arch_set_ioport_access(region->start, region->size, enable);
+    if (r < 0) {
+        region->status = r;
+    } else {
+        region->status = 1;
+    }
 }
 
-void kvm_remove_ioperm_data(unsigned long start_port, unsigned long num)
+int kvm_add_ioport_region(unsigned long start, unsigned long size)
 {
-    struct ioperm_data *data;
-
-    data = QLIST_FIRST(&ioperm_head);
-    while (data) {
-        struct ioperm_data *next = QLIST_NEXT(data, entries);
+    KVMIOPortRegion *region = qemu_mallocz(sizeof(KVMIOPortRegion));
+    CPUState *env;
+    int r = 0;
 
-        if (data->start_port == start_port && data->num == num) {
-            QLIST_REMOVE(data, entries);
-            qemu_free(data);
+    region->start = start;
+    region->size = size;
+    region->status = 1;
+    QLIST_INSERT_HEAD(&ioport_regions, region, entry);
+
+    if (qemu_system_ready) {
+        for (env = first_cpu; env != NULL; env = env->next_cpu) {
+            on_vcpu(env, do_set_ioport_access, region);
+            if (region->status < 0) {
+                r = region->status;
+                kvm_remove_ioport_region(start, size);
+                break;
+            }
         }
-
-        data = next;
     }
+    return r;
 }
 
-void kvm_ioperm(CPUState *env, void *data)
+int kvm_remove_ioport_region(unsigned long start, unsigned long size)
 {
-    if (kvm_enabled() && qemu_system_ready) {
-        on_vcpu(env, kvm_arch_do_ioperm, data);
+    KVMIOPortRegion *region, *tmp;
+    CPUState *env;
+    int r = -ENOENT;
+
+    QLIST_FOREACH_SAFE(region, &ioport_regions, entry, tmp) {
+        if (region->start == start && region->size == size) {
+            region->status = 0;
+        }
+        if (qemu_system_ready) {
+            for (env = first_cpu; env != NULL; env = env->next_cpu) {
+                on_vcpu(env, do_set_ioport_access, region);
+            }
+        }
+        QLIST_REMOVE(region, entry);
+        qemu_free(region);
+        r = 0;
     }
+    return r;
 }
+#endif /* CONFIG_KVM_DEVICE_ASSIGNMENT */
 
-#endif
+int kvm_update_ioport_access(CPUState *env)
+{
+#ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
+    KVMIOPortRegion *region;
+    int r;
+
+    assert(qemu_cpu_is_self(env));
+
+    QLIST_FOREACH(region, &ioport_regions, entry) {
+        bool enable = region->status > 0;
+
+        r = kvm_arch_set_ioport_access(region->start, region->size, enable);
+        if (r < 0) {
+            return r;
+        }
+    }
+#endif /* CONFIG_KVM_DEVICE_ASSIGNMENT */
+    return 0;
+}
 
 int kvm_set_boot_cpu_id(KVMState *s, uint32_t id)
 {
diff --git a/qemu-kvm.h b/qemu-kvm.h
index 57dc6aa..094aef2 100644
--- a/qemu-kvm.h
+++ b/qemu-kvm.h
@@ -282,31 +282,20 @@ void kvm_tpr_access_report(CPUState *env, uint64_t rip, int is_write);
 
 int kvm_arch_init_irq_routing(void);
 
-#ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
-struct ioperm_data;
+int kvm_add_ioport_region(unsigned long start, unsigned long size);
+int kvm_remove_ioport_region(unsigned long start, unsigned long size);
 
-void kvm_ioperm(CPUState *env, void *data);
-void kvm_add_ioperm_data(struct ioperm_data *data);
-void kvm_remove_ioperm_data(unsigned long start_port, unsigned long num);
-void kvm_arch_do_ioperm(void *_data);
-#endif
+int kvm_update_ioport_access(CPUState *env);
+int kvm_arch_set_ioport_access(unsigned long start, unsigned long size,
+                               bool enable);
 
 #ifdef CONFIG_KVM
-#include "qemu-queue.h"
-
 extern int kvm_irqchip;
 extern int kvm_pit;
 extern int kvm_pit_reinject;
 extern int kvm_nested;
 extern unsigned int kvm_shadow_memory;
 
-struct ioperm_data {
-    unsigned long start_port;
-    unsigned long num;
-    int turn_on;
-    QLIST_ENTRY(ioperm_data) entries;
-};
-
 int kvm_handle_tpr_access(CPUState *env);
 
 #else
-- 
1.7.1

Re: [PATCH 00/10] qemu-kvm: Cleanup and switch to upstream - Part III

[Alexander Graf]

On 20.05.2011, at 19:12, Jan Kiszka wrote:

> This is a rather short round as the next and final one cannot be split
> up very well.
> 
> We start with three code cleanup patches, then work towards using
> upstream kvm_cpu_exec, and finally rework the core's PIO access
> management used for device assignment.
> 
> Please review/merge.

So with this set applied, s390 and ppc targets should work in qemu-kvm.git?


Alex

Re: [PATCH 00/10] qemu-kvm: Cleanup and switch to upstream - Part III

[Jan Kiszka]

On 2011-05-20 19:14, Alexander Graf wrote:
> 
> On 20.05.2011, at 19:12, Jan Kiszka wrote:
> 
>> This is a rather short round as the next and final one cannot be split
>> up very well.
>>
>> We start with three code cleanup patches, then work towards using
>> upstream kvm_cpu_exec, and finally rework the core's PIO access
>> management used for device assignment.
>>
>> Please review/merge.
> 
> So with this set applied, s390 and ppc targets should work in qemu-kvm.git?

I bet not yet. Part IV will finally switch over to upstream vcpu loop,
and that should make them work.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [PATCH 09/10] qemu-kvm: Use upstream kvm_cpu_exec

[Christoph Hellwig]

On Fri, May 20, 2011 at 07:12:39PM +0200, Jan Kiszka wrote:
> Upstream's and qemu-kvm's kvm_cpu_exec are not logically equivalent so

s/not/now/?

Re: [PATCH 09/10] qemu-kvm: Use upstream kvm_cpu_exec

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 313 bytes --]

On 2011-05-20 19:17, Christoph Hellwig wrote:
> On Fri, May 20, 2011 at 07:12:39PM +0200, Jan Kiszka wrote:
>> Upstream's and qemu-kvm's kvm_cpu_exec are not logically equivalent so
> 
> s/not/now/?

Oops, of course.

If there is no other need to repost, this should be fixed on merge.

Thanks,
Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 259 bytes --]

Re: [PATCH 00/10] qemu-kvm: Cleanup and switch to upstream - Part III

[Avi Kivity]

On 05/20/2011 08:12 PM, Jan Kiszka wrote:
> This is a rather short round as the next and final one cannot be split
> up very well.
>
> We start with three code cleanup patches, then work towards using
> upstream kvm_cpu_exec, and finally rework the core's PIO access
> management used for device assignment.
>
> Please review/merge.
>

Thanks, applied.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.