[PATCH v11 00/19] Initial support for multi-process Qemu

[PATCH v11 00/19] Initial support for multi-process Qemu

[Jagannathan Raman]

Hello,

This is the v11 of the patchset. Thank you very much for the
review of the v10 of the series. We are glad to hear the
patchset is getting in a better shape.

We made changes to the following patches in this series.

[PATCH v10 06/19] multi-process: define MPQemuMsg format and transmission functions
[PATCH v10 08/19] multi-process: Associate fd of a PCIDevice with its object
[PATCH v10 10/19] multi-process: introduce proxy object

To touch upon the history of this project, we posted the
Proof Of Concept patches before the BoF session in 2018.
Subsequently, we posted 9 versions on the qemu-devel mailist.
You can find them by following the links below ([1] - [9]).

Following people contributed to the design and
implementation of this project:
Jagannathan Raman <jag.raman@oracle.com>
Elena Ufimtseva <elena.ufimtseva@oracle.com>
John G Johnson <john.g.johnson@oracle.com>
Stefan Hajnoczi <stefanha@redhat.com>
Konrad Wilk <konrad.wilk@oracle.com>
Kanth Ghatraju <kanth.ghatraju@oracle.com>

We would like to thank QEMU community for your feedback in the
design and implementation of this project.

Qemu wiki page:
https://wiki.qemu.org/Features/MultiProcessQEMU

For the full concept writeup about QEMU multi-process, please refer to
docs/devel/qemu-multiprocess.rst. Also see docs/qemu-multiprocess.txt for
usage information.


We welcome all your ideas, concerns, and questions for this patchset.
[POC]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg566538.html
[1]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg602285.html
[2]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg624877.html
[3]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg642000.html
[4]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg655118.html
[5]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg682429.html
[6]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg697484.html
[7]: https://patchew.org/QEMU/cover.1593273671.git.elena.ufimtseva@oracle.com/
[8]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg727007.html
[9]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg734275.html
[10]: https://www.mail-archive.com/qemu-devel@nongnu.org/msg747638.html

Elena Ufimtseva (7):
  multi-process: add qio channel function to transmit
  multi-process: define MPQemuMsg format and transmission functions
  multi-process: introduce proxy object
  multi-process: add proxy communication functions
  multi-process: Forward PCI config space acceses to the remote process
  multi-process: perform device reset in the remote process
  multi-process: add configure and usage information

Jagannathan Raman (11):
  memory: alloc RAM from file at offset
  multi-process: Add config option for multi-process QEMU
  multi-process: setup PCI host bridge for remote device
  multi-process: setup a machine object for remote device process
  multi-process: Initialize message handler in remote device
  multi-process: Associate fd of a PCIDevice with its object
  multi-process: setup memory manager for remote device
  multi-process: PCI BAR read/write handling for proxy & remote
    endpoints
  multi-process: Synchronize remote memory
  multi-process: create IOHUB object to handle irq
  multi-process: Retrieve PCI info from remote process

John G Johnson (1):
  multi-process: add the concept description to
    docs/devel/qemu-multiprocess

 MAINTAINERS                     |  26 ++
 backends/hostmem-memfd.c        |   2 +-
 configure                       |  10 +
 docs/devel/index.rst            |   1 +
 docs/devel/multi-process.rst    | 966 ++++++++++++++++++++++++++++++++++++++++
 docs/multi-process.rst          |  67 +++
 hw/i386/meson.build             |   5 +
 hw/i386/remote-iohub.c          | 123 +++++
 hw/i386/remote-memory.c         |  58 +++
 hw/i386/remote-msg.c            | 241 ++++++++++
 hw/i386/remote-obj.c            | 154 +++++++
 hw/i386/remote.c                |  79 ++++
 hw/misc/ivshmem.c               |   3 +-
 hw/pci-host/meson.build         |   1 +
 hw/pci-host/remote.c            |  75 ++++
 hw/pci/memory-sync.c            | 210 +++++++++
 hw/pci/meson.build              |   3 +
 hw/pci/proxy.c                  | 377 ++++++++++++++++
 include/exec/memory.h           |   2 +
 include/exec/ram_addr.h         |   2 +-
 include/hw/i386/remote-iohub.h  |  42 ++
 include/hw/i386/remote-memory.h |  19 +
 include/hw/i386/remote-obj.h    |  42 ++
 include/hw/i386/remote.h        |  40 ++
 include/hw/pci-host/remote.h    |  31 ++
 include/hw/pci/memory-sync.h    |  27 ++
 include/hw/pci/pci_ids.h        |   3 +
 include/hw/pci/proxy.h          |  53 +++
 include/io/channel.h            |  24 +
 include/io/mpqemu-link.h        |  98 ++++
 include/qemu/mmap-alloc.h       |   3 +-
 io/channel.c                    |  45 ++
 io/meson.build                  |   2 +
 io/mpqemu-link.c                | 303 +++++++++++++
 meson.build                     |   1 +
 scripts/mpqemu-launcher.py      |  49 ++
 softmmu/memory.c                |   3 +-
 softmmu/physmem.c               |  11 +-
 util/mmap-alloc.c               |   7 +-
 util/oslib-posix.c              |   2 +-
 40 files changed, 3197 insertions(+), 13 deletions(-)
 create mode 100644 docs/devel/multi-process.rst
 create mode 100644 docs/multi-process.rst
 create mode 100644 hw/i386/remote-iohub.c
 create mode 100644 hw/i386/remote-memory.c
 create mode 100644 hw/i386/remote-msg.c
 create mode 100644 hw/i386/remote-obj.c
 create mode 100644 hw/i386/remote.c
 create mode 100644 hw/pci-host/remote.c
 create mode 100644 hw/pci/memory-sync.c
 create mode 100644 hw/pci/proxy.c
 create mode 100644 include/hw/i386/remote-iohub.h
 create mode 100644 include/hw/i386/remote-memory.h
 create mode 100644 include/hw/i386/remote-obj.h
 create mode 100644 include/hw/i386/remote.h
 create mode 100644 include/hw/pci-host/remote.h
 create mode 100644 include/hw/pci/memory-sync.h
 create mode 100644 include/hw/pci/proxy.h
 create mode 100644 include/io/mpqemu-link.h
 create mode 100644 io/mpqemu-link.c
 create mode 100755 scripts/mpqemu-launcher.py

-- 
1.8.3.1

[PATCH v11 01/19] memory: alloc RAM from file at offset

[Jagannathan Raman]

Allow RAM MemoryRegion to be created from an offset in a file, instead
of allocating at offset of 0 by default. This is needed to synchronize
RAM between QEMU & remote process.

Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 backends/hostmem-memfd.c  |  2 +-
 hw/misc/ivshmem.c         |  3 ++-
 include/exec/memory.h     |  2 ++
 include/exec/ram_addr.h   |  2 +-
 include/qemu/mmap-alloc.h |  3 ++-
 softmmu/memory.c          |  3 ++-
 softmmu/physmem.c         | 11 +++++++----
 util/mmap-alloc.c         |  7 ++++---
 util/oslib-posix.c        |  2 +-
 9 files changed, 22 insertions(+), 13 deletions(-)

diff --git a/backends/hostmem-memfd.c b/backends/hostmem-memfd.c
index e5626d4..69b0ae3 100644
--- a/backends/hostmem-memfd.c
+++ b/backends/hostmem-memfd.c
@@ -55,7 +55,7 @@ memfd_backend_memory_alloc(HostMemoryBackend *backend, Error **errp)
     name = host_memory_backend_get_name(backend);
     memory_region_init_ram_from_fd(&backend->mr, OBJECT(backend),
                                    name, backend->size,
-                                   backend->share, fd, errp);
+                                   backend->share, fd, 0, errp);
     g_free(name);
 }
 
diff --git a/hw/misc/ivshmem.c b/hw/misc/ivshmem.c
index e321e5c..8d3e1ee 100644
--- a/hw/misc/ivshmem.c
+++ b/hw/misc/ivshmem.c
@@ -494,7 +494,8 @@ static void process_msg_shmem(IVShmemState *s, int fd, Error **errp)
 
     /* mmap the region and map into the BAR2 */
     memory_region_init_ram_from_fd(&s->server_bar2, OBJECT(s),
-                                   "ivshmem.bar2", size, true, fd, &local_err);
+                                   "ivshmem.bar2", size, true, fd, 0,
+                                   &local_err);
     if (local_err) {
         error_propagate(errp, local_err);
         return;
diff --git a/include/exec/memory.h b/include/exec/memory.h
index 622207b..5eb2ece 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -934,6 +934,7 @@ void memory_region_init_ram_from_file(MemoryRegion *mr,
  * @size: size of the region.
  * @share: %true if memory must be mmaped with the MAP_SHARED flag
  * @fd: the fd to mmap.
+ * @offset: offset within the file referenced by fd
  * @errp: pointer to Error*, to store an error if it happens.
  *
  * Note that this function does not do anything to cause the data in the
@@ -945,6 +946,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
                                     uint64_t size,
                                     bool share,
                                     int fd,
+                                    ram_addr_t offset,
                                     Error **errp);
 #endif
 
diff --git a/include/exec/ram_addr.h b/include/exec/ram_addr.h
index c6d2ef1..d465a48 100644
--- a/include/exec/ram_addr.h
+++ b/include/exec/ram_addr.h
@@ -121,7 +121,7 @@ RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr,
                                    Error **errp);
 RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr,
                                  uint32_t ram_flags, int fd,
-                                 Error **errp);
+                                 off_t offset, Error **errp);
 
 RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
                                   MemoryRegion *mr, Error **errp);
diff --git a/include/qemu/mmap-alloc.h b/include/qemu/mmap-alloc.h
index e786266..4f57985 100644
--- a/include/qemu/mmap-alloc.h
+++ b/include/qemu/mmap-alloc.h
@@ -25,7 +25,8 @@ void *qemu_ram_mmap(int fd,
                     size_t size,
                     size_t align,
                     bool shared,
-                    bool is_pmem);
+                    bool is_pmem,
+                    off_t start);
 
 void qemu_ram_munmap(int fd, void *ptr, size_t size);
 
diff --git a/softmmu/memory.c b/softmmu/memory.c
index 403ff3a..f7d81ee 100644
--- a/softmmu/memory.c
+++ b/softmmu/memory.c
@@ -1591,6 +1591,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
                                     uint64_t size,
                                     bool share,
                                     int fd,
+                                    ram_addr_t offset,
                                     Error **errp)
 {
     Error *err = NULL;
@@ -1600,7 +1601,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
     mr->destructor = memory_region_destructor_ram;
     mr->ram_block = qemu_ram_alloc_from_fd(size, mr,
                                            share ? RAM_SHARED : 0,
-                                           fd, &err);
+                                           fd, offset, &err);
     mr->dirty_log_mask = tcg_enabled() ? (1 << DIRTY_MEMORY_CODE) : 0;
     if (err) {
         mr->size = int128_zero();
diff --git a/softmmu/physmem.c b/softmmu/physmem.c
index e319fb2..6203ea3 100644
--- a/softmmu/physmem.c
+++ b/softmmu/physmem.c
@@ -1461,6 +1461,7 @@ static void *file_ram_alloc(RAMBlock *block,
                             ram_addr_t memory,
                             int fd,
                             bool truncate,
+                            off_t offset,
                             Error **errp)
 {
     void *area;
@@ -1511,7 +1512,8 @@ static void *file_ram_alloc(RAMBlock *block,
     }
 
     area = qemu_ram_mmap(fd, memory, block->mr->align,
-                         block->flags & RAM_SHARED, block->flags & RAM_PMEM);
+                         block->flags & RAM_SHARED, block->flags & RAM_PMEM,
+                         offset);
     if (area == MAP_FAILED) {
         error_setg_errno(errp, errno,
                          "unable to map backing store for guest RAM");
@@ -1943,7 +1945,7 @@ static void ram_block_add(RAMBlock *new_block, Error **errp, bool shared)
 #ifdef CONFIG_POSIX
 RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr,
                                  uint32_t ram_flags, int fd,
-                                 Error **errp)
+                                 off_t offset, Error **errp)
 {
     RAMBlock *new_block;
     Error *local_err = NULL;
@@ -1996,7 +1998,8 @@ RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr,
     new_block->used_length = size;
     new_block->max_length = size;
     new_block->flags = ram_flags;
-    new_block->host = file_ram_alloc(new_block, size, fd, !file_size, errp);
+    new_block->host = file_ram_alloc(new_block, size, fd, !file_size, offset,
+                                     errp);
     if (!new_block->host) {
         g_free(new_block);
         return NULL;
@@ -2026,7 +2029,7 @@ RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr,
         return NULL;
     }
 
-    block = qemu_ram_alloc_from_fd(size, mr, ram_flags, fd, errp);
+    block = qemu_ram_alloc_from_fd(size, mr, ram_flags, fd, 0, errp);
     if (!block) {
         if (created) {
             unlink(mem_path);
diff --git a/util/mmap-alloc.c b/util/mmap-alloc.c
index 27dcccd..a28f702 100644
--- a/util/mmap-alloc.c
+++ b/util/mmap-alloc.c
@@ -86,7 +86,8 @@ void *qemu_ram_mmap(int fd,
                     size_t size,
                     size_t align,
                     bool shared,
-                    bool is_pmem)
+                    bool is_pmem,
+                    off_t start)
 {
     int flags;
     int map_sync_flags = 0;
@@ -147,7 +148,7 @@ void *qemu_ram_mmap(int fd,
     offset = QEMU_ALIGN_UP((uintptr_t)guardptr, align) - (uintptr_t)guardptr;
 
     ptr = mmap(guardptr + offset, size, PROT_READ | PROT_WRITE,
-               flags | map_sync_flags, fd, 0);
+               flags | map_sync_flags, fd, start);
 
     if (ptr == MAP_FAILED && map_sync_flags) {
         if (errno == ENOTSUP) {
@@ -172,7 +173,7 @@ void *qemu_ram_mmap(int fd,
          * we will remove these flags to handle compatibility.
          */
         ptr = mmap(guardptr + offset, size, PROT_READ | PROT_WRITE,
-                   flags, fd, 0);
+                   flags, fd, start);
     }
 
     if (ptr == MAP_FAILED) {
diff --git a/util/oslib-posix.c b/util/oslib-posix.c
index f15234b..93874df 100644
--- a/util/oslib-posix.c
+++ b/util/oslib-posix.c
@@ -227,7 +227,7 @@ void *qemu_memalign(size_t alignment, size_t size)
 void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment, bool shared)
 {
     size_t align = QEMU_VMALLOC_ALIGN;
-    void *ptr = qemu_ram_mmap(-1, size, align, shared, false);
+    void *ptr = qemu_ram_mmap(-1, size, align, shared, false, 0);
 
     if (ptr == MAP_FAILED) {
         return NULL;
-- 
1.8.3.1

[PATCH v11 02/19] multi-process: Add config option for multi-process QEMU

[Jagannathan Raman]

Add a configuration option to separate multi-process code

Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 configure   | 10 ++++++++++
 meson.build |  1 +
 2 files changed, 11 insertions(+)

diff --git a/configure b/configure
index f839c2a..2025e89 100755
--- a/configure
+++ b/configure
@@ -447,6 +447,7 @@ meson=""
 ninja=""
 skip_meson=no
 gettext=""
+mpqemu="no"
 
 bogus_os="no"
 malloc_trim="auto"
@@ -1520,6 +1521,10 @@ for opt do
   ;;
   --disable-libdaxctl) libdaxctl=no
   ;;
+  --enable-mpqemu) mpqemu=yes
+  ;;
+  --disable-mpqemu) mpqemu=no
+  ;;
   *)
       echo "ERROR: unknown option $opt"
       echo "Try '$0 --help' for more information"
@@ -1838,6 +1843,8 @@ disabled with --disable-FEATURE, default is enabled if available:
   xkbcommon       xkbcommon support
   rng-none        dummy RNG, avoid using /dev/(u)random and getrandom()
   libdaxctl       libdaxctl support
+  mpqemu          multi-process QEMU support
+
 
 NOTE: The object files are built at the place where configure is launched
 EOF
@@ -6737,6 +6744,9 @@ fi
 if test "$have_mlockall" = "yes" ; then
   echo "HAVE_MLOCKALL=y" >> $config_host_mak
 fi
+if test "$mpqemu" = "yes" ; then
+  echo "CONFIG_MPQEMU=y" >> $config_host_mak
+fi
 if test "$fuzzing" = "yes" ; then
   # If LIB_FUZZING_ENGINE is set, assume we are running on OSS-Fuzz, and the
   # needed CFLAGS have already been provided
diff --git a/meson.build b/meson.build
index 1a4a482..bbcf524 100644
--- a/meson.build
+++ b/meson.build
@@ -2091,6 +2091,7 @@ endif
 summary_info += {'thread sanitizer':  config_host.has_key('CONFIG_TSAN')}
 summary_info += {'rng-none':          config_host.has_key('CONFIG_RNG_NONE')}
 summary_info += {'Linux keyring':     config_host.has_key('CONFIG_SECRET_KEYRING')}
+summary_info += {'Multi-Process QEMU': config_host.has_key('CONFIG_MPQEMU')}
 summary(summary_info, bool_yn: true)
 
 if not supported_cpus.contains(cpu)
-- 
1.8.3.1

[PATCH v11 03/19] multi-process: setup PCI host bridge for remote device

[Jagannathan Raman]

PCI host bridge is setup for the remote device process. It is
implemented using remote-pcihost object. It is an extension of the PCI
host bridge setup by QEMU.
Remote-pcihost configures a PCI bus which could be used by the remote
PCI device to latch on to.

Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                  |  8 +++++
 hw/pci-host/meson.build      |  1 +
 hw/pci-host/remote.c         | 75 ++++++++++++++++++++++++++++++++++++++++++++
 include/hw/pci-host/remote.h | 30 ++++++++++++++++++
 4 files changed, 114 insertions(+)
 create mode 100644 hw/pci-host/remote.c
 create mode 100644 include/hw/pci-host/remote.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 47dd38a..bb1d701 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3095,6 +3095,14 @@ S: Maintained
 F: hw/semihosting/
 F: include/hw/semihosting/
 
+Multi-process QEMU
+M: Jagannathan Raman <jag.raman@oracle.com>
+M: Elena Ufimtseva <elena.ufimtseva@oracle.com>
+M: John G Johnson <john.g.johnson@oracle.com>
+S: Maintained
+F: hw/pci-host/remote.c
+F: include/hw/pci-host/remote.h
+
 Build and test automation
 -------------------------
 Build and test automation
diff --git a/hw/pci-host/meson.build b/hw/pci-host/meson.build
index e6d1b89..1bcecf6 100644
--- a/hw/pci-host/meson.build
+++ b/hw/pci-host/meson.build
@@ -9,6 +9,7 @@ pci_ss.add(when: 'CONFIG_PCI_EXPRESS_XILINX', if_true: files('xilinx-pcie.c'))
 pci_ss.add(when: 'CONFIG_PCI_I440FX', if_true: files('i440fx.c'))
 pci_ss.add(when: 'CONFIG_PCI_SABRE', if_true: files('sabre.c'))
 pci_ss.add(when: 'CONFIG_XEN_IGD_PASSTHROUGH', if_true: files('xen_igd_pt.c'))
+pci_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
 
 # PPC devices
 pci_ss.add(when: 'CONFIG_PREP_PCI', if_true: files('prep.c'))
diff --git a/hw/pci-host/remote.c b/hw/pci-host/remote.c
new file mode 100644
index 0000000..11325e2
--- /dev/null
+++ b/hw/pci-host/remote.c
@@ -0,0 +1,75 @@
+/*
+ * Remote PCI host device
+ *
+ * Unlike PCI host devices that model physical hardware, the purpose
+ * of this PCI host is to host multi-process QEMU devices.
+ *
+ * Multi-process QEMU extends the PCI host of a QEMU machine into a
+ * remote process. Any PCI device attached to the remote process is
+ * visible in the QEMU guest. This allows existing QEMU device models
+ * to be reused in the remote process.
+ *
+ * This PCI host is purely a container for PCI devices. It's fake in the
+ * sense that the guest never sees this PCI host and has no way of
+ * accessing it. Its job is just to provide the environment that QEMU
+ * PCI device models need when running in a remote process.
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/pci/pci.h"
+#include "hw/pci/pci_host.h"
+#include "hw/pci/pcie_host.h"
+#include "hw/qdev-properties.h"
+#include "hw/pci-host/remote.h"
+#include "exec/memory.h"
+
+static const char *remote_pcihost_root_bus_path(PCIHostState *host_bridge,
+                                                PCIBus *rootbus)
+{
+    return "0000:00";
+}
+
+static void remote_pcihost_realize(DeviceState *dev, Error **errp)
+{
+    PCIHostState *pci = PCI_HOST_BRIDGE(dev);
+    RemotePCIHost *s = REMOTE_HOST_DEVICE(dev);
+
+    pci->bus = pci_root_bus_new(DEVICE(s), "remote-pci",
+                                s->mr_pci_mem, s->mr_sys_io,
+                                0, TYPE_PCIE_BUS);
+}
+
+static void remote_pcihost_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
+
+    hc->root_bus_path = remote_pcihost_root_bus_path;
+    dc->realize = remote_pcihost_realize;
+
+    dc->user_creatable = false;
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->fw_name = "pci";
+}
+
+static const TypeInfo remote_pcihost_info = {
+    .name = TYPE_REMOTE_HOST_DEVICE,
+    .parent = TYPE_PCIE_HOST_BRIDGE,
+    .instance_size = sizeof(RemotePCIHost),
+    .class_init = remote_pcihost_class_init,
+};
+
+static void remote_pcihost_register(void)
+{
+    type_register_static(&remote_pcihost_info);
+}
+
+type_init(remote_pcihost_register)
diff --git a/include/hw/pci-host/remote.h b/include/hw/pci-host/remote.h
new file mode 100644
index 0000000..bab6d3c
--- /dev/null
+++ b/include/hw/pci-host/remote.h
@@ -0,0 +1,30 @@
+/*
+ * PCI Host for remote device
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef REMOTE_PCIHOST_H
+#define REMOTE_PCIHOST_H
+
+#include "exec/memory.h"
+#include "hw/pci/pcie_host.h"
+
+#define TYPE_REMOTE_HOST_DEVICE "remote-pcihost"
+#define REMOTE_HOST_DEVICE(obj) \
+    OBJECT_CHECK(RemotePCIHost, (obj), TYPE_REMOTE_HOST_DEVICE)
+
+typedef struct RemotePCIHost {
+    /*< private >*/
+    PCIExpressHost parent_obj;
+    /*< public >*/
+
+    MemoryRegion *mr_pci_mem;
+    MemoryRegion *mr_sys_io;
+} RemotePCIHost;
+
+#endif
-- 
1.8.3.1

[PATCH v11 04/19] multi-process: setup a machine object for remote device process

[Jagannathan Raman]

remote-machine object sets up various subsystems of the remote
device process. Instantiate PCI host bridge object and initialize RAM, IO &
PCI memory regions.

Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                  |  2 ++
 hw/i386/meson.build          |  1 +
 hw/i386/remote.c             | 69 ++++++++++++++++++++++++++++++++++++++++++++
 include/hw/i386/remote.h     | 28 ++++++++++++++++++
 include/hw/pci-host/remote.h |  1 +
 5 files changed, 101 insertions(+)
 create mode 100644 hw/i386/remote.c
 create mode 100644 include/hw/i386/remote.h

diff --git a/MAINTAINERS b/MAINTAINERS
index bb1d701..12945ee 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3102,6 +3102,8 @@ M: John G Johnson <john.g.johnson@oracle.com>
 S: Maintained
 F: hw/pci-host/remote.c
 F: include/hw/pci-host/remote.h
+F: hw/i386/remote.c
+F: include/hw/i386/remote.h
 
 Build and test automation
 -------------------------
diff --git a/hw/i386/meson.build b/hw/i386/meson.build
index e5d109f..d9d4f40 100644
--- a/hw/i386/meson.build
+++ b/hw/i386/meson.build
@@ -24,6 +24,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
   'pc_sysfw.c',
   'acpi-build.c',
   'port92.c'))
+i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
 
 subdir('kvm')
 subdir('xen')
diff --git a/hw/i386/remote.c b/hw/i386/remote.c
new file mode 100644
index 0000000..a67be33
--- /dev/null
+++ b/hw/i386/remote.c
@@ -0,0 +1,69 @@
+/*
+ * Machine for remote device
+ *
+ *  This machine type is used by the remote device process in multi-process
+ *  QEMU. QEMU device models depend on parent busses, interrupt controllers,
+ *  memory regions, etc. The remote machine type offers this environment so
+ *  that QEMU device models can be used as remote devices.
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/i386/remote.h"
+#include "exec/address-spaces.h"
+#include "exec/memory.h"
+#include "qapi/error.h"
+
+static void remote_machine_init(MachineState *machine)
+{
+    MemoryRegion *system_memory, *system_io, *pci_memory;
+    RemoteMachineState *s = REMOTE_MACHINE(machine);
+    RemotePCIHost *rem_host;
+
+    system_memory = get_system_memory();
+    system_io = get_system_io();
+
+    pci_memory = g_new(MemoryRegion, 1);
+    memory_region_init(pci_memory, NULL, "pci", UINT64_MAX);
+
+    rem_host = REMOTE_HOST_DEVICE(qdev_new(TYPE_REMOTE_HOST_DEVICE));
+
+    rem_host->mr_pci_mem = pci_memory;
+    rem_host->mr_sys_mem = system_memory;
+    rem_host->mr_sys_io = system_io;
+
+    s->host = rem_host;
+
+    object_property_add_child(OBJECT(s), "remote-device", OBJECT(rem_host));
+    memory_region_add_subregion_overlap(system_memory, 0x0, pci_memory, -1);
+
+    qdev_realize(DEVICE(rem_host), sysbus_get_default(), &error_fatal);
+}
+
+static void remote_machine_class_init(ObjectClass *oc, void *data)
+{
+    MachineClass *mc = MACHINE_CLASS(oc);
+
+    mc->init = remote_machine_init;
+}
+
+static const TypeInfo remote_machine = {
+    .name = TYPE_REMOTE_MACHINE,
+    .parent = TYPE_MACHINE,
+    .instance_size = sizeof(RemoteMachineState),
+    .class_init = remote_machine_class_init,
+};
+
+static void remote_machine_register_types(void)
+{
+    type_register_static(&remote_machine);
+}
+
+type_init(remote_machine_register_types);
diff --git a/include/hw/i386/remote.h b/include/hw/i386/remote.h
new file mode 100644
index 0000000..d312972
--- /dev/null
+++ b/include/hw/i386/remote.h
@@ -0,0 +1,28 @@
+/*
+ * Remote machine configuration
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef REMOTE_MACHINE_H
+#define REMOTE_MACHINE_H
+
+#include "qom/object.h"
+#include "hw/boards.h"
+#include "hw/pci-host/remote.h"
+
+typedef struct RemoteMachineState {
+    MachineState parent_obj;
+
+    RemotePCIHost *host;
+} RemoteMachineState;
+
+#define TYPE_REMOTE_MACHINE "x-remote-machine"
+#define REMOTE_MACHINE(obj) \
+    OBJECT_CHECK(RemoteMachineState, (obj), TYPE_REMOTE_MACHINE)
+
+#endif
diff --git a/include/hw/pci-host/remote.h b/include/hw/pci-host/remote.h
index bab6d3c..cc0fff4 100644
--- a/include/hw/pci-host/remote.h
+++ b/include/hw/pci-host/remote.h
@@ -25,6 +25,7 @@ typedef struct RemotePCIHost {
 
     MemoryRegion *mr_pci_mem;
     MemoryRegion *mr_sys_io;
+    MemoryRegion *mr_sys_mem;
 } RemotePCIHost;
 
 #endif
-- 
1.8.3.1

[PATCH v11 05/19] multi-process: add qio channel function to transmit

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

The entire array of the memory regions and file handlers.
Will be used in the next patch.

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 include/io/channel.h | 24 ++++++++++++++++++++++++
 io/channel.c         | 45 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 69 insertions(+)

diff --git a/include/io/channel.h b/include/io/channel.h
index 3c04f0e..6807506 100644
--- a/include/io/channel.h
+++ b/include/io/channel.h
@@ -773,5 +773,29 @@ void qio_channel_set_aio_fd_handler(QIOChannel *ioc,
                                     IOHandler *io_read,
                                     IOHandler *io_write,
                                     void *opaque);
+/**
+ * qio_channel_writev_full_all:
+ * @ioc: the channel object
+ * @iov: the array of memory regions to write data from
+ * @niov: the length of the @iov array
+ * @fds: an array of file handles to send
+ * @nfds: number of file handles in @fds
+ * @errp: pointer to a NULL-initialized error object
+ *
+ *
+ * Behaves like qio_channel_writev_full but will attempt
+ * to send all data passed (file handles and memory regions).
+ * The function will wait for all requested data
+ * to be written, yielding from the current coroutine
+ * if required.
+ *
+ * Returns: 0 if all bytes were written, or -1 on error
+ */
+
+int qio_channel_writev_full_all(QIOChannel *ioc,
+                           const struct iovec *iov,
+                           size_t niov,
+                           int *fds, size_t nfds,
+                           Error **errp);
 
 #endif /* QIO_CHANNEL_H */
diff --git a/io/channel.c b/io/channel.c
index e4376eb..22c10c5 100644
--- a/io/channel.c
+++ b/io/channel.c
@@ -190,6 +190,51 @@ int qio_channel_writev_all(QIOChannel *ioc,
     return ret;
 }
 
+int qio_channel_writev_full_all(QIOChannel *ioc,
+                                const struct iovec *iov,
+                                size_t niov,
+                                int *fds, size_t nfds,
+                                Error **errp)
+{
+    int ret = -1;
+    struct iovec *local_iov = g_new(struct iovec, niov);
+    struct iovec *local_iov_head = local_iov;
+    unsigned int nlocal_iov = niov;
+
+    nlocal_iov = iov_copy(local_iov, nlocal_iov,
+                          iov, niov,
+                          0, iov_size(iov, niov));
+
+    while (nlocal_iov > 0) {
+        ssize_t len;
+        len = qio_channel_writev_full(ioc, local_iov, nlocal_iov, fds,
+                                      nfds, errp);
+        if (len == QIO_CHANNEL_ERR_BLOCK) {
+            if (qemu_in_coroutine()) {
+                qio_channel_yield(ioc, G_IO_OUT);
+            } else {
+                qio_channel_wait(ioc, G_IO_OUT);
+            }
+            continue;
+        }
+        if (len < 0) {
+            goto cleanup;
+        }
+
+        iov_discard_front(&local_iov, &nlocal_iov, len);
+
+        if (len > 0) {
+            fds = NULL;
+            nfds = 0;
+        }
+    }
+
+    ret = 0;
+ cleanup:
+    g_free(local_iov_head);
+    return ret;
+}
+
 ssize_t qio_channel_readv(QIOChannel *ioc,
                           const struct iovec *iov,
                           size_t niov,
-- 
1.8.3.1

[PATCH v11 06/19] multi-process: define MPQemuMsg format and transmission functions

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

Defines MPQemuMsg, which is the message that is sent to the remote
process. This message is sent over QIOChannel and is used to
command the remote process to perform various tasks.
Define transmission functions used by proxy and by remote.
There are certain restrictions on where its safe to use these
functions:
  - From main loop in co-routine context. Will block the main loop if not in
    co-routine context;
  - From vCPU thread with no co-routine context and if the channel is not part
    of the main loop handling;
  - From IOThread within co-routine context, outside of co-routine context will
    block IOThread;

Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
---
 MAINTAINERS              |   2 +
 include/io/mpqemu-link.h |  60 ++++++++++++
 io/meson.build           |   2 +
 io/mpqemu-link.c         | 237 +++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 301 insertions(+)
 create mode 100644 include/io/mpqemu-link.h
 create mode 100644 io/mpqemu-link.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 12945ee..42b5da5 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3104,6 +3104,8 @@ F: hw/pci-host/remote.c
 F: include/hw/pci-host/remote.h
 F: hw/i386/remote.c
 F: include/hw/i386/remote.h
+F: io/mpqemu-link.c
+F: include/io/mpqemu-link.h
 
 Build and test automation
 -------------------------
diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
new file mode 100644
index 0000000..3505e06
--- /dev/null
+++ b/include/io/mpqemu-link.h
@@ -0,0 +1,60 @@
+/*
+ * Communication channel between QEMU and remote device process
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef MPQEMU_LINK_H
+#define MPQEMU_LINK_H
+
+#include "qom/object.h"
+#include "qemu/thread.h"
+#include "io/channel.h"
+
+#define REMOTE_MAX_FDS 8
+
+#define MPQEMU_MSG_HDR_SIZE offsetof(MPQemuMsg, data.u64)
+
+/**
+ * MPQemuCmd:
+ *
+ * MPQemuCmd enum type to specify the command to be executed on the remote
+ * device.
+ */
+typedef enum {
+    MPQEMU_CMD_INIT,
+    MPQEMU_CMD_MAX,
+} MPQemuCmd;
+
+/**
+ * MPQemuMsg:
+ * @cmd: The remote command
+ * @size: Size of the data to be shared
+ * @data: Structured data
+ * @fds: File descriptors to be shared with remote device
+ *
+ * MPQemuMsg Format of the message sent to the remote device from QEMU.
+ *
+ */
+typedef struct {
+    int cmd;
+    size_t size;
+
+    union {
+        uint64_t u64;
+    } data;
+
+    int fds[REMOTE_MAX_FDS];
+    int num_fds;
+} QEMU_PACKED MPQemuMsg;
+
+void mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp);
+void mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp);
+
+bool mpqemu_msg_valid(MPQemuMsg *msg);
+
+#endif
diff --git a/io/meson.build b/io/meson.build
index bcd8b1e..d25529c 100644
--- a/io/meson.build
+++ b/io/meson.build
@@ -13,3 +13,5 @@ io_ss.add(files(
   'net-listener.c',
   'task.c',
 ))
+
+io_ss.add(when: 'CONFIG_MPQEMU', if_true: files('mpqemu-link.c'))
diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
new file mode 100644
index 0000000..d646aad
--- /dev/null
+++ b/io/mpqemu-link.c
@@ -0,0 +1,237 @@
+/*
+ * Communication channel between QEMU and remote device process
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "qemu/module.h"
+#include "io/mpqemu-link.h"
+#include "qapi/error.h"
+#include "qemu/iov.h"
+#include "qemu/error-report.h"
+#include "qemu/main-loop.h"
+
+/*
+ * Send message over the ioc QIOChannel.
+ * This function is safe to call from:
+ * - From main loop in co-routine context. Will block the main loop if not in
+ *   co-routine context;
+ * - From vCPU thread with no co-routine context and if the channel is not part
+ *   of the main loop handling;
+ * - From IOThread within co-routine context, outside of co-routine context
+ *   will block IOThread;
+ */
+void mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp)
+{
+    bool iolock = qemu_mutex_iothread_locked();
+    bool iothread = qemu_get_current_aio_context() == qemu_get_aio_context() ?
+                    false : true;
+    Error *local_err = NULL;
+    struct iovec send[2] =  {0};
+    int *fds = NULL;
+    size_t nfds = 0;
+
+    send[0].iov_base = msg;
+    send[0].iov_len = MPQEMU_MSG_HDR_SIZE;
+
+    send[1].iov_base = (void *)&msg->data;
+    send[1].iov_len = msg->size;
+
+    if (msg->num_fds) {
+        nfds = msg->num_fds;
+        fds = msg->fds;
+    }
+    /*
+     * Dont use in IOThread out of co-routine context as
+     * it will block IOThread.
+     */
+    if (iothread) {
+        assert(qemu_in_coroutine());
+    }
+    /*
+     * Skip unlocking/locking iothread when in IOThread running
+     * in co-routine context. Co-routine context is asserted above
+     * for IOThread case.
+     * Also skip this while in a co-routine in the main context.
+     */
+    if (iolock && !iothread && !qemu_in_coroutine()) {
+        qemu_mutex_unlock_iothread();
+    }
+
+    (void)qio_channel_writev_full_all(ioc, send, G_N_ELEMENTS(send), fds, nfds,
+                                      &local_err);
+
+    if (iolock && !iothread && !qemu_in_coroutine()) {
+        /* See above comment why skip locking here. */
+        qemu_mutex_lock_iothread();
+    }
+
+    if (errp) {
+        error_propagate(errp, local_err);
+    } else if (local_err) {
+        error_report_err(local_err);
+    }
+
+    return;
+}
+
+/*
+ * Read message from the ioc QIOChannel.
+ * This function is safe to call from:
+ * - From main loop in co-routine context. Will block the main loop if not in
+ *   co-routine context;
+ * - From vCPU thread with no co-routine context and if the channel is not part
+ *   of the main loop handling;
+ * - From IOThread within co-routine context, outside of co-routine context
+ *   will block IOThread;
+ */
+static ssize_t mpqemu_read(QIOChannel *ioc, void *buf, size_t len, int **fds,
+                           size_t *nfds, Error **errp)
+{
+    struct iovec iov = { .iov_base = buf, .iov_len = len };
+    bool iolock = qemu_mutex_iothread_locked();
+    bool iothread = qemu_get_current_aio_context() == qemu_get_aio_context()
+                        ? false : true;
+    struct iovec *iovp = &iov;
+    Error *local_err = NULL;
+    unsigned int niov = 1;
+    size_t *l_nfds = nfds;
+    int **l_fds = fds;
+    ssize_t bytes = 0;
+    size_t size;
+
+    size = iov.iov_len;
+
+    /*
+     * Dont use in IOThread out of co-routine context as
+     * it will block IOThread.
+     */
+    if (iothread) {
+        assert(qemu_in_coroutine());
+    }
+
+    while (size > 0) {
+        bytes = qio_channel_readv_full(ioc, iovp, niov, l_fds, l_nfds,
+                                       &local_err);
+        if (bytes == QIO_CHANNEL_ERR_BLOCK) {
+            /*
+             * Skip unlocking/locking iothread when in IOThread running
+             * in co-routine context. Co-routine context is asserted above
+             * for IOThread case.
+             * Also skip this while in a co-routine in the main context.
+             */
+            if (iolock && !iothread && !qemu_in_coroutine()) {
+                qemu_mutex_unlock_iothread();
+            }
+            if (qemu_in_coroutine()) {
+                qio_channel_yield(ioc, G_IO_IN);
+            } else {
+                qio_channel_wait(ioc, G_IO_IN);
+            }
+            /* See above comment why skip locking here. */
+            if (iolock && !iothread && !qemu_in_coroutine()) {
+                qemu_mutex_lock_iothread();
+            }
+            continue;
+        }
+
+        if (bytes <= 0) {
+            error_propagate(errp, local_err);
+            return -EIO;
+        }
+
+        l_fds = NULL;
+        l_nfds = NULL;
+
+        size -= bytes;
+
+        (void)iov_discard_front(&iovp, &niov, bytes);
+    }
+
+    return len - size;
+}
+
+void mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp)
+{
+    Error *local_err = NULL;
+    int *fds = NULL;
+    size_t nfds = 0;
+    ssize_t len;
+
+    len = mpqemu_read(ioc, (void *)msg, MPQEMU_MSG_HDR_SIZE, &fds, &nfds,
+                      &local_err);
+    if (!local_err) {
+        if (len == -EIO) {
+            error_setg(&local_err, "Connection closed.");
+            goto fail;
+        }
+        if (len < 0) {
+            error_setg(&local_err, "Message length is less than 0");
+            goto fail;
+        }
+        if (len != MPQEMU_MSG_HDR_SIZE) {
+            error_setg(&local_err, "Message header corrupted");
+            goto fail;
+        }
+    } else {
+        goto fail;
+    }
+
+    if (msg->size > sizeof(msg->data)) {
+        error_setg(&local_err, "Invalid size for message");
+        goto fail;
+    }
+
+    if (mpqemu_read(ioc, (void *)&msg->data, msg->size, NULL, NULL,
+                    &local_err) < 0) {
+        goto fail;
+    }
+
+    msg->num_fds = nfds;
+    if (nfds) {
+        memcpy(msg->fds, fds, nfds * sizeof(int));
+    }
+
+fail:
+    while (local_err && nfds) {
+        close(fds[nfds - 1]);
+        nfds--;
+    }
+
+    g_free(fds);
+
+    if (errp) {
+        error_propagate(errp, local_err);
+    } else if (local_err) {
+        error_report_err(local_err);
+    }
+}
+
+bool mpqemu_msg_valid(MPQemuMsg *msg)
+{
+    if (msg->cmd >= MPQEMU_CMD_MAX && msg->cmd < 0) {
+        return false;
+    }
+
+    /* Verify FDs. */
+    if (msg->num_fds >= REMOTE_MAX_FDS) {
+        return false;
+    }
+
+    if (msg->num_fds > 0) {
+        for (int i = 0; i < msg->num_fds; i++) {
+            if (fcntl(msg->fds[i], F_GETFL) == -1) {
+                return false;
+            }
+        }
+    }
+
+    return true;
+}
-- 
1.8.3.1

[PATCH v11 07/19] multi-process: Initialize message handler in remote device

[Jagannathan Raman]

Initializes the message handler function in the remote process. It is
called whenever there's an event pending on QIOChannel that registers
this function.

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS              |  1 +
 hw/i386/meson.build      |  1 +
 hw/i386/remote-msg.c     | 61 ++++++++++++++++++++++++++++++++++++++++++++++++
 include/hw/i386/remote.h |  9 +++++++
 4 files changed, 72 insertions(+)
 create mode 100644 hw/i386/remote-msg.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 42b5da5..7314536 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3106,6 +3106,7 @@ F: hw/i386/remote.c
 F: include/hw/i386/remote.h
 F: io/mpqemu-link.c
 F: include/io/mpqemu-link.h
+F: hw/i386/remote-msg.c
 
 Build and test automation
 -------------------------
diff --git a/hw/i386/meson.build b/hw/i386/meson.build
index d9d4f40..3a2aa6e 100644
--- a/hw/i386/meson.build
+++ b/hw/i386/meson.build
@@ -25,6 +25,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
   'acpi-build.c',
   'port92.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
+i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-msg.c'))
 
 subdir('kvm')
 subdir('xen')
diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
new file mode 100644
index 0000000..83fd7bd
--- /dev/null
+++ b/hw/i386/remote-msg.c
@@ -0,0 +1,61 @@
+/*
+ * Copyright © 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL-v2, version 2 or later.
+ *
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/i386/remote.h"
+#include "io/channel.h"
+#include "io/mpqemu-link.h"
+#include "qapi/error.h"
+#include "sysemu/runstate.h"
+
+void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
+{
+    RemoteCommDev *com = (RemoteCommDev *)data;
+    PCIDevice *pci_dev = NULL;
+
+    pci_dev = com->dev;
+    for (;;) {
+        MPQemuMsg msg = {0};
+        Error *local_err = NULL;
+
+        if (!com->ioc) {
+            error_report("ERROR: No channel available");
+            break;
+        }
+        mpqemu_msg_recv(&msg, com->ioc, &local_err);
+        if (local_err) {
+            error_report_err(local_err);
+            break;
+        }
+
+        if (!mpqemu_msg_valid(&msg)) {
+            error_report("Received invalid message from proxy"
+                         "in remote process pid=%d", getpid());
+            break;
+        }
+
+        switch (msg.cmd) {
+        default:
+            error_setg(&local_err,
+                       "Unknown command (%d) received for device %s (pid=%d)",
+                       msg.cmd, DEVICE(pci_dev)->id, getpid());
+        }
+
+        if (local_err) {
+            error_report_err(local_err);
+            qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
+            break;
+        }
+    }
+    qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
+
+    return;
+}
diff --git a/include/hw/i386/remote.h b/include/hw/i386/remote.h
index d312972..3073db6 100644
--- a/include/hw/i386/remote.h
+++ b/include/hw/i386/remote.h
@@ -14,6 +14,7 @@
 #include "qom/object.h"
 #include "hw/boards.h"
 #include "hw/pci-host/remote.h"
+#include "io/channel.h"
 
 typedef struct RemoteMachineState {
     MachineState parent_obj;
@@ -21,8 +22,16 @@ typedef struct RemoteMachineState {
     RemotePCIHost *host;
 } RemoteMachineState;
 
+/* Used to pass to co-routine device and ioc. */
+typedef struct RemoteCommDev {
+    PCIDevice *dev;
+    QIOChannel *ioc;
+} RemoteCommDev;
+
 #define TYPE_REMOTE_MACHINE "x-remote-machine"
 #define REMOTE_MACHINE(obj) \
     OBJECT_CHECK(RemoteMachineState, (obj), TYPE_REMOTE_MACHINE)
 
+void coroutine_fn mpqemu_remote_msg_loop_co(void *data);
+
 #endif
-- 
1.8.3.1

[PATCH v11 08/19] multi-process: Associate fd of a PCIDevice with its object

[Jagannathan Raman]

Associate the file descriptor for a PCIDevice in remote process with
DeviceState object.

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
---
 MAINTAINERS                  |   2 +
 hw/i386/meson.build          |   1 +
 hw/i386/remote-msg.c         |   1 +
 hw/i386/remote-obj.c         | 154 +++++++++++++++++++++++++++++++++++++++++++
 include/hw/i386/remote-obj.h |  42 ++++++++++++
 5 files changed, 200 insertions(+)
 create mode 100644 hw/i386/remote-obj.c
 create mode 100644 include/hw/i386/remote-obj.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 7314536..487ab13 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3107,6 +3107,8 @@ F: include/hw/i386/remote.h
 F: io/mpqemu-link.c
 F: include/io/mpqemu-link.h
 F: hw/i386/remote-msg.c
+F: include/hw/i386/remote-obj.h
+F: hw/i386/remote-obj.c
 
 Build and test automation
 -------------------------
diff --git a/hw/i386/meson.build b/hw/i386/meson.build
index 3a2aa6e..9735f6b 100644
--- a/hw/i386/meson.build
+++ b/hw/i386/meson.build
@@ -26,6 +26,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
   'port92.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-msg.c'))
+i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-obj.c'))
 
 subdir('kvm')
 subdir('xen')
diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
index 83fd7bd..6451b77 100644
--- a/hw/i386/remote-msg.c
+++ b/hw/i386/remote-msg.c
@@ -56,6 +56,7 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
         }
     }
     qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
+    g_free(com);
 
     return;
 }
diff --git a/hw/i386/remote-obj.c b/hw/i386/remote-obj.c
new file mode 100644
index 0000000..a9a732a
--- /dev/null
+++ b/hw/i386/remote-obj.c
@@ -0,0 +1,154 @@
+/*
+ * Copyright © 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL-v2, version 2 or later.
+ *
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/i386/remote-obj.h"
+#include "qemu/error-report.h"
+#include "qom/object_interfaces.h"
+#include "hw/qdev-core.h"
+#include "io/channel.h"
+#include "hw/qdev-core.h"
+#include "hw/i386/remote.h"
+#include "io/channel-util.h"
+#include "qapi/error.h"
+#include "sysemu/sysemu.h"
+#include "hw/pci/pci.h"
+
+static void remote_object_set_fd(Object *obj, const char *str, Error **errp)
+{
+    RemoteObject *o = REMOTE_OBJECT(obj);
+
+    o->fd = atoi(str);
+}
+
+static void remote_object_set_devid(Object *obj, const char *str, Error **errp)
+{
+    RemoteObject *o = REMOTE_OBJECT(obj);
+
+    g_free(o->devid);
+
+    o->devid = g_strdup(str);
+}
+
+static void property_release_remote_object(Object *obj, const char *name,
+                                           void *opaque)
+{
+    Object *remote_object = OBJECT(opaque);
+
+    object_unref(remote_object);
+}
+
+static void remote_object_machine_done(Notifier *notifier, void *data)
+{
+    RemoteObject *o = container_of(notifier, RemoteObject, machine_done);
+    DeviceState *dev = NULL;
+    QIOChannel *ioc = NULL;
+    Coroutine *co = NULL;
+    RemoteCommDev *comdev = NULL;
+    Error *err = NULL;
+
+    dev = qdev_find_recursive(sysbus_get_default(), o->devid);
+    if (!dev || !object_dynamic_cast(OBJECT(dev), TYPE_PCI_DEVICE)) {
+        error_report("%s is not a PCI device", o->devid);
+        return;
+    }
+
+    ioc = qio_channel_new_fd(o->fd, &err);
+    if (!ioc) {
+        error_report_err(err);
+        return;
+    }
+    qio_channel_set_blocking(ioc, false, NULL);
+
+    object_property_add(OBJECT(dev), "remote-object", "object", NULL, NULL,
+                        property_release_remote_object, (void *)OBJECT(o));
+    /* co-routine should free this. */
+    comdev = g_new0(RemoteCommDev, 1);
+    *comdev = (RemoteCommDev) {
+        .ioc = ioc,
+        .dev = PCI_DEVICE(dev),
+    };
+
+    co = qemu_coroutine_create(mpqemu_remote_msg_loop_co, comdev);
+    qemu_coroutine_enter(co);
+}
+
+static void remote_object_init(Object *obj)
+{
+    RemoteObjectClass *k = REMOTE_OBJECT_GET_CLASS(obj);
+    RemoteObject *o = REMOTE_OBJECT(obj);
+
+    if (k->nr_devs >= k->max_devs) {
+        error_report("Reached maximum number of devices: %u", k->max_devs);
+        return;
+    }
+
+    o->ioc = NULL;
+    o->fd = -1;
+    o->devid = NULL;
+
+    k->nr_devs++;
+
+    object_property_add_str(obj, "fd", NULL, remote_object_set_fd);
+    object_property_set_description(obj, "fd",
+                                    "file descriptor for the object");
+    object_property_add_str(obj, "devid", NULL, remote_object_set_devid);
+    object_property_set_description(obj, "devid",
+                                    "id of device to associate");
+
+    o->machine_done.notify = remote_object_machine_done;
+    qemu_add_machine_init_done_notifier(&o->machine_done);
+}
+
+static void remote_object_finalize(Object *obj)
+{
+    RemoteObjectClass *k = REMOTE_OBJECT_GET_CLASS(obj);
+    RemoteObject *o = REMOTE_OBJECT(obj);
+
+    if (o->ioc) {
+        qio_channel_shutdown(o->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
+        qio_channel_close(o->ioc, NULL);
+    }
+
+    object_unref(OBJECT(o->ioc));
+
+    k->nr_devs--;
+    g_free(o->devid);
+}
+
+static void remote_object_class_init(ObjectClass *klass, void *data)
+{
+    RemoteObjectClass *k = REMOTE_OBJECT_CLASS(klass);
+
+    k->max_devs = 1;
+    k->nr_devs = 0;
+}
+
+static const TypeInfo remote_object_info = {
+    .name = TYPE_REMOTE_OBJECT,
+    .parent = TYPE_OBJECT,
+    .instance_size = sizeof(RemoteObject),
+    .instance_init = remote_object_init,
+    .instance_finalize = remote_object_finalize,
+    .class_size = sizeof(RemoteObjectClass),
+    .class_init = remote_object_class_init,
+    .interfaces = (InterfaceInfo[]) {
+        { TYPE_USER_CREATABLE },
+        { }
+    }
+};
+
+static void register_types(void)
+{
+    type_register_static(&remote_object_info);
+}
+
+type_init(register_types);
diff --git a/include/hw/i386/remote-obj.h b/include/hw/i386/remote-obj.h
new file mode 100644
index 0000000..0e95813
--- /dev/null
+++ b/include/hw/i386/remote-obj.h
@@ -0,0 +1,42 @@
+/*
+ * Copyright © 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL-v2, version 2 or later.
+ *
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef REMOTE_OBJECT_H
+#define REMOTE_OBJECT_H
+
+#include "io/channel.h"
+#include "qemu/notify.h"
+
+#define TYPE_REMOTE_OBJECT "x-remote-object"
+#define REMOTE_OBJECT(obj) \
+    OBJECT_CHECK(RemoteObject, (obj), TYPE_REMOTE_OBJECT)
+#define REMOTE_OBJECT_GET_CLASS(obj) \
+    OBJECT_GET_CLASS(RemoteObjectClass, (obj), TYPE_REMOTE_OBJECT)
+#define REMOTE_OBJECT_CLASS(klass) \
+    OBJECT_CLASS_CHECK(RemoteObjectClass, (klass), TYPE_REMOTE_OBJECT)
+
+typedef struct {
+    ObjectClass parent_class;
+
+    unsigned int nr_devs;
+    unsigned int max_devs;
+} RemoteObjectClass;
+
+typedef struct {
+    /* private */
+    Object parent;
+
+    Notifier machine_done;
+
+    int fd;
+    char *devid;
+    QIOChannel *ioc;
+} RemoteObject;
+
+#endif
-- 
1.8.3.1

[PATCH v11 09/19] multi-process: setup memory manager for remote device

[Jagannathan Raman]

SyncSysMemMsg message format is defined. It is used to send
file descriptors of the RAM regions to remote device.
RAM on the remote device is configured with a set of file descriptors.
Old RAM regions are deleted and new regions, each with an fd, is
added to the RAM.

Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                     |  2 ++
 hw/i386/meson.build             |  1 +
 hw/i386/remote-memory.c         | 58 +++++++++++++++++++++++++++++++++++++++++
 include/hw/i386/remote-memory.h | 19 ++++++++++++++
 include/io/mpqemu-link.h        | 13 +++++++++
 io/mpqemu-link.c                | 11 ++++++++
 6 files changed, 104 insertions(+)
 create mode 100644 hw/i386/remote-memory.c
 create mode 100644 include/hw/i386/remote-memory.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 487ab13..4aa00ec 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3109,6 +3109,8 @@ F: include/io/mpqemu-link.h
 F: hw/i386/remote-msg.c
 F: include/hw/i386/remote-obj.h
 F: hw/i386/remote-obj.c
+F: include/hw/i386/remote-memory.h
+F: hw/i386/remote-memory.c
 
 Build and test automation
 -------------------------
diff --git a/hw/i386/meson.build b/hw/i386/meson.build
index 9735f6b..b698b36 100644
--- a/hw/i386/meson.build
+++ b/hw/i386/meson.build
@@ -27,6 +27,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-msg.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-obj.c'))
+i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-memory.c'))
 
 subdir('kvm')
 subdir('xen')
diff --git a/hw/i386/remote-memory.c b/hw/i386/remote-memory.c
new file mode 100644
index 0000000..27bc8a7
--- /dev/null
+++ b/hw/i386/remote-memory.c
@@ -0,0 +1,58 @@
+/*
+ * Memory manager for remote device
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/i386/remote-memory.h"
+#include "exec/address-spaces.h"
+#include "exec/ram_addr.h"
+#include "qapi/error.h"
+
+void remote_sysmem_reconfig(MPQemuMsg *msg, Error **errp)
+{
+    SyncSysmemMsg *sysmem_info = &msg->data.sync_sysmem;
+    MemoryRegion *sysmem, *subregion, *next;
+    static unsigned int suffix;
+    Error *local_err = NULL;
+    char *name;
+    int region;
+
+    sysmem = get_system_memory();
+
+    memory_region_transaction_begin();
+
+    QTAILQ_FOREACH_SAFE(subregion, &sysmem->subregions, subregions_link, next) {
+        if (subregion->ram) {
+            memory_region_del_subregion(sysmem, subregion);
+            object_unparent(OBJECT(subregion));
+        }
+    }
+
+    for (region = 0; region < msg->num_fds; region++) {
+        subregion = g_new(MemoryRegion, 1);
+        name = g_strdup_printf("remote-mem-%u", suffix++);
+        memory_region_init_ram_from_fd(subregion, NULL,
+                                       name, sysmem_info->sizes[region],
+                                       RAM_SHARED, msg->fds[region],
+                                       sysmem_info->offsets[region],
+                                       &local_err);
+        g_free(name);
+        if (local_err) {
+            error_propagate(errp, local_err);
+            break;
+        }
+
+        memory_region_add_subregion(sysmem, sysmem_info->gpas[region],
+                                    subregion);
+    }
+
+    memory_region_transaction_commit();
+}
diff --git a/include/hw/i386/remote-memory.h b/include/hw/i386/remote-memory.h
new file mode 100644
index 0000000..aab3e4d
--- /dev/null
+++ b/include/hw/i386/remote-memory.h
@@ -0,0 +1,19 @@
+/*
+ * Memory manager for remote device
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef REMOTE_MEMORY_H
+#define REMOTE_MEMORY_H
+
+#include "exec/hwaddr.h"
+#include "io/mpqemu-link.h"
+
+void remote_sysmem_reconfig(MPQemuMsg *msg, Error **errp);
+
+#endif
diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
index 3505e06..e33181a 100644
--- a/include/io/mpqemu-link.h
+++ b/include/io/mpqemu-link.h
@@ -14,6 +14,7 @@
 #include "qom/object.h"
 #include "qemu/thread.h"
 #include "io/channel.h"
+#include "exec/hwaddr.h"
 
 #define REMOTE_MAX_FDS 8
 
@@ -24,12 +25,22 @@
  *
  * MPQemuCmd enum type to specify the command to be executed on the remote
  * device.
+ *
+ * SYNC_SYSMEM      Shares QEMU's RAM with remote device's RAM
  */
 typedef enum {
     MPQEMU_CMD_INIT,
+    SYNC_SYSMEM,
+    RET_MSG,
     MPQEMU_CMD_MAX,
 } MPQemuCmd;
 
+typedef struct {
+    hwaddr gpas[REMOTE_MAX_FDS];
+    uint64_t sizes[REMOTE_MAX_FDS];
+    off_t offsets[REMOTE_MAX_FDS];
+} SyncSysmemMsg;
+
 /**
  * MPQemuMsg:
  * @cmd: The remote command
@@ -40,12 +51,14 @@ typedef enum {
  * MPQemuMsg Format of the message sent to the remote device from QEMU.
  *
  */
+
 typedef struct {
     int cmd;
     size_t size;
 
     union {
         uint64_t u64;
+        SyncSysmemMsg sync_sysmem;
     } data;
 
     int fds[REMOTE_MAX_FDS];
diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
index d646aad..b30e27c 100644
--- a/io/mpqemu-link.c
+++ b/io/mpqemu-link.c
@@ -233,5 +233,16 @@ bool mpqemu_msg_valid(MPQemuMsg *msg)
         }
     }
 
+     /* Verify message specific fields. */
+    switch (msg->cmd) {
+    case SYNC_SYSMEM:
+        if (msg->num_fds == 0 || msg->size != sizeof(SyncSysmemMsg)) {
+            return false;
+        }
+        break;
+    default:
+        break;
+    }
+
     return true;
 }
-- 
1.8.3.1

[PATCH v11 10/19] multi-process: introduce proxy object

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

Defines a PCI Device proxy object as a child of TYPE_PCI_DEVICE.

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
---
 MAINTAINERS            |  2 ++
 hw/pci/meson.build     |  1 +
 hw/pci/proxy.c         | 98 ++++++++++++++++++++++++++++++++++++++++++++++++++
 include/hw/pci/proxy.h | 36 +++++++++++++++++++
 4 files changed, 137 insertions(+)
 create mode 100644 hw/pci/proxy.c
 create mode 100644 include/hw/pci/proxy.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 4aa00ec..202db86 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3111,6 +3111,8 @@ F: include/hw/i386/remote-obj.h
 F: hw/i386/remote-obj.c
 F: include/hw/i386/remote-memory.h
 F: hw/i386/remote-memory.c
+F: hw/pci/proxy.c
+F: include/hw/pci/proxy.h
 
 Build and test automation
 -------------------------
diff --git a/hw/pci/meson.build b/hw/pci/meson.build
index 5c4bbac..0df3017 100644
--- a/hw/pci/meson.build
+++ b/hw/pci/meson.build
@@ -12,6 +12,7 @@ pci_ss.add(files(
 # allow plugging PCIe devices into PCI buses, include them even if
 # CONFIG_PCI_EXPRESS=n.
 pci_ss.add(files('pcie.c', 'pcie_aer.c'))
+pci_ss.add(when: 'CONFIG_MPQEMU', if_true: files('proxy.c'))
 softmmu_ss.add(when: 'CONFIG_PCI_EXPRESS', if_true: files('pcie_port.c', 'pcie_host.c'))
 softmmu_ss.add_all(when: 'CONFIG_PCI', if_true: pci_ss)
 
diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
new file mode 100644
index 0000000..7a12f23
--- /dev/null
+++ b/hw/pci/proxy.c
@@ -0,0 +1,98 @@
+/*
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/pci/proxy.h"
+#include "hw/pci/pci.h"
+#include "qapi/error.h"
+#include "io/channel-util.h"
+#include "hw/qdev-properties.h"
+#include "monitor/monitor.h"
+#include "migration/blocker.h"
+
+static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
+{
+    pdev->ioc = qio_channel_new_fd(fd, errp);
+}
+
+static Property proxy_properties[] = {
+    DEFINE_PROP_STRING("fd", PCIProxyDev, fd),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pci_proxy_dev_realize(PCIDevice *device, Error **errp)
+{
+    PCIProxyDev *dev = PCI_PROXY_DEV(device);
+    int fd;
+
+    if (dev->fd) {
+        fd = monitor_fd_param(monitor_cur(), dev->fd, errp);
+        if (fd == -1) {
+            error_prepend(errp, "proxy: unable to parse fd: ");
+            return;
+        }
+        proxy_set_socket(dev, fd, errp);
+    } else {
+        error_setg(errp, "fd parameter not specified for %s",
+                   DEVICE(device)->id);
+        return;
+    }
+
+    error_setg(&dev->migration_blocker, "%s does not support migration",
+               TYPE_PCI_PROXY_DEV);
+    if (migrate_add_blocker(dev->migration_blocker, errp)) {
+        error_free(dev->migration_blocker);
+        error_free(*errp);
+        dev->migration_blocker = NULL;
+        error_setg(errp, "Failed to set migration blocker");
+    }
+
+    qemu_mutex_init(&dev->io_mutex);
+    qio_channel_set_blocking(dev->ioc, true, NULL);
+}
+
+static void pci_proxy_dev_exit(PCIDevice *pdev)
+{
+    PCIProxyDev *dev = PCI_PROXY_DEV(pdev);
+
+    qio_channel_close(dev->ioc, NULL);
+
+    migrate_del_blocker(dev->migration_blocker);
+
+    error_free(dev->migration_blocker);
+}
+
+static void pci_proxy_dev_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+
+    k->realize = pci_proxy_dev_realize;
+    k->exit = pci_proxy_dev_exit;
+    device_class_set_props(dc, proxy_properties);
+}
+
+static const TypeInfo pci_proxy_dev_type_info = {
+    .name          = TYPE_PCI_PROXY_DEV,
+    .parent        = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(PCIProxyDev),
+    .class_init    = pci_proxy_dev_class_init,
+    .interfaces = (InterfaceInfo[]) {
+        { INTERFACE_CONVENTIONAL_PCI_DEVICE },
+        { },
+    },
+};
+
+static void pci_proxy_dev_register_types(void)
+{
+    type_register_static(&pci_proxy_dev_type_info);
+}
+
+type_init(pci_proxy_dev_register_types)
diff --git a/include/hw/pci/proxy.h b/include/hw/pci/proxy.h
new file mode 100644
index 0000000..923432a
--- /dev/null
+++ b/include/hw/pci/proxy.h
@@ -0,0 +1,36 @@
+/*
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef PROXY_H
+#define PROXY_H
+
+#include "hw/pci/pci.h"
+#include "io/channel.h"
+
+#define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev"
+
+#define PCI_PROXY_DEV(obj) \
+            OBJECT_CHECK(PCIProxyDev, (obj), TYPE_PCI_PROXY_DEV)
+typedef struct PCIProxyDev PCIProxyDev;
+
+struct PCIProxyDev {
+    PCIDevice parent_dev;
+    char *fd;
+
+    /*
+     * Mutex used to protect the QIOChannel fd from
+     * the concurrent access by the VCPUs since proxy
+     * blocks while awaiting for the replies from the
+     * process remote.
+     */
+    QemuMutex io_mutex;
+    QIOChannel *ioc;
+    Error *migration_blocker;
+};
+
+#endif /* PROXY_H */
-- 
1.8.3.1

[PATCH v11 11/19] multi-process: add proxy communication functions

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 include/io/mpqemu-link.h |  4 ++++
 io/mpqemu-link.c         | 38 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 42 insertions(+)

diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
index e33181a..4bea5da 100644
--- a/include/io/mpqemu-link.h
+++ b/include/io/mpqemu-link.h
@@ -15,6 +15,8 @@
 #include "qemu/thread.h"
 #include "io/channel.h"
 #include "exec/hwaddr.h"
+#include "io/channel-socket.h"
+#include "hw/pci/proxy.h"
 
 #define REMOTE_MAX_FDS 8
 
@@ -65,6 +67,8 @@ typedef struct {
     int num_fds;
 } QEMU_PACKED MPQemuMsg;
 
+uint64_t mpqemu_msg_send_and_await_reply(MPQemuMsg *msg, PCIProxyDev *pdev,
+                                         Error **errp);
 void mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp);
 void mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp);
 
diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
index b30e27c..1339cc7 100644
--- a/io/mpqemu-link.c
+++ b/io/mpqemu-link.c
@@ -17,6 +17,7 @@
 #include "qemu/iov.h"
 #include "qemu/error-report.h"
 #include "qemu/main-loop.h"
+#include "io/channel.h"
 
 /*
  * Send message over the ioc QIOChannel.
@@ -214,6 +215,43 @@ fail:
     }
 }
 
+/*
+ * Called from VCPU thread in non-coroutine context.
+ */
+uint64_t mpqemu_msg_send_and_await_reply(MPQemuMsg *msg, PCIProxyDev *pdev,
+                                         Error **errp)
+{
+    MPQemuMsg msg_reply = {0};
+    uint64_t ret = UINT64_MAX;
+    Error *local_err = NULL;
+
+    qemu_mutex_lock(&pdev->io_mutex);
+    mpqemu_msg_send(msg, pdev->ioc, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        goto exit_send;
+    }
+
+    mpqemu_msg_recv(&msg_reply, pdev->ioc, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        goto exit_send;
+    }
+
+    if (!mpqemu_msg_valid(&msg_reply) || msg_reply.cmd != RET_MSG) {
+        error_setg(errp, "ERROR: Invalid reply received for command %d",
+                         msg->cmd);
+        goto exit_send;
+    } else {
+        ret = msg_reply.data.u64;
+    }
+
+ exit_send:
+    qemu_mutex_unlock(&pdev->io_mutex);
+
+    return ret;
+}
+
 bool mpqemu_msg_valid(MPQemuMsg *msg)
 {
     if (msg->cmd >= MPQEMU_CMD_MAX && msg->cmd < 0) {
-- 
1.8.3.1

[PATCH v11 12/19] multi-process: Forward PCI config space acceses to the remote process

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

The Proxy Object sends the PCI config space accesses as messages
to the remote process over the communication channel

TODO:
Investigate if the local PCI config writes can be dropped.
Without the proxy local PCI config space writes for the device,
the driver in the guest times out on the probing.
We have tried to only refer to the remote for the PCI config writes,
but the driver timeout in the guest forced as to left this
as it is (removing local PCI config only).

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 hw/i386/remote-msg.c     | 62 ++++++++++++++++++++++++++++++++++++++++++++++++
 hw/pci/proxy.c           | 50 ++++++++++++++++++++++++++++++++++++++
 include/io/mpqemu-link.h |  9 +++++++
 io/mpqemu-link.c         |  6 +++++
 4 files changed, 127 insertions(+)

diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
index 6451b77..94937db 100644
--- a/hw/i386/remote-msg.c
+++ b/hw/i386/remote-msg.c
@@ -15,6 +15,12 @@
 #include "io/mpqemu-link.h"
 #include "qapi/error.h"
 #include "sysemu/runstate.h"
+#include "hw/pci/pci.h"
+
+static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
+                                 MPQemuMsg *msg);
+static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
+                                MPQemuMsg *msg);
 
 void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
 {
@@ -43,6 +49,12 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
         }
 
         switch (msg.cmd) {
+        case PCI_CONFIG_WRITE:
+            process_config_write(com->ioc, pci_dev, &msg);
+            break;
+        case PCI_CONFIG_READ:
+            process_config_read(com->ioc, pci_dev, &msg);
+            break;
         default:
             error_setg(&local_err,
                        "Unknown command (%d) received for device %s (pid=%d)",
@@ -60,3 +72,53 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
 
     return;
 }
+
+static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
+                                 MPQemuMsg *msg)
+{
+    ConfDataMsg *conf = (ConfDataMsg *)&msg->data.conf_data;
+    MPQemuMsg ret = { 0 };
+    Error *local_err = NULL;
+
+    if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) {
+        error_report("Bad address received when writing PCI config, pid %d",
+                     getpid());
+        ret.data.u64 = UINT64_MAX;
+    } else {
+        pci_default_write_config(dev, conf->addr, conf->val, conf->l);
+    }
+
+    ret.cmd = RET_MSG;
+    ret.size = sizeof(ret.data.u64);
+
+    mpqemu_msg_send(&ret, ioc, &local_err);
+    if (local_err) {
+        error_report("Could not send message to proxy from pid %d",
+                     getpid());
+    }
+}
+
+static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
+                                MPQemuMsg *msg)
+{
+    ConfDataMsg *conf = (ConfDataMsg *)&msg->data.conf_data;
+    MPQemuMsg ret = { 0 };
+    Error *local_err = NULL;
+
+    if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) {
+        error_report("Bad address received when reading PCI config, pid %d",
+                     getpid());
+        ret.data.u64 = UINT64_MAX;
+    } else {
+        ret.data.u64 = pci_default_read_config(dev, conf->addr, conf->l);
+    }
+
+    ret.cmd = RET_MSG;
+    ret.size = sizeof(ret.data.u64);
+
+    mpqemu_msg_send(&ret, ioc, &local_err);
+    if (local_err) {
+        error_report("Could not send message to proxy from pid %d",
+                     getpid());
+    }
+}
diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
index 7a12f23..27714fe 100644
--- a/hw/pci/proxy.c
+++ b/hw/pci/proxy.c
@@ -16,6 +16,8 @@
 #include "hw/qdev-properties.h"
 #include "monitor/monitor.h"
 #include "migration/blocker.h"
+#include "io/mpqemu-link.h"
+#include "qemu/error-report.h"
 
 static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
 {
@@ -69,6 +71,51 @@ static void pci_proxy_dev_exit(PCIDevice *pdev)
     error_free(dev->migration_blocker);
 }
 
+static int config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val,
+                          int l, unsigned int op)
+{
+    MPQemuMsg msg = { 0 };
+    uint64_t ret = -EINVAL;
+    Error *local_err = NULL;
+
+    msg.cmd = op;
+    msg.data.conf_data.addr = addr;
+    msg.data.conf_data.val = (op == PCI_CONFIG_WRITE) ? *val : 0;
+    msg.data.conf_data.l = l;
+    msg.size = sizeof(ConfDataMsg);
+
+    ret = mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err);
+    if (local_err) {
+        error_report_err(local_err);
+    }
+    if (op == PCI_CONFIG_READ) {
+        *val = (uint32_t)ret;
+    }
+
+    return ret;
+}
+
+static uint32_t pci_proxy_read_config(PCIDevice *d, uint32_t addr, int len)
+{
+    uint32_t val;
+
+    (void)config_op_send(PCI_PROXY_DEV(d), addr, &val, len, PCI_CONFIG_READ);
+
+    return val;
+}
+
+static void pci_proxy_write_config(PCIDevice *d, uint32_t addr, uint32_t val,
+                                   int l)
+{
+    /*
+     * Some of the functions access the copy of the remote device
+     * PCI config space, therefore maintain it updated.
+     */
+    pci_default_write_config(d, addr, val, l);
+
+    (void)config_op_send(PCI_PROXY_DEV(d), addr, &val, l, PCI_CONFIG_WRITE);
+}
+
 static void pci_proxy_dev_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
@@ -76,6 +123,9 @@ static void pci_proxy_dev_class_init(ObjectClass *klass, void *data)
 
     k->realize = pci_proxy_dev_realize;
     k->exit = pci_proxy_dev_exit;
+    k->config_read = pci_proxy_read_config;
+    k->config_write = pci_proxy_write_config;
+
     device_class_set_props(dc, proxy_properties);
 }
 
diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
index 4bea5da..459d345 100644
--- a/include/io/mpqemu-link.h
+++ b/include/io/mpqemu-link.h
@@ -34,6 +34,8 @@ typedef enum {
     MPQEMU_CMD_INIT,
     SYNC_SYSMEM,
     RET_MSG,
+    PCI_CONFIG_WRITE,
+    PCI_CONFIG_READ,
     MPQEMU_CMD_MAX,
 } MPQemuCmd;
 
@@ -43,6 +45,12 @@ typedef struct {
     off_t offsets[REMOTE_MAX_FDS];
 } SyncSysmemMsg;
 
+typedef struct {
+    uint32_t addr;
+    uint32_t val;
+    int l;
+} ConfDataMsg;
+
 /**
  * MPQemuMsg:
  * @cmd: The remote command
@@ -60,6 +68,7 @@ typedef struct {
 
     union {
         uint64_t u64;
+        ConfDataMsg conf_data;
         SyncSysmemMsg sync_sysmem;
     } data;
 
diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
index 1339cc7..5ef82da 100644
--- a/io/mpqemu-link.c
+++ b/io/mpqemu-link.c
@@ -278,6 +278,12 @@ bool mpqemu_msg_valid(MPQemuMsg *msg)
             return false;
         }
         break;
+    case PCI_CONFIG_WRITE:
+    case PCI_CONFIG_READ:
+        if (msg->size != sizeof(ConfDataMsg)) {
+            return false;
+        }
+        break;
     default:
         break;
     }
-- 
1.8.3.1

[PATCH v11 13/19] multi-process: PCI BAR read/write handling for proxy & remote endpoints

[Jagannathan Raman]

Proxy device object implements handler for PCI BAR writes and reads.
The handler uses BAR_WRITE/BAR_READ message to communicate to the
remote process with the BAR address and value to be written/read.
The remote process implements handler for BAR_WRITE/BAR_READ
message.

Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 hw/i386/remote-msg.c     | 87 ++++++++++++++++++++++++++++++++++++++++++++++++
 hw/pci/proxy.c           | 60 +++++++++++++++++++++++++++++++++
 include/hw/pci/proxy.h   | 10 ++++++
 include/io/mpqemu-link.h | 10 ++++++
 io/mpqemu-link.c         |  6 ++++
 5 files changed, 173 insertions(+)

diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
index 94937db..9b9ca7e 100644
--- a/hw/i386/remote-msg.c
+++ b/hw/i386/remote-msg.c
@@ -16,11 +16,14 @@
 #include "qapi/error.h"
 #include "sysemu/runstate.h"
 #include "hw/pci/pci.h"
+#include "exec/memattrs.h"
 
 static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
                                  MPQemuMsg *msg);
 static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
                                 MPQemuMsg *msg);
+static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp);
+static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp);
 
 void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
 {
@@ -55,6 +58,12 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
         case PCI_CONFIG_READ:
             process_config_read(com->ioc, pci_dev, &msg);
             break;
+        case BAR_WRITE:
+            process_bar_write(com->ioc, &msg, &local_err);
+            break;
+        case BAR_READ:
+            process_bar_read(com->ioc, &msg, &local_err);
+            break;
         default:
             error_setg(&local_err,
                        "Unknown command (%d) received for device %s (pid=%d)",
@@ -122,3 +131,81 @@ static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
                      getpid());
     }
 }
+
+static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp)
+{
+    BarAccessMsg *bar_access = &msg->data.bar_access;
+    AddressSpace *as =
+        bar_access->memory ? &address_space_memory : &address_space_io;
+    MPQemuMsg ret = { 0 };
+    MemTxResult res;
+    uint64_t val;
+    Error *local_err = NULL;
+
+    if (!is_power_of_2(bar_access->size) ||
+       (bar_access->size > sizeof(uint64_t))) {
+        ret.data.u64 = UINT64_MAX;
+        goto fail;
+    }
+
+    val = cpu_to_le64(bar_access->val);
+
+    res = address_space_rw(as, bar_access->addr, MEMTXATTRS_UNSPECIFIED,
+                           (void *)&val, bar_access->size, true);
+
+    if (res != MEMTX_OK) {
+        error_setg(errp, "Could not perform address space write operation,"
+                   " inaccessible address: %lx in pid %d.",
+                   bar_access->addr, getpid());
+        ret.data.u64 = -1;
+    }
+
+fail:
+    ret.cmd = RET_MSG;
+    ret.size = sizeof(ret.data.u64);
+
+    mpqemu_msg_send(&ret, ioc, &local_err);
+    if (local_err) {
+        error_setg(errp, "Error while sending message to proxy "
+                   "in remote process pid=%d", getpid());
+    }
+}
+
+static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp)
+{
+    BarAccessMsg *bar_access = &msg->data.bar_access;
+    MPQemuMsg ret = { 0 };
+    AddressSpace *as;
+    MemTxResult res;
+    uint64_t val = 0;
+    Error *local_err = NULL;
+
+    as = bar_access->memory ? &address_space_memory : &address_space_io;
+
+    if (!is_power_of_2(bar_access->size) ||
+       (bar_access->size > sizeof(uint64_t))) {
+        val = UINT64_MAX;
+        goto fail;
+    }
+
+    res = address_space_rw(as, bar_access->addr, MEMTXATTRS_UNSPECIFIED,
+                           (void *)&val, bar_access->size, false);
+
+    if (res != MEMTX_OK) {
+        error_setg(errp, "Could not perform address space read operation,"
+                   " inaccessible address: %lx in pid %d.",
+                   bar_access->addr, getpid());
+        val = UINT64_MAX;
+    }
+
+fail:
+    ret.cmd = RET_MSG;
+    ret.data.u64 = le64_to_cpu(val);
+    ret.size = sizeof(ret.data.u64);
+
+    mpqemu_msg_send(&ret, ioc, &local_err);
+    if (local_err) {
+        error_setg(errp, "Error while sending message to proxy "
+                   "in remote process pid=%d", getpid());
+    }
+}
diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
index 27714fe..4ae4919 100644
--- a/hw/pci/proxy.c
+++ b/hw/pci/proxy.c
@@ -146,3 +146,63 @@ static void pci_proxy_dev_register_types(void)
 }
 
 type_init(pci_proxy_dev_register_types)
+
+static void send_bar_access_msg(PCIProxyDev *pdev, MemoryRegion *mr,
+                                bool write, hwaddr addr, uint64_t *val,
+                                unsigned size, bool memory)
+{
+    MPQemuMsg msg = { 0 };
+    long ret = -EINVAL;
+    Error *local_err = NULL;
+
+    msg.size = sizeof(BarAccessMsg);
+    msg.data.bar_access.addr = mr->addr + addr;
+    msg.data.bar_access.size = size;
+    msg.data.bar_access.memory = memory;
+
+    if (write) {
+        msg.cmd = BAR_WRITE;
+        msg.data.bar_access.val = *val;
+    } else {
+        msg.cmd = BAR_READ;
+    }
+
+    ret = mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err);
+    if (local_err) {
+        error_report_err(local_err);
+    }
+
+    if (!write) {
+        *val = ret;
+    }
+}
+
+static void proxy_bar_write(void *opaque, hwaddr addr, uint64_t val,
+                            unsigned size)
+{
+    ProxyMemoryRegion *pmr = opaque;
+
+    send_bar_access_msg(pmr->dev, &pmr->mr, true, addr, &val, size,
+                        pmr->memory);
+}
+
+static uint64_t proxy_bar_read(void *opaque, hwaddr addr, unsigned size)
+{
+    ProxyMemoryRegion *pmr = opaque;
+    uint64_t val;
+
+    send_bar_access_msg(pmr->dev, &pmr->mr, false, addr, &val, size,
+                        pmr->memory);
+
+    return val;
+}
+
+const MemoryRegionOps proxy_mr_ops = {
+    .read = proxy_bar_read,
+    .write = proxy_bar_write,
+    .endianness = DEVICE_NATIVE_ENDIAN,
+    .impl = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+    },
+};
diff --git a/include/hw/pci/proxy.h b/include/hw/pci/proxy.h
index 923432a..e29c61b 100644
--- a/include/hw/pci/proxy.h
+++ b/include/hw/pci/proxy.h
@@ -16,8 +16,17 @@
 
 #define PCI_PROXY_DEV(obj) \
             OBJECT_CHECK(PCIProxyDev, (obj), TYPE_PCI_PROXY_DEV)
+
 typedef struct PCIProxyDev PCIProxyDev;
 
+typedef struct ProxyMemoryRegion {
+    PCIProxyDev *dev;
+    MemoryRegion mr;
+    bool memory;
+    bool present;
+    uint8_t type;
+} ProxyMemoryRegion;
+
 struct PCIProxyDev {
     PCIDevice parent_dev;
     char *fd;
@@ -31,6 +40,7 @@ struct PCIProxyDev {
     QemuMutex io_mutex;
     QIOChannel *ioc;
     Error *migration_blocker;
+    ProxyMemoryRegion region[PCI_NUM_REGIONS];
 };
 
 #endif /* PROXY_H */
diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
index 459d345..234ac82 100644
--- a/include/io/mpqemu-link.h
+++ b/include/io/mpqemu-link.h
@@ -36,6 +36,8 @@ typedef enum {
     RET_MSG,
     PCI_CONFIG_WRITE,
     PCI_CONFIG_READ,
+    BAR_WRITE,
+    BAR_READ,
     MPQEMU_CMD_MAX,
 } MPQemuCmd;
 
@@ -51,6 +53,13 @@ typedef struct {
     int l;
 } ConfDataMsg;
 
+typedef struct {
+    hwaddr addr;
+    uint64_t val;
+    unsigned size;
+    bool memory;
+} BarAccessMsg;
+
 /**
  * MPQemuMsg:
  * @cmd: The remote command
@@ -70,6 +79,7 @@ typedef struct {
         uint64_t u64;
         ConfDataMsg conf_data;
         SyncSysmemMsg sync_sysmem;
+        BarAccessMsg bar_access;
     } data;
 
     int fds[REMOTE_MAX_FDS];
diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
index 5ef82da..7e297ef 100644
--- a/io/mpqemu-link.c
+++ b/io/mpqemu-link.c
@@ -284,6 +284,12 @@ bool mpqemu_msg_valid(MPQemuMsg *msg)
             return false;
         }
         break;
+    case BAR_WRITE:
+    case BAR_READ:
+        if ((msg->size != sizeof(BarAccessMsg)) || (msg->num_fds != 0)) {
+            return false;
+        }
+        break;
     default:
         break;
     }
-- 
1.8.3.1

[PATCH v11 14/19] multi-process: Synchronize remote memory

[Jagannathan Raman]

Add memory-listener object which is used to keep the view of the RAM
in sync between QEMU and remote process.
A MemoryListener is registered for system-memory AddressSpace. The
listener sends SYNC_SYSMEM message to the remote process when memory
listener commits the changes to memory, the remote process receives
the message and processes it in the handler for SYNC_SYSMEM message.

Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                  |   2 +
 hw/i386/remote-msg.c         |   5 ++
 hw/pci/memory-sync.c         | 210 +++++++++++++++++++++++++++++++++++++++++++
 hw/pci/meson.build           |   2 +
 hw/pci/proxy.c               |   6 ++
 include/hw/pci/memory-sync.h |  27 ++++++
 include/hw/pci/proxy.h       |   2 +
 7 files changed, 254 insertions(+)
 create mode 100644 hw/pci/memory-sync.c
 create mode 100644 include/hw/pci/memory-sync.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 202db86..8ee717c 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3113,6 +3113,8 @@ F: include/hw/i386/remote-memory.h
 F: hw/i386/remote-memory.c
 F: hw/pci/proxy.c
 F: include/hw/pci/proxy.h
+F: hw/pci/memory-sync.c
+F: include/hw/pci/memory-sync.h
 
 Build and test automation
 -------------------------
diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
index 9b9ca7e..3465a9d 100644
--- a/hw/i386/remote-msg.c
+++ b/hw/i386/remote-msg.c
@@ -17,6 +17,7 @@
 #include "sysemu/runstate.h"
 #include "hw/pci/pci.h"
 #include "exec/memattrs.h"
+#include "hw/i386/remote-memory.h"
 
 static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
                                  MPQemuMsg *msg);
@@ -64,6 +65,10 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
         case BAR_READ:
             process_bar_read(com->ioc, &msg, &local_err);
             break;
+        case SYNC_SYSMEM:
+            remote_sysmem_reconfig(&msg, &local_err);
+            break;
+
         default:
             error_setg(&local_err,
                        "Unknown command (%d) received for device %s (pid=%d)",
diff --git a/hw/pci/memory-sync.c b/hw/pci/memory-sync.c
new file mode 100644
index 0000000..26a8784
--- /dev/null
+++ b/hw/pci/memory-sync.c
@@ -0,0 +1,210 @@
+/*
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "qemu/compiler.h"
+#include "qemu/int128.h"
+#include "qemu/range.h"
+#include "exec/memory.h"
+#include "exec/cpu-common.h"
+#include "cpu.h"
+#include "exec/ram_addr.h"
+#include "exec/address-spaces.h"
+#include "io/mpqemu-link.h"
+#include "hw/pci/memory-sync.h"
+
+static void proxy_ml_begin(MemoryListener *listener)
+{
+    RemoteMemSync *sync = container_of(listener, RemoteMemSync, listener);
+    int mrs;
+
+    for (mrs = 0; mrs < sync->n_mr_sections; mrs++) {
+        memory_region_unref(sync->mr_sections[mrs].mr);
+    }
+
+    g_free(sync->mr_sections);
+    sync->mr_sections = NULL;
+    sync->n_mr_sections = 0;
+}
+
+static int get_fd_from_hostaddr(uint64_t host, ram_addr_t *offset)
+{
+    MemoryRegion *mr;
+    ram_addr_t off;
+
+    /**
+     * Assumes that the host address is a valid address as it's
+     * coming from the MemoryListener system. In the case host
+     * address is not valid, the following call would return
+     * the default subregion of "system_memory" region, and
+     * not NULL. So it's not possible to check for NULL here.
+     */
+    mr = memory_region_from_host((void *)(uintptr_t)host, &off);
+
+    if (offset) {
+        *offset = off;
+    }
+
+    return memory_region_get_fd(mr);
+}
+
+static bool proxy_mrs_can_merge(uint64_t host, uint64_t prev_host, size_t size)
+{
+    bool merge;
+    int fd1, fd2;
+
+    fd1 = get_fd_from_hostaddr(host, NULL);
+
+    fd2 = get_fd_from_hostaddr(prev_host, NULL);
+
+    merge = (fd1 == fd2);
+
+    merge &= ((prev_host + size) == host);
+
+    return merge;
+}
+
+static bool try_merge(RemoteMemSync *sync, MemoryRegionSection *section)
+{
+    uint64_t mrs_size, mrs_gpa, mrs_page;
+    MemoryRegionSection *prev_sec;
+    bool merged = false;
+    uintptr_t mrs_host;
+    RAMBlock *mrs_rb;
+
+    if (!sync->n_mr_sections) {
+        return false;
+    }
+
+    mrs_rb = section->mr->ram_block;
+    mrs_page = (uint64_t)qemu_ram_pagesize(mrs_rb);
+    mrs_size = int128_get64(section->size);
+    mrs_gpa = section->offset_within_address_space;
+    mrs_host = (uintptr_t)memory_region_get_ram_ptr(section->mr) +
+               section->offset_within_region;
+
+    if (get_fd_from_hostaddr(mrs_host, NULL) < 0) {
+        return true;
+    }
+
+    mrs_host = mrs_host & ~(mrs_page - 1);
+    mrs_gpa = mrs_gpa & ~(mrs_page - 1);
+    mrs_size = ROUND_UP(mrs_size, mrs_page);
+
+    prev_sec = sync->mr_sections + (sync->n_mr_sections - 1);
+    uint64_t prev_gpa_start = prev_sec->offset_within_address_space;
+    uint64_t prev_size = int128_get64(prev_sec->size);
+    uint64_t prev_gpa_end   = range_get_last(prev_gpa_start, prev_size);
+    uint64_t prev_host_start =
+        (uintptr_t)memory_region_get_ram_ptr(prev_sec->mr) +
+        prev_sec->offset_within_region;
+    uint64_t prev_host_end = range_get_last(prev_host_start, prev_size);
+
+    if (mrs_gpa <= (prev_gpa_end + 1)) {
+        g_assert(mrs_gpa > prev_gpa_start);
+
+        if ((section->mr == prev_sec->mr) &&
+            proxy_mrs_can_merge(mrs_host, prev_host_start,
+                                (mrs_gpa - prev_gpa_start))) {
+            uint64_t max_end = MAX(prev_host_end, mrs_host + mrs_size);
+            merged = true;
+            prev_sec->offset_within_address_space =
+                MIN(prev_gpa_start, mrs_gpa);
+            prev_sec->offset_within_region =
+                MIN(prev_host_start, mrs_host) -
+                (uintptr_t)memory_region_get_ram_ptr(prev_sec->mr);
+            prev_sec->size = int128_make64(max_end - MIN(prev_host_start,
+                                                         mrs_host));
+        }
+    }
+
+    return merged;
+}
+
+static void proxy_ml_region_addnop(MemoryListener *listener,
+                                   MemoryRegionSection *section)
+{
+    RemoteMemSync *sync = container_of(listener, RemoteMemSync, listener);
+
+    if (!(memory_region_is_ram(section->mr) &&
+          !memory_region_is_rom(section->mr))) {
+        return;
+    }
+
+    if (try_merge(sync, section)) {
+        return;
+    }
+
+    ++sync->n_mr_sections;
+    sync->mr_sections = g_renew(MemoryRegionSection, sync->mr_sections,
+                                sync->n_mr_sections);
+    sync->mr_sections[sync->n_mr_sections - 1] = *section;
+    sync->mr_sections[sync->n_mr_sections - 1].fv = NULL;
+    memory_region_ref(section->mr);
+}
+
+static void proxy_ml_commit(MemoryListener *listener)
+{
+    RemoteMemSync *sync = container_of(listener, RemoteMemSync, listener);
+    MPQemuMsg msg;
+    MemoryRegionSection *section;
+    ram_addr_t offset;
+    uintptr_t host_addr;
+    int region;
+    Error *local_err = NULL;
+
+    memset(&msg, 0, sizeof(MPQemuMsg));
+
+    msg.cmd = SYNC_SYSMEM;
+    msg.num_fds = sync->n_mr_sections;
+    msg.size = sizeof(SyncSysmemMsg);
+    if (msg.num_fds > REMOTE_MAX_FDS) {
+        error_report("Number of fds is more than %d", REMOTE_MAX_FDS);
+        return;
+    }
+
+    for (region = 0; region < sync->n_mr_sections; region++) {
+        section = &sync->mr_sections[region];
+        msg.data.sync_sysmem.gpas[region] =
+            section->offset_within_address_space;
+        msg.data.sync_sysmem.sizes[region] = int128_get64(section->size);
+        host_addr = (uintptr_t)memory_region_get_ram_ptr(section->mr) +
+                    section->offset_within_region;
+        msg.fds[region] = get_fd_from_hostaddr(host_addr, &offset);
+        msg.data.sync_sysmem.offsets[region] = offset;
+    }
+    mpqemu_msg_send(&msg, sync->ioc, &local_err);
+    if (local_err) {
+        error_report("Error in sending command %d", msg.cmd);
+    }
+}
+
+void deconfigure_memory_sync(RemoteMemSync *sync)
+{
+    memory_listener_unregister(&sync->listener);
+
+    proxy_ml_begin(&sync->listener);
+}
+
+void configure_memory_sync(RemoteMemSync *sync, QIOChannel *ioc)
+{
+    sync->n_mr_sections = 0;
+    sync->mr_sections = NULL;
+
+    sync->ioc = ioc;
+
+    sync->listener.begin = proxy_ml_begin;
+    sync->listener.commit = proxy_ml_commit;
+    sync->listener.region_add = proxy_ml_region_addnop;
+    sync->listener.region_nop = proxy_ml_region_addnop;
+    sync->listener.priority = 10;
+
+    memory_listener_register(&sync->listener, &address_space_memory);
+}
diff --git a/hw/pci/meson.build b/hw/pci/meson.build
index 0df3017..7947a62 100644
--- a/hw/pci/meson.build
+++ b/hw/pci/meson.build
@@ -18,3 +18,5 @@ softmmu_ss.add_all(when: 'CONFIG_PCI', if_true: pci_ss)
 
 softmmu_ss.add(when: 'CONFIG_PCI', if_false: files('pci-stub.c'))
 softmmu_ss.add(when: 'CONFIG_ALL', if_true: files('pci-stub.c'))
+
+specific_ss.add(when: 'CONFIG_MPQEMU', if_true: files('memory-sync.c'))
diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
index 4ae4919..908c82e 100644
--- a/hw/pci/proxy.c
+++ b/hw/pci/proxy.c
@@ -18,6 +18,8 @@
 #include "migration/blocker.h"
 #include "io/mpqemu-link.h"
 #include "qemu/error-report.h"
+#include "hw/pci/memory-sync.h"
+#include "qom/object.h"
 
 static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
 {
@@ -58,6 +60,8 @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp)
 
     qemu_mutex_init(&dev->io_mutex);
     qio_channel_set_blocking(dev->ioc, true, NULL);
+
+    configure_memory_sync(&dev->sync, dev->ioc);
 }
 
 static void pci_proxy_dev_exit(PCIDevice *pdev)
@@ -69,6 +73,8 @@ static void pci_proxy_dev_exit(PCIDevice *pdev)
     migrate_del_blocker(dev->migration_blocker);
 
     error_free(dev->migration_blocker);
+
+    deconfigure_memory_sync(&dev->sync);
 }
 
 static int config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val,
diff --git a/include/hw/pci/memory-sync.h b/include/hw/pci/memory-sync.h
new file mode 100644
index 0000000..785f76a
--- /dev/null
+++ b/include/hw/pci/memory-sync.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef MEMORY_SYNC_H
+#define MEMORY_SYNC_H
+
+#include "exec/memory.h"
+#include "io/channel.h"
+
+typedef struct RemoteMemSync {
+    MemoryListener listener;
+
+    int n_mr_sections;
+    MemoryRegionSection *mr_sections;
+
+    QIOChannel *ioc;
+} RemoteMemSync;
+
+void configure_memory_sync(RemoteMemSync *sync, QIOChannel *ioc);
+void deconfigure_memory_sync(RemoteMemSync *sync);
+
+#endif
diff --git a/include/hw/pci/proxy.h b/include/hw/pci/proxy.h
index e29c61b..52f3c43 100644
--- a/include/hw/pci/proxy.h
+++ b/include/hw/pci/proxy.h
@@ -11,6 +11,7 @@
 
 #include "hw/pci/pci.h"
 #include "io/channel.h"
+#include "hw/pci/memory-sync.h"
 
 #define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev"
 
@@ -40,6 +41,7 @@ struct PCIProxyDev {
     QemuMutex io_mutex;
     QIOChannel *ioc;
     Error *migration_blocker;
+    RemoteMemSync sync;
     ProxyMemoryRegion region[PCI_NUM_REGIONS];
 };
 
-- 
1.8.3.1

[PATCH v11 15/19] multi-process: create IOHUB object to handle irq

[Jagannathan Raman]

IOHUB object is added to manage PCI IRQs. It uses KVM_IRQFD
ioctl to create irqfd to injecting PCI interrupts to the guest.
IOHUB object forwards the irqfd to the remote process. Remote process
uses this fd to directly send interrupts to the guest, bypassing QEMU.

Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                    |   2 +
 hw/i386/meson.build            |   1 +
 hw/i386/remote-iohub.c         | 123 +++++++++++++++++++++++++++++++++++++++++
 hw/i386/remote-msg.c           |   4 ++
 hw/i386/remote.c               |  10 ++++
 hw/pci/proxy.c                 |  58 +++++++++++++++++++
 include/hw/i386/remote-iohub.h |  42 ++++++++++++++
 include/hw/i386/remote.h       |   3 +
 include/hw/pci/pci_ids.h       |   3 +
 include/hw/pci/proxy.h         |   5 ++
 include/io/mpqemu-link.h       |   1 +
 io/mpqemu-link.c               |   5 ++
 12 files changed, 257 insertions(+)
 create mode 100644 hw/i386/remote-iohub.c
 create mode 100644 include/hw/i386/remote-iohub.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 8ee717c..f33de17 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3115,6 +3115,8 @@ F: hw/pci/proxy.c
 F: include/hw/pci/proxy.h
 F: hw/pci/memory-sync.c
 F: include/hw/pci/memory-sync.h
+F: hw/i386/remote-iohub.c
+F: include/hw/i386/remote-iohub.h
 
 Build and test automation
 -------------------------
diff --git a/hw/i386/meson.build b/hw/i386/meson.build
index b698b36..1442775 100644
--- a/hw/i386/meson.build
+++ b/hw/i386/meson.build
@@ -28,6 +28,7 @@ i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-msg.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-obj.c'))
 i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-memory.c'))
+i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote-iohub.c'))
 
 subdir('kvm')
 subdir('xen')
diff --git a/hw/i386/remote-iohub.c b/hw/i386/remote-iohub.c
new file mode 100644
index 0000000..e59424e
--- /dev/null
+++ b/hw/i386/remote-iohub.c
@@ -0,0 +1,123 @@
+/*
+ * Remote IO Hub
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+
+#include "hw/pci/pci.h"
+#include "hw/pci/pci_ids.h"
+#include "hw/pci/pci_bus.h"
+#include "qemu/thread.h"
+#include "hw/boards.h"
+#include "hw/i386/remote.h"
+#include "hw/i386/remote-iohub.h"
+#include "qemu/main-loop.h"
+
+void remote_iohub_init(RemoteIOHubState *iohub)
+{
+    int pirq;
+
+    memset(&iohub->irqfds, 0, sizeof(iohub->irqfds));
+    memset(&iohub->resamplefds, 0, sizeof(iohub->resamplefds));
+
+    for (pirq = 0; pirq < REMOTE_IOHUB_NB_PIRQS; pirq++) {
+        qemu_mutex_init(&iohub->irq_level_lock[pirq]);
+        iohub->irq_level[pirq] = 0;
+        event_notifier_init_fd(&iohub->irqfds[pirq], -1);
+        event_notifier_init_fd(&iohub->resamplefds[pirq], -1);
+    }
+}
+
+void remote_iohub_finalize(RemoteIOHubState *iohub)
+{
+    int pirq;
+
+    for (pirq = 0; pirq < REMOTE_IOHUB_NB_PIRQS; pirq++) {
+        qemu_set_fd_handler(event_notifier_get_fd(&iohub->resamplefds[pirq]),
+                            NULL, NULL, NULL);
+        event_notifier_cleanup(&iohub->irqfds[pirq]);
+        event_notifier_cleanup(&iohub->resamplefds[pirq]);
+        qemu_mutex_destroy(&iohub->irq_level_lock[pirq]);
+    }
+}
+
+int remote_iohub_map_irq(PCIDevice *pci_dev, int intx)
+{
+    return pci_dev->devfn;
+}
+
+void remote_iohub_set_irq(void *opaque, int pirq, int level)
+{
+    RemoteIOHubState *iohub = opaque;
+
+    assert(pirq >= 0);
+    assert(pirq < PCI_DEVFN_MAX);
+
+    qemu_mutex_lock(&iohub->irq_level_lock[pirq]);
+
+    if (level) {
+        if (++iohub->irq_level[pirq] == 1) {
+            event_notifier_set(&iohub->irqfds[pirq]);
+        }
+    } else if (iohub->irq_level[pirq] > 0) {
+        iohub->irq_level[pirq]--;
+    }
+
+    qemu_mutex_unlock(&iohub->irq_level_lock[pirq]);
+}
+
+static void intr_resample_handler(void *opaque)
+{
+    ResampleToken *token = opaque;
+    RemoteIOHubState *iohub = token->iohub;
+    int pirq, s;
+
+    pirq = token->pirq;
+
+    s = event_notifier_test_and_clear(&iohub->resamplefds[pirq]);
+
+    assert(s >= 0);
+
+    qemu_mutex_lock(&iohub->irq_level_lock[pirq]);
+
+    if (iohub->irq_level[pirq]) {
+        event_notifier_set(&iohub->irqfds[pirq]);
+    }
+
+    qemu_mutex_unlock(&iohub->irq_level_lock[pirq]);
+}
+
+void process_set_irqfd_msg(PCIDevice *pci_dev, MPQemuMsg *msg)
+{
+    RemoteMachineState *machine = REMOTE_MACHINE(current_machine);
+    RemoteIOHubState *iohub = &machine->iohub;
+    int pirq, intx;
+
+    intx = pci_get_byte(pci_dev->config + PCI_INTERRUPT_PIN) - 1;
+
+    pirq = remote_iohub_map_irq(pci_dev, intx);
+
+    if (event_notifier_get_fd(&iohub->irqfds[pirq]) != -1) {
+        qemu_set_fd_handler(event_notifier_get_fd(&iohub->resamplefds[pirq]),
+                            NULL, NULL, NULL);
+        event_notifier_cleanup(&iohub->irqfds[pirq]);
+        event_notifier_cleanup(&iohub->resamplefds[pirq]);
+        memset(&iohub->token[pirq], 0, sizeof(ResampleToken));
+    }
+
+    event_notifier_init_fd(&iohub->irqfds[pirq], msg->fds[0]);
+    event_notifier_init_fd(&iohub->resamplefds[pirq], msg->fds[1]);
+
+    iohub->token[pirq].iohub = iohub;
+    iohub->token[pirq].pirq = pirq;
+
+    qemu_set_fd_handler(msg->fds[1], intr_resample_handler, NULL,
+                        &iohub->token[pirq]);
+}
diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
index 3465a9d..a318a9f 100644
--- a/hw/i386/remote-msg.c
+++ b/hw/i386/remote-msg.c
@@ -18,6 +18,7 @@
 #include "hw/pci/pci.h"
 #include "exec/memattrs.h"
 #include "hw/i386/remote-memory.h"
+#include "hw/i386/remote-iohub.h"
 
 static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
                                  MPQemuMsg *msg);
@@ -68,6 +69,9 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
         case SYNC_SYSMEM:
             remote_sysmem_reconfig(&msg, &local_err);
             break;
+        case SET_IRQFD:
+            process_set_irqfd_msg(pci_dev, &msg);
+            break;
 
         default:
             error_setg(&local_err,
diff --git a/hw/i386/remote.c b/hw/i386/remote.c
index a67be33..95de652 100644
--- a/hw/i386/remote.c
+++ b/hw/i386/remote.c
@@ -20,12 +20,15 @@
 #include "exec/address-spaces.h"
 #include "exec/memory.h"
 #include "qapi/error.h"
+#include "hw/pci/pci_host.h"
+#include "hw/i386/remote-iohub.h"
 
 static void remote_machine_init(MachineState *machine)
 {
     MemoryRegion *system_memory, *system_io, *pci_memory;
     RemoteMachineState *s = REMOTE_MACHINE(machine);
     RemotePCIHost *rem_host;
+    PCIHostState *pci_host;
 
     system_memory = get_system_memory();
     system_io = get_system_io();
@@ -45,6 +48,13 @@ static void remote_machine_init(MachineState *machine)
     memory_region_add_subregion_overlap(system_memory, 0x0, pci_memory, -1);
 
     qdev_realize(DEVICE(rem_host), sysbus_get_default(), &error_fatal);
+
+    pci_host = PCI_HOST_BRIDGE(rem_host);
+
+    remote_iohub_init(&s->iohub);
+
+    pci_bus_irqs(pci_host->bus, remote_iohub_set_irq, remote_iohub_map_irq,
+                 &s->iohub, REMOTE_IOHUB_NB_PIRQS);
 }
 
 static void remote_machine_class_init(ObjectClass *oc, void *data)
diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
index 908c82e..93cffcc 100644
--- a/hw/pci/proxy.c
+++ b/hw/pci/proxy.c
@@ -20,6 +20,9 @@
 #include "qemu/error-report.h"
 #include "hw/pci/memory-sync.h"
 #include "qom/object.h"
+#include "qemu/event_notifier.h"
+#include "sysemu/kvm.h"
+#include "util/event_notifier-posix.c"
 
 static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
 {
@@ -31,6 +34,56 @@ static Property proxy_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static void proxy_intx_update(PCIDevice *pci_dev)
+{
+    PCIProxyDev *dev = PCI_PROXY_DEV(pci_dev);
+    PCIINTxRoute route;
+    int pin = pci_get_byte(pci_dev->config + PCI_INTERRUPT_PIN) - 1;
+
+    if (dev->virq != -1) {
+        kvm_irqchip_remove_irqfd_notifier_gsi(kvm_state, &dev->intr, dev->virq);
+        dev->virq = -1;
+    }
+
+    route = pci_device_route_intx_to_irq(pci_dev, pin);
+
+    dev->virq = route.irq;
+
+    if (dev->virq != -1) {
+        kvm_irqchip_add_irqfd_notifier_gsi(kvm_state, &dev->intr,
+                                           &dev->resample, dev->virq);
+    }
+}
+
+static void setup_irqfd(PCIProxyDev *dev)
+{
+    PCIDevice *pci_dev = PCI_DEVICE(dev);
+    MPQemuMsg msg;
+    Error *local_err = NULL;
+
+    event_notifier_init(&dev->intr, 0);
+    event_notifier_init(&dev->resample, 0);
+
+    memset(&msg, 0, sizeof(MPQemuMsg));
+    msg.cmd = SET_IRQFD;
+    msg.num_fds = 2;
+    msg.fds[0] = event_notifier_get_fd(&dev->intr);
+    msg.fds[1] = event_notifier_get_fd(&dev->resample);
+    msg.size = 0;
+
+    mpqemu_msg_send(&msg, dev->ioc, &local_err);
+    if (local_err) {
+        error_report("Error to send cmd to remote process %d",
+                     msg.cmd);
+    }
+
+    dev->virq = -1;
+
+    proxy_intx_update(pci_dev);
+
+    pci_device_set_intx_routing_notifier(pci_dev, proxy_intx_update);
+}
+
 static void pci_proxy_dev_realize(PCIDevice *device, Error **errp)
 {
     PCIProxyDev *dev = PCI_PROXY_DEV(device);
@@ -62,6 +115,8 @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp)
     qio_channel_set_blocking(dev->ioc, true, NULL);
 
     configure_memory_sync(&dev->sync, dev->ioc);
+
+    setup_irqfd(dev);
 }
 
 static void pci_proxy_dev_exit(PCIDevice *pdev)
@@ -75,6 +130,9 @@ static void pci_proxy_dev_exit(PCIDevice *pdev)
     error_free(dev->migration_blocker);
 
     deconfigure_memory_sync(&dev->sync);
+
+    event_notifier_cleanup(&dev->intr);
+    event_notifier_cleanup(&dev->resample);
 }
 
 static int config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val,
diff --git a/include/hw/i386/remote-iohub.h b/include/hw/i386/remote-iohub.h
new file mode 100644
index 0000000..cd59476
--- /dev/null
+++ b/include/hw/i386/remote-iohub.h
@@ -0,0 +1,42 @@
+/*
+ * IO Hub for remote device
+ *
+ * Copyright © 2018, 2020 Oracle and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef REMOTE_IOHUB_H
+#define REMOTE_IOHUB_H
+
+#include "hw/pci/pci.h"
+#include "qemu/event_notifier.h"
+#include "qemu/thread-posix.h"
+#include "io/mpqemu-link.h"
+
+#define REMOTE_IOHUB_NB_PIRQS    PCI_DEVFN_MAX
+
+typedef struct ResampleToken {
+    void *iohub;
+    int pirq;
+} ResampleToken;
+
+typedef struct RemoteIOHubState {
+    PCIDevice d;
+    EventNotifier irqfds[REMOTE_IOHUB_NB_PIRQS];
+    EventNotifier resamplefds[REMOTE_IOHUB_NB_PIRQS];
+    unsigned int irq_level[REMOTE_IOHUB_NB_PIRQS];
+    ResampleToken token[REMOTE_IOHUB_NB_PIRQS];
+    QemuMutex irq_level_lock[REMOTE_IOHUB_NB_PIRQS];
+} RemoteIOHubState;
+
+int remote_iohub_map_irq(PCIDevice *pci_dev, int intx);
+void remote_iohub_set_irq(void *opaque, int pirq, int level);
+void process_set_irqfd_msg(PCIDevice *pci_dev, MPQemuMsg *msg);
+
+void remote_iohub_init(RemoteIOHubState *iohub);
+void remote_iohub_finalize(RemoteIOHubState *iohub);
+
+#endif
diff --git a/include/hw/i386/remote.h b/include/hw/i386/remote.h
index 3073db6..dec072f 100644
--- a/include/hw/i386/remote.h
+++ b/include/hw/i386/remote.h
@@ -15,11 +15,14 @@
 #include "hw/boards.h"
 #include "hw/pci-host/remote.h"
 #include "io/channel.h"
+#include "hw/i386/remote-iohub.h"
 
 typedef struct RemoteMachineState {
     MachineState parent_obj;
 
     RemotePCIHost *host;
+
+    RemoteIOHubState iohub;
 } RemoteMachineState;
 
 /* Used to pass to co-routine device and ioc. */
diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
index 11f8ab7..bd0c17d 100644
--- a/include/hw/pci/pci_ids.h
+++ b/include/hw/pci/pci_ids.h
@@ -192,6 +192,9 @@
 #define PCI_DEVICE_ID_SUN_SIMBA          0x5000
 #define PCI_DEVICE_ID_SUN_SABRE          0xa000
 
+#define PCI_VENDOR_ID_ORACLE             0x108e
+#define PCI_DEVICE_ID_REMOTE_IOHUB       0xb000
+
 #define PCI_VENDOR_ID_CMD                0x1095
 #define PCI_DEVICE_ID_CMD_646            0x0646
 
diff --git a/include/hw/pci/proxy.h b/include/hw/pci/proxy.h
index 52f3c43..1482217 100644
--- a/include/hw/pci/proxy.h
+++ b/include/hw/pci/proxy.h
@@ -12,6 +12,7 @@
 #include "hw/pci/pci.h"
 #include "io/channel.h"
 #include "hw/pci/memory-sync.h"
+#include "qemu/event_notifier.h"
 
 #define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev"
 
@@ -42,6 +43,10 @@ struct PCIProxyDev {
     QIOChannel *ioc;
     Error *migration_blocker;
     RemoteMemSync sync;
+    int virq;
+    EventNotifier intr;
+    EventNotifier resample;
+
     ProxyMemoryRegion region[PCI_NUM_REGIONS];
 };
 
diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
index 234ac82..4cf81f5 100644
--- a/include/io/mpqemu-link.h
+++ b/include/io/mpqemu-link.h
@@ -38,6 +38,7 @@ typedef enum {
     PCI_CONFIG_READ,
     BAR_WRITE,
     BAR_READ,
+    SET_IRQFD,
     MPQEMU_CMD_MAX,
 } MPQemuCmd;
 
diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
index 7e297ef..2d1e576 100644
--- a/io/mpqemu-link.c
+++ b/io/mpqemu-link.c
@@ -290,6 +290,11 @@ bool mpqemu_msg_valid(MPQemuMsg *msg)
             return false;
         }
         break;
+    case SET_IRQFD:
+        if (msg->size || (msg->num_fds != 2)) {
+            return false;
+        }
+        break;
     default:
         break;
     }
-- 
1.8.3.1

[PATCH v11 16/19] multi-process: Retrieve PCI info from remote process

[Jagannathan Raman]

Retrieve PCI configuration info about the remote device and
configure the Proxy PCI object based on the returned information

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 hw/pci/proxy.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 85 insertions(+)

diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
index 93cffcc..ccffddb 100644
--- a/hw/pci/proxy.c
+++ b/hw/pci/proxy.c
@@ -24,6 +24,8 @@
 #include "sysemu/kvm.h"
 #include "util/event_notifier-posix.c"
 
+static void probe_pci_info(PCIDevice *dev, Error **errp);
+
 static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
 {
     pdev->ioc = qio_channel_new_fd(fd, errp);
@@ -87,6 +89,7 @@ static void setup_irqfd(PCIProxyDev *dev)
 static void pci_proxy_dev_realize(PCIDevice *device, Error **errp)
 {
     PCIProxyDev *dev = PCI_PROXY_DEV(device);
+    uint8_t *pci_conf = device->config;
     int fd;
 
     if (dev->fd) {
@@ -114,9 +117,14 @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp)
     qemu_mutex_init(&dev->io_mutex);
     qio_channel_set_blocking(dev->ioc, true, NULL);
 
+    pci_conf[PCI_LATENCY_TIMER] = 0xff;
+    pci_conf[PCI_INTERRUPT_PIN] = 0x01;
+
     configure_memory_sync(&dev->sync, dev->ioc);
 
     setup_irqfd(dev);
+
+    probe_pci_info(PCI_DEVICE(dev), errp);
 }
 
 static void pci_proxy_dev_exit(PCIDevice *pdev)
@@ -270,3 +278,80 @@ const MemoryRegionOps proxy_mr_ops = {
         .max_access_size = 8,
     },
 };
+
+static void probe_pci_info(PCIDevice *dev, Error **errp)
+{
+    PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(dev);
+    uint32_t orig_val, new_val, base_class, val;
+    PCIProxyDev *pdev = PCI_PROXY_DEV(dev);
+    DeviceClass *dc = DEVICE_CLASS(pc);
+    uint8_t type;
+    int i, size;
+    char *name;
+
+    config_op_send(pdev, PCI_VENDOR_ID, &val, 2, PCI_CONFIG_READ);
+    pc->vendor_id = (uint16_t)val;
+
+    config_op_send(pdev, PCI_DEVICE_ID, &val, 2, PCI_CONFIG_READ);
+    pc->device_id = (uint16_t)val;
+
+    config_op_send(pdev, PCI_CLASS_DEVICE, &val, 2, PCI_CONFIG_READ);
+    pc->class_id = (uint16_t)val;
+
+    config_op_send(pdev, PCI_SUBSYSTEM_ID, &val, 2, PCI_CONFIG_READ);
+    pc->subsystem_id = (uint16_t)val;
+
+    base_class = pc->class_id >> 4;
+    switch (base_class) {
+    case PCI_BASE_CLASS_BRIDGE:
+        set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+        break;
+    case PCI_BASE_CLASS_STORAGE:
+        set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
+        break;
+    case PCI_BASE_CLASS_NETWORK:
+        set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
+        break;
+    case PCI_BASE_CLASS_INPUT:
+        set_bit(DEVICE_CATEGORY_INPUT, dc->categories);
+        break;
+    case PCI_BASE_CLASS_DISPLAY:
+        set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories);
+        break;
+    case PCI_BASE_CLASS_PROCESSOR:
+        set_bit(DEVICE_CATEGORY_CPU, dc->categories);
+        break;
+    default:
+        set_bit(DEVICE_CATEGORY_MISC, dc->categories);
+        break;
+    }
+
+    for (i = 0; i < PCI_NUM_REGIONS; i++) {
+        config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &orig_val, 4,
+                       PCI_CONFIG_READ);
+        new_val = 0xffffffff;
+        config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &new_val, 4,
+                       PCI_CONFIG_WRITE);
+        config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &new_val, 4,
+                       PCI_CONFIG_READ);
+        size = (~(new_val & 0xFFFFFFF0)) + 1;
+        config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &orig_val, 4,
+                       PCI_CONFIG_WRITE);
+        type = (new_val & 0x1) ?
+                   PCI_BASE_ADDRESS_SPACE_IO : PCI_BASE_ADDRESS_SPACE_MEMORY;
+
+        if (size) {
+            pdev->region[i].dev = pdev;
+            pdev->region[i].present = true;
+            if (type == PCI_BASE_ADDRESS_SPACE_MEMORY) {
+                pdev->region[i].memory = true;
+            }
+            name = g_strdup_printf("bar-region-%d", i);
+            memory_region_init_io(&pdev->region[i].mr, OBJECT(pdev),
+                                  &proxy_mr_ops, &pdev->region[i],
+                                  name, size);
+            pci_register_bar(dev, i, type, &pdev->region[i].mr);
+            g_free(name);
+        }
+    }
+}
-- 
1.8.3.1

[PATCH v11 17/19] multi-process: perform device reset in the remote process

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

Perform device reset in the remote process when QEMU performs
device reset. This is required to reset the internal state
(like registers, etc...) of emulated devices

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 hw/i386/remote-msg.c     | 23 ++++++++++++++++++++++-
 hw/pci/proxy.c           | 20 ++++++++++++++++++++
 include/io/mpqemu-link.h |  1 +
 3 files changed, 43 insertions(+), 1 deletion(-)

diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
index a318a9f..82f1b4f 100644
--- a/hw/i386/remote-msg.c
+++ b/hw/i386/remote-msg.c
@@ -19,6 +19,7 @@
 #include "exec/memattrs.h"
 #include "hw/i386/remote-memory.h"
 #include "hw/i386/remote-iohub.h"
+#include "sysemu/reset.h"
 
 static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
                                  MPQemuMsg *msg);
@@ -26,6 +27,8 @@ static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
                                 MPQemuMsg *msg);
 static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp);
 static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp);
+static void process_device_reset_msg(QIOChannel *ioc, PCIDevice *dev,
+                                     Error **errp);
 
 void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
 {
@@ -72,7 +75,9 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
         case SET_IRQFD:
             process_set_irqfd_msg(pci_dev, &msg);
             break;
-
+        case DEVICE_RESET:
+            process_device_reset_msg(com->ioc, pci_dev, &local_err);
+            break;
         default:
             error_setg(&local_err,
                        "Unknown command (%d) received for device %s (pid=%d)",
@@ -218,3 +223,19 @@ fail:
                    "in remote process pid=%d", getpid());
     }
 }
+
+static void process_device_reset_msg(QIOChannel *ioc, PCIDevice *dev,
+                                     Error **errp)
+{
+    DeviceClass *dc = DEVICE_GET_CLASS(dev);
+    DeviceState *s = DEVICE(dev);
+    MPQemuMsg ret = { 0 };
+
+    if (dc->reset) {
+        dc->reset(s);
+    }
+
+    ret.cmd = RET_MSG;
+
+    mpqemu_msg_send(&ret, ioc, errp);
+}
diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
index ccffddb..e5b74bb 100644
--- a/hw/pci/proxy.c
+++ b/hw/pci/proxy.c
@@ -25,6 +25,7 @@
 #include "util/event_notifier-posix.c"
 
 static void probe_pci_info(PCIDevice *dev, Error **errp);
+static void proxy_device_reset(DeviceState *dev);
 
 static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
 {
@@ -198,6 +199,8 @@ static void pci_proxy_dev_class_init(ObjectClass *klass, void *data)
     k->config_read = pci_proxy_read_config;
     k->config_write = pci_proxy_write_config;
 
+    dc->reset = proxy_device_reset;
+
     device_class_set_props(dc, proxy_properties);
 }
 
@@ -355,3 +358,20 @@ static void probe_pci_info(PCIDevice *dev, Error **errp)
         }
     }
 }
+
+static void proxy_device_reset(DeviceState *dev)
+{
+    PCIProxyDev *pdev = PCI_PROXY_DEV(dev);
+    MPQemuMsg msg = { 0 };
+    Error *local_err = NULL;
+
+    msg.cmd = DEVICE_RESET;
+    msg.size = 0;
+
+    (void)mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err);
+    if (local_err) {
+        error_report("Failed to send DEVICE_RESET to the remote process");
+        error_free(local_err);
+    }
+
+}
diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
index 4cf81f5..7462968 100644
--- a/include/io/mpqemu-link.h
+++ b/include/io/mpqemu-link.h
@@ -39,6 +39,7 @@ typedef enum {
     BAR_WRITE,
     BAR_READ,
     SET_IRQFD,
+    DEVICE_RESET,
     MPQEMU_CMD_MAX,
 } MPQemuCmd;
 
-- 
1.8.3.1

[PATCH v11 18/19] multi-process: add the concept description to docs/devel/qemu-multiprocess

[Jagannathan Raman]

From: John G Johnson <john.g.johnson@oracle.com>

Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                  |   1 +
 docs/devel/index.rst         |   1 +
 docs/devel/multi-process.rst | 966 +++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 968 insertions(+)
 create mode 100644 docs/devel/multi-process.rst

diff --git a/MAINTAINERS b/MAINTAINERS
index f33de17..9a911e0 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3117,6 +3117,7 @@ F: hw/pci/memory-sync.c
 F: include/hw/pci/memory-sync.h
 F: hw/i386/remote-iohub.c
 F: include/hw/i386/remote-iohub.h
+F: docs/devel/multi-process.rst
 
 Build and test automation
 -------------------------
diff --git a/docs/devel/index.rst b/docs/devel/index.rst
index 77baae5..06480ce 100644
--- a/docs/devel/index.rst
+++ b/docs/devel/index.rst
@@ -34,3 +34,4 @@ Contents:
    clocks
    qom
    block-coroutine-wrapper
+   multi-process
diff --git a/docs/devel/multi-process.rst b/docs/devel/multi-process.rst
new file mode 100644
index 0000000..6969932
--- /dev/null
+++ b/docs/devel/multi-process.rst
@@ -0,0 +1,966 @@
+This is the design document for multi-process QEMU. It does not
+necessarily reflect the status of the current implementation, which
+may lack features or be considerably different from what is described
+in this document. This document is still useful as a description of
+the goals and general direction of this feature.
+
+Please refer to the following wiki for latest details:
+https://wiki.qemu.org/Features/MultiProcessQEMU
+
+Multi-process QEMU
+===================
+
+QEMU is often used as the hypervisor for virtual machines running in the
+Oracle cloud. Since one of the advantages of cloud computing is the
+ability to run many VMs from different tenants in the same cloud
+infrastructure, a guest that compromised its hypervisor could
+potentially use the hypervisor's access privileges to access data it is
+not authorized for.
+
+QEMU can be susceptible to security attacks because it is a large,
+monolithic program that provides many features to the VMs it services.
+Many of these features can be configured out of QEMU, but even a reduced
+configuration QEMU has a large amount of code a guest can potentially
+attack. Separating QEMU reduces the attack surface by aiding to
+limit each component in the system to only access the resources that
+it needs to perform its job.
+
+QEMU services
+-------------
+
+QEMU can be broadly described as providing three main services. One is a
+VM control point, where VMs can be created, migrated, re-configured, and
+destroyed. A second is to emulate the CPU instructions within the VM,
+often accelerated by HW virtualization features such as Intel's VT
+extensions. Finally, it provides IO services to the VM by emulating HW
+IO devices, such as disk and network devices.
+
+A multi-process QEMU
+~~~~~~~~~~~~~~~~~~~~
+
+A multi-process QEMU involves separating QEMU services into separate
+host processes. Each of these processes can be given only the privileges
+it needs to provide its service, e.g., a disk service could be given
+access only to the disk images it provides, and not be allowed to
+access other files, or any network devices. An attacker who compromised
+this service would not be able to use this exploit to access files or
+devices beyond what the disk service was given access to.
+
+A QEMU control process would remain, but in multi-process mode, will
+have no direct interfaces to the VM. During VM execution, it would still
+provide the user interface to hot-plug devices or live migrate the VM.
+
+A first step in creating a multi-process QEMU is to separate IO services
+from the main QEMU program, which would continue to provide CPU
+emulation. i.e., the control process would also be the CPU emulation
+process. In a later phase, CPU emulation could be separated from the
+control process.
+
+Separating IO services
+----------------------
+
+Separating IO services into individual host processes is a good place to
+begin for a couple of reasons. One is the sheer number of IO devices QEMU
+can emulate provides a large surface of interfaces which could potentially
+be exploited, and, indeed, have been a source of exploits in the past.
+Another is the modular nature of QEMU device emulation code provides
+interface points where the QEMU functions that perform device emulation
+can be separated from the QEMU functions that manage the emulation of
+guest CPU instructions. The devices emulated in the separate process are
+referred to as remote devices.
+
+QEMU device emulation
+~~~~~~~~~~~~~~~~~~~~~
+
+QEMU uses an object oriented SW architecture for device emulation code.
+Configured objects are all compiled into the QEMU binary, then objects
+are instantiated by name when used by the guest VM. For example, the
+code to emulate a device named "foo" is always present in QEMU, but its
+instantiation code is only run when the device is included in the target
+VM. (e.g., via the QEMU command line as *-device foo*)
+
+The object model is hierarchical, so device emulation code names its
+parent object (such as "pci-device" for a PCI device) and QEMU will
+instantiate a parent object before calling the device's instantiation
+code.
+
+Current separation models
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+In order to separate the device emulation code from the CPU emulation
+code, the device object code must run in a different process. There are
+a couple of existing QEMU features that can run emulation code
+separately from the main QEMU process. These are examined below.
+
+vhost user model
+^^^^^^^^^^^^^^^^
+
+Virtio guest device drivers can be connected to vhost user applications
+in order to perform their IO operations. This model uses special virtio
+device drivers in the guest and vhost user device objects in QEMU, but
+once the QEMU vhost user code has configured the vhost user application,
+mission-mode IO is performed by the application. The vhost user
+application is a daemon process that can be contacted via a known UNIX
+domain socket.
+
+vhost socket
+''''''''''''
+
+As mentioned above, one of the tasks of the vhost device object within
+QEMU is to contact the vhost application and send it configuration
+information about this device instance. As part of the configuration
+process, the application can also be sent other file descriptors over
+the socket, which then can be used by the vhost user application in
+various ways, some of which are described below.
+
+vhost MMIO store acceleration
+'''''''''''''''''''''''''''''
+
+VMs are often run using HW virtualization features via the KVM kernel
+driver. This driver allows QEMU to accelerate the emulation of guest CPU
+instructions by running the guest in a virtual HW mode. When the guest
+executes instructions that cannot be executed by virtual HW mode,
+execution returns to the KVM driver so it can inform QEMU to emulate the
+instructions in SW.
+
+One of the events that can cause a return to QEMU is when a guest device
+driver accesses an IO location. QEMU then dispatches the memory
+operation to the corresponding QEMU device object. In the case of a
+vhost user device, the memory operation would need to be sent over a
+socket to the vhost application. This path is accelerated by the QEMU
+virtio code by setting up an eventfd file descriptor that the vhost
+application can directly receive MMIO store notifications from the KVM
+driver, instead of needing them to be sent to the QEMU process first.
+
+vhost interrupt acceleration
+''''''''''''''''''''''''''''
+
+Another optimization used by the vhost application is the ability to
+directly inject interrupts into the VM via the KVM driver, again,
+bypassing the need to send the interrupt back to the QEMU process first.
+The QEMU virtio setup code configures the KVM driver with an eventfd
+that triggers the device interrupt in the guest when the eventfd is
+written. This irqfd file descriptor is then passed to the vhost user
+application program.
+
+vhost access to guest memory
+''''''''''''''''''''''''''''
+
+The vhost application is also allowed to directly access guest memory,
+instead of needing to send the data as messages to QEMU. This is also
+done with file descriptors sent to the vhost user application by QEMU.
+These descriptors can be passed to ``mmap()`` by the vhost application
+to map the guest address space into the vhost application.
+
+IOMMUs introduce another level of complexity, since the address given to
+the guest virtio device to DMA to or from is not a guest physical
+address. This case is handled by having vhost code within QEMU register
+as a listener for IOMMU mapping changes. The vhost application maintains
+a cache of IOMMMU translations: sending translation requests back to
+QEMU on cache misses, and in turn receiving flush requests from QEMU
+when mappings are purged.
+
+applicability to device separation
+''''''''''''''''''''''''''''''''''
+
+Much of the vhost model can be re-used by separated device emulation. In
+particular, the ideas of using a socket between QEMU and the device
+emulation application, using a file descriptor to inject interrupts into
+the VM via KVM, and allowing the application to ``mmap()`` the guest
+should be re used.
+
+There are, however, some notable differences between how a vhost
+application works and the needs of separated device emulation. The most
+basic is that vhost uses custom virtio device drivers which always
+trigger IO with MMIO stores. A separated device emulation model must
+work with existing IO device models and guest device drivers. MMIO loads
+break vhost store acceleration since they are synchronous - guest
+progress cannot continue until the load has been emulated. By contrast,
+stores are asynchronous, the guest can continue after the store event
+has been sent to the vhost application.
+
+Another difference is that in the vhost user model, a single daemon can
+support multiple QEMU instances. This is contrary to the security regime
+desired, in which the emulation application should only be allowed to
+access the files or devices the VM it's running on behalf of can access.
+#### qemu-io model
+
+Qemu-io is a test harness used to test changes to the QEMU block backend
+object code. (e.g., the code that implements disk images for disk driver
+emulation) Qemu-io is not a device emulation application per se, but it
+does compile the QEMU block objects into a separate binary from the main
+QEMU one. This could be useful for disk device emulation, since its
+emulation applications will need to include the QEMU block objects.
+
+New separation model based on proxy objects
+-------------------------------------------
+
+A different model based on proxy objects in the QEMU program
+communicating with remote emulation programs could provide separation
+while minimizing the changes needed to the device emulation code. The
+rest of this section is a discussion of how a proxy object model would
+work.
+
+Remote emulation processes
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The remote emulation process will run the QEMU object hierarchy without
+modification. The device emulation objects will be also be based on the
+QEMU code, because for anything but the simplest device, it would not be
+a tractable to re-implement both the object model and the many device
+backends that QEMU has.
+
+The processes will communicate with the QEMU process over UNIX domain
+sockets. The processes can be executed either as standalone processes,
+or be executed by QEMU. In both cases, the host backends the emulation
+processes will provide are specified on its command line, as they would
+be for QEMU. For example:
+
+::
+
+    disk-proc -blockdev driver=file,node-name=file0,filename=disk-file0  \
+    -blockdev driver=qcow2,node-name=drive0,file=file0
+
+would indicate process *disk-proc* uses a qcow2 emulated disk named
+*file0* as its backend.
+
+Emulation processes may emulate more than one guest controller. A common
+configuration might be to put all controllers of the same device class
+(e.g., disk, network, etc.) in a single process, so that all backends of
+the same type can be managed by a single QMP monitor.
+
+communication with QEMU
+^^^^^^^^^^^^^^^^^^^^^^^
+
+The first argument to the remote emulation process will be a Unix domain
+socket that connects with the Proxy object. This is a required argument.
+
+::
+
+    disk-proc <socket number> <backend list>
+
+remote process QMP monitor
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Remote emulation processes can be monitored via QMP, similar to QEMU
+itself. The QMP monitor socket is specified the same as for a QEMU
+process:
+
+::
+
+    disk-proc -qmp unix:/tmp/disk-mon,server
+
+can be monitored over the UNIX socket path */tmp/disk-mon*.
+
+QEMU command line
+~~~~~~~~~~~~~~~~~
+
+Each remote device emulated in a remote process on the host is
+represented as a *-device* of type *pci-proxy-dev*. A socket
+sub-option to this option specifies the Unix socket that connects
+to the remote process. An *id* sub-option is required, and it should
+be the same id as used in the remote process.
+
+::
+
+    qemu-system-x86_64 ... -device pci-proxy-dev,id=lsi0,socket=3
+
+can be used to add a device emulated in a remote process
+
+
+QEMU management of remote processes
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+QEMU is not aware of the type of type of the remote PCI device. It is
+a pass through device as far as QEMU is concerned.
+
+communication with emulation process
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+primary channel
+'''''''''''''''
+
+The primary channel (referred to as com in the code) is used to bootstrap
+the remote process. It is also used to pass on device-agnostic commands
+like reset.
+
+per-device channels
+'''''''''''''''''''
+
+Each remote device communicates with QEMU using a dedicated communication
+channel. The proxy object sets up this channel using the primary
+channel during its initialization.
+
+QEMU device proxy objects
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+QEMU has an object model based on sub-classes inherited from the
+"object" super-class. The sub-classes that are of interest here are the
+"device" and "bus" sub-classes whose child sub-classes make up the
+device tree of a QEMU emulated system.
+
+The proxy object model will use device proxy objects to replace the
+device emulation code within the QEMU process. These objects will live
+in the same place in the object and bus hierarchies as the objects they
+replace. i.e., the proxy object for an LSI SCSI controller will be a
+sub-class of the "pci-device" class, and will have the same PCI bus
+parent and the same SCSI bus child objects as the LSI controller object
+it replaces.
+
+It is worth noting that the same proxy object is used to mediate with
+all types of remote PCI devices.
+
+object initialization
+^^^^^^^^^^^^^^^^^^^^^
+
+The Proxy device objects are initialized in the exact same manner in
+which any other QEMU device would be initialized.
+
+In addition, the Proxy objects perform the following two tasks:
+- Parses the "socket" sub option and connects to the remote process
+using this channel
+- Uses the "id" sub-option to connect to the emulated device on the
+separate process
+
+class\_init
+'''''''''''
+
+The ``class_init()`` method of a proxy object will, in general behave
+similarly to the object it replaces, including setting any static
+properties and methods needed by the proxy.
+
+instance\_init / realize
+''''''''''''''''''''''''
+
+The ``instance_init()`` and ``realize()`` functions would only need to
+perform tasks related to being a proxy, such are registering its own
+MMIO handlers, or creating a child bus that other proxy devices can be
+attached to later.
+
+Other tasks will be device-specific. For example, PCI device objects
+will initialize the PCI config space in order to make a valid PCI device
+tree within the QEMU process.
+
+address space registration
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Most devices are driven by guest device driver accesses to IO addresses
+or ports. The QEMU device emulation code uses QEMU's memory region
+function calls (such as ``memory_region_init_io()``) to add callback
+functions that QEMU will invoke when the guest accesses the device's
+areas of the IO address space. When a guest driver does access the
+device, the VM will exit HW virtualization mode and return to QEMU,
+which will then lookup and execute the corresponding callback function.
+
+A proxy object would need to mirror the memory region calls the actual
+device emulator would perform in its initialization code, but with its
+own callbacks. When invoked by QEMU as a result of a guest IO operation,
+they will forward the operation to the device emulation process.
+
+PCI config space
+^^^^^^^^^^^^^^^^
+
+PCI devices also have a configuration space that can be accessed by the
+guest driver. Guest accesses to this space is not handled by the device
+emulation object, but by its PCI parent object. Much of this space is
+read-only, but certain registers (especially BAR and MSI-related ones)
+need to be propagated to the emulation process.
+
+PCI parent proxy
+''''''''''''''''
+
+One way to propagate guest PCI config accesses is to create a
+"pci-device-proxy" class that can serve as the parent of a PCI device
+proxy object. This class's parent would be "pci-device" and it would
+override the PCI parent's ``config_read()`` and ``config_write()``
+methods with ones that forward these operations to the emulation
+program.
+
+interrupt receipt
+^^^^^^^^^^^^^^^^^
+
+A proxy for a device that generates interrupts will need to create a
+socket to receive interrupt indications from the emulation process. An
+incoming interrupt indication would then be sent up to its bus parent to
+be injected into the guest. For example, a PCI device object may use
+``pci_set_irq()``.
+
+live migration
+^^^^^^^^^^^^^^
+
+The proxy will register to save and restore any *vmstate* it needs over
+a live migration event. The device proxy does not need to manage the
+remote device's *vmstate*; that will be handled by the remote process
+proxy (see below).
+
+QEMU remote device operation
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Generic device operations, such as DMA, will be performed by the remote
+process proxy by sending messages to the remote process.
+
+DMA operations
+^^^^^^^^^^^^^^
+
+DMA operations would be handled much like vhost applications do. One of
+the initial messages sent to the emulation process is a guest memory
+table. Each entry in this table consists of a file descriptor and size
+that the emulation process can ``mmap()`` to directly access guest
+memory, similar to ``vhost_user_set_mem_table()``. Note guest memory
+must be backed by file descriptors, such as when QEMU is given the
+*-mem-path* command line option.
+
+IOMMU operations
+^^^^^^^^^^^^^^^^
+
+When the emulated system includes an IOMMU, the remote process proxy in
+QEMU will need to create a socket for IOMMU requests from the emulation
+process. It will handle those requests with an
+``address_space_get_iotlb_entry()`` call. In order to handle IOMMU
+unmaps, the remote process proxy will also register as a listener on the
+device's DMA address space. When an IOMMU memory region is created
+within the DMA address space, an IOMMU notifier for unmaps will be added
+to the memory region that will forward unmaps to the emulation process
+over the IOMMU socket.
+
+device hot-plug via QMP
+^^^^^^^^^^^^^^^^^^^^^^^
+
+An QMP "device\_add" command can add a device emulated by a remote
+process. It will also have "rid" option to the command, just as the
+*-device* command line option does. The remote process may either be one
+started at QEMU startup, or be one added by the "add-process" QMP
+command described above. In either case, the remote process proxy will
+forward the new device's JSON description to the corresponding emulation
+process.
+
+live migration
+^^^^^^^^^^^^^^
+
+The remote process proxy will also register for live migration
+notifications with ``vmstate_register()``. When called to save state,
+the proxy will send the remote process a secondary socket file
+descriptor to save the remote process's device *vmstate* over. The
+incoming byte stream length and data will be saved as the proxy's
+*vmstate*. When the proxy is resumed on its new host, this *vmstate*
+will be extracted, and a secondary socket file descriptor will be sent
+to the new remote process through which it receives the *vmstate* in
+order to restore the devices there.
+
+device emulation in remote process
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The parts of QEMU that the emulation program will need include the
+object model; the memory emulation objects; the device emulation objects
+of the targeted device, and any dependent devices; and, the device's
+backends. It will also need code to setup the machine environment,
+handle requests from the QEMU process, and route machine-level requests
+(such as interrupts or IOMMU mappings) back to the QEMU process.
+
+initialization
+^^^^^^^^^^^^^^
+
+The process initialization sequence will follow the same sequence
+followed by QEMU. It will first initialize the backend objects, then
+device emulation objects. The JSON descriptions sent by the QEMU process
+will drive which objects need to be created.
+
+-  address spaces
+
+Before the device objects are created, the initial address spaces and
+memory regions must be configured with ``memory_map_init()``. This
+creates a RAM memory region object (*system\_memory*) and an IO memory
+region object (*system\_io*).
+
+-  RAM
+
+RAM memory region creation will follow how ``pc_memory_init()`` creates
+them, but must use ``memory_region_init_ram_from_fd()`` instead of
+``memory_region_allocate_system_memory()``. The file descriptors needed
+will be supplied by the guest memory table from above. Those RAM regions
+would then be added to the *system\_memory* memory region with
+``memory_region_add_subregion()``.
+
+-  PCI
+
+IO initialization will be driven by the JSON descriptions sent from the
+QEMU process. For a PCI device, a PCI bus will need to be created with
+``pci_root_bus_new()``, and a PCI memory region will need to be created
+and added to the *system\_memory* memory region with
+``memory_region_add_subregion_overlap()``. The overlap version is
+required for architectures where PCI memory overlaps with RAM memory.
+
+MMIO handling
+^^^^^^^^^^^^^
+
+The device emulation objects will use ``memory_region_init_io()`` to
+install their MMIO handlers, and ``pci_register_bar()`` to associate
+those handlers with a PCI BAR, as they do within QEMU currently.
+
+In order to use ``address_space_rw()`` in the emulation process to
+handle MMIO requests from QEMU, the PCI physical addresses must be the
+same in the QEMU process and the device emulation process. In order to
+accomplish that, guest BAR programming must also be forwarded from QEMU
+to the emulation process.
+
+interrupt injection
+^^^^^^^^^^^^^^^^^^^
+
+When device emulation wants to inject an interrupt into the VM, the
+request climbs the device's bus object hierarchy until the point where a
+bus object knows how to signal the interrupt to the guest. The details
+depend on the type of interrupt being raised.
+
+-  PCI pin interrupts
+
+On x86 systems, there is an emulated IOAPIC object attached to the root
+PCI bus object, and the root PCI object forwards interrupt requests to
+it. The IOAPIC object, in turn, calls the KVM driver to inject the
+corresponding interrupt into the VM. The simplest way to handle this in
+an emulation process would be to setup the root PCI bus driver (via
+``pci_bus_irqs()``) to send a interrupt request back to the QEMU
+process, and have the device proxy object reflect it up the PCI tree
+there.
+
+-  PCI MSI/X interrupts
+
+PCI MSI/X interrupts are implemented in HW as DMA writes to a
+CPU-specific PCI address. In QEMU on x86, a KVM APIC object receives
+these DMA writes, then calls into the KVM driver to inject the interrupt
+into the VM. A simple emulation process implementation would be to send
+the MSI DMA address from QEMU as a message at initialization, then
+install an address space handler at that address which forwards the MSI
+message back to QEMU.
+
+DMA operations
+^^^^^^^^^^^^^^
+
+When a emulation object wants to DMA into or out of guest memory, it
+first must use dma\_memory\_map() to convert the DMA address to a local
+virtual address. The emulation process memory region objects setup above
+will be used to translate the DMA address to a local virtual address the
+device emulation code can access.
+
+IOMMU
+^^^^^
+
+When an IOMMU is in use in QEMU, DMA translation uses IOMMU memory
+regions to translate the DMA address to a guest physical address before
+that physical address can be translated to a local virtual address. The
+emulation process will need similar functionality.
+
+-  IOTLB cache
+
+The emulation process will maintain a cache of recent IOMMU translations
+(the IOTLB). When the translate() callback of an IOMMU memory region is
+invoked, the IOTLB cache will be searched for an entry that will map the
+DMA address to a guest PA. On a cache miss, a message will be sent back
+to QEMU requesting the corresponding translation entry, which be both be
+used to return a guest address and be added to the cache.
+
+-  IOTLB purge
+
+The IOMMU emulation will also need to act on unmap requests from QEMU.
+These happen when the guest IOMMU driver purges an entry from the
+guest's translation table.
+
+live migration
+^^^^^^^^^^^^^^
+
+When a remote process receives a live migration indication from QEMU, it
+will set up a channel using the received file descriptor with
+``qio_channel_socket_new_fd()``. This channel will be used to create a
+*QEMUfile* that can be passed to ``qemu_save_device_state()`` to send
+the process's device state back to QEMU. This method will be reversed on
+restore - the channel will be passed to ``qemu_loadvm_state()`` to
+restore the device state.
+
+Accelerating device emulation
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The messages that are required to be sent between QEMU and the emulation
+process can add considerable latency to IO operations. The optimizations
+described below attempt to ameliorate this effect by allowing the
+emulation process to communicate directly with the kernel KVM driver.
+The KVM file descriptors created would be passed to the emulation process
+via initialization messages, much like the guest memory table is done.
+#### MMIO acceleration
+
+Vhost user applications can receive guest virtio driver stores directly
+from KVM. The issue with the eventfd mechanism used by vhost user is
+that it does not pass any data with the event indication, so it cannot
+handle guest loads or guest stores that carry store data. This concept
+could, however, be expanded to cover more cases.
+
+The expanded idea would require a new type of KVM device:
+*KVM\_DEV\_TYPE\_USER*. This device has two file descriptors: a master
+descriptor that QEMU can use for configuration, and a slave descriptor
+that the emulation process can use to receive MMIO notifications. QEMU
+would create both descriptors using the KVM driver, and pass the slave
+descriptor to the emulation process via an initialization message.
+
+data structures
+^^^^^^^^^^^^^^^
+
+-  guest physical range
+
+The guest physical range structure describes the address range that a
+device will respond to. It includes the base and length of the range, as
+well as which bus the range resides on (e.g., on an x86machine, it can
+specify whether the range refers to memory or IO addresses).
+
+A device can have multiple physical address ranges it responds to (e.g.,
+a PCI device can have multiple BARs), so the structure will also include
+an enumerated identifier to specify which of the device's ranges is
+being referred to.
+
++--------+----------------------------+
+| Name   | Description                |
++========+============================+
+| addr   | range base address         |
++--------+----------------------------+
+| len    | range length               |
++--------+----------------------------+
+| bus    | addr type (memory or IO)   |
++--------+----------------------------+
+| id     | range ID (e.g., PCI BAR)   |
++--------+----------------------------+
+
+-  MMIO request structure
+
+This structure describes an MMIO operation. It includes which guest
+physical range the MMIO was within, the offset within that range, the
+MMIO type (e.g., load or store), and its length and data. It also
+includes a sequence number that can be used to reply to the MMIO, and
+the CPU that issued the MMIO.
+
++----------+------------------------+
+| Name     | Description            |
++==========+========================+
+| rid      | range MMIO is within   |
++----------+------------------------+
+| offset   | offset withing *rid*   |
++----------+------------------------+
+| type     | e.g., load or store    |
++----------+------------------------+
+| len      | MMIO length            |
++----------+------------------------+
+| data     | store data             |
++----------+------------------------+
+| seq      | sequence ID            |
++----------+------------------------+
+
+-  MMIO request queues
+
+MMIO request queues are FIFO arrays of MMIO request structures. There
+are two queues: pending queue is for MMIOs that haven't been read by the
+emulation program, and the sent queue is for MMIOs that haven't been
+acknowledged. The main use of the second queue is to validate MMIO
+replies from the emulation program.
+
+-  scoreboard
+
+Each CPU in the VM is emulated in QEMU by a separate thread, so multiple
+MMIOs may be waiting to be consumed by an emulation program and multiple
+threads may be waiting for MMIO replies. The scoreboard would contain a
+wait queue and sequence number for the per-CPU threads, allowing them to
+be individually woken when the MMIO reply is received from the emulation
+program. It also tracks the number of posted MMIO stores to the device
+that haven't been replied to, in order to satisfy the PCI constraint
+that a load to a device will not complete until all previous stores to
+that device have been completed.
+
+-  device shadow memory
+
+Some MMIO loads do not have device side-effects. These MMIOs can be
+completed without sending a MMIO request to the emulation program if the
+emulation program shares a shadow image of the device's memory image
+with the KVM driver.
+
+The emulation program will ask the KVM driver to allocate memory for the
+shadow image, and will then use ``mmap()`` to directly access it. The
+emulation program can control KVM access to the shadow image by sending
+KVM an access map telling it which areas of the image have no
+side-effects (and can be completed immediately), and which require a
+MMIO request to the emulation program. The access map can also inform
+the KVM drive which size accesses are allowed to the image.
+
+master descriptor
+^^^^^^^^^^^^^^^^^
+
+The master descriptor is used by QEMU to configure the new KVM device.
+The descriptor would be returned by the KVM driver when QEMU issues a
+*KVM\_CREATE\_DEVICE* ``ioctl()`` with a *KVM\_DEV\_TYPE\_USER* type.
+
+KVM\_DEV\_TYPE\_USER device ops
+
+
+The *KVM\_DEV\_TYPE\_USER* operations vector will be registered by a
+``kvm_register_device_ops()`` call when the KVM system in initialized by
+``kvm_init()``. These device ops are called by the KVM driver when QEMU
+executes certain ``ioctl()`` operations on its KVM file descriptor. They
+include:
+
+-  create
+
+This routine is called when QEMU issues a *KVM\_CREATE\_DEVICE*
+``ioctl()`` on its per-VM file descriptor. It will allocate and
+initialize a KVM user device specific data structure, and assign the
+*kvm\_device* private field to it.
+
+-  ioctl
+
+This routine is invoked when QEMU issues an ``ioctl()`` on the master
+descriptor. The ``ioctl()`` commands supported are defined by the KVM
+device type. *KVM\_DEV\_TYPE\_USER* ones will need several commands:
+
+*KVM\_DEV\_USER\_SLAVE\_FD* creates the slave file descriptor that will
+be passed to the device emulation program. Only one slave can be created
+by each master descriptor. The file operations performed by this
+descriptor are described below.
+
+The *KVM\_DEV\_USER\_PA\_RANGE* command configures a guest physical
+address range that the slave descriptor will receive MMIO notifications
+for. The range is specified by a guest physical range structure
+argument. For buses that assign addresses to devices dynamically, this
+command can be executed while the guest is running, such as the case
+when a guest changes a device's PCI BAR registers.
+
+*KVM\_DEV\_USER\_PA\_RANGE* will use ``kvm_io_bus_register_dev()`` to
+register *kvm\_io\_device\_ops* callbacks to be invoked when the guest
+performs a MMIO operation within the range. When a range is changed,
+``kvm_io_bus_unregister_dev()`` is used to remove the previous
+instantiation.
+
+*KVM\_DEV\_USER\_TIMEOUT* will configure a timeout value that specifies
+how long KVM will wait for the emulation process to respond to a MMIO
+indication.
+
+-  destroy
+
+This routine is called when the VM instance is destroyed. It will need
+to destroy the slave descriptor; and free any memory allocated by the
+driver, as well as the *kvm\_device* structure itself.
+
+slave descriptor
+^^^^^^^^^^^^^^^^
+
+The slave descriptor will have its own file operations vector, which
+responds to system calls on the descriptor performed by the device
+emulation program.
+
+-  read
+
+A read returns any pending MMIO requests from the KVM driver as MMIO
+request structures. Multiple structures can be returned if there are
+multiple MMIO operations pending. The MMIO requests are moved from the
+pending queue to the sent queue, and if there are threads waiting for
+space in the pending to add new MMIO operations, they will be woken
+here.
+
+-  write
+
+A write also consists of a set of MMIO requests. They are compared to
+the MMIO requests in the sent queue. Matches are removed from the sent
+queue, and any threads waiting for the reply are woken. If a store is
+removed, then the number of posted stores in the per-CPU scoreboard is
+decremented. When the number is zero, and a non side-effect load was
+waiting for posted stores to complete, the load is continued.
+
+-  ioctl
+
+There are several ioctl()s that can be performed on the slave
+descriptor.
+
+A *KVM\_DEV\_USER\_SHADOW\_SIZE* ``ioctl()`` causes the KVM driver to
+allocate memory for the shadow image. This memory can later be
+``mmap()``\ ed by the emulation process to share the emulation's view of
+device memory with the KVM driver.
+
+A *KVM\_DEV\_USER\_SHADOW\_CTRL* ``ioctl()`` controls access to the
+shadow image. It will send the KVM driver a shadow control map, which
+specifies which areas of the image can complete guest loads without
+sending the load request to the emulation program. It will also specify
+the size of load operations that are allowed.
+
+-  poll
+
+An emulation program will use the ``poll()`` call with a *POLLIN* flag
+to determine if there are MMIO requests waiting to be read. It will
+return if the pending MMIO request queue is not empty.
+
+-  mmap
+
+This call allows the emulation program to directly access the shadow
+image allocated by the KVM driver. As device emulation updates device
+memory, changes with no side-effects will be reflected in the shadow,
+and the KVM driver can satisfy guest loads from the shadow image without
+needing to wait for the emulation program.
+
+kvm\_io\_device ops
+^^^^^^^^^^^^^^^^^^^
+
+Each KVM per-CPU thread can handle MMIO operation on behalf of the guest
+VM. KVM will use the MMIO's guest physical address to search for a
+matching *kvm\_io\_device* to see if the MMIO can be handled by the KVM
+driver instead of exiting back to QEMU. If a match is found, the
+corresponding callback will be invoked.
+
+-  read
+
+This callback is invoked when the guest performs a load to the device.
+Loads with side-effects must be handled synchronously, with the KVM
+driver putting the QEMU thread to sleep waiting for the emulation
+process reply before re-starting the guest. Loads that do not have
+side-effects may be optimized by satisfying them from the shadow image,
+if there are no outstanding stores to the device by this CPU. PCI memory
+ordering demands that a load cannot complete before all older stores to
+the same device have been completed.
+
+-  write
+
+Stores can be handled asynchronously unless the pending MMIO request
+queue is full. In this case, the QEMU thread must sleep waiting for
+space in the queue. Stores will increment the number of posted stores in
+the per-CPU scoreboard, in order to implement the PCI ordering
+constraint above.
+
+interrupt acceleration
+^^^^^^^^^^^^^^^^^^^^^^
+
+This performance optimization would work much like a vhost user
+application does, where the QEMU process sets up *eventfds* that cause
+the device's corresponding interrupt to be triggered by the KVM driver.
+These irq file descriptors are sent to the emulation process at
+initialization, and are used when the emulation code raises a device
+interrupt.
+
+intx acceleration
+'''''''''''''''''
+
+Traditional PCI pin interrupts are level based, so, in addition to an
+irq file descriptor, a re-sampling file descriptor needs to be sent to
+the emulation program. This second file descriptor allows multiple
+devices sharing an irq to be notified when the interrupt has been
+acknowledged by the guest, so they can re-trigger the interrupt if their
+device has not de-asserted its interrupt.
+
+intx irq descriptor
+
+
+The irq descriptors are created by the proxy object
+``using event_notifier_init()`` to create the irq and re-sampling
+*eventds*, and ``kvm_vm_ioctl(KVM_IRQFD)`` to bind them to an interrupt.
+The interrupt route can be found with
+``pci_device_route_intx_to_irq()``.
+
+intx routing changes
+
+
+Intx routing can be changed when the guest programs the APIC the device
+pin is connected to. The proxy object in QEMU will use
+``pci_device_set_intx_routing_notifier()`` to be informed of any guest
+changes to the route. This handler will broadly follow the VFIO
+interrupt logic to change the route: de-assigning the existing irq
+descriptor from its route, then assigning it the new route. (see
+``vfio_intx_update()``)
+
+MSI/X acceleration
+''''''''''''''''''
+
+MSI/X interrupts are sent as DMA transactions to the host. The interrupt
+data contains a vector that is programmed by the guest, A device may have
+multiple MSI interrupts associated with it, so multiple irq descriptors
+may need to be sent to the emulation program.
+
+MSI/X irq descriptor
+
+
+This case will also follow the VFIO example. For each MSI/X interrupt,
+an *eventfd* is created, a virtual interrupt is allocated by
+``kvm_irqchip_add_msi_route()``, and the virtual interrupt is bound to
+the eventfd with ``kvm_irqchip_add_irqfd_notifier()``.
+
+MSI/X config space changes
+
+
+The guest may dynamically update several MSI-related tables in the
+device's PCI config space. These include per-MSI interrupt enables and
+vector data. Additionally, MSIX tables exist in device memory space, not
+config space. Much like the BAR case above, the proxy object must look
+at guest config space programming to keep the MSI interrupt state
+consistent between QEMU and the emulation program.
+
+--------------
+
+Disaggregated CPU emulation
+---------------------------
+
+After IO services have been disaggregated, a second phase would be to
+separate a process to handle CPU instruction emulation from the main
+QEMU control function. There are no object separation points for this
+code, so the first task would be to create one.
+
+Host access controls
+--------------------
+
+Separating QEMU relies on the host OS's access restriction mechanisms to
+enforce that the differing processes can only access the objects they
+are entitled to. There are a couple types of mechanisms usually provided
+by general purpose OSs.
+
+Discretionary access control
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Discretionary access control allows each user to control who can access
+their files. In Linux, this type of control is usually too coarse for
+QEMU separation, since it only provides three separate access controls:
+one for the same user ID, the second for users IDs with the same group
+ID, and the third for all other user IDs. Each device instance would
+need a separate user ID to provide access control, which is likely to be
+unwieldy for dynamically created VMs.
+
+Mandatory access control
+~~~~~~~~~~~~~~~~~~~~~~~~
+
+Mandatory access control allows the OS to add an additional set of
+controls on top of discretionary access for the OS to control. It also
+adds other attributes to processes and files such as types, roles, and
+categories, and can establish rules for how processes and files can
+interact.
+
+Type enforcement
+^^^^^^^^^^^^^^^^
+
+Type enforcement assigns a *type* attribute to processes and files, and
+allows rules to be written on what operations a process with a given
+type can perform on a file with a given type. QEMU separation could take
+advantage of type enforcement by running the emulation processes with
+different types, both from the main QEMU process, and from the emulation
+processes of different classes of devices.
+
+For example, guest disk images and disk emulation processes could have
+types separate from the main QEMU process and non-disk emulation
+processes, and the type rules could prevent processes other than disk
+emulation ones from accessing guest disk images. Similarly, network
+emulation processes can have a type separate from the main QEMU process
+and non-network emulation process, and only that type can access the
+host tun/tap device used to provide guest networking.
+
+Category enforcement
+^^^^^^^^^^^^^^^^^^^^
+
+Category enforcement assigns a set of numbers within a given range to
+the process or file. The process is granted access to the file if the
+process's set is a superset of the file's set. This enforcement can be
+used to separate multiple instances of devices in the same class.
+
+For example, if there are multiple disk devices provides to a guest,
+each device emulation process could be provisioned with a separate
+category. The different device emulation processes would not be able to
+access each other's backing disk images.
+
+Alternatively, categories could be used in lieu of the type enforcement
+scheme described above. In this scenario, different categories would be
+used to prevent device emulation processes in different classes from
+accessing resources assigned to other classes.
-- 
1.8.3.1

[PATCH v11 19/19] multi-process: add configure and usage information

[Jagannathan Raman]

From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 MAINTAINERS                |  2 ++
 docs/multi-process.rst     | 67 ++++++++++++++++++++++++++++++++++++++++++++++
 scripts/mpqemu-launcher.py | 49 +++++++++++++++++++++++++++++++++
 3 files changed, 118 insertions(+)
 create mode 100644 docs/multi-process.rst
 create mode 100755 scripts/mpqemu-launcher.py

diff --git a/MAINTAINERS b/MAINTAINERS
index 9a911e0..d12aba7 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3118,6 +3118,8 @@ F: include/hw/pci/memory-sync.h
 F: hw/i386/remote-iohub.c
 F: include/hw/i386/remote-iohub.h
 F: docs/devel/multi-process.rst
+F: scripts/mpqemu-launcher.py
+F: scripts/mpqemu-launcher-perf-mode.py
 
 Build and test automation
 -------------------------
diff --git a/docs/multi-process.rst b/docs/multi-process.rst
new file mode 100644
index 0000000..c4b022c
--- /dev/null
+++ b/docs/multi-process.rst
@@ -0,0 +1,67 @@
+Multi-process QEMU
+==================
+
+This document describes how to configure and use multi-process qemu.
+For the design document refer to docs/devel/qemu-multiprocess.
+
+1) Configuration
+----------------
+
+To enable support for multi-process add --enable-mpqemu
+to the list of options for the "configure" script.
+
+
+2) Usage
+--------
+
+Multi-process QEMU requires an orchestrator to launch. Please refer to a
+light-weight python based orchestrator for mpqemu in
+scripts/mpqemu-launcher.py to lauch QEMU in multi-process mode.
+
+Following is a description of command-line used to launch mpqemu.
+
+* Orchestrator:
+
+  - The Orchestrator creates a unix socketpair
+
+  - It launches the remote process and passes one of the
+    sockets to it via command-line.
+
+  - It then launches QEMU and specifies the other socket as an option
+    to the Proxy device object
+
+* Remote Process:
+
+  - QEMU can enter remote process mode by using the "remote" machine
+    option.
+
+  - The orchestrator creates a "remote-object" with details about
+    the device and the file descriptor for the device
+
+  - The remaining options are no different from how one launches QEMU with
+    devices.
+
+  - Example command-line for the remote process is as follows:
+
+      /usr/bin/qemu-system-x86_64                                        \
+      -machine x-remote                                                  \
+      -device lsi53c895a,id=lsi0                                         \
+      -drive id=drive_image2,file=/build/ol7-nvme-test-1.qcow2           \
+      -device scsi-hd,id=drive2,drive=drive_image2,bus=lsi0.0,scsi-id=0  \
+      -object x-remote-object,id=robj1,devid=lsi1,fd=4,
+
+* QEMU:
+
+  - Since parts of the RAM are shared between QEMU & remote process, a
+    memory-backend-memfd is required to facilitate this, as follows:
+
+    -object memory-backend-memfd,id=mem,size=2G
+
+  - A "x-pci-proxy-dev" device is created for each of the PCI devices emulated
+    in the remote process. A "socket" sub-option specifies the other end of
+    unix channel created by orchestrator. The "id" sub-option must be specified
+    and should be the same as the "id" specified for the remote PCI device
+
+  - Example commandline for QEMU is as follows:
+
+      -device x-pci-proxy-dev,id=lsi0,socket=3
diff --git a/scripts/mpqemu-launcher.py b/scripts/mpqemu-launcher.py
new file mode 100755
index 0000000..6e0ef22
--- /dev/null
+++ b/scripts/mpqemu-launcher.py
@@ -0,0 +1,49 @@
+#!/usr/bin/env python3
+import socket
+import os
+import subprocess
+import time
+
+PROC_QEMU='/usr/bin/qemu-system-x86_64'
+
+proxy, remote = socket.socketpair(socket.AF_UNIX, socket.SOCK_STREAM)
+
+remote_cmd = [ PROC_QEMU,                                                      \
+               '-machine', 'x-remote',                                         \
+               '-device', 'lsi53c895a,id=lsi1',                                \
+               '-drive', 'id=drive_image1,file=/build/ol7-nvme-test-1.qcow2',  \
+               '-device', 'scsi-hd,id=drive1,drive=drive_image1,bus=lsi1.0,'   \
+                              'scsi-id=0',                                     \
+               '-object',                                                      \
+               'x-remote-object,id=robj1,devid=lsi1,fd='+str(remote.fileno()), \
+               '-nographic',                                                   \
+             ]
+
+proxy_cmd = [ PROC_QEMU,                                                       \
+              '-name', 'OL7.4',                                                \
+              '-machine', 'q35,accel=kvm',                                     \
+              '-smp', 'sockets=1,cores=1,threads=1',                           \
+              '-m', '2048',                                                    \
+              '-object', 'memory-backend-memfd,id=sysmem-file,size=2G',        \
+              '-numa', 'node,memdev=sysmem-file',                              \
+              '-device', 'virtio-scsi-pci,id=virtio_scsi_pci0',                \
+              '-drive', 'id=drive_image1,if=none,format=qcow2,'                \
+                            'file=/home/ol7-hdd-1.qcow2',                      \
+              '-device', 'scsi-hd,id=image1,drive=drive_image1,'               \
+                             'bus=virtio_scsi_pci0.0',                         \
+              '-boot', 'd',                                                    \
+              '-vnc', ':0',                                                    \
+              '-device', 'x-pci-proxy-dev,id=lsi1,fd='+str(proxy.fileno()),    \
+            ]
+
+
+pid = os.fork();
+
+if pid:
+    # In Proxy
+    print('Launching QEMU with Proxy object');
+    process = subprocess.Popen(proxy_cmd, pass_fds=[proxy.fileno()])
+else:
+    # In remote
+    print('Launching Remote process');
+    process = subprocess.Popen(remote_cmd, pass_fds=[remote.fileno()])
-- 
1.8.3.1

Re: [PATCH v11 06/19] multi-process: define MPQemuMsg format and transmission functions

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 1341 bytes --]

On Thu, Oct 15, 2020 at 02:04:59PM -0400, Jagannathan Raman wrote:
> +void mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp)
> +{
> +    Error *local_err = NULL;
> +    int *fds = NULL;
> +    size_t nfds = 0;
> +    ssize_t len;
> +
> +    len = mpqemu_read(ioc, (void *)msg, MPQEMU_MSG_HDR_SIZE, &fds, &nfds,
> +                      &local_err);
> +    if (!local_err) {
> +        if (len == -EIO) {
> +            error_setg(&local_err, "Connection closed.");
> +            goto fail;
> +        }
> +        if (len < 0) {
> +            error_setg(&local_err, "Message length is less than 0");
> +            goto fail;
> +        }
> +        if (len != MPQEMU_MSG_HDR_SIZE) {
> +            error_setg(&local_err, "Message header corrupted");
> +            goto fail;
> +        }
> +    } else {
> +        goto fail;
> +    }
> +
> +    if (msg->size > sizeof(msg->data)) {
> +        error_setg(&local_err, "Invalid size for message");
> +        goto fail;
> +    }
> +
> +    if (mpqemu_read(ioc, (void *)&msg->data, msg->size, NULL, NULL,
> +                    &local_err) < 0) {
> +        goto fail;
> +    }
> +
> +    msg->num_fds = nfds;
> +    if (nfds) {
> +        memcpy(msg->fds, fds, nfds * sizeof(int));

I can't find anything that limits nfds to REMOTE_MAX_FDS. This looks
like a buffer overflow.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v11 08/19] multi-process: Associate fd of a PCIDevice with its object

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 804 bytes --]

On Thu, Oct 15, 2020 at 02:05:01PM -0400, Jagannathan Raman wrote:
> Associate the file descriptor for a PCIDevice in remote process with
> DeviceState object.
> 
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> ---
>  MAINTAINERS                  |   2 +
>  hw/i386/meson.build          |   1 +
>  hw/i386/remote-msg.c         |   1 +
>  hw/i386/remote-obj.c         | 154 +++++++++++++++++++++++++++++++++++++++++++
>  include/hw/i386/remote-obj.h |  42 ++++++++++++
>  5 files changed, 200 insertions(+)
>  create mode 100644 hw/i386/remote-obj.c
>  create mode 100644 include/hw/i386/remote-obj.h

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v11 10/19] multi-process: introduce proxy object

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 769 bytes --]

On Thu, Oct 15, 2020 at 02:05:03PM -0400, Jagannathan Raman wrote:
> From: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> 
> Defines a PCI Device proxy object as a child of TYPE_PCI_DEVICE.
> 
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> ---
>  MAINTAINERS            |  2 ++
>  hw/pci/meson.build     |  1 +
>  hw/pci/proxy.c         | 98 ++++++++++++++++++++++++++++++++++++++++++++++++++
>  include/hw/pci/proxy.h | 36 +++++++++++++++++++
>  4 files changed, 137 insertions(+)
>  create mode 100644 hw/pci/proxy.c
>  create mode 100644 include/hw/pci/proxy.h

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v11 00/19] Initial support for multi-process Qemu

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 641 bytes --]

On Thu, Oct 15, 2020 at 02:04:53PM -0400, Jagannathan Raman wrote:
> Hello,
> 
> This is the v11 of the patchset. Thank you very much for the
> review of the v10 of the series. We are glad to hear the
> patchset is getting in a better shape.
> 
> We made changes to the following patches in this series.
> 
> [PATCH v10 06/19] multi-process: define MPQemuMsg format and transmission functions
> [PATCH v10 08/19] multi-process: Associate fd of a PCIDevice with its object
> [PATCH v10 10/19] multi-process: introduce proxy object

I have reviewed these patches and posted a comment about a possible
buffer overflow.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v11 04/19] multi-process: setup a machine object for remote device process

[Philippe Mathieu-Daudé]

On 10/15/20 8:04 PM, Jagannathan Raman wrote:
> remote-machine object sets up various subsystems of the remote
> device process. Instantiate PCI host bridge object and initialize RAM, IO &
> PCI memory regions.
> 
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>   MAINTAINERS                  |  2 ++
>   hw/i386/meson.build          |  1 +
>   hw/i386/remote.c             | 69 ++++++++++++++++++++++++++++++++++++++++++++
>   include/hw/i386/remote.h     | 28 ++++++++++++++++++
>   include/hw/pci-host/remote.h |  1 +
>   5 files changed, 101 insertions(+)
>   create mode 100644 hw/i386/remote.c
>   create mode 100644 include/hw/i386/remote.h
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index bb1d701..12945ee 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -3102,6 +3102,8 @@ M: John G Johnson <john.g.johnson@oracle.com>
>   S: Maintained
>   F: hw/pci-host/remote.c
>   F: include/hw/pci-host/remote.h
> +F: hw/i386/remote.c
> +F: include/hw/i386/remote.h
>   
>   Build and test automation
>   -------------------------
> diff --git a/hw/i386/meson.build b/hw/i386/meson.build
> index e5d109f..d9d4f40 100644
> --- a/hw/i386/meson.build
> +++ b/hw/i386/meson.build
> @@ -24,6 +24,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
>     'pc_sysfw.c',
>     'acpi-build.c',
>     'port92.c'))
> +i386_ss.add(when: 'CONFIG_MPQEMU', if_true: files('remote.c'))
>   
>   subdir('kvm')
>   subdir('xen')
> diff --git a/hw/i386/remote.c b/hw/i386/remote.c
> new file mode 100644
> index 0000000..a67be33
> --- /dev/null
> +++ b/hw/i386/remote.c
> @@ -0,0 +1,69 @@
> +/*
> + * Machine for remote device
> + *
> + *  This machine type is used by the remote device process in multi-process
> + *  QEMU. QEMU device models depend on parent busses, interrupt controllers,
> + *  memory regions, etc. The remote machine type offers this environment so
> + *  that QEMU device models can be used as remote devices.
> + *
> + * Copyright © 2018, 2020 Oracle and/or its affiliates.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#include "qemu/osdep.h"
> +#include "qemu-common.h"
> +
> +#include "hw/i386/remote.h"
> +#include "exec/address-spaces.h"
> +#include "exec/memory.h"
> +#include "qapi/error.h"
> +
> +static void remote_machine_init(MachineState *machine)
> +{
> +    MemoryRegion *system_memory, *system_io, *pci_memory;
> +    RemoteMachineState *s = REMOTE_MACHINE(machine);
> +    RemotePCIHost *rem_host;
> +
> +    system_memory = get_system_memory();
> +    system_io = get_system_io();
> +
> +    pci_memory = g_new(MemoryRegion, 1);
> +    memory_region_init(pci_memory, NULL, "pci", UINT64_MAX);
> +
> +    rem_host = REMOTE_HOST_DEVICE(qdev_new(TYPE_REMOTE_HOST_DEVICE));
> +
> +    rem_host->mr_pci_mem = pci_memory;
> +    rem_host->mr_sys_mem = system_memory;
> +    rem_host->mr_sys_io = system_io;
> +
> +    s->host = rem_host;
> +
> +    object_property_add_child(OBJECT(s), "remote-device", OBJECT(rem_host));
> +    memory_region_add_subregion_overlap(system_memory, 0x0, pci_memory, -1);
> +
> +    qdev_realize(DEVICE(rem_host), sysbus_get_default(), &error_fatal);
> +}

Nothing is x86-specific in this machine. Why not make it a generic
machine, like the 'none' one?

Re: [PATCH v11 05/19] multi-process: add qio channel function to transmit

[Philippe Mathieu-Daudé]

On 10/15/20 8:04 PM, Jagannathan Raman wrote:
> From: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> 
> The entire array of the memory regions and file handlers.

I don't understand the description... Did you forgot the
verb?

> Will be used in the next patch.
> 
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>   include/io/channel.h | 24 ++++++++++++++++++++++++
>   io/channel.c         | 45 +++++++++++++++++++++++++++++++++++++++++++++
>   2 files changed, 69 insertions(+)

Re: [PATCH v11 02/19] multi-process: Add config option for multi-process QEMU

[Philippe Mathieu-Daudé]

On 10/15/20 8:04 PM, Jagannathan Raman wrote:
> Add a configuration option to separate multi-process code
> 
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>   configure   | 10 ++++++++++
>   meson.build |  1 +
>   2 files changed, 11 insertions(+)
> 
> diff --git a/configure b/configure
> index f839c2a..2025e89 100755
> --- a/configure
> +++ b/configure
> @@ -447,6 +447,7 @@ meson=""
>   ninja=""
>   skip_meson=no
>   gettext=""
> +mpqemu="no"
>   
>   bogus_os="no"
>   malloc_trim="auto"
> @@ -1520,6 +1521,10 @@ for opt do
>     ;;
>     --disable-libdaxctl) libdaxctl=no
>     ;;
> +  --enable-mpqemu) mpqemu=yes
> +  ;;
> +  --disable-mpqemu) mpqemu=no
> +  ;;
>     *)
>         echo "ERROR: unknown option $opt"
>         echo "Try '$0 --help' for more information"
> @@ -1838,6 +1843,8 @@ disabled with --disable-FEATURE, default is enabled if available:
>     xkbcommon       xkbcommon support
>     rng-none        dummy RNG, avoid using /dev/(u)random and getrandom()
>     libdaxctl       libdaxctl support
> +  mpqemu          multi-process QEMU support

When I first heard about this, I was expecting something
else. It could be better named as "remote-device emulation"
or maybe "manifold (device) emulation" :) Anyway too late
I presume.

Re: [PATCH v11 12/19] multi-process: Forward PCI config space acceses to the remote process

[Philippe Mathieu-Daudé]

On 10/15/20 8:05 PM, Jagannathan Raman wrote:
> From: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> 
> The Proxy Object sends the PCI config space accesses as messages
> to the remote process over the communication channel
> 
> TODO:
> Investigate if the local PCI config writes can be dropped.
> Without the proxy local PCI config space writes for the device,
> the driver in the guest times out on the probing.
> We have tried to only refer to the remote for the PCI config writes,
> but the driver timeout in the guest forced as to left this
> as it is (removing local PCI config only).

I expect this TODO to be lost in git history. Better to
commit it in the code and remove it once resolved IMHO.

> 
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>   hw/i386/remote-msg.c     | 62 ++++++++++++++++++++++++++++++++++++++++++++++++
>   hw/pci/proxy.c           | 50 ++++++++++++++++++++++++++++++++++++++
>   include/io/mpqemu-link.h |  9 +++++++
>   io/mpqemu-link.c         |  6 +++++
>   4 files changed, 127 insertions(+)

Please have a look at scripts/git.orderfile, it helps
reviewers to avoid scrolling down/up the patches.

> 
> diff --git a/hw/i386/remote-msg.c b/hw/i386/remote-msg.c
> index 6451b77..94937db 100644
> --- a/hw/i386/remote-msg.c
> +++ b/hw/i386/remote-msg.c
> @@ -15,6 +15,12 @@
>   #include "io/mpqemu-link.h"
>   #include "qapi/error.h"
>   #include "sysemu/runstate.h"
> +#include "hw/pci/pci.h"
> +
> +static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
> +                                 MPQemuMsg *msg);
> +static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
> +                                MPQemuMsg *msg);
>   
>   void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
>   {
> @@ -43,6 +49,12 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
>           }
>   
>           switch (msg.cmd) {
> +        case PCI_CONFIG_WRITE:
> +            process_config_write(com->ioc, pci_dev, &msg);
> +            break;
> +        case PCI_CONFIG_READ:
> +            process_config_read(com->ioc, pci_dev, &msg);
> +            break;
>           default:
>               error_setg(&local_err,
>                          "Unknown command (%d) received for device %s (pid=%d)",
> @@ -60,3 +72,53 @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data)
>   
>       return;
>   }
> +
> +static void process_config_write(QIOChannel *ioc, PCIDevice *dev,
> +                                 MPQemuMsg *msg)
> +{
> +    ConfDataMsg *conf = (ConfDataMsg *)&msg->data.conf_data;
> +    MPQemuMsg ret = { 0 };
> +    Error *local_err = NULL;
> +
> +    if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) {
> +        error_report("Bad address received when writing PCI config, pid %d",
> +                     getpid());
> +        ret.data.u64 = UINT64_MAX;
> +    } else {
> +        pci_default_write_config(dev, conf->addr, conf->val, conf->l);
> +    }
> +
> +    ret.cmd = RET_MSG;
> +    ret.size = sizeof(ret.data.u64);
> +
> +    mpqemu_msg_send(&ret, ioc, &local_err);
> +    if (local_err) {
> +        error_report("Could not send message to proxy from pid %d",
> +                     getpid());
> +    }
> +}
> +
> +static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
> +                                MPQemuMsg *msg)
> +{
> +    ConfDataMsg *conf = (ConfDataMsg *)&msg->data.conf_data;
> +    MPQemuMsg ret = { 0 };
> +    Error *local_err = NULL;
> +
> +    if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) {
> +        error_report("Bad address received when reading PCI config, pid %d",
> +                     getpid());
> +        ret.data.u64 = UINT64_MAX;
> +    } else {
> +        ret.data.u64 = pci_default_read_config(dev, conf->addr, conf->l);

Isn't it endianess issue here? It might makes sense to
declare ConfDataMsg and PCI_CONFIG_READ/PCI_CONFIG_WRITE
packets in little endian.

> +    }
> +
> +    ret.cmd = RET_MSG;
> +    ret.size = sizeof(ret.data.u64);
> +
> +    mpqemu_msg_send(&ret, ioc, &local_err);
> +    if (local_err) {
> +        error_report("Could not send message to proxy from pid %d",
> +                     getpid());
> +    }
> +}
> diff --git a/hw/pci/proxy.c b/hw/pci/proxy.c
> index 7a12f23..27714fe 100644
> --- a/hw/pci/proxy.c
> +++ b/hw/pci/proxy.c
> @@ -16,6 +16,8 @@
>   #include "hw/qdev-properties.h"
>   #include "monitor/monitor.h"
>   #include "migration/blocker.h"
> +#include "io/mpqemu-link.h"
> +#include "qemu/error-report.h"
>   
>   static void proxy_set_socket(PCIProxyDev *pdev, int fd, Error **errp)
>   {
> @@ -69,6 +71,51 @@ static void pci_proxy_dev_exit(PCIDevice *pdev)
>       error_free(dev->migration_blocker);
>   }
>   
> +static int config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val,
> +                          int l, unsigned int op)
> +{
> +    MPQemuMsg msg = { 0 };
> +    uint64_t ret = -EINVAL;
> +    Error *local_err = NULL;
> +
> +    msg.cmd = op;
> +    msg.data.conf_data.addr = addr;
> +    msg.data.conf_data.val = (op == PCI_CONFIG_WRITE) ? *val : 0;
> +    msg.data.conf_data.l = l;
> +    msg.size = sizeof(ConfDataMsg);
> +
> +    ret = mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err);
> +    if (local_err) {
> +        error_report_err(local_err);
> +    }
> +    if (op == PCI_CONFIG_READ) {
> +        *val = (uint32_t)ret;
> +    }
> +
> +    return ret;
> +}
> +
> +static uint32_t pci_proxy_read_config(PCIDevice *d, uint32_t addr, int len)
> +{
> +    uint32_t val;
> +
> +    (void)config_op_send(PCI_PROXY_DEV(d), addr, &val, len, PCI_CONFIG_READ);
> +
> +    return val;
> +}
> +
> +static void pci_proxy_write_config(PCIDevice *d, uint32_t addr, uint32_t val,
> +                                   int l)
> +{
> +    /*
> +     * Some of the functions access the copy of the remote device
> +     * PCI config space, therefore maintain it updated.
> +     */
> +    pci_default_write_config(d, addr, val, l);
> +
> +    (void)config_op_send(PCI_PROXY_DEV(d), addr, &val, l, PCI_CONFIG_WRITE);
> +}
> +
>   static void pci_proxy_dev_class_init(ObjectClass *klass, void *data)
>   {
>       DeviceClass *dc = DEVICE_CLASS(klass);
> @@ -76,6 +123,9 @@ static void pci_proxy_dev_class_init(ObjectClass *klass, void *data)
>   
>       k->realize = pci_proxy_dev_realize;
>       k->exit = pci_proxy_dev_exit;
> +    k->config_read = pci_proxy_read_config;
> +    k->config_write = pci_proxy_write_config;
> +
>       device_class_set_props(dc, proxy_properties);
>   }
>   
> diff --git a/include/io/mpqemu-link.h b/include/io/mpqemu-link.h
> index 4bea5da..459d345 100644
> --- a/include/io/mpqemu-link.h
> +++ b/include/io/mpqemu-link.h
> @@ -34,6 +34,8 @@ typedef enum {
>       MPQEMU_CMD_INIT,
>       SYNC_SYSMEM,
>       RET_MSG,
> +    PCI_CONFIG_WRITE,
> +    PCI_CONFIG_READ,
>       MPQEMU_CMD_MAX,
>   } MPQemuCmd;
>   
> @@ -43,6 +45,12 @@ typedef struct {
>       off_t offsets[REMOTE_MAX_FDS];
>   } SyncSysmemMsg;
>   
> +typedef struct {
> +    uint32_t addr;
> +    uint32_t val;
> +    int l;
> +} ConfDataMsg;

Maybe name this PciConfDataMsg?

> +
>   /**
>    * MPQemuMsg:
>    * @cmd: The remote command
> @@ -60,6 +68,7 @@ typedef struct {
>   
>       union {
>           uint64_t u64;
> +        ConfDataMsg conf_data;
>           SyncSysmemMsg sync_sysmem;
>       } data;
>   
> diff --git a/io/mpqemu-link.c b/io/mpqemu-link.c
> index 1339cc7..5ef82da 100644
> --- a/io/mpqemu-link.c
> +++ b/io/mpqemu-link.c
> @@ -278,6 +278,12 @@ bool mpqemu_msg_valid(MPQemuMsg *msg)
>               return false;
>           }
>           break;
> +    case PCI_CONFIG_WRITE:
> +    case PCI_CONFIG_READ:
> +        if (msg->size != sizeof(ConfDataMsg)) {
> +            return false;
> +        }
> +        break;
>       default:
>           break;
>       }
> 

Re: [PATCH v11 12/19] multi-process: Forward PCI config space acceses to the remote process

[Philippe Mathieu-Daudé]

On 10/23/20 6:59 PM, Philippe Mathieu-Daudé wrote:
> On 10/15/20 8:05 PM, Jagannathan Raman wrote:
>> From: Elena Ufimtseva <elena.ufimtseva@oracle.com>
>>
>> The Proxy Object sends the PCI config space accesses as messages
>> to the remote process over the communication channel
...

>> +static void process_config_read(QIOChannel *ioc, PCIDevice *dev,
>> +                                MPQemuMsg *msg)
>> +{
>> +    ConfDataMsg *conf = (ConfDataMsg *)&msg->data.conf_data;
>> +    MPQemuMsg ret = { 0 };
>> +    Error *local_err = NULL;
>> +
>> +    if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) {
>> +        error_report("Bad address received when reading PCI config, 
>> pid %d",
>> +                     getpid());
>> +        ret.data.u64 = UINT64_MAX;
>> +    } else {
>> +        ret.data.u64 = pci_default_read_config(dev, conf->addr, 
>> conf->l);
> 
> Isn't it endianess issue here? It might makes sense to
> declare ConfDataMsg and PCI_CONFIG_READ/PCI_CONFIG_WRITE
> packets in little endian.

Bah Stefan told me to look at the specs which are included in
the latest patch (docs/multi-process.rst). As I process patches
of this series in order I haven't looked at it yet. Still,
why not put it first in your series?

Regards,

Phil.

Re: [PATCH v11 19/19] multi-process: add configure and usage information

[Philippe Mathieu-Daudé]

Hi Jagannathan,

On 10/15/20 8:05 PM, Jagannathan Raman wrote:
> From: Elena Ufimtseva <elena.ufimtseva@oracle.com>

Documentation is scarce ;)

> 
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>  MAINTAINERS                |  2 ++
>  docs/multi-process.rst     | 67 ++++++++++++++++++++++++++++++++++++++++++++++
>  scripts/mpqemu-launcher.py | 49 +++++++++++++++++++++++++++++++++
>  3 files changed, 118 insertions(+)
>  create mode 100644 docs/multi-process.rst
>  create mode 100755 scripts/mpqemu-launcher.py
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 9a911e0..d12aba7 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -3118,6 +3118,8 @@ F: include/hw/pci/memory-sync.h
>  F: hw/i386/remote-iohub.c
>  F: include/hw/i386/remote-iohub.h
>  F: docs/devel/multi-process.rst
> +F: scripts/mpqemu-launcher.py
> +F: scripts/mpqemu-launcher-perf-mode.py

This one was in v7, Stefan asked about it, then the script
disappeared in v8 =)
https://www.mail-archive.com/qemu-devel@nongnu.org/msg718984.html

>  
>  Build and test automation
>  -------------------------
> diff --git a/docs/multi-process.rst b/docs/multi-process.rst
> new file mode 100644
> index 0000000..c4b022c
> --- /dev/null
> +++ b/docs/multi-process.rst
> @@ -0,0 +1,67 @@
> +Multi-process QEMU
> +==================
> +
> +This document describes how to configure and use multi-process qemu.
> +For the design document refer to docs/devel/qemu-multiprocess.
> +
> +1) Configuration
> +----------------
> +
> +To enable support for multi-process add --enable-mpqemu
> +to the list of options for the "configure" script.
> +
> +
> +2) Usage
> +--------
> +
> +Multi-process QEMU requires an orchestrator to launch. Please refer to a
> +light-weight python based orchestrator for mpqemu in
> +scripts/mpqemu-launcher.py to lauch QEMU in multi-process mode.
> +
> +Following is a description of command-line used to launch mpqemu.
> +
> +* Orchestrator:
> +
> +  - The Orchestrator creates a unix socketpair
> +
> +  - It launches the remote process and passes one of the
> +    sockets to it via command-line.
> +
> +  - It then launches QEMU and specifies the other socket as an option
> +    to the Proxy device object
> +
> +* Remote Process:
> +
> +  - QEMU can enter remote process mode by using the "remote" machine
> +    option.
> +
> +  - The orchestrator creates a "remote-object" with details about
> +    the device and the file descriptor for the device
> +
> +  - The remaining options are no different from how one launches QEMU with
> +    devices.
> +
> +  - Example command-line for the remote process is as follows:
> +
> +      /usr/bin/qemu-system-x86_64                                        \
> +      -machine x-remote                                                  \
> +      -device lsi53c895a,id=lsi0                                         \
> +      -drive id=drive_image2,file=/build/ol7-nvme-test-1.qcow2           \
> +      -device scsi-hd,id=drive2,drive=drive_image2,bus=lsi0.0,scsi-id=0  \
> +      -object x-remote-object,id=robj1,devid=lsi1,fd=4,
> +
> +* QEMU:
> +
> +  - Since parts of the RAM are shared between QEMU & remote process, a
> +    memory-backend-memfd is required to facilitate this, as follows:
> +
> +    -object memory-backend-memfd,id=mem,size=2G
> +
> +  - A "x-pci-proxy-dev" device is created for each of the PCI devices emulated
> +    in the remote process. A "socket" sub-option specifies the other end of
> +    unix channel created by orchestrator. The "id" sub-option must be specified
> +    and should be the same as the "id" specified for the remote PCI device
> +
> +  - Example commandline for QEMU is as follows:
> +
> +      -device x-pci-proxy-dev,id=lsi0,socket=3
> diff --git a/scripts/mpqemu-launcher.py b/scripts/mpqemu-launcher.py
> new file mode 100755
> index 0000000..6e0ef22
> --- /dev/null
> +++ b/scripts/mpqemu-launcher.py
> @@ -0,0 +1,49 @@
> +#!/usr/bin/env python3
> +import socket
> +import os
> +import subprocess
> +import time
> +
> +PROC_QEMU='/usr/bin/qemu-system-x86_64'

If this is a (multiarch) test, then ...

> +
> +proxy, remote = socket.socketpair(socket.AF_UNIX, socket.SOCK_STREAM)
> +
> +remote_cmd = [ PROC_QEMU,                                                      \
> +               '-machine', 'x-remote',                                         \
> +               '-device', 'lsi53c895a,id=lsi1',                                \

... I'd move it to tests/integration/multiproc-x86-lsi53c895a.py ...

> +               '-drive', 'id=drive_image1,file=/build/ol7-nvme-test-1.qcow2',  \

... use avocado.utils.vmimage (see tests/acceptance/boot_linux.py)
to download a prebuilt image, ...

> +               '-device', 'scsi-hd,id=drive1,drive=drive_image1,bus=lsi1.0,'   \
> +                              'scsi-id=0',                                     \
> +               '-object',                                                      \
> +               'x-remote-object,id=robj1,devid=lsi1,fd='+str(remote.fileno()), \
> +               '-nographic',                                                   \
> +             ]
> +
> +proxy_cmd = [ PROC_QEMU,                                                       \
> +              '-name', 'OL7.4',                                                \
> +              '-machine', 'q35,accel=kvm',                                     \
> +              '-smp', 'sockets=1,cores=1,threads=1',                           \
> +              '-m', '2048',                                                    \
> +              '-object', 'memory-backend-memfd,id=sysmem-file,size=2G',        \
> +              '-numa', 'node,memdev=sysmem-file',                              \
> +              '-device', 'virtio-scsi-pci,id=virtio_scsi_pci0',                \
> +              '-drive', 'id=drive_image1,if=none,format=qcow2,'                \
> +                            'file=/home/ol7-hdd-1.qcow2',                      \
> +              '-device', 'scsi-hd,id=image1,drive=drive_image1,'               \
> +                             'bus=virtio_scsi_pci0.0',                         \
> +              '-boot', 'd',                                                    \
> +              '-vnc', ':0',                                                    \
> +              '-device', 'x-pci-proxy-dev,id=lsi1,fd='+str(proxy.fileno()),    \
> +            ]
> +
> +
> +pid = os.fork();
> +
> +if pid:
> +    # In Proxy
> +    print('Launching QEMU with Proxy object');
> +    process = subprocess.Popen(proxy_cmd, pass_fds=[proxy.fileno()])
> +else:
> +    # In remote
> +    print('Launching Remote process');
> +    process = subprocess.Popen(remote_cmd, pass_fds=[remote.fileno()])
> 

... and do something within the guest to be sure MultiProc works :)

Regards,

Phil.

Re: [PATCH v11 05/19] multi-process: add qio channel function to transmit

[Marc-André Lureau]

[-- Attachment #1: Type: text/plain, Size: 3984 bytes --]

On Thu, Oct 15, 2020 at 10:05 PM Jagannathan Raman <jag.raman@oracle.com>
wrote:

> From: Elena Ufimtseva <elena.ufimtseva@oracle.com>
>
> The entire array of the memory regions and file handlers.
> Will be used in the next patch.
>
> Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
> Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
> Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>  include/io/channel.h | 24 ++++++++++++++++++++++++
>  io/channel.c         | 45 +++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 69 insertions(+)
>
> diff --git a/include/io/channel.h b/include/io/channel.h
> index 3c04f0e..6807506 100644
> --- a/include/io/channel.h
> +++ b/include/io/channel.h
> @@ -773,5 +773,29 @@ void qio_channel_set_aio_fd_handler(QIOChannel *ioc,
>                                      IOHandler *io_read,
>                                      IOHandler *io_write,
>                                      void *opaque);
> +/**
> + * qio_channel_writev_full_all:
> + * @ioc: the channel object
> + * @iov: the array of memory regions to write data from
> + * @niov: the length of the @iov array
> + * @fds: an array of file handles to send
> + * @nfds: number of file handles in @fds
> + * @errp: pointer to a NULL-initialized error object
> + *
> + *
> + * Behaves like qio_channel_writev_full but will attempt
> + * to send all data passed (file handles and memory regions).
> + * The function will wait for all requested data
> + * to be written, yielding from the current coroutine
> + * if required.
> + *
> + * Returns: 0 if all bytes were written, or -1 on error
> + */
> +
> +int qio_channel_writev_full_all(QIOChannel *ioc,
> +                           const struct iovec *iov,
> +                           size_t niov,
> +                           int *fds, size_t nfds,
> +                           Error **errp);
>
>  #endif /* QIO_CHANNEL_H */
> diff --git a/io/channel.c b/io/channel.c
> index e4376eb..22c10c5 100644
> --- a/io/channel.c
> +++ b/io/channel.c
> @@ -190,6 +190,51 @@ int qio_channel_writev_all(QIOChannel *ioc,
>      return ret;
>  }
>
> +int qio_channel_writev_full_all(QIOChannel *ioc,
> +                                const struct iovec *iov,
> +                                size_t niov,
> +                                int *fds, size_t nfds,
> +                                Error **errp)
> +{
>

Please make qio_channel_writev_all() call qio_channel_writev_full_all() to
avoid logic duplication.

thanks

+    int ret = -1;
> +    struct iovec *local_iov = g_new(struct iovec, niov);
> +    struct iovec *local_iov_head = local_iov;
> +    unsigned int nlocal_iov = niov;
> +
> +    nlocal_iov = iov_copy(local_iov, nlocal_iov,
> +                          iov, niov,
> +                          0, iov_size(iov, niov));
> +
> +    while (nlocal_iov > 0) {
> +        ssize_t len;
> +        len = qio_channel_writev_full(ioc, local_iov, nlocal_iov, fds,
> +                                      nfds, errp);
> +        if (len == QIO_CHANNEL_ERR_BLOCK) {
> +            if (qemu_in_coroutine()) {
> +                qio_channel_yield(ioc, G_IO_OUT);
> +            } else {
> +                qio_channel_wait(ioc, G_IO_OUT);
> +            }
> +            continue;
> +        }
> +        if (len < 0) {
> +            goto cleanup;
> +        }
> +
> +        iov_discard_front(&local_iov, &nlocal_iov, len);
> +
> +        if (len > 0) {
> +            fds = NULL;
> +            nfds = 0;
> +        }
> +    }
> +
> +    ret = 0;
> + cleanup:
> +    g_free(local_iov_head);
> +    return ret;
> +}
> +
>  ssize_t qio_channel_readv(QIOChannel *ioc,
>                            const struct iovec *iov,
>                            size_t niov,
> --
> 1.8.3.1
>
>

-- 
Marc-André Lureau

[-- Attachment #2: Type: text/html, Size: 5493 bytes --]