[Qemu-devel] [PATCH qemu v6 00/15] spapr: vfio: Enable Dynamic DMA windows (DDW)

[Qemu-devel] [PATCH qemu v6 00/15] spapr: vfio: Enable Dynamic DMA windows (DDW)

[Alexey Kardashevskiy]

(cut-n-paste from kernel patchset)

Each Partitionable Endpoint (IOMMU group) has an address range on a PCI bus
where devices are allowed to do DMA. These ranges are called DMA windows.
By default, there is a single DMA window, 1 or 2GB big, mapped at zero
on a PCI bus.

PAPR defines a DDW RTAS API which allows pseries guests
querying the hypervisor about DDW support and capabilities (page size mask
for now). A pseries guest may request an additional (to the default)
DMA windows using this RTAS API.
The existing pseries Linux guests request an additional window as big as
the guest RAM and map the entire guest window which effectively creates
direct mapping of the guest memory to a PCI bus.

This patchset reworks PPC64 IOMMU code and adds necessary structures
to support big windows.

Once a Linux guest discovers the presence of DDW, it does:
1. query hypervisor about number of available windows and page size masks;
2. create a window with the biggest possible page size (today 4K/64K/16M);
3. map the entire guest RAM via H_PUT_TCE* hypercalls;
4. switche dma_ops to direct_dma_ops on the selected PE.

Once this is done, H_PUT_TCE is not called anymore for 64bit devices and
the guest does not waste time on DMA map/unmap operations.

Note that 32bit devices won't use DDW and will keep using the default
DMA window so KVM optimizations will be required (to be posted later).

This patchset adds DDW support for pseries. The host kernel changes are
required, posted as:

[PATCH kernel v8 00/31] powerpc/iommu/vfio: Enable Dynamic DMA windows

This patchset is based on git://github.com/dgibson/qemu.git spapr-next branch.

TODO: fix commit logss :)

Please comment. Thanks!

Changes:
v6:
* spapr-pci-vfio-host-bridge is now a synonim of spapr-pci-host-bridge -
same PHB can host emulated and VFIO devices
* changed patches order
* lot of small changes

v5:
* TCE tables got "enabled" state and are persistent, i.e. not recreated
every reboot
* added v2 of SPAPR_TCE_IOMMU
* fixed migration for emulated PHB with enabled DDW
* huge pile of other changes

v4:
* reimplemented the whole thing
* machine reset and ddw-reset RTAS call both remove all TCE tables and
create the default one
* IOMMU group id is not needed to use VFIO PHB anymore, multiple groups
are supported on the same VFIO container and virtual PHB

v3:
* removed "reset" from API now
* reworked machine versions
* applied multiple comments
* includes David's machine QOM rework as this patchset adds a new machine type

v2:
* tested on emulated PHB
* removed "ddw" machine property, now it is PHB property
* disabled by default
* defined "pseries-2.2" machine which enables DDW by default
* fixed reset() and reference counting




Alexey Kardashevskiy (15):
  linux headers update for DDW on SPAPR
  vmstate: Define VARRAY with VMS_ALLOC
  spapr_pci: Make find_phb()/find_dev() public
  vfio: spapr: Move SPAPR-related code to a separate file
  spapr_pci_vfio: Enable multiple groups per container
  spapr_pci: Convert finish_realize() to
    dma_capabilities_update()+dma_init_window()
  vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)
  spapr_iommu: Introduce "enabled" state for TCE table
  spapr_iommu: Add root memory region
  spapr_pci_vfio: Remove unnecessary cast to sPAPRPHBVFIOState
  spapr_pci: Do complete reset of DMA config when resetting PHB
  spapr_vfio_pci: Remove redundant spapr-pci-vfio-host-bridge
  spapr: Add pseries-2.4 machine
  spapr_pci/spapr_pci_vfio: Support Dynamic DMA Windows (DDW)
  vfio: Enable DDW ioctls to VFIO IOMMU driver

 hw/ppc/Makefile.objs          |   3 +
 hw/ppc/spapr.c                |  32 ++++-
 hw/ppc/spapr_iommu.c          | 144 ++++++++++++++------
 hw/ppc/spapr_pci.c            | 209 ++++++++++++++++++----------
 hw/ppc/spapr_pci_vfio.c       | 147 ++++++++++++--------
 hw/ppc/spapr_rtas_ddw.c       | 300 ++++++++++++++++++++++++++++++++++++++++
 hw/ppc/spapr_vio.c            |   9 +-
 hw/vfio/Makefile.objs         |   1 +
 hw/vfio/common.c              | 180 +++++-------------------
 hw/vfio/spapr.c               | 310 ++++++++++++++++++++++++++++++++++++++++++
 include/hw/pci-host/spapr.h   |  49 +++++--
 include/hw/ppc/spapr.h        |  30 +++-
 include/hw/vfio/vfio-common.h |  16 +++
 include/hw/vfio/vfio.h        |   2 +-
 include/migration/vmstate.h   |  10 ++
 linux-headers/linux/vfio.h    |  88 +++++++++++-
 trace-events                  |   5 +
 17 files changed, 1187 insertions(+), 348 deletions(-)
 create mode 100644 hw/ppc/spapr_rtas_ddw.c
 create mode 100644 hw/vfio/spapr.c

-- 
2.0.0

[Qemu-devel] [PATCH qemu v6 01/15] linux headers update for DDW on SPAPR

[Alexey Kardashevskiy]

Since the changes are not in upstream yet, no tag or branch is specified here.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 linux-headers/linux/vfio.h | 88 ++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 85 insertions(+), 3 deletions(-)

diff --git a/linux-headers/linux/vfio.h b/linux-headers/linux/vfio.h
index 95ba870..ce05371 100644
--- a/linux-headers/linux/vfio.h
+++ b/linux-headers/linux/vfio.h
@@ -36,6 +36,8 @@
 /* Two-stage IOMMU */
 #define VFIO_TYPE1_NESTING_IOMMU	6	/* Implies v2 */
 
+#define VFIO_SPAPR_TCE_v2_IOMMU		7
+
 /*
  * The IOCTL interface is designed for extensibility by embedding the
  * structure length (argsz) and flags into structures passed between
@@ -441,6 +443,23 @@ struct vfio_iommu_type1_dma_unmap {
 /* -------- Additional API for SPAPR TCE (Server POWERPC) IOMMU -------- */
 
 /*
+ * The SPAPR TCE DDW info struct provides the information about
+ * the details of Dynamic DMA window capability.
+ *
+ * @pgsizes contains a page size bitmask, 4K/64K/16M are supported.
+ * @max_dynamic_windows_supported tells the maximum number of windows
+ * which the platform can create.
+ * @levels tells the maximum number of levels in multi-level IOMMU tables;
+ * this allows splitting a table into smaller chunks which reduces
+ * the amount of physically contiguous memory required for the table.
+ */
+struct vfio_iommu_spapr_tce_ddw_info {
+	__u64 pgsizes;			/* Bitmap of supported page sizes */
+	__u32 max_dynamic_windows_supported;
+	__u32 levels;
+};
+
+/*
  * The SPAPR TCE info struct provides the information about the PCI bus
  * address ranges available for DMA, these values are programmed into
  * the hardware so the guest has to know that information.
@@ -450,14 +469,17 @@ struct vfio_iommu_type1_dma_unmap {
  * addresses too so the window works as a filter rather than an offset
  * for IOVA addresses.
  *
- * A flag will need to be added if other page sizes are supported,
- * so as defined here, it is always 4k.
+ * Flags supported:
+ * - VFIO_IOMMU_SPAPR_INFO_DDW: informs the userspace that dynamic DMA windows
+ *   (DDW) support is present. @ddw is only supported when DDW is present.
  */
 struct vfio_iommu_spapr_tce_info {
 	__u32 argsz;
-	__u32 flags;			/* reserved for future use */
+	__u32 flags;
+#define VFIO_IOMMU_SPAPR_INFO_DDW	(1 << 0)	/* DDW supported */
 	__u32 dma32_window_start;	/* 32 bit window start (bytes) */
 	__u32 dma32_window_size;	/* 32 bit window size (bytes) */
+	struct vfio_iommu_spapr_tce_ddw_info ddw;
 };
 
 #define VFIO_IOMMU_SPAPR_TCE_GET_INFO	_IO(VFIO_TYPE, VFIO_BASE + 12)
@@ -493,6 +515,66 @@ struct vfio_eeh_pe_op {
 
 #define VFIO_EEH_PE_OP			_IO(VFIO_TYPE, VFIO_BASE + 21)
 
+/**
+ * VFIO_IOMMU_SPAPR_REGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 17, struct vfio_iommu_spapr_register_memory)
+ *
+ * Registers user space memory where DMA is allowed. It pins
+ * user pages and does the locked memory accounting so
+ * subsequent VFIO_IOMMU_MAP_DMA/VFIO_IOMMU_UNMAP_DMA calls
+ * get faster.
+ */
+struct vfio_iommu_spapr_register_memory {
+	__u32	argsz;
+	__u32	flags;
+	__u64	vaddr;				/* Process virtual address */
+	__u64	size;				/* Size of mapping (bytes) */
+};
+#define VFIO_IOMMU_SPAPR_REGISTER_MEMORY	_IO(VFIO_TYPE, VFIO_BASE + 17)
+
+/**
+ * VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 18, struct vfio_iommu_spapr_register_memory)
+ *
+ * Unregisters user space memory registered with
+ * VFIO_IOMMU_SPAPR_REGISTER_MEMORY.
+ * Uses vfio_iommu_spapr_register_memory for parameters.
+ */
+#define VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY	_IO(VFIO_TYPE, VFIO_BASE + 18)
+
+/**
+ * VFIO_IOMMU_SPAPR_TCE_CREATE - _IOWR(VFIO_TYPE, VFIO_BASE + 19, struct vfio_iommu_spapr_tce_create)
+ *
+ * Creates an additional TCE table and programs it (sets a new DMA window)
+ * to every IOMMU group in the container. It receives page shift, window
+ * size and number of levels in the TCE table being created.
+ *
+ * It allocates and returns an offset on a PCI bus of the new DMA window.
+ */
+struct vfio_iommu_spapr_tce_create {
+	__u32 argsz;
+	__u32 flags;
+	/* in */
+	__u32 page_shift;
+	__u64 window_size;
+	__u32 levels;
+	/* out */
+	__u64 start_addr;
+};
+#define VFIO_IOMMU_SPAPR_TCE_CREATE	_IO(VFIO_TYPE, VFIO_BASE + 19)
+
+/**
+ * VFIO_IOMMU_SPAPR_TCE_REMOVE - _IOW(VFIO_TYPE, VFIO_BASE + 20, struct vfio_iommu_spapr_tce_remove)
+ *
+ * Unprograms a TCE table from all groups in the container and destroys it.
+ * It receives a PCI bus offset as a window id.
+ */
+struct vfio_iommu_spapr_tce_remove {
+	__u32 argsz;
+	__u32 flags;
+	/* in */
+	__u64 start_addr;
+};
+#define VFIO_IOMMU_SPAPR_TCE_REMOVE	_IO(VFIO_TYPE, VFIO_BASE + 20)
+
 /* ***************************************************************** */
 
 #endif /* VFIO_H */
-- 
2.0.0

[Qemu-devel] [PATCH qemu v6 02/15] vmstate: Define VARRAY with VMS_ALLOC

[Alexey Kardashevskiy]

This allows dynamic allocation for migrating arrays.

Already existing VMSTATE_VARRAY_UINT32 requires an array to be
pre-allocated, however there are cases when the size is not known in
advance and there is no real need to enforce it.

This defines another variant of VMSTATE_VARRAY_UINT32 with WMS_ALLOC
flag which tells the receiving side to allocate memory for the array
before receiving the data.

The first user of it is a dynamic DMA window which existence and size
are totally dynamic.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 include/migration/vmstate.h | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/include/migration/vmstate.h b/include/migration/vmstate.h
index bc7616a..73b9d67 100644
--- a/include/migration/vmstate.h
+++ b/include/migration/vmstate.h
@@ -299,6 +299,16 @@ extern const VMStateInfo vmstate_info_bitmap;
     .offset     = vmstate_offset_pointer(_state, _field, _type),     \
 }
 
+#define VMSTATE_VARRAY_UINT32_ALLOC(_field, _state, _field_num, _version, _info, _type) {\
+    .name       = (stringify(_field)),                               \
+    .version_id = (_version),                                        \
+    .num_offset = vmstate_offset_value(_state, _field_num, uint32_t),\
+    .info       = &(_info),                                          \
+    .size       = sizeof(_type),                                     \
+    .flags      = VMS_VARRAY_UINT32|VMS_POINTER|VMS_ALLOC,           \
+    .offset     = vmstate_offset_pointer(_state, _field, _type),     \
+}
+
 #define VMSTATE_VARRAY_UINT16_UNSAFE(_field, _state, _field_num, _version, _info, _type) {\
     .name       = (stringify(_field)),                               \
     .version_id = (_version),                                        \
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 02/15] vmstate: Define VARRAY with VMS_ALLOC

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 914 bytes --]

On Sat, Apr 11, 2015 at 01:24:31AM +1000, Alexey Kardashevskiy wrote:
> This allows dynamic allocation for migrating arrays.
> 
> Already existing VMSTATE_VARRAY_UINT32 requires an array to be
> pre-allocated, however there are cases when the size is not known in
> advance and there is no real need to enforce it.
> 
> This defines another variant of VMSTATE_VARRAY_UINT32 with WMS_ALLOC
> flag which tells the receiving side to allocate memory for the array
> before receiving the data.
> 
> The first user of it is a dynamic DMA window which existence and size
> are totally dynamic.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

[Qemu-devel] [PATCH qemu v6 03/15] spapr_pci: Make find_phb()/find_dev() public

[Alexey Kardashevskiy]

This makes find_phb()/find_dev() public and changed its names
to spapr_pci_find_phb()/spapr_pci_find_dev() as they are going to
be used from other parts of QEMU such as VFIO DDW (dynamic DMA window)
or VFIO PCI error injection or VFIO EEH handling - in all these
cases there are RTAS calls which are addressed to BUID+config_addr
in IEEE1275 format.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
---
 hw/ppc/spapr_pci.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index 609a8ae..52c5c73 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -426,7 +426,7 @@ static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
     addr = rtas_ld(args, 0);
     option = rtas_ld(args, 3);
 
-    sphb = find_phb(spapr, buid);
+    sphb = spapr_pci_find_phb(spapr, buid);
     if (!sphb) {
         goto param_error_exit;
     }
@@ -461,7 +461,7 @@ static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
     }
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
-    sphb = find_phb(spapr, buid);
+    sphb = spapr_pci_find_phb(spapr, buid);
     if (!sphb) {
         goto param_error_exit;
     }
@@ -479,7 +479,7 @@ static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
     switch (option) {
     case RTAS_GET_PE_ADDR:
         addr = rtas_ld(args, 0);
-        pdev = find_dev(spapr, buid, addr);
+        pdev = spapr_pci_find_dev(spapr, buid, addr);
         if (!pdev) {
             goto param_error_exit;
         }
@@ -516,7 +516,7 @@ static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
     }
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
-    sphb = find_phb(spapr, buid);
+    sphb = spapr_pci_find_phb(spapr, buid);
     if (!sphb) {
         goto param_error_exit;
     }
@@ -562,7 +562,7 @@ static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
     option = rtas_ld(args, 3);
-    sphb = find_phb(spapr, buid);
+    sphb = spapr_pci_find_phb(spapr, buid);
     if (!sphb) {
         goto param_error_exit;
     }
@@ -596,7 +596,7 @@ static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
     }
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
-    sphb = find_phb(spapr, buid);
+    sphb = spapr_pci_find_phb(spapr, buid);
     if (!sphb) {
         goto param_error_exit;
     }
@@ -631,7 +631,7 @@ static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
     }
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
-    sphb = find_phb(spapr, buid);
+    sphb = spapr_pci_find_phb(spapr, buid);
     if (!sphb) {
         goto param_error_exit;
     }
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 03/15] spapr_pci: Make find_phb()/find_dev() public

[Michael Roth]

Quoting Alexey Kardashevskiy (2015-04-10 10:24:32)
> This makes find_phb()/find_dev() public and changed its names
> to spapr_pci_find_phb()/spapr_pci_find_dev() as they are going to
> be used from other parts of QEMU such as VFIO DDW (dynamic DMA window)
> or VFIO PCI error injection or VFIO EEH handling - in all these
> cases there are RTAS calls which are addressed to BUID+config_addr
> in IEEE1275 format.

ee954280da8d9ea7afd28b4c288da15fb8b4efca introduced more find_dev/find_phb
callers that didn't seem to get covered by the conversions here.

Noticed this was staged in spapr-next already, so may want to squash those
in.

> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> ---
>  hw/ppc/spapr_pci.c | 14 +++++++-------
>  1 file changed, 7 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> index 609a8ae..52c5c73 100644
> --- a/hw/ppc/spapr_pci.c
> +++ b/hw/ppc/spapr_pci.c
> @@ -426,7 +426,7 @@ static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
>      addr = rtas_ld(args, 0);
>      option = rtas_ld(args, 3);
> 
> -    sphb = find_phb(spapr, buid);
> +    sphb = spapr_pci_find_phb(spapr, buid);
>      if (!sphb) {
>          goto param_error_exit;
>      }
> @@ -461,7 +461,7 @@ static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
>      }
> 
>      buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
> -    sphb = find_phb(spapr, buid);
> +    sphb = spapr_pci_find_phb(spapr, buid);
>      if (!sphb) {
>          goto param_error_exit;
>      }
> @@ -479,7 +479,7 @@ static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
>      switch (option) {
>      case RTAS_GET_PE_ADDR:
>          addr = rtas_ld(args, 0);
> -        pdev = find_dev(spapr, buid, addr);
> +        pdev = spapr_pci_find_dev(spapr, buid, addr);
>          if (!pdev) {
>              goto param_error_exit;
>          }
> @@ -516,7 +516,7 @@ static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
>      }
> 
>      buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
> -    sphb = find_phb(spapr, buid);
> +    sphb = spapr_pci_find_phb(spapr, buid);
>      if (!sphb) {
>          goto param_error_exit;
>      }
> @@ -562,7 +562,7 @@ static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
> 
>      buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
>      option = rtas_ld(args, 3);
> -    sphb = find_phb(spapr, buid);
> +    sphb = spapr_pci_find_phb(spapr, buid);
>      if (!sphb) {
>          goto param_error_exit;
>      }
> @@ -596,7 +596,7 @@ static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
>      }
> 
>      buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
> -    sphb = find_phb(spapr, buid);
> +    sphb = spapr_pci_find_phb(spapr, buid);
>      if (!sphb) {
>          goto param_error_exit;
>      }
> @@ -631,7 +631,7 @@ static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
>      }
> 
>      buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
> -    sphb = find_phb(spapr, buid);
> +    sphb = spapr_pci_find_phb(spapr, buid);
>      if (!sphb) {
>          goto param_error_exit;
>      }
> -- 
> 2.0.0

[Qemu-devel] [PATCH qemu v6 04/15] vfio: spapr: Move SPAPR-related code to a separate file

[Alexey Kardashevskiy]

This moves SPAPR bits to a separate file to avoid pollution of x86 code.

This is a mechanical patch.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
---
 hw/vfio/Makefile.objs         |   1 +
 hw/vfio/common.c              | 134 ++-----------------------
 hw/vfio/spapr.c               | 226 ++++++++++++++++++++++++++++++++++++++++++
 include/hw/vfio/vfio-common.h |  13 +++
 4 files changed, 246 insertions(+), 128 deletions(-)
 create mode 100644 hw/vfio/spapr.c

diff --git a/hw/vfio/Makefile.objs b/hw/vfio/Makefile.objs
index e31f30e..b987ffb 100644
--- a/hw/vfio/Makefile.objs
+++ b/hw/vfio/Makefile.objs
@@ -1,4 +1,5 @@
 ifeq ($(CONFIG_LINUX), y)
 obj-$(CONFIG_SOFTMMU) += common.o
 obj-$(CONFIG_PCI) += pci.o
+obj-$(CONFIG_SOFTMMU) += spapr.o
 endif
diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index b012620..3e4c685 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -190,8 +190,8 @@ const MemoryRegionOps vfio_region_ops = {
 /*
  * DMA - Mapping and unmapping for the "type1" IOMMU interface used on x86
  */
-static int vfio_dma_unmap(VFIOContainer *container,
-                          hwaddr iova, ram_addr_t size)
+int vfio_dma_unmap(VFIOContainer *container,
+                   hwaddr iova, ram_addr_t size)
 {
     struct vfio_iommu_type1_dma_unmap unmap = {
         .argsz = sizeof(unmap),
@@ -208,8 +208,8 @@ static int vfio_dma_unmap(VFIOContainer *container,
     return 0;
 }
 
-static int vfio_dma_map(VFIOContainer *container, hwaddr iova,
-                        ram_addr_t size, void *vaddr, bool readonly)
+int vfio_dma_map(VFIOContainer *container, hwaddr iova,
+                 ram_addr_t size, void *vaddr, bool readonly)
 {
     struct vfio_iommu_type1_dma_map map = {
         .argsz = sizeof(map),
@@ -238,7 +238,7 @@ static int vfio_dma_map(VFIOContainer *container, hwaddr iova,
     return -errno;
 }
 
-static bool vfio_listener_skipped_section(MemoryRegionSection *section)
+bool vfio_listener_skipped_section(MemoryRegionSection *section)
 {
     return (!memory_region_is_ram(section->mr) &&
             !memory_region_is_iommu(section->mr)) ||
@@ -251,64 +251,6 @@ static bool vfio_listener_skipped_section(MemoryRegionSection *section)
            section->offset_within_address_space & (1ULL << 63);
 }
 
-static void vfio_iommu_map_notify(Notifier *n, void *data)
-{
-    VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
-    VFIOContainer *container = giommu->container;
-    IOMMUTLBEntry *iotlb = data;
-    MemoryRegion *mr;
-    hwaddr xlat;
-    hwaddr len = iotlb->addr_mask + 1;
-    void *vaddr;
-    int ret;
-
-    trace_vfio_iommu_map_notify(iotlb->iova,
-                                iotlb->iova + iotlb->addr_mask);
-
-    /*
-     * The IOMMU TLB entry we have just covers translation through
-     * this IOMMU to its immediate target.  We need to translate
-     * it the rest of the way through to memory.
-     */
-    mr = address_space_translate(&address_space_memory,
-                                 iotlb->translated_addr,
-                                 &xlat, &len, iotlb->perm & IOMMU_WO);
-    if (!memory_region_is_ram(mr)) {
-        error_report("iommu map to non memory area %"HWADDR_PRIx"",
-                     xlat);
-        return;
-    }
-    /*
-     * Translation truncates length to the IOMMU page size,
-     * check that it did not truncate too much.
-     */
-    if (len & iotlb->addr_mask) {
-        error_report("iommu has granularity incompatible with target AS");
-        return;
-    }
-
-    if ((iotlb->perm & IOMMU_RW) != IOMMU_NONE) {
-        vaddr = memory_region_get_ram_ptr(mr) + xlat;
-        ret = vfio_dma_map(container, iotlb->iova,
-                           iotlb->addr_mask + 1, vaddr,
-                           !(iotlb->perm & IOMMU_WO) || mr->readonly);
-        if (ret) {
-            error_report("vfio_dma_map(%p, 0x%"HWADDR_PRIx", "
-                         "0x%"HWADDR_PRIx", %p) = %d (%m)",
-                         container, iotlb->iova,
-                         iotlb->addr_mask + 1, vaddr, ret);
-        }
-    } else {
-        ret = vfio_dma_unmap(container, iotlb->iova, iotlb->addr_mask + 1);
-        if (ret) {
-            error_report("vfio_dma_unmap(%p, 0x%"HWADDR_PRIx", "
-                         "0x%"HWADDR_PRIx") = %d (%m)",
-                         container, iotlb->iova,
-                         iotlb->addr_mask + 1, ret);
-        }
-    }
-}
-
 static void vfio_listener_region_add(MemoryListener *listener,
                                      MemoryRegionSection *section)
 {
@@ -344,45 +286,6 @@ static void vfio_listener_region_add(MemoryListener *listener,
 
     memory_region_ref(section->mr);
 
-    if (memory_region_is_iommu(section->mr)) {
-        VFIOGuestIOMMU *giommu;
-
-        trace_vfio_listener_region_add_iommu(iova,
-                    int128_get64(int128_sub(llend, int128_one())));
-        /*
-         * FIXME: We should do some checking to see if the
-         * capabilities of the host VFIO IOMMU are adequate to model
-         * the guest IOMMU
-         *
-         * FIXME: For VFIO iommu types which have KVM acceleration to
-         * avoid bouncing all map/unmaps through qemu this way, this
-         * would be the right place to wire that up (tell the KVM
-         * device emulation the VFIO iommu handles to use).
-         */
-        /*
-         * This assumes that the guest IOMMU is empty of
-         * mappings at this point.
-         *
-         * One way of doing this is:
-         * 1. Avoid sharing IOMMUs between emulated devices or different
-         * IOMMU groups.
-         * 2. Implement VFIO_IOMMU_ENABLE in the host kernel to fail if
-         * there are some mappings in IOMMU.
-         *
-         * VFIO on SPAPR does that. Other IOMMU models may do that different,
-         * they must make sure there are no existing mappings or
-         * loop through existing mappings to map them into VFIO.
-         */
-        giommu = g_malloc0(sizeof(*giommu));
-        giommu->iommu = section->mr;
-        giommu->container = container;
-        giommu->n.notify = vfio_iommu_map_notify;
-        QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
-        memory_region_register_iommu_notifier(giommu->iommu, &giommu->n);
-
-        return;
-    }
-
     /* Here we assume that memory_region_is_ram(section->mr)==true */
 
     end = int128_get64(llend);
@@ -435,27 +338,6 @@ static void vfio_listener_region_del(MemoryListener *listener,
         return;
     }
 
-    if (memory_region_is_iommu(section->mr)) {
-        VFIOGuestIOMMU *giommu;
-
-        QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) {
-            if (giommu->iommu == section->mr) {
-                memory_region_unregister_iommu_notifier(&giommu->n);
-                QLIST_REMOVE(giommu, giommu_next);
-                g_free(giommu);
-                break;
-            }
-        }
-
-        /*
-         * FIXME: We assume the one big unmap below is adequate to
-         * remove any individual page mappings in the IOMMU which
-         * might have been copied into VFIO. This works for a page table
-         * based IOMMU where a big unmap flattens a large range of IO-PTEs.
-         * That may not be true for all IOMMU types.
-         */
-    }
-
     iova = TARGET_PAGE_ALIGN(section->offset_within_address_space);
     end = (section->offset_within_address_space + int128_get64(section->size)) &
           TARGET_PAGE_MASK;
@@ -721,11 +603,7 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
             goto free_container_exit;
         }
 
-        container->iommu_data.type1.listener = vfio_memory_listener;
-        container->iommu_data.release = vfio_listener_release;
-
-        memory_listener_register(&container->iommu_data.type1.listener,
-                                 container->space->as);
+        spapr_memory_listener_register(container);
 
     } else {
         error_report("vfio: No available IOMMU models");
diff --git a/hw/vfio/spapr.c b/hw/vfio/spapr.c
new file mode 100644
index 0000000..5f79194
--- /dev/null
+++ b/hw/vfio/spapr.c
@@ -0,0 +1,226 @@
+/*
+ * QEMU sPAPR VFIO IOMMU
+ *
+ * Copyright (c) 2015 Alexey Kardashevskiy, IBM Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License,
+ *  or (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "hw/vfio/vfio-common.h"
+#include "qemu/error-report.h"
+#include "trace.h"
+
+static void vfio_iommu_map_notify(Notifier *n, void *data)
+{
+    VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
+    VFIOContainer *container = giommu->container;
+    IOMMUTLBEntry *iotlb = data;
+    MemoryRegion *mr;
+    hwaddr xlat;
+    hwaddr len = iotlb->addr_mask + 1;
+    void *vaddr;
+    int ret;
+
+    trace_vfio_iommu_map_notify(iotlb->iova,
+                                iotlb->iova + iotlb->addr_mask);
+
+    /*
+     * The IOMMU TLB entry we have just covers translation through
+     * this IOMMU to its immediate target.  We need to translate
+     * it the rest of the way through to memory.
+     */
+    mr = address_space_translate(&address_space_memory,
+                                 iotlb->translated_addr,
+                                 &xlat, &len, iotlb->perm & IOMMU_WO);
+    if (!memory_region_is_ram(mr)) {
+        error_report("iommu map to non memory area %"HWADDR_PRIx,
+                     xlat);
+        return;
+    }
+    /*
+     * Translation truncates length to the IOMMU page size,
+     * check that it did not truncate too much.
+     */
+    if (len & iotlb->addr_mask) {
+        error_report("iommu has granularity incompatible with target AS");
+        return;
+    }
+
+    if ((iotlb->perm & IOMMU_RW) != IOMMU_NONE) {
+        vaddr = memory_region_get_ram_ptr(mr) + xlat;
+        ret = vfio_dma_map(container, iotlb->iova,
+                           iotlb->addr_mask + 1, vaddr,
+                           !(iotlb->perm & IOMMU_WO) || mr->readonly);
+        if (ret) {
+            error_report("vfio_dma_map(%p, 0x%"HWADDR_PRIx", "
+                         "0x%"HWADDR_PRIx", %p) = %d (%m)",
+                         container, iotlb->iova,
+                         iotlb->addr_mask + 1, vaddr, ret);
+        }
+    } else {
+        ret = vfio_dma_unmap(container, iotlb->iova, iotlb->addr_mask + 1);
+        if (ret) {
+            error_report("vfio_dma_unmap(%p, 0x%"HWADDR_PRIx", "
+                         "0x%"HWADDR_PRIx") = %d (%m)",
+                         container, iotlb->iova,
+                         iotlb->addr_mask + 1, ret);
+        }
+    }
+}
+
+static void vfio_spapr_listener_region_add(MemoryListener *listener,
+                                     MemoryRegionSection *section)
+{
+    VFIOContainer *container = container_of(listener, VFIOContainer,
+                                            iommu_data.spapr.listener);
+    hwaddr iova;
+    Int128 llend;
+    VFIOGuestIOMMU *giommu;
+
+    if (vfio_listener_skipped_section(section)) {
+        trace_vfio_listener_region_add_skip(
+            section->offset_within_address_space,
+            section->offset_within_address_space +
+            int128_get64(int128_sub(section->size, int128_one())));
+        return;
+    }
+
+    if (unlikely((section->offset_within_address_space & ~TARGET_PAGE_MASK) !=
+                 (section->offset_within_region & ~TARGET_PAGE_MASK))) {
+        error_report("%s received unaligned region", __func__);
+        return;
+    }
+
+    iova = TARGET_PAGE_ALIGN(section->offset_within_address_space);
+    llend = int128_make64(section->offset_within_address_space);
+    llend = int128_add(llend, section->size);
+    llend = int128_and(llend, int128_exts64(TARGET_PAGE_MASK));
+
+    if (int128_ge(int128_make64(iova), llend)) {
+        return;
+    }
+
+    memory_region_ref(section->mr);
+
+    trace_vfio_listener_region_add_iommu(iova,
+         int128_get64(int128_sub(llend, int128_one())));
+    /*
+     * FIXME: We should do some checking to see if the
+     * capabilities of the host VFIO IOMMU are adequate to model
+     * the guest IOMMU
+     *
+     * FIXME: For VFIO iommu types which have KVM acceleration to
+     * avoid bouncing all map/unmaps through qemu this way, this
+     * would be the right place to wire that up (tell the KVM
+     * device emulation the VFIO iommu handles to use).
+     */
+    /*
+     * This assumes that the guest IOMMU is empty of
+     * mappings at this point.
+     *
+     * One way of doing this is:
+     * 1. Avoid sharing IOMMUs between emulated devices or different
+     * IOMMU groups.
+     * 2. Implement VFIO_IOMMU_ENABLE in the host kernel to fail if
+     * there are some mappings in IOMMU.
+     *
+     * VFIO on SPAPR does that. Other IOMMU models may do that different,
+     * they must make sure there are no existing mappings or
+     * loop through existing mappings to map them into VFIO.
+     */
+    giommu = g_malloc0(sizeof(*giommu));
+    giommu->iommu = section->mr;
+    giommu->container = container;
+    giommu->n.notify = vfio_iommu_map_notify;
+    QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
+    memory_region_register_iommu_notifier(giommu->iommu, &giommu->n);
+}
+
+static void vfio_spapr_listener_region_del(MemoryListener *listener,
+                                     MemoryRegionSection *section)
+{
+    VFIOContainer *container = container_of(listener, VFIOContainer,
+                                            iommu_data.spapr.listener);
+    hwaddr iova, end;
+    int ret;
+    VFIOGuestIOMMU *giommu;
+
+    if (vfio_listener_skipped_section(section)) {
+        trace_vfio_listener_region_del_skip(
+            section->offset_within_address_space,
+            section->offset_within_address_space +
+            int128_get64(int128_sub(section->size, int128_one())));
+        return;
+    }
+
+    if (unlikely((section->offset_within_address_space & ~TARGET_PAGE_MASK) !=
+                 (section->offset_within_region & ~TARGET_PAGE_MASK))) {
+        error_report("%s received unaligned region", __func__);
+        return;
+    }
+
+    QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) {
+        if (giommu->iommu == section->mr) {
+            memory_region_unregister_iommu_notifier(&giommu->n);
+            QLIST_REMOVE(giommu, giommu_next);
+            g_free(giommu);
+            break;
+        }
+    }
+
+    /*
+     * FIXME: We assume the one big unmap below is adequate to
+     * remove any individual page mappings in the IOMMU which
+     * might have been copied into VFIO. This works for a page table
+     * based IOMMU where a big unmap flattens a large range of IO-PTEs.
+     * That may not be true for all IOMMU types.
+     */
+
+    iova = TARGET_PAGE_ALIGN(section->offset_within_address_space);
+    end = (section->offset_within_address_space + int128_get64(section->size)) &
+        TARGET_PAGE_MASK;
+
+    if (iova >= end) {
+        return;
+    }
+
+    trace_vfio_listener_region_del(iova, end - 1);
+
+    ret = vfio_dma_unmap(container, iova, end - iova);
+    memory_region_unref(section->mr);
+    if (ret) {
+        error_report("vfio_dma_unmap(%p, 0x%"HWADDR_PRIx", "
+                     "0x%"HWADDR_PRIx") = %d (%m)",
+                     container, iova, end - iova, ret);
+    }
+}
+
+static const MemoryListener vfio_spapr_memory_listener = {
+    .region_add = vfio_spapr_listener_region_add,
+    .region_del = vfio_spapr_listener_region_del,
+};
+
+static void vfio_spapr_listener_release(VFIOContainer *container)
+{
+    memory_listener_unregister(&container->iommu_data.spapr.listener);
+}
+
+void spapr_memory_listener_register(VFIOContainer *container)
+{
+    container->iommu_data.spapr.listener = vfio_spapr_memory_listener;
+    container->iommu_data.release = vfio_spapr_listener_release;
+
+    memory_listener_register(&container->iommu_data.spapr.listener,
+                             container->space->as);
+}
diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
index cfdfb6a..d0b831c 100644
--- a/include/hw/vfio/vfio-common.h
+++ b/include/hw/vfio/vfio-common.h
@@ -69,6 +69,10 @@ typedef struct VFIOType1 {
     bool initialized;
 } VFIOType1;
 
+typedef struct VFIOSPAPR {
+    MemoryListener listener;
+} VFIOSPAPR;
+
 typedef struct VFIOContainer {
     VFIOAddressSpace *space;
     int fd; /* /dev/vfio/vfio, empowered by the attached groups */
@@ -76,6 +80,7 @@ typedef struct VFIOContainer {
         /* enable abstraction to support various iommu backends */
         union {
             VFIOType1 type1;
+            VFIOSPAPR spapr;
         };
         void (*release)(struct VFIOContainer *);
     } iommu_data;
@@ -145,4 +150,12 @@ extern const MemoryRegionOps vfio_region_ops;
 extern QLIST_HEAD(vfio_group_head, VFIOGroup) vfio_group_list;
 extern QLIST_HEAD(vfio_as_head, VFIOAddressSpace) vfio_address_spaces;
 
+extern int vfio_dma_map(VFIOContainer *container, hwaddr iova,
+                        ram_addr_t size, void *vaddr, bool readonly);
+extern int vfio_dma_unmap(VFIOContainer *container,
+                          hwaddr iova, ram_addr_t size);
+bool vfio_listener_skipped_section(MemoryRegionSection *section);
+
+extern void spapr_memory_listener_register(VFIOContainer *container);
+
 #endif /* !HW_VFIO_VFIO_COMMON_H */
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 04/15] vfio: spapr: Move SPAPR-related code to a separate file

[Thomas Huth]

Am Sat, 11 Apr 2015 01:24:33 +1000
schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:

> This moves SPAPR bits to a separate file to avoid pollution of x86 code.
> 
> This is a mechanical patch.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> ---
>  hw/vfio/Makefile.objs         |   1 +
>  hw/vfio/common.c              | 134 ++-----------------------
>  hw/vfio/spapr.c               | 226 ++++++++++++++++++++++++++++++++++++++++++
>  include/hw/vfio/vfio-common.h |  13 +++
>  4 files changed, 246 insertions(+), 128 deletions(-)
>  create mode 100644 hw/vfio/spapr.c
> 
> diff --git a/hw/vfio/Makefile.objs b/hw/vfio/Makefile.objs
> index e31f30e..b987ffb 100644
> --- a/hw/vfio/Makefile.objs
> +++ b/hw/vfio/Makefile.objs
> @@ -1,4 +1,5 @@
>  ifeq ($(CONFIG_LINUX), y)
>  obj-$(CONFIG_SOFTMMU) += common.o
>  obj-$(CONFIG_PCI) += pci.o
> +obj-$(CONFIG_SOFTMMU) += spapr.o

Would it be possible / make sense to use CONFIG_PSERIES instead of
CONFIG_SOFTMMU here instead?

 Thomas

Re: [Qemu-devel] [PATCH qemu v6 04/15] vfio: spapr: Move SPAPR-related code to a separate file

[Alexey Kardashevskiy]

On 04/16/2015 05:50 PM, Thomas Huth wrote:
> Am Sat, 11 Apr 2015 01:24:33 +1000
> schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:
>
>> This moves SPAPR bits to a separate file to avoid pollution of x86 code.
>>
>> This is a mechanical patch.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
>> ---
>>   hw/vfio/Makefile.objs         |   1 +
>>   hw/vfio/common.c              | 134 ++-----------------------
>>   hw/vfio/spapr.c               | 226 ++++++++++++++++++++++++++++++++++++++++++
>>   include/hw/vfio/vfio-common.h |  13 +++
>>   4 files changed, 246 insertions(+), 128 deletions(-)
>>   create mode 100644 hw/vfio/spapr.c
>>
>> diff --git a/hw/vfio/Makefile.objs b/hw/vfio/Makefile.objs
>> index e31f30e..b987ffb 100644
>> --- a/hw/vfio/Makefile.objs
>> +++ b/hw/vfio/Makefile.objs
>> @@ -1,4 +1,5 @@
>>   ifeq ($(CONFIG_LINUX), y)
>>   obj-$(CONFIG_SOFTMMU) += common.o
>>   obj-$(CONFIG_PCI) += pci.o
>> +obj-$(CONFIG_SOFTMMU) += spapr.o
>
> Would it be possible / make sense to use CONFIG_PSERIES instead of
> CONFIG_SOFTMMU here instead?

No. This file has some exports and CONFIG_PSERIES is only for makefiles, 
the source code does not get to see these config options so it is 
impossible to define extern symbols and stubs.



-- 
Alexey

[Qemu-devel] [PATCH qemu v6 05/15] spapr_pci_vfio: Enable multiple groups per container

[Alexey Kardashevskiy]

This enables multiple IOMMU groups in one VFIO container which means
that multiple devices from different groups can share the same IOMMU
table (or tables if DDW).

This removes a group id from vfio_container_ioctl(). The kernel support
is required for this; if the host kernel does not have the support,
it will allow only one group per container. The PHB's "iommuid" property
is ignored. The ioctl is called for every container attached to
the address space. At the moment there is just one container anyway.

If there is no container attached to the address space,
vfio_container_do_ioctl() returns -1.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/ppc/spapr_pci_vfio.c | 17 ++++++-----------
 hw/vfio/common.c        | 20 ++++++--------------
 include/hw/vfio/vfio.h  |  2 +-
 3 files changed, 13 insertions(+), 26 deletions(-)

diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index 99a1be5..f8b503e 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -35,12 +35,7 @@ static void spapr_phb_vfio_finish_realize(sPAPRPHBState *sphb, Error **errp)
     sPAPRTCETable *tcet;
     uint32_t liobn = svphb->phb.dma_liobn;
 
-    if (svphb->iommugroupid == -1) {
-        error_setg(errp, "Wrong IOMMU group ID %d", svphb->iommugroupid);
-        return;
-    }
-
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as, svphb->iommugroupid,
+    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
                                VFIO_CHECK_EXTENSION,
                                (void *) VFIO_SPAPR_TCE_IOMMU);
     if (ret != 1) {
@@ -49,7 +44,7 @@ static void spapr_phb_vfio_finish_realize(sPAPRPHBState *sphb, Error **errp)
         return;
     }
 
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as, svphb->iommugroupid,
+    ret = vfio_container_ioctl(&sphb->iommu_as,
                                VFIO_IOMMU_SPAPR_TCE_GET_INFO, &info);
     if (ret) {
         error_setg_errno(errp, -ret,
@@ -116,7 +111,7 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
         return RTAS_OUT_PARAM_ERROR;
     }
 
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as, svphb->iommugroupid,
+    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_HW_ERROR;
@@ -132,7 +127,7 @@ static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
     int ret;
 
     op.op = VFIO_EEH_PE_GET_STATE;
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as, svphb->iommugroupid,
+    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_PARAM_ERROR;
@@ -162,7 +157,7 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
         return RTAS_OUT_PARAM_ERROR;
     }
 
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as, svphb->iommugroupid,
+    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_HW_ERROR;
@@ -178,7 +173,7 @@ static int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
     int ret;
 
     op.op = VFIO_EEH_PE_CONFIGURE;
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as, svphb->iommugroupid,
+    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_PARAM_ERROR;
diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index 3e4c685..369e564 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -793,34 +793,26 @@ void vfio_put_base_device(VFIODevice *vbasedev)
     close(vbasedev->fd);
 }
 
-static int vfio_container_do_ioctl(AddressSpace *as, int32_t groupid,
+static int vfio_container_do_ioctl(AddressSpace *as,
                                    int req, void *param)
 {
-    VFIOGroup *group;
     VFIOContainer *container;
     int ret = -1;
+    VFIOAddressSpace *space = vfio_get_address_space(as);
 
-    group = vfio_get_group(groupid, as);
-    if (!group) {
-        error_report("vfio: group %d not registered", groupid);
-        return ret;
-    }
-
-    container = group->container;
-    if (group->container) {
+    QLIST_FOREACH(container, &space->containers, next) {
         ret = ioctl(container->fd, req, param);
         if (ret < 0) {
             error_report("vfio: failed to ioctl %d to container: ret=%d, %s",
                          _IOC_NR(req) - VFIO_BASE, ret, strerror(errno));
+            return -errno;
         }
     }
 
-    vfio_put_group(group);
-
     return ret;
 }
 
-int vfio_container_ioctl(AddressSpace *as, int32_t groupid,
+int vfio_container_ioctl(AddressSpace *as,
                          int req, void *param)
 {
     /* We allow only certain ioctls to the container */
@@ -835,5 +827,5 @@ int vfio_container_ioctl(AddressSpace *as, int32_t groupid,
         return -1;
     }
 
-    return vfio_container_do_ioctl(as, groupid, req, param);
+    return vfio_container_do_ioctl(as, req, param);
 }
diff --git a/include/hw/vfio/vfio.h b/include/hw/vfio/vfio.h
index 0b26cd8..76b5744 100644
--- a/include/hw/vfio/vfio.h
+++ b/include/hw/vfio/vfio.h
@@ -3,7 +3,7 @@
 
 #include "qemu/typedefs.h"
 
-extern int vfio_container_ioctl(AddressSpace *as, int32_t groupid,
+extern int vfio_container_ioctl(AddressSpace *as,
                                 int req, void *param);
 
 #endif
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 05/15] spapr_pci_vfio: Enable multiple groups per container

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 1027 bytes --]

On Sat, Apr 11, 2015 at 01:24:34AM +1000, Alexey Kardashevskiy wrote:
> This enables multiple IOMMU groups in one VFIO container which means
> that multiple devices from different groups can share the same IOMMU
> table (or tables if DDW).
> 
> This removes a group id from vfio_container_ioctl(). The kernel support
> is required for this; if the host kernel does not have the support,
> it will allow only one group per container. The PHB's "iommuid" property
> is ignored. The ioctl is called for every container attached to
> the address space. At the moment there is just one container anyway.
> 
> If there is no container attached to the address space,
> vfio_container_do_ioctl() returns -1.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

[Qemu-devel] [PATCH qemu v6 06/15] spapr_pci: Convert finish_realize() to dma_capabilities_update()+dma_init_window()

[Alexey Kardashevskiy]

This reworks finish_realize() which used to finalize DMA setup with
an assumption that it will not change later.

New callbacks supports various window parameters such as page and
windows sizes. The new callback return error code rather than Error**.

This is a mechanical change so no change in behaviour is expected.
This is a part of getting rid of spapr-pci-vfio-host-bridge type.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/ppc/spapr_pci.c          | 59 ++++++++++++++++++++++++++-------------------
 hw/ppc/spapr_pci_vfio.c     | 47 +++++++++++++++---------------------
 include/hw/pci-host/spapr.h |  8 +++++-
 3 files changed, 61 insertions(+), 53 deletions(-)

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index 52c5c73..8c0d2eb 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -741,6 +741,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
     int i;
     PCIBus *bus;
     uint64_t msi_window_size = 4096;
+    sPAPRTCETable *tcet;
 
     if (sphb->index != (uint32_t)-1) {
         hwaddr windows_base;
@@ -880,33 +881,40 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         sphb->lsi_table[i].irq = irq;
     }
 
-    if (!info->finish_realize) {
-        error_setg(errp, "finish_realize not defined");
-        return;
-    }
-
-    info->finish_realize(sphb, errp);
-
-    sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
-}
-
-static void spapr_phb_finish_realize(sPAPRPHBState *sphb, Error **errp)
-{
-    sPAPRTCETable *tcet;
-    uint32_t nb_table;
-
-    nb_table = SPAPR_PCI_DMA32_SIZE >> SPAPR_TCE_PAGE_SHIFT;
-    tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn,
-                               0, SPAPR_TCE_PAGE_SHIFT, nb_table, false);
+    info->dma_capabilities_update(sphb);
+    info->dma_init_window(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
+                          sphb->dma32_window_size);
+    tcet = spapr_tce_find_by_liobn(sphb->dma_liobn);
     if (!tcet) {
-        error_setg(errp, "Unable to create TCE table for %s",
-                   sphb->dtbusname);
-        return ;
+        error_setg(errp, "failed to create TCE table");
+        return;
     }
-
-    /* Register default 32bit DMA window */
-    memory_region_add_subregion(&sphb->iommu_root, 0,
+    memory_region_add_subregion(&sphb->iommu_root, tcet->bus_offset,
                                 spapr_tce_get_iommu(tcet));
+
+    sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
+}
+
+static int spapr_phb_dma_capabilities_update(sPAPRPHBState *sphb)
+{
+    sphb->dma32_window_start = 0;
+    sphb->dma32_window_size = SPAPR_PCI_DMA32_SIZE;
+
+    return 0;
+}
+
+static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
+                                     uint32_t liobn, uint32_t page_shift,
+                                     uint64_t window_size)
+{
+    uint64_t bus_offset = sphb->dma32_window_start;
+    sPAPRTCETable *tcet;
+
+    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, bus_offset, page_shift,
+                               window_size >> page_shift,
+                               false);
+
+    return tcet ? 0 : -1;
 }
 
 static int spapr_phb_children_reset(Object *child, void *opaque)
@@ -1057,7 +1065,8 @@ static void spapr_phb_class_init(ObjectClass *klass, void *data)
     dc->vmsd = &vmstate_spapr_pci;
     set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
     dc->cannot_instantiate_with_device_add_yet = false;
-    spc->finish_realize = spapr_phb_finish_realize;
+    spc->dma_capabilities_update = spapr_phb_dma_capabilities_update;
+    spc->dma_init_window = spapr_phb_dma_init_window;
 }
 
 static const TypeInfo spapr_phb_info = {
diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index f8b503e..0ce8e61 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -27,43 +27,35 @@ static Property spapr_phb_vfio_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
-static void spapr_phb_vfio_finish_realize(sPAPRPHBState *sphb, Error **errp)
+static int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb)
 {
-    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
     struct vfio_iommu_spapr_tce_info info = { .argsz = sizeof(info) };
     int ret;
-    sPAPRTCETable *tcet;
-    uint32_t liobn = svphb->phb.dma_liobn;
-
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
-                               VFIO_CHECK_EXTENSION,
-                               (void *) VFIO_SPAPR_TCE_IOMMU);
-    if (ret != 1) {
-        error_setg_errno(errp, -ret,
-                         "spapr-vfio: SPAPR extension is not supported");
-        return;
-    }
 
     ret = vfio_container_ioctl(&sphb->iommu_as,
                                VFIO_IOMMU_SPAPR_TCE_GET_INFO, &info);
     if (ret) {
-        error_setg_errno(errp, -ret,
-                         "spapr-vfio: get info from container failed");
-        return;
+        return ret;
     }
 
-    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, info.dma32_window_start,
-                               SPAPR_TCE_PAGE_SHIFT,
-                               info.dma32_window_size >> SPAPR_TCE_PAGE_SHIFT,
+    sphb->dma32_window_start = info.dma32_window_start;
+    sphb->dma32_window_size = info.dma32_window_size;
+
+    return ret;
+}
+
+static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
+                                          uint32_t liobn, uint32_t page_shift,
+                                          uint64_t window_size)
+{
+    uint64_t bus_offset = sphb->dma32_window_start;
+    sPAPRTCETable *tcet;
+
+    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, bus_offset, page_shift,
+                               window_size >> page_shift,
                                true);
-    if (!tcet) {
-        error_setg(errp, "spapr-vfio: failed to create VFIO TCE table");
-        return;
-    }
 
-    /* Register default 32bit DMA window */
-    memory_region_add_subregion(&sphb->iommu_root, tcet->bus_offset,
-                                spapr_tce_get_iommu(tcet));
+    return tcet ? 0 : -1;
 }
 
 static void spapr_phb_vfio_reset(DeviceState *qdev)
@@ -189,7 +181,8 @@ static void spapr_phb_vfio_class_init(ObjectClass *klass, void *data)
 
     dc->props = spapr_phb_vfio_properties;
     dc->reset = spapr_phb_vfio_reset;
-    spc->finish_realize = spapr_phb_vfio_finish_realize;
+    spc->dma_capabilities_update = spapr_phb_vfio_dma_capabilities_update;
+    spc->dma_init_window = spapr_phb_vfio_dma_init_window;
     spc->eeh_set_option = spapr_phb_vfio_eeh_set_option;
     spc->eeh_get_state = spapr_phb_vfio_eeh_get_state;
     spc->eeh_reset = spapr_phb_vfio_eeh_reset;
diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
index 5b497ce..3074145 100644
--- a/include/hw/pci-host/spapr.h
+++ b/include/hw/pci-host/spapr.h
@@ -48,7 +48,10 @@ typedef struct sPAPRPHBVFIOState sPAPRPHBVFIOState;
 struct sPAPRPHBClass {
     PCIHostBridgeClass parent_class;
 
-    void (*finish_realize)(sPAPRPHBState *sphb, Error **errp);
+    int (*dma_capabilities_update)(sPAPRPHBState *sphb);
+    int (*dma_init_window)(sPAPRPHBState *sphb,
+                           uint32_t liobn, uint32_t page_shift,
+                           uint64_t window_size);
     int (*eeh_set_option)(sPAPRPHBState *sphb, unsigned int addr, int option);
     int (*eeh_get_state)(sPAPRPHBState *sphb, int *state);
     int (*eeh_reset)(sPAPRPHBState *sphb, int option);
@@ -89,6 +92,9 @@ struct sPAPRPHBState {
     int32_t msi_devs_num;
     spapr_pci_msi_mig *msi_devs;
 
+    uint32_t dma32_window_start;
+    uint32_t dma32_window_size;
+
     QLIST_ENTRY(sPAPRPHBState) list;
 };
 
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 06/15] spapr_pci: Convert finish_realize() to dma_capabilities_update()+dma_init_window()

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 794 bytes --]

On Sat, Apr 11, 2015 at 01:24:35AM +1000, Alexey Kardashevskiy wrote:
> This reworks finish_realize() which used to finalize DMA setup with
> an assumption that it will not change later.
> 
> New callbacks supports various window parameters such as page and
> windows sizes. The new callback return error code rather than Error**.
> 
> This is a mechanical change so no change in behaviour is expected.
> This is a part of getting rid of spapr-pci-vfio-host-bridge type.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

[Qemu-devel] [PATCH qemu v6 07/15] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)

[Alexey Kardashevskiy]

This makes use of the new "memory registering" feature. The idea is
to provide the userspace ability to notify the host kernel about pages
which are going to be used for DMA. Having this information, the host
kernel can pin them all once per user process, do locked pages
accounting (once) and not spent time on doing that in real time with
possible failures which cannot be handled nicely in some cases.

This adds a guest RAM memory listener which notifies a VFIO container
about memory which needs to be pinned/unpinned. VFIO MMIO regions
(i.e. "skip dump" regions) are skipped.

The feature is only enabled for SPAPR IOMMU v2. The host kernel changes
are required. Since v2 does not need/support VFIO_IOMMU_ENABLE, this does
not call it when v2 is detected and enabled.

This does not change the guest visible interface.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
Changes:
v6:
* fixed commit log (s/guest/userspace/), added note about no guest visible
change
* fixed error checking if ram registration failed
* added alignment check for section->offset_within_region

v5:
* simplified the patch
* added trace points
* added round_up() for the size
* SPAPR IOMMU v2 used
---
 hw/vfio/common.c              | 26 +++++++++----
 hw/vfio/spapr.c               | 88 ++++++++++++++++++++++++++++++++++++++++++-
 include/hw/vfio/vfio-common.h |  5 ++-
 trace-events                  |  1 +
 4 files changed, 109 insertions(+), 11 deletions(-)

diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index 369e564..9e3e0b0 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -577,14 +577,18 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
 
         container->iommu_data.type1.initialized = true;
 
-    } else if (ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_IOMMU)) {
+    } else if (ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_IOMMU) ||
+               ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_v2_IOMMU)) {
+        bool v2 = !!ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_v2_IOMMU);
+
         ret = ioctl(group->fd, VFIO_GROUP_SET_CONTAINER, &fd);
         if (ret) {
             error_report("vfio: failed to set group container: %m");
             ret = -errno;
             goto free_container_exit;
         }
-        ret = ioctl(fd, VFIO_SET_IOMMU, VFIO_SPAPR_TCE_IOMMU);
+        ret = ioctl(fd, VFIO_SET_IOMMU,
+                v2 ? VFIO_SPAPR_TCE_v2_IOMMU : VFIO_SPAPR_TCE_IOMMU);
         if (ret) {
             error_report("vfio: failed to set iommu for container: %m");
             ret = -errno;
@@ -596,14 +600,20 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
          * when container fd is closed so we do not call it explicitly
          * in this file.
          */
-        ret = ioctl(fd, VFIO_IOMMU_ENABLE);
-        if (ret) {
-            error_report("vfio: failed to enable container: %m");
-            ret = -errno;
-            goto free_container_exit;
+        if (!v2) {
+            ret = ioctl(fd, VFIO_IOMMU_ENABLE);
+            if (ret) {
+                error_report("vfio: failed to enable container: %m");
+                ret = -errno;
+                goto free_container_exit;
+            }
         }
 
-        spapr_memory_listener_register(container);
+        ret = spapr_memory_listener_register(container, v2 ? 2 : 1);
+        if (ret) {
+            error_report("vfio: RAM memory listener initialization failed for container");
+            goto listener_release_exit;
+        }
 
     } else {
         error_report("vfio: No available IOMMU models");
diff --git a/hw/vfio/spapr.c b/hw/vfio/spapr.c
index 5f79194..31353f1 100644
--- a/hw/vfio/spapr.c
+++ b/hw/vfio/spapr.c
@@ -17,6 +17,9 @@
  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <sys/ioctl.h>
+#include <linux/vfio.h>
+
 #include "hw/vfio/vfio-common.h"
 #include "qemu/error-report.h"
 #include "trace.h"
@@ -211,16 +214,97 @@ static const MemoryListener vfio_spapr_memory_listener = {
     .region_del = vfio_spapr_listener_region_del,
 };
 
+static void vfio_ram_do_region(VFIOContainer *container,
+                              MemoryRegionSection *section, unsigned long req)
+{
+    int ret;
+    struct vfio_iommu_spapr_register_memory reg = { .argsz = sizeof(reg) };
+
+    if (!memory_region_is_ram(section->mr) ||
+        memory_region_is_skip_dump(section->mr)) {
+        return;
+    }
+
+    if (unlikely((section->offset_within_region & (getpagesize() - 1)))) {
+        error_report("%s received unaligned region", __func__);
+        return;
+    }
+
+    reg.vaddr = (__u64) memory_region_get_ram_ptr(section->mr) +
+        section->offset_within_region;
+    reg.size = ROUND_UP(int128_get64(section->size), TARGET_PAGE_SIZE);
+
+    ret = ioctl(container->fd, req, &reg);
+    trace_vfio_ram_register(_IOC_NR(req) - VFIO_BASE, reg.vaddr, reg.size,
+            ret ? -errno : 0);
+    if (!ret) {
+        return;
+    }
+
+    /*
+     * On the initfn path, store the first error in the container so we
+     * can gracefully fail.  Runtime, there's not much we can do other
+     * than throw a hardware error.
+     */
+    if (!container->iommu_data.spapr.ram_reg_initialized) {
+        if (!container->iommu_data.spapr.ram_reg_error) {
+            container->iommu_data.spapr.ram_reg_error = -errno;
+        }
+    } else {
+        hw_error("vfio: RAM registering failed, unable to continue");
+    }
+}
+
+static void vfio_spapr_ram_listener_region_add(MemoryListener *listener,
+                                               MemoryRegionSection *section)
+{
+    VFIOContainer *container = container_of(listener, VFIOContainer,
+                                            iommu_data.spapr.ramlistener);
+    memory_region_ref(section->mr);
+    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_REGISTER_MEMORY);
+}
+
+static void vfio_spapr_ram_listener_region_del(MemoryListener *listener,
+                                               MemoryRegionSection *section)
+{
+    VFIOContainer *container = container_of(listener, VFIOContainer,
+                                            iommu_data.spapr.ramlistener);
+    memory_region_unref(section->mr);
+    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY);
+}
+
+static const MemoryListener vfio_spapr_ram_memory_listener = {
+    .region_add = vfio_spapr_ram_listener_region_add,
+    .region_del = vfio_spapr_ram_listener_region_del,
+};
+
 static void vfio_spapr_listener_release(VFIOContainer *container)
 {
     memory_listener_unregister(&container->iommu_data.spapr.listener);
 }
 
-void spapr_memory_listener_register(VFIOContainer *container)
+static void vfio_spapr_listener_release_v2(VFIOContainer *container)
+{
+    memory_listener_unregister(&container->iommu_data.spapr.listener);
+    vfio_spapr_listener_release(container);
+}
+
+int spapr_memory_listener_register(VFIOContainer *container, int ver)
 {
     container->iommu_data.spapr.listener = vfio_spapr_memory_listener;
     container->iommu_data.release = vfio_spapr_listener_release;
-
     memory_listener_register(&container->iommu_data.spapr.listener,
                              container->space->as);
+    if (ver < 2) {
+        return 0;
+    }
+
+    container->iommu_data.spapr.ramlistener = vfio_spapr_ram_memory_listener;
+    container->iommu_data.release = vfio_spapr_listener_release_v2;
+    memory_listener_register(&container->iommu_data.spapr.ramlistener,
+                             &address_space_memory);
+
+    container->iommu_data.spapr.ram_reg_initialized = true;
+
+    return container->iommu_data.spapr.ram_reg_error;
 }
diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
index d0b831c..b5ef446 100644
--- a/include/hw/vfio/vfio-common.h
+++ b/include/hw/vfio/vfio-common.h
@@ -71,6 +71,9 @@ typedef struct VFIOType1 {
 
 typedef struct VFIOSPAPR {
     MemoryListener listener;
+    MemoryListener ramlistener;
+    int ram_reg_error;
+    bool ram_reg_initialized;
 } VFIOSPAPR;
 
 typedef struct VFIOContainer {
@@ -156,6 +159,6 @@ extern int vfio_dma_unmap(VFIOContainer *container,
                           hwaddr iova, ram_addr_t size);
 bool vfio_listener_skipped_section(MemoryRegionSection *section);
 
-extern void spapr_memory_listener_register(VFIOContainer *container);
+extern int spapr_memory_listener_register(VFIOContainer *container, int ver);
 
 #endif /* !HW_VFIO_VFIO_COMMON_H */
diff --git a/trace-events b/trace-events
index 1231ba4..2739140 100644
--- a/trace-events
+++ b/trace-events
@@ -1563,6 +1563,7 @@ vfio_disconnect_container(int fd) "close container->fd=%d"
 vfio_put_group(int fd) "close group->fd=%d"
 vfio_get_device(const char * name, unsigned int flags, unsigned int num_regions, unsigned int num_irqs) "Device %s flags: %u, regions: %u, irqs: %u"
 vfio_put_base_device(int fd) "close vdev->fd=%d"
+vfio_ram_register(int req, uint64_t va, uint64_t size, int ret) "req=%d va=%"PRIx64" size=%"PRIx64" ret=%d"
 
 #hw/acpi/memory_hotplug.c
 mhp_acpi_invalid_slot_selected(uint32_t slot) "0x%"PRIx32
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 07/15] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)

[Thomas Huth]

Am Sat, 11 Apr 2015 01:24:36 +1000
schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:

> This makes use of the new "memory registering" feature. The idea is
> to provide the userspace ability to notify the host kernel about pages
> which are going to be used for DMA. Having this information, the host
> kernel can pin them all once per user process, do locked pages
> accounting (once) and not spent time on doing that in real time with
> possible failures which cannot be handled nicely in some cases.
> 
> This adds a guest RAM memory listener which notifies a VFIO container
> about memory which needs to be pinned/unpinned. VFIO MMIO regions
> (i.e. "skip dump" regions) are skipped.
> 
> The feature is only enabled for SPAPR IOMMU v2. The host kernel changes
> are required. Since v2 does not need/support VFIO_IOMMU_ENABLE, this does
> not call it when v2 is detected and enabled.
> 
> This does not change the guest visible interface.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> Changes:
> v6:
> * fixed commit log (s/guest/userspace/), added note about no guest visible
> change
> * fixed error checking if ram registration failed
> * added alignment check for section->offset_within_region
> 
> v5:
> * simplified the patch
> * added trace points
> * added round_up() for the size
> * SPAPR IOMMU v2 used
> ---
>  hw/vfio/common.c              | 26 +++++++++----
>  hw/vfio/spapr.c               | 88 ++++++++++++++++++++++++++++++++++++++++++-
>  include/hw/vfio/vfio-common.h |  5 ++-
>  trace-events                  |  1 +
>  4 files changed, 109 insertions(+), 11 deletions(-)
[...]
> diff --git a/hw/vfio/spapr.c b/hw/vfio/spapr.c
> index 5f79194..31353f1 100644
> --- a/hw/vfio/spapr.c
> +++ b/hw/vfio/spapr.c
> @@ -17,6 +17,9 @@
>   *  along with this program; if not, see <http://www.gnu.org/licenses/>.
>   */
>  
> +#include <sys/ioctl.h>
> +#include <linux/vfio.h>
> +
>  #include "hw/vfio/vfio-common.h"
>  #include "qemu/error-report.h"
>  #include "trace.h"
> @@ -211,16 +214,97 @@ static const MemoryListener vfio_spapr_memory_listener = {
>      .region_del = vfio_spapr_listener_region_del,
>  };
>  
> +static void vfio_ram_do_region(VFIOContainer *container,
> +                              MemoryRegionSection *section, unsigned long req)
> +{
> +    int ret;
> +    struct vfio_iommu_spapr_register_memory reg = { .argsz = sizeof(reg) };
> +
> +    if (!memory_region_is_ram(section->mr) ||
> +        memory_region_is_skip_dump(section->mr)) {
> +        return;
> +    }
> +
> +    if (unlikely((section->offset_within_region & (getpagesize() - 1)))) {
> +        error_report("%s received unaligned region", __func__);
> +        return;
> +    }
> +
> +    reg.vaddr = (__u64) memory_region_get_ram_ptr(section->mr) +
> +        section->offset_within_region;
> +    reg.size = ROUND_UP(int128_get64(section->size), TARGET_PAGE_SIZE);
> +
> +    ret = ioctl(container->fd, req, &reg);
> +    trace_vfio_ram_register(_IOC_NR(req) - VFIO_BASE, reg.vaddr, reg.size,
> +            ret ? -errno : 0);
> +    if (!ret) {

Since this function does not return an error code (maybe it should?), would it
make sense to print out at least an error message here that something went wrong?
(most people won't have the tracing enabled, so the error might go unnoticed otherwise)

> +        return;
> +    }
> +
> +    /*
> +     * On the initfn path, store the first error in the container so we
> +     * can gracefully fail.  Runtime, there's not much we can do other
> +     * than throw a hardware error.
> +     */
> +    if (!container->iommu_data.spapr.ram_reg_initialized) {
> +        if (!container->iommu_data.spapr.ram_reg_error) {
> +            container->iommu_data.spapr.ram_reg_error = -errno;
> +        }
> +    } else {
> +        hw_error("vfio: RAM registering failed, unable to continue");
> +    }
> +}

 Thomas

Re: [Qemu-devel] [PATCH qemu v6 07/15] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)

[Alexey Kardashevskiy]

On 04/16/2015 08:07 PM, Thomas Huth wrote:
> Am Sat, 11 Apr 2015 01:24:36 +1000
> schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:
>
>> This makes use of the new "memory registering" feature. The idea is
>> to provide the userspace ability to notify the host kernel about pages
>> which are going to be used for DMA. Having this information, the host
>> kernel can pin them all once per user process, do locked pages
>> accounting (once) and not spent time on doing that in real time with
>> possible failures which cannot be handled nicely in some cases.
>>
>> This adds a guest RAM memory listener which notifies a VFIO container
>> about memory which needs to be pinned/unpinned. VFIO MMIO regions
>> (i.e. "skip dump" regions) are skipped.
>>
>> The feature is only enabled for SPAPR IOMMU v2. The host kernel changes
>> are required. Since v2 does not need/support VFIO_IOMMU_ENABLE, this does
>> not call it when v2 is detected and enabled.
>>
>> This does not change the guest visible interface.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> ---
>> Changes:
>> v6:
>> * fixed commit log (s/guest/userspace/), added note about no guest visible
>> change
>> * fixed error checking if ram registration failed
>> * added alignment check for section->offset_within_region
>>
>> v5:
>> * simplified the patch
>> * added trace points
>> * added round_up() for the size
>> * SPAPR IOMMU v2 used
>> ---
>>   hw/vfio/common.c              | 26 +++++++++----
>>   hw/vfio/spapr.c               | 88 ++++++++++++++++++++++++++++++++++++++++++-
>>   include/hw/vfio/vfio-common.h |  5 ++-
>>   trace-events                  |  1 +
>>   4 files changed, 109 insertions(+), 11 deletions(-)
> [...]
>> diff --git a/hw/vfio/spapr.c b/hw/vfio/spapr.c
>> index 5f79194..31353f1 100644
>> --- a/hw/vfio/spapr.c
>> +++ b/hw/vfio/spapr.c
>> @@ -17,6 +17,9 @@
>>    *  along with this program; if not, see <http://www.gnu.org/licenses/>.
>>    */
>>
>> +#include <sys/ioctl.h>
>> +#include <linux/vfio.h>
>> +
>>   #include "hw/vfio/vfio-common.h"
>>   #include "qemu/error-report.h"
>>   #include "trace.h"
>> @@ -211,16 +214,97 @@ static const MemoryListener vfio_spapr_memory_listener = {
>>       .region_del = vfio_spapr_listener_region_del,
>>   };
>>
>> +static void vfio_ram_do_region(VFIOContainer *container,
>> +                              MemoryRegionSection *section, unsigned long req)
>> +{
>> +    int ret;
>> +    struct vfio_iommu_spapr_register_memory reg = { .argsz = sizeof(reg) };
>> +
>> +    if (!memory_region_is_ram(section->mr) ||
>> +        memory_region_is_skip_dump(section->mr)) {
>> +        return;
>> +    }
>> +
>> +    if (unlikely((section->offset_within_region & (getpagesize() - 1)))) {
>> +        error_report("%s received unaligned region", __func__);
>> +        return;
>> +    }
>> +
>> +    reg.vaddr = (__u64) memory_region_get_ram_ptr(section->mr) +
>> +        section->offset_within_region;
>> +    reg.size = ROUND_UP(int128_get64(section->size), TARGET_PAGE_SIZE);
>> +
>> +    ret = ioctl(container->fd, req, &reg);
>> +    trace_vfio_ram_register(_IOC_NR(req) - VFIO_BASE, reg.vaddr, reg.size,
>> +            ret ? -errno : 0);
>> +    if (!ret) {
>
> Since this function does not return an error code (maybe it should?),


It is called from listener callbacks which do not return codes.


> would it
> make sense to print out at least an error message here that something went wrong?
> (most people won't have the tracing enabled, so the error might go unnoticed otherwise)


!ret means ret==0 which means "good". The error case handling is few lines 
below and it won't go unnoticed.



>
>> +        return;
>> +    }
>> +
>> +    /*
>> +     * On the initfn path, store the first error in the container so we
>> +     * can gracefully fail.  Runtime, there's not much we can do other
>> +     * than throw a hardware error.
>> +     */
>> +    if (!container->iommu_data.spapr.ram_reg_initialized) {
>> +        if (!container->iommu_data.spapr.ram_reg_error) {
>> +            container->iommu_data.spapr.ram_reg_error = -errno;
>> +        }
>> +    } else {
>> +        hw_error("vfio: RAM registering failed, unable to continue");
>> +    }
>> +}
>
>   Thomas
>


-- 
Alexey

Re: [Qemu-devel] [PATCH qemu v6 07/15] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 10876 bytes --]

On Sat, Apr 11, 2015 at 01:24:36AM +1000, Alexey Kardashevskiy wrote:
> This makes use of the new "memory registering" feature. The idea is
> to provide the userspace ability to notify the host kernel about pages
> which are going to be used for DMA. Having this information, the host
> kernel can pin them all once per user process, do locked pages
> accounting (once) and not spent time on doing that in real time with
> possible failures which cannot be handled nicely in some cases.
> 
> This adds a guest RAM memory listener which notifies a VFIO container
> about memory which needs to be pinned/unpinned. VFIO MMIO regions
> (i.e. "skip dump" regions) are skipped.
> 
> The feature is only enabled for SPAPR IOMMU v2. The host kernel changes
> are required. Since v2 does not need/support VFIO_IOMMU_ENABLE, this does
> not call it when v2 is detected and enabled.
> 
> This does not change the guest visible interface.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

Albeit with some nits described below.

> ---
> Changes:
> v6:
> * fixed commit log (s/guest/userspace/), added note about no guest visible
> change
> * fixed error checking if ram registration failed
> * added alignment check for section->offset_within_region
> 
> v5:
> * simplified the patch
> * added trace points
> * added round_up() for the size
> * SPAPR IOMMU v2 used
> ---
>  hw/vfio/common.c              | 26 +++++++++----
>  hw/vfio/spapr.c               | 88 ++++++++++++++++++++++++++++++++++++++++++-
>  include/hw/vfio/vfio-common.h |  5 ++-
>  trace-events                  |  1 +
>  4 files changed, 109 insertions(+), 11 deletions(-)
> 
> diff --git a/hw/vfio/common.c b/hw/vfio/common.c
> index 369e564..9e3e0b0 100644
> --- a/hw/vfio/common.c
> +++ b/hw/vfio/common.c
> @@ -577,14 +577,18 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
>  
>          container->iommu_data.type1.initialized = true;
>  
> -    } else if (ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_IOMMU)) {
> +    } else if (ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_IOMMU) ||
> +               ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_v2_IOMMU)) {
> +        bool v2 = !!ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_v2_IOMMU);
> +
>          ret = ioctl(group->fd, VFIO_GROUP_SET_CONTAINER, &fd);
>          if (ret) {
>              error_report("vfio: failed to set group container: %m");
>              ret = -errno;
>              goto free_container_exit;
>          }
> -        ret = ioctl(fd, VFIO_SET_IOMMU, VFIO_SPAPR_TCE_IOMMU);
> +        ret = ioctl(fd, VFIO_SET_IOMMU,
> +                v2 ? VFIO_SPAPR_TCE_v2_IOMMU : VFIO_SPAPR_TCE_IOMMU);
>          if (ret) {
>              error_report("vfio: failed to set iommu for container: %m");
>              ret = -errno;
> @@ -596,14 +600,20 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
>           * when container fd is closed so we do not call it explicitly
>           * in this file.
>           */
> -        ret = ioctl(fd, VFIO_IOMMU_ENABLE);
> -        if (ret) {
> -            error_report("vfio: failed to enable container: %m");
> -            ret = -errno;
> -            goto free_container_exit;
> +        if (!v2) {
> +            ret = ioctl(fd, VFIO_IOMMU_ENABLE);
> +            if (ret) {
> +                error_report("vfio: failed to enable container: %m");
> +                ret = -errno;
> +                goto free_container_exit;
> +            }
>          }
>  
> -        spapr_memory_listener_register(container);
> +        ret = spapr_memory_listener_register(container, v2 ? 2 : 1);
> +        if (ret) {
> +            error_report("vfio: RAM memory listener initialization failed for container");
> +            goto listener_release_exit;
> +        }
>  
>      } else {
>          error_report("vfio: No available IOMMU models");
> diff --git a/hw/vfio/spapr.c b/hw/vfio/spapr.c
> index 5f79194..31353f1 100644
> --- a/hw/vfio/spapr.c
> +++ b/hw/vfio/spapr.c
> @@ -17,6 +17,9 @@
>   *  along with this program; if not, see <http://www.gnu.org/licenses/>.
>   */
>  
> +#include <sys/ioctl.h>
> +#include <linux/vfio.h>
> +
>  #include "hw/vfio/vfio-common.h"
>  #include "qemu/error-report.h"
>  #include "trace.h"
> @@ -211,16 +214,97 @@ static const MemoryListener vfio_spapr_memory_listener = {
>      .region_del = vfio_spapr_listener_region_del,
>  };
>  
> +static void vfio_ram_do_region(VFIOContainer *container,
> +                              MemoryRegionSection *section, unsigned long req)
> +{
> +    int ret;
> +    struct vfio_iommu_spapr_register_memory reg = { .argsz = sizeof(reg) };
> +
> +    if (!memory_region_is_ram(section->mr) ||
> +        memory_region_is_skip_dump(section->mr)) {
> +        return;
> +    }
> +
> +    if (unlikely((section->offset_within_region & (getpagesize() - 1)))) {
> +        error_report("%s received unaligned region", __func__);
> +        return;
> +    }
> +
> +    reg.vaddr = (__u64) memory_region_get_ram_ptr(section->mr) +
> +        section->offset_within_region;
> +    reg.size = ROUND_UP(int128_get64(section->size), TARGET_PAGE_SIZE);
> +
> +    ret = ioctl(container->fd, req, &reg);
> +    trace_vfio_ram_register(_IOC_NR(req) - VFIO_BASE, reg.vaddr, reg.size,
> +            ret ? -errno : 0);
> +    if (!ret) {
> +        return;
> +    }
> +
> +    /*
> +     * On the initfn path, store the first error in the container so we
> +     * can gracefully fail.  Runtime, there's not much we can do other
> +     * than throw a hardware error.
> +     */
> +    if (!container->iommu_data.spapr.ram_reg_initialized) {
> +        if (!container->iommu_data.spapr.ram_reg_error) {
> +            container->iommu_data.spapr.ram_reg_error = -errno;
> +        }
> +    } else {
> +        hw_error("vfio: RAM registering failed, unable to continue");
> +    }
> +}
> +
> +static void vfio_spapr_ram_listener_region_add(MemoryListener *listener,
> +                                               MemoryRegionSection *section)
> +{
> +    VFIOContainer *container = container_of(listener, VFIOContainer,
> +                                            iommu_data.spapr.ramlistener);
> +    memory_region_ref(section->mr);
> +    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_REGISTER_MEMORY);
> +}
> +
> +static void vfio_spapr_ram_listener_region_del(MemoryListener *listener,
> +                                               MemoryRegionSection *section)
> +{
> +    VFIOContainer *container = container_of(listener, VFIOContainer,
> +                                            iommu_data.spapr.ramlistener);
> +    memory_region_unref(section->mr);

You're probably safe because of the QBL, but it would be more
obviously correct if you did the ioctl unregister before unreferencing
the MR.

> +    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY);
> +}
> +
> +static const MemoryListener vfio_spapr_ram_memory_listener = {
> +    .region_add = vfio_spapr_ram_listener_region_add,
> +    .region_del = vfio_spapr_ram_listener_region_del,
> +};
> +
>  static void vfio_spapr_listener_release(VFIOContainer *container)
>  {
>      memory_listener_unregister(&container->iommu_data.spapr.listener);
>  }
>  
> -void spapr_memory_listener_register(VFIOContainer *container)
> +static void vfio_spapr_listener_release_v2(VFIOContainer *container)
> +{
> +    memory_listener_unregister(&container->iommu_data.spapr.listener);
> +    vfio_spapr_listener_release(container);
> +}
> +
> +int spapr_memory_listener_register(VFIOContainer *container, int ver)
>  {
>      container->iommu_data.spapr.listener = vfio_spapr_memory_listener;
>      container->iommu_data.release = vfio_spapr_listener_release;
> -
>      memory_listener_register(&container->iommu_data.spapr.listener,
>                               container->space->as);
> +    if (ver < 2) {
> +        return 0;
> +    }

I wonder if it would make sense to store the IOMMU type value into the
VFIOContainer (from non-arch specific code).  It would avoid the
ad-hoc passing of version here and also allow for some sanity checks.

> +
> +    container->iommu_data.spapr.ramlistener = vfio_spapr_ram_memory_listener;
> +    container->iommu_data.release = vfio_spapr_listener_release_v2;
> +    memory_listener_register(&container->iommu_data.spapr.ramlistener,
> +                             &address_space_memory);
> +
> +    container->iommu_data.spapr.ram_reg_initialized = true;
> +
> +    return container->iommu_data.spapr.ram_reg_error;
>  }
> diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
> index d0b831c..b5ef446 100644
> --- a/include/hw/vfio/vfio-common.h
> +++ b/include/hw/vfio/vfio-common.h
> @@ -71,6 +71,9 @@ typedef struct VFIOType1 {
>  
>  typedef struct VFIOSPAPR {
>      MemoryListener listener;
> +    MemoryListener ramlistener;

The names "listener" and "ramlistener" don't exactly give a clear hint
as to what the difference between them is.  Maybe "register_listener"
for the new one since its purpose is to register dma memory.  I forget
what the purpose of the old one is, and so what a better name for it
might be.

> +    int ram_reg_error;
> +    bool ram_reg_initialized;
>  } VFIOSPAPR;
>  
>  typedef struct VFIOContainer {
> @@ -156,6 +159,6 @@ extern int vfio_dma_unmap(VFIOContainer *container,
>                            hwaddr iova, ram_addr_t size);
>  bool vfio_listener_skipped_section(MemoryRegionSection *section);
>  
> -extern void spapr_memory_listener_register(VFIOContainer *container);
> +extern int spapr_memory_listener_register(VFIOContainer *container, int ver);
>  
>  #endif /* !HW_VFIO_VFIO_COMMON_H */
> diff --git a/trace-events b/trace-events
> index 1231ba4..2739140 100644
> --- a/trace-events
> +++ b/trace-events
> @@ -1563,6 +1563,7 @@ vfio_disconnect_container(int fd) "close container->fd=%d"
>  vfio_put_group(int fd) "close group->fd=%d"
>  vfio_get_device(const char * name, unsigned int flags, unsigned int num_regions, unsigned int num_irqs) "Device %s flags: %u, regions: %u, irqs: %u"
>  vfio_put_base_device(int fd) "close vdev->fd=%d"
> +vfio_ram_register(int req, uint64_t va, uint64_t size, int ret) "req=%d va=%"PRIx64" size=%"PRIx64" ret=%d"
>  
>  #hw/acpi/memory_hotplug.c
>  mhp_acpi_invalid_slot_selected(uint32_t slot) "0x%"PRIx32

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [Qemu-devel] [PATCH qemu v6 07/15] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)

[Alexey Kardashevskiy]

On 04/22/2015 03:53 PM, David Gibson wrote:
> On Sat, Apr 11, 2015 at 01:24:36AM +1000, Alexey Kardashevskiy wrote:
>> This makes use of the new "memory registering" feature. The idea is
>> to provide the userspace ability to notify the host kernel about pages
>> which are going to be used for DMA. Having this information, the host
>> kernel can pin them all once per user process, do locked pages
>> accounting (once) and not spent time on doing that in real time with
>> possible failures which cannot be handled nicely in some cases.
>>
>> This adds a guest RAM memory listener which notifies a VFIO container
>> about memory which needs to be pinned/unpinned. VFIO MMIO regions
>> (i.e. "skip dump" regions) are skipped.
>>
>> The feature is only enabled for SPAPR IOMMU v2. The host kernel changes
>> are required. Since v2 does not need/support VFIO_IOMMU_ENABLE, this does
>> not call it when v2 is detected and enabled.
>>
>> This does not change the guest visible interface.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
>
> Albeit with some nits described below.
>
>> ---
>> Changes:
>> v6:
>> * fixed commit log (s/guest/userspace/), added note about no guest visible
>> change
>> * fixed error checking if ram registration failed
>> * added alignment check for section->offset_within_region
>>
>> v5:
>> * simplified the patch
>> * added trace points
>> * added round_up() for the size
>> * SPAPR IOMMU v2 used
>> ---
>>   hw/vfio/common.c              | 26 +++++++++----
>>   hw/vfio/spapr.c               | 88 ++++++++++++++++++++++++++++++++++++++++++-
>>   include/hw/vfio/vfio-common.h |  5 ++-
>>   trace-events                  |  1 +
>>   4 files changed, 109 insertions(+), 11 deletions(-)
>>
>> diff --git a/hw/vfio/common.c b/hw/vfio/common.c
>> index 369e564..9e3e0b0 100644
>> --- a/hw/vfio/common.c
>> +++ b/hw/vfio/common.c
>> @@ -577,14 +577,18 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
>>
>>           container->iommu_data.type1.initialized = true;
>>
>> -    } else if (ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_IOMMU)) {
>> +    } else if (ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_IOMMU) ||
>> +               ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_v2_IOMMU)) {
>> +        bool v2 = !!ioctl(fd, VFIO_CHECK_EXTENSION, VFIO_SPAPR_TCE_v2_IOMMU);
>> +
>>           ret = ioctl(group->fd, VFIO_GROUP_SET_CONTAINER, &fd);
>>           if (ret) {
>>               error_report("vfio: failed to set group container: %m");
>>               ret = -errno;
>>               goto free_container_exit;
>>           }
>> -        ret = ioctl(fd, VFIO_SET_IOMMU, VFIO_SPAPR_TCE_IOMMU);
>> +        ret = ioctl(fd, VFIO_SET_IOMMU,
>> +                v2 ? VFIO_SPAPR_TCE_v2_IOMMU : VFIO_SPAPR_TCE_IOMMU);
>>           if (ret) {
>>               error_report("vfio: failed to set iommu for container: %m");
>>               ret = -errno;
>> @@ -596,14 +600,20 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as)
>>            * when container fd is closed so we do not call it explicitly
>>            * in this file.
>>            */
>> -        ret = ioctl(fd, VFIO_IOMMU_ENABLE);
>> -        if (ret) {
>> -            error_report("vfio: failed to enable container: %m");
>> -            ret = -errno;
>> -            goto free_container_exit;
>> +        if (!v2) {
>> +            ret = ioctl(fd, VFIO_IOMMU_ENABLE);
>> +            if (ret) {
>> +                error_report("vfio: failed to enable container: %m");
>> +                ret = -errno;
>> +                goto free_container_exit;
>> +            }
>>           }
>>
>> -        spapr_memory_listener_register(container);
>> +        ret = spapr_memory_listener_register(container, v2 ? 2 : 1);
>> +        if (ret) {
>> +            error_report("vfio: RAM memory listener initialization failed for container");
>> +            goto listener_release_exit;
>> +        }
>>
>>       } else {
>>           error_report("vfio: No available IOMMU models");
>> diff --git a/hw/vfio/spapr.c b/hw/vfio/spapr.c
>> index 5f79194..31353f1 100644
>> --- a/hw/vfio/spapr.c
>> +++ b/hw/vfio/spapr.c
>> @@ -17,6 +17,9 @@
>>    *  along with this program; if not, see <http://www.gnu.org/licenses/>.
>>    */
>>
>> +#include <sys/ioctl.h>
>> +#include <linux/vfio.h>
>> +
>>   #include "hw/vfio/vfio-common.h"
>>   #include "qemu/error-report.h"
>>   #include "trace.h"
>> @@ -211,16 +214,97 @@ static const MemoryListener vfio_spapr_memory_listener = {
>>       .region_del = vfio_spapr_listener_region_del,
>>   };
>>
>> +static void vfio_ram_do_region(VFIOContainer *container,
>> +                              MemoryRegionSection *section, unsigned long req)
>> +{
>> +    int ret;
>> +    struct vfio_iommu_spapr_register_memory reg = { .argsz = sizeof(reg) };
>> +
>> +    if (!memory_region_is_ram(section->mr) ||
>> +        memory_region_is_skip_dump(section->mr)) {
>> +        return;
>> +    }
>> +
>> +    if (unlikely((section->offset_within_region & (getpagesize() - 1)))) {
>> +        error_report("%s received unaligned region", __func__);
>> +        return;
>> +    }
>> +
>> +    reg.vaddr = (__u64) memory_region_get_ram_ptr(section->mr) +
>> +        section->offset_within_region;
>> +    reg.size = ROUND_UP(int128_get64(section->size), TARGET_PAGE_SIZE);
>> +
>> +    ret = ioctl(container->fd, req, &reg);
>> +    trace_vfio_ram_register(_IOC_NR(req) - VFIO_BASE, reg.vaddr, reg.size,
>> +            ret ? -errno : 0);
>> +    if (!ret) {
>> +        return;
>> +    }
>> +
>> +    /*
>> +     * On the initfn path, store the first error in the container so we
>> +     * can gracefully fail.  Runtime, there's not much we can do other
>> +     * than throw a hardware error.
>> +     */
>> +    if (!container->iommu_data.spapr.ram_reg_initialized) {
>> +        if (!container->iommu_data.spapr.ram_reg_error) {
>> +            container->iommu_data.spapr.ram_reg_error = -errno;
>> +        }
>> +    } else {
>> +        hw_error("vfio: RAM registering failed, unable to continue");
>> +    }
>> +}
>> +
>> +static void vfio_spapr_ram_listener_region_add(MemoryListener *listener,
>> +                                               MemoryRegionSection *section)
>> +{
>> +    VFIOContainer *container = container_of(listener, VFIOContainer,
>> +                                            iommu_data.spapr.ramlistener);
>> +    memory_region_ref(section->mr);
>> +    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_REGISTER_MEMORY);
>> +}
>> +
>> +static void vfio_spapr_ram_listener_region_del(MemoryListener *listener,
>> +                                               MemoryRegionSection *section)
>> +{
>> +    VFIOContainer *container = container_of(listener, VFIOContainer,
>> +                                            iommu_data.spapr.ramlistener);
>> +    memory_region_unref(section->mr);
>
> You're probably safe because of the QBL, but it would be more
> obviously correct if you did the ioctl unregister before unreferencing
> the MR.


I rather expect the called to pass a referenced section but I'll do what 
you suggested, cannot hurt.


>
>> +    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY);
>> +}
>> +
>> +static const MemoryListener vfio_spapr_ram_memory_listener = {
>> +    .region_add = vfio_spapr_ram_listener_region_add,
>> +    .region_del = vfio_spapr_ram_listener_region_del,
>> +};
>> +
>>   static void vfio_spapr_listener_release(VFIOContainer *container)
>>   {
>>       memory_listener_unregister(&container->iommu_data.spapr.listener);
>>   }
>>
>> -void spapr_memory_listener_register(VFIOContainer *container)
>> +static void vfio_spapr_listener_release_v2(VFIOContainer *container)
>> +{
>> +    memory_listener_unregister(&container->iommu_data.spapr.listener);
>> +    vfio_spapr_listener_release(container);
>> +}
>> +
>> +int spapr_memory_listener_register(VFIOContainer *container, int ver)
>>   {
>>       container->iommu_data.spapr.listener = vfio_spapr_memory_listener;
>>       container->iommu_data.release = vfio_spapr_listener_release;
>> -
>>       memory_listener_register(&container->iommu_data.spapr.listener,
>>                                container->space->as);
>> +    if (ver < 2) {
>> +        return 0;
>> +    }
>
> I wonder if it would make sense to store the IOMMU type value into the
> VFIOContainer (from non-arch specific code).  It would avoid the
> ad-hoc passing of version here and also allow for some sanity checks.


What kind of checks? Cannot think of any which would make sense to do here 
and not in spapr_(pci|rtas)*.c



>> +
>> +    container->iommu_data.spapr.ramlistener = vfio_spapr_ram_memory_listener;
>> +    container->iommu_data.release = vfio_spapr_listener_release_v2;
>> +    memory_listener_register(&container->iommu_data.spapr.ramlistener,
>> +                             &address_space_memory);
>> +
>> +    container->iommu_data.spapr.ram_reg_initialized = true;
>> +
>> +    return container->iommu_data.spapr.ram_reg_error;
>>   }
>> diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
>> index d0b831c..b5ef446 100644
>> --- a/include/hw/vfio/vfio-common.h
>> +++ b/include/hw/vfio/vfio-common.h
>> @@ -71,6 +71,9 @@ typedef struct VFIOType1 {
>>
>>   typedef struct VFIOSPAPR {
>>       MemoryListener listener;
>> +    MemoryListener ramlistener;
>
> The names "listener" and "ramlistener" don't exactly give a clear hint
> as to what the difference between them is.  Maybe "register_listener"
> for the new one since its purpose is to register dma memory.


It is listening for RAM changes. If/when we decide to add something else to 
this listener, we won't have to change its name :) Still change?


> I forget
> what the purpose of the old one is, and so what a better name for it
> might be.

It is listening on PHB address space for IOMMU table updates and translates 
those to MAP/UNMAP ioctls to VFIO containers.



>
>> +    int ram_reg_error;
>> +    bool ram_reg_initialized;
>>   } VFIOSPAPR;
>>
>>   typedef struct VFIOContainer {
>> @@ -156,6 +159,6 @@ extern int vfio_dma_unmap(VFIOContainer *container,
>>                             hwaddr iova, ram_addr_t size);
>>   bool vfio_listener_skipped_section(MemoryRegionSection *section);
>>
>> -extern void spapr_memory_listener_register(VFIOContainer *container);
>> +extern int spapr_memory_listener_register(VFIOContainer *container, int ver);
>>
>>   #endif /* !HW_VFIO_VFIO_COMMON_H */
>> diff --git a/trace-events b/trace-events
>> index 1231ba4..2739140 100644
>> --- a/trace-events
>> +++ b/trace-events
>> @@ -1563,6 +1563,7 @@ vfio_disconnect_container(int fd) "close container->fd=%d"
>>   vfio_put_group(int fd) "close group->fd=%d"
>>   vfio_get_device(const char * name, unsigned int flags, unsigned int num_regions, unsigned int num_irqs) "Device %s flags: %u, regions: %u, irqs: %u"
>>   vfio_put_base_device(int fd) "close vdev->fd=%d"
>> +vfio_ram_register(int req, uint64_t va, uint64_t size, int ret) "req=%d va=%"PRIx64" size=%"PRIx64" ret=%d"
>>
>>   #hw/acpi/memory_hotplug.c
>>   mhp_acpi_invalid_slot_selected(uint32_t slot) "0x%"PRIx32
>


-- 
Alexey

Re: [Qemu-devel] [PATCH qemu v6 07/15] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering)

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 6248 bytes --]

On Wed, Apr 22, 2015 at 07:38:27PM +1000, Alexey Kardashevskiy wrote:
> On 04/22/2015 03:53 PM, David Gibson wrote:
> >On Sat, Apr 11, 2015 at 01:24:36AM +1000, Alexey Kardashevskiy wrote:
> >>This makes use of the new "memory registering" feature. The idea is
> >>to provide the userspace ability to notify the host kernel about pages
> >>which are going to be used for DMA. Having this information, the host
> >>kernel can pin them all once per user process, do locked pages
> >>accounting (once) and not spent time on doing that in real time with
> >>possible failures which cannot be handled nicely in some cases.
> >>
> >>This adds a guest RAM memory listener which notifies a VFIO container
> >>about memory which needs to be pinned/unpinned. VFIO MMIO regions
> >>(i.e. "skip dump" regions) are skipped.
> >>
> >>The feature is only enabled for SPAPR IOMMU v2. The host kernel changes
> >>are required. Since v2 does not need/support VFIO_IOMMU_ENABLE, this does
> >>not call it when v2 is detected and enabled.
> >>
> >>This does not change the guest visible interface.
> >>
> >>Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> >
> >Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> >
> >Albeit with some nits described below.

[snip]
> >>+    vfio_ram_do_region(container, section, VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY);
> >>+}
> >>+
> >>+static const MemoryListener vfio_spapr_ram_memory_listener = {
> >>+    .region_add = vfio_spapr_ram_listener_region_add,
> >>+    .region_del = vfio_spapr_ram_listener_region_del,
> >>+};
> >>+
> >>  static void vfio_spapr_listener_release(VFIOContainer *container)
> >>  {
> >>      memory_listener_unregister(&container->iommu_data.spapr.listener);
> >>  }
> >>
> >>-void spapr_memory_listener_register(VFIOContainer *container)
> >>+static void vfio_spapr_listener_release_v2(VFIOContainer *container)
> >>+{
> >>+    memory_listener_unregister(&container->iommu_data.spapr.listener);
> >>+    vfio_spapr_listener_release(container);
> >>+}
> >>+
> >>+int spapr_memory_listener_register(VFIOContainer *container, int ver)
> >>  {
> >>      container->iommu_data.spapr.listener = vfio_spapr_memory_listener;
> >>      container->iommu_data.release = vfio_spapr_listener_release;
> >>-
> >>      memory_listener_register(&container->iommu_data.spapr.listener,
> >>                               container->space->as);
> >>+    if (ver < 2) {
> >>+        return 0;
> >>+    }
> >
> >I wonder if it would make sense to store the IOMMU type value into the
> >VFIOContainer (from non-arch specific code).  It would avoid the
> >ad-hoc passing of version here and also allow for some sanity checks.
> 
> 
> What kind of checks? Cannot think of any which would make sense to do here
> and not in spapr_(pci|rtas)*.c

Well, all I was thinking of was something like
      assert(container->iommu_type == SPAPR_TCE_TYPE);
In spapr (or x86) callbacks from the core vfio code to make sure we
haven't somehow gotten here with a Type1 container.

> >>+
> >>+    container->iommu_data.spapr.ramlistener = vfio_spapr_ram_memory_listener;
> >>+    container->iommu_data.release = vfio_spapr_listener_release_v2;
> >>+    memory_listener_register(&container->iommu_data.spapr.ramlistener,
> >>+                             &address_space_memory);
> >>+
> >>+    container->iommu_data.spapr.ram_reg_initialized = true;
> >>+
> >>+    return container->iommu_data.spapr.ram_reg_error;
> >>  }
> >>diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
> >>index d0b831c..b5ef446 100644
> >>--- a/include/hw/vfio/vfio-common.h
> >>+++ b/include/hw/vfio/vfio-common.h
> >>@@ -71,6 +71,9 @@ typedef struct VFIOType1 {
> >>
> >>  typedef struct VFIOSPAPR {
> >>      MemoryListener listener;
> >>+    MemoryListener ramlistener;
> >
> >The names "listener" and "ramlistener" don't exactly give a clear hint
> >as to what the difference between them is.  Maybe "register_listener"
> >for the new one since its purpose is to register dma memory.
> 
> 
> It is listening for RAM changes. If/when we decide to add something else to
> this listener, we won't have to change its name :) Still change?

Ah, I see.  I was forgetting that the other listener wasn't really
about RAM (I think an understandable error for a *Memory*Listener).

But actually I'm having trouble thinking of names that are clearly
better than the current ones.  So, I guess I don't really care.

> 
> 
> >I forget
> >what the purpose of the old one is, and so what a better name for it
> >might be.
> 
> It is listening on PHB address space for IOMMU table updates and translates
> those to MAP/UNMAP ioctls to VFIO containers.
> 
> 
> 
> >
> >>+    int ram_reg_error;
> >>+    bool ram_reg_initialized;
> >>  } VFIOSPAPR;
> >>
> >>  typedef struct VFIOContainer {
> >>@@ -156,6 +159,6 @@ extern int vfio_dma_unmap(VFIOContainer *container,
> >>                            hwaddr iova, ram_addr_t size);
> >>  bool vfio_listener_skipped_section(MemoryRegionSection *section);
> >>
> >>-extern void spapr_memory_listener_register(VFIOContainer *container);
> >>+extern int spapr_memory_listener_register(VFIOContainer *container, int ver);
> >>
> >>  #endif /* !HW_VFIO_VFIO_COMMON_H */
> >>diff --git a/trace-events b/trace-events
> >>index 1231ba4..2739140 100644
> >>--- a/trace-events
> >>+++ b/trace-events
> >>@@ -1563,6 +1563,7 @@ vfio_disconnect_container(int fd) "close container->fd=%d"
> >>  vfio_put_group(int fd) "close group->fd=%d"
> >>  vfio_get_device(const char * name, unsigned int flags, unsigned int num_regions, unsigned int num_irqs) "Device %s flags: %u, regions: %u, irqs: %u"
> >>  vfio_put_base_device(int fd) "close vdev->fd=%d"
> >>+vfio_ram_register(int req, uint64_t va, uint64_t size, int ret) "req=%d va=%"PRIx64" size=%"PRIx64" ret=%d"
> >>
> >>  #hw/acpi/memory_hotplug.c
> >>  mhp_acpi_invalid_slot_selected(uint32_t slot) "0x%"PRIx32
> >
> 
> 

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

[Qemu-devel] [PATCH qemu v6 08/15] spapr_iommu: Introduce "enabled" state for TCE table

[Alexey Kardashevskiy]

Currently TCE tables are created once at start and their size never
changes. We are going to change that by introducing a Dynamic DMA windows
support where DMA configuration may change during the guest execution.

This changes spapr_tce_new_table() to create an empty stub object. Only
LIOBN is assigned by the time of creation. It still will be called once
at the owner object (VIO or PHB) creation.

This introduces an "enabled" state for TCE table objects with two
helper functions - spapr_tce_table_enable()/spapr_tce_table_disable().
spapr_tce_table_enable() receives TCE table parameters and allocates
a guest view of the TCE table (in the user space or KVM).
spapr_tce_table_disable() disposes the table.

Follow up patches will disable+enable tables on reset (system reset
or DDW reset).

No visible change in behaviour is expected except the actual table
will be reallocated every reset. We might optimize this later.

The other way to implement this would be dynamically create/remove
the TCE table QOM objects but this would make migration impossible
as migration expects all QOM objects to exist at the receiver
so we have to have TCE table objects created when migration begins.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
Changes:
v6:
* got rid of set_props()
---
 hw/ppc/spapr_iommu.c    | 104 +++++++++++++++++++++++++++++++-----------------
 hw/ppc/spapr_pci.c      |  16 +++++---
 hw/ppc/spapr_pci_vfio.c |  10 ++---
 hw/ppc/spapr_vio.c      |   9 ++---
 include/hw/ppc/spapr.h  |  11 ++---
 5 files changed, 93 insertions(+), 57 deletions(-)

diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
index a14cdc4..64f20f2 100644
--- a/hw/ppc/spapr_iommu.c
+++ b/hw/ppc/spapr_iommu.c
@@ -126,8 +126,47 @@ static MemoryRegionIOMMUOps spapr_iommu_ops = {
 static int spapr_tce_table_realize(DeviceState *dev)
 {
     sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
+
+    QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
+
+    vmstate_register(DEVICE(tcet), tcet->liobn, &vmstate_spapr_tce_table,
+                     tcet);
+
+    return 0;
+}
+
+sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)
+{
+    sPAPRTCETable *tcet;
+    char tmp[64];
+
+    if (spapr_tce_find_by_liobn(liobn)) {
+        fprintf(stderr, "Attempted to create TCE table with duplicate"
+                " LIOBN 0x%x\n", liobn);
+        return NULL;
+    }
+
+    tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
+    tcet->liobn = liobn;
+
+    snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
+    object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
+
+    object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
+
+    trace_spapr_iommu_new_table(tcet->liobn, tcet, tcet->table, tcet->fd);
+
+    return tcet;
+}
+
+static void spapr_tce_table_do_enable(sPAPRTCETable *tcet)
+{
     uint64_t window_size = (uint64_t)tcet->nb_table << tcet->page_shift;
 
+    if (!tcet->nb_table) {
+        return;
+    }
+
     if (kvm_enabled() && !(window_size >> 32)) {
         tcet->table = kvmppc_create_spapr_tce(tcet->liobn,
                                               window_size,
@@ -140,65 +179,56 @@ static int spapr_tce_table_realize(DeviceState *dev)
         tcet->table = g_malloc0(table_size);
     }
 
-    trace_spapr_iommu_new_table(tcet->liobn, tcet, tcet->table, tcet->fd);
-
-    memory_region_init_iommu(&tcet->iommu, OBJECT(dev), &spapr_iommu_ops,
+    memory_region_init_iommu(&tcet->iommu, OBJECT(tcet), &spapr_iommu_ops,
                              "iommu-spapr",
                              (uint64_t)tcet->nb_table << tcet->page_shift);
 
-    QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
-
-    vmstate_register(DEVICE(tcet), tcet->liobn, &vmstate_spapr_tce_table,
-                     tcet);
-
-    return 0;
+    tcet->enabled = true;
 }
 
-sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
-                                   uint64_t bus_offset,
-                                   uint32_t page_shift,
-                                   uint32_t nb_table,
-                                   bool vfio_accel)
+void spapr_tce_table_enable(sPAPRTCETable *tcet,
+                            uint64_t bus_offset, uint32_t page_shift,
+                            uint32_t nb_table, bool vfio_accel)
 {
-    sPAPRTCETable *tcet;
-    char tmp[64];
-
-    if (spapr_tce_find_by_liobn(liobn)) {
-        fprintf(stderr, "Attempted to create TCE table with duplicate"
-                " LIOBN 0x%x\n", liobn);
-        return NULL;
-    }
-
-    if (!nb_table) {
-        return NULL;
+    if (tcet->enabled) {
+        return;
     }
 
-    tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
-    tcet->liobn = liobn;
     tcet->bus_offset = bus_offset;
     tcet->page_shift = page_shift;
     tcet->nb_table = nb_table;
     tcet->vfio_accel = vfio_accel;
 
-    snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
-    object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
-
-    object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
-
-    return tcet;
+    spapr_tce_table_do_enable(tcet);
 }
 
-static void spapr_tce_table_unrealize(DeviceState *dev, Error **errp)
+void spapr_tce_table_disable(sPAPRTCETable *tcet)
 {
-    sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
-
-    QLIST_REMOVE(tcet, list);
+    if (!tcet->enabled) {
+        return;
+    }
 
     if (!kvm_enabled() ||
         (kvmppc_remove_spapr_tce(tcet->table, tcet->fd,
                                  tcet->nb_table) != 0)) {
+        tcet->fd = -1;
         g_free(tcet->table);
     }
+    tcet->table = NULL;
+    tcet->enabled = false;
+    tcet->bus_offset = 0;
+    tcet->page_shift = 0;
+    tcet->nb_table = 0;
+    tcet->vfio_accel = false;
+}
+
+static void spapr_tce_table_unrealize(DeviceState *dev, Error **errp)
+{
+    sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
+
+    QLIST_REMOVE(tcet, list);
+
+    spapr_tce_table_disable(tcet);
 }
 
 MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet)
diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index 8c0d2eb..c3410b8 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -881,6 +881,12 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         sphb->lsi_table[i].irq = irq;
     }
 
+    tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn);
+    if (!tcet) {
+            error_setg(errp, "failed to create TCE table");
+            return;
+    }
+
     info->dma_capabilities_update(sphb);
     info->dma_init_window(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
                           sphb->dma32_window_size);
@@ -908,13 +914,13 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
                                      uint64_t window_size)
 {
     uint64_t bus_offset = sphb->dma32_window_start;
-    sPAPRTCETable *tcet;
+    sPAPRTCETable *tcet = spapr_tce_find_by_liobn(liobn);
 
-    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, bus_offset, page_shift,
-                               window_size >> page_shift,
-                               false);
+    spapr_tce_table_enable(tcet, bus_offset, page_shift,
+                           window_size >> page_shift,
+                           false);
 
-    return tcet ? 0 : -1;
+    return 0;
 }
 
 static int spapr_phb_children_reset(Object *child, void *opaque)
diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index 0ce8e61..a428166 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -49,13 +49,13 @@ static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
                                           uint64_t window_size)
 {
     uint64_t bus_offset = sphb->dma32_window_start;
-    sPAPRTCETable *tcet;
+    sPAPRTCETable *tcet = spapr_tce_find_by_liobn(liobn);
 
-    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, bus_offset, page_shift,
-                               window_size >> page_shift,
-                               true);
+    spapr_tce_table_enable(tcet, bus_offset, page_shift,
+                           window_size >> page_shift,
+                           true);
 
-    return tcet ? 0 : -1;
+    return 0;
 }
 
 static void spapr_phb_vfio_reset(DeviceState *qdev)
diff --git a/hw/ppc/spapr_vio.c b/hw/ppc/spapr_vio.c
index 174033d..3e28835 100644
--- a/hw/ppc/spapr_vio.c
+++ b/hw/ppc/spapr_vio.c
@@ -479,11 +479,10 @@ static void spapr_vio_busdev_realize(DeviceState *qdev, Error **errp)
         memory_region_add_subregion_overlap(&dev->mrroot, 0, &dev->mrbypass, 1);
         address_space_init(&dev->as, &dev->mrroot, qdev->id);
 
-        dev->tcet = spapr_tce_new_table(qdev, liobn,
-                                        0,
-                                        SPAPR_TCE_PAGE_SHIFT,
-                                        pc->rtce_window_size >>
-                                        SPAPR_TCE_PAGE_SHIFT, false);
+        dev->tcet = spapr_tce_new_table(qdev, liobn);
+        spapr_tce_table_enable(dev->tcet, 0, SPAPR_TCE_PAGE_SHIFT,
+                               pc->rtce_window_size >> SPAPR_TCE_PAGE_SHIFT,
+                               false);
         dev->tcet->vdev = dev;
         memory_region_add_subregion_overlap(&dev->mrroot, 0,
                                             spapr_tce_get_iommu(dev->tcet), 2);
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index 7d9ab9d..074d837 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -498,6 +498,7 @@ typedef struct sPAPRTCETable sPAPRTCETable;
 
 struct sPAPRTCETable {
     DeviceState parent;
+    bool enabled;
     uint32_t liobn;
     uint32_t nb_table;
     uint64_t bus_offset;
@@ -515,11 +516,11 @@ sPAPRTCETable *spapr_tce_find_by_liobn(uint32_t liobn);
 void spapr_events_init(sPAPREnvironment *spapr);
 void spapr_events_fdt_skel(void *fdt, uint32_t epow_irq);
 int spapr_h_cas_compose_response(target_ulong addr, target_ulong size);
-sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
-                                   uint64_t bus_offset,
-                                   uint32_t page_shift,
-                                   uint32_t nb_table,
-                                   bool vfio_accel);
+sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
+void spapr_tce_table_enable(sPAPRTCETable *tcet,
+                            uint64_t bus_offset, uint32_t page_shift,
+                            uint32_t nb_table, bool vfio_accel);
+void spapr_tce_table_disable(sPAPRTCETable *tcet);
 MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet);
 int spapr_dma_dt(void *fdt, int node_off, const char *propname,
                  uint32_t liobn, uint64_t window, uint32_t size);
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 08/15] spapr_iommu: Introduce "enabled" state for TCE table

[Thomas Huth]

Am Sat, 11 Apr 2015 01:24:37 +1000
schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:

> Currently TCE tables are created once at start and their size never
> changes. We are going to change that by introducing a Dynamic DMA windows
> support where DMA configuration may change during the guest execution.
> 
> This changes spapr_tce_new_table() to create an empty stub object. Only
> LIOBN is assigned by the time of creation. It still will be called once
> at the owner object (VIO or PHB) creation.
> 
> This introduces an "enabled" state for TCE table objects with two
> helper functions - spapr_tce_table_enable()/spapr_tce_table_disable().
> spapr_tce_table_enable() receives TCE table parameters and allocates
> a guest view of the TCE table (in the user space or KVM).
> spapr_tce_table_disable() disposes the table.
> 
> Follow up patches will disable+enable tables on reset (system reset
> or DDW reset).
> 
> No visible change in behaviour is expected except the actual table
> will be reallocated every reset. We might optimize this later.
> 
> The other way to implement this would be dynamically create/remove
> the TCE table QOM objects but this would make migration impossible
> as migration expects all QOM objects to exist at the receiver
> so we have to have TCE table objects created when migration begins.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> Changes:
> v6:
> * got rid of set_props()
> ---
>  hw/ppc/spapr_iommu.c    | 104 +++++++++++++++++++++++++++++++-----------------
>  hw/ppc/spapr_pci.c      |  16 +++++---
>  hw/ppc/spapr_pci_vfio.c |  10 ++---
>  hw/ppc/spapr_vio.c      |   9 ++---
>  include/hw/ppc/spapr.h  |  11 ++---
>  5 files changed, 93 insertions(+), 57 deletions(-)
> 
> diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
> index a14cdc4..64f20f2 100644
> --- a/hw/ppc/spapr_iommu.c
> +++ b/hw/ppc/spapr_iommu.c
> @@ -126,8 +126,47 @@ static MemoryRegionIOMMUOps spapr_iommu_ops = {
>  static int spapr_tce_table_realize(DeviceState *dev)
>  {
>      sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
> +
> +    QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
> +
> +    vmstate_register(DEVICE(tcet), tcet->liobn, &vmstate_spapr_tce_table,
> +                     tcet);
> +
> +    return 0;
> +}
> +
> +sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)
> +{
> +    sPAPRTCETable *tcet;
> +    char tmp[64];

Hm, any reason for that value "64"? 64 bytes seem much more than enough
here, if I count correctly, you only need 19 bytes max.

> +    if (spapr_tce_find_by_liobn(liobn)) {
> +        fprintf(stderr, "Attempted to create TCE table with duplicate"
> +                " LIOBN 0x%x\n", liobn);
> +        return NULL;
> +    }
> +
> +    tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
> +    tcet->liobn = liobn;
> +
> +    snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
> +    object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
> +
> +    object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
> +
> +    trace_spapr_iommu_new_table(tcet->liobn, tcet, tcet->table, tcet->fd);
> +
> +    return tcet;
> +}

 Thomas

Re: [Qemu-devel] [PATCH qemu v6 08/15] spapr_iommu: Introduce "enabled" state for TCE table

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 11987 bytes --]

On Sat, Apr 11, 2015 at 01:24:37AM +1000, Alexey Kardashevskiy wrote:
> Currently TCE tables are created once at start and their size never
> changes. We are going to change that by introducing a Dynamic DMA windows
> support where DMA configuration may change during the guest execution.
> 
> This changes spapr_tce_new_table() to create an empty stub object. Only
> LIOBN is assigned by the time of creation. It still will be called once
> at the owner object (VIO or PHB) creation.
> 
> This introduces an "enabled" state for TCE table objects with two
> helper functions - spapr_tce_table_enable()/spapr_tce_table_disable().
> spapr_tce_table_enable() receives TCE table parameters and allocates
> a guest view of the TCE table (in the user space or KVM).
> spapr_tce_table_disable() disposes the table.
> 
> Follow up patches will disable+enable tables on reset (system reset
> or DDW reset).
> 
> No visible change in behaviour is expected except the actual table
> will be reallocated every reset. We might optimize this later.
> 
> The other way to implement this would be dynamically create/remove
> the TCE table QOM objects but this would make migration impossible
> as migration expects all QOM objects to exist at the receiver
> so we have to have TCE table objects created when migration begins.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> Changes:
> v6:
> * got rid of set_props()
> ---
>  hw/ppc/spapr_iommu.c    | 104 +++++++++++++++++++++++++++++++-----------------
>  hw/ppc/spapr_pci.c      |  16 +++++---
>  hw/ppc/spapr_pci_vfio.c |  10 ++---
>  hw/ppc/spapr_vio.c      |   9 ++---
>  include/hw/ppc/spapr.h  |  11 ++---
>  5 files changed, 93 insertions(+), 57 deletions(-)
> 
> diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
> index a14cdc4..64f20f2 100644
> --- a/hw/ppc/spapr_iommu.c
> +++ b/hw/ppc/spapr_iommu.c
> @@ -126,8 +126,47 @@ static MemoryRegionIOMMUOps spapr_iommu_ops = {
>  static int spapr_tce_table_realize(DeviceState *dev)
>  {
>      sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
> +
> +    QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
> +
> +    vmstate_register(DEVICE(tcet), tcet->liobn, &vmstate_spapr_tce_table,
> +                     tcet);
> +
> +    return 0;
> +}
> +
> +sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)
> +{
> +    sPAPRTCETable *tcet;
> +    char tmp[64];
> +
> +    if (spapr_tce_find_by_liobn(liobn)) {
> +        fprintf(stderr, "Attempted to create TCE table with duplicate"
> +                " LIOBN 0x%x\n", liobn);
> +        return NULL;
> +    }
> +
> +    tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
> +    tcet->liobn = liobn;
> +
> +    snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
> +    object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
> +
> +    object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
> +
> +    trace_spapr_iommu_new_table(tcet->liobn, tcet, tcet->table, tcet->fd);
> +
> +    return tcet;
> +}
> +
> +static void spapr_tce_table_do_enable(sPAPRTCETable *tcet)

AFAICT there's only one caller of this, so it's not clear why this
isn't just open-coded in spapr_tce_table_enable().

> +{
>      uint64_t window_size = (uint64_t)tcet->nb_table << tcet->page_shift;
>  
> +    if (!tcet->nb_table) {
> +        return;
> +    }
> +
>      if (kvm_enabled() && !(window_size >> 32)) {
>          tcet->table = kvmppc_create_spapr_tce(tcet->liobn,
>                                                window_size,
> @@ -140,65 +179,56 @@ static int spapr_tce_table_realize(DeviceState *dev)
>          tcet->table = g_malloc0(table_size);
>      }
>  
> -    trace_spapr_iommu_new_table(tcet->liobn, tcet, tcet->table, tcet->fd);
> -
> -    memory_region_init_iommu(&tcet->iommu, OBJECT(dev), &spapr_iommu_ops,
> +    memory_region_init_iommu(&tcet->iommu, OBJECT(tcet), &spapr_iommu_ops,
>                               "iommu-spapr",
>                               (uint64_t)tcet->nb_table << tcet->page_shift);
>  
> -    QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
> -
> -    vmstate_register(DEVICE(tcet), tcet->liobn, &vmstate_spapr_tce_table,
> -                     tcet);
> -
> -    return 0;
> +    tcet->enabled = true;
>  }
>  
> -sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
> -                                   uint64_t bus_offset,
> -                                   uint32_t page_shift,
> -                                   uint32_t nb_table,
> -                                   bool vfio_accel)
> +void spapr_tce_table_enable(sPAPRTCETable *tcet,
> +                            uint64_t bus_offset, uint32_t page_shift,
> +                            uint32_t nb_table, bool vfio_accel)
>  {
> -    sPAPRTCETable *tcet;
> -    char tmp[64];
> -
> -    if (spapr_tce_find_by_liobn(liobn)) {
> -        fprintf(stderr, "Attempted to create TCE table with duplicate"
> -                " LIOBN 0x%x\n", liobn);
> -        return NULL;
> -    }
> -
> -    if (!nb_table) {
> -        return NULL;
> +    if (tcet->enabled) {
> +        return;
>      }
>  
> -    tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
> -    tcet->liobn = liobn;
>      tcet->bus_offset = bus_offset;
>      tcet->page_shift = page_shift;
>      tcet->nb_table = nb_table;
>      tcet->vfio_accel = vfio_accel;
>  
> -    snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
> -    object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
> -
> -    object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
> -
> -    return tcet;
> +    spapr_tce_table_do_enable(tcet);
>  }
>  
> -static void spapr_tce_table_unrealize(DeviceState *dev, Error **errp)
> +void spapr_tce_table_disable(sPAPRTCETable *tcet)
>  {
> -    sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
> -
> -    QLIST_REMOVE(tcet, list);
> +    if (!tcet->enabled) {
> +        return;
> +    }
>  
>      if (!kvm_enabled() ||
>          (kvmppc_remove_spapr_tce(tcet->table, tcet->fd,
>                                   tcet->nb_table) != 0)) {
> +        tcet->fd = -1;
>          g_free(tcet->table);
>      }
> +    tcet->table = NULL;
> +    tcet->enabled = false;
> +    tcet->bus_offset = 0;
> +    tcet->page_shift = 0;
> +    tcet->nb_table = 0;
> +    tcet->vfio_accel = false;
> +}
> +
> +static void spapr_tce_table_unrealize(DeviceState *dev, Error **errp)
> +{
> +    sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
> +
> +    QLIST_REMOVE(tcet, list);
> +
> +    spapr_tce_table_disable(tcet);
>  }
>  
>  MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet)
> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> index 8c0d2eb..c3410b8 100644
> --- a/hw/ppc/spapr_pci.c
> +++ b/hw/ppc/spapr_pci.c
> @@ -881,6 +881,12 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>          sphb->lsi_table[i].irq = irq;
>      }
>  
> +    tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn);
> +    if (!tcet) {
> +            error_setg(errp, "failed to create TCE table");
> +            return;
> +    }
> +
>      info->dma_capabilities_update(sphb);
>      info->dma_init_window(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
>                            sphb->dma32_window_size);
> @@ -908,13 +914,13 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
>                                       uint64_t window_size)
>  {
>      uint64_t bus_offset = sphb->dma32_window_start;
> -    sPAPRTCETable *tcet;
> +    sPAPRTCETable *tcet = spapr_tce_find_by_liobn(liobn);
>  
> -    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, bus_offset, page_shift,
> -                               window_size >> page_shift,
> -                               false);
> +    spapr_tce_table_enable(tcet, bus_offset, page_shift,
> +                           window_size >> page_shift,
> +                           false);
>  
> -    return tcet ? 0 : -1;
> +    return 0;
>  }
>  
>  static int spapr_phb_children_reset(Object *child, void *opaque)
> diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
> index 0ce8e61..a428166 100644
> --- a/hw/ppc/spapr_pci_vfio.c
> +++ b/hw/ppc/spapr_pci_vfio.c
> @@ -49,13 +49,13 @@ static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
>                                            uint64_t window_size)
>  {
>      uint64_t bus_offset = sphb->dma32_window_start;
> -    sPAPRTCETable *tcet;
> +    sPAPRTCETable *tcet = spapr_tce_find_by_liobn(liobn);
>  
> -    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, bus_offset, page_shift,
> -                               window_size >> page_shift,
> -                               true);
> +    spapr_tce_table_enable(tcet, bus_offset, page_shift,
> +                           window_size >> page_shift,
> +                           true);
>  
> -    return tcet ? 0 : -1;
> +    return 0;
>  }
>  
>  static void spapr_phb_vfio_reset(DeviceState *qdev)
> diff --git a/hw/ppc/spapr_vio.c b/hw/ppc/spapr_vio.c
> index 174033d..3e28835 100644
> --- a/hw/ppc/spapr_vio.c
> +++ b/hw/ppc/spapr_vio.c
> @@ -479,11 +479,10 @@ static void spapr_vio_busdev_realize(DeviceState *qdev, Error **errp)
>          memory_region_add_subregion_overlap(&dev->mrroot, 0, &dev->mrbypass, 1);
>          address_space_init(&dev->as, &dev->mrroot, qdev->id);
>  
> -        dev->tcet = spapr_tce_new_table(qdev, liobn,
> -                                        0,
> -                                        SPAPR_TCE_PAGE_SHIFT,
> -                                        pc->rtce_window_size >>
> -                                        SPAPR_TCE_PAGE_SHIFT, false);
> +        dev->tcet = spapr_tce_new_table(qdev, liobn);
> +        spapr_tce_table_enable(dev->tcet, 0, SPAPR_TCE_PAGE_SHIFT,
> +                               pc->rtce_window_size >> SPAPR_TCE_PAGE_SHIFT,
> +                               false);
>          dev->tcet->vdev = dev;
>          memory_region_add_subregion_overlap(&dev->mrroot, 0,
>                                              spapr_tce_get_iommu(dev->tcet), 2);
> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
> index 7d9ab9d..074d837 100644
> --- a/include/hw/ppc/spapr.h
> +++ b/include/hw/ppc/spapr.h
> @@ -498,6 +498,7 @@ typedef struct sPAPRTCETable sPAPRTCETable;
>  
>  struct sPAPRTCETable {
>      DeviceState parent;
> +    bool enabled;
>      uint32_t liobn;
>      uint32_t nb_table;
>      uint64_t bus_offset;
> @@ -515,11 +516,11 @@ sPAPRTCETable *spapr_tce_find_by_liobn(uint32_t liobn);
>  void spapr_events_init(sPAPREnvironment *spapr);
>  void spapr_events_fdt_skel(void *fdt, uint32_t epow_irq);
>  int spapr_h_cas_compose_response(target_ulong addr, target_ulong size);
> -sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
> -                                   uint64_t bus_offset,
> -                                   uint32_t page_shift,
> -                                   uint32_t nb_table,
> -                                   bool vfio_accel);
> +sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
> +void spapr_tce_table_enable(sPAPRTCETable *tcet,
> +                            uint64_t bus_offset, uint32_t page_shift,
> +                            uint32_t nb_table, bool vfio_accel);
> +void spapr_tce_table_disable(sPAPRTCETable *tcet);
>  MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet);
>  int spapr_dma_dt(void *fdt, int node_off, const char *propname,
>                   uint32_t liobn, uint64_t window, uint32_t size);

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [Qemu-devel] [PATCH qemu v6 08/15] spapr_iommu: Introduce "enabled" state for TCE table

[Alexey Kardashevskiy]

On 04/22/2015 04:14 PM, David Gibson wrote:
> On Sat, Apr 11, 2015 at 01:24:37AM +1000, Alexey Kardashevskiy wrote:
>> Currently TCE tables are created once at start and their size never
>> changes. We are going to change that by introducing a Dynamic DMA windows
>> support where DMA configuration may change during the guest execution.
>>
>> This changes spapr_tce_new_table() to create an empty stub object. Only
>> LIOBN is assigned by the time of creation. It still will be called once
>> at the owner object (VIO or PHB) creation.
>>
>> This introduces an "enabled" state for TCE table objects with two
>> helper functions - spapr_tce_table_enable()/spapr_tce_table_disable().
>> spapr_tce_table_enable() receives TCE table parameters and allocates
>> a guest view of the TCE table (in the user space or KVM).
>> spapr_tce_table_disable() disposes the table.
>>
>> Follow up patches will disable+enable tables on reset (system reset
>> or DDW reset).
>>
>> No visible change in behaviour is expected except the actual table
>> will be reallocated every reset. We might optimize this later.
>>
>> The other way to implement this would be dynamically create/remove
>> the TCE table QOM objects but this would make migration impossible
>> as migration expects all QOM objects to exist at the receiver
>> so we have to have TCE table objects created when migration begins.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> ---
>> Changes:
>> v6:
>> * got rid of set_props()
>> ---
>>   hw/ppc/spapr_iommu.c    | 104 +++++++++++++++++++++++++++++++-----------------
>>   hw/ppc/spapr_pci.c      |  16 +++++---
>>   hw/ppc/spapr_pci_vfio.c |  10 ++---
>>   hw/ppc/spapr_vio.c      |   9 ++---
>>   include/hw/ppc/spapr.h  |  11 ++---
>>   5 files changed, 93 insertions(+), 57 deletions(-)
>>
>> diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
>> index a14cdc4..64f20f2 100644
>> --- a/hw/ppc/spapr_iommu.c
>> +++ b/hw/ppc/spapr_iommu.c
>> @@ -126,8 +126,47 @@ static MemoryRegionIOMMUOps spapr_iommu_ops = {
>>   static int spapr_tce_table_realize(DeviceState *dev)
>>   {
>>       sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
>> +
>> +    QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
>> +
>> +    vmstate_register(DEVICE(tcet), tcet->liobn, &vmstate_spapr_tce_table,
>> +                     tcet);
>> +
>> +    return 0;
>> +}
>> +
>> +sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)
>> +{
>> +    sPAPRTCETable *tcet;
>> +    char tmp[64];
>> +
>> +    if (spapr_tce_find_by_liobn(liobn)) {
>> +        fprintf(stderr, "Attempted to create TCE table with duplicate"
>> +                " LIOBN 0x%x\n", liobn);
>> +        return NULL;
>> +    }
>> +
>> +    tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
>> +    tcet->liobn = liobn;
>> +
>> +    snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
>> +    object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
>> +
>> +    object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
>> +
>> +    trace_spapr_iommu_new_table(tcet->liobn, tcet, tcet->table, tcet->fd);
>> +
>> +    return tcet;
>> +}
>> +
>> +static void spapr_tce_table_do_enable(sPAPRTCETable *tcet)
>
> AFAICT there's only one caller of this, so it's not clear why this
> isn't just open-coded in spapr_tce_table_enable().


There is another call in "[PATCH qemu v6 14/15] spapr_pci/spapr_pci_vfio: 
Support Dynamic DMA Windows (DDW)", in spapr_tce_table_post_load().

Should I put a note in the commit log?


-- 
Alexey

[Qemu-devel] [PATCH qemu v6 09/15] spapr_iommu: Add root memory region

[Alexey Kardashevskiy]

We are going to have multiple DMA windows at different offsets on
a PCI bus. For the sake of migration, we will have as many TCE table
objects pre-created as many windows supported.
So we need a way to map windows dynamically onto a PCI bus
when migration of a table is completed but at this stage a TCE table
object does not have access to a PHB to ask it to map a DMA window
backed by just migrated TCE table.

This adds a "root" memory region (UINT64_MAX long) to the TCE object.
This new region is mapped on a PCI bus with enabled overlapping as
there will be one root MR per TCE table, each of them mapped at 0.
The actual IOMMU memory region is a subregion of the root region and
a TCE table enables/disables this subregion and maps it at
the specific offset inside the root MR which is 1:1 mapping of
a PCI address space.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/ppc/spapr_iommu.c   | 9 ++++++++-
 hw/ppc/spapr_pci.c     | 2 +-
 include/hw/ppc/spapr.h | 2 +-
 3 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
index 64f20f2..e655bf4 100644
--- a/hw/ppc/spapr_iommu.c
+++ b/hw/ppc/spapr_iommu.c
@@ -150,6 +150,8 @@ sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)
     tcet->liobn = liobn;
 
     snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
+    memory_region_init(&tcet->root, OBJECT(tcet), tmp, UINT64_MAX);
+
     object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
 
     object_property_set_bool(OBJECT(tcet), true, "realized", NULL);
@@ -183,6 +185,8 @@ static void spapr_tce_table_do_enable(sPAPRTCETable *tcet)
                              "iommu-spapr",
                              (uint64_t)tcet->nb_table << tcet->page_shift);
 
+    memory_region_add_subregion(&tcet->root, tcet->bus_offset, &tcet->iommu);
+
     tcet->enabled = true;
 }
 
@@ -208,6 +212,8 @@ void spapr_tce_table_disable(sPAPRTCETable *tcet)
         return;
     }
 
+    memory_region_del_subregion(&tcet->root, &tcet->iommu);
+
     if (!kvm_enabled() ||
         (kvmppc_remove_spapr_tce(tcet->table, tcet->fd,
                                  tcet->nb_table) != 0)) {
@@ -215,6 +221,7 @@ void spapr_tce_table_disable(sPAPRTCETable *tcet)
         g_free(tcet->table);
     }
     tcet->table = NULL;
+    object_unref(OBJECT(&tcet->iommu));
     tcet->enabled = false;
     tcet->bus_offset = 0;
     tcet->page_shift = 0;
@@ -233,7 +240,7 @@ static void spapr_tce_table_unrealize(DeviceState *dev, Error **errp)
 
 MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet)
 {
-    return &tcet->iommu;
+    return &tcet->root;
 }
 
 static void spapr_tce_reset(DeviceState *dev)
diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index c3410b8..664687c 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -895,7 +895,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         error_setg(errp, "failed to create TCE table");
         return;
     }
-    memory_region_add_subregion(&sphb->iommu_root, tcet->bus_offset,
+    memory_region_add_subregion(&sphb->iommu_root, 0,
                                 spapr_tce_get_iommu(tcet));
 
     sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index 074d837..c8ac03f 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -507,7 +507,7 @@ struct sPAPRTCETable {
     bool bypass;
     bool vfio_accel;
     int fd;
-    MemoryRegion iommu;
+    MemoryRegion root, iommu;
     struct VIOsPAPRDevice *vdev; /* for @bypass migration compatibility only */
     QLIST_ENTRY(sPAPRTCETable) list;
 };
-- 
2.0.0

[Qemu-devel] [PATCH qemu v6 10/15] spapr_pci_vfio: Remove unnecessary cast to sPAPRPHBVFIOState

[Alexey Kardashevskiy]

EEH handlers only need AddressSpace to get to the right VFIO container
to call ioctl() so remove unnecessary cast of sPAPRPHBState to
sPAPRPHBVFIOState.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/ppc/spapr_pci_vfio.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index a428166..a5b97d0 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -66,7 +66,6 @@ static void spapr_phb_vfio_reset(DeviceState *qdev)
 static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
                                          unsigned int addr, int option)
 {
-    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
 
@@ -103,7 +102,7 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
         return RTAS_OUT_PARAM_ERROR;
     }
 
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
+    ret = vfio_container_ioctl(&sphb->iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_HW_ERROR;
@@ -114,12 +113,11 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
 
 static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
 {
-    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
 
     op.op = VFIO_EEH_PE_GET_STATE;
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
+    ret = vfio_container_ioctl(&sphb->iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_PARAM_ERROR;
@@ -131,7 +129,6 @@ static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
 
 static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
 {
-    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
 
@@ -149,7 +146,7 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
         return RTAS_OUT_PARAM_ERROR;
     }
 
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
+    ret = vfio_container_ioctl(&sphb->iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_HW_ERROR;
@@ -160,12 +157,11 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
 
 static int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
 {
-    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
 
     op.op = VFIO_EEH_PE_CONFIGURE;
-    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
+    ret = vfio_container_ioctl(&sphb->iommu_as,
                                VFIO_EEH_PE_OP, &op);
     if (ret < 0) {
         return RTAS_OUT_PARAM_ERROR;
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 10/15] spapr_pci_vfio: Remove unnecessary cast to sPAPRPHBVFIOState

[Thomas Huth]

Am Sat, 11 Apr 2015 01:24:39 +1000
schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:

> EEH handlers only need AddressSpace to get to the right VFIO container
> to call ioctl() so remove unnecessary cast of sPAPRPHBState to
> sPAPRPHBVFIOState.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
>  hw/ppc/spapr_pci_vfio.c | 12 ++++--------
>  1 file changed, 4 insertions(+), 8 deletions(-)
> 
> diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
> index a428166..a5b97d0 100644
> --- a/hw/ppc/spapr_pci_vfio.c
> +++ b/hw/ppc/spapr_pci_vfio.c
> @@ -66,7 +66,6 @@ static void spapr_phb_vfio_reset(DeviceState *qdev)
>  static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>                                           unsigned int addr, int option)
>  {
> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>      struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>      int ret;
>  
> @@ -103,7 +102,7 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>          return RTAS_OUT_PARAM_ERROR;
>      }
>  
> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>                                 VFIO_EEH_PE_OP, &op);
>      if (ret < 0) {
>          return RTAS_OUT_HW_ERROR;
> @@ -114,12 +113,11 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>  
>  static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
>  {
> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>      struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>      int ret;
>  
>      op.op = VFIO_EEH_PE_GET_STATE;
> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>                                 VFIO_EEH_PE_OP, &op);
>      if (ret < 0) {
>          return RTAS_OUT_PARAM_ERROR;
> @@ -131,7 +129,6 @@ static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
>  
>  static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
>  {
> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>      struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>      int ret;
>  
> @@ -149,7 +146,7 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
>          return RTAS_OUT_PARAM_ERROR;
>      }
>  
> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>                                 VFIO_EEH_PE_OP, &op);
>      if (ret < 0) {
>          return RTAS_OUT_HW_ERROR;
> @@ -160,12 +157,11 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
>  
>  static int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
>  {
> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>      struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>      int ret;
>  
>      op.op = VFIO_EEH_PE_CONFIGURE;
> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>                                 VFIO_EEH_PE_OP, &op);
>      if (ret < 0) {
>          return RTAS_OUT_PARAM_ERROR;

Patch looks fine to me, so:

Reviewed-by: Thomas Huth <thuth@redhat.com>

But maybe you could even squash this into the patch where you removed the
iommugroupid parameters (patch #5) ... so that you don't have to touch all
these lines twice?

Re: [Qemu-devel] [PATCH qemu v6 10/15] spapr_pci_vfio: Remove unnecessary cast to sPAPRPHBVFIOState

[Alexey Kardashevskiy]

On 04/16/2015 08:31 PM, Thomas Huth wrote:
> Am Sat, 11 Apr 2015 01:24:39 +1000
> schrieb Alexey Kardashevskiy <aik@ozlabs.ru>:
>
>> EEH handlers only need AddressSpace to get to the right VFIO container
>> to call ioctl() so remove unnecessary cast of sPAPRPHBState to
>> sPAPRPHBVFIOState.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> ---
>>   hw/ppc/spapr_pci_vfio.c | 12 ++++--------
>>   1 file changed, 4 insertions(+), 8 deletions(-)
>>
>> diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
>> index a428166..a5b97d0 100644
>> --- a/hw/ppc/spapr_pci_vfio.c
>> +++ b/hw/ppc/spapr_pci_vfio.c
>> @@ -66,7 +66,6 @@ static void spapr_phb_vfio_reset(DeviceState *qdev)
>>   static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>>                                            unsigned int addr, int option)
>>   {
>> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>>       struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>>       int ret;
>>
>> @@ -103,7 +102,7 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>>           return RTAS_OUT_PARAM_ERROR;
>>       }
>>
>> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
>> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>>                                  VFIO_EEH_PE_OP, &op);
>>       if (ret < 0) {
>>           return RTAS_OUT_HW_ERROR;
>> @@ -114,12 +113,11 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>>
>>   static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
>>   {
>> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>>       struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>>       int ret;
>>
>>       op.op = VFIO_EEH_PE_GET_STATE;
>> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
>> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>>                                  VFIO_EEH_PE_OP, &op);
>>       if (ret < 0) {
>>           return RTAS_OUT_PARAM_ERROR;
>> @@ -131,7 +129,6 @@ static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
>>
>>   static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
>>   {
>> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>>       struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>>       int ret;
>>
>> @@ -149,7 +146,7 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
>>           return RTAS_OUT_PARAM_ERROR;
>>       }
>>
>> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
>> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>>                                  VFIO_EEH_PE_OP, &op);
>>       if (ret < 0) {
>>           return RTAS_OUT_HW_ERROR;
>> @@ -160,12 +157,11 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
>>
>>   static int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
>>   {
>> -    sPAPRPHBVFIOState *svphb = SPAPR_PCI_VFIO_HOST_BRIDGE(sphb);
>>       struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
>>       int ret;
>>
>>       op.op = VFIO_EEH_PE_CONFIGURE;
>> -    ret = vfio_container_ioctl(&svphb->phb.iommu_as,
>> +    ret = vfio_container_ioctl(&sphb->iommu_as,
>>                                  VFIO_EEH_PE_OP, &op);
>>       if (ret < 0) {
>>           return RTAS_OUT_PARAM_ERROR;
>
> Patch looks fine to me, so:
>
> Reviewed-by: Thomas Huth <thuth@redhat.com>
>
> But maybe you could even squash this into the patch where you removed the
> iommugroupid parameters (patch #5) ... so that you don't have to touch all
> these lines twice?

May be... Patch #5 enables multiple groups per container, and this one 
prepares for spapr-vfio-phb merge so they solve different issues in general 
and I was told multiple times the keeping patches simpler is a good thing 
(to reasonable extent) :)

Anyway in the next respin I'll remove more of EEH stuff, coming soon!


-- 
Alexey

Re: [Qemu-devel] [PATCH qemu v6 10/15] spapr_pci_vfio: Remove unnecessary cast to sPAPRPHBVFIOState

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 547 bytes --]

On Sat, Apr 11, 2015 at 01:24:39AM +1000, Alexey Kardashevskiy wrote:
> EEH handlers only need AddressSpace to get to the right VFIO container
> to call ioctl() so remove unnecessary cast of sPAPRPHBState to
> sPAPRPHBVFIOState.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

[Qemu-devel] [PATCH qemu v6 11/15] spapr_pci: Do complete reset of DMA config when resetting PHB

[Alexey Kardashevskiy]

On a system reset, DMA configuration has to reset too. At the moment
it clears the table content. This is enough for the single table case
but with DDW, we will also have to disable all DMA windows except
the default one. Furthermore according to sPAPR, if the guest removed
the default window and created a huge one at the same zero offset on
a PCI bus, the reset handler has to recreate the default window with
the default properties (2GB big, 4K pages).

This reworks SPAPR PHB code to disable the existing DMA window on reset
and then configure and enable the default window.
Without DDW that means that the same window will be disabled and then
enabled with no other change in behaviour.

This changes the table creation to do it in one place in PHB (VFIO PHB
just inherits the behaviour from PHB). The actual table allocation is
done from the reset handler and this is where finish_realize() is called.

This disables all DMA windows on a PHB reset. It does not make any
difference now as there is just one DMA window but it will later with DDW
patches.

spapr_phb_dma_reset() and spapr_phb_dma_remove_window() will be used
in following patch so make it public.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/ppc/spapr_pci.c          | 45 ++++++++++++++++++++++++++++++++++++---------
 hw/ppc/spapr_pci_vfio.c     |  6 ------
 include/hw/pci-host/spapr.h |  3 +++
 3 files changed, 39 insertions(+), 15 deletions(-)

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index 664687c..3d40f5b 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -736,7 +736,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
     SysBusDevice *s = SYS_BUS_DEVICE(dev);
     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
     PCIHostState *phb = PCI_HOST_BRIDGE(s);
-    sPAPRPHBClass *info = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(s);
     char *namebuf;
     int i;
     PCIBus *bus;
@@ -887,14 +886,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
             return;
     }
 
-    info->dma_capabilities_update(sphb);
-    info->dma_init_window(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
-                          sphb->dma32_window_size);
-    tcet = spapr_tce_find_by_liobn(sphb->dma_liobn);
-    if (!tcet) {
-        error_setg(errp, "failed to create TCE table");
-        return;
-    }
     memory_region_add_subregion(&sphb->iommu_root, 0,
                                 spapr_tce_get_iommu(tcet));
 
@@ -923,6 +914,40 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
     return 0;
 }
 
+int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
+                                sPAPRTCETable *tcet)
+{
+    spapr_tce_table_disable(tcet);
+
+    return 0;
+}
+
+static int spapr_phb_disable_dma_windows(Object *child, void *opaque)
+{
+    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
+    sPAPRTCETable *tcet = (sPAPRTCETable *)
+        object_dynamic_cast(child, TYPE_SPAPR_TCE_TABLE);
+
+    if (tcet) {
+        spapr_phb_dma_remove_window(sphb, tcet);
+    }
+
+    return 0;
+}
+
+int spapr_phb_dma_reset(sPAPRPHBState *sphb)
+{
+    const uint32_t liobn = SPAPR_PCI_LIOBN(sphb->index, 0);
+    sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
+
+    spc->dma_capabilities_update(sphb); /* Refresh @has_vfio status */
+    object_child_foreach(OBJECT(sphb), spapr_phb_disable_dma_windows, sphb);
+    spc->dma_init_window(sphb, liobn, SPAPR_TCE_PAGE_SHIFT,
+                         sphb->dma32_window_size);
+
+    return 0;
+}
+
 static int spapr_phb_children_reset(Object *child, void *opaque)
 {
     DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
@@ -936,6 +961,8 @@ static int spapr_phb_children_reset(Object *child, void *opaque)
 
 static void spapr_phb_reset(DeviceState *qdev)
 {
+    spapr_phb_dma_reset(SPAPR_PCI_HOST_BRIDGE(qdev));
+
     /* Reset the IOMMU state */
     object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
 }
diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index a5b97d0..f89e053 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -58,11 +58,6 @@ static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
     return 0;
 }
 
-static void spapr_phb_vfio_reset(DeviceState *qdev)
-{
-    /* Do nothing */
-}
-
 static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
                                          unsigned int addr, int option)
 {
@@ -176,7 +171,6 @@ static void spapr_phb_vfio_class_init(ObjectClass *klass, void *data)
     sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_CLASS(klass);
 
     dc->props = spapr_phb_vfio_properties;
-    dc->reset = spapr_phb_vfio_reset;
     spc->dma_capabilities_update = spapr_phb_vfio_dma_capabilities_update;
     spc->dma_init_window = spapr_phb_vfio_dma_init_window;
     spc->eeh_set_option = spapr_phb_vfio_eeh_set_option;
diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
index 3074145..7fda78e 100644
--- a/include/hw/pci-host/spapr.h
+++ b/include/hw/pci-host/spapr.h
@@ -140,5 +140,8 @@ void spapr_pci_rtas_init(void);
 sPAPRPHBState *spapr_pci_find_phb(sPAPREnvironment *spapr, uint64_t buid);
 PCIDevice *spapr_pci_find_dev(sPAPREnvironment *spapr, uint64_t buid,
                               uint32_t config_addr);
+int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
+                                sPAPRTCETable *tcet);
+int spapr_phb_dma_reset(sPAPRPHBState *sphb);
 
 #endif /* __HW_SPAPR_PCI_H__ */
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 11/15] spapr_pci: Do complete reset of DMA config when resetting PHB

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 6385 bytes --]

On Sat, Apr 11, 2015 at 01:24:40AM +1000, Alexey Kardashevskiy wrote:
> On a system reset, DMA configuration has to reset too. At the moment
> it clears the table content. This is enough for the single table case
> but with DDW, we will also have to disable all DMA windows except
> the default one. Furthermore according to sPAPR, if the guest removed
> the default window and created a huge one at the same zero offset on
> a PCI bus, the reset handler has to recreate the default window with
> the default properties (2GB big, 4K pages).
> 
> This reworks SPAPR PHB code to disable the existing DMA window on reset
> and then configure and enable the default window.
> Without DDW that means that the same window will be disabled and then
> enabled with no other change in behaviour.
> 
> This changes the table creation to do it in one place in PHB (VFIO PHB
> just inherits the behaviour from PHB). The actual table allocation is
> done from the reset handler and this is where finish_realize() is called.

Looks like your commit message needs an update - you already got rid
of finish_realize() at this point in the series.

> This disables all DMA windows on a PHB reset. It does not make any
> difference now as there is just one DMA window but it will later with DDW
> patches.
> 
> spapr_phb_dma_reset() and spapr_phb_dma_remove_window() will be used
> in following patch so make it public.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
>  hw/ppc/spapr_pci.c          | 45 ++++++++++++++++++++++++++++++++++++---------
>  hw/ppc/spapr_pci_vfio.c     |  6 ------
>  include/hw/pci-host/spapr.h |  3 +++
>  3 files changed, 39 insertions(+), 15 deletions(-)
> 
> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> index 664687c..3d40f5b 100644
> --- a/hw/ppc/spapr_pci.c
> +++ b/hw/ppc/spapr_pci.c
> @@ -736,7 +736,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>      SysBusDevice *s = SYS_BUS_DEVICE(dev);
>      sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
>      PCIHostState *phb = PCI_HOST_BRIDGE(s);
> -    sPAPRPHBClass *info = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(s);
>      char *namebuf;
>      int i;
>      PCIBus *bus;
> @@ -887,14 +886,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>              return;
>      }
>  
> -    info->dma_capabilities_update(sphb);
> -    info->dma_init_window(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
> -                          sphb->dma32_window_size);
> -    tcet = spapr_tce_find_by_liobn(sphb->dma_liobn);
> -    if (!tcet) {
> -        error_setg(errp, "failed to create TCE table");
> -        return;
> -    }
>      memory_region_add_subregion(&sphb->iommu_root, 0,
>                                  spapr_tce_get_iommu(tcet));
>  
> @@ -923,6 +914,40 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
>      return 0;
>  }
>  
> +int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
> +                                sPAPRTCETable *tcet)

The only caller of this is just below, why is it not static?

> +{
> +    spapr_tce_table_disable(tcet);
> +
> +    return 0;
> +}
> +
> +static int spapr_phb_disable_dma_windows(Object *child, void *opaque)
> +{
> +    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
> +    sPAPRTCETable *tcet = (sPAPRTCETable *)
> +        object_dynamic_cast(child, TYPE_SPAPR_TCE_TABLE);
> +
> +    if (tcet) {
> +        spapr_phb_dma_remove_window(sphb, tcet);
> +    }
> +
> +    return 0;
> +}
> +
> +int spapr_phb_dma_reset(sPAPRPHBState *sphb)
> +{
> +    const uint32_t liobn = SPAPR_PCI_LIOBN(sphb->index, 0);
> +    sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
> +
> +    spc->dma_capabilities_update(sphb); /* Refresh @has_vfio status */
> +    object_child_foreach(OBJECT(sphb), spapr_phb_disable_dma_windows, sphb);
> +    spc->dma_init_window(sphb, liobn, SPAPR_TCE_PAGE_SHIFT,
> +                         sphb->dma32_window_size);
> +
> +    return 0;
> +}
> +
>  static int spapr_phb_children_reset(Object *child, void *opaque)
>  {
>      DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
> @@ -936,6 +961,8 @@ static int spapr_phb_children_reset(Object *child, void *opaque)
>  
>  static void spapr_phb_reset(DeviceState *qdev)
>  {
> +    spapr_phb_dma_reset(SPAPR_PCI_HOST_BRIDGE(qdev));
> +
>      /* Reset the IOMMU state */
>      object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
>  }
> diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
> index a5b97d0..f89e053 100644
> --- a/hw/ppc/spapr_pci_vfio.c
> +++ b/hw/ppc/spapr_pci_vfio.c
> @@ -58,11 +58,6 @@ static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
>      return 0;
>  }
>  
> -static void spapr_phb_vfio_reset(DeviceState *qdev)
> -{
> -    /* Do nothing */
> -}
> -
>  static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>                                           unsigned int addr, int option)
>  {
> @@ -176,7 +171,6 @@ static void spapr_phb_vfio_class_init(ObjectClass *klass, void *data)
>      sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_CLASS(klass);
>  
>      dc->props = spapr_phb_vfio_properties;
> -    dc->reset = spapr_phb_vfio_reset;
>      spc->dma_capabilities_update = spapr_phb_vfio_dma_capabilities_update;
>      spc->dma_init_window = spapr_phb_vfio_dma_init_window;
>      spc->eeh_set_option = spapr_phb_vfio_eeh_set_option;
> diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
> index 3074145..7fda78e 100644
> --- a/include/hw/pci-host/spapr.h
> +++ b/include/hw/pci-host/spapr.h
> @@ -140,5 +140,8 @@ void spapr_pci_rtas_init(void);
>  sPAPRPHBState *spapr_pci_find_phb(sPAPREnvironment *spapr, uint64_t buid);
>  PCIDevice *spapr_pci_find_dev(sPAPREnvironment *spapr, uint64_t buid,
>                                uint32_t config_addr);
> +int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
> +                                sPAPRTCETable *tcet);
> +int spapr_phb_dma_reset(sPAPRPHBState *sphb);
>  
>  #endif /* __HW_SPAPR_PCI_H__ */

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [Qemu-devel] [PATCH qemu v6 11/15] spapr_pci: Do complete reset of DMA config when resetting PHB

[Alexey Kardashevskiy]

On 04/22/2015 04:39 PM, David Gibson wrote:
> On Sat, Apr 11, 2015 at 01:24:40AM +1000, Alexey Kardashevskiy wrote:
>> On a system reset, DMA configuration has to reset too. At the moment
>> it clears the table content. This is enough for the single table case
>> but with DDW, we will also have to disable all DMA windows except
>> the default one. Furthermore according to sPAPR, if the guest removed
>> the default window and created a huge one at the same zero offset on
>> a PCI bus, the reset handler has to recreate the default window with
>> the default properties (2GB big, 4K pages).
>>
>> This reworks SPAPR PHB code to disable the existing DMA window on reset
>> and then configure and enable the default window.
>> Without DDW that means that the same window will be disabled and then
>> enabled with no other change in behaviour.
>>
>> This changes the table creation to do it in one place in PHB (VFIO PHB
>> just inherits the behaviour from PHB). The actual table allocation is
>> done from the reset handler and this is where finish_realize() is called.
>
> Looks like your commit message needs an update - you already got rid
> of finish_realize() at this point in the series.
>
>> This disables all DMA windows on a PHB reset. It does not make any
>> difference now as there is just one DMA window but it will later with DDW
>> patches.
>>
>> spapr_phb_dma_reset() and spapr_phb_dma_remove_window() will be used
>> in following patch so make it public.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> ---
>>   hw/ppc/spapr_pci.c          | 45 ++++++++++++++++++++++++++++++++++++---------
>>   hw/ppc/spapr_pci_vfio.c     |  6 ------
>>   include/hw/pci-host/spapr.h |  3 +++
>>   3 files changed, 39 insertions(+), 15 deletions(-)
>>
>> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
>> index 664687c..3d40f5b 100644
>> --- a/hw/ppc/spapr_pci.c
>> +++ b/hw/ppc/spapr_pci.c
>> @@ -736,7 +736,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>>       SysBusDevice *s = SYS_BUS_DEVICE(dev);
>>       sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
>>       PCIHostState *phb = PCI_HOST_BRIDGE(s);
>> -    sPAPRPHBClass *info = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(s);
>>       char *namebuf;
>>       int i;
>>       PCIBus *bus;
>> @@ -887,14 +886,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>>               return;
>>       }
>>
>> -    info->dma_capabilities_update(sphb);
>> -    info->dma_init_window(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
>> -                          sphb->dma32_window_size);
>> -    tcet = spapr_tce_find_by_liobn(sphb->dma_liobn);
>> -    if (!tcet) {
>> -        error_setg(errp, "failed to create TCE table");
>> -        return;
>> -    }
>>       memory_region_add_subregion(&sphb->iommu_root, 0,
>>                                   spapr_tce_get_iommu(tcet));
>>
>> @@ -923,6 +914,40 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
>>       return 0;
>>   }
>>
>> +int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
>> +                                sPAPRTCETable *tcet)
>
> The only caller of this is just below, why is it not static?


rtas_ibm_remove_pe_dma_window() will call it in "[PATCH qemu v6 14/15] 
spapr_pci/spapr_pci_vfio: Support Dynamic DMA Windows (DDW)". I do not see 
much sense in making it static first and then making it public in later 
patch, should I fix this anyway (do not care that much) or updated commit 
log will be enough?




>
>> +{
>> +    spapr_tce_table_disable(tcet);
>> +
>> +    return 0;
>> +}
>> +
>> +static int spapr_phb_disable_dma_windows(Object *child, void *opaque)
>> +{
>> +    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
>> +    sPAPRTCETable *tcet = (sPAPRTCETable *)
>> +        object_dynamic_cast(child, TYPE_SPAPR_TCE_TABLE);
>> +
>> +    if (tcet) {
>> +        spapr_phb_dma_remove_window(sphb, tcet);
>> +    }
>> +
>> +    return 0;
>> +}
>> +
>> +int spapr_phb_dma_reset(sPAPRPHBState *sphb)
>> +{
>> +    const uint32_t liobn = SPAPR_PCI_LIOBN(sphb->index, 0);
>> +    sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
>> +
>> +    spc->dma_capabilities_update(sphb); /* Refresh @has_vfio status */
>> +    object_child_foreach(OBJECT(sphb), spapr_phb_disable_dma_windows, sphb);
>> +    spc->dma_init_window(sphb, liobn, SPAPR_TCE_PAGE_SHIFT,
>> +                         sphb->dma32_window_size);
>> +
>> +    return 0;
>> +}
>> +
>>   static int spapr_phb_children_reset(Object *child, void *opaque)
>>   {
>>       DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
>> @@ -936,6 +961,8 @@ static int spapr_phb_children_reset(Object *child, void *opaque)
>>
>>   static void spapr_phb_reset(DeviceState *qdev)
>>   {
>> +    spapr_phb_dma_reset(SPAPR_PCI_HOST_BRIDGE(qdev));
>> +
>>       /* Reset the IOMMU state */
>>       object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
>>   }
>> diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
>> index a5b97d0..f89e053 100644
>> --- a/hw/ppc/spapr_pci_vfio.c
>> +++ b/hw/ppc/spapr_pci_vfio.c
>> @@ -58,11 +58,6 @@ static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
>>       return 0;
>>   }
>>
>> -static void spapr_phb_vfio_reset(DeviceState *qdev)
>> -{
>> -    /* Do nothing */
>> -}
>> -
>>   static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
>>                                            unsigned int addr, int option)
>>   {
>> @@ -176,7 +171,6 @@ static void spapr_phb_vfio_class_init(ObjectClass *klass, void *data)
>>       sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_CLASS(klass);
>>
>>       dc->props = spapr_phb_vfio_properties;
>> -    dc->reset = spapr_phb_vfio_reset;
>>       spc->dma_capabilities_update = spapr_phb_vfio_dma_capabilities_update;
>>       spc->dma_init_window = spapr_phb_vfio_dma_init_window;
>>       spc->eeh_set_option = spapr_phb_vfio_eeh_set_option;
>> diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
>> index 3074145..7fda78e 100644
>> --- a/include/hw/pci-host/spapr.h
>> +++ b/include/hw/pci-host/spapr.h
>> @@ -140,5 +140,8 @@ void spapr_pci_rtas_init(void);
>>   sPAPRPHBState *spapr_pci_find_phb(sPAPREnvironment *spapr, uint64_t buid);
>>   PCIDevice *spapr_pci_find_dev(sPAPREnvironment *spapr, uint64_t buid,
>>                                 uint32_t config_addr);
>> +int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
>> +                                sPAPRTCETable *tcet);
>> +int spapr_phb_dma_reset(sPAPRPHBState *sphb);
>>
>>   #endif /* __HW_SPAPR_PCI_H__ */
>


-- 
Alexey

[Qemu-devel] [PATCH qemu v6 12/15] spapr_vfio_pci: Remove redundant spapr-pci-vfio-host-bridge

[Alexey Kardashevskiy]

sPAPRTCETable is handling 2 TCE tables already:

1) guest view of the TCE table - emulated devices use only this table;

2) hardware IOMMU table - VFIO PCI devices use it for actual work but
it does not replace 1) and it is not visible to the guest.
The initialization of this table is driven by vfio-pci device,
DMA map/unmap requests are handled via MemoryListener so there is very
little to do in spapr-pci-vfio-host-bridge.

This moves VFIO bits to the generic spapr-pci-host-bridge which allows
putting emulated and VFIO devices on the same PHB. It is still possible
to create multiple PHBs and avoid sharing PHB resouces for emulated and
VFIO devices.

If there is no VFIO-PCI device attaches, no special ioctls will be called.
If there are some VFIO-PCI devices attached, PHB may refuse to attach
another VFIO-PCI device if a VFIO container on the host kernel side
does not support
container sharing.

This changes spapr-pci-host-bridge to support properties of
spapr-pci-vfio-host-bridge. This makes spapr-pci-vfio-host-bridge type
equal to spapr-pci-host-bridge.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---

How do create aliases for types?
---
 hw/ppc/spapr_pci.c          | 78 ++++++++++++++-------------------------------
 hw/ppc/spapr_pci_vfio.c     | 35 ++++----------------
 include/hw/pci-host/spapr.h | 25 ++++++---------
 3 files changed, 41 insertions(+), 97 deletions(-)

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index 3d40f5b..d097cce 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -413,7 +413,6 @@ static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
                                     target_ulong rets)
 {
     sPAPRPHBState *sphb;
-    sPAPRPHBClass *spc;
     uint32_t addr, option;
     uint64_t buid;
     int ret;
@@ -427,16 +426,11 @@ static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
     option = rtas_ld(args, 3);
 
     sphb = spapr_pci_find_phb(spapr, buid);
-    if (!sphb) {
+    if (!sphb || !sphb->has_vfio) {
         goto param_error_exit;
     }
 
-    spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
-    if (!spc->eeh_set_option) {
-        goto param_error_exit;
-    }
-
-    ret = spc->eeh_set_option(sphb, addr, option);
+    ret = spapr_phb_vfio_eeh_set_option(sphb, addr, option);
     rtas_st(rets, 0, ret);
     return;
 
@@ -451,7 +445,6 @@ static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
                                            target_ulong rets)
 {
     sPAPRPHBState *sphb;
-    sPAPRPHBClass *spc;
     PCIDevice *pdev;
     uint32_t addr, option;
     uint64_t buid;
@@ -462,12 +455,7 @@ static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
     sphb = spapr_pci_find_phb(spapr, buid);
-    if (!sphb) {
-        goto param_error_exit;
-    }
-
-    spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
-    if (!spc->eeh_set_option) {
+    if (!sphb || !sphb->has_vfio) {
         goto param_error_exit;
     }
 
@@ -507,7 +495,6 @@ static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
                                             target_ulong rets)
 {
     sPAPRPHBState *sphb;
-    sPAPRPHBClass *spc;
     uint64_t buid;
     int state, ret;
 
@@ -517,16 +504,11 @@ static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
     sphb = spapr_pci_find_phb(spapr, buid);
-    if (!sphb) {
+    if (!sphb || !sphb->has_vfio) {
         goto param_error_exit;
     }
 
-    spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
-    if (!spc->eeh_get_state) {
-        goto param_error_exit;
-    }
-
-    ret = spc->eeh_get_state(sphb, &state);
+    ret = spapr_phb_vfio_eeh_get_state(sphb, &state);
     rtas_st(rets, 0, ret);
     if (ret != RTAS_OUT_SUCCESS) {
         return;
@@ -551,7 +533,6 @@ static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
                                     target_ulong rets)
 {
     sPAPRPHBState *sphb;
-    sPAPRPHBClass *spc;
     uint32_t option;
     uint64_t buid;
     int ret;
@@ -563,16 +544,11 @@ static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
     option = rtas_ld(args, 3);
     sphb = spapr_pci_find_phb(spapr, buid);
-    if (!sphb) {
+    if (!sphb || !sphb->has_vfio) {
         goto param_error_exit;
     }
 
-    spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
-    if (!spc->eeh_reset) {
-        goto param_error_exit;
-    }
-
-    ret = spc->eeh_reset(sphb, option);
+    ret = spapr_phb_vfio_eeh_reset(sphb, option);
     rtas_st(rets, 0, ret);
     return;
 
@@ -587,7 +563,6 @@ static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
                                   target_ulong rets)
 {
     sPAPRPHBState *sphb;
-    sPAPRPHBClass *spc;
     uint64_t buid;
     int ret;
 
@@ -597,16 +572,11 @@ static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
     sphb = spapr_pci_find_phb(spapr, buid);
-    if (!sphb) {
+    if (!sphb || !sphb->has_vfio) {
         goto param_error_exit;
     }
 
-    spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
-    if (!spc->eeh_configure) {
-        goto param_error_exit;
-    }
-
-    ret = spc->eeh_configure(sphb);
+    ret = spapr_phb_vfio_eeh_configure(sphb);
     rtas_st(rets, 0, ret);
     return;
 
@@ -622,7 +592,6 @@ static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
                                        target_ulong rets)
 {
     sPAPRPHBState *sphb;
-    sPAPRPHBClass *spc;
     int option;
     uint64_t buid;
 
@@ -632,12 +601,7 @@ static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
 
     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
     sphb = spapr_pci_find_phb(spapr, buid);
-    if (!sphb) {
-        goto param_error_exit;
-    }
-
-    spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
-    if (!spc->eeh_set_option) {
+    if (!sphb || !sphb->has_vfio) {
         goto param_error_exit;
     }
 
@@ -742,6 +706,11 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
     uint64_t msi_window_size = 4096;
     sPAPRTCETable *tcet;
 
+    if ((sphb->iommugroupid != -1) &&
+        object_dynamic_cast(OBJECT(sphb), TYPE_SPAPR_PCI_VFIO_HOST_BRIDGE)) {
+        error_report("Warning: iommugroupid shall not be used");
+    }
+
     if (sphb->index != (uint32_t)-1) {
         hwaddr windows_base;
 
@@ -894,9 +863,14 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
 
 static int spapr_phb_dma_capabilities_update(sPAPRPHBState *sphb)
 {
+    int ret;
+
     sphb->dma32_window_start = 0;
     sphb->dma32_window_size = SPAPR_PCI_DMA32_SIZE;
 
+    ret = spapr_phb_vfio_dma_capabilities_update(sphb);
+    sphb->has_vfio = (ret == 0);
+
     return 0;
 }
 
@@ -909,7 +883,7 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
 
     spapr_tce_table_enable(tcet, bus_offset, page_shift,
                            window_size >> page_shift,
-                           false);
+                           sphb->has_vfio);
 
     return 0;
 }
@@ -938,12 +912,11 @@ static int spapr_phb_disable_dma_windows(Object *child, void *opaque)
 int spapr_phb_dma_reset(sPAPRPHBState *sphb)
 {
     const uint32_t liobn = SPAPR_PCI_LIOBN(sphb->index, 0);
-    sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
 
-    spc->dma_capabilities_update(sphb); /* Refresh @has_vfio status */
+    spapr_phb_dma_capabilities_update(sphb); /* Refresh @has_vfio status */
     object_child_foreach(OBJECT(sphb), spapr_phb_disable_dma_windows, sphb);
-    spc->dma_init_window(sphb, liobn, SPAPR_TCE_PAGE_SHIFT,
-                         sphb->dma32_window_size);
+    spapr_phb_dma_init_window(sphb, liobn, SPAPR_TCE_PAGE_SHIFT,
+                              sphb->dma32_window_size);
 
     return 0;
 }
@@ -1089,7 +1062,6 @@ static void spapr_phb_class_init(ObjectClass *klass, void *data)
 {
     PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
     DeviceClass *dc = DEVICE_CLASS(klass);
-    sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_CLASS(klass);
 
     hc->root_bus_path = spapr_phb_root_bus_path;
     dc->realize = spapr_phb_realize;
@@ -1098,8 +1070,6 @@ static void spapr_phb_class_init(ObjectClass *klass, void *data)
     dc->vmsd = &vmstate_spapr_pci;
     set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
     dc->cannot_instantiate_with_device_add_yet = false;
-    spc->dma_capabilities_update = spapr_phb_dma_capabilities_update;
-    spc->dma_init_window = spapr_phb_dma_init_window;
 }
 
 static const TypeInfo spapr_phb_info = {
diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index f89e053..6f91b39 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -23,11 +23,11 @@
 #include "hw/vfio/vfio.h"
 
 static Property spapr_phb_vfio_properties[] = {
-    DEFINE_PROP_INT32("iommu", sPAPRPHBVFIOState, iommugroupid, -1),
+    DEFINE_PROP_INT32("iommu", sPAPRPHBState, iommugroupid, -1),
     DEFINE_PROP_END_OF_LIST(),
 };
 
-static int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb)
+int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb)
 {
     struct vfio_iommu_spapr_tce_info info = { .argsz = sizeof(info) };
     int ret;
@@ -44,21 +44,7 @@ static int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb)
     return ret;
 }
 
-static int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
-                                          uint32_t liobn, uint32_t page_shift,
-                                          uint64_t window_size)
-{
-    uint64_t bus_offset = sphb->dma32_window_start;
-    sPAPRTCETable *tcet = spapr_tce_find_by_liobn(liobn);
-
-    spapr_tce_table_enable(tcet, bus_offset, page_shift,
-                           window_size >> page_shift,
-                           true);
-
-    return 0;
-}
-
-static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
+int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
                                          unsigned int addr, int option)
 {
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
@@ -106,7 +92,7 @@ static int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
     return RTAS_OUT_SUCCESS;
 }
 
-static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
+int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
 {
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
@@ -122,7 +108,7 @@ static int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state)
     return RTAS_OUT_SUCCESS;
 }
 
-static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
+int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
 {
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
@@ -150,7 +136,7 @@ static int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option)
     return RTAS_OUT_SUCCESS;
 }
 
-static int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
+int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
 {
     struct vfio_eeh_pe_op op = { .argsz = sizeof(op) };
     int ret;
@@ -168,21 +154,14 @@ static int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb)
 static void spapr_phb_vfio_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
-    sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_CLASS(klass);
 
     dc->props = spapr_phb_vfio_properties;
-    spc->dma_capabilities_update = spapr_phb_vfio_dma_capabilities_update;
-    spc->dma_init_window = spapr_phb_vfio_dma_init_window;
-    spc->eeh_set_option = spapr_phb_vfio_eeh_set_option;
-    spc->eeh_get_state = spapr_phb_vfio_eeh_get_state;
-    spc->eeh_reset = spapr_phb_vfio_eeh_reset;
-    spc->eeh_configure = spapr_phb_vfio_eeh_configure;
 }
 
 static const TypeInfo spapr_phb_vfio_info = {
     .name          = TYPE_SPAPR_PCI_VFIO_HOST_BRIDGE,
     .parent        = TYPE_SPAPR_PCI_HOST_BRIDGE,
-    .instance_size = sizeof(sPAPRPHBVFIOState),
+    .instance_size = sizeof(sPAPRPHBState),
     .class_init    = spapr_phb_vfio_class_init,
     .class_size    = sizeof(sPAPRPHBClass),
 };
diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
index 7fda78e..484291c 100644
--- a/include/hw/pci-host/spapr.h
+++ b/include/hw/pci-host/spapr.h
@@ -47,15 +47,6 @@ typedef struct sPAPRPHBVFIOState sPAPRPHBVFIOState;
 
 struct sPAPRPHBClass {
     PCIHostBridgeClass parent_class;
-
-    int (*dma_capabilities_update)(sPAPRPHBState *sphb);
-    int (*dma_init_window)(sPAPRPHBState *sphb,
-                           uint32_t liobn, uint32_t page_shift,
-                           uint64_t window_size);
-    int (*eeh_set_option)(sPAPRPHBState *sphb, unsigned int addr, int option);
-    int (*eeh_get_state)(sPAPRPHBState *sphb, int *state);
-    int (*eeh_reset)(sPAPRPHBState *sphb, int option);
-    int (*eeh_configure)(sPAPRPHBState *sphb);
 };
 
 typedef struct spapr_pci_msi {
@@ -94,16 +85,12 @@ struct sPAPRPHBState {
 
     uint32_t dma32_window_start;
     uint32_t dma32_window_size;
+    bool has_vfio;
+    int32_t iommugroupid; /* obsolete */
 
     QLIST_ENTRY(sPAPRPHBState) list;
 };
 
-struct sPAPRPHBVFIOState {
-    sPAPRPHBState phb;
-
-    int32_t iommugroupid;
-};
-
 #define SPAPR_PCI_MAX_INDEX          255
 
 #define SPAPR_PCI_BASE_BUID          0x800000020000000ULL
@@ -144,4 +131,12 @@ int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
                                 sPAPRTCETable *tcet);
 int spapr_phb_dma_reset(sPAPRPHBState *sphb);
 
+int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb);
+int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
+                                  unsigned int addr, int option);
+int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state);
+int spapr_phb_vfio_eeh_reset(sPAPRPHBState *sphb, int option);
+int spapr_phb_vfio_eeh_configure(sPAPRPHBState *sphb);
+
+
 #endif /* __HW_SPAPR_PCI_H__ */
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 12/15] spapr_vfio_pci: Remove redundant spapr-pci-vfio-host-bridge

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 1640 bytes --]

On Sat, Apr 11, 2015 at 01:24:41AM +1000, Alexey Kardashevskiy wrote:
> sPAPRTCETable is handling 2 TCE tables already:
> 
> 1) guest view of the TCE table - emulated devices use only this table;
> 
> 2) hardware IOMMU table - VFIO PCI devices use it for actual work but
> it does not replace 1) and it is not visible to the guest.
> The initialization of this table is driven by vfio-pci device,
> DMA map/unmap requests are handled via MemoryListener so there is very
> little to do in spapr-pci-vfio-host-bridge.
> 
> This moves VFIO bits to the generic spapr-pci-host-bridge which allows
> putting emulated and VFIO devices on the same PHB. It is still possible
> to create multiple PHBs and avoid sharing PHB resouces for emulated and
> VFIO devices.
> 
> If there is no VFIO-PCI device attaches, no special ioctls will be called.
> If there are some VFIO-PCI devices attached, PHB may refuse to attach
> another VFIO-PCI device if a VFIO container on the host kernel side
> does not support
> container sharing.
> 
> This changes spapr-pci-host-bridge to support properties of
> spapr-pci-vfio-host-bridge. This makes spapr-pci-vfio-host-bridge type
> equal to spapr-pci-host-bridge.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

> ---
> 
> How do create aliases for types?

^^ I assume this is left over from some notes of your own?

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [Qemu-devel] [PATCH qemu v6 12/15] spapr_vfio_pci: Remove redundant spapr-pci-vfio-host-bridge

[Alexey Kardashevskiy]

On 04/22/2015 04:47 PM, David Gibson wrote:
> On Sat, Apr 11, 2015 at 01:24:41AM +1000, Alexey Kardashevskiy wrote:
>> sPAPRTCETable is handling 2 TCE tables already:
>>
>> 1) guest view of the TCE table - emulated devices use only this table;
>>
>> 2) hardware IOMMU table - VFIO PCI devices use it for actual work but
>> it does not replace 1) and it is not visible to the guest.
>> The initialization of this table is driven by vfio-pci device,
>> DMA map/unmap requests are handled via MemoryListener so there is very
>> little to do in spapr-pci-vfio-host-bridge.
>>
>> This moves VFIO bits to the generic spapr-pci-host-bridge which allows
>> putting emulated and VFIO devices on the same PHB. It is still possible
>> to create multiple PHBs and avoid sharing PHB resouces for emulated and
>> VFIO devices.
>>
>> If there is no VFIO-PCI device attaches, no special ioctls will be called.
>> If there are some VFIO-PCI devices attached, PHB may refuse to attach
>> another VFIO-PCI device if a VFIO container on the host kernel side
>> does not support
>> container sharing.
>>
>> This changes spapr-pci-host-bridge to support properties of
>> spapr-pci-vfio-host-bridge. This makes spapr-pci-vfio-host-bridge type
>> equal to spapr-pci-host-bridge.
>>
>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
>
>> ---
>>
>> How do create aliases for types?
>
> ^^ I assume this is left over from some notes of your own?

Nope, this is a question to the public. I ended up having 2 different QOM 
types (one is derived from another and has an extra property - "iommuid") 
while I could (probably) have just one (spapr-pci-host-bridge) and the 
other (spapr-pci-vfio-host-bridge) would be an alias as we have aliases for 
machines (when "pseries" is an alias for "pseries-2.3", etc). I am fine 
with the existing approach as it does not add new property to 
spapr-pci-host-bridge (otherwise it would have to), but still.



-- 
Alexey

[Qemu-devel] [PATCH qemu v6 13/15] spapr: Add pseries-2.4 machine

[Alexey Kardashevskiy]

The next patch implements dynamic DMA windows and disables them by default
for older pseries machines.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/ppc/spapr.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)

diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index 7febff7..b28209f 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -1852,12 +1852,15 @@ static const TypeInfo spapr_machine_2_2_info = {
 
 static void spapr_machine_2_3_class_init(ObjectClass *oc, void *data)
 {
+    static GlobalProperty compat_props[] = {
+        SPAPR_COMPAT_2_2,
+        { /* end of list */ }
+    };
     MachineClass *mc = MACHINE_CLASS(oc);
 
     mc->name = "pseries-2.3";
     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.3";
-    mc->alias = "pseries";
-    mc->is_default = 1;
+    mc->compat_props = compat_props;
 }
 
 static const TypeInfo spapr_machine_2_3_info = {
@@ -1866,12 +1869,29 @@ static const TypeInfo spapr_machine_2_3_info = {
     .class_init    = spapr_machine_2_3_class_init,
 };
 
+static void spapr_machine_2_4_class_init(ObjectClass *oc, void *data)
+{
+    MachineClass *mc = MACHINE_CLASS(oc);
+
+    mc->name = "pseries-2.4";
+    mc->desc = "pSeries Logical Partition (PAPR compliant) v2.4";
+    mc->alias = "pseries";
+    mc->is_default = 1;
+}
+
+static const TypeInfo spapr_machine_2_4_info = {
+    .name          = TYPE_SPAPR_MACHINE "2.4",
+    .parent        = TYPE_SPAPR_MACHINE,
+    .class_init    = spapr_machine_2_4_class_init,
+};
+
 static void spapr_machine_register_types(void)
 {
     type_register_static(&spapr_machine_info);
     type_register_static(&spapr_machine_2_1_info);
     type_register_static(&spapr_machine_2_2_info);
     type_register_static(&spapr_machine_2_3_info);
+    type_register_static(&spapr_machine_2_4_info);
 }
 
 type_init(spapr_machine_register_types)
-- 
2.0.0

Re: [Qemu-devel] [PATCH qemu v6 13/15] spapr: Add pseries-2.4 machine

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 494 bytes --]

On Sat, Apr 11, 2015 at 01:24:42AM +1000, Alexey Kardashevskiy wrote:
> The next patch implements dynamic DMA windows and disables them by default
> for older pseries machines.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

[Qemu-devel] [PATCH qemu v6 14/15] spapr_pci/spapr_pci_vfio: Support Dynamic DMA Windows (DDW)

[Alexey Kardashevskiy]

This adds support for Dynamic DMA Windows (DDW) option defined by
the SPAPR specification which allows to have additional DMA window(s)

This implements DDW for emulated and VFIO devices. As all TCE root regions
are mapped at 0 and 64bit long (and actual tables are child regions),
this replaces memory_region_add_subregion() with _overlap() to make
QEMU memory API happy.

This reserves RTAS token numbers for DDW calls.

This implements helpers to interact with VFIO kernel interface.

This changes the TCE table migration descriptor to support dynamic
tables as from now on, PHB will create as many stub TCE table objects
as PHB can possibly support but not all of them might be initialized at
the time of migration because DDW might or might not be requested by
the guest.

The "ddw" property is enabled by default on a PHB but for compatibility
the pseries-2.3 machine and older disable it.

This implements DDW for VFIO. The host kernel support is required.
This adds a "levels" property to PHB to control the number of levels
in the actual TCE table allocated by the host kernel, 0 is the default
value to tell QEMU to calculate the correct value. Current hardware
supports up to 5 levels.

The existing linux guests try creating one additional huge DMA window
with 64K or 16MB pages and map the entire guest RAM to. If succeeded,
the guest switches to dma_direct_ops and never calls TCE hypercalls
(H_PUT_TCE,...) again. This enables VFIO devices to use the entire RAM
and not waste time on map/unmap later.

This adds 4 RTAS handlers:
* ibm,query-pe-dma-window
* ibm,create-pe-dma-window
* ibm,remove-pe-dma-window
* ibm,reset-pe-dma-window
These are registered from type_init() callback.

These RTAS handlers are implemented in a separate file to avoid polluting
spapr_iommu.c with PCI.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
Changes:
v6:
* rework as there is no more special device for VFIO PHB

v5:
* total rework
* enabled for machines >2.3
* fixed migration
* merged rtas handlers here

v4:
* reset handler is back in generalized form

v3:
* removed reset
* windows_num is now 1 or bigger rather than 0-based value and it is only
changed in PHB code, not in RTAS
* added page mask check in create()
* added SPAPR_PCI_DDW_MAX_WINDOWS to track how many windows are already
created

v2:
* tested on hacked emulated E1000
* implemented DDW reset on the PHB reset
* spapr_pci_ddw_remove/spapr_pci_ddw_reset are public for reuse by VFIO
---
 hw/ppc/Makefile.objs        |   3 +
 hw/ppc/spapr.c              |  10 +-
 hw/ppc/spapr_iommu.c        |  35 +++++-
 hw/ppc/spapr_pci.c          |  67 ++++++++--
 hw/ppc/spapr_pci_vfio.c     |  80 ++++++++++++
 hw/ppc/spapr_rtas_ddw.c     | 300 ++++++++++++++++++++++++++++++++++++++++++++
 include/hw/pci-host/spapr.h |  21 ++++
 include/hw/ppc/spapr.h      |  17 ++-
 trace-events                |   4 +
 9 files changed, 522 insertions(+), 15 deletions(-)
 create mode 100644 hw/ppc/spapr_rtas_ddw.c

diff --git a/hw/ppc/Makefile.objs b/hw/ppc/Makefile.objs
index 437955d..c6b344f 100644
--- a/hw/ppc/Makefile.objs
+++ b/hw/ppc/Makefile.objs
@@ -7,6 +7,9 @@ obj-$(CONFIG_PSERIES) += spapr_pci.o spapr_rtc.o
 ifeq ($(CONFIG_PCI)$(CONFIG_PSERIES)$(CONFIG_LINUX), yyy)
 obj-y += spapr_pci_vfio.o
 endif
+ifeq ($(CONFIG_PCI)$(CONFIG_PSERIES), yy)
+obj-y += spapr_rtas_ddw.o
+endif
 # PowerPC 4xx boards
 obj-y += ppc405_boards.o ppc4xx_devs.o ppc405_uc.o ppc440_bamboo.o
 obj-y += ppc4xx_pci.o
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index b28209f..fd7fdb3 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -1801,7 +1801,15 @@ static const TypeInfo spapr_machine_info = {
     },
 };
 
+#define SPAPR_COMPAT_2_3 \
+        {\
+            .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
+            .property = "ddw",\
+            .value    = stringify(off),\
+        }
+
 #define SPAPR_COMPAT_2_2 \
+        SPAPR_COMPAT_2_3, \
         {\
             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
             .property = "mem_win_size",\
@@ -1853,7 +1861,7 @@ static const TypeInfo spapr_machine_2_2_info = {
 static void spapr_machine_2_3_class_init(ObjectClass *oc, void *data)
 {
     static GlobalProperty compat_props[] = {
-        SPAPR_COMPAT_2_2,
+        SPAPR_COMPAT_2_3,
         { /* end of list */ }
     };
     MachineClass *mc = MACHINE_CLASS(oc);
diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
index e655bf4..b985db3 100644
--- a/hw/ppc/spapr_iommu.c
+++ b/hw/ppc/spapr_iommu.c
@@ -90,6 +90,15 @@ static IOMMUTLBEntry spapr_tce_translate_iommu(MemoryRegion *iommu, hwaddr addr,
     return ret;
 }
 
+static void spapr_tce_table_pre_save(void *opaque)
+{
+    sPAPRTCETable *tcet = SPAPR_TCE_TABLE(opaque);
+
+    tcet->migtable = tcet->table;
+}
+
+static void spapr_tce_table_do_enable(sPAPRTCETable *tcet);
+
 static int spapr_tce_table_post_load(void *opaque, int version_id)
 {
     sPAPRTCETable *tcet = SPAPR_TCE_TABLE(opaque);
@@ -98,22 +107,42 @@ static int spapr_tce_table_post_load(void *opaque, int version_id)
         spapr_vio_set_bypass(tcet->vdev, tcet->bypass);
     }
 
+    if (!tcet->migtable) {
+        return 0;
+    }
+
+    if (tcet->enabled) {
+        if (!tcet->table) {
+            tcet->enabled = false;
+            spapr_tce_table_do_enable(tcet);
+        }
+        memcpy(tcet->table, tcet->migtable,
+               tcet->nb_table * sizeof(tcet->table[0]));
+        free(tcet->migtable);
+        tcet->migtable = NULL;
+    }
+
     return 0;
 }
 
 static const VMStateDescription vmstate_spapr_tce_table = {
     .name = "spapr_iommu",
-    .version_id = 2,
+    .version_id = 3,
     .minimum_version_id = 2,
+    .pre_save = spapr_tce_table_pre_save,
     .post_load = spapr_tce_table_post_load,
     .fields      = (VMStateField []) {
         /* Sanity check */
         VMSTATE_UINT32_EQUAL(liobn, sPAPRTCETable),
-        VMSTATE_UINT32_EQUAL(nb_table, sPAPRTCETable),
 
         /* IOMMU state */
+        VMSTATE_BOOL_V(enabled, sPAPRTCETable, 3),
+        VMSTATE_UINT64_V(bus_offset, sPAPRTCETable, 3),
+        VMSTATE_UINT32_V(page_shift, sPAPRTCETable, 3),
+        VMSTATE_UINT32(nb_table, sPAPRTCETable),
         VMSTATE_BOOL(bypass, sPAPRTCETable),
-        VMSTATE_VARRAY_UINT32(table, sPAPRTCETable, nb_table, 0, vmstate_info_uint64, uint64_t),
+        VMSTATE_VARRAY_UINT32_ALLOC(migtable, sPAPRTCETable, nb_table, 0,
+                                    vmstate_info_uint64, uint64_t),
 
         VMSTATE_END_OF_LIST()
     },
diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index d097cce..15ade90 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -849,15 +849,17 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         sphb->lsi_table[i].irq = irq;
     }
 
-    tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn);
-    if (!tcet) {
-            error_setg(errp, "failed to create TCE table");
+    for (i = 0; i < SPAPR_PCI_DMA_MAX_WINDOWS; ++i) {
+        tcet = spapr_tce_new_table(DEVICE(sphb),
+                                   SPAPR_PCI_LIOBN(sphb->index, i));
+        if (!tcet) {
+            error_setg(errp, "spapr_tce_new_table failed");
             return;
+        }
+        memory_region_add_subregion_overlap(&sphb->iommu_root, 0,
+                                            spapr_tce_get_iommu(tcet), 0);
     }
 
-    memory_region_add_subregion(&sphb->iommu_root, 0,
-                                spapr_tce_get_iommu(tcet));
-
     sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
 }
 
@@ -867,6 +869,9 @@ static int spapr_phb_dma_capabilities_update(sPAPRPHBState *sphb)
 
     sphb->dma32_window_start = 0;
     sphb->dma32_window_size = SPAPR_PCI_DMA32_SIZE;
+    sphb->windows_supported = SPAPR_PCI_DMA_MAX_WINDOWS;
+    sphb->page_size_mask = (1 << 12) | (1 << 16) | (1 << 24);
+    sphb->dma64_window_size = pow2ceil(ram_size);
 
     ret = spapr_phb_vfio_dma_capabilities_update(sphb);
     sphb->has_vfio = (ret == 0);
@@ -874,12 +879,30 @@ static int spapr_phb_dma_capabilities_update(sPAPRPHBState *sphb)
     return 0;
 }
 
-static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
-                                     uint32_t liobn, uint32_t page_shift,
-                                     uint64_t window_size)
+int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
+                              uint32_t liobn, uint32_t page_shift,
+                              uint64_t window_size)
 {
     uint64_t bus_offset = sphb->dma32_window_start;
     sPAPRTCETable *tcet = spapr_tce_find_by_liobn(liobn);
+    int ret;
+
+    if (SPAPR_PCI_DMA_WINDOW_NUM(liobn) && !sphb->ddw_enabled) {
+        return -1;
+    }
+
+    if (sphb->ddw_enabled) {
+        if (sphb->has_vfio) {
+            ret = spapr_phb_vfio_dma_init_window(sphb,
+                                                 page_shift, window_size,
+                                                 &bus_offset);
+            if (ret) {
+                return ret;
+            }
+        } else if (SPAPR_PCI_DMA_WINDOW_NUM(liobn)) {
+            bus_offset = SPAPR_PCI_DMA64_START;
+        }
+    }
 
     spapr_tce_table_enable(tcet, bus_offset, page_shift,
                            window_size >> page_shift,
@@ -891,9 +914,14 @@ static int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
 int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
                                 sPAPRTCETable *tcet)
 {
+    int ret;
+
+    if (sphb->has_vfio && sphb->ddw_enabled) {
+        ret = spapr_phb_vfio_dma_remove_window(sphb, tcet);
+    }
     spapr_tce_table_disable(tcet);
 
-    return 0;
+    return ret;
 }
 
 static int spapr_phb_disable_dma_windows(Object *child, void *opaque)
@@ -950,6 +978,8 @@ static Property spapr_phb_properties[] = {
     DEFINE_PROP_UINT64("io_win_addr", sPAPRPHBState, io_win_addr, -1),
     DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState, io_win_size,
                        SPAPR_PCI_IO_WIN_SIZE),
+    DEFINE_PROP_BOOL("ddw", sPAPRPHBState, ddw_enabled, true),
+    DEFINE_PROP_UINT8("levels", sPAPRPHBState, levels, 0),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -1140,6 +1170,15 @@ int spapr_populate_pci_dt(sPAPRPHBState *phb,
     uint32_t interrupt_map_mask[] = {
         cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
     uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7];
+    uint32_t ddw_applicable[] = {
+        cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW),
+        cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW),
+        cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW)
+    };
+    uint32_t ddw_extensions[] = {
+        cpu_to_be32(1),
+        cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW)
+    };
     sPAPRTCETable *tcet;
 
     /* Start populating the FDT */
@@ -1170,6 +1209,14 @@ int spapr_populate_pci_dt(sPAPRPHBState *phb,
     _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type", 0x1));
     _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pe-total-#msi", XICS_IRQS));
 
+    /* Dynamic DMA window */
+    if (phb->ddw_enabled) {
+        _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-applicable", &ddw_applicable,
+                         sizeof(ddw_applicable)));
+        _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-extensions",
+                         &ddw_extensions, sizeof(ddw_extensions)));
+    }
+
     /* Build the interrupt-map, this must matches what is done
      * in pci_spapr_map_irq
      */
diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index 6f91b39..7372d91 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -41,6 +41,86 @@ int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb)
     sphb->dma32_window_start = info.dma32_window_start;
     sphb->dma32_window_size = info.dma32_window_size;
 
+    if (sphb->ddw_enabled && (info.flags & VFIO_IOMMU_SPAPR_INFO_DDW)) {
+        sphb->windows_supported = info.ddw.max_dynamic_windows_supported;
+        sphb->page_size_mask = info.ddw.pgsizes;
+        sphb->dma64_window_size = pow2ceil(ram_size);
+        sphb->max_levels = info.ddw.levels;
+    } else {
+        /* If VFIO_IOMMU_INFO_DDW is not set, disable DDW */
+        sphb->ddw_enabled = false;
+    }
+
+    return ret;
+}
+
+static int spapr_phb_vfio_levels(uint32_t entries)
+{
+    unsigned pages = (entries * sizeof(uint64_t)) / getpagesize();
+    int levels;
+
+    if (pages <= 64) {
+        levels = 1;
+    } else if (pages <= 64*64) {
+        levels = 2;
+    } else if (pages <= 64*64*64) {
+        levels = 3;
+    } else {
+        levels = 4;
+    }
+
+    return levels;
+}
+
+int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
+                                   uint32_t page_shift,
+                                   uint64_t window_size,
+                                   uint64_t *bus_offset)
+{
+    int ret;
+    struct vfio_iommu_spapr_tce_create create = {
+        .argsz = sizeof(create),
+        .page_shift = page_shift,
+        .window_size = window_size,
+        .levels = sphb->levels,
+        .start_addr = 0,
+    };
+
+    /*
+     * Dynamic windows are supported, that means that there is no
+     * pre-created window and we have to create one.
+     */
+    if (!create.levels) {
+        create.levels = spapr_phb_vfio_levels(create.window_size >>
+                                              page_shift);
+    }
+
+    if (create.levels > sphb->max_levels) {
+        return -EINVAL;
+    }
+
+    ret = vfio_container_ioctl(&sphb->iommu_as,
+                               VFIO_IOMMU_SPAPR_TCE_CREATE, &create);
+    if (ret) {
+        return ret;
+    }
+    *bus_offset = create.start_addr;
+
+    return 0;
+}
+
+int spapr_phb_vfio_dma_remove_window(sPAPRPHBState *sphb,
+                                            sPAPRTCETable *tcet)
+{
+    struct vfio_iommu_spapr_tce_remove remove = {
+        .argsz = sizeof(remove),
+        .start_addr = tcet->bus_offset
+    };
+    int ret;
+
+    ret = vfio_container_ioctl(&sphb->iommu_as,
+                               VFIO_IOMMU_SPAPR_TCE_REMOVE, &remove);
+
     return ret;
 }
 
diff --git a/hw/ppc/spapr_rtas_ddw.c b/hw/ppc/spapr_rtas_ddw.c
new file mode 100644
index 0000000..7ab7572
--- /dev/null
+++ b/hw/ppc/spapr_rtas_ddw.c
@@ -0,0 +1,300 @@
+/*
+ * QEMU sPAPR Dynamic DMA windows support
+ *
+ * Copyright (c) 2014 Alexey Kardashevskiy, IBM Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License,
+ *  or (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "qemu/error-report.h"
+#include "hw/ppc/spapr.h"
+#include "hw/pci-host/spapr.h"
+#include "trace.h"
+
+static int spapr_phb_get_active_win_num_cb(Object *child, void *opaque)
+{
+    sPAPRTCETable *tcet;
+
+    tcet = (sPAPRTCETable *) object_dynamic_cast(child, TYPE_SPAPR_TCE_TABLE);
+    if (tcet && tcet->enabled) {
+        ++*(unsigned *)opaque;
+    }
+    return 0;
+}
+
+static unsigned spapr_phb_get_active_win_num(sPAPRPHBState *sphb)
+{
+    unsigned ret = 0;
+
+    object_child_foreach(OBJECT(sphb), spapr_phb_get_active_win_num_cb, &ret);
+
+    return ret;
+}
+
+static int spapr_phb_get_free_liobn_cb(Object *child, void *opaque)
+{
+    sPAPRTCETable *tcet;
+
+    tcet = (sPAPRTCETable *) object_dynamic_cast(child, TYPE_SPAPR_TCE_TABLE);
+    if (tcet && !tcet->enabled) {
+        *(uint32_t *)opaque = tcet->liobn;
+        return 1;
+    }
+    return 0;
+}
+
+static unsigned spapr_phb_get_free_liobn(sPAPRPHBState *sphb)
+{
+    uint32_t liobn = 0;
+
+    object_child_foreach(OBJECT(sphb), spapr_phb_get_free_liobn_cb, &liobn);
+
+    return liobn;
+}
+
+static uint32_t spapr_query_mask(struct ppc_one_seg_page_size *sps,
+                                 uint64_t page_mask)
+{
+    int i, j;
+    uint32_t mask = 0;
+    const struct { int shift; uint32_t mask; } masks[] = {
+        { 12, RTAS_DDW_PGSIZE_4K },
+        { 16, RTAS_DDW_PGSIZE_64K },
+        { 24, RTAS_DDW_PGSIZE_16M },
+        { 25, RTAS_DDW_PGSIZE_32M },
+        { 26, RTAS_DDW_PGSIZE_64M },
+        { 27, RTAS_DDW_PGSIZE_128M },
+        { 28, RTAS_DDW_PGSIZE_256M },
+        { 34, RTAS_DDW_PGSIZE_16G },
+    };
+
+    for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
+        for (j = 0; j < ARRAY_SIZE(masks); ++j) {
+            if ((sps[i].page_shift == masks[j].shift) &&
+                    (page_mask & (1ULL << masks[j].shift))) {
+                mask |= masks[j].mask;
+            }
+        }
+    }
+
+    return mask;
+}
+
+static void rtas_ibm_query_pe_dma_window(PowerPCCPU *cpu,
+                                         sPAPREnvironment *spapr,
+                                         uint32_t token, uint32_t nargs,
+                                         target_ulong args,
+                                         uint32_t nret, target_ulong rets)
+{
+    CPUPPCState *env = &cpu->env;
+    sPAPRPHBState *sphb;
+    uint64_t buid;
+    uint32_t avail, addr, pgmask = 0;
+    unsigned current;
+
+    if ((nargs != 3) || (nret != 5)) {
+        goto param_error_exit;
+    }
+
+    buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
+    addr = rtas_ld(args, 0);
+    sphb = spapr_pci_find_phb(spapr, buid);
+    if (!sphb || !sphb->ddw_enabled) {
+        goto param_error_exit;
+    }
+
+    current = spapr_phb_get_active_win_num(sphb);
+    avail = (sphb->windows_supported > current) ?
+            (sphb->windows_supported - current) : 0;
+
+    /* Work out supported page masks */
+    pgmask = spapr_query_mask(env->sps.sps, sphb->page_size_mask);
+
+    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
+    rtas_st(rets, 1, avail);
+
+    /*
+     * This is "Largest contiguous block of TCEs allocated specifically
+     * for (that is, are reserved for) this PE".
+     * Return the maximum number as all RAM was in 4K pages.
+     */
+    rtas_st(rets, 2, sphb->dma64_window_size >> SPAPR_TCE_PAGE_SHIFT);
+    rtas_st(rets, 3, pgmask);
+    rtas_st(rets, 4, 0); /* DMA migration mask, not supported */
+
+    trace_spapr_iommu_ddw_query(buid, addr, avail, sphb->dma64_window_size,
+                                pgmask);
+    return;
+
+param_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
+}
+
+static void rtas_ibm_create_pe_dma_window(PowerPCCPU *cpu,
+                                          sPAPREnvironment *spapr,
+                                          uint32_t token, uint32_t nargs,
+                                          target_ulong args,
+                                          uint32_t nret, target_ulong rets)
+{
+    sPAPRPHBState *sphb;
+    sPAPRTCETable *tcet = NULL;
+    uint32_t addr, page_shift, window_shift, liobn;
+    uint64_t buid;
+    long ret;
+
+    if ((nargs != 5) || (nret != 4)) {
+        goto param_error_exit;
+    }
+
+    buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
+    addr = rtas_ld(args, 0);
+    sphb = spapr_pci_find_phb(spapr, buid);
+    if (!sphb || !sphb->ddw_enabled) {
+        goto param_error_exit;
+    }
+
+    page_shift = rtas_ld(args, 3);
+    window_shift = rtas_ld(args, 4);
+    liobn = spapr_phb_get_free_liobn(sphb);
+
+    if (!liobn || !(sphb->page_size_mask & (1ULL << page_shift))) {
+        goto hw_error_exit;
+    }
+
+    ret = spapr_phb_dma_init_window(sphb, liobn, page_shift,
+                                    1ULL << window_shift);
+    tcet = spapr_tce_find_by_liobn(liobn);
+    trace_spapr_iommu_ddw_create(buid, addr, 1ULL << page_shift,
+                                 1ULL << window_shift,
+                                 tcet ? tcet->bus_offset : 0xbaadf00d,
+                                 liobn, ret);
+    if (ret || !tcet) {
+        goto hw_error_exit;
+    }
+
+    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
+    rtas_st(rets, 1, liobn);
+    rtas_st(rets, 2, tcet->bus_offset >> 32);
+    rtas_st(rets, 3, tcet->bus_offset & ((uint32_t) -1));
+
+    return;
+
+hw_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
+    return;
+
+param_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
+}
+
+static void rtas_ibm_remove_pe_dma_window(PowerPCCPU *cpu,
+                                          sPAPREnvironment *spapr,
+                                          uint32_t token, uint32_t nargs,
+                                          target_ulong args,
+                                          uint32_t nret, target_ulong rets)
+{
+    sPAPRPHBState *sphb;
+    sPAPRTCETable *tcet;
+    uint32_t liobn;
+    long ret;
+
+    if ((nargs != 1) || (nret != 1)) {
+        goto param_error_exit;
+    }
+
+    liobn = rtas_ld(args, 0);
+    tcet = spapr_tce_find_by_liobn(liobn);
+    if (!tcet) {
+        goto param_error_exit;
+    }
+
+    sphb = SPAPR_PCI_HOST_BRIDGE(OBJECT(tcet)->parent);
+    if (!sphb || !sphb->ddw_enabled) {
+        goto param_error_exit;
+    }
+
+    ret = spapr_phb_dma_remove_window(sphb, tcet);
+    trace_spapr_iommu_ddw_remove(liobn, ret);
+    if (ret) {
+        goto hw_error_exit;
+    }
+
+    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
+    return;
+
+hw_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
+    return;
+
+param_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
+}
+
+static void rtas_ibm_reset_pe_dma_window(PowerPCCPU *cpu,
+                                         sPAPREnvironment *spapr,
+                                         uint32_t token, uint32_t nargs,
+                                         target_ulong args,
+                                         uint32_t nret, target_ulong rets)
+{
+    sPAPRPHBState *sphb;
+    uint64_t buid;
+    uint32_t addr;
+    long ret;
+
+    if ((nargs != 3) || (nret != 1)) {
+        goto param_error_exit;
+    }
+
+    buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
+    addr = rtas_ld(args, 0);
+    sphb = spapr_pci_find_phb(spapr, buid);
+    if (!sphb || !sphb->ddw_enabled) {
+        goto param_error_exit;
+    }
+
+    ret = spapr_phb_dma_reset(sphb);
+    trace_spapr_iommu_ddw_reset(buid, addr, ret);
+    if (ret) {
+        goto hw_error_exit;
+    }
+
+    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
+
+    return;
+
+hw_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
+    return;
+
+param_error_exit:
+    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
+}
+
+static void spapr_rtas_ddw_init(void)
+{
+    spapr_rtas_register(RTAS_IBM_QUERY_PE_DMA_WINDOW,
+                        "ibm,query-pe-dma-window",
+                        rtas_ibm_query_pe_dma_window);
+    spapr_rtas_register(RTAS_IBM_CREATE_PE_DMA_WINDOW,
+                        "ibm,create-pe-dma-window",
+                        rtas_ibm_create_pe_dma_window);
+    spapr_rtas_register(RTAS_IBM_REMOVE_PE_DMA_WINDOW,
+                        "ibm,remove-pe-dma-window",
+                        rtas_ibm_remove_pe_dma_window);
+    spapr_rtas_register(RTAS_IBM_RESET_PE_DMA_WINDOW,
+                        "ibm,reset-pe-dma-window",
+                        rtas_ibm_reset_pe_dma_window);
+}
+
+type_init(spapr_rtas_ddw_init)
diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
index 484291c..1d2ea8d 100644
--- a/include/hw/pci-host/spapr.h
+++ b/include/hw/pci-host/spapr.h
@@ -87,6 +87,12 @@ struct sPAPRPHBState {
     uint32_t dma32_window_size;
     bool has_vfio;
     int32_t iommugroupid; /* obsolete */
+    bool ddw_enabled;
+    uint32_t windows_supported;
+    uint64_t page_size_mask;
+    uint64_t dma64_window_size;
+    uint8_t max_levels;
+    uint8_t levels;
 
     QLIST_ENTRY(sPAPRPHBState) list;
 };
@@ -109,6 +115,12 @@ struct sPAPRPHBState {
 
 #define SPAPR_PCI_DMA32_SIZE         0x40000000
 
+/* Default 64bit dynamic window offset */
+#define SPAPR_PCI_DMA64_START        0x8000000000000000ULL
+
+/* Maximum allowed number of DMA windows for emulated PHB */
+#define SPAPR_PCI_DMA_MAX_WINDOWS    2
+
 static inline qemu_irq spapr_phb_lsi_qirq(struct sPAPRPHBState *phb, int pin)
 {
     return xics_get_qirq(spapr->icp, phb->lsi_table[pin].irq);
@@ -127,11 +139,20 @@ void spapr_pci_rtas_init(void);
 sPAPRPHBState *spapr_pci_find_phb(sPAPREnvironment *spapr, uint64_t buid);
 PCIDevice *spapr_pci_find_dev(sPAPREnvironment *spapr, uint64_t buid,
                               uint32_t config_addr);
+int spapr_phb_dma_init_window(sPAPRPHBState *sphb,
+                              uint32_t liobn, uint32_t page_shift,
+                              uint64_t window_size);
 int spapr_phb_dma_remove_window(sPAPRPHBState *sphb,
                                 sPAPRTCETable *tcet);
 int spapr_phb_dma_reset(sPAPRPHBState *sphb);
 
 int spapr_phb_vfio_dma_capabilities_update(sPAPRPHBState *sphb);
+int spapr_phb_vfio_dma_init_window(sPAPRPHBState *sphb,
+                                   uint32_t page_shift,
+                                   uint64_t window_size,
+                                   uint64_t *bus_offset);
+int spapr_phb_vfio_dma_remove_window(sPAPRPHBState *sphb,
+                                     sPAPRTCETable *tcet);
 int spapr_phb_vfio_eeh_set_option(sPAPRPHBState *sphb,
                                   unsigned int addr, int option);
 int spapr_phb_vfio_eeh_get_state(sPAPRPHBState *sphb, int *state);
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index c8ac03f..873c661 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -381,6 +381,16 @@ int spapr_allocate_irq_block(int num, bool lsi, bool msi);
 #define RTAS_OUT_NOT_SUPPORTED      -3
 #define RTAS_OUT_NOT_AUTHORIZED     -9002
 
+/* DDW pagesize mask values from ibm,query-pe-dma-window */
+#define RTAS_DDW_PGSIZE_4K       0x01
+#define RTAS_DDW_PGSIZE_64K      0x02
+#define RTAS_DDW_PGSIZE_16M      0x04
+#define RTAS_DDW_PGSIZE_32M      0x08
+#define RTAS_DDW_PGSIZE_64M      0x10
+#define RTAS_DDW_PGSIZE_128M     0x20
+#define RTAS_DDW_PGSIZE_256M     0x40
+#define RTAS_DDW_PGSIZE_16G      0x80
+
 /* RTAS tokens */
 #define RTAS_TOKEN_BASE      0x2000
 
@@ -422,8 +432,12 @@ int spapr_allocate_irq_block(int num, bool lsi, bool msi);
 #define RTAS_IBM_SET_SLOT_RESET                 (RTAS_TOKEN_BASE + 0x23)
 #define RTAS_IBM_CONFIGURE_PE                   (RTAS_TOKEN_BASE + 0x24)
 #define RTAS_IBM_SLOT_ERROR_DETAIL              (RTAS_TOKEN_BASE + 0x25)
+#define RTAS_IBM_QUERY_PE_DMA_WINDOW            (RTAS_TOKEN_BASE + 0x26)
+#define RTAS_IBM_CREATE_PE_DMA_WINDOW           (RTAS_TOKEN_BASE + 0x27)
+#define RTAS_IBM_REMOVE_PE_DMA_WINDOW           (RTAS_TOKEN_BASE + 0x28)
+#define RTAS_IBM_RESET_PE_DMA_WINDOW            (RTAS_TOKEN_BASE + 0x29)
 
-#define RTAS_TOKEN_MAX                          (RTAS_TOKEN_BASE + 0x26)
+#define RTAS_TOKEN_MAX                          (RTAS_TOKEN_BASE + 0x2A)
 
 /* RTAS ibm,get-system-parameter token values */
 #define RTAS_SYSPARM_SPLPAR_CHARACTERISTICS      20
@@ -504,6 +518,7 @@ struct sPAPRTCETable {
     uint64_t bus_offset;
     uint32_t page_shift;
     uint64_t *table;
+    uint64_t *migtable;
     bool bypass;
     bool vfio_accel;
     int fd;
diff --git a/trace-events b/trace-events
index 2739140..fd8ea7a 100644
--- a/trace-events
+++ b/trace-events
@@ -1344,6 +1344,10 @@ spapr_iommu_pci_indirect(uint64_t liobn, uint64_t ioba, uint64_t tce, uint64_t i
 spapr_iommu_pci_stuff(uint64_t liobn, uint64_t ioba, uint64_t tce_value, uint64_t npages, uint64_t ret) "liobn=%"PRIx64" ioba=0x%"PRIx64" tcevalue=0x%"PRIx64" npages=%"PRId64" ret=%"PRId64
 spapr_iommu_xlate(uint64_t liobn, uint64_t ioba, uint64_t tce, unsigned perm, unsigned pgsize) "liobn=%"PRIx64" 0x%"PRIx64" -> 0x%"PRIx64" perm=%u mask=%x"
 spapr_iommu_new_table(uint64_t liobn, void *tcet, void *table, int fd) "liobn=%"PRIx64" tcet=%p table=%p fd=%d"
+spapr_iommu_ddw_query(uint64_t buid, uint32_t cfgaddr, unsigned wa, uint64_t win_size, uint32_t pgmask) "buid=%"PRIx64" addr=%"PRIx32", %u windows available, max window size=%"PRIx64", mask=%"PRIx32
+spapr_iommu_ddw_create(uint64_t buid, uint32_t cfgaddr, unsigned long long pg_size, unsigned long long req_size, uint64_t start, uint32_t liobn, long ret) "buid=%"PRIx64" addr=%"PRIx32", page size=0x%llx, requested=0x%llx, start addr=%"PRIx64", liobn=%"PRIx32", ret = %ld"
+spapr_iommu_ddw_remove(uint32_t liobn, long ret) "liobn=%"PRIx32", ret = %ld"
+spapr_iommu_ddw_reset(uint64_t buid, uint32_t cfgaddr, long ret) "buid=%"PRIx64" addr=%"PRIx32", ret = %ld"
 
 # hw/ppc/ppc.c
 ppc_tb_adjust(uint64_t offs1, uint64_t offs2, int64_t diff, int64_t seconds) "adjusted from 0x%"PRIx64" to 0x%"PRIx64", diff %"PRId64" (%"PRId64"s)"
-- 
2.0.0

[Qemu-devel] [PATCH qemu v6 15/15] vfio: Enable DDW ioctls to VFIO IOMMU driver

[Alexey Kardashevskiy]

This enables DDW RTAS-related ioctls in VFIO.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---
 hw/vfio/common.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index 9e3e0b0..f915127 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -830,6 +830,8 @@ int vfio_container_ioctl(AddressSpace *as,
     case VFIO_CHECK_EXTENSION:
     case VFIO_IOMMU_SPAPR_TCE_GET_INFO:
     case VFIO_EEH_PE_OP:
+    case VFIO_IOMMU_SPAPR_TCE_CREATE:
+    case VFIO_IOMMU_SPAPR_TCE_REMOVE:
         break;
     default:
         /* Return an error on unknown requests */
-- 
2.0.0