qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [RFC PATCH v2 00/32] CXL 2.0 Support
@ 2021-01-05 16:52 Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 01/32] Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy Ben Widawsky
                   ` (32 more replies)
  0 siblings, 33 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Fixes since v1 [1]:
 * Defer introducing some commands/registers not yet used (Ben)
 * Add stubbed device_reg_init_common() (Ben)
 * Improve assertions in DVSEC creation (Jonathan)
 * Use 'n' for HDM register offsets (Jonathan)
 * Correct revision ID for extensions (Jonathan)
 * Minor cleanups and clarifications (Jonathan)
 * Remove error codes not yet used (Jonathan)
 * Fix interrupt enable bit width (Jonathan)
 * Add comment for weird register size (Jonathan)
 * Break out register alignment checks (Jonathan)
 * Use the reg alignment helper (Jonathan)
 * Rename error codes to match spec
 * Fix cap count mid series (Jonathan)

New since v1 [1]:
 * Entirely reworked framework for firmware handling
 * Implemented more device commands
 * CEL support

(There are some new patches that I just named 'v2' for simplicity's sake)

Introduce emulation of Compute Express Link 2.0
(https://www.computeexpresslink.org/).

The emulation has been critical to get the Linux enabling started [2], it would
be an ideal place to land regression tests for different topology handling, and
there may be applications for this emulation as a way for a guest to manipulate
its address space relative to different performance memories.

Three of the five CXL component types are emulated with some level of functionality:
host bridge, root port, and memory device. Upstream ports and downstream ports
aren't implemented (the two components needed to make up a switch).

CXL 2.0 is built on top of PCIe (see spec for details). As a result, much of the
implementation utilizes existing PCI paradigms. To implement the host bridge,
I've chosen to use PXB (PCI Expander Bridge). It seemed to be the most natural
fit even though it doesn't directly map to how hardware will work. For
persistent capacity of the memory device, I utilized the memory subsystem
(hw/mem).

We have 3 reasons why this work is valuable:
1. OS driver development and testing
2. OS driver regression testing
3. Possible guest support for HDMs

As mentioned above there are three benefits to carrying this enabling in
upstream QEMU:

1. Linux driver feature development benefits from emulation both due to
a lack of initial hardware availability, but also, as is seen with
NVDIMM/PMEM emulation, there is value in being able to share
topologies with system-software developers even after hardware is
available.

2. The Linux kernel's unit test suite for NVDIMM/PMEM ended up injecting fake
resources via custom modules (nfit_test). In retrospect a QEMU emulation of
nfit_test capabilities would have made the test environment more portable, and
allowed for easier community contributions of example configurations.

3. This is still being fleshed out, but in short it provides a standardized
mechanism for the guest to provide feedback to the host about size and placement
needs of the memory. After the host gives the guest a physical window mapping to
the CXL device, the emulated HDM decoders allow the guest a way to tell the host
how much it wants and where. There are likely simpler ways to do this, but
they'd require inventing a new interface and you'd need to have diverging driver
code in the guest programming of the HDM decoder vs. the host. Since we've
already done this work, why not use it?

There is quite a long list of work to do for full spec compliance, but I don't
believe that any of it precludes merging. Off the top of my head:
- Main host bridge support (WIP)
- Interleaving
- Better Tests
- Huge swaths of firmware functionality
- Hot plug support
- Emulating volatile capacity

The flow of the patches in general is to define all the data structures and
registers associated with the various components in a top down manner. Host
bridge, component, ports, devices. Then, the actual implementation is done in
the same order.

The summary is:
1-8: Put infrastructure in place for emulation of the components.
9-12: Implement device mailboxes
13-15: Create the concept of a CXL bus and plumb into PXB
16-22: Implement host bridges
23: Implement a root port
24: Implement a memory device
25: Implement HDM decoders
26-30: ACPI bits
31: Start working on enabling the main host bridge
32: Basic test case

[1]: https://lore.kernel.org/qemu-devel/20201111054724.794888-1-ben.widawsky@intel.com/
[2]: https://lore.kernel.org/linux-cxl/20201209002418.1976362-1-ben.widawsky@intel.com/

Ben Widawsky (31):
  hw/pci/cxl: Add a CXL component type (interface)
  hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)
  hw/cxl/device: Introduce a CXL device (8.2.8)
  hw/cxl/device: Implement the CAP array (8.2.8.1-2)
  hw/cxl/device: Add device status (8.2.8.3)
  hw/cxl/device: Implement basic mailbox (8.2.8.4)
  hw/cxl/device: Add memory devices (8.2.8.5)
  hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1)
  hw/cxl/device: Placeholder for firmware commands (8.2.9.2)
  hw/cxl/device: Timestamp implementation (8.2.9.3)
  hw/cxl/device: Add log commands (8.2.9.4) + CEL
  hw/pxb: Use a type for realizing expanders
  hw/pci/cxl: Create a CXL bus type
  hw/pxb: Allow creation of a CXL PXB (host bridge)
  qtest: allow DSDT acpi table changes
  acpi/pci: Consolidate host bridge setup
  tests/acpi: remove stale allowed tables
  hw/pci: Plumb _UID through host bridges
  hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)
  acpi/pxb/cxl: Reserve host bridge MMIO
  hw/pxb/cxl: Add "windows" for host bridges
  hw/cxl/rp: Add a root port
  hw/cxl/device: Add a memory device (8.2.8.5)
  hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12)
  acpi/cxl: Add _OSC implementation (9.14.2)
  tests/acpi: allow CEDT table addition
  acpi/cxl: Create the CEDT (9.14.1)
  Temp: acpi/cxl: Add ACPI0017 (CEDT awareness)
  tests/acpi: Add new CEDT files
  WIP: i386/cxl: Initialize a host bridge
  qtest/cxl: Add very basic sanity tests

Jonathan Cameron (1):
  Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy.

 MAINTAINERS                               |   6 +
 hw/Kconfig                                |   1 +
 hw/acpi/Kconfig                           |   5 +
 hw/acpi/cxl.c                             | 173 ++++++++++
 hw/acpi/meson.build                       |   1 +
 hw/arm/virt.c                             |   1 +
 hw/core/machine.c                         |  26 ++
 hw/core/numa.c                            |   3 +
 hw/cxl/Kconfig                            |   3 +
 hw/cxl/cxl-component-utils.c              | 193 +++++++++++
 hw/cxl/cxl-device-utils.c                 | 278 ++++++++++++++++
 hw/cxl/cxl-mailbox-utils.c                | 386 ++++++++++++++++++++++
 hw/cxl/meson.build                        |   5 +
 hw/i386/acpi-build.c                      |  87 ++++-
 hw/i386/microvm.c                         |   1 +
 hw/i386/pc.c                              |   2 +
 hw/mem/Kconfig                            |   5 +
 hw/mem/cxl_type3.c                        | 334 +++++++++++++++++++
 hw/mem/meson.build                        |   1 +
 hw/meson.build                            |   1 +
 hw/pci-bridge/Kconfig                     |   5 +
 hw/pci-bridge/cxl_root_port.c             | 231 +++++++++++++
 hw/pci-bridge/meson.build                 |   1 +
 hw/pci-bridge/pci_expander_bridge.c       | 209 +++++++++++-
 hw/pci-bridge/pcie_root_port.c            |   6 +-
 hw/pci/pci.c                              |  32 +-
 hw/pci/pcie.c                             |  30 ++
 hw/ppc/spapr.c                            |   2 +
 include/hw/acpi/cxl.h                     |  27 ++
 include/hw/boards.h                       |   2 +
 include/hw/cxl/cxl.h                      |  30 ++
 include/hw/cxl/cxl_component.h            | 187 +++++++++++
 include/hw/cxl/cxl_device.h               | 210 ++++++++++++
 include/hw/cxl/cxl_pci.h                  | 160 +++++++++
 include/hw/pci/pci.h                      |  15 +
 include/hw/pci/pci_bridge.h               |  25 ++
 include/hw/pci/pci_bus.h                  |   8 +
 include/hw/pci/pci_ids.h                  |   1 +
 include/standard-headers/linux/pci_regs.h |   1 +
 monitor/hmp-cmds.c                        |  15 +
 qapi/machine.json                         |   1 +
 tests/data/acpi/pc/CEDT                   | Bin 0 -> 36 bytes
 tests/data/acpi/pc/DSDT                   | Bin 5065 -> 5065 bytes
 tests/data/acpi/pc/DSDT.acpihmat          | Bin 6390 -> 6390 bytes
 tests/data/acpi/pc/DSDT.bridge            | Bin 6924 -> 6924 bytes
 tests/data/acpi/pc/DSDT.cphp              | Bin 5529 -> 5529 bytes
 tests/data/acpi/pc/DSDT.dimmpxm           | Bin 6719 -> 6719 bytes
 tests/data/acpi/pc/DSDT.hpbridge          | Bin 5026 -> 5026 bytes
 tests/data/acpi/pc/DSDT.hpbrroot          | Bin 3084 -> 3084 bytes
 tests/data/acpi/pc/DSDT.ipmikcs           | Bin 5137 -> 5137 bytes
 tests/data/acpi/pc/DSDT.memhp             | Bin 6424 -> 6424 bytes
 tests/data/acpi/pc/DSDT.numamem           | Bin 5071 -> 5071 bytes
 tests/data/acpi/pc/DSDT.roothp            | Bin 5261 -> 5261 bytes
 tests/data/acpi/q35/CEDT                  | Bin 0 -> 36 bytes
 tests/data/acpi/q35/DSDT                  | Bin 7801 -> 7801 bytes
 tests/data/acpi/q35/DSDT.acpihmat         | Bin 9126 -> 9126 bytes
 tests/data/acpi/q35/DSDT.bridge           | Bin 7819 -> 7819 bytes
 tests/data/acpi/q35/DSDT.cphp             | Bin 8265 -> 8265 bytes
 tests/data/acpi/q35/DSDT.dimmpxm          | Bin 9455 -> 9455 bytes
 tests/data/acpi/q35/DSDT.ipmibt           | Bin 7876 -> 7876 bytes
 tests/data/acpi/q35/DSDT.memhp            | Bin 9160 -> 9160 bytes
 tests/data/acpi/q35/DSDT.mmio64           | Bin 8932 -> 8932 bytes
 tests/data/acpi/q35/DSDT.numamem          | Bin 7807 -> 7807 bytes
 tests/qtest/cxl-test.c                    |  93 ++++++
 tests/qtest/meson.build                   |   4 +
 65 files changed, 2777 insertions(+), 30 deletions(-)
 create mode 100644 hw/acpi/cxl.c
 create mode 100644 hw/cxl/Kconfig
 create mode 100644 hw/cxl/cxl-component-utils.c
 create mode 100644 hw/cxl/cxl-device-utils.c
 create mode 100644 hw/cxl/cxl-mailbox-utils.c
 create mode 100644 hw/cxl/meson.build
 create mode 100644 hw/mem/cxl_type3.c
 create mode 100644 hw/pci-bridge/cxl_root_port.c
 create mode 100644 include/hw/acpi/cxl.h
 create mode 100644 include/hw/cxl/cxl.h
 create mode 100644 include/hw/cxl/cxl_component.h
 create mode 100644 include/hw/cxl/cxl_device.h
 create mode 100644 include/hw/cxl/cxl_pci.h
 create mode 100644 tests/data/acpi/pc/CEDT
 create mode 100644 tests/data/acpi/q35/CEDT
 create mode 100644 tests/qtest/cxl-test.c

-- 
2.30.0



^ permalink raw reply	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 01/32] Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy.
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 02/32] hw/pci/cxl: Add a CXL component type (interface) Ben Widawsky
                   ` (31 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

From: Jonathan Cameron <Jonathan.Cameron@huawei.com>

This hasn't yet been added to the linux kernel tree, so for purposes
of this RFC just add it locally.

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 include/standard-headers/linux/pci_regs.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/include/standard-headers/linux/pci_regs.h b/include/standard-headers/linux/pci_regs.h
index a95d55f9f2..5d0b79b9da 100644
--- a/include/standard-headers/linux/pci_regs.h
+++ b/include/standard-headers/linux/pci_regs.h
@@ -723,6 +723,7 @@
 #define PCI_EXT_CAP_ID_DPC	0x1D	/* Downstream Port Containment */
 #define PCI_EXT_CAP_ID_L1SS	0x1E	/* L1 PM Substates */
 #define PCI_EXT_CAP_ID_PTM	0x1F	/* Precision Time Measurement */
+#define PCI_EXT_CAP_ID_DVSEC	0x23    /* Designated Vendor-Specific */
 #define PCI_EXT_CAP_ID_DLF	0x25	/* Data Link Feature */
 #define PCI_EXT_CAP_ID_PL_16GT	0x26	/* Physical Layer 16.0 GT/s */
 #define PCI_EXT_CAP_ID_MAX	PCI_EXT_CAP_ID_PL_16GT
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 02/32] hw/pci/cxl: Add a CXL component type (interface)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 01/32] Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 03/32] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5) Ben Widawsky
                   ` (30 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

A CXL component is a hardware entity that implements CXL component
registers from the CXL 2.0 spec (8.2.3). Currently these represent 3
general types.
1. Host Bridge
2. Ports (root, upstream, downstream)
3. Devices (memory, other)

A CXL component can be conceptually thought of as a PCIe device with
extra functionality when enumerated and enabled. For this reason, CXL
does here, and will continue to add on to existing PCI code paths.

Host bridges will typically need to be handled specially and so they can
implement this newly introduced interface or not. All other components
should implement this interface. Implementing this interface allows the
core pci code to treat these devices as special where appropriate.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/pci/pci.c         | 10 ++++++++++
 include/hw/pci/pci.h |  8 ++++++++
 2 files changed, 18 insertions(+)

diff --git a/hw/pci/pci.c b/hw/pci/pci.c
index a6b0c5602e..50fc61c81f 100644
--- a/hw/pci/pci.c
+++ b/hw/pci/pci.c
@@ -194,6 +194,11 @@ static const TypeInfo pci_bus_info = {
     .class_init = pci_bus_class_init,
 };
 
+static const TypeInfo cxl_interface_info = {
+    .name          = INTERFACE_CXL_DEVICE,
+    .parent        = TYPE_INTERFACE,
+};
+
 static const TypeInfo pcie_interface_info = {
     .name          = INTERFACE_PCIE_DEVICE,
     .parent        = TYPE_INTERFACE,
@@ -2115,6 +2120,10 @@ static void pci_qdev_realize(DeviceState *qdev, Error **errp)
         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
     }
 
+    if (object_class_dynamic_cast(klass, INTERFACE_CXL_DEVICE)) {
+        pci_dev->cap_present |= QEMU_PCIE_CAP_CXL;
+    }
+
     pci_dev = do_pci_register_device(pci_dev,
                                      object_get_typename(OBJECT(qdev)),
                                      pci_dev->devfn, errp);
@@ -2841,6 +2850,7 @@ static void pci_register_types(void)
     type_register_static(&pci_bus_info);
     type_register_static(&pcie_bus_info);
     type_register_static(&conventional_pci_interface_info);
+    type_register_static(&cxl_interface_info);
     type_register_static(&pcie_interface_info);
     type_register_static(&pci_device_type_info);
 }
diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h
index 259f9c992d..f6b1eff446 100644
--- a/include/hw/pci/pci.h
+++ b/include/hw/pci/pci.h
@@ -194,6 +194,8 @@ enum {
     QEMU_PCIE_LNKSTA_DLLLA = (1 << QEMU_PCIE_LNKSTA_DLLLA_BITNR),
 #define QEMU_PCIE_EXTCAP_INIT_BITNR 9
     QEMU_PCIE_EXTCAP_INIT = (1 << QEMU_PCIE_EXTCAP_INIT_BITNR),
+#define QEMU_PCIE_CXL_BITNR 10
+    QEMU_PCIE_CAP_CXL = (1 << QEMU_PCIE_CXL_BITNR),
 };
 
 #define TYPE_PCI_DEVICE "pci-device"
@@ -201,6 +203,12 @@ typedef struct PCIDeviceClass PCIDeviceClass;
 DECLARE_OBJ_CHECKERS(PCIDevice, PCIDeviceClass,
                      PCI_DEVICE, TYPE_PCI_DEVICE)
 
+/*
+ * Implemented by devices that can be plugged on CXL buses. In the spec, this is
+ * actually a "CXL Component, but we name it device to match the PCI naming.
+ */
+#define INTERFACE_CXL_DEVICE "cxl-device"
+
 /* Implemented by devices that can be plugged on PCI Express buses */
 #define INTERFACE_PCIE_DEVICE "pci-express-device"
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 03/32] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 01/32] Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 02/32] hw/pci/cxl: Add a CXL component type (interface) Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 04/32] hw/cxl/device: Introduce a CXL device (8.2.8) Ben Widawsky
                   ` (29 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

A CXL 2.0 component is any entity in the CXL topology. All components
have a analogous function in PCIe. Except for the CXL host bridge, all
have a PCIe config space that is accessible via the common PCIe
mechanisms. CXL components are enumerated via DVSEC fields in the
extended PCIe header space. CXL components will minimally implement some
subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL
2.0 specification. Two headers and a utility library are introduced to
support the minimum functionality needed to enumerate components.

The cxl_pci header manages bits associated with PCI, specifically the
DVSEC and related fields. The cxl_component.h variant has data
structures and APIs that are useful for drivers implementing any of the
CXL 2.0 components. The library takes care of making use of the DVSEC
bits and the CXL.[mem|cache] regisetrs.

None of the mechanisms required to enumerate a CXL capable hostbridge
are introduced at this point.

Note that the CXL.mem and CXL.cache registers used are always 4B wide.
It's possible in the future that this constraint will not hold.

v2: Improve assertions in DVSEC creation (Jonathan)
    Use 'n' for HDM register offsets (Jonathan)
    Correct revision ID for extensions (Jonathan)
    Minor cleanups and clarifications (Jonathan)

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 MAINTAINERS                    |   6 +
 hw/Kconfig                     |   1 +
 hw/cxl/Kconfig                 |   3 +
 hw/cxl/cxl-component-utils.c   | 193 +++++++++++++++++++++++++++++++++
 hw/cxl/meson.build             |   3 +
 hw/meson.build                 |   1 +
 include/hw/cxl/cxl.h           |  17 +++
 include/hw/cxl/cxl_component.h | 187 ++++++++++++++++++++++++++++++++
 include/hw/cxl/cxl_pci.h       | 138 +++++++++++++++++++++++
 9 files changed, 549 insertions(+)
 create mode 100644 hw/cxl/Kconfig
 create mode 100644 hw/cxl/cxl-component-utils.c
 create mode 100644 hw/cxl/meson.build
 create mode 100644 include/hw/cxl/cxl.h
 create mode 100644 include/hw/cxl/cxl_component.h
 create mode 100644 include/hw/cxl/cxl_pci.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 42fedf91e7..1f5fc033b4 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -2225,6 +2225,12 @@ F: qapi/block*.json
 F: qapi/transaction.json
 T: git https://repo.or.cz/qemu/armbru.git block-next
 
+Compute Express Link
+M: Ben Widawsky <ben.widawsky@intel.com>
+S: Supported
+F: hw/cxl/
+F: include/hw/cxl/
+
 Dirty Bitmaps
 M: Eric Blake <eblake@redhat.com>
 M: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
diff --git a/hw/Kconfig b/hw/Kconfig
index 5ad3c6b5a4..c03650c5ed 100644
--- a/hw/Kconfig
+++ b/hw/Kconfig
@@ -6,6 +6,7 @@ source audio/Kconfig
 source block/Kconfig
 source char/Kconfig
 source core/Kconfig
+source cxl/Kconfig
 source display/Kconfig
 source dma/Kconfig
 source gpio/Kconfig
diff --git a/hw/cxl/Kconfig b/hw/cxl/Kconfig
new file mode 100644
index 0000000000..8e67519b16
--- /dev/null
+++ b/hw/cxl/Kconfig
@@ -0,0 +1,3 @@
+config CXL
+    bool
+    default y if PCI_EXPRESS
diff --git a/hw/cxl/cxl-component-utils.c b/hw/cxl/cxl-component-utils.c
new file mode 100644
index 0000000000..a95c16505c
--- /dev/null
+++ b/hw/cxl/cxl-component-utils.c
@@ -0,0 +1,193 @@
+/*
+ * CXL Utility library for components
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
+
+static uint64_t cxl_cache_mem_read_reg(void *opaque, hwaddr offset,
+                                       unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    uint32_t *cache_mem = cregs->cache_mem_registers;
+
+    if (size != 4) {
+        qemu_log_mask(LOG_UNIMP, "%uB component register read (RAZ)\n", size);
+        return 0;
+    }
+
+    if (cregs->special_ops && cregs->special_ops->read) {
+        return cregs->special_ops->read(cxl_cstate, offset, size);
+    } else {
+        return cache_mem[offset >> 2];
+    }
+}
+
+static void cxl_cache_mem_write_reg(void *opaque, hwaddr offset, uint64_t value,
+                                    unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size != 4) {
+        qemu_log_mask(LOG_UNIMP, "%uB component register write (WI)\n", size);
+        return;
+    }
+
+    if (cregs->special_ops && cregs->special_ops->write) {
+        cregs->special_ops->write(cxl_cstate, offset, value, size);
+    }
+}
+
+static const MemoryRegionOps cache_mem_ops = {
+    .read = cxl_cache_mem_read_reg,
+    .write = cxl_cache_mem_write_reg,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 4,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 4,
+    },
+};
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type)
+{
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    memory_region_init(&cregs->component_registers, obj, type, 0x10000);
+    memory_region_init_io(&cregs->io, obj, NULL, cregs, ".io", 0x1000);
+    memory_region_init_io(&cregs->cache_mem, obj, &cache_mem_ops, cregs,
+                          ".cache_mem", 0x1000);
+
+    memory_region_add_subregion(&cregs->component_registers, 0, &cregs->io);
+    memory_region_add_subregion(&cregs->component_registers, 0x1000,
+                                &cregs->cache_mem);
+}
+
+static void ras_init_common(uint32_t *reg_state)
+{
+    reg_state[R_CXL_RAS_UNC_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_UNC_ERR_MASK] = 0x1efff;
+    reg_state[R_CXL_RAS_UNC_ERR_SEVERITY] = 0x1efff;
+    reg_state[R_CXL_RAS_COR_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_COR_ERR_MASK] = 0x3f;
+    reg_state[R_CXL_RAS_ERR_CAP_CTRL] = 0; /* CXL switches and devices must set */
+}
+
+static void hdm_init_common(uint32_t *reg_state)
+{
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_GLOBAL_CONTROL, HDM_DECODER_ENABLE, 0);
+}
+
+void cxl_component_register_init_common(uint32_t *reg_state, enum reg_type type)
+{
+    int caps = 0;
+    switch (type) {
+    case CXL2_DOWNSTREAM_PORT:
+    case CXL2_DEVICE:
+        /* CAP, RAS, Link */
+        caps = 2;
+        break;
+    case CXL2_UPSTREAM_PORT:
+    case CXL2_TYPE3_DEVICE:
+    case CXL2_LOGICAL_DEVICE:
+        /* + HDM */
+        caps = 3;
+        break;
+    case CXL2_ROOT_PORT:
+        /* + Extended Security, + Snoop */
+        caps = 5;
+        break;
+    default:
+        abort();
+    }
+
+    memset(reg_state, 0, 0x1000);
+
+    /* CXL Capability Header Register */
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ID, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ARRAY_SIZE, caps);
+
+
+#define init_cap_reg(reg, id, version)                                        \
+    do {                                                                      \
+        int which = R_CXL_##reg##_CAPABILITY_HEADER;                          \
+        reg_state[which] = FIELD_DP32(reg_state[which],                       \
+                                      CXL_##reg##_CAPABILITY_HEADER, ID, id); \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER,       \
+                       VERSION, version);                                     \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER, PTR,  \
+                       CXL_##reg##_REGISTERS_OFFSET);                         \
+    } while (0)
+
+    init_cap_reg(RAS, 2, 1);
+    ras_init_common(reg_state);
+
+    init_cap_reg(LINK, 4, 2);
+
+    if (caps < 4) {
+        return;
+    }
+
+    init_cap_reg(HDM, 5, 1);
+    hdm_init_common(reg_state);
+
+    if (caps < 6) {
+        return;
+    }
+
+    init_cap_reg(EXTSEC, 6, 1);
+    init_cap_reg(SNOOP, 8, 1);
+
+#undef init_cap_reg
+}
+
+/*
+ * Helper to creates a DVSEC header for a CXL entity. The caller is responsible
+ * for tracking the valid offset.
+ *
+ * This function will build the DVSEC header on behalf of the caller and then
+ * copy in the remaining data for the vendor specific bits.
+ */
+void cxl_component_create_dvsec(CXLComponentState *cxl, uint16_t length,
+                                uint16_t type, uint8_t rev, uint8_t *body)
+{
+    PCIDevice *pdev = cxl->pdev;
+    uint16_t offset = cxl->dvsec_offset;
+
+    assert(offset >= PCI_CFG_SPACE_SIZE &&
+           ((offset + length) < PCI_CFG_SPACE_EXP_SIZE));
+    assert((length & 0xf000) == 0);
+    assert((rev & ~0xf) == 0);
+
+    /* Create the DVSEC in the MCFG space */
+    pcie_add_capability(pdev, PCI_EXT_CAP_ID_DVSEC, 1, offset, length);
+    pci_set_long(pdev->config + offset + PCIE_DVSEC_HEADER1_OFFSET,
+                 (length << 20) | (rev << 16) | CXL_VENDOR_ID);
+    pci_set_word(pdev->config + offset + PCIE_DVSEC_ID_OFFSET, type);
+    memcpy(pdev->config + offset + sizeof(struct dvsec_header),
+           body + sizeof(struct dvsec_header),
+           length - sizeof(struct dvsec_header));
+
+    /* Update state for future DVSEC additions */
+    range_init_nofail(&cxl->dvsecs[type], cxl->dvsec_offset, length);
+    cxl->dvsec_offset += length;
+}
diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
new file mode 100644
index 0000000000..00c3876a0f
--- /dev/null
+++ b/hw/cxl/meson.build
@@ -0,0 +1,3 @@
+softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
+  'cxl-component-utils.c',
+))
diff --git a/hw/meson.build b/hw/meson.build
index 010de7219c..3e440c341a 100644
--- a/hw/meson.build
+++ b/hw/meson.build
@@ -6,6 +6,7 @@ subdir('block')
 subdir('char')
 subdir('core')
 subdir('cpu')
+subdir('cxl')
 subdir('display')
 subdir('dma')
 subdir('gpio')
diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
new file mode 100644
index 0000000000..55f6cc30a5
--- /dev/null
+++ b/include/hw/cxl/cxl.h
@@ -0,0 +1,17 @@
+/*
+ * QEMU CXL Support
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_H
+#define CXL_H
+
+#include "cxl_pci.h"
+#include "cxl_component.h"
+
+#endif
+
diff --git a/include/hw/cxl/cxl_component.h b/include/hw/cxl/cxl_component.h
new file mode 100644
index 0000000000..8b40360c22
--- /dev/null
+++ b/include/hw/cxl/cxl_component.h
@@ -0,0 +1,187 @@
+/*
+ * QEMU CXL Component
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_COMPONENT_H
+#define CXL_COMPONENT_H
+
+/* CXL 2.0 - 8.2.4 */
+#define CXL2_COMPONENT_IO_REGION_SIZE 0x1000
+#define CXL2_COMPONENT_CM_REGION_SIZE 0x1000
+#define CXL2_COMPONENT_BLOCK_SIZE 0x10000
+
+#include "qemu/range.h"
+#include "qemu/typedefs.h"
+#include "hw/register.h"
+
+enum reg_type {
+    CXL2_DEVICE,
+    CXL2_TYPE3_DEVICE,
+    CXL2_LOGICAL_DEVICE,
+    CXL2_ROOT_PORT,
+    CXL2_UPSTREAM_PORT,
+    CXL2_DOWNSTREAM_PORT
+};
+
+/*
+ * Capability registers are defined at the top of the CXL.cache/mem region and
+ * are packed. For our purposes we will always define the caps in the same
+ * order.
+ * CXL 2.0 - 8.2.5 Table 142 for details.
+ */
+
+/* CXL 2.0 - 8.2.5.1 */
+REG32(CXL_CAPABILITY_HEADER, 0)
+    FIELD(CXL_CAPABILITY_HEADER, ID, 0, 16)
+    FIELD(CXL_CAPABILITY_HEADER, VERSION, 16, 4)
+    FIELD(CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 20, 4)
+    FIELD(CXL_CAPABILITY_HEADER, ARRAY_SIZE, 24, 8)
+
+#define CXLx_CAPABILITY_HEADER(type, offset)                  \
+    REG32(CXL_##type##_CAPABILITY_HEADER, offset)             \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, ID, 0, 16)      \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, VERSION, 16, 4) \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, PTR, 20, 12)
+CXLx_CAPABILITY_HEADER(RAS, 0x4)
+CXLx_CAPABILITY_HEADER(LINK, 0x8)
+CXLx_CAPABILITY_HEADER(HDM, 0xc)
+CXLx_CAPABILITY_HEADER(EXTSEC, 0x10)
+CXLx_CAPABILITY_HEADER(SNOOP, 0x14)
+
+/*
+ * Capability structures contain the actual registers that the CXL component
+ * implements. Some of these are specific to certain types of components, but
+ * this implementation leaves enough space regardless.
+ */
+/* 8.2.5.9 - CXL RAS Capability Structure */
+#define CXL_RAS_REGISTERS_OFFSET 0x80 /* Give ample space for caps before this */
+#define CXL_RAS_REGISTERS_SIZE   0x58
+REG32(CXL_RAS_UNC_ERR_STATUS, CXL_RAS_REGISTERS_OFFSET)
+REG32(CXL_RAS_UNC_ERR_MASK, CXL_RAS_REGISTERS_OFFSET + 0x4)
+REG32(CXL_RAS_UNC_ERR_SEVERITY, CXL_RAS_REGISTERS_OFFSET + 0x8)
+REG32(CXL_RAS_COR_ERR_STATUS, CXL_RAS_REGISTERS_OFFSET + 0xc)
+REG32(CXL_RAS_COR_ERR_MASK, CXL_RAS_REGISTERS_OFFSET + 0x10)
+REG32(CXL_RAS_ERR_CAP_CTRL, CXL_RAS_REGISTERS_OFFSET + 0x14)
+/* Offset 0x18 - 0x58 reserved for RAS logs */
+
+/* 8.2.5.10 - CXL Security Capability Structure */
+#define CXL_SEC_REGISTERS_OFFSET (CXL_RAS_REGISTERS_OFFSET + CXL_RAS_REGISTERS_SIZE)
+#define CXL_SEC_REGISTERS_SIZE   0 /* We don't implement 1.1 downstream ports */
+
+/* 8.2.5.11 - CXL Link Capability Structure */
+#define CXL_LINK_REGISTERS_OFFSET (CXL_SEC_REGISTERS_OFFSET + CXL_SEC_REGISTERS_SIZE)
+#define CXL_LINK_REGISTERS_SIZE   0x38
+
+/* 8.2.5.12 - CXL HDM Decoder Capability Structure */
+#define HDM_DECODE_MAX 10 /* 8.2.5.12.1 */
+#define CXL_HDM_REGISTERS_OFFSET \
+    (CXL_LINK_REGISTERS_OFFSET + CXL_LINK_REGISTERS_SIZE) /* 8.2.5.12 */
+#define CXL_HDM_REGISTERS_SIZE (0x20 + HDM_DECODE_MAX * 10)
+#define HDM_DECODER_INIT(n)                                                    \
+  REG32(CXL_HDM_DECODER##n##_BASE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x10)                          \
+            FIELD(CXL_HDM_DECODER##n##_BASE_LO, L, 28, 4)                      \
+  REG32(CXL_HDM_DECODER##n##_BASE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x14)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x18)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x1C)                          \
+  REG32(CXL_HDM_DECODER##n##_CTRL,                                             \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x20)                          \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IG, 0, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IW, 4, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, LOCK_ON_COMMIT, 8, 1)             \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMIT, 9, 1)                     \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMITTED, 10, 1)                 \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, ERROR, 11, 1)                     \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, TYPE, 12, 1)                      \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_LO, 0x24)                             \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_HI, 0x28)
+
+REG32(CXL_HDM_DECODER_CAPABILITY, 0)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 4, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTERLEAVE_256B, 8, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTELEAVE_4K, 9, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, POISON_ON_ERR_CAP, 10, 1)
+REG32(CXL_HDM_DECODER_GLOBAL_CONTROL, 0)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, POISON_ON_ERR_EN, 0, 1)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, HDM_DECODER_ENABLE, 1, 1)
+
+HDM_DECODER_INIT(0);
+
+/* 8.2.5.13 - CXL Extended Security Capability Structure (Root complex only) */
+#define EXTSEC_ENTRY_MAX        256
+#define CXL_EXTSEC_REGISTERS_OFFSET (CXL_HDM_REGISTERS_OFFSET + CXL_HDM_REGISTERS_SIZE)
+#define CXL_EXTSEC_REGISTERS_SIZE   (8 * EXTSEC_ENTRY_MAX + 4)
+
+/* 8.2.5.14 - CXL IDE Capability Structure */
+#define CXL_IDE_REGISTERS_OFFSET (CXL_EXTSEC_REGISTERS_OFFSET + CXL_EXTSEC_REGISTERS_SIZE)
+#define CXL_IDE_REGISTERS_SIZE   0
+
+/* 8.2.5.15 - CXL Snoop Filter Capability Structure */
+#define CXL_SNOOP_REGISTERS_OFFSET (CXL_IDE_REGISTERS_OFFSET + CXL_IDE_REGISTERS_SIZE)
+#define CXL_SNOOP_REGISTERS_SIZE   0x8
+
+_Static_assert((CXL_SNOOP_REGISTERS_OFFSET + CXL_SNOOP_REGISTERS_SIZE) < 0x1000,
+               "No space for registers");
+
+typedef struct component_registers {
+    /*
+     * Main memory region to be registered with QEMU core.
+     */
+    MemoryRegion component_registers;
+
+    /*
+     * 8.2.4 Table 141:
+     *   0x0000 - 0x0fff CXL.io registers
+     *   0x1000 - 0x1fff CXL.cache and CXL.mem
+     *   0x2000 - 0xdfff Implementation specific
+     *   0xe000 - 0xe3ff CXL ARB/MUX registers
+     *   0xe400 - 0xffff RSVD
+     */
+    uint32_t io_registers[CXL2_COMPONENT_IO_REGION_SIZE >> 2];
+    MemoryRegion io;
+
+    uint32_t cache_mem_registers[CXL2_COMPONENT_CM_REGION_SIZE >> 2];
+    MemoryRegion cache_mem;
+
+    MemoryRegion impl_specific;
+    MemoryRegion arb_mux;
+    MemoryRegion rsvd;
+
+    /* special_ops is used for any component that needs any specific handling */
+    MemoryRegionOps *special_ops;
+} ComponentRegisters;
+
+/*
+ * A CXL component represents all entities in a CXL hierarchy. This includes,
+ * host bridges, root ports, upstream/downstream switch ports, and devices
+ */
+typedef struct cxl_component {
+    ComponentRegisters crb;
+    union {
+        struct {
+            Range dvsecs[CXL20_MAX_DVSEC];
+            uint16_t dvsec_offset;
+            struct PCIDevice *pdev;
+        };
+    };
+} CXLComponentState;
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type);
+void cxl_component_register_init_common(uint32_t *reg_state,
+                                        enum reg_type type);
+
+void cxl_component_create_dvsec(CXLComponentState *cxl_cstate, uint16_t length,
+                                uint16_t type, uint8_t rev, uint8_t *body);
+
+#endif
diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
new file mode 100644
index 0000000000..a53c2e5ae7
--- /dev/null
+++ b/include/hw/cxl/cxl_pci.h
@@ -0,0 +1,138 @@
+/*
+ * QEMU CXL PCI interfaces
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_PCI_H
+#define CXL_PCI_H
+
+#include "hw/pci/pci.h"
+#include "hw/pci/pcie.h"
+
+#define CXL_VENDOR_ID 0x1e98
+
+#define PCIE_DVSEC_HEADER1_OFFSET 0x4 /* Offset from start of extend cap */
+#define PCIE_DVSEC_ID_OFFSET 0x8
+
+#define PCIE_CXL_DEVICE_DVSEC_LENGTH 0x38
+#define PCIE_CXL1_DEVICE_DVSEC_REVID 0
+#define PCIE_CXL2_DEVICE_DVSEC_REVID 1
+
+#define EXTENSIONS_PORT_DVSEC_LENGTH 0x28
+#define EXTENSIONS_PORT_DVSEC_REVID 0
+
+#define GPF_PORT_DVSEC_LENGTH 0x10
+#define GPF_PORT_DVSEC_REVID  0
+
+#define PCIE_FLEXBUS_PORT_DVSEC_LENGTH_2_0 0x14
+#define PCIE_FLEXBUS_PORT_DVSEC_REVID_2_0  1
+
+#define REG_LOC_DVSEC_LENGTH 0x24
+#define REG_LOC_DVSEC_REVID  0
+
+enum {
+    PCIE_CXL_DEVICE_DVSEC      = 0,
+    NON_CXL_FUNCTION_MAP_DVSEC = 2,
+    EXTENSIONS_PORT_DVSEC      = 3,
+    GPF_PORT_DVSEC             = 4,
+    GPF_DEVICE_DVSEC           = 5,
+    PCIE_FLEXBUS_PORT_DVSEC    = 7,
+    REG_LOC_DVSEC              = 8,
+    MLD_DVSEC                  = 9,
+    CXL20_MAX_DVSEC
+};
+
+struct dvsec_header {
+    uint32_t cap_hdr;
+    uint32_t dv_hdr1;
+    uint16_t dv_hdr2;
+} __attribute__((__packed__));
+_Static_assert(sizeof(struct dvsec_header) == 10,
+               "dvsec header size incorrect");
+
+/*
+ * CXL 2.0 devices must implement certain DVSEC IDs, and can [optionally]
+ * implement others.
+ *
+ * CXL 2.0 Device: 0, [2], 5, 8
+ * CXL 2.0 RP: 3, 4, 7, 8
+ * CXL 2.0 Upstream Port: [2], 7, 8
+ * CXL 2.0 Downstream Port: 3, 4, 7, 8
+ */
+
+/* CXL 2.0 - 8.1.5 (ID 0003) */
+struct extensions_dvsec_port {
+    struct dvsec_header hdr;
+    uint16_t status;
+    uint16_t control;
+    uint8_t alt_bus_base;
+    uint8_t alt_bus_limit;
+    uint16_t alt_memory_base;
+    uint16_t alt_memory_limit;
+    uint16_t alt_prefetch_base;
+    uint16_t alt_prefetch_limit;
+    uint32_t alt_prefetch_base_high;
+    uint32_t alt_prefetch_base_low;
+    uint32_t rcrb_base;
+    uint32_t rcrb_base_high;
+};
+_Static_assert(sizeof(struct extensions_dvsec_port) == 0x28,
+               "extensions dvsec port size incorrect");
+#define PORT_CONTROL_OVERRIDE_OFFSET 0xc
+#define PORT_CONTROL_UNMASK_SBR      1
+#define PORT_CONTROL_ALT_MEMID_EN    4
+
+/* CXL 2.0 - 8.1.6 GPF DVSEC (ID 0004) */
+struct dvsec_port_gpf {
+    struct dvsec_header hdr;
+    uint16_t rsvd;
+    uint16_t phase1_ctrl;
+    uint16_t phase2_ctrl;
+};
+_Static_assert(sizeof(struct dvsec_port_gpf) == 0x10,
+               "dvsec port GPF size incorrect");
+
+/* CXL 2.0 - 8.1.8/8.2.1.3 Flexbus DVSEC (ID 0007) */
+struct dvsec_port_flexbus {
+    struct dvsec_header hdr;
+    uint16_t cap;
+    uint16_t ctrl;
+    uint16_t status;
+    uint32_t rcvd_mod_ts_data;
+};
+_Static_assert(sizeof(struct dvsec_port_flexbus) == 0x14,
+               "dvsec port flexbus size incorrect");
+
+/* CXL 2.0 - 8.1.9 Register Locator DVSEC (ID 0008) */
+struct dvsec_register_locator {
+    struct dvsec_header hdr;
+    uint16_t rsvd;
+    uint32_t reg0_base_lo;
+    uint32_t reg0_base_hi;
+    uint32_t reg1_base_lo;
+    uint32_t reg1_base_hi;
+    uint32_t reg2_base_lo;
+    uint32_t reg2_base_hi;
+};
+_Static_assert(sizeof(struct dvsec_register_locator) == 0x24,
+               "dvsec register locator size incorrect");
+
+/* BAR Equivalence Indicator */
+#define BEI_BAR_10H 0
+#define BEI_BAR_14H 1
+#define BEI_BAR_18H 2
+#define BEI_BAR_1cH 3
+#define BEI_BAR_20H 4
+#define BEI_BAR_24H 5
+
+/* Register Block Identifier */
+#define RBI_EMPTY          0
+#define RBI_COMPONENT_REG  (1 << 8)
+#define RBI_BAR_VIRT_ACL   (2 << 8)
+#define RBI_CXL_DEVICE_REG (3 << 8)
+
+#endif
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 04/32] hw/cxl/device: Introduce a CXL device (8.2.8)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (2 preceding siblings ...)
  2021-01-05 16:52 ` [RFC PATCH v2 03/32] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5) Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2) Ben Widawsky
                   ` (28 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

A CXL device is a type of CXL component. Conceptually, a CXL device
would be a leaf node in a CXL topology. From an emulation perspective,
CXL devices are the most complex and so the actual implementation is
reserved for discrete commits.

This new device type is specifically catered towards the eventually
implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0
specification.

v2: Remove error codes not yet used (Jonathan)
    Fix interrupt enable bit width (Jonathan)
    Add comment for weird register size (Jonathan)
    Remove commands/registers not yet used (Ben)

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 include/hw/cxl/cxl.h        |   1 +
 include/hw/cxl/cxl_device.h | 151 ++++++++++++++++++++++++++++++++++++
 2 files changed, 152 insertions(+)
 create mode 100644 include/hw/cxl/cxl_device.h

diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
index 55f6cc30a5..23f52c4cf9 100644
--- a/include/hw/cxl/cxl.h
+++ b/include/hw/cxl/cxl.h
@@ -12,6 +12,7 @@
 
 #include "cxl_pci.h"
 #include "cxl_component.h"
+#include "cxl_device.h"
 
 #endif
 
diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
new file mode 100644
index 0000000000..38af43c1a6
--- /dev/null
+++ b/include/hw/cxl/cxl_device.h
@@ -0,0 +1,151 @@
+/*
+ * QEMU CXL Devices
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_DEVICE_H
+#define CXL_DEVICE_H
+
+#include "hw/register.h"
+
+/*
+ * The following is how a CXL device's MMIO space is laid out. The only
+ * requirement from the spec is that the capabilities array and the capability
+ * headers start at offset 0 and are contiguously packed. The headers themselves
+ * provide offsets to the register fields. For this emulation, registers will
+ * start at offset 0x80 (m == 0x80). No secondary mailbox is implemented which
+ * means that n = m + sizeof(mailbox registers) + sizeof(device registers).
+ *
+ * This is roughly described in 8.2.8 Figure 138 of the CXL 2.0 spec.
+ *
+ * n + PAYLOAD_SIZE_MAX  +---------------------------------+
+ *                       |                                 |
+ *                  ^    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |         Command Payload         |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  n    +---------------------------------+
+ *                  ^    |                                 |
+ *                  |    |    Device Capability Registers  |
+ *                  |    |    x, mailbox, y                |
+ *                  |    |                                 |
+ *                  m    +---------------------------------+
+ *                  ^    |     Device Capability Header y  |
+ *                  |    +---------------------------------+
+ *                  |    | Device Capability Header Mailbox|
+ *                  |    +------------- --------------------
+ *                  |    |     Device Capability Header x  |
+ *                  |    +---------------------------------+
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |      Device Cap Array[0..n]     |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  0    +---------------------------------+
+ */
+
+#define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
+#define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
+
+#define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */
+#define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */
+
+#define CXL_MAILBOX_REGISTERS_OFFSET \
+    (CXL_DEVICE_REGISTERS_OFFSET + CXL_DEVICE_REGISTERS_LENGTH)
+#define CXL_MAILBOX_REGISTERS_SIZE 0x20
+#define CXL_MAILBOX_PAYLOAD_SHIFT 11
+#define CXL_MAILBOX_MAX_PAYLOAD_SIZE (1 << CXL_MAILBOX_PAYLOAD_SHIFT)
+#define CXL_MAILBOX_REGISTERS_LENGTH \
+    (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
+
+typedef struct cxl_device_state {
+    /* Boss container and caps registers */
+    MemoryRegion device_registers;
+
+    struct {
+        MemoryRegion caps;
+        union {
+            uint8_t caps_reg_state[CXL_DEVICE_CAP_REG_SIZE * 4]; /* ARRAY + 3 CAPS */
+            uint32_t caps_reg_state32[CXL_DEVICE_CAP_REG_SIZE >> 2];
+        };
+    };
+    MemoryRegion device;
+    MemoryRegion mailbox;
+
+    MemoryRegion *pmem;
+    MemoryRegion *vmem;
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);
+
+/* Set up default values for the register block */
+void cxl_device_register_init_common(CXLDeviceState *dev);
+
+/* CXL 2.0 - 8.2.8.1 */
+REG32(CXL_DEV_CAP_ARRAY, 0) /* 48b!?!?! */
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_ID, 0, 16)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_VERSION, 16, 8)
+REG32(CXL_DEV_CAP_ARRAY2, 4) /* We're going to pretend it's 64b */
+    FIELD(CXL_DEV_CAP_ARRAY2, CAP_COUNT, 0, 16)
+
+/*
+ * In the 8.2.8.2, this is listed as a 128b register, but in 8.2.8, it says:
+ * > No registers defined in Section 8.2.8 are larger than 64-bits wide so that
+ * > is the maximum access size allowed for these registers. If this rule is not
+ * > followed, the behavior is undefined
+ *
+ * Here we've chosen to make it 4 dwords. The spec allows any pow2 multiple
+ * access to be used for a register (2 qwords, 8 words, 128 bytes).
+ */
+#define CXL_DEVICE_CAPABILITY_HEADER_REGISTER(n, offset)                            \
+    REG32(CXL_DEV_##n##_CAP_HDR0, offset)                 \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_ID, 0, 16)      \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_VERSION, 16, 8) \
+    REG32(CXL_DEV_##n##_CAP_HDR1, offset + 4)             \
+        FIELD(CXL_DEV_##n##_CAP_HDR1, CAP_OFFSET, 0, 32)  \
+    REG32(CXL_DEV_##n##_CAP_HDR2, offset + 8)             \
+        FIELD(CXL_DEV_##n##_CAP_HDR2, CAP_LENGTH, 0, 32)
+
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
+                                               CXL_DEVICE_CAP_REG_SIZE)
+
+REG32(CXL_DEV_MAILBOX_CAP, 0)
+    FIELD(CXL_DEV_MAILBOX_CAP, PAYLOAD_SIZE, 0, 5)
+    FIELD(CXL_DEV_MAILBOX_CAP, INT_CAP, 5, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, BG_INT_CAP, 6, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, MSI_N, 7, 4)
+
+REG32(CXL_DEV_MAILBOX_CTRL, 4)
+    FIELD(CXL_DEV_MAILBOX_CTRL, DOORBELL, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
+
+/* XXX: actually a 64b register */
+REG32(CXL_DEV_MAILBOX_STS, 0x10)
+    FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_STS, ERRNO, 32, 16)
+    FIELD(CXL_DEV_MAILBOX_STS, VENDOR_ERRNO, 48, 16)
+
+/* XXX: actually a 64b register */
+REG32(CXL_DEV_BG_CMD_STS, 0x18)
+    FIELD(CXL_DEV_BG_CMD_STS, BG, 0, 16)
+    FIELD(CXL_DEV_BG_CMD_STS, DONE, 16, 7)
+    FIELD(CXL_DEV_BG_CMD_STS, ERRNO, 32, 16)
+    FIELD(CXL_DEV_BG_CMD_STS, VENDOR_ERRNO, 48, 16)
+
+REG32(CXL_DEV_CMD_PAYLOAD, 0x20)
+
+#endif
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (3 preceding siblings ...)
  2021-01-05 16:52 ` [RFC PATCH v2 04/32] hw/cxl/device: Introduce a CXL device (8.2.8) Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-06 13:28   ` Jonathan Cameron
  2021-01-05 16:52 ` [RFC PATCH v2 06/32] hw/cxl/device: Add device status (8.2.8.3) Ben Widawsky
                   ` (27 subsequent siblings)
  32 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This implements all device MMIO up to the first capability .That
includes the CXL Device Capabilities Array Register, as well as all of
the CXL Device Capability Header Registers. The latter are filled in as
they are implemented in the following patches.

v2: Break out register alignment checks (Jonathan)

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-device-utils.c | 72 +++++++++++++++++++++++++++++++++++++++
 hw/cxl/meson.build        |  1 +
 2 files changed, 73 insertions(+)
 create mode 100644 hw/cxl/cxl-device-utils.c

diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
new file mode 100644
index 0000000000..d1b1371e66
--- /dev/null
+++ b/hw/cxl/cxl-device-utils.c
@@ -0,0 +1,72 @@
+/*
+ * CXL Utility library for devices
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "hw/cxl/cxl.h"
+
+static int cxl_device_check_register_alignment(hwaddr offset, unsigned size)
+{
+    if (unlikely(offset & (size - 1))) {
+        return 1;
+    }
+
+    return 0;
+}
+
+static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    CXLDeviceState *cxl_dstate = opaque;
+
+    if (cxl_device_check_register_alignment(offset, size)) {
+        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+        return 0;
+    }
+
+    return ldn_le_p(cxl_dstate->caps_reg_state + offset, size);
+}
+
+static const MemoryRegionOps caps_ops = {
+    .read = caps_reg_read,
+    .write = NULL,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
+{
+    /* This will be a BAR, so needs to be rounded up to pow2 for PCI spec */
+    memory_region_init(
+        &cxl_dstate->device_registers, obj, "device-registers",
+        pow2ceil(CXL_MAILBOX_REGISTERS_LENGTH + CXL_MAILBOX_REGISTERS_OFFSET));
+
+    memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
+                          "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
+
+    memory_region_add_subregion(&cxl_dstate->device_registers, 0,
+                                &cxl_dstate->caps);
+}
+
+void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
+{
+    uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
+    const int cap_count = 0;
+
+    /* CXL Device Capabilities Array Register */
+    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
+    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
+    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY2, CAP_COUNT, cap_count);
+}
diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
index 00c3876a0f..47154d6850 100644
--- a/hw/cxl/meson.build
+++ b/hw/cxl/meson.build
@@ -1,3 +1,4 @@
 softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
   'cxl-component-utils.c',
+  'cxl-device-utils.c',
 ))
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 06/32] hw/cxl/device: Add device status (8.2.8.3)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (4 preceding siblings ...)
  2021-01-05 16:52 ` [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2) Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-05 16:52 ` [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4) Ben Widawsky
                   ` (26 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This implements the CXL device status registers from 8.2.8.3.1 in the
CXL 2.0 specification. It is capability ID 0001h.

v2: Use the reg alignment helper (Jonathan)
    Add stubbed device_reg_init_common() (Ben)

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-device-utils.c   | 38 ++++++++++++++++++++++++++++++++++++-
 include/hw/cxl/cxl_device.h | 16 ++++++++++++++++
 2 files changed, 53 insertions(+), 1 deletion(-)

diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
index d1b1371e66..b86e5466bd 100644
--- a/hw/cxl/cxl-device-utils.c
+++ b/hw/cxl/cxl-device-utils.c
@@ -32,6 +32,32 @@ static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
     return ldn_le_p(cxl_dstate->caps_reg_state + offset, size);
 }
 
+static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    uint64_t retval = 0;
+
+    if (cxl_device_check_register_alignment(offset, size)) {
+        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+        return 0;
+    }
+
+    return ldn_le_p(&retval, size);
+}
+
+static const MemoryRegionOps dev_ops = {
+    .read = dev_reg_read,
+    .write = NULL,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
 static const MemoryRegionOps caps_ops = {
     .read = caps_reg_read,
     .write = NULL,
@@ -55,18 +81,28 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
 
     memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
                           "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
+    memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
+                          "device-status", CXL_DEVICE_REGISTERS_LENGTH);
 
     memory_region_add_subregion(&cxl_dstate->device_registers, 0,
                                 &cxl_dstate->caps);
+    memory_region_add_subregion(&cxl_dstate->device_registers,
+                                CXL_DEVICE_REGISTERS_OFFSET,
+                                &cxl_dstate->device);
 }
 
+static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
+
 void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 {
     uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
-    const int cap_count = 0;
+    const int cap_count = 1;
 
     /* CXL Device Capabilities Array Register */
     ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
     ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
     ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY2, CAP_COUNT, cap_count);
+
+    cxl_device_cap_init(cxl_dstate, DEVICE, 1);
+    device_reg_init_common(cxl_dstate);
 }
diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
index 38af43c1a6..3380fc9e7a 100644
--- a/include/hw/cxl/cxl_device.h
+++ b/include/hw/cxl/cxl_device.h
@@ -122,6 +122,22 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
                                                CXL_DEVICE_CAP_REG_SIZE)
 
+#define cxl_device_cap_init(dstate, reg, cap_id)                                   \
+    do {                                                                           \
+        uint32_t *cap_hdrs = dstate->caps_reg_state32;                             \
+        int which = R_CXL_DEV_##reg##_CAP_HDR0;                                    \
+        cap_hdrs[which] =                                                          \
+            FIELD_DP32(cap_hdrs[which], CXL_DEV_##reg##_CAP_HDR0, CAP_ID, cap_id); \
+        cap_hdrs[which] = FIELD_DP32(                                              \
+            cap_hdrs[which], CXL_DEV_##reg##_CAP_HDR0, CAP_VERSION, 1);            \
+        cap_hdrs[which + 1] =                                                      \
+            FIELD_DP32(cap_hdrs[which + 1], CXL_DEV_##reg##_CAP_HDR1,              \
+                       CAP_OFFSET, CXL_##reg##_REGISTERS_OFFSET);                  \
+        cap_hdrs[which + 2] =                                                      \
+            FIELD_DP32(cap_hdrs[which + 2], CXL_DEV_##reg##_CAP_HDR2,              \
+                       CAP_LENGTH, CXL_##reg##_REGISTERS_LENGTH);                  \
+    } while (0)
+
 REG32(CXL_DEV_MAILBOX_CAP, 0)
     FIELD(CXL_DEV_MAILBOX_CAP, PAYLOAD_SIZE, 0, 5)
     FIELD(CXL_DEV_MAILBOX_CAP, INT_CAP, 5, 1)
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (5 preceding siblings ...)
  2021-01-05 16:52 ` [RFC PATCH v2 06/32] hw/cxl/device: Add device status (8.2.8.3) Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-06 13:21   ` Jonathan Cameron
  2021-01-05 16:52 ` [RFC PATCH v2 08/32] hw/cxl/device: Add memory devices (8.2.8.5) Ben Widawsky
                   ` (25 subsequent siblings)
  32 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This is the beginning of implementing mailbox support for CXL 2.0
devices.

v2: Use register alignment helper (Ben)
    Minor cleanups (Jonathan)
    Rename error codes to match spec (Jonathan)
    Update cap count from 1 to 2 (Jonathan)
    Add infra to support CEL (Ben)
    Add more of the actual mailbox handling from later patch (Ben)

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-device-utils.c   | 122 ++++++++++++++++++++++++-
 hw/cxl/cxl-mailbox-utils.c  | 173 ++++++++++++++++++++++++++++++++++++
 hw/cxl/meson.build          |   1 +
 include/hw/cxl/cxl.h        |   3 +
 include/hw/cxl/cxl_device.h |  27 +++++-
 5 files changed, 322 insertions(+), 4 deletions(-)
 create mode 100644 hw/cxl/cxl-mailbox-utils.c

diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
index b86e5466bd..642e3c2617 100644
--- a/hw/cxl/cxl-device-utils.c
+++ b/hw/cxl/cxl-device-utils.c
@@ -44,6 +44,108 @@ static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
     return ldn_le_p(&retval, size);
 }
 
+static uint64_t mailbox_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    CXLDeviceState *cxl_dstate = opaque;
+
+    if (cxl_device_check_register_alignment(offset, size)) {
+        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+        return 0;
+    }
+
+    return ldn_le_p(cxl_dstate->mbox_reg_state + offset, size);
+}
+
+static void mailbox_mem_writel(uint32_t *reg_state, hwaddr offset,
+                               uint64_t value)
+{
+    switch (offset) {
+    case A_CXL_DEV_MAILBOX_CTRL:
+        /* fallthrough */
+    case A_CXL_DEV_MAILBOX_CAP:
+        /* RO register */
+        break;
+    default:
+        qemu_log_mask(LOG_UNIMP,
+                      "%s Unexpected 32-bit access to 0x%" PRIx64 " (WI)\n",
+                      __func__, offset);
+        break;
+    }
+
+    stl_le_p((uint8_t *)reg_state + offset, value);
+}
+
+static void mailbox_mem_writeq(uint8_t *reg_state, hwaddr offset,
+                               uint64_t value)
+{
+    switch (offset) {
+    case A_CXL_DEV_MAILBOX_CMD:
+        break;
+    case A_CXL_DEV_BG_CMD_STS:
+        /* BG not supported */
+        /* fallthrough */
+    case A_CXL_DEV_MAILBOX_STS:
+        /* Read only register, will get updated by the state machine */
+        return;
+    default:
+        qemu_log_mask(LOG_UNIMP,
+                      "%s Unexpected 64-bit access to 0x%" PRIx64 " (WI)\n",
+                      __func__, offset);
+        return;
+    }
+
+    stq_le_p(reg_state + offset, value);
+}
+
+static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
+                              unsigned size)
+{
+    CXLDeviceState *cxl_dstate = opaque;
+
+    /*
+     * Lock is needed to prevent concurrent writes as well as to prevent writes
+     * coming in while the firmware is processing. Without background commands
+     * or the second mailbox implemented, this serves no purpose since the
+     * memory access is synchronized at a higher level (per memory region).
+     */
+    RCU_READ_LOCK_GUARD();
+
+    switch (size) {
+    case 4:
+        if (unlikely(offset & (sizeof(uint32_t) - 1))) {
+            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+            return;
+        }
+        mailbox_mem_writel(cxl_dstate->mbox_reg_state32, offset, value);
+        break;
+    case 8:
+        if (unlikely(offset & (sizeof(uint64_t) - 1))) {
+            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+            return;
+        }
+        mailbox_mem_writeq(cxl_dstate->mbox_reg_state, offset, value);
+        break;
+    }
+
+    if (ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
+                         DOORBELL))
+        cxl_process_mailbox(cxl_dstate);
+}
+
+static const MemoryRegionOps mailbox_ops = {
+    .read = mailbox_reg_read,
+    .write = mailbox_reg_write,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
 static const MemoryRegionOps dev_ops = {
     .read = dev_reg_read,
     .write = NULL,
@@ -83,20 +185,33 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
                           "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
     memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
                           "device-status", CXL_DEVICE_REGISTERS_LENGTH);
+    memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
+                          "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
 
     memory_region_add_subregion(&cxl_dstate->device_registers, 0,
                                 &cxl_dstate->caps);
     memory_region_add_subregion(&cxl_dstate->device_registers,
                                 CXL_DEVICE_REGISTERS_OFFSET,
                                 &cxl_dstate->device);
+    memory_region_add_subregion(&cxl_dstate->device_registers,
+                                CXL_MAILBOX_REGISTERS_OFFSET,
+                                &cxl_dstate->mailbox);
 }
 
 static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
 
+static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
+{
+    /* 2048 payload size, with no interrupt or background support */
+    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CAP,
+                     PAYLOAD_SIZE, CXL_MAILBOX_PAYLOAD_SHIFT);
+    cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
+}
+
 void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 {
     uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
-    const int cap_count = 1;
+    const int cap_count = 2;
 
     /* CXL Device Capabilities Array Register */
     ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
@@ -105,4 +220,9 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 
     cxl_device_cap_init(cxl_dstate, DEVICE, 1);
     device_reg_init_common(cxl_dstate);
+
+    cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
+    mailbox_reg_init_common(cxl_dstate);
+
+    assert(cxl_initialize_mailbox(cxl_dstate) == 0);
 }
diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
new file mode 100644
index 0000000000..3a39c936de
--- /dev/null
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -0,0 +1,173 @@
+/*
+ * CXL Utility library for mailbox interface
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "hw/cxl/cxl.h"
+#include "hw/pci/pci.h"
+#include "qemu/log.h"
+#include "qemu/uuid.h"
+
+/*
+ * How to add a new command, example. The command set FOO, with cmd BAR.
+ *  1. Add the command set and cmd to the enum.
+ *     FOO    = 0x7f,
+ *          #define BAR 0
+ *  2. Forward declare the handler.
+ *     declare_mailbox_handler(FOO_BAR);
+ *  3. Add the command to the cxl_cmd_set[][]
+ *     CXL_CMD(FOO, BAR, 0, 0),
+ *  4. Implement your handler
+ *     define_mailbox_handler(FOO_BAR) { ... return CXL_MBOX_SUCCESS; }
+ *
+ *
+ *  Writing the handler:
+ *    The handler will provide the &struct cxl_cmd, the &CXLDeviceState, and the
+ *    in/out length of the payload. The handler is responsible for consuming the
+ *    payload from cmd->payload and operating upon it as necessary. It must then
+ *    fill the output data into cmd->payload (overwriting what was there),
+ *    setting the length, and returning a valid return code.
+ *
+ *  XXX: The handler need not worry about endianess. The payload is read out of
+ *  a register interface that already deals with it.
+ */
+
+/* 8.2.8.4.5.1 Command Return Codes */
+typedef enum {
+    CXL_MBOX_SUCCESS = 0x0,
+    CXL_MBOX_BG_STARTED = 0x1,
+    CXL_MBOX_INVALID_INPUT = 0x2,
+    CXL_MBOX_UNSUPPORTED = 0x3,
+    CXL_MBOX_INTERNAL_ERROR = 0x4,
+    CXL_MBOX_RETRY_REQUIRED = 0x5,
+    CXL_MBOX_BUSY = 0x6,
+    CXL_MBOX_MEDIA_DISABLED = 0x7,
+    CXL_MBOX_FW_XFER_IN_PROGRESS = 0x8,
+    CXL_MBOX_FW_XFER_OUT_OF_ORDER = 0x9,
+    CXL_MBOX_FW_AUTH_FAILED = 0xa,
+    CXL_MBOX_FW_INVALID_SLOT = 0xb,
+    CXL_MBOX_FW_ROLLEDBACK = 0xc,
+    CXL_MBOX_FW_REST_REQD = 0xd,
+    CXL_MBOX_INVALID_HANDLE = 0xe,
+    CXL_MBOX_INVALID_PA = 0xf,
+    CXL_MBOX_INJECT_POISON_LIMIT = 0x10,
+    CXL_MBOX_PERMANENT_MEDIA_FAILURE = 0x11,
+    CXL_MBOX_ABORTED = 0x12,
+    CXL_MBOX_INVALID_SECURITY_STATE = 0x13,
+    CXL_MBOX_INCORRECT_PASSPHRASE = 0x14,
+    CXL_MBOX_UNSUPPORTED_MAILBOX = 0x15,
+    CXL_MBOX_INVALID_PAYLOAD_LENGTH = 0x16,
+    CXL_MBOX_MAX = 0x17
+} ret_code;
+
+struct cxl_cmd;
+typedef ret_code (*opcode_handler)(struct cxl_cmd *cmd,
+                                   CXLDeviceState *cxl_dstate, uint16_t *len);
+struct cxl_cmd {
+    const char *name;
+    opcode_handler handler;
+    ssize_t in;
+    uint16_t effect; /* Reported in CEL */
+    uint8_t *payload;
+};
+
+#define define_mailbox_handler(name)                \
+    static ret_code cmd_##name(struct cxl_cmd *cmd, \
+                               CXLDeviceState *cxl_dstate, uint16_t *len)
+#define declare_mailbox_handler(name) define_mailbox_handler(name)
+
+#define CXL_CMD(s, c, in, cel_effect) \
+    [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
+
+static struct cxl_cmd cxl_cmd_set[256][256] = {};
+
+#undef CXL_CMD
+
+QemuUUID cel_uuid;
+
+void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
+{
+    uint16_t ret = CXL_MBOX_SUCCESS;
+    struct cxl_cmd *cxl_cmd;
+    uint64_t status_reg;
+    opcode_handler h;
+
+    /*
+     * current state of mailbox interface
+     *  mbox_cap_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CAP];
+     *  mbox_ctrl_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CTRL];
+     *  status_reg = *(uint64_t *)&cxl_dstate->reg_state[A_CXL_DEV_MAILBOX_STS];
+     */
+    uint64_t command_reg =
+        *(uint64_t *)&cxl_dstate->mbox_reg_state[A_CXL_DEV_MAILBOX_CMD];
+
+    /* Check if we have to do anything */
+    if (!ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
+                          DOORBELL)) {
+        qemu_log_mask(LOG_UNIMP, "Corrupt internal state for firmware\n");
+        return;
+    }
+
+    uint8_t set = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET);
+    uint8_t cmd = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND);
+    uint16_t len = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH);
+    cxl_cmd = &cxl_cmd_set[set][cmd];
+    h = cxl_cmd->handler;
+    if (!h) {
+        goto handled;
+    }
+
+    if (len != cxl_cmd->in) {
+        ret = CXL_MBOX_INVALID_PAYLOAD_LENGTH;
+    }
+
+    cxl_cmd->payload = cxl_dstate->mbox_reg_state + A_CXL_DEV_CMD_PAYLOAD;
+    ret = (*h)(cxl_cmd, cxl_dstate, &len);
+    assert(len <= cxl_dstate->payload_size);
+
+handled:
+    /*
+     * Set the return code
+     * XXX: it's a 64b register, but we're not setting the vendor, so we can get
+     * away with this
+     */
+    status_reg = FIELD_DP64(0, CXL_DEV_MAILBOX_STS, ERRNO, ret);
+
+    /*
+     * Set the return length
+     */
+    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET, 0);
+    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND, 0);
+    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH, len);
+
+    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_CMD, command_reg);
+    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_STS, status_reg);
+
+    /* Tell the host we're done */
+    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
+                     DOORBELL, 0);
+}
+
+int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate)
+{
+    const char *cel_uuidstr = "0da9c0b5-bf41-4b78-8f79-96b1623b3f17";
+
+    for (int i = 0; i < 256; i++) {
+        for (int j = 0; j < 256; j++) {
+            if (cxl_cmd_set[i][j].handler) {
+                struct cxl_cmd *c = &cxl_cmd_set[i][j];
+
+                cxl_dstate->cel_log[cxl_dstate->cel_size].opcode = (i << 8) | j;
+                cxl_dstate->cel_log[cxl_dstate->cel_size].effect = c->effect;
+                cxl_dstate->cel_size++;
+            }
+        }
+    }
+
+    return qemu_uuid_parse(cel_uuidstr, &cel_uuid);
+}
diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
index 47154d6850..0eca715d10 100644
--- a/hw/cxl/meson.build
+++ b/hw/cxl/meson.build
@@ -1,4 +1,5 @@
 softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
   'cxl-component-utils.c',
   'cxl-device-utils.c',
+  'cxl-mailbox-utils.c',
 ))
diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
index 23f52c4cf9..362cda40de 100644
--- a/include/hw/cxl/cxl.h
+++ b/include/hw/cxl/cxl.h
@@ -14,5 +14,8 @@
 #include "cxl_component.h"
 #include "cxl_device.h"
 
+#define COMPONENT_REG_BAR_IDX 0
+#define DEVICE_REG_BAR_IDX 2
+
 #endif
 
diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
index 3380fc9e7a..bdf990cec2 100644
--- a/include/hw/cxl/cxl_device.h
+++ b/include/hw/cxl/cxl_device.h
@@ -70,9 +70,10 @@
     (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
 
 typedef struct cxl_device_state {
-    /* Boss container and caps registers */
+    /* Main register container */
     MemoryRegion device_registers;
 
+    MemoryRegion device;
     struct {
         MemoryRegion caps;
         union {
@@ -80,8 +81,19 @@ typedef struct cxl_device_state {
             uint32_t caps_reg_state32[CXL_DEVICE_CAP_REG_SIZE >> 2];
         };
     };
-    MemoryRegion device;
-    MemoryRegion mailbox;
+    struct {
+        MemoryRegion mailbox;
+        uint16_t payload_size;
+        union {
+            uint8_t mbox_reg_state[CXL_MAILBOX_REGISTERS_LENGTH];
+            uint32_t mbox_reg_state32[CXL_MAILBOX_REGISTERS_LENGTH >> 2];
+        };
+        struct {
+            uint16_t opcode;
+            uint16_t effect;
+        } cel_log[1 << 16];
+        size_t cel_size;
+    };
 
     MemoryRegion *pmem;
     MemoryRegion *vmem;
@@ -122,6 +134,9 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
                                                CXL_DEVICE_CAP_REG_SIZE)
 
+int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
+void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
+
 #define cxl_device_cap_init(dstate, reg, cap_id)                                   \
     do {                                                                           \
         uint32_t *cap_hdrs = dstate->caps_reg_state32;                             \
@@ -149,6 +164,12 @@ REG32(CXL_DEV_MAILBOX_CTRL, 4)
     FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
     FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
 
+/* XXX: actually a 64b register */
+REG32(CXL_DEV_MAILBOX_CMD, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
+
 /* XXX: actually a 64b register */
 REG32(CXL_DEV_MAILBOX_STS, 0x10)
     FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 08/32] hw/cxl/device: Add memory devices (8.2.8.5)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (6 preceding siblings ...)
  2021-01-05 16:52 ` [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4) Ben Widawsky
@ 2021-01-05 16:52 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 09/32] hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1) Ben Widawsky
                   ` (24 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:52 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Memory devices implement extra capabilities on top of CXL devices. This
adds support for that.

A large part of memory devices is the mailbox/command interface. All of
the mailbox handling is done in the mailbox-utils library. Longer term,
new CXL devices that are being emulated may want to handle commands
differently, and therefore would need a mechanism to opt in/out of the
specific generic handlers. As such, this is considered sufficient for
now, but may need more depth in the future.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-device-utils.c   | 52 ++++++++++++++++++++++++++++++++++++-
 hw/cxl/cxl-mailbox-utils.c  | 24 +++++++++++++++++
 include/hw/cxl/cxl_device.h | 16 ++++++++++++
 3 files changed, 91 insertions(+), 1 deletion(-)

diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
index 642e3c2617..c515d45d20 100644
--- a/hw/cxl/cxl-device-utils.c
+++ b/hw/cxl/cxl-device-utils.c
@@ -132,6 +132,45 @@ static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
         cxl_process_mailbox(cxl_dstate);
 }
 
+static uint64_t mdev_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    uint64_t retval = 0;
+
+    retval = FIELD_DP64(retval, CXL_MEM_DEV_STS, MEDIA_STATUS, 1);
+    retval = FIELD_DP64(retval, CXL_MEM_DEV_STS, MBOX_READY, 1);
+
+    switch (size) {
+    case 4:
+        if (unlikely(offset & (sizeof(uint32_t) - 1))) {
+            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+            return 0;
+        }
+        break;
+    case 8:
+        if (unlikely(offset & (sizeof(uint64_t) - 1))) {
+            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
+            return 0;
+        }
+        break;
+    }
+
+    return ldn_le_p(&retval, size);
+}
+
+static const MemoryRegionOps mdev_ops = {
+    .read = mdev_reg_read,
+    .write = NULL,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
 static const MemoryRegionOps mailbox_ops = {
     .read = mailbox_reg_read,
     .write = mailbox_reg_write,
@@ -187,6 +226,9 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
                           "device-status", CXL_DEVICE_REGISTERS_LENGTH);
     memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
                           "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
+    memory_region_init_io(&cxl_dstate->memory_device, obj, &mdev_ops,
+                          cxl_dstate, "memory device caps",
+                          CXL_MEMORY_DEVICE_REGISTERS_LENGTH);
 
     memory_region_add_subregion(&cxl_dstate->device_registers, 0,
                                 &cxl_dstate->caps);
@@ -196,6 +238,9 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
     memory_region_add_subregion(&cxl_dstate->device_registers,
                                 CXL_MAILBOX_REGISTERS_OFFSET,
                                 &cxl_dstate->mailbox);
+    memory_region_add_subregion(&cxl_dstate->device_registers,
+                                CXL_MEMORY_DEVICE_REGISTERS_OFFSET,
+                                &cxl_dstate->memory_device);
 }
 
 static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
@@ -208,10 +253,12 @@ static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
     cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
 }
 
+static void memdev_reg_init_common(CXLDeviceState *cxl_dstate) { }
+
 void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 {
     uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
-    const int cap_count = 2;
+    const int cap_count = 3;
 
     /* CXL Device Capabilities Array Register */
     ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
@@ -224,5 +271,8 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
     cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
     mailbox_reg_init_common(cxl_dstate);
 
+    cxl_device_cap_init(cxl_dstate, MEMORY_DEVICE, 0x4000);
+    memdev_reg_init_common(cxl_dstate);
+
     assert(cxl_initialize_mailbox(cxl_dstate) == 0);
 }
diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
index 3a39c936de..0e94f5a95f 100644
--- a/hw/cxl/cxl-mailbox-utils.c
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -81,6 +81,30 @@ struct cxl_cmd {
                                CXLDeviceState *cxl_dstate, uint16_t *len)
 #define declare_mailbox_handler(name) define_mailbox_handler(name)
 
+#define define_mailbox_handler_zeroed(name, size)                         \
+    uint16_t __zero##name = size;                                         \
+    static ret_code cmd_##name(struct cxl_cmd *cmd,                       \
+                               CXLDeviceState *cxl_dstate, uint16_t *len) \
+    {                                                                     \
+        *len = __zero##name;                                              \
+        memset(cmd->payload, 0, *len);                                    \
+        return CXL_MBOX_SUCCESS;                                          \
+    }
+#define define_mailbox_handler_const(name, data)                          \
+    static ret_code cmd_##name(struct cxl_cmd *cmd,                       \
+                               CXLDeviceState *cxl_dstate, uint16_t *len) \
+    {                                                                     \
+        *len = sizeof(data);                                              \
+        memcpy(cmd->payload, data, *len);                                 \
+        return CXL_MBOX_SUCCESS;                                          \
+    }
+#define define_mailbox_handler_nop(name)                                  \
+    static ret_code cmd_##name(struct cxl_cmd *cmd,                       \
+                               CXLDeviceState *cxl_dstate, uint16_t *len) \
+    {                                                                     \
+        return CXL_MBOX_SUCCESS;                                          \
+    }
+
 #define CXL_CMD(s, c, in, cel_effect) \
     [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
 
diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
index bdf990cec2..dd3f4572aa 100644
--- a/include/hw/cxl/cxl_device.h
+++ b/include/hw/cxl/cxl_device.h
@@ -69,11 +69,16 @@
 #define CXL_MAILBOX_REGISTERS_LENGTH \
     (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
 
+#define CXL_MEMORY_DEVICE_REGISTERS_OFFSET \
+    (CXL_MAILBOX_REGISTERS_OFFSET + CXL_MAILBOX_REGISTERS_LENGTH)
+#define CXL_MEMORY_DEVICE_REGISTERS_LENGTH 0x8
+
 typedef struct cxl_device_state {
     /* Main register container */
     MemoryRegion device_registers;
 
     MemoryRegion device;
+    MemoryRegion memory_device;
     struct {
         MemoryRegion caps;
         union {
@@ -133,6 +138,9 @@ REG32(CXL_DEV_CAP_ARRAY2, 4) /* We're going to pretend it's 64b */
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
                                                CXL_DEVICE_CAP_REG_SIZE)
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MEMORY_DEVICE,
+                                      CXL_DEVICE_CAP_HDR1_OFFSET +
+                                          CXL_DEVICE_CAP_REG_SIZE * 2)
 
 int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
@@ -185,4 +193,12 @@ REG32(CXL_DEV_BG_CMD_STS, 0x18)
 
 REG32(CXL_DEV_CMD_PAYLOAD, 0x20)
 
+/* XXX: actually a 64b registers */
+REG32(CXL_MEM_DEV_STS, 0)
+    FIELD(CXL_MEM_DEV_STS, FATAL, 0, 1)
+    FIELD(CXL_MEM_DEV_STS, FW_HALT, 1, 1)
+    FIELD(CXL_MEM_DEV_STS, MEDIA_STATUS, 2, 2)
+    FIELD(CXL_MEM_DEV_STS, MBOX_READY, 4, 1)
+    FIELD(CXL_MEM_DEV_STS, RESET_NEEDED, 5, 3)
+
 #endif
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 09/32] hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (7 preceding siblings ...)
  2021-01-05 16:52 ` [RFC PATCH v2 08/32] hw/cxl/device: Add memory devices (8.2.8.5) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 10/32] hw/cxl/device: Placeholder for firmware commands Ben Widawsky
                   ` (23 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Using the previously implemented stubbed helpers, it is now possible to
easily add the missing, required commands to the implementation.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-mailbox-utils.c | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)

diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
index 0e94f5a95f..890e486ea1 100644
--- a/hw/cxl/cxl-mailbox-utils.c
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -37,6 +37,14 @@
  *  a register interface that already deals with it.
  */
 
+enum {
+    EVENTS      = 0x01,
+        #define GET_RECORDS   0x0
+        #define CLEAR_RECORDS   0x1
+        #define GET_INTERRUPT_POLICY   0x2
+        #define SET_INTERRUPT_POLICY   0x3
+};
+
 /* 8.2.8.4.5.1 Command Return Codes */
 typedef enum {
     CXL_MBOX_SUCCESS = 0x0,
@@ -105,10 +113,23 @@ struct cxl_cmd {
         return CXL_MBOX_SUCCESS;                                          \
     }
 
+define_mailbox_handler_zeroed(EVENTS_GET_RECORDS, 0x20);
+define_mailbox_handler_nop(EVENTS_CLEAR_RECORDS);
+define_mailbox_handler_zeroed(EVENTS_GET_INTERRUPT_POLICY, 4);
+define_mailbox_handler_nop(EVENTS_SET_INTERRUPT_POLICY);
+
+#define IMMEDIATE_CONFIG_CHANGE (1 << 1)
+#define IMMEDIATE_LOG_CHANGE (1 << 4)
+
 #define CXL_CMD(s, c, in, cel_effect) \
     [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
 
-static struct cxl_cmd cxl_cmd_set[256][256] = {};
+static struct cxl_cmd cxl_cmd_set[256][256] = {
+    CXL_CMD(EVENTS, GET_RECORDS, 1, 0),
+    CXL_CMD(EVENTS, CLEAR_RECORDS, ~0, IMMEDIATE_LOG_CHANGE),
+    CXL_CMD(EVENTS, GET_INTERRUPT_POLICY, 0, 0),
+    CXL_CMD(EVENTS, SET_INTERRUPT_POLICY, 4, IMMEDIATE_CONFIG_CHANGE),
+};
 
 #undef CXL_CMD
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 10/32] hw/cxl/device: Placeholder for firmware commands
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (8 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 09/32] hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 11/32] hw/cxl/device: Timestamp implementation (8.2.9.3) Ben Widawsky
                   ` (22 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams




^ permalink raw reply	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 11/32] hw/cxl/device: Timestamp implementation (8.2.9.3)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (9 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 10/32] hw/cxl/device: Placeholder for firmware commands Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 17:12   ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 12/32] hw/cxl/device: Add log commands (8.2.9.4) + CEL Ben Widawsky
                   ` (21 subsequent siblings)
  32 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Per spec, timestamp appears to be a free-running counter from a value
set by the host via the Set Timestamp command (0301h). There are
references to the epoch, which seem like a red herring. Therefore, the
implementation implements the timestamp as freerunning counter from the
last value that was issued by the Set Timestamp command.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-mailbox-utils.c  | 52 +++++++++++++++++++++++++++++++++++++
 include/hw/cxl/cxl_device.h |  6 +++++
 2 files changed, 58 insertions(+)

diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
index 890e486ea1..8b956d9e94 100644
--- a/hw/cxl/cxl-mailbox-utils.c
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -43,6 +43,9 @@ enum {
         #define CLEAR_RECORDS   0x1
         #define GET_INTERRUPT_POLICY   0x2
         #define SET_INTERRUPT_POLICY   0x3
+    TIMESTAMP   = 0x03,
+        #define GET           0x0
+        #define SET           0x1
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -117,8 +120,11 @@ define_mailbox_handler_zeroed(EVENTS_GET_RECORDS, 0x20);
 define_mailbox_handler_nop(EVENTS_CLEAR_RECORDS);
 define_mailbox_handler_zeroed(EVENTS_GET_INTERRUPT_POLICY, 4);
 define_mailbox_handler_nop(EVENTS_SET_INTERRUPT_POLICY);
+declare_mailbox_handler(TIMESTAMP_GET);
+declare_mailbox_handler(TIMESTAMP_SET);
 
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
+#define IMMEDIATE_POLICY_CHANGE (1 << 3)
 #define IMMEDIATE_LOG_CHANGE (1 << 4)
 
 #define CXL_CMD(s, c, in, cel_effect) \
@@ -129,10 +135,56 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
     CXL_CMD(EVENTS, CLEAR_RECORDS, ~0, IMMEDIATE_LOG_CHANGE),
     CXL_CMD(EVENTS, GET_INTERRUPT_POLICY, 0, 0),
     CXL_CMD(EVENTS, SET_INTERRUPT_POLICY, 4, IMMEDIATE_CONFIG_CHANGE),
+    CXL_CMD(TIMESTAMP, GET, 0, 0),
+    CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
 };
 
 #undef CXL_CMD
 
+/*
+ * 8.2.9.3.1
+ */
+define_mailbox_handler(TIMESTAMP_GET)
+{
+    struct timespec ts;
+
+    clock_gettime(CLOCK_REALTIME, &ts);
+
+    cxl_dstate->timestamp.set = true;
+    cxl_dstate->timestamp.last_set =
+        ts.tv_sec * NANOSECONDS_PER_SECOND + ts.tv_nsec;
+    cxl_dstate->timestamp.host_set = *(uint64_t *)cmd->payload;
+
+    *len = 8;
+    return CXL_MBOX_SUCCESS;
+}
+
+/*
+ * 8.2.9.3.2
+ */
+define_mailbox_handler(TIMESTAMP_SET)
+{
+    struct timespec ts;
+    uint64_t delta;
+
+    if (!cxl_dstate->timestamp.set) {
+        *(uint64_t *)cmd->payload = 0;
+        goto done;
+    }
+
+    /* First find the delta from the last time the host set the time. */
+    clock_gettime(CLOCK_REALTIME, &ts);
+    delta = (ts.tv_sec * NANOSECONDS_PER_SECOND + ts.tv_nsec) -
+            cxl_dstate->timestamp.host_set;
+
+    /* Then adjust the actual time */
+    *(uint64_t *)cmd->payload = cxl_dstate->timestamp.host_set + delta;
+
+done:
+    *len = 8;
+    return CXL_MBOX_SUCCESS;
+}
+
 QemuUUID cel_uuid;
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
index dd3f4572aa..25974b2416 100644
--- a/include/hw/cxl/cxl_device.h
+++ b/include/hw/cxl/cxl_device.h
@@ -100,6 +100,12 @@ typedef struct cxl_device_state {
         size_t cel_size;
     };
 
+    struct {
+        bool set;
+        uint64_t last_set;
+        uint64_t host_set;
+    } timestamp;
+
     MemoryRegion *pmem;
     MemoryRegion *vmem;
 } CXLDeviceState;
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 12/32] hw/cxl/device: Add log commands (8.2.9.4) + CEL
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (10 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 11/32] hw/cxl/device: Timestamp implementation (8.2.9.3) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 13/32] hw/pxb: Use a type for realizing expanders Ben Widawsky
                   ` (20 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

CXL specification provides for the ability to obtain logs from the
device. Logs are either spec defined, like the "Command Effects Log"
(CEL), or vendor specific. UUIDs are defined for all log types.

The CEL is a mechanism to provide information to the host about which
commands are supported. It is useful both to determine which spec'd
optional commands are supported, as well as provide a list of vendor
specified commands that might be used. The CEL is already created as
part of mailbox initialization, but here it is now exported to hosts
that use these log commands.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/cxl/cxl-mailbox-utils.c | 75 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 75 insertions(+)

diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
index 8b956d9e94..f68ec5b5b9 100644
--- a/hw/cxl/cxl-mailbox-utils.c
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -46,6 +46,9 @@ enum {
     TIMESTAMP   = 0x03,
         #define GET           0x0
         #define SET           0x1
+    LOGS        = 0x04,
+        #define GET_SUPPORTED 0x0
+        #define GET_LOG       0x1
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -122,6 +125,8 @@ define_mailbox_handler_zeroed(EVENTS_GET_INTERRUPT_POLICY, 4);
 define_mailbox_handler_nop(EVENTS_SET_INTERRUPT_POLICY);
 declare_mailbox_handler(TIMESTAMP_GET);
 declare_mailbox_handler(TIMESTAMP_SET);
+declare_mailbox_handler(LOGS_GET_SUPPORTED);
+declare_mailbox_handler(LOGS_GET_LOG);
 
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
 #define IMMEDIATE_POLICY_CHANGE (1 << 3)
@@ -137,6 +142,8 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
     CXL_CMD(EVENTS, SET_INTERRUPT_POLICY, 4, IMMEDIATE_CONFIG_CHANGE),
     CXL_CMD(TIMESTAMP, GET, 0, 0),
     CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
+    CXL_CMD(LOGS, GET_SUPPORTED, 0, 0),
+    CXL_CMD(LOGS, GET_LOG, 0x18, 0),
 };
 
 #undef CXL_CMD
@@ -187,6 +194,74 @@ done:
 
 QemuUUID cel_uuid;
 
+/* 8.2.9.4.1 */
+define_mailbox_handler(LOGS_GET_SUPPORTED)
+{
+    struct {
+        uint16_t entries;
+        uint8_t rsvd[6];
+        struct {
+            QemuUUID uuid;
+            uint32_t size;
+        } log_entries[1];
+    } __attribute__((packed)) *supported_logs;
+    _Static_assert(sizeof(*supported_logs) == 0x1c, "Bad supported log size");
+
+    supported_logs = (void *)cmd->payload;
+
+    supported_logs->entries = 1;
+    supported_logs->log_entries[0].uuid = cel_uuid;
+    supported_logs->log_entries[0].size = 4 * cxl_dstate->cel_size;
+
+    *len = sizeof(*supported_logs);
+    return CXL_MBOX_SUCCESS;
+}
+
+/* 8.2.9.4.2 */
+define_mailbox_handler(LOGS_GET_LOG)
+{
+    struct {
+        union {
+            struct {
+                QemuUUID uuid;
+                uint32_t offset;
+                uint32_t length;
+            };
+        };
+    } __attribute__((packed, __aligned__(16))) *get_log;
+
+    get_log = (void *)cmd->payload;
+
+    /*
+     * 8.2.9.4.2
+     *   The device shall return Invalid Parameter if the Offset or Length
+     *   fields attempt to access beyond the size of the log as reported by Get
+     *   Supported Logs.
+     *
+     * XXX: Spec is wrong, "Invalid Parameter" isn't a thing.
+     * XXX: Spec doesn't address incorrect UUID incorrectness.
+     *
+     * The CEL buffer is large enough to fit all commands in the emulation, so
+     * the only possible failure would be if the mailbox itself isn't big
+     * enough.
+     */
+    if (get_log->offset + get_log->length > cxl_dstate->payload_size) {
+        return CXL_MBOX_INVALID_INPUT;
+    }
+
+    if (!qemu_uuid_is_equal(&get_log->uuid, &cel_uuid)) {
+        return CXL_MBOX_UNSUPPORTED;
+    }
+
+    /* Store off everything to local variables so we can wipe out the payload */
+    *len = get_log->length;
+
+    memmove(cmd->payload, cxl_dstate->cel_log + get_log->offset,
+           get_log->length);
+
+    return CXL_MBOX_SUCCESS;
+}
+
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
 {
     uint16_t ret = CXL_MBOX_SUCCESS;
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 13/32] hw/pxb: Use a type for realizing expanders
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (11 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 12/32] hw/cxl/device: Add log commands (8.2.9.4) + CEL Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 14/32] hw/pci/cxl: Create a CXL bus type Ben Widawsky
                   ` (19 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This opens up the possibility for more types of expanders (other than
PCI and PCIe). We'll need this to create a CXL expander.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/pci-bridge/pci_expander_bridge.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index aedded1064..232b7ce305 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -24,6 +24,8 @@
 #include "hw/boards.h"
 #include "qom/object.h"
 
+enum BusType { PCI, PCIE };
+
 #define TYPE_PXB_BUS "pxb-bus"
 typedef struct PXBBus PXBBus;
 DECLARE_INSTANCE_CHECKER(PXBBus, PXB_BUS,
@@ -214,7 +216,8 @@ static gint pxb_compare(gconstpointer a, gconstpointer b)
            0;
 }
 
-static void pxb_dev_realize_common(PCIDevice *dev, bool pcie, Error **errp)
+static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
+                                   Error **errp)
 {
     PXBDev *pxb = convert_to_pxb(dev);
     DeviceState *ds, *bds = NULL;
@@ -239,7 +242,7 @@ static void pxb_dev_realize_common(PCIDevice *dev, bool pcie, Error **errp)
     }
 
     ds = qdev_new(TYPE_PXB_HOST);
-    if (pcie) {
+    if (type == PCIE) {
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
     } else {
         bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
@@ -287,7 +290,7 @@ static void pxb_dev_realize(PCIDevice *dev, Error **errp)
         return;
     }
 
-    pxb_dev_realize_common(dev, false, errp);
+    pxb_dev_realize_common(dev, PCI, errp);
 }
 
 static void pxb_dev_exitfn(PCIDevice *pci_dev)
@@ -339,7 +342,7 @@ static void pxb_pcie_dev_realize(PCIDevice *dev, Error **errp)
         return;
     }
 
-    pxb_dev_realize_common(dev, true, errp);
+    pxb_dev_realize_common(dev, PCIE, errp);
 }
 
 static void pxb_pcie_dev_class_init(ObjectClass *klass, void *data)
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 14/32] hw/pci/cxl: Create a CXL bus type
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (12 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 13/32] hw/pxb: Use a type for realizing expanders Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 15/32] hw/pxb: Allow creation of a CXL PXB (host bridge) Ben Widawsky
                   ` (18 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

The easiest way to differentiate a CXL bus, and a PCIE bus is using a
flag. A CXL bus, in hardware, is backward compatible with PCIE, and
therefore the code tries pretty hard to keep them in sync as much as
possible.

The other way to implement this would be to try to cast the bus to the
correct type. This is less code and useful for debugging via simply
looking at the flags.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/pci-bridge/pci_expander_bridge.c | 9 ++++++++-
 include/hw/pci/pci_bus.h            | 7 +++++++
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index 232b7ce305..88c45dc3b5 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -24,7 +24,7 @@
 #include "hw/boards.h"
 #include "qom/object.h"
 
-enum BusType { PCI, PCIE };
+enum BusType { PCI, PCIE, CXL };
 
 #define TYPE_PXB_BUS "pxb-bus"
 typedef struct PXBBus PXBBus;
@@ -35,6 +35,10 @@ DECLARE_INSTANCE_CHECKER(PXBBus, PXB_BUS,
 DECLARE_INSTANCE_CHECKER(PXBBus, PXB_PCIE_BUS,
                          TYPE_PXB_PCIE_BUS)
 
+#define TYPE_PXB_CXL_BUS "pxb-cxl-bus"
+DECLARE_INSTANCE_CHECKER(PXBBus, PXB_CXL_BUS,
+                         TYPE_PXB_CXL_BUS)
+
 struct PXBBus {
     /*< private >*/
     PCIBus parent_obj;
@@ -244,6 +248,9 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
     ds = qdev_new(TYPE_PXB_HOST);
     if (type == PCIE) {
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
+    } else if (type == CXL) {
+        bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_CXL_BUS);
+        bus->flags |= PCI_BUS_CXL;
     } else {
         bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
         bds = qdev_new("pci-bridge");
diff --git a/include/hw/pci/pci_bus.h b/include/hw/pci/pci_bus.h
index 347440d42c..eb94e7e85c 100644
--- a/include/hw/pci/pci_bus.h
+++ b/include/hw/pci/pci_bus.h
@@ -24,6 +24,8 @@ enum PCIBusFlags {
     PCI_BUS_IS_ROOT                                         = 0x0001,
     /* PCIe extended configuration space is accessible on this bus */
     PCI_BUS_EXTENDED_CONFIG_SPACE                           = 0x0002,
+    /* This is a CXL Type BUS */
+    PCI_BUS_CXL                                             = 0x0004,
 };
 
 struct PCIBus {
@@ -53,6 +55,11 @@ struct PCIBus {
     Notifier machine_done;
 };
 
+static inline bool pci_bus_is_cxl(PCIBus *bus)
+{
+    return !!(bus->flags & PCI_BUS_CXL);
+}
+
 static inline bool pci_bus_is_root(PCIBus *bus)
 {
     return !!(bus->flags & PCI_BUS_IS_ROOT);
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 15/32] hw/pxb: Allow creation of a CXL PXB (host bridge)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (13 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 14/32] hw/pci/cxl: Create a CXL bus type Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 16/32] qtest: allow DSDT acpi table changes Ben Widawsky
                   ` (17 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This works like adding a typical pxb device, except the name is
'pxb-cxl' instead of 'pxb-pcie'. An example command line would be as
follows:
  -device pxb-cxl,id=cxl.0,bus="pcie.0",bus_nr=1

A CXL PXB is backward compatible with PCIe. What this means in practice
is that an operating system that is unaware of CXL should still be able
to enumerate this topology as if it were PCIe.

One can create multiple CXL PXB host bridges, but a host bridge can only
be connected to the main root bus. Host bridges cannot appear elsewhere
in the topology.

Note that as of this patch, the ACPI tables needed for the host bridge
(specifically, an ACPI object in _SB named ACPI0016 and the CEDT) aren't
created. So while this patch internally creates it, it cannot be
properly used by an operating system or other system software.

Upcoming patches will allow creating multiple host bridges.

v2: Remove vendor and device ID (Ben)

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/pci-bridge/pci_expander_bridge.c | 67 ++++++++++++++++++++++++++++-
 hw/pci/pci.c                        |  7 +++
 include/hw/pci/pci.h                |  6 +++
 3 files changed, 78 insertions(+), 2 deletions(-)

diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index 88c45dc3b5..b42592e1ff 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -56,6 +56,10 @@ DECLARE_INSTANCE_CHECKER(PXBDev, PXB_DEV,
 DECLARE_INSTANCE_CHECKER(PXBDev, PXB_PCIE_DEV,
                          TYPE_PXB_PCIE_DEVICE)
 
+#define TYPE_PXB_CXL_DEVICE "pxb-cxl"
+DECLARE_INSTANCE_CHECKER(PXBDev, PXB_CXL_DEV,
+                         TYPE_PXB_CXL_DEVICE)
+
 struct PXBDev {
     /*< private >*/
     PCIDevice parent_obj;
@@ -67,6 +71,11 @@ struct PXBDev {
 
 static PXBDev *convert_to_pxb(PCIDevice *dev)
 {
+    /* A CXL PXB's parent bus is PCIe, so the normal check won't work */
+    if (object_dynamic_cast(OBJECT(dev), TYPE_PXB_CXL_DEVICE)) {
+        return PXB_CXL_DEV(dev);
+    }
+
     return pci_bus_is_express(pci_get_bus(dev))
         ? PXB_PCIE_DEV(dev) : PXB_DEV(dev);
 }
@@ -111,11 +120,20 @@ static const TypeInfo pxb_pcie_bus_info = {
     .class_init    = pxb_bus_class_init,
 };
 
+static const TypeInfo pxb_cxl_bus_info = {
+    .name          = TYPE_PXB_CXL_BUS,
+    .parent        = TYPE_CXL_BUS,
+    .instance_size = sizeof(PXBBus),
+    .class_init    = pxb_bus_class_init,
+};
+
 static const char *pxb_host_root_bus_path(PCIHostState *host_bridge,
                                           PCIBus *rootbus)
 {
-    PXBBus *bus = pci_bus_is_express(rootbus) ?
-                  PXB_PCIE_BUS(rootbus) : PXB_BUS(rootbus);
+    PXBBus *bus = pci_bus_is_cxl(rootbus) ?
+                      PXB_CXL_BUS(rootbus) :
+                      pci_bus_is_express(rootbus) ? PXB_PCIE_BUS(rootbus) :
+                                                    PXB_BUS(rootbus);
 
     snprintf(bus->bus_path, 8, "0000:%02x", pxb_bus_num(rootbus));
     return bus->bus_path;
@@ -380,13 +398,58 @@ static const TypeInfo pxb_pcie_dev_info = {
     },
 };
 
+static void pxb_cxl_dev_realize(PCIDevice *dev, Error **errp)
+{
+    /* A CXL PXB's parent bus is still PCIe */
+    if (!pci_bus_is_express(pci_get_bus(dev))) {
+        error_setg(errp, "pxb-cxl devices cannot reside on a PCI bus");
+        return;
+    }
+
+    pxb_dev_realize_common(dev, CXL, errp);
+}
+
+static void pxb_cxl_dev_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc   = DEVICE_CLASS(klass);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+
+    k->realize             = pxb_cxl_dev_realize;
+    k->exit                = pxb_dev_exitfn;
+    /*
+     * XXX: These types of bridges don't actually show up in the hierarchy so
+     * vendor, device, class, etc. ids are intentionally left out.
+     */
+
+    dc->desc = "CXL Host Bridge";
+    device_class_set_props(dc, pxb_dev_properties);
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+
+    /* Host bridges aren't hotpluggable. FIXME: spec reference */
+    dc->hotpluggable = false;
+}
+
+static const TypeInfo pxb_cxl_dev_info = {
+    .name          = TYPE_PXB_CXL_DEVICE,
+    .parent        = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(PXBDev),
+    .class_init    = pxb_cxl_dev_class_init,
+    .interfaces =
+        (InterfaceInfo[]){
+            { INTERFACE_CONVENTIONAL_PCI_DEVICE },
+            {},
+        },
+};
+
 static void pxb_register_types(void)
 {
     type_register_static(&pxb_bus_info);
     type_register_static(&pxb_pcie_bus_info);
+    type_register_static(&pxb_cxl_bus_info);
     type_register_static(&pxb_host_info);
     type_register_static(&pxb_dev_info);
     type_register_static(&pxb_pcie_dev_info);
+    type_register_static(&pxb_cxl_dev_info);
 }
 
 type_init(pxb_register_types)
diff --git a/hw/pci/pci.c b/hw/pci/pci.c
index 50fc61c81f..fb454e2317 100644
--- a/hw/pci/pci.c
+++ b/hw/pci/pci.c
@@ -222,6 +222,12 @@ static const TypeInfo pcie_bus_info = {
     .class_init = pcie_bus_class_init,
 };
 
+static const TypeInfo cxl_bus_info = {
+    .name       = TYPE_CXL_BUS,
+    .parent     = TYPE_PCIE_BUS,
+    .class_init = pcie_bus_class_init,
+};
+
 static PCIBus *pci_find_bus_nr(PCIBus *bus, int bus_num);
 static void pci_update_mappings(PCIDevice *d);
 static void pci_irq_handler(void *opaque, int irq_num, int level);
@@ -2849,6 +2855,7 @@ static void pci_register_types(void)
 {
     type_register_static(&pci_bus_info);
     type_register_static(&pcie_bus_info);
+    type_register_static(&cxl_bus_info);
     type_register_static(&conventional_pci_interface_info);
     type_register_static(&cxl_interface_info);
     type_register_static(&pcie_interface_info);
diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h
index f6b1eff446..c464597e53 100644
--- a/include/hw/pci/pci.h
+++ b/include/hw/pci/pci.h
@@ -405,6 +405,7 @@ typedef PCIINTxRoute (*pci_route_irq_fn)(void *opaque, int pin);
 #define TYPE_PCI_BUS "PCI"
 OBJECT_DECLARE_TYPE(PCIBus, PCIBusClass, PCI_BUS)
 #define TYPE_PCIE_BUS "PCIE"
+#define TYPE_CXL_BUS "CXL"
 
 bool pci_bus_is_express(PCIBus *bus);
 
@@ -753,6 +754,11 @@ static inline void pci_irq_pulse(PCIDevice *pci_dev)
     pci_irq_deassert(pci_dev);
 }
 
+static inline int pci_is_cxl(const PCIDevice *d)
+{
+    return d->cap_present & QEMU_PCIE_CAP_CXL;
+}
+
 static inline int pci_is_express(const PCIDevice *d)
 {
     return d->cap_present & QEMU_PCI_CAP_EXPRESS;
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 16/32] qtest: allow DSDT acpi table changes
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (14 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 15/32] hw/pxb: Allow creation of a CXL PXB (host bridge) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 17/32] acpi/pci: Consolidate host bridge setup Ben Widawsky
                   ` (16 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 tests/qtest/bios-tables-test-allowed-diff.h | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/tests/qtest/bios-tables-test-allowed-diff.h b/tests/qtest/bios-tables-test-allowed-diff.h
index dfb8523c8b..5c695cdf37 100644
--- a/tests/qtest/bios-tables-test-allowed-diff.h
+++ b/tests/qtest/bios-tables-test-allowed-diff.h
@@ -1 +1,22 @@
 /* List of comma-separated changed AML files to ignore */
+"tests/data/acpi/pc/DSDT",
+"tests/data/acpi/pc/DSDT.acpihmat",
+"tests/data/acpi/pc/DSDT.bridge",
+"tests/data/acpi/pc/DSDT.cphp",
+"tests/data/acpi/pc/DSDT.dimmpxm",
+"tests/data/acpi/pc/DSDT.hpbridge",
+"tests/data/acpi/pc/DSDT.hpbrroot",
+"tests/data/acpi/pc/DSDT.ipmikcs",
+"tests/data/acpi/pc/DSDT.memhp",
+"tests/data/acpi/pc/DSDT.numamem",
+"tests/data/acpi/pc/DSDT.roothp",
+"tests/data/acpi/q35/DSDT",
+"tests/data/acpi/q35/DSDT.acpihmat",
+"tests/data/acpi/q35/DSDT.bridge",
+"tests/data/acpi/q35/DSDT.cphp",
+"tests/data/acpi/q35/DSDT.dimmpxm",
+"tests/data/acpi/q35/DSDT.ipmibt",
+"tests/data/acpi/q35/DSDT.memhp",
+"tests/data/acpi/q35/DSDT.mmio64",
+"tests/data/acpi/q35/DSDT.numamem",
+"tests/data/acpi/q35/DSDT.tis",
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 17/32] acpi/pci: Consolidate host bridge setup
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (15 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 16/32] qtest: allow DSDT acpi table changes Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 18/32] tests/acpi: remove stale allowed tables Ben Widawsky
                   ` (15 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This cleanup will make it easier to add support for CXL to the mix.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/i386/acpi-build.c | 31 +++++++++++++++++--------------
 1 file changed, 17 insertions(+), 14 deletions(-)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index f18b71dea9..3c214bba67 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -1194,6 +1194,20 @@ static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
     aml_append(table, scope);
 }
 
+enum { PCI, PCIE };
+static void init_pci_acpi(Aml *dev, int uid, int type)
+{
+    if (type == PCI) {
+        aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
+        aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
+    } else {
+        aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
+        aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
+        aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
+        aml_append(dev, build_q35_osc_method());
+    }
+}
+
 static void
 build_dsdt(GArray *table_data, BIOSLinker *linker,
            AcpiPmInfo *pm, AcpiMiscInfo *misc,
@@ -1222,9 +1236,8 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
     if (misc->is_piix4) {
         sb_scope = aml_scope("_SB");
         dev = aml_device("PCI0");
-        aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
+        init_pci_acpi(dev, 0, PCI);
         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
-        aml_append(dev, aml_name_decl("_UID", aml_int(0)));
         aml_append(sb_scope, dev);
         aml_append(dsdt, sb_scope);
 
@@ -1238,11 +1251,8 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
     } else {
         sb_scope = aml_scope("_SB");
         dev = aml_device("PCI0");
-        aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
-        aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
+        init_pci_acpi(dev, 0, PCIE);
         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
-        aml_append(dev, aml_name_decl("_UID", aml_int(0)));
-        aml_append(dev, build_q35_osc_method());
         aml_append(sb_scope, dev);
 
         if (pm->smi_on_cpuhp) {
@@ -1345,15 +1355,8 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
 
             scope = aml_scope("\\_SB");
             dev = aml_device("PC%.02X", bus_num);
-            aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
-            if (pci_bus_is_express(bus)) {
-                aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
-                aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
-                aml_append(dev, build_q35_osc_method());
-            } else {
-                aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
-            }
+            init_pci_acpi(dev, bus_num, pci_bus_is_express(bus) ? PCIE : PCI);
 
             if (numa_node != NUMA_NODE_UNASSIGNED) {
                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 18/32] tests/acpi: remove stale allowed tables
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (16 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 17/32] acpi/pci: Consolidate host bridge setup Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 19/32] hw/pci: Plumb _UID through host bridges Ben Widawsky
                   ` (14 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

     Scope (_SB)
     {
         Device (PCI0)
         {
             Name (_HID, EisaId ("PNP0A03") /* PCI Bus */)  // _HID: Hardware ID
-            Name (_ADR, Zero)  // _ADR: Address
             Name (_UID, Zero)  // _UID: Unique ID
+            Name (_ADR, Zero)  // _ADR: Address

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 tests/data/acpi/pc/DSDT                     | Bin 5065 -> 5065 bytes
 tests/data/acpi/pc/DSDT.acpihmat            | Bin 6390 -> 6390 bytes
 tests/data/acpi/pc/DSDT.bridge              | Bin 6924 -> 6924 bytes
 tests/data/acpi/pc/DSDT.cphp                | Bin 5529 -> 5529 bytes
 tests/data/acpi/pc/DSDT.dimmpxm             | Bin 6719 -> 6719 bytes
 tests/data/acpi/pc/DSDT.hpbridge            | Bin 5026 -> 5026 bytes
 tests/data/acpi/pc/DSDT.hpbrroot            | Bin 3084 -> 3084 bytes
 tests/data/acpi/pc/DSDT.ipmikcs             | Bin 5137 -> 5137 bytes
 tests/data/acpi/pc/DSDT.memhp               | Bin 6424 -> 6424 bytes
 tests/data/acpi/pc/DSDT.numamem             | Bin 5071 -> 5071 bytes
 tests/data/acpi/pc/DSDT.roothp              | Bin 5261 -> 5261 bytes
 tests/data/acpi/q35/DSDT                    | Bin 7801 -> 7801 bytes
 tests/data/acpi/q35/DSDT.acpihmat           | Bin 9126 -> 9126 bytes
 tests/data/acpi/q35/DSDT.bridge             | Bin 7819 -> 7819 bytes
 tests/data/acpi/q35/DSDT.cphp               | Bin 8265 -> 8265 bytes
 tests/data/acpi/q35/DSDT.dimmpxm            | Bin 9455 -> 9455 bytes
 tests/data/acpi/q35/DSDT.ipmibt             | Bin 7876 -> 7876 bytes
 tests/data/acpi/q35/DSDT.memhp              | Bin 9160 -> 9160 bytes
 tests/data/acpi/q35/DSDT.mmio64             | Bin 8932 -> 8932 bytes
 tests/data/acpi/q35/DSDT.numamem            | Bin 7807 -> 7807 bytes
 tests/qtest/bios-tables-test-allowed-diff.h |  21 --------------------
 21 files changed, 21 deletions(-)

diff --git a/tests/data/acpi/pc/DSDT b/tests/data/acpi/pc/DSDT
index f6173df1d598767a79aa34ad7585ad7d45c5d4f3..b516745128e3f1a297b6327e9057026a2d16229c 100644
GIT binary patch
delta 20
bcmX@9eo}oxJ7=h;3j;^Iqf5}n36{bDOsEE~

delta 20
bcmX@9eo}oxJEx;d5CcbisHe-u36{bDOlAhI

diff --git a/tests/data/acpi/pc/DSDT.acpihmat b/tests/data/acpi/pc/DSDT.acpihmat
index 67f3f7249eaaa9404ebf0f2d0a324b8c8e3bd445..aeae285c6434ae6cf3c53660e34425727a497871 100644
GIT binary patch
delta 20
bcmexn_|0%aJ7=h;3j;^Iqf5}n3271lRUHRT

delta 20
bcmexn_|0%aJEx;d5CcbisHe-u3271lRNDtm

diff --git a/tests/data/acpi/pc/DSDT.bridge b/tests/data/acpi/pc/DSDT.bridge
index 643390f4c4138b37fc481656d3f555d0eeedcb02..4cd26a87dd11d96e10bf6de786b9d56ebfe0a4f9 100644
GIT binary patch
delta 20
bcmeA%>oJ?q&Kc_I!oU&l=n}MXLX8vvMneXi

delta 20
bcmeA%>oJ?q&gtk9#J~|B>glp^LX8vvMgaz#

diff --git a/tests/data/acpi/pc/DSDT.cphp b/tests/data/acpi/pc/DSDT.cphp
index 1ddcf7d8812f5d8d4d38fe7e7b35fd5885806046..fecb784812cbb2308ef58acf4a2c580f56d35c39 100644
GIT binary patch
delta 20
bcmbQKJyUx^J7=h;3j;^Iqf5}n37nz;MY;wk

delta 20
bcmbQKJyUx^JEx;d5CcbisHe-u37nz;MR*1%

diff --git a/tests/data/acpi/pc/DSDT.dimmpxm b/tests/data/acpi/pc/DSDT.dimmpxm
index c44385cc01879324738ffb7f997b8cdd762cbf97..f2c31e150ead16e4931367a6dab42704950a21e9 100644
GIT binary patch
delta 20
bcmdmQvfpGvJ7=h;3j;^Iqf5}n3F{>RP4WjY

delta 20
bcmdmQvfpGvJEx;d5CcbisHe-u3F{>RO|S<r

diff --git a/tests/data/acpi/pc/DSDT.hpbridge b/tests/data/acpi/pc/DSDT.hpbridge
index 4ecf1eb13bf49499f729b53a6d0114672a76e28d..7a8955cdbc52c025a2fd8f160cf8aff9442c985b 100644
GIT binary patch
delta 20
bcmZ3azDRvSJ7=h;3j;^Iqf5}n2|~gEMvw+M

delta 20
bcmZ3azDRvSJEx;d5CcbisHe-u2|~gEMotDf

diff --git a/tests/data/acpi/pc/DSDT.hpbrroot b/tests/data/acpi/pc/DSDT.hpbrroot
index a3046226ec1dcb234b726029b3790dfedb3b9221..88d23fca4743c2ee57493e7d77d6297a60964d3c 100644
GIT binary patch
delta 20
bcmeB?=#iMv&Kc_I!oU&l=n}MXLJc<nLHq_$

delta 20
bcmeB?=#iMv&gtk9#J~|B>glp^LJc<nLAnM}

diff --git a/tests/data/acpi/pc/DSDT.ipmikcs b/tests/data/acpi/pc/DSDT.ipmikcs
index f1638c5d079a9442c09390426a913010df6efd8d..d670ae793b5778c095a7f8c79ff1a046889d1a56 100644
GIT binary patch
delta 20
bcmbQJF;QbeJ7=h;3j;^Iqf5}n35~)4MGOXr

delta 20
bcmbQJF;QbeJEx;d5CcbisHe-u35~)4M9Kz;

diff --git a/tests/data/acpi/pc/DSDT.memhp b/tests/data/acpi/pc/DSDT.memhp
index 4c19e45e66918c61674785c99e4474e58866f125..a7de3d9fd94e62e8fc357fe3093bf7f394a39219 100644
GIT binary patch
delta 20
bcmbPXG{a~@J7=h;3j;^Iqf5}n2^|suN0A1$

delta 20
bcmbPXG{a~@JEx;d5CcbisHe-u2^|suM^6T}

diff --git a/tests/data/acpi/pc/DSDT.numamem b/tests/data/acpi/pc/DSDT.numamem
index 40cfd933259af05ac2aee07fca32f22122255211..57958b6cec216c1fb8731f4ed2da67f0fad7484a 100644
GIT binary patch
delta 20
bcmX@FeqMb-J7=h;3j;^Iqf5}n3HHJOO_&D2

delta 20
bcmX@FeqMb-JEx;d5CcbisHe-u3HHJOO;!fL

diff --git a/tests/data/acpi/pc/DSDT.roothp b/tests/data/acpi/pc/DSDT.roothp
index 078fc8031b479cc77b6527a2b7b4bd576b6e6028..624d0e367693fe267a4237a5fc97295cee2ebd60 100644
GIT binary patch
delta 20
bcmeCx?A4sm&Kc_I!oU&l=n}MX!e3zkMUV#m

delta 20
bcmeCx?A4sm&gtk9#J~|B>glp^!e3zkMNS6(

diff --git a/tests/data/acpi/q35/DSDT b/tests/data/acpi/q35/DSDT
index d25cd7072932886d6967f4023faac1e1fa6e836c..17e2aebde98e0a3161d93e9b2e200737b13699ac 100644
GIT binary patch
delta 21
dcmexq^V4R+<cTvI**M}IU4j@kOEJdF0sv{z2gd*a

delta 19
bcmexq^V4R+WEMx4Aclz(n>R}_#>)Z#RX+z<

diff --git a/tests/data/acpi/q35/DSDT.acpihmat b/tests/data/acpi/q35/DSDT.acpihmat
index 722e06af83abcde203a2b96a8ec81fd3bab9fc98..7b3d659352a0923822f6a5db1dbd0a6ad853c446 100644
GIT binary patch
delta 21
dcmZ4HzRZ2X<cTvI**M}IU4j@kOELB+0RUdw2WbER

delta 19
bcmZ4HzRZ2XWEMx4Aclz(n>R}__9y`WOK1lA

diff --git a/tests/data/acpi/q35/DSDT.bridge b/tests/data/acpi/q35/DSDT.bridge
index 06bac139d668ddfc7914e258b471a303c9dbd192..5961b55b1067c3090b2f1f4cd3386d71efee241d 100644
GIT binary patch
delta 21
ccmeCS?Y5mTdE(4QHja2lmmr4CQjCSN09fk={{R30

delta 19
acmeCS?Y5mTnZ?m1h+*Qy=FL)!g|Yxf4F-?^

diff --git a/tests/data/acpi/q35/DSDT.cphp b/tests/data/acpi/q35/DSDT.cphp
index 2b933ac482e6883efccbd7d6c96089602f2c0b4d..09c92d52f92bb346ed807945b9638cad958446f8 100644
GIT binary patch
delta 21
dcmX@<aMEGI<cTvI**M}IU4j@kOEK!p0{~)+2SES;

delta 19
bcmX@<aMEGIWEMx4Aclz(n>R}_>dONFPN@dc

diff --git a/tests/data/acpi/q35/DSDT.dimmpxm b/tests/data/acpi/q35/DSDT.dimmpxm
index bd8f8305b028ef20f9b6d1a0c69ac428d027e3d1..1da97afb32dddafefe7f27934acbcb7d56a67489 100644
GIT binary patch
delta 21
dcmaFw`QCHF<cTvI**M}IU4j@kOEF$m1^{az2uT0{

delta 19
bcmaFw`QCHFWEMx4Aclz(n>R}_UR4GFR)YuH

diff --git a/tests/data/acpi/q35/DSDT.ipmibt b/tests/data/acpi/q35/DSDT.ipmibt
index a8f868e23c25688ab1c0371016c071f23e9d732f..c7e68432b66e7b4d03284c882c65bbf3066825dc 100644
GIT binary patch
delta 21
dcmX?Nd&G9a<cTvI**M}IU4j@kOEIpJ1ps122dV%7

delta 19
bcmX?Nd&G9aWEMx4Aclz(n>R}_u95`+PJ;(K

diff --git a/tests/data/acpi/q35/DSDT.memhp b/tests/data/acpi/q35/DSDT.memhp
index 9a802e4c67022386442976d5cb997ea3fc57b58f..3af457dd550461b2d2ea85aa85d7740452913b34 100644
GIT binary patch
delta 21
dcmX@%e!_ji<cTvI**M}IU4j@kOEIof0sv%g2hRWi

delta 19
bcmX@%e!_jiWEMx4Aclz(n>R}_u2TX4P;>`i

diff --git a/tests/data/acpi/q35/DSDT.mmio64 b/tests/data/acpi/q35/DSDT.mmio64
index 948c2dc7264c31932b490ca00691a7c4d9aefdb0..a4d20f676ac173e6846dcd4e076220d512215963 100644
GIT binary patch
delta 21
dcmaFj`owj@<cTvI**M}IU4j@kOEI2O1ORBc2p#|c

delta 19
bcmaFj`owj@WEMx4Aclz(n>R}_o>Bw=R96SD

diff --git a/tests/data/acpi/q35/DSDT.numamem b/tests/data/acpi/q35/DSDT.numamem
index 44ec1b0af400da6d298284aa959aa38add7e6dd5..bbab0d10a2a064528519fa69e90c799430129b75 100644
GIT binary patch
delta 21
dcmexw^WSE|<cTvI**M}IU4j@kOEIR(0sv~w2iX7s

delta 19
bcmexw^WSE|WEMx4Aclz(n>R}_rpf{URwD;$

diff --git a/tests/qtest/bios-tables-test-allowed-diff.h b/tests/qtest/bios-tables-test-allowed-diff.h
index 5c695cdf37..dfb8523c8b 100644
--- a/tests/qtest/bios-tables-test-allowed-diff.h
+++ b/tests/qtest/bios-tables-test-allowed-diff.h
@@ -1,22 +1 @@
 /* List of comma-separated changed AML files to ignore */
-"tests/data/acpi/pc/DSDT",
-"tests/data/acpi/pc/DSDT.acpihmat",
-"tests/data/acpi/pc/DSDT.bridge",
-"tests/data/acpi/pc/DSDT.cphp",
-"tests/data/acpi/pc/DSDT.dimmpxm",
-"tests/data/acpi/pc/DSDT.hpbridge",
-"tests/data/acpi/pc/DSDT.hpbrroot",
-"tests/data/acpi/pc/DSDT.ipmikcs",
-"tests/data/acpi/pc/DSDT.memhp",
-"tests/data/acpi/pc/DSDT.numamem",
-"tests/data/acpi/pc/DSDT.roothp",
-"tests/data/acpi/q35/DSDT",
-"tests/data/acpi/q35/DSDT.acpihmat",
-"tests/data/acpi/q35/DSDT.bridge",
-"tests/data/acpi/q35/DSDT.cphp",
-"tests/data/acpi/q35/DSDT.dimmpxm",
-"tests/data/acpi/q35/DSDT.ipmibt",
-"tests/data/acpi/q35/DSDT.memhp",
-"tests/data/acpi/q35/DSDT.mmio64",
-"tests/data/acpi/q35/DSDT.numamem",
-"tests/data/acpi/q35/DSDT.tis",
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 19/32] hw/pci: Plumb _UID through host bridges
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (17 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 18/32] tests/acpi: remove stale allowed tables Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 20/32] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142) Ben Widawsky
                   ` (13 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Currently, QEMU makes _UID equivalent to the bus number (_BBN). While
there is nothing wrong with doing it this way, CXL spec has a heavy
reliance on _UID to identify host bridges and there is no link to the
bus number. Having a distinct UID solves two problems. The first is it
gets us around the limitation of 256 (current max bus number). The
second is it allows us to replicate hardware configurations where bus
number and uid aren't equivalent. The latter has benefits for our
development and debugging using QEMU.

The other way to do this would be to implement the expanded bus
numbering, but having an explicit uid makes more sense when trying to
replicate real hardware configurations.

The QEMU commandline to utilize this would be:
  -device pxb-cxl,id=cxl.0,bus="pcie.0",bus_nr=1,uid=x

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>

--

I'm guessing this patch will be somewhat controversial. For early CXL
work, this can be dropped without too much heartache.
---
 hw/i386/acpi-build.c                |  3 ++-
 hw/pci-bridge/pci_expander_bridge.c | 19 +++++++++++++++++++
 hw/pci/pci.c                        | 11 +++++++++++
 include/hw/pci/pci.h                |  1 +
 include/hw/pci/pci_bus.h            |  1 +
 5 files changed, 34 insertions(+), 1 deletion(-)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 3c214bba67..26e4ddd025 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -1343,6 +1343,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
         QLIST_FOREACH(bus, &bus->child, sibling) {
             uint8_t bus_num = pci_bus_num(bus);
             uint8_t numa_node = pci_bus_numa_node(bus);
+            int32_t uid = pci_bus_uid(bus);
 
             /* look only for expander root buses */
             if (!pci_bus_is_root(bus)) {
@@ -1356,7 +1357,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
             scope = aml_scope("\\_SB");
             dev = aml_device("PC%.02X", bus_num);
             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
-            init_pci_acpi(dev, bus_num, pci_bus_is_express(bus) ? PCIE : PCI);
+            init_pci_acpi(dev, uid, pci_bus_is_express(bus) ? PCIE : PCI);
 
             if (numa_node != NUMA_NODE_UNASSIGNED) {
                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index b42592e1ff..5021b60435 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -67,6 +67,7 @@ struct PXBDev {
 
     uint8_t bus_nr;
     uint16_t numa_node;
+    int32_t uid;
 };
 
 static PXBDev *convert_to_pxb(PCIDevice *dev)
@@ -98,12 +99,20 @@ static uint16_t pxb_bus_numa_node(PCIBus *bus)
     return pxb->numa_node;
 }
 
+static int32_t pxb_bus_uid(PCIBus *bus)
+{
+    PXBDev *pxb = convert_to_pxb(bus->parent_dev);
+
+    return pxb->uid;
+}
+
 static void pxb_bus_class_init(ObjectClass *class, void *data)
 {
     PCIBusClass *pbc = PCI_BUS_CLASS(class);
 
     pbc->bus_num = pxb_bus_num;
     pbc->numa_node = pxb_bus_numa_node;
+    pbc->uid = pxb_bus_uid;
 }
 
 static const TypeInfo pxb_bus_info = {
@@ -329,6 +338,7 @@ static Property pxb_dev_properties[] = {
     /* Note: 0 is not a legal PXB bus number. */
     DEFINE_PROP_UINT8("bus_nr", PXBDev, bus_nr, 0),
     DEFINE_PROP_UINT16("numa_node", PXBDev, numa_node, NUMA_NODE_UNASSIGNED),
+    DEFINE_PROP_INT32("uid", PXBDev, uid, -1),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -400,12 +410,21 @@ static const TypeInfo pxb_pcie_dev_info = {
 
 static void pxb_cxl_dev_realize(PCIDevice *dev, Error **errp)
 {
+    PXBDev *pxb = convert_to_pxb(dev);
+
     /* A CXL PXB's parent bus is still PCIe */
     if (!pci_bus_is_express(pci_get_bus(dev))) {
         error_setg(errp, "pxb-cxl devices cannot reside on a PCI bus");
         return;
     }
 
+    if (pxb->uid < 0) {
+        error_setg(errp, "pxb-cxl devices must have a valid uid (0-2147483647)");
+        return;
+    }
+
+    /* FIXME: Check that uid doesn't collide with UIDs of other host bridges */
+
     pxb_dev_realize_common(dev, CXL, errp);
 }
 
diff --git a/hw/pci/pci.c b/hw/pci/pci.c
index fb454e2317..d2eae06e18 100644
--- a/hw/pci/pci.c
+++ b/hw/pci/pci.c
@@ -170,6 +170,11 @@ static uint16_t pcibus_numa_node(PCIBus *bus)
     return NUMA_NODE_UNASSIGNED;
 }
 
+static int32_t pcibus_uid(PCIBus *bus)
+{
+    return -1;
+}
+
 static void pci_bus_class_init(ObjectClass *klass, void *data)
 {
     BusClass *k = BUS_CLASS(klass);
@@ -184,6 +189,7 @@ static void pci_bus_class_init(ObjectClass *klass, void *data)
 
     pbc->bus_num = pcibus_num;
     pbc->numa_node = pcibus_numa_node;
+    pbc->uid = pcibus_uid;
 }
 
 static const TypeInfo pci_bus_info = {
@@ -530,6 +536,11 @@ int pci_bus_numa_node(PCIBus *bus)
     return PCI_BUS_GET_CLASS(bus)->numa_node(bus);
 }
 
+int pci_bus_uid(PCIBus *bus)
+{
+    return PCI_BUS_GET_CLASS(bus)->uid(bus);
+}
+
 static int get_pci_config_device(QEMUFile *f, void *pv, size_t size,
                                  const VMStateField *field)
 {
diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h
index c464597e53..d65c2bceae 100644
--- a/include/hw/pci/pci.h
+++ b/include/hw/pci/pci.h
@@ -462,6 +462,7 @@ static inline int pci_dev_bus_num(const PCIDevice *dev)
 }
 
 int pci_bus_numa_node(PCIBus *bus);
+int pci_bus_uid(PCIBus *bus);
 void pci_for_each_device(PCIBus *bus, int bus_num,
                          void (*fn)(PCIBus *bus, PCIDevice *d, void *opaque),
                          void *opaque);
diff --git a/include/hw/pci/pci_bus.h b/include/hw/pci/pci_bus.h
index eb94e7e85c..3c9fbc55bb 100644
--- a/include/hw/pci/pci_bus.h
+++ b/include/hw/pci/pci_bus.h
@@ -17,6 +17,7 @@ struct PCIBusClass {
 
     int (*bus_num)(PCIBus *bus);
     uint16_t (*numa_node)(PCIBus *bus);
+    int32_t (*uid)(PCIBus *bus);
 };
 
 enum PCIBusFlags {
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 20/32] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (18 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 19/32] hw/pci: Plumb _UID through host bridges Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 21/32] acpi/pxb/cxl: Reserve host bridge MMIO Ben Widawsky
                   ` (12 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

CXL host bridges themselves may have MMIO. Since host bridges don't have
a BAR they are treated as special for MMIO.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>

--

It's arbitrarily chosen here to pick 0xD0000000 as the base for the host
bridge MMIO. I'm not sure what the right way to find free space for
platform hardcoded things like this is.
---
 hw/pci-bridge/pci_expander_bridge.c | 53 ++++++++++++++++++++++++++++-
 include/hw/cxl/cxl.h                |  2 ++
 2 files changed, 54 insertions(+), 1 deletion(-)

diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index 5021b60435..226a8a5fff 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -17,6 +17,7 @@
 #include "hw/pci/pci_host.h"
 #include "hw/qdev-properties.h"
 #include "hw/pci/pci_bridge.h"
+#include "hw/cxl/cxl.h"
 #include "qemu/range.h"
 #include "qemu/error-report.h"
 #include "qemu/module.h"
@@ -70,6 +71,12 @@ struct PXBDev {
     int32_t uid;
 };
 
+typedef struct CXLHost {
+    PCIHostState parent_obj;
+
+    CXLComponentState cxl_cstate;
+} CXLHost;
+
 static PXBDev *convert_to_pxb(PCIDevice *dev)
 {
     /* A CXL PXB's parent bus is PCIe, so the normal check won't work */
@@ -85,6 +92,9 @@ static GList *pxb_dev_list;
 
 #define TYPE_PXB_HOST "pxb-host"
 
+#define TYPE_PXB_CXL_HOST "pxb-cxl-host"
+#define PXB_CXL_HOST(obj) OBJECT_CHECK(CXLHost, (obj), TYPE_PXB_CXL_HOST)
+
 static int pxb_bus_num(PCIBus *bus)
 {
     PXBDev *pxb = convert_to_pxb(bus->parent_dev);
@@ -198,6 +208,46 @@ static const TypeInfo pxb_host_info = {
     .class_init    = pxb_host_class_init,
 };
 
+static void pxb_cxl_realize(DeviceState *dev, Error **errp)
+{
+    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
+    PCIHostState *phb = PCI_HOST_BRIDGE(dev);
+    CXLHost *cxl = PXB_CXL_HOST(dev);
+    CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
+    struct MemoryRegion *mr = &cxl_cstate->crb.component_registers;
+
+    cxl_component_register_block_init(OBJECT(dev), cxl_cstate,
+                                      TYPE_PXB_CXL_HOST);
+    sysbus_init_mmio(sbd, mr);
+
+    /* FIXME: support multiple host bridges. */
+    sysbus_mmio_map(sbd, 0, CXL_HOST_BASE +
+                            memory_region_size(mr) * pci_bus_uid(phb->bus));
+}
+
+static void pxb_cxl_host_class_init(ObjectClass *class, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(class);
+    PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(class);
+
+    hc->root_bus_path = pxb_host_root_bus_path;
+    dc->fw_name = "cxl";
+    dc->realize = pxb_cxl_realize;
+    /* Reason: Internal part of the pxb/pxb-pcie device, not usable by itself */
+    dc->user_creatable = false;
+}
+
+/*
+ * This is a device to handle the MMIO for a CXL host bridge. It does nothing
+ * else.
+ */
+static const TypeInfo cxl_host_info = {
+    .name          = TYPE_PXB_CXL_HOST,
+    .parent        = TYPE_PCI_HOST_BRIDGE,
+    .instance_size = sizeof(CXLHost),
+    .class_init    = pxb_cxl_host_class_init,
+};
+
 /*
  * Registers the PXB bus as a child of pci host root bus.
  */
@@ -272,7 +322,7 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
         dev_name = dev->qdev.id;
     }
 
-    ds = qdev_new(TYPE_PXB_HOST);
+    ds = qdev_new(type == CXL ? TYPE_PXB_CXL_HOST : TYPE_PXB_HOST);
     if (type == PCIE) {
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
     } else if (type == CXL) {
@@ -466,6 +516,7 @@ static void pxb_register_types(void)
     type_register_static(&pxb_pcie_bus_info);
     type_register_static(&pxb_cxl_bus_info);
     type_register_static(&pxb_host_info);
+    type_register_static(&cxl_host_info);
     type_register_static(&pxb_dev_info);
     type_register_static(&pxb_pcie_dev_info);
     type_register_static(&pxb_cxl_dev_info);
diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
index 362cda40de..6bc344f205 100644
--- a/include/hw/cxl/cxl.h
+++ b/include/hw/cxl/cxl.h
@@ -17,5 +17,7 @@
 #define COMPONENT_REG_BAR_IDX 0
 #define DEVICE_REG_BAR_IDX 2
 
+#define CXL_HOST_BASE 0xD0000000
+
 #endif
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 21/32] acpi/pxb/cxl: Reserve host bridge MMIO
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (19 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 20/32] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 22/32] hw/pxb/cxl: Add "windows" for host bridges Ben Widawsky
                   ` (11 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

For all host bridges, reserve MMIO space with _CRS. The MMIO for the
host bridge lives in a magically hard coded space in the system's
physical address space. The standard mechanism to tell the OS about
regions which can't be used for host bridges is _CRS.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/i386/acpi-build.c | 22 +++++++++++++++++++---
 1 file changed, 19 insertions(+), 3 deletions(-)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 26e4ddd025..16cde677a0 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -28,6 +28,7 @@
 #include "qemu/bitmap.h"
 #include "qemu/error-report.h"
 #include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
 #include "hw/core/cpu.h"
 #include "target/i386/cpu.h"
 #include "hw/misc/pvpanic.h"
@@ -1194,7 +1195,7 @@ static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
     aml_append(table, scope);
 }
 
-enum { PCI, PCIE };
+enum { PCI, PCIE, CXL };
 static void init_pci_acpi(Aml *dev, int uid, int type)
 {
     if (type == PCI) {
@@ -1344,20 +1345,28 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
             uint8_t bus_num = pci_bus_num(bus);
             uint8_t numa_node = pci_bus_numa_node(bus);
             int32_t uid = pci_bus_uid(bus);
+            int type;
 
             /* look only for expander root buses */
             if (!pci_bus_is_root(bus)) {
                 continue;
             }
 
+            type = pci_bus_is_cxl(bus) ? CXL :
+                                         pci_bus_is_express(bus) ? PCIE : PCI;
+
             if (bus_num < root_bus_limit) {
                 root_bus_limit = bus_num - 1;
             }
 
             scope = aml_scope("\\_SB");
-            dev = aml_device("PC%.02X", bus_num);
+            if (type == CXL) {
+                dev = aml_device("CXL%.01X", pci_bus_uid(bus));
+            } else {
+                dev = aml_device("PC%.02X", bus_num);
+            }
             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
-            init_pci_acpi(dev, uid, pci_bus_is_express(bus) ? PCIE : PCI);
+            init_pci_acpi(dev, uid, type);
 
             if (numa_node != NUMA_NODE_UNASSIGNED) {
                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
@@ -1368,6 +1377,13 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
             aml_append(dev, aml_name_decl("_CRS", crs));
             aml_append(scope, dev);
             aml_append(dsdt, scope);
+
+            /* Handle the ranges for the PXB expanders */
+            if (type == CXL) {
+                uint64_t base = CXL_HOST_BASE + uid * 0x10000;
+                crs_range_insert(crs_range_set.mem_ranges, base,
+                                 base + 0x10000 - 1);
+            }
         }
     }
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 22/32] hw/pxb/cxl: Add "windows" for host bridges
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (20 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 21/32] acpi/pxb/cxl: Reserve host bridge MMIO Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 23/32] hw/cxl/rp: Add a root port Ben Widawsky
                   ` (10 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

In a bare metal CXL capable system, system firmware will program
physical address ranges on the host. This is done by programming
internal registers that aren't typically known to OS. These address
ranges might be contiguous or interleaved across host bridges.

For a QEMU guest a new construct is introduced allowing passing a memory
backend to the host bridge for this same purpose. Each memory backend
needs to be passed to the host bridge as well as any device that will be
emulating that memory (not implemented here).

I'm hopeful the interleaving work in the link can be re-purposed here
(see Link).

An example to create a host bridges with a 512M window at 0x4c0000000
 -object memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M
 -device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52,uid=0,len-memory-base=1,memory-base\[0\]=0x4c0000000,memory\[0\]=cxl-mem1

Link: https://lists.nongnu.org/archive/html/qemu-devel/2020-08/msg03680.html
Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/pci-bridge/pci_expander_bridge.c | 65 +++++++++++++++++++++++++++--
 include/hw/cxl/cxl.h                |  1 +
 2 files changed, 62 insertions(+), 4 deletions(-)

diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index 226a8a5fff..af1450c69d 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -69,12 +69,19 @@ struct PXBDev {
     uint8_t bus_nr;
     uint16_t numa_node;
     int32_t uid;
+    struct cxl_dev {
+        HostMemoryBackend *memory_window[CXL_WINDOW_MAX];
+
+        uint32_t num_windows;
+        hwaddr *window_base[CXL_WINDOW_MAX];
+    } cxl;
 };
 
 typedef struct CXLHost {
     PCIHostState parent_obj;
 
     CXLComponentState cxl_cstate;
+    PXBDev *dev;
 } CXLHost;
 
 static PXBDev *convert_to_pxb(PCIDevice *dev)
@@ -213,16 +220,31 @@ static void pxb_cxl_realize(DeviceState *dev, Error **errp)
     SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
     PCIHostState *phb = PCI_HOST_BRIDGE(dev);
     CXLHost *cxl = PXB_CXL_HOST(dev);
+    struct cxl_dev *cxl_dev = &cxl->dev->cxl;
     CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
     struct MemoryRegion *mr = &cxl_cstate->crb.component_registers;
+    int uid = pci_bus_uid(phb->bus);
 
     cxl_component_register_block_init(OBJECT(dev), cxl_cstate,
                                       TYPE_PXB_CXL_HOST);
     sysbus_init_mmio(sbd, mr);
 
-    /* FIXME: support multiple host bridges. */
-    sysbus_mmio_map(sbd, 0, CXL_HOST_BASE +
-                            memory_region_size(mr) * pci_bus_uid(phb->bus));
+    sysbus_mmio_map(sbd, 0, CXL_HOST_BASE + memory_region_size(mr) * uid);
+
+    /*
+     * A CXL host bridge can exist without a fixed memory window, but it would
+     * only operate in legacy PCIe mode.
+     */
+    if (!cxl_dev->memory_window[uid]) {
+        warn_report(
+            "CXL expander bridge created without window. Consider using %s",
+            "memdev[0]=<memory_backend>");
+        return;
+    }
+
+    mr = host_memory_backend_get_memory(cxl_dev->memory_window[uid]);
+    sysbus_init_mmio(sbd, mr);
+    sysbus_mmio_map(sbd, 1 + uid, *cxl_dev->window_base[uid]);
 }
 
 static void pxb_cxl_host_class_init(ObjectClass *class, void *data)
@@ -328,6 +350,7 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
     } else if (type == CXL) {
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_CXL_BUS);
         bus->flags |= PCI_BUS_CXL;
+        PXB_CXL_HOST(ds)->dev = PXB_CXL_DEV(dev);
     } else {
         bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
         bds = qdev_new("pci-bridge");
@@ -389,6 +412,8 @@ static Property pxb_dev_properties[] = {
     DEFINE_PROP_UINT8("bus_nr", PXBDev, bus_nr, 0),
     DEFINE_PROP_UINT16("numa_node", PXBDev, numa_node, NUMA_NODE_UNASSIGNED),
     DEFINE_PROP_INT32("uid", PXBDev, uid, -1),
+    DEFINE_PROP_ARRAY("window-base", PXBDev, cxl.num_windows, cxl.window_base,
+                      qdev_prop_uint64, hwaddr),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -460,7 +485,9 @@ static const TypeInfo pxb_pcie_dev_info = {
 
 static void pxb_cxl_dev_realize(PCIDevice *dev, Error **errp)
 {
-    PXBDev *pxb = convert_to_pxb(dev);
+    PXBDev *pxb = PXB_CXL_DEV(dev);
+    struct cxl_dev *cxl = &pxb->cxl;
+    int count = 0;
 
     /* A CXL PXB's parent bus is still PCIe */
     if (!pci_bus_is_express(pci_get_bus(dev))) {
@@ -476,6 +503,23 @@ static void pxb_cxl_dev_realize(PCIDevice *dev, Error **errp)
     /* FIXME: Check that uid doesn't collide with UIDs of other host bridges */
 
     pxb_dev_realize_common(dev, CXL, errp);
+
+    for (unsigned i = 0; i < CXL_WINDOW_MAX; i++) {
+        if (!cxl->memory_window[i]) {
+            continue;
+        }
+
+        count++;
+    }
+
+    if (!count) {
+        warn_report("memory-windows should be set when creating CXL host bridges");
+    }
+
+    if (count != cxl->num_windows) {
+        error_setg(errp, "window bases count (%d) must match window count (%d)",
+                   cxl->num_windows, count);
+    }
 }
 
 static void pxb_cxl_dev_class_init(ObjectClass *klass, void *data)
@@ -496,6 +540,19 @@ static void pxb_cxl_dev_class_init(ObjectClass *klass, void *data)
 
     /* Host bridges aren't hotpluggable. FIXME: spec reference */
     dc->hotpluggable = false;
+
+    /*
+     * Below is moral equivalent of:
+     *   DEFINE_PROP_ARRAY("memdev", PXBDev, window_count, windows,
+     *                     qdev_prop_memory_device, HostMemoryBackend)
+     */
+    for (unsigned i = 0; i < CXL_WINDOW_MAX; i++) {
+        g_autofree char *name = g_strdup_printf("memdev[%u]", i);
+        object_class_property_add_link(klass, name, TYPE_MEMORY_BACKEND,
+                offsetof(PXBDev, cxl.memory_window[i]),
+                qdev_prop_allow_set_link_before_realize,
+                OBJ_PROP_LINK_STRONG);
+    }
 }
 
 static const TypeInfo pxb_cxl_dev_info = {
diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
index 6bc344f205..b1e5f4a8fa 100644
--- a/include/hw/cxl/cxl.h
+++ b/include/hw/cxl/cxl.h
@@ -18,6 +18,7 @@
 #define DEVICE_REG_BAR_IDX 2
 
 #define CXL_HOST_BASE 0xD0000000
+#define CXL_WINDOW_MAX 10
 
 #endif
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 23/32] hw/cxl/rp: Add a root port
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (21 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 22/32] hw/pxb/cxl: Add "windows" for host bridges Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5) Ben Widawsky
                   ` (9 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This adds just enough of a root port implementation to be able to
enumerate root ports (creating the required DVSEC entries). What's not
here yet is the MMIO nor the ability to write some of the DVSEC entries.

This can be added with the qemu commandline by adding a rootport to a
specific CXL host bridge. For example:
  -device cxl-rp,id=rp0,bus="cxl.0",addr=0.0,chassis=4

Like the host bridge patch, the ACPI tables aren't generated at this
point and so system software cannot use it.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/pci-bridge/Kconfig          |   5 +
 hw/pci-bridge/cxl_root_port.c  | 231 +++++++++++++++++++++++++++++++++
 hw/pci-bridge/meson.build      |   1 +
 hw/pci-bridge/pcie_root_port.c |   6 +-
 hw/pci/pci.c                   |   4 +-
 5 files changed, 245 insertions(+), 2 deletions(-)
 create mode 100644 hw/pci-bridge/cxl_root_port.c

diff --git a/hw/pci-bridge/Kconfig b/hw/pci-bridge/Kconfig
index a51ec716f5..a821b531da 100644
--- a/hw/pci-bridge/Kconfig
+++ b/hw/pci-bridge/Kconfig
@@ -27,3 +27,8 @@ config DEC_PCI
 
 config SIMBA
     bool
+
+config CXL
+    bool
+    default y if PCI_EXPRESS && PXB
+    depends on PCI_EXPRESS && MSI_NONBROKEN && PXB
diff --git a/hw/pci-bridge/cxl_root_port.c b/hw/pci-bridge/cxl_root_port.c
new file mode 100644
index 0000000000..6c3b215bb3
--- /dev/null
+++ b/hw/pci-bridge/cxl_root_port.c
@@ -0,0 +1,231 @@
+/*
+ * CXL 2.0 Root Port Implementation
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "qemu/range.h"
+#include "hw/pci/pci_bridge.h"
+#include "hw/pci/pcie_port.h"
+#include "hw/qdev-properties.h"
+#include "hw/sysbus.h"
+#include "qapi/error.h"
+#include "hw/cxl/cxl.h"
+
+#define CXL_ROOT_PORT_DID 0x7075
+
+/* Copied from the gen root port which we derive */
+#define GEN_PCIE_ROOT_PORT_AER_OFFSET 0x100
+#define GEN_PCIE_ROOT_PORT_ACS_OFFSET \
+    (GEN_PCIE_ROOT_PORT_AER_OFFSET + PCI_ERR_SIZEOF)
+#define CXL_ROOT_PORT_DVSEC_OFFSET \
+    (GEN_PCIE_ROOT_PORT_ACS_OFFSET + PCI_ACS_SIZEOF)
+
+typedef struct CXLRootPort {
+    /*< private >*/
+    PCIESlot parent_obj;
+
+    CXLComponentState cxl_cstate;
+    PCIResReserve res_reserve;
+} CXLRootPort;
+
+#define TYPE_CXL_ROOT_PORT "cxl-rp"
+DECLARE_INSTANCE_CHECKER(CXLRootPort, CXL_ROOT_PORT, TYPE_CXL_ROOT_PORT)
+
+static void latch_registers(CXLRootPort *crp)
+{
+    uint32_t *reg_state = crp->cxl_cstate.crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_ROOT_PORT);
+}
+
+static void build_dvsecs(CXLComponentState *cxl)
+{
+    uint8_t *dvsec;
+
+    dvsec = (uint8_t *)&(struct extensions_dvsec_port){ 0 };
+    cxl_component_create_dvsec(cxl, EXTENSIONS_PORT_DVSEC_LENGTH,
+                               EXTENSIONS_PORT_DVSEC,
+                               EXTENSIONS_PORT_DVSEC_REVID, dvsec);
+
+    dvsec = (uint8_t *)&(struct dvsec_port_gpf){
+        .rsvd        = 0,
+        .phase1_ctrl = 1, /* 1μs timeout */
+        .phase2_ctrl = 1, /* 1μs timeout */
+    };
+    cxl_component_create_dvsec(cxl, GPF_PORT_DVSEC_LENGTH, GPF_PORT_DVSEC,
+                               GPF_PORT_DVSEC_REVID, dvsec);
+
+    dvsec = (uint8_t *)&(struct dvsec_port_flexbus){
+        .cap              = 0x26, /* IO, Mem, non-MLD */
+        .ctrl             = 0,
+        .status           = 0x26, /* same */
+        .rcvd_mod_ts_data = 0xef, /* WTF? */
+    };
+    cxl_component_create_dvsec(cxl, PCIE_FLEXBUS_PORT_DVSEC_LENGTH_2_0,
+                               PCIE_FLEXBUS_PORT_DVSEC,
+                               PCIE_FLEXBUS_PORT_DVSEC_REVID_2_0, dvsec);
+
+    dvsec = (uint8_t *)&(struct dvsec_register_locator){
+        .rsvd         = 0,
+        .reg0_base_lo = RBI_COMPONENT_REG | COMPONENT_REG_BAR_IDX,
+        .reg0_base_hi = 0,
+    };
+    cxl_component_create_dvsec(cxl, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
+                               REG_LOC_DVSEC_REVID, dvsec);
+}
+
+static void cxl_rp_realize(DeviceState *dev, Error **errp)
+{
+    PCIDevice *pci_dev     = PCI_DEVICE(dev);
+    PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
+    CXLRootPort *crp       = CXL_ROOT_PORT(dev);
+    CXLComponentState *cxl_cstate = &crp->cxl_cstate;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    MemoryRegion *component_bar = &cregs->component_registers;
+    Error *local_err = NULL;
+
+    rpc->parent_realize(dev, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return;
+    }
+
+    int rc =
+        pci_bridge_qemu_reserve_cap_init(pci_dev, 0, crp->res_reserve, errp);
+    if (rc < 0) {
+        rpc->parent_class.exit(pci_dev);
+        return;
+    }
+
+    if (!crp->res_reserve.io || crp->res_reserve.io == -1) {
+        pci_word_test_and_clear_mask(pci_dev->wmask + PCI_COMMAND,
+                                     PCI_COMMAND_IO);
+        pci_dev->wmask[PCI_IO_BASE]  = 0;
+        pci_dev->wmask[PCI_IO_LIMIT] = 0;
+    }
+
+    cxl_cstate->dvsec_offset = CXL_ROOT_PORT_DVSEC_OFFSET;
+    cxl_cstate->pdev = pci_dev;
+    build_dvsecs(&crp->cxl_cstate);
+
+    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
+                                      TYPE_CXL_ROOT_PORT);
+
+    pci_register_bar(pci_dev, COMPONENT_REG_BAR_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY |
+                         PCI_BASE_ADDRESS_MEM_TYPE_64,
+                     component_bar);
+}
+
+static void cxl_rp_reset(DeviceState *dev)
+{
+    PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
+    CXLRootPort *crp = CXL_ROOT_PORT(dev);
+
+    rpc->parent_reset(dev);
+
+    latch_registers(crp);
+}
+
+static Property gen_rp_props[] = {
+    DEFINE_PROP_UINT32("bus-reserve", CXLRootPort, res_reserve.bus, -1),
+    DEFINE_PROP_SIZE("io-reserve", CXLRootPort, res_reserve.io, -1),
+    DEFINE_PROP_SIZE("mem-reserve", CXLRootPort, res_reserve.mem_non_pref, -1),
+    DEFINE_PROP_SIZE("pref32-reserve", CXLRootPort, res_reserve.mem_pref_32,
+                     -1),
+    DEFINE_PROP_SIZE("pref64-reserve", CXLRootPort, res_reserve.mem_pref_64,
+                     -1),
+    DEFINE_PROP_END_OF_LIST()
+};
+
+static void cxl_rp_dvsec_write_config(PCIDevice *dev, uint32_t addr,
+                                      uint32_t val, int len)
+{
+    CXLRootPort *crp = CXL_ROOT_PORT(dev);
+
+    if (range_contains(&crp->cxl_cstate.dvsecs[EXTENSIONS_PORT_DVSEC], addr)) {
+        uint8_t *reg = &dev->config[addr];
+        addr -= crp->cxl_cstate.dvsecs[EXTENSIONS_PORT_DVSEC].lob;
+        if (addr == PORT_CONTROL_OVERRIDE_OFFSET) {
+            if (pci_get_word(reg) & PORT_CONTROL_UNMASK_SBR) {
+                /* unmask SBR */
+            }
+            if (pci_get_word(reg) & PORT_CONTROL_ALT_MEMID_EN) {
+                /* Alt Memory & ID Space Enable */
+            }
+        }
+    }
+}
+
+static void cxl_rp_write_config(PCIDevice *d, uint32_t address, uint32_t val,
+                                int len)
+{
+    uint16_t slt_ctl, slt_sta;
+
+    pcie_cap_slot_get(d, &slt_ctl, &slt_sta);
+    pci_bridge_write_config(d, address, val, len);
+    pcie_cap_flr_write_config(d, address, val, len);
+    pcie_cap_slot_write_config(d, slt_ctl, slt_sta, address, val, len);
+    pcie_aer_write_config(d, address, val, len);
+
+    cxl_rp_dvsec_write_config(d, address, val, len);
+}
+
+static void cxl_root_port_class_init(ObjectClass *oc, void *data)
+{
+    DeviceClass *dc        = DEVICE_CLASS(oc);
+    PCIDeviceClass *k      = PCI_DEVICE_CLASS(oc);
+    PCIERootPortClass *rpc = PCIE_ROOT_PORT_CLASS(oc);
+
+    k->vendor_id = PCI_VENDOR_ID_INTEL;
+    k->device_id = CXL_ROOT_PORT_DID;
+    dc->desc     = "CXL Root Port";
+    k->revision  = 0;
+    device_class_set_props(dc, gen_rp_props);
+    k->config_write = cxl_rp_write_config;
+
+    device_class_set_parent_realize(dc, cxl_rp_realize, &rpc->parent_realize);
+    device_class_set_parent_reset(dc, cxl_rp_reset, &rpc->parent_reset);
+
+    rpc->aer_offset = GEN_PCIE_ROOT_PORT_AER_OFFSET;
+    rpc->acs_offset = GEN_PCIE_ROOT_PORT_ACS_OFFSET;
+
+    /*
+     * Explain
+     */
+    dc->hotpluggable = false;
+}
+
+static const TypeInfo cxl_root_port_info = {
+    .name = TYPE_CXL_ROOT_PORT,
+    .parent = TYPE_PCIE_ROOT_PORT,
+    .instance_size = sizeof(CXLRootPort),
+    .class_init = cxl_root_port_class_init,
+    .interfaces = (InterfaceInfo[]) {
+        { INTERFACE_CXL_DEVICE },
+        { }
+    },
+};
+
+static void cxl_register(void)
+{
+    type_register_static(&cxl_root_port_info);
+}
+
+type_init(cxl_register);
diff --git a/hw/pci-bridge/meson.build b/hw/pci-bridge/meson.build
index daab8acf2a..b6d26a03d5 100644
--- a/hw/pci-bridge/meson.build
+++ b/hw/pci-bridge/meson.build
@@ -5,6 +5,7 @@ pci_ss.add(when: 'CONFIG_IOH3420', if_true: files('ioh3420.c'))
 pci_ss.add(when: 'CONFIG_PCIE_PORT', if_true: files('pcie_root_port.c', 'gen_pcie_root_port.c', 'pcie_pci_bridge.c'))
 pci_ss.add(when: 'CONFIG_PXB', if_true: files('pci_expander_bridge.c'))
 pci_ss.add(when: 'CONFIG_XIO3130', if_true: files('xio3130_upstream.c', 'xio3130_downstream.c'))
+pci_ss.add(when: 'CONFIG_CXL', if_true: files('cxl_root_port.c'))
 
 # NewWorld PowerMac
 pci_ss.add(when: 'CONFIG_DEC_PCI', if_true: files('dec.c'))
diff --git a/hw/pci-bridge/pcie_root_port.c b/hw/pci-bridge/pcie_root_port.c
index f1cfe9d14a..460e48269d 100644
--- a/hw/pci-bridge/pcie_root_port.c
+++ b/hw/pci-bridge/pcie_root_port.c
@@ -67,7 +67,11 @@ static void rp_realize(PCIDevice *d, Error **errp)
     int rc;
 
     pci_config_set_interrupt_pin(d->config, 1);
-    pci_bridge_initfn(d, TYPE_PCIE_BUS);
+    if (d->cap_present & QEMU_PCIE_CAP_CXL) {
+        pci_bridge_initfn(d, TYPE_CXL_BUS);
+    } else {
+        pci_bridge_initfn(d, TYPE_PCIE_BUS);
+    }
     pcie_port_init_reg(d);
 
     rc = pci_bridge_ssvid_init(d, rpc->ssvid_offset, dc->vendor_id,
diff --git a/hw/pci/pci.c b/hw/pci/pci.c
index d2eae06e18..ecc98938f9 100644
--- a/hw/pci/pci.c
+++ b/hw/pci/pci.c
@@ -2692,7 +2692,9 @@ static void pci_device_class_base_init(ObjectClass *klass, void *data)
             object_class_dynamic_cast(klass, INTERFACE_CONVENTIONAL_PCI_DEVICE);
         ObjectClass *pcie =
             object_class_dynamic_cast(klass, INTERFACE_PCIE_DEVICE);
-        assert(conventional || pcie);
+        ObjectClass *cxl =
+            object_class_dynamic_cast(klass, INTERFACE_CXL_DEVICE);
+        assert(conventional || pcie || cxl);
     }
 }
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (22 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 23/32] hw/cxl/rp: Add a root port Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-27 21:03   ` Igor Mammedov
  2021-01-05 16:53 ` [RFC PATCH v2 25/32] hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12) Ben Widawsky
                   ` (8 subsequent siblings)
  32 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

A CXL memory device (AKA Type 3) is a CXL component that contains some
combination of volatile and persistent memory. It also implements the
previously defined mailbox interface as well as the memory device
firmware interface.

The following example will create a 256M device in a 512M window:

-object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
-device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/core/numa.c             |   3 +
 hw/cxl/cxl-mailbox-utils.c |  41 ++++++
 hw/i386/pc.c               |   1 +
 hw/mem/Kconfig             |   5 +
 hw/mem/cxl_type3.c         | 262 +++++++++++++++++++++++++++++++++++++
 hw/mem/meson.build         |   1 +
 hw/pci/pcie.c              |  30 +++++
 include/hw/cxl/cxl.h       |   2 +
 include/hw/cxl/cxl_pci.h   |  22 ++++
 include/hw/pci/pci_ids.h   |   1 +
 monitor/hmp-cmds.c         |  15 +++
 qapi/machine.json          |   1 +
 12 files changed, 384 insertions(+)
 create mode 100644 hw/mem/cxl_type3.c

diff --git a/hw/core/numa.c b/hw/core/numa.c
index 68cee65f61..cd7df371e6 100644
--- a/hw/core/numa.c
+++ b/hw/core/numa.c
@@ -770,6 +770,9 @@ static void numa_stat_memory_devices(NumaNodeMem node_mem[])
                 node_mem[pcdimm_info->node].node_plugged_mem +=
                     pcdimm_info->size;
                 break;
+            case MEMORY_DEVICE_INFO_KIND_CXL:
+                /* FINISHME */
+                break;
             case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
                 vpi = value->u.virtio_pmem.data;
                 /* TODO: once we support numa, assign to right node */
diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
index f68ec5b5b9..eeb10b8943 100644
--- a/hw/cxl/cxl-mailbox-utils.c
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -49,6 +49,8 @@ enum {
     LOGS        = 0x04,
         #define GET_SUPPORTED 0x0
         #define GET_LOG       0x1
+    IDENTIFY    = 0x40,
+        #define MEMORY_DEVICE 0x0
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -127,6 +129,7 @@ declare_mailbox_handler(TIMESTAMP_GET);
 declare_mailbox_handler(TIMESTAMP_SET);
 declare_mailbox_handler(LOGS_GET_SUPPORTED);
 declare_mailbox_handler(LOGS_GET_LOG);
+declare_mailbox_handler(IDENTIFY_MEMORY_DEVICE);
 
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
 #define IMMEDIATE_POLICY_CHANGE (1 << 3)
@@ -144,6 +147,7 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
     CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
     CXL_CMD(LOGS, GET_SUPPORTED, 0, 0),
     CXL_CMD(LOGS, GET_LOG, 0x18, 0),
+    CXL_CMD(IDENTIFY, MEMORY_DEVICE, 0, 0),
 };
 
 #undef CXL_CMD
@@ -262,6 +266,43 @@ define_mailbox_handler(LOGS_GET_LOG)
     return CXL_MBOX_SUCCESS;
 }
 
+/* 8.2.9.5.1.1 */
+define_mailbox_handler(IDENTIFY_MEMORY_DEVICE)
+{
+    struct {
+        char fw_revision[0x10];
+        uint64_t total_capacity;
+        uint64_t volatile_capacity;
+        uint64_t persistent_capacity;
+        uint64_t partition_align;
+        uint16_t info_event_log_size;
+        uint16_t warning_event_log_size;
+        uint16_t failure_event_log_size;
+        uint16_t fatal_event_log_size;
+        uint32_t lsa_size;
+        uint8_t poison_list_max_mer[3];
+        uint16_t inject_poison_limit;
+        uint8_t poison_caps;
+        uint8_t qos_telemetry_caps;
+    } __attribute__((packed)) *id;
+    _Static_assert(sizeof(*id) == 0x43, "Bad identify size");
+
+    if (memory_region_size(cxl_dstate->pmem) < (256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    id = (void *)cmd->payload;
+    memset(id, 0, sizeof(*id));
+
+    /* PMEM only */
+    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
+    id->total_capacity = memory_region_size(cxl_dstate->pmem);
+    id->persistent_capacity = memory_region_size(cxl_dstate->pmem);
+
+    *len = sizeof(*id);
+    return CXL_MBOX_SUCCESS;
+}
+
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
 {
     uint16_t ret = CXL_MBOX_SUCCESS;
diff --git a/hw/i386/pc.c b/hw/i386/pc.c
index 5458f61d10..5d41809b37 100644
--- a/hw/i386/pc.c
+++ b/hw/i386/pc.c
@@ -79,6 +79,7 @@
 #include "acpi-build.h"
 #include "hw/mem/pc-dimm.h"
 #include "hw/mem/nvdimm.h"
+#include "hw/cxl/cxl.h"
 #include "qapi/error.h"
 #include "qapi/qapi-visit-common.h"
 #include "qapi/visitor.h"
diff --git a/hw/mem/Kconfig b/hw/mem/Kconfig
index a0ef2cf648..7d9d1ced3e 100644
--- a/hw/mem/Kconfig
+++ b/hw/mem/Kconfig
@@ -10,3 +10,8 @@ config NVDIMM
     default y
     depends on (PC || PSERIES || ARM_VIRT)
     select MEM_DEVICE
+
+config CXL_MEM_DEVICE
+    bool
+    default y if CXL
+    select MEM_DEVICE
diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
new file mode 100644
index 0000000000..3985bb8d0b
--- /dev/null
+++ b/hw/mem/cxl_type3.c
@@ -0,0 +1,262 @@
+#include "qemu/osdep.h"
+#include "qemu/units.h"
+#include "qemu/error-report.h"
+#include "hw/mem/memory-device.h"
+#include "hw/mem/pc-dimm.h"
+#include "hw/pci/pci.h"
+#include "hw/qdev-properties.h"
+#include "qapi/error.h"
+#include "qemu/log.h"
+#include "qemu/module.h"
+#include "qemu/range.h"
+#include "qemu/rcu.h"
+#include "sysemu/hostmem.h"
+#include "hw/cxl/cxl.h"
+
+typedef struct cxl_type3_dev {
+    /* Private */
+    PCIDevice parent_obj;
+
+    /* Properties */
+    uint64_t size;
+    HostMemoryBackend *hostmem;
+
+    /* State */
+    CXLComponentState cxl_cstate;
+    CXLDeviceState cxl_dstate;
+} CXLType3Dev;
+
+#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
+
+static void build_dvsecs(CXLType3Dev *ct3d)
+{
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    uint8_t *dvsec;
+
+    dvsec = (uint8_t *)&(struct dvsec_device){
+        .cap = 0x1e,
+        .ctrl = 0x6,
+        .status2 = 0x2,
+        .range1_size_hi = 0,
+        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
+        .range1_base_hi = 0,
+        .range1_base_lo = 0,
+    };
+    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
+                               PCIE_CXL_DEVICE_DVSEC,
+                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
+
+    dvsec = (uint8_t *)&(struct dvsec_register_locator){
+        .rsvd         = 0,
+        .reg0_base_lo = RBI_COMPONENT_REG | COMPONENT_REG_BAR_IDX,
+        .reg0_base_hi = 0,
+        .reg1_base_lo = RBI_CXL_DEVICE_REG | DEVICE_REG_BAR_IDX,
+        .reg1_base_hi = 0,
+    };
+    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
+                               REG_LOC_DVSEC_REVID, dvsec);
+}
+
+static void ct3_instance_init(Object *obj)
+{
+    /* MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); */
+}
+
+static void ct3_finalize(Object *obj)
+{
+    CXLType3Dev *ct3d = CT3(obj);
+
+    g_free(ct3d->cxl_dstate.pmem);
+}
+
+static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
+{
+    MemoryRegionSection mrs;
+    MemoryRegion *mr;
+    uint64_t offset = 0;
+    size_t remaining_size;
+
+    if (!ct3d->hostmem) {
+        error_setg(errp, "memdev property must be set");
+        return;
+    }
+
+    /* FIXME: need to check mr is the host bridge's MR */
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+
+    /* Create our new subregion */
+    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
+
+    /* Find the first free space in the window */
+    WITH_RCU_READ_LOCK_GUARD()
+    {
+        mrs = memory_region_find(mr, offset, 1);
+        while (mrs.mr && mrs.mr != mr) {
+            offset += memory_region_size(mrs.mr);
+            mrs = memory_region_find(mr, offset, 1);
+        }
+    }
+
+    remaining_size = memory_region_size(mr) - offset;
+    if (remaining_size < ct3d->size) {
+        g_free(ct3d->cxl_dstate.pmem);
+        error_setg(errp,
+                   "Not enough free space (%zd) required for device (%" PRId64  ")",
+                   remaining_size, ct3d->size);
+    }
+
+    /* Register our subregion as non-volatile */
+    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
+                           "cxl_type3-memory", ct3d->size, errp);
+    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
+
+#ifdef SET_PMEM_PADDR
+    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
+#endif
+}
+
+static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
+                                              Error **errp)
+{
+    CXLType3Dev *ct3d = CT3(md);
+
+    if (!ct3d->cxl_dstate.pmem) {
+        cxl_setup_memory(ct3d, errp);
+    }
+
+    return ct3d->cxl_dstate.pmem;
+}
+
+static void ct3_realize(PCIDevice *pci_dev, Error **errp)
+{
+    CXLType3Dev *ct3d = CT3(pci_dev);
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    ComponentRegisters *regs = &cxl_cstate->crb;
+    MemoryRegion *mr = &regs->component_registers;
+    uint8_t *pci_conf = pci_dev->config;
+
+    if (!ct3d->cxl_dstate.pmem) {
+        cxl_setup_memory(ct3d, errp);
+    }
+
+    pci_config_set_prog_interface(pci_conf, 0x10);
+    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
+
+    pcie_endpoint_cap_init(pci_dev, 0x80);
+    cxl_cstate->dvsec_offset = 0x100;
+
+    ct3d->cxl_cstate.pdev = pci_dev;
+    build_dvsecs(ct3d);
+
+    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
+                                      TYPE_CXL_TYPE3_DEV);
+
+    pci_register_bar(
+        pci_dev, COMPONENT_REG_BAR_IDX,
+        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
+
+    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
+    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY |
+                         PCI_BASE_ADDRESS_MEM_TYPE_64,
+                     &ct3d->cxl_dstate.device_registers);
+}
+
+static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
+{
+    CXLType3Dev *ct3d = CT3(md);
+
+    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
+}
+
+static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
+{
+    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
+}
+
+static void ct3d_reset(DeviceState *dev)
+{
+    CXLType3Dev *ct3d = CT3(dev);
+    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
+    cxl_device_register_init_common(&ct3d->cxl_dstate);
+}
+
+static Property ct3_props[] = {
+    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
+    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
+                     HostMemoryBackend *),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
+                                        MemoryDeviceInfo *info)
+{
+    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
+    const DeviceClass *dc = DEVICE_GET_CLASS(md);
+    const DeviceState *dev = DEVICE(md);
+    CXLType3Dev *ct3d = CT3(md);
+
+    if (dev->id) {
+        di->has_id = true;
+        di->id = g_strdup(dev->id);
+    }
+    di->hotplugged = dev->hotplugged;
+    di->hotpluggable = dc->hotpluggable;
+    di->addr = cxl_md_get_addr(md);
+    di->slot = 0;
+    di->node = 0;
+    di->size = memory_device_get_region_size(md, NULL);
+    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
+
+
+    info->u.cxl.data = di;
+    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
+}
+
+static void ct3_class_init(ObjectClass *oc, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(oc);
+    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
+    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
+
+    pc->realize = ct3_realize;
+    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
+    pc->vendor_id = PCI_VENDOR_ID_INTEL;
+    pc->device_id = 0xd93; /* LVF for now */
+    pc->revision = 1;
+
+    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
+    dc->desc = "CXL PMEM Device (Type 3)";
+    dc->reset = ct3d_reset;
+    device_class_set_props(dc, ct3_props);
+
+    mdc->get_memory_region = cxl_md_get_memory_region;
+    mdc->get_addr = cxl_md_get_addr;
+    mdc->fill_device_info = pc_dimm_md_fill_device_info;
+    mdc->get_plugged_size = memory_device_get_region_size;
+    mdc->set_addr = cxl_md_set_addr;
+}
+
+static const TypeInfo ct3d_info = {
+    .name = TYPE_CXL_TYPE3_DEV,
+    .parent = TYPE_PCI_DEVICE,
+    .class_init = ct3_class_init,
+    .instance_size = sizeof(CXLType3Dev),
+    .instance_init = ct3_instance_init,
+    .instance_finalize = ct3_finalize,
+    .interfaces = (InterfaceInfo[]) {
+        { TYPE_MEMORY_DEVICE },
+        { INTERFACE_CXL_DEVICE },
+        { INTERFACE_PCIE_DEVICE },
+        {}
+    },
+};
+
+static void ct3d_registers(void)
+{
+    type_register_static(&ct3d_info);
+}
+
+type_init(ct3d_registers);
diff --git a/hw/mem/meson.build b/hw/mem/meson.build
index 0d22f2b572..d13c3ed117 100644
--- a/hw/mem/meson.build
+++ b/hw/mem/meson.build
@@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
 mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
 mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
 mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
+mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
 
 softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
index d4010cf8f3..1ecf6f6a55 100644
--- a/hw/pci/pcie.c
+++ b/hw/pci/pcie.c
@@ -20,6 +20,7 @@
 
 #include "qemu/osdep.h"
 #include "qapi/error.h"
+#include "hw/mem/memory-device.h"
 #include "hw/pci/pci_bridge.h"
 #include "hw/pci/pcie.h"
 #include "hw/pci/msix.h"
@@ -27,6 +28,8 @@
 #include "hw/pci/pci_bus.h"
 #include "hw/pci/pcie_regs.h"
 #include "hw/pci/pcie_port.h"
+#include "hw/cxl/cxl.h"
+#include "hw/boards.h"
 #include "qemu/range.h"
 
 //#define DEBUG_PCIE
@@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
     }
 
     pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
+
+#ifdef CXL_MEM_DEVICE
+    /*
+     * FIXME:
+     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
+     *    HotplugHandler *hotplug_ctrl;
+     *   Error *local_err = NULL;
+     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
+     *  if (hotplug_ctrl) {
+     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
+     *      if (local_err) {
+     *          error_propagate(errp, local_err);
+     *          return;
+     *      }
+     *  }
+     */
+
+    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
+        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
+                               NULL, errp);
+    }
+#endif
 }
 
 void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
@@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
         pcie_cap_slot_event(hotplug_pdev,
                             PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
     }
+
+#ifdef CXL_MEM_DEVICE
+    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
+        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
+#endif
 }
 
 void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
index b1e5f4a8fa..809ed7de60 100644
--- a/include/hw/cxl/cxl.h
+++ b/include/hw/cxl/cxl.h
@@ -17,6 +17,8 @@
 #define COMPONENT_REG_BAR_IDX 0
 #define DEVICE_REG_BAR_IDX 2
 
+#define TYPE_CXL_TYPE3_DEV "cxl-type3"
+
 #define CXL_HOST_BASE 0xD0000000
 #define CXL_WINDOW_MAX 10
 
diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
index a53c2e5ae7..9ec28c9feb 100644
--- a/include/hw/cxl/cxl_pci.h
+++ b/include/hw/cxl/cxl_pci.h
@@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
  * CXL 2.0 Downstream Port: 3, 4, 7, 8
  */
 
+/* CXL 2.0 - 8.1.3 (ID 0001) */
+struct dvsec_device {
+    struct dvsec_header hdr;
+    uint16_t cap;
+    uint16_t ctrl;
+    uint16_t status;
+    uint16_t ctrl2;
+    uint16_t status2;
+    uint16_t lock;
+    uint16_t cap2;
+    uint32_t range1_size_hi;
+    uint32_t range1_size_lo;
+    uint32_t range1_base_hi;
+    uint32_t range1_base_lo;
+    uint32_t range2_size_hi;
+    uint32_t range2_size_lo;
+    uint32_t range2_base_hi;
+    uint32_t range2_base_lo;
+};
+_Static_assert(sizeof(struct dvsec_device) == 0x38,
+               "dvsec device size incorrect");
+
 /* CXL 2.0 - 8.1.5 (ID 0003) */
 struct extensions_dvsec_port {
     struct dvsec_header hdr;
diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
index 11f8ab7149..76bf3ed590 100644
--- a/include/hw/pci/pci_ids.h
+++ b/include/hw/pci/pci_ids.h
@@ -53,6 +53,7 @@
 #define PCI_BASE_CLASS_MEMORY            0x05
 #define PCI_CLASS_MEMORY_RAM             0x0500
 #define PCI_CLASS_MEMORY_FLASH           0x0501
+#define PCI_CLASS_MEMORY_CXL             0x0502
 #define PCI_CLASS_MEMORY_OTHER           0x0580
 
 #define PCI_BASE_CLASS_BRIDGE            0x06
diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
index 0dd594f92b..0f67bc61ce 100644
--- a/monitor/hmp-cmds.c
+++ b/monitor/hmp-cmds.c
@@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
                 monitor_printf(mon, "  hotpluggable: %s\n",
                                di->hotpluggable ? "true" : "false");
                 break;
+            case MEMORY_DEVICE_INFO_KIND_CXL:
+                di = value->u.cxl.data;
+                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
+                               MemoryDeviceInfoKind_str(value->type),
+                               di->id ? di->id : "");
+                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
+                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
+                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
+                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
+                monitor_printf(mon, "  memdev: %s\n", di->memdev);
+                monitor_printf(mon, "  hotplugged: %s\n",
+                               di->hotplugged ? "true" : "false");
+                monitor_printf(mon, "  hotpluggable: %s\n",
+                               di->hotpluggable ? "true" : "false");
+                break;
             case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
                 vpi = value->u.virtio_pmem.data;
                 monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
diff --git a/qapi/machine.json b/qapi/machine.json
index 330189efe3..aa96d662bd 100644
--- a/qapi/machine.json
+++ b/qapi/machine.json
@@ -1394,6 +1394,7 @@
 { 'union': 'MemoryDeviceInfo',
   'data': { 'dimm': 'PCDIMMDeviceInfo',
             'nvdimm': 'PCDIMMDeviceInfo',
+            'cxl': 'PCDIMMDeviceInfo',
             'virtio-pmem': 'VirtioPMEMDeviceInfo',
             'virtio-mem': 'VirtioMEMDeviceInfo'
           }
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 25/32] hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (23 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 26/32] acpi/cxl: Add _OSC implementation (9.14.2) Ben Widawsky
                   ` (7 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

A device's volatile and persistent memory are known Host Defined Memory
(HDM) regions. The mechanism by which the device is programmed to claim
the addresses associated with those regions is through dedicated logic
known as the HDM decoder. In order to allow the OS to properly program
the HDMs, the HDM decoders must be modeled.

There are two ways the HDM decoders can be implemented, the legacy
mechanism is through the PCIe DVSEC programming from CXL 1.1 (8.1.3.8),
and MMIO is found in 8.2.5.12 of the spec. For now, 8.1.3.8 is not
implemented.

Much of CXL device logic is implemented in cxl-utils. The HDM decoder
however is implemented directly by the device implementation. The
generic cxl-utils probably should be the correct place to put this since
HDM decoders aren't unique to a type3 device. It is however easier at
the moment, and requires less design consideration to simply implement
it in the device, and figure out how to consolidate it later.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/mem/cxl_type3.c | 82 +++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 77 insertions(+), 5 deletions(-)

diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
index 3985bb8d0b..dee5a8884b 100644
--- a/hw/mem/cxl_type3.c
+++ b/hw/mem/cxl_type3.c
@@ -57,6 +57,71 @@ static void build_dvsecs(CXLType3Dev *ct3d)
                                REG_LOC_DVSEC_REVID, dvsec);
 }
 
+static void hdm_decoder_commit(CXLType3Dev *ct3d, int which)
+{
+    MemoryRegion *pmem = ct3d->cxl_dstate.pmem;
+    MemoryRegion *mr = host_memory_backend_get_memory(ct3d->hostmem);
+    Range window, device;
+    ComponentRegisters *cregs = &ct3d->cxl_cstate.crb;
+    uint32_t *cache_mem = cregs->cache_mem_registers;
+    uint64_t offset, size;
+    Error *err = NULL;
+
+    assert(which == 0);
+
+    ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, COMMIT, 0);
+    ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, ERROR, 0);
+
+    offset = ((uint64_t)cache_mem[R_CXL_HDM_DECODER0_BASE_HI] << 32) |
+             cache_mem[R_CXL_HDM_DECODER0_BASE_LO];
+    size = ((uint64_t)cache_mem[R_CXL_HDM_DECODER0_SIZE_HI] << 32) |
+           cache_mem[R_CXL_HDM_DECODER0_SIZE_LO];
+
+    range_init_nofail(&window, mr->addr, memory_region_size(mr));
+    range_init_nofail(&device, offset, size);
+
+    if (!range_contains_range(&window, &device)) {
+        ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, ERROR, 1);
+        return;
+    }
+
+    memory_region_ram_resize(pmem, size, &err);
+    if (err) {
+        ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, ERROR, 1);
+        return;
+    }
+
+    offset -= mr->addr;
+    memory_region_add_subregion(mr, offset, pmem);
+
+    ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, COMMITTED, 1);
+}
+
+static void ct3d_reg_write(void *opaque, hwaddr offset, uint64_t value, unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    CXLType3Dev *ct3d = container_of(cxl_cstate, CXLType3Dev, cxl_cstate);
+    uint32_t *cache_mem = cregs->cache_mem_registers;
+    bool should_commit = false;
+    int which_hdm = -1;
+
+    assert(size == 4);
+
+    switch (offset) {
+    case A_CXL_HDM_DECODER0_CTRL:
+        should_commit = FIELD_EX32(value, CXL_HDM_DECODER0_CTRL, COMMIT);
+        which_hdm = 0;
+        break;
+    default:
+        break;
+    }
+
+    stl_le_p((uint8_t *)cache_mem + offset, value);
+    if (should_commit)
+        hdm_decoder_commit(ct3d, which_hdm);
+}
+
 static void ct3_instance_init(Object *obj)
 {
     /* MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); */
@@ -65,7 +130,10 @@ static void ct3_instance_init(Object *obj)
 static void ct3_finalize(Object *obj)
 {
     CXLType3Dev *ct3d = CT3(obj);
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    ComponentRegisters *regs = &cxl_cstate->crb;
 
+    g_free((void *)regs->special_ops);
     g_free(ct3d->cxl_dstate.pmem);
 }
 
@@ -81,11 +149,12 @@ static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
         return;
     }
 
-    /* FIXME: need to check mr is the host bridge's MR */
-    mr = host_memory_backend_get_memory(ct3d->hostmem);
-
     /* Create our new subregion */
     ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
+    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
+
+    /* FIXME: need to check mr is the host bridge's MR */
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
 
     /* Find the first free space in the window */
     WITH_RCU_READ_LOCK_GUARD()
@@ -108,8 +177,6 @@ static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
     /* Register our subregion as non-volatile */
     memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
                            "cxl_type3-memory", ct3d->size, errp);
-    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
-
 #ifdef SET_PMEM_PADDR
     memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
 #endif
@@ -148,6 +215,11 @@ static void ct3_realize(PCIDevice *pci_dev, Error **errp)
     ct3d->cxl_cstate.pdev = pci_dev;
     build_dvsecs(ct3d);
 
+#ifndef SET_PMEM_PADDR
+    regs->special_ops = g_new0(MemoryRegionOps, 1);
+    regs->special_ops->write = ct3d_reg_write;
+#endif
+
     cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
                                       TYPE_CXL_TYPE3_DEV);
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 26/32] acpi/cxl: Add _OSC implementation (9.14.2)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (24 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 25/32] hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 27/32] tests/acpi: allow CEDT table addition Ben Widawsky
                   ` (6 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

CXL 2.0 specification adds 2 new dwords to the existing _OSC definition
from PCIe. The new dwords are accessed with a new uuid. This
implementation supports what is in the specification.

We are currently in the process of trying to define a new definition for
_OSC. See later work for an explanation.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/acpi/Kconfig       |   5 ++
 hw/acpi/cxl.c         | 104 ++++++++++++++++++++++++++++++++++++++++++
 hw/acpi/meson.build   |   1 +
 hw/i386/acpi-build.c  |  12 ++++-
 include/hw/acpi/cxl.h |  23 ++++++++++
 5 files changed, 144 insertions(+), 1 deletion(-)
 create mode 100644 hw/acpi/cxl.c
 create mode 100644 include/hw/acpi/cxl.h

diff --git a/hw/acpi/Kconfig b/hw/acpi/Kconfig
index 1932f66af8..b27907953e 100644
--- a/hw/acpi/Kconfig
+++ b/hw/acpi/Kconfig
@@ -5,6 +5,7 @@ config ACPI_X86
     bool
     select ACPI
     select ACPI_NVDIMM
+    select ACPI_CXL
     select ACPI_CPU_HOTPLUG
     select ACPI_MEMORY_HOTPLUG
     select ACPI_HMAT
@@ -42,3 +43,7 @@ config ACPI_VMGENID
     depends on PC
 
 config ACPI_HW_REDUCED
+
+config ACPI_CXL
+    bool
+    depends on ACPI
diff --git a/hw/acpi/cxl.c b/hw/acpi/cxl.c
new file mode 100644
index 0000000000..7124d5a1a3
--- /dev/null
+++ b/hw/acpi/cxl.c
@@ -0,0 +1,104 @@
+/*
+ * CXL ACPI Implementation
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>
+ */
+
+#include "qemu/osdep.h"
+#include "hw/cxl/cxl.h"
+#include "hw/acpi/acpi.h"
+#include "hw/acpi/aml-build.h"
+#include "hw/acpi/bios-linker-loader.h"
+#include "hw/acpi/cxl.h"
+#include "qapi/error.h"
+#include "qemu/uuid.h"
+
+static Aml *__build_cxl_osc_method(void)
+{
+    Aml *method, *if_uuid, *else_uuid, *if_arg1_not_1, *if_cxl, *if_caps_masked;
+    Aml *a_ctrl = aml_local(0);
+    Aml *a_cdw1 = aml_name("CDW1");
+
+    method = aml_method("_OSC", 4, AML_NOTSERIALIZED);
+    aml_append(method, aml_create_dword_field(aml_arg(3), aml_int(0), "CDW1"));
+
+    /* 9.14.2.1.4 */
+    if_uuid = aml_if(
+        aml_lor(aml_equal(aml_arg(0),
+                          aml_touuid("33DB4D5B-1FF7-401C-9657-7441C03DD766")),
+                aml_equal(aml_arg(0),
+                          aml_touuid("68F2D50B-C469-4D8A-BD3D-941A103FD3FC"))));
+    aml_append(if_uuid, aml_create_dword_field(aml_arg(3), aml_int(4), "CDW2"));
+    aml_append(if_uuid, aml_create_dword_field(aml_arg(3), aml_int(8), "CDW3"));
+
+    aml_append(if_uuid, aml_store(aml_name("CDW3"), a_ctrl));
+
+    /* This is all the same as what's used for PCIe */
+    aml_append(if_uuid,
+               aml_and(aml_name("CTRL"), aml_int(0x1F), aml_name("CTRL")));
+
+    if_arg1_not_1 = aml_if(aml_lnot(aml_equal(aml_arg(1), aml_int(0x1))));
+    /* Unknown revision */
+    aml_append(if_arg1_not_1, aml_or(a_cdw1, aml_int(0x08), a_cdw1));
+    aml_append(if_uuid, if_arg1_not_1);
+
+    if_caps_masked = aml_if(aml_lnot(aml_equal(aml_name("CDW3"), a_ctrl)));
+    /* Capability bits were masked */
+    aml_append(if_caps_masked, aml_or(a_cdw1, aml_int(0x10), a_cdw1));
+    aml_append(if_uuid, if_caps_masked);
+
+    aml_append(if_uuid, aml_store(aml_name("CDW2"), aml_name("SUPP")));
+    aml_append(if_uuid, aml_store(aml_name("CDW3"), aml_name("CTRL")));
+
+    if_cxl = aml_if(aml_equal(
+        aml_arg(0), aml_touuid("68F2D50B-C469-4D8A-BD3D-941A103FD3FC")));
+    /* CXL support field */
+    aml_append(if_cxl, aml_create_dword_field(aml_arg(3), aml_int(12), "CDW4"));
+    /* CXL capabilities */
+    aml_append(if_cxl, aml_create_dword_field(aml_arg(3), aml_int(16), "CDW5"));
+    aml_append(if_cxl, aml_store(aml_name("CDW4"), aml_name("SUPC")));
+    aml_append(if_cxl, aml_store(aml_name("CDW5"), aml_name("CTRC")));
+
+    /* CXL 2.0 Port/Device Register access */
+    aml_append(if_cxl,
+               aml_or(aml_name("CDW5"), aml_int(0x1), aml_name("CDW5")));
+    aml_append(if_uuid, if_cxl);
+
+    /* Update DWORD3 (the return value) */
+    aml_append(if_uuid, aml_store(a_ctrl, aml_name("CDW3")));
+
+    aml_append(if_uuid, aml_return(aml_arg(3)));
+    aml_append(method, if_uuid);
+
+    else_uuid = aml_else();
+
+    /* unrecognized uuid */
+    aml_append(else_uuid,
+               aml_or(aml_name("CDW1"), aml_int(0x4), aml_name("CDW1")));
+    aml_append(else_uuid, aml_return(aml_arg(3)));
+    aml_append(method, else_uuid);
+
+    return method;
+}
+
+void build_cxl_osc_method(Aml *dev)
+{
+    aml_append(dev, aml_name_decl("SUPP", aml_int(0)));
+    aml_append(dev, aml_name_decl("CTRL", aml_int(0)));
+    aml_append(dev, aml_name_decl("SUPC", aml_int(0)));
+    aml_append(dev, aml_name_decl("CTRC", aml_int(0)));
+    aml_append(dev, __build_cxl_osc_method());
+}
diff --git a/hw/acpi/meson.build b/hw/acpi/meson.build
index dd69577212..9f5c5ced28 100644
--- a/hw/acpi/meson.build
+++ b/hw/acpi/meson.build
@@ -10,6 +10,7 @@ acpi_ss.add(when: 'CONFIG_ACPI_CPU_HOTPLUG', if_true: files('cpu_hotplug.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_MEMORY_HOTPLUG', if_true: files('memory_hotplug.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_NVDIMM', if_true: files('nvdimm.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_PCI', if_true: files('pci.c'))
+acpi_ss.add(when: 'CONFIG_ACPI_CXL', if_true: files('cxl.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_VMGENID', if_true: files('vmgenid.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_HW_REDUCED', if_true: files('generic_event_device.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_HMAT', if_true: files('hmat.c'))
diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 16cde677a0..3eb07b9741 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -66,6 +66,7 @@
 #include "hw/acpi/aml-build.h"
 #include "hw/acpi/utils.h"
 #include "hw/acpi/pci.h"
+#include "hw/acpi/cxl.h"
 
 #include "qom/qom-qobject.h"
 #include "hw/i386/amd_iommu.h"
@@ -1201,11 +1202,20 @@ static void init_pci_acpi(Aml *dev, int uid, int type)
     if (type == PCI) {
         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
         aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
-    } else {
+    } else if (type == PCIE) {
         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
         aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
         aml_append(dev, build_q35_osc_method());
+    } else /* CXL */ {
+        struct Aml *pkg = aml_package(2);
+
+        aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0016")));
+        aml_append(pkg, aml_eisaid("PNP0A08"));
+        aml_append(pkg, aml_eisaid("PNP0A03"));
+        aml_append(dev, aml_name_decl("_CID", pkg));
+        aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
+        build_cxl_osc_method(dev);
     }
 }
 
diff --git a/include/hw/acpi/cxl.h b/include/hw/acpi/cxl.h
new file mode 100644
index 0000000000..7b8f3b8a2e
--- /dev/null
+++ b/include/hw/acpi/cxl.h
@@ -0,0 +1,23 @@
+/*
+ * Copyright (C) 2020 Intel 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/>.
+ */
+
+#ifndef HW_ACPI_CXL_H
+#define HW_ACPI_CXL_H
+
+void build_cxl_osc_method(Aml *dev);
+
+#endif
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 27/32] tests/acpi: allow CEDT table addition
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (25 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 26/32] acpi/cxl: Add _OSC implementation (9.14.2) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 28/32] acpi/cxl: Create the CEDT (9.14.1) Ben Widawsky
                   ` (5 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 tests/data/acpi/pc/CEDT                     | 0
 tests/data/acpi/q35/CEDT                    | 0
 tests/qtest/bios-tables-test-allowed-diff.h | 2 ++
 3 files changed, 2 insertions(+)
 create mode 100644 tests/data/acpi/pc/CEDT
 create mode 100644 tests/data/acpi/q35/CEDT

diff --git a/tests/data/acpi/pc/CEDT b/tests/data/acpi/pc/CEDT
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/tests/data/acpi/q35/CEDT b/tests/data/acpi/q35/CEDT
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/tests/qtest/bios-tables-test-allowed-diff.h b/tests/qtest/bios-tables-test-allowed-diff.h
index dfb8523c8b..9b07f1e1ff 100644
--- a/tests/qtest/bios-tables-test-allowed-diff.h
+++ b/tests/qtest/bios-tables-test-allowed-diff.h
@@ -1 +1,3 @@
 /* List of comma-separated changed AML files to ignore */
+"tests/data/acpi/pc/CEDT",
+"tests/data/acpi/q35/CEDT",
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 28/32] acpi/cxl: Create the CEDT (9.14.1)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (26 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 27/32] tests/acpi: allow CEDT table addition Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 29/32] Temp: acpi/cxl: Add ACPI0017 (CEDT awareness) Ben Widawsky
                   ` (4 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

The CXL Early Discovery Table is defined in the CXL 2.0 specification as
a way for the OS to get CXL specific information from the system
firmware.

As of CXL 2.0 spec, only 1 sub structure is defined, the CXL Host Bridge
Structure (CHBS) which is primarily useful for telling the OS exactly
where the MMIO for the host bridge is.

v2: Update CHBS to spec released definition

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/acpi/cxl.c                       | 69 +++++++++++++++++++++++++++++
 hw/i386/acpi-build.c                |  6 ++-
 hw/pci-bridge/pci_expander_bridge.c | 21 +--------
 include/hw/acpi/cxl.h               |  4 ++
 include/hw/pci/pci_bridge.h         | 25 +++++++++++
 5 files changed, 104 insertions(+), 21 deletions(-)

diff --git a/hw/acpi/cxl.c b/hw/acpi/cxl.c
index 7124d5a1a3..68db0fe3a8 100644
--- a/hw/acpi/cxl.c
+++ b/hw/acpi/cxl.c
@@ -18,14 +18,83 @@
  */
 
 #include "qemu/osdep.h"
+#include "hw/sysbus.h"
+#include "hw/pci/pci_bridge.h"
+#include "hw/pci/pci_host.h"
 #include "hw/cxl/cxl.h"
+#include "hw/mem/memory-device.h"
 #include "hw/acpi/acpi.h"
 #include "hw/acpi/aml-build.h"
 #include "hw/acpi/bios-linker-loader.h"
 #include "hw/acpi/cxl.h"
+#include "hw/acpi/cxl.h"
 #include "qapi/error.h"
 #include "qemu/uuid.h"
 
+static void cedt_build_chbs(GArray *table_data, PXBDev *cxl)
+{
+    SysBusDevice *sbd = SYS_BUS_DEVICE(cxl->cxl.cxl_host_bridge);
+    struct MemoryRegion *mr = sbd->mmio[0].memory;
+
+    /* Type */
+    build_append_int_noprefix(table_data, 0, 1);
+
+    /* Reserved */
+    build_append_int_noprefix(table_data, 0, 1);
+
+    /* Record Length */
+    build_append_int_noprefix(table_data, 32, 2);
+
+    /* UID */
+    build_append_int_noprefix(table_data, cxl->uid, 4);
+
+    /* Version */
+    build_append_int_noprefix(table_data, 1, 4);
+
+    /* Reserved */
+    build_append_int_noprefix(table_data, 0, 4);
+
+    /* Base */
+    build_append_int_noprefix(table_data, mr->addr, 8);
+
+    /* Length */
+    build_append_int_noprefix(table_data, memory_region_size(mr), 8);
+}
+
+static int cxl_foreach_pxb_hb(Object *obj, void *opaque)
+{
+    Aml *cedt = opaque;
+
+    if (object_dynamic_cast(obj, TYPE_PXB_CXL_DEVICE)) {
+        PXBDev *pxb = PXB_CXL_DEV(obj);
+
+        cedt_build_chbs(cedt->buf, pxb);
+    }
+
+    return 0;
+}
+
+void cxl_build_cedt(GArray *table_offsets, GArray *table_data,
+                    BIOSLinker *linker)
+{
+    const int cedt_start = table_data->len;
+    Aml *cedt;
+
+    cedt = init_aml_allocator();
+
+    /* reserve space for CEDT header */
+    acpi_add_table(table_offsets, table_data);
+    acpi_data_push(cedt->buf, sizeof(AcpiTableHeader));
+
+    object_child_foreach_recursive(object_get_root(), cxl_foreach_pxb_hb, cedt);
+
+    /* copy AML table into ACPI tables blob and patch header there */
+    g_array_append_vals(table_data, cedt->buf->data, cedt->buf->len);
+    build_header(linker, table_data, (void *)(table_data->data + cedt_start),
+                 "CEDT", table_data->len - cedt_start, 1, NULL, NULL);
+    free_aml_allocator();
+}
+
 static Aml *__build_cxl_osc_method(void)
 {
     Aml *method, *if_uuid, *else_uuid, *if_arg1_not_1, *if_cxl, *if_caps_masked;
diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 3eb07b9741..49242eb8f3 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -75,6 +75,8 @@
 #include "hw/acpi/ipmi.h"
 #include "hw/acpi/hmat.h"
 
+#include "hw/acpi/cxl.h"
+
 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
  * a little bit, there should be plenty of free space since the DSDT
@@ -1371,7 +1373,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
 
             scope = aml_scope("\\_SB");
             if (type == CXL) {
-                dev = aml_device("CXL%.01X", pci_bus_uid(bus));
+                dev = aml_device("CXL%.01X", uid);
             } else {
                 dev = aml_device("PC%.02X", bus_num);
             }
@@ -2277,6 +2279,8 @@ void acpi_build(AcpiBuildTables *tables, MachineState *machine)
                           machine->nvdimms_state, machine->ram_slots);
     }
 
+    cxl_build_cedt(table_offsets, tables_blob, tables->linker);
+
     acpi_add_table(table_offsets, tables_blob);
     build_waet(tables_blob, tables->linker);
 
diff --git a/hw/pci-bridge/pci_expander_bridge.c b/hw/pci-bridge/pci_expander_bridge.c
index af1450c69d..6458d5b76e 100644
--- a/hw/pci-bridge/pci_expander_bridge.c
+++ b/hw/pci-bridge/pci_expander_bridge.c
@@ -57,26 +57,6 @@ DECLARE_INSTANCE_CHECKER(PXBDev, PXB_DEV,
 DECLARE_INSTANCE_CHECKER(PXBDev, PXB_PCIE_DEV,
                          TYPE_PXB_PCIE_DEVICE)
 
-#define TYPE_PXB_CXL_DEVICE "pxb-cxl"
-DECLARE_INSTANCE_CHECKER(PXBDev, PXB_CXL_DEV,
-                         TYPE_PXB_CXL_DEVICE)
-
-struct PXBDev {
-    /*< private >*/
-    PCIDevice parent_obj;
-    /*< public >*/
-
-    uint8_t bus_nr;
-    uint16_t numa_node;
-    int32_t uid;
-    struct cxl_dev {
-        HostMemoryBackend *memory_window[CXL_WINDOW_MAX];
-
-        uint32_t num_windows;
-        hwaddr *window_base[CXL_WINDOW_MAX];
-    } cxl;
-};
-
 typedef struct CXLHost {
     PCIHostState parent_obj;
 
@@ -351,6 +331,7 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_CXL_BUS);
         bus->flags |= PCI_BUS_CXL;
         PXB_CXL_HOST(ds)->dev = PXB_CXL_DEV(dev);
+        PXB_CXL_DEV(dev)->cxl.cxl_host_bridge = ds;
     } else {
         bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
         bds = qdev_new("pci-bridge");
diff --git a/include/hw/acpi/cxl.h b/include/hw/acpi/cxl.h
index 7b8f3b8a2e..db2063f8c9 100644
--- a/include/hw/acpi/cxl.h
+++ b/include/hw/acpi/cxl.h
@@ -18,6 +18,10 @@
 #ifndef HW_ACPI_CXL_H
 #define HW_ACPI_CXL_H
 
+#include "hw/acpi/bios-linker-loader.h"
+
+void cxl_build_cedt(GArray *table_offsets, GArray *table_data,
+                    BIOSLinker *linker);
 void build_cxl_osc_method(Aml *dev);
 
 #endif
diff --git a/include/hw/pci/pci_bridge.h b/include/hw/pci/pci_bridge.h
index a94d350034..50dd7fdf33 100644
--- a/include/hw/pci/pci_bridge.h
+++ b/include/hw/pci/pci_bridge.h
@@ -28,6 +28,7 @@
 
 #include "hw/pci/pci.h"
 #include "hw/pci/pci_bus.h"
+#include "hw/cxl/cxl.h"
 #include "qom/object.h"
 
 typedef struct PCIBridgeWindows PCIBridgeWindows;
@@ -81,6 +82,30 @@ struct PCIBridge {
 #define PCI_BRIDGE_DEV_PROP_MSI        "msi"
 #define PCI_BRIDGE_DEV_PROP_SHPC       "shpc"
 
+struct PXBDev {
+    /*< private >*/
+    PCIDevice parent_obj;
+    /*< public >*/
+
+    uint8_t bus_nr;
+    uint16_t numa_node;
+    int32_t uid;
+
+    struct cxl_dev {
+        HostMemoryBackend *memory_window[CXL_WINDOW_MAX];
+
+        uint32_t num_windows;
+        hwaddr *window_base[CXL_WINDOW_MAX];
+
+        void *cxl_host_bridge; /* Pointer to a CXLHost */
+    } cxl;
+};
+
+typedef struct PXBDev PXBDev;
+#define TYPE_PXB_CXL_DEVICE "pxb-cxl"
+DECLARE_INSTANCE_CHECKER(PXBDev, PXB_CXL_DEV,
+                         TYPE_PXB_CXL_DEVICE)
+
 int pci_bridge_ssvid_init(PCIDevice *dev, uint8_t offset,
                           uint16_t svid, uint16_t ssid,
                           Error **errp);
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 29/32] Temp: acpi/cxl: Add ACPI0017 (CEDT awareness)
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (27 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 28/32] acpi/cxl: Create the CEDT (9.14.1) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 30/32] tests/acpi: Add new CEDT files Ben Widawsky
                   ` (3 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This represents Intel's proposal for how the system firmware can notify
Linux that the CEDT exists and provides a driver attach point. It is not
in the CXL 2.0 specification as of now.

CXL 2.0 specification adds an _HID, ACPI0016, for CXL capable host
bridges, with a _CID of PNP0A08 (PCIe host bridge). CXL aware software
is able to use this initiate the proper _OSC method, and get the _UID
which is referenced by the CEDT. Therefore the existence of an ACPI0016
device allows a CXL aware driver perform the necessary actions. For a
CXL capable OS, this works. For a CXL unaware OS, this works.

The motivation for ACPI0017 is to provide the possibility of having a
Linux CXL module that can work on a legacy Linux kernel.  Linux core
PCI/ACPI which won't be built as a module, will see the _CID of PNP0A08
and bind a driver to it. If we later loaded a driver for ACPI0016, Linux
won't be able to bind it to the hardware because it has already bound
the PNP0A08 driver. The ACPI0017 device is an opportunity to have an
object to bind a driver will be used by a Linux driver to walk the CXL
topology and do everything that we would have preferred to do with
ACPI0016.

There is another motivation for an ACPI0017 device which isn't
implemented here. An operating system needs an attach point for a
non-volatile region provider that understands cross-hostbridge
interleaving. Since QEMU emulation doesn't support interleaving yet,
this is more important on the OS side, for now.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 hw/i386/acpi-build.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 49242eb8f3..470dd7ddf4 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -1221,6 +1221,19 @@ static void init_pci_acpi(Aml *dev, int uid, int type)
     }
 }
 
+static void build_acpi0017(Aml *table)
+{
+    Aml *dev;
+    Aml *scope;
+
+    scope =  aml_scope("_SB");
+    dev = aml_device("CXLM");
+    aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0017")));
+
+    aml_append(scope, dev);
+    aml_append(table, scope);
+}
+
 static void
 build_dsdt(GArray *table_data, BIOSLinker *linker,
            AcpiPmInfo *pm, AcpiMiscInfo *misc,
@@ -1237,6 +1250,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
     int root_bus_limit = 0xFF;
     PCIBus *bus = NULL;
     TPMIf *tpm = tpm_find();
+    bool cxl_present = false;
     int i;
     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
 
@@ -1392,6 +1406,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
 
             /* Handle the ranges for the PXB expanders */
             if (type == CXL) {
+                cxl_present = true;
                 uint64_t base = CXL_HOST_BASE + uid * 0x10000;
                 crs_range_insert(crs_range_set.mem_ranges, base,
                                  base + 0x10000 - 1);
@@ -1399,6 +1414,10 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
         }
     }
 
+    if (cxl_present) {
+        build_acpi0017(dsdt);
+    }
+
     /*
      * At this point crs_range_set has all the ranges used by pci
      * busses *other* than PCI0.  These ranges will be excluded from
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 30/32] tests/acpi: Add new CEDT files
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (28 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 29/32] Temp: acpi/cxl: Add ACPI0017 (CEDT awareness) Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 31/32] WIP: i386/cxl: Initialize a host bridge Ben Widawsky
                   ` (2 subsequent siblings)
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 tests/data/acpi/pc/CEDT                     | Bin 0 -> 36 bytes
 tests/data/acpi/q35/CEDT                    | Bin 0 -> 36 bytes
 tests/qtest/bios-tables-test-allowed-diff.h |   2 --
 3 files changed, 2 deletions(-)

diff --git a/tests/data/acpi/pc/CEDT b/tests/data/acpi/pc/CEDT
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..ebf9b54b0b27d9efca53359c3c2e560511f0e165 100644
GIT binary patch
literal 36
kcmZ>EbqP^nU|?X};N<V@5v<@85#a0$6k`O6f!H7#0Fb2z0RR91

literal 0
HcmV?d00001

diff --git a/tests/data/acpi/q35/CEDT b/tests/data/acpi/q35/CEDT
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..ebf9b54b0b27d9efca53359c3c2e560511f0e165 100644
GIT binary patch
literal 36
kcmZ>EbqP^nU|?X};N<V@5v<@85#a0$6k`O6f!H7#0Fb2z0RR91

literal 0
HcmV?d00001

diff --git a/tests/qtest/bios-tables-test-allowed-diff.h b/tests/qtest/bios-tables-test-allowed-diff.h
index 9b07f1e1ff..dfb8523c8b 100644
--- a/tests/qtest/bios-tables-test-allowed-diff.h
+++ b/tests/qtest/bios-tables-test-allowed-diff.h
@@ -1,3 +1 @@
 /* List of comma-separated changed AML files to ignore */
-"tests/data/acpi/pc/CEDT",
-"tests/data/acpi/q35/CEDT",
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 31/32] WIP: i386/cxl: Initialize a host bridge
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (29 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 30/32] tests/acpi: Add new CEDT files Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-05 16:53 ` [RFC PATCH v2 32/32] qtest/cxl: Add very basic sanity tests Ben Widawsky
  2021-01-08 18:44 ` [RFC PATCH v2 00/32] CXL 2.0 Support Jonathan Cameron
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

This patch allows initializing the primary host bridge as a CXL capable
hostbridge.

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>

--
This patch is WIP.
---
 hw/arm/virt.c        |  1 +
 hw/core/machine.c    | 26 ++++++++++++++++++++++++++
 hw/i386/acpi-build.c |  8 +++++++-
 hw/i386/microvm.c    |  1 +
 hw/i386/pc.c         |  1 +
 hw/ppc/spapr.c       |  2 ++
 include/hw/boards.h  |  2 ++
 include/hw/cxl/cxl.h |  4 ++++
 8 files changed, 44 insertions(+), 1 deletion(-)

diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index 96985917d3..c9d6636efa 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -2481,6 +2481,7 @@ static void virt_machine_class_init(ObjectClass *oc, void *data)
     hc->unplug_request = virt_machine_device_unplug_request_cb;
     hc->unplug = virt_machine_device_unplug_cb;
     mc->nvdimm_supported = true;
+    mc->cxl_supported = false;
     mc->auto_enable_numa_with_memhp = true;
     mc->auto_enable_numa_with_memdev = true;
     mc->default_ram_id = "mach-virt.ram";
diff --git a/hw/core/machine.c b/hw/core/machine.c
index de3b8f1b31..c739803854 100644
--- a/hw/core/machine.c
+++ b/hw/core/machine.c
@@ -30,6 +30,7 @@
 #include "sysemu/qtest.h"
 #include "hw/pci/pci.h"
 #include "hw/mem/nvdimm.h"
+#include "hw/cxl/cxl.h"
 #include "migration/global_state.h"
 #include "migration/vmstate.h"
 
@@ -502,6 +503,20 @@ static void machine_set_nvdimm_persistence(Object *obj, const char *value,
     nvdimms_state->persistence_string = g_strdup(value);
 }
 
+static bool machine_get_cxl(Object *obj, Error **errp)
+{
+    MachineState *ms = MACHINE(obj);
+
+    return ms->cxl_devices_state->is_enabled;
+}
+
+static void machine_set_cxl(Object *obj, bool value, Error **errp)
+{
+    MachineState *ms = MACHINE(obj);
+
+    ms->cxl_devices_state->is_enabled = value;
+}
+
 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
 {
     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
@@ -903,6 +918,16 @@ static void machine_initfn(Object *obj)
                                         "Valid values are cpu, mem-ctrl");
     }
 
+    if (mc->cxl_supported) {
+        Object *obj = OBJECT(ms);
+
+        ms->cxl_devices_state = g_new0(CXLState, 1);
+        object_property_add_bool(obj, "cxl", machine_get_cxl, machine_set_cxl);
+        object_property_set_description(obj, "cxl",
+                                        "Set on/off to enable/disable "
+                                        "CXL instantiation");
+    }
+
     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
         ms->numa_state = g_new0(NumaState, 1);
         object_property_add_bool(obj, "hmat",
@@ -939,6 +964,7 @@ static void machine_finalize(Object *obj)
     g_free(ms->device_memory);
     g_free(ms->nvdimms_state);
     g_free(ms->numa_state);
+    g_free(ms->cxl_devices_state);
 }
 
 bool machine_usb(MachineState *machine)
diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 470dd7ddf4..90d52e7774 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -53,6 +53,7 @@
 #include "sysemu/numa.h"
 #include "sysemu/reset.h"
 #include "hw/hyperv/vmbus-bridge.h"
+#include "hw/cxl/cxl.h"
 
 /* Supported chipsets: */
 #include "hw/southbridge/piix.h"
@@ -1277,8 +1278,13 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
         build_piix4_pci0_int(dsdt);
     } else {
         sb_scope = aml_scope("_SB");
+        /*
+         * XXX: CXL spec calls this "CXL0", but that would require lots of
+         * changes throughout and so even for CXL enabled, we call it "PCI0"
+         */
         dev = aml_device("PCI0");
-        init_pci_acpi(dev, 0, PCIE);
+        init_pci_acpi(dev, 0,
+                machine->cxl_devices_state->is_enabled ? CXL : PCIE);
         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
         aml_append(sb_scope, dev);
 
diff --git a/hw/i386/microvm.c b/hw/i386/microvm.c
index edf2b0f061..970b299a69 100644
--- a/hw/i386/microvm.c
+++ b/hw/i386/microvm.c
@@ -688,6 +688,7 @@ static void microvm_class_init(ObjectClass *oc, void *data)
     mc->auto_enable_numa_with_memdev = false;
     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
     mc->nvdimm_supported = false;
+    mc->cxl_supported = false;
     mc->default_ram_id = "microvm.ram";
 
     /* Avoid relying too much on kernel components */
diff --git a/hw/i386/pc.c b/hw/i386/pc.c
index 5d41809b37..7350eeea9c 100644
--- a/hw/i386/pc.c
+++ b/hw/i386/pc.c
@@ -1725,6 +1725,7 @@ static void pc_machine_class_init(ObjectClass *oc, void *data)
     hc->unplug = pc_machine_device_unplug_cb;
     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
     mc->nvdimm_supported = true;
+    mc->cxl_supported = true;
     mc->default_ram_id = "pc.ram";
 
     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index 489cefcb81..b5ea497403 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -4442,6 +4442,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
     mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power9_v2.0");
     mc->has_hotpluggable_cpus = true;
     mc->nvdimm_supported = true;
+    mc->cxl_supported = false;
     smc->resize_hpt_default = SPAPR_RESIZE_HPT_ENABLED;
     fwc->get_dev_path = spapr_get_fw_dev_path;
     nc->nmi_monitor_handler = spapr_nmi;
@@ -4602,6 +4603,7 @@ static void spapr_machine_4_2_class_options(MachineClass *mc)
     smc->default_caps.caps[SPAPR_CAP_FWNMI] = SPAPR_CAP_OFF;
     smc->rma_limit = 16 * GiB;
     mc->nvdimm_supported = false;
+    mc->cxl_supported = false;
 }
 
 DEFINE_SPAPR_MACHINE(4_2, "4.2", false);
diff --git a/include/hw/boards.h b/include/hw/boards.h
index 17b1f3f0b9..808f73e134 100644
--- a/include/hw/boards.h
+++ b/include/hw/boards.h
@@ -206,6 +206,7 @@ struct MachineClass {
     bool ignore_boot_device_suffixes;
     bool smbus_no_migration_support;
     bool nvdimm_supported;
+    bool cxl_supported;
     bool numa_mem_supported;
     bool auto_enable_numa;
     const char *default_ram_id;
@@ -292,6 +293,7 @@ struct MachineState {
     CPUArchIdList *possible_cpus;
     CpuTopology smp;
     struct NVDIMMState *nvdimms_state;
+    struct CXLState *cxl_devices_state;
     struct NumaState *numa_state;
 };
 
diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
index 809ed7de60..6961e47076 100644
--- a/include/hw/cxl/cxl.h
+++ b/include/hw/cxl/cxl.h
@@ -22,5 +22,9 @@
 #define CXL_HOST_BASE 0xD0000000
 #define CXL_WINDOW_MAX 10
 
+typedef struct CXLState {
+    bool is_enabled;
+} CXLState;
+
 #endif
 
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* [RFC PATCH v2 32/32] qtest/cxl: Add very basic sanity tests
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (30 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 31/32] WIP: i386/cxl: Initialize a host bridge Ben Widawsky
@ 2021-01-05 16:53 ` Ben Widawsky
  2021-01-08 18:44 ` [RFC PATCH v2 00/32] CXL 2.0 Support Jonathan Cameron
  32 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 16:53 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Ben Widawsky, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
---
 tests/qtest/cxl-test.c  | 93 +++++++++++++++++++++++++++++++++++++++++
 tests/qtest/meson.build |  4 ++
 2 files changed, 97 insertions(+)
 create mode 100644 tests/qtest/cxl-test.c

diff --git a/tests/qtest/cxl-test.c b/tests/qtest/cxl-test.c
new file mode 100644
index 0000000000..00eca14faa
--- /dev/null
+++ b/tests/qtest/cxl-test.c
@@ -0,0 +1,93 @@
+/*
+ * QTest testcase for CXL
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "libqtest-single.h"
+
+#define QEMU_PXB_CMD "-machine q35 -object memory-backend-file,id=cxl-mem1," \
+                     "share,mem-path=%s,size=512M "                          \
+                     "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52,uid=0,"  \
+                     "len-window-base=1,window-base[0]=0x4c0000000,memdev[0]=cxl-mem1"
+#define QEMU_RP "-device cxl-rp,id=rp0,bus=cxl.0,addr=0.0,chassis=0,slot=0"
+
+#define QEMU_T3D "-device cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"
+
+static void cxl_basic_hb(void)
+{
+    qtest_start("-machine q35,cxl");
+    qtest_end();
+}
+
+static void cxl_basic_pxb(void)
+{
+    qtest_start("-machine q35 -device pxb-cxl,bus=pcie.0,uid=0");
+    qtest_end();
+}
+
+static void cxl_pxb_with_window(void)
+{
+    GString *cmdline;
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    cmdline = g_string_new(NULL);
+    g_string_printf(cmdline, QEMU_PXB_CMD, tmpfs);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+
+    g_string_free(cmdline, TRUE);
+}
+
+static void cxl_root_port(void)
+{
+    GString *cmdline;
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    cmdline = g_string_new(NULL);
+    g_string_printf(cmdline, QEMU_PXB_CMD " %s", tmpfs, QEMU_RP);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+
+    g_string_free(cmdline, TRUE);
+}
+
+static void cxl_t3d(void)
+{
+    GString *cmdline;
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    cmdline = g_string_new(NULL);
+    g_string_printf(cmdline, QEMU_PXB_CMD " %s %s", tmpfs, QEMU_RP, QEMU_T3D);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+
+    g_string_free(cmdline, TRUE);
+}
+
+int main(int argc, char **argv)
+{
+    g_test_init(&argc, &argv, NULL);
+
+    qtest_add_func("/pci/cxl/basic_hostbridge", cxl_basic_hb);
+    qtest_add_func("/pci/cxl/basic_pxb", cxl_basic_pxb);
+    qtest_add_func("/pci/cxl/pxb_with_window", cxl_pxb_with_window);
+    qtest_add_func("/pci/cxl/root_port", cxl_root_port);
+    qtest_add_func("/pci/cxl/type3_device", cxl_t3d);
+
+    return g_test_run();
+}
diff --git a/tests/qtest/meson.build b/tests/qtest/meson.build
index 6a67c538be..6ab14fc79a 100644
--- a/tests/qtest/meson.build
+++ b/tests/qtest/meson.build
@@ -22,6 +22,9 @@ qtests_pci = \
   (config_all_devices.has_key('CONFIG_VGA') ? ['display-vga-test'] : []) +                  \
   (config_all_devices.has_key('CONFIG_IVSHMEM_DEVICE') ? ['ivshmem-test'] : [])
 
+qtests_cxl = \
+  (config_all_devices.has_key('CONFIG_CXL') ? ['cxl-test'] : [])
+
 qtests_i386 = \
   (slirp.found() ? ['pxe-test', 'test-netfilter'] : []) +             \
   (config_host.has_key('CONFIG_POSIX') ? ['test-filter-mirror'] : []) +                     \
@@ -47,6 +50,7 @@ qtests_i386 = \
   (config_all_devices.has_key('CONFIG_TPM_TIS_ISA') ? ['tpm-tis-swtpm-test'] : []) +        \
   (config_all_devices.has_key('CONFIG_RTL8139_PCI') ? ['rtl8139-test'] : []) +              \
   qtests_pci +                                                                              \
+  qtests_cxl +                                                                              \
   ['fdc-test',
    'ide-test',
    'hd-geo-test',
-- 
2.30.0



^ permalink raw reply related	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 11/32] hw/cxl/device: Timestamp implementation (8.2.9.3)
  2021-01-05 16:53 ` [RFC PATCH v2 11/32] hw/cxl/device: Timestamp implementation (8.2.9.3) Ben Widawsky
@ 2021-01-05 17:12   ` Ben Widawsky
  0 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-05 17:12 UTC (permalink / raw)
  To: qemu-devel
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	Jonathan Cameron, Prashant V Agarwal, Dan Williams

On 21-01-05 08:53:02, Ben Widawsky wrote:
> Per spec, timestamp appears to be a free-running counter from a value
> set by the host via the Set Timestamp command (0301h). There are
> references to the epoch, which seem like a red herring. Therefore, the
> implementation implements the timestamp as freerunning counter from the
> last value that was issued by the Set Timestamp command.
> 
> Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> ---
>  hw/cxl/cxl-mailbox-utils.c  | 52 +++++++++++++++++++++++++++++++++++++
>  include/hw/cxl/cxl_device.h |  6 +++++
>  2 files changed, 58 insertions(+)
> 
> diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> index 890e486ea1..8b956d9e94 100644
> --- a/hw/cxl/cxl-mailbox-utils.c
> +++ b/hw/cxl/cxl-mailbox-utils.c
> @@ -43,6 +43,9 @@ enum {
>          #define CLEAR_RECORDS   0x1
>          #define GET_INTERRUPT_POLICY   0x2
>          #define SET_INTERRUPT_POLICY   0x3
> +    TIMESTAMP   = 0x03,
> +        #define GET           0x0
> +        #define SET           0x1
>  };
>  
>  /* 8.2.8.4.5.1 Command Return Codes */
> @@ -117,8 +120,11 @@ define_mailbox_handler_zeroed(EVENTS_GET_RECORDS, 0x20);
>  define_mailbox_handler_nop(EVENTS_CLEAR_RECORDS);
>  define_mailbox_handler_zeroed(EVENTS_GET_INTERRUPT_POLICY, 4);
>  define_mailbox_handler_nop(EVENTS_SET_INTERRUPT_POLICY);
> +declare_mailbox_handler(TIMESTAMP_GET);
> +declare_mailbox_handler(TIMESTAMP_SET);
>  
>  #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
> +#define IMMEDIATE_POLICY_CHANGE (1 << 3)
>  #define IMMEDIATE_LOG_CHANGE (1 << 4)
>  
>  #define CXL_CMD(s, c, in, cel_effect) \
> @@ -129,10 +135,56 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
>      CXL_CMD(EVENTS, CLEAR_RECORDS, ~0, IMMEDIATE_LOG_CHANGE),
>      CXL_CMD(EVENTS, GET_INTERRUPT_POLICY, 0, 0),
>      CXL_CMD(EVENTS, SET_INTERRUPT_POLICY, 4, IMMEDIATE_CONFIG_CHANGE),
> +    CXL_CMD(TIMESTAMP, GET, 0, 0),
> +    CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
>  };
>  
>  #undef CXL_CMD
>  
> +/*
> + * 8.2.9.3.1
> + */
> +define_mailbox_handler(TIMESTAMP_GET)
> +{
> +    struct timespec ts;
> +
> +    clock_gettime(CLOCK_REALTIME, &ts);
> +
> +    cxl_dstate->timestamp.set = true;
> +    cxl_dstate->timestamp.last_set =
> +        ts.tv_sec * NANOSECONDS_PER_SECOND + ts.tv_nsec;
> +    cxl_dstate->timestamp.host_set = *(uint64_t *)cmd->payload;
> +
> +    *len = 8;
> +    return CXL_MBOX_SUCCESS;
> +}
> +
> +/*
> + * 8.2.9.3.2
> + */
> +define_mailbox_handler(TIMESTAMP_SET)
> +{
> +    struct timespec ts;
> +    uint64_t delta;
> +
> +    if (!cxl_dstate->timestamp.set) {
> +        *(uint64_t *)cmd->payload = 0;
> +        goto done;
> +    }
> +
> +    /* First find the delta from the last time the host set the time. */
> +    clock_gettime(CLOCK_REALTIME, &ts);
> +    delta = (ts.tv_sec * NANOSECONDS_PER_SECOND + ts.tv_nsec) -
> +            cxl_dstate->timestamp.host_set;
> +
> +    /* Then adjust the actual time */
> +    *(uint64_t *)cmd->payload = cxl_dstate->timestamp.host_set + delta;
> +
> +done:
> +    *len = 8;
> +    return CXL_MBOX_SUCCESS;
> +}
> +

Set/get are reversed here. I missed the fixup before sending.

>  QemuUUID cel_uuid;
>  
>  void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> index dd3f4572aa..25974b2416 100644
> --- a/include/hw/cxl/cxl_device.h
> +++ b/include/hw/cxl/cxl_device.h
> @@ -100,6 +100,12 @@ typedef struct cxl_device_state {
>          size_t cel_size;
>      };
>  
> +    struct {
> +        bool set;
> +        uint64_t last_set;
> +        uint64_t host_set;
> +    } timestamp;
> +
>      MemoryRegion *pmem;
>      MemoryRegion *vmem;
>  } CXLDeviceState;
> -- 
> 2.30.0
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-05 16:52 ` [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4) Ben Widawsky
@ 2021-01-06 13:21   ` Jonathan Cameron
  2021-01-06 16:31     ` Ben Widawsky
  2021-01-06 17:40     ` [Linuxarm] " Jonathan Cameron
  0 siblings, 2 replies; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-06 13:21 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, linuxarm, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On Tue, 5 Jan 2021 08:52:58 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> This is the beginning of implementing mailbox support for CXL 2.0
> devices.
> 
> v2: Use register alignment helper (Ben)
>     Minor cleanups (Jonathan)
>     Rename error codes to match spec (Jonathan)
>     Update cap count from 1 to 2 (Jonathan)
>     Add infra to support CEL (Ben)
>     Add more of the actual mailbox handling from later patch (Ben)
> 
> Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>

Hi Ben,

I hacked support in for ARM64 to give this a spin and ran into an
interesting problem around read sizes.

The mailbox registers space allows 4 or 8 byte reads, but in the kernel
driver (I think I have the right version from your github) you do
the payload drain with
memcpy_from_io()

If the size of the payload is not a multiple of 8 bytes, on ARM64 that
results in byte reads and an exception.  This happens with some of the
existing calls which happen to have non multiple of 8 payload sizes.

I hacked below to allow 1 byte reads from that region but that's probably
not the right fix.  I found a statement in the CXL spec saying maximum read
size from this register block was 8 bytes but couldn't immediately see a minimum.
(I haven't looked that hard yet though!)

Various approaches in kernel could also be used:
1) Change the payload drain to have specific handling for the end few bytes.
2) Pad the various structures to ensure payloads are always 8 byte multiples
in length (nasty).

> ---
>  hw/cxl/cxl-device-utils.c   | 122 ++++++++++++++++++++++++-
>  hw/cxl/cxl-mailbox-utils.c  | 173 ++++++++++++++++++++++++++++++++++++
>  hw/cxl/meson.build          |   1 +
>  include/hw/cxl/cxl.h        |   3 +
>  include/hw/cxl/cxl_device.h |  27 +++++-
>  5 files changed, 322 insertions(+), 4 deletions(-)
>  create mode 100644 hw/cxl/cxl-mailbox-utils.c
> 
> diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> index b86e5466bd..642e3c2617 100644
> --- a/hw/cxl/cxl-device-utils.c
> +++ b/hw/cxl/cxl-device-utils.c
> @@ -44,6 +44,108 @@ static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
>      return ldn_le_p(&retval, size);
>  }
>  
> +static uint64_t mailbox_reg_read(void *opaque, hwaddr offset, unsigned size)
> +{
> +    CXLDeviceState *cxl_dstate = opaque;
> +
> +    if (cxl_device_check_register_alignment(offset, size)) {
> +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> +        return 0;
> +    }
> +
> +    return ldn_le_p(cxl_dstate->mbox_reg_state + offset, size);
> +}
> +
> +static void mailbox_mem_writel(uint32_t *reg_state, hwaddr offset,
> +                               uint64_t value)
> +{
> +    switch (offset) {
> +    case A_CXL_DEV_MAILBOX_CTRL:
> +        /* fallthrough */
> +    case A_CXL_DEV_MAILBOX_CAP:
> +        /* RO register */
> +        break;
> +    default:
> +        qemu_log_mask(LOG_UNIMP,
> +                      "%s Unexpected 32-bit access to 0x%" PRIx64 " (WI)\n",
> +                      __func__, offset);
> +        break;
> +    }
> +
> +    stl_le_p((uint8_t *)reg_state + offset, value);
> +}
> +
> +static void mailbox_mem_writeq(uint8_t *reg_state, hwaddr offset,
> +                               uint64_t value)
> +{
> +    switch (offset) {
> +    case A_CXL_DEV_MAILBOX_CMD:
> +        break;
> +    case A_CXL_DEV_BG_CMD_STS:
> +        /* BG not supported */
> +        /* fallthrough */
> +    case A_CXL_DEV_MAILBOX_STS:
> +        /* Read only register, will get updated by the state machine */
> +        return;
> +    default:
> +        qemu_log_mask(LOG_UNIMP,
> +                      "%s Unexpected 64-bit access to 0x%" PRIx64 " (WI)\n",
> +                      __func__, offset);
> +        return;
> +    }
> +
> +    stq_le_p(reg_state + offset, value);
> +}
> +
> +static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
> +                              unsigned size)
> +{
> +    CXLDeviceState *cxl_dstate = opaque;
> +
> +    /*
> +     * Lock is needed to prevent concurrent writes as well as to prevent writes
> +     * coming in while the firmware is processing. Without background commands
> +     * or the second mailbox implemented, this serves no purpose since the
> +     * memory access is synchronized at a higher level (per memory region).
> +     */
> +    RCU_READ_LOCK_GUARD();
> +
> +    switch (size) {
> +    case 4:
> +        if (unlikely(offset & (sizeof(uint32_t) - 1))) {
> +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> +            return;
> +        }
> +        mailbox_mem_writel(cxl_dstate->mbox_reg_state32, offset, value);
> +        break;
> +    case 8:
> +        if (unlikely(offset & (sizeof(uint64_t) - 1))) {
> +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> +            return;
> +        }
> +        mailbox_mem_writeq(cxl_dstate->mbox_reg_state, offset, value);
> +        break;
> +    }
> +
> +    if (ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> +                         DOORBELL))
> +        cxl_process_mailbox(cxl_dstate);
> +}
> +
> +static const MemoryRegionOps mailbox_ops = {
> +    .read = mailbox_reg_read,
> +    .write = mailbox_reg_write,
> +    .endianness = DEVICE_LITTLE_ENDIAN,
> +    .valid = {
> +        .min_access_size = 4,

Hack was to set this to 1 (and the impl below - no idea if both are needed).

> +        .max_access_size = 8,
> +    },
> +    .impl = {
> +        .min_access_size = 4,
> +        .max_access_size = 8,
> +    },
> +};
> +
>  static const MemoryRegionOps dev_ops = {
>      .read = dev_reg_read,
>      .write = NULL,
> @@ -83,20 +185,33 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
>                            "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
>      memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
>                            "device-status", CXL_DEVICE_REGISTERS_LENGTH);
> +    memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
> +                          "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
>  
>      memory_region_add_subregion(&cxl_dstate->device_registers, 0,
>                                  &cxl_dstate->caps);
>      memory_region_add_subregion(&cxl_dstate->device_registers,
>                                  CXL_DEVICE_REGISTERS_OFFSET,
>                                  &cxl_dstate->device);
> +    memory_region_add_subregion(&cxl_dstate->device_registers,
> +                                CXL_MAILBOX_REGISTERS_OFFSET,
> +                                &cxl_dstate->mailbox);
>  }
>  
>  static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
>  
> +static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
> +{
> +    /* 2048 payload size, with no interrupt or background support */
> +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CAP,
> +                     PAYLOAD_SIZE, CXL_MAILBOX_PAYLOAD_SHIFT);
> +    cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
> +}
> +
>  void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
>  {
>      uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> -    const int cap_count = 1;
> +    const int cap_count = 2;
>  
>      /* CXL Device Capabilities Array Register */
>      ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> @@ -105,4 +220,9 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
>  
>      cxl_device_cap_init(cxl_dstate, DEVICE, 1);
>      device_reg_init_common(cxl_dstate);
> +
> +    cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
> +    mailbox_reg_init_common(cxl_dstate);
> +
> +    assert(cxl_initialize_mailbox(cxl_dstate) == 0);
>  }
> diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> new file mode 100644
> index 0000000000..3a39c936de
> --- /dev/null
> +++ b/hw/cxl/cxl-mailbox-utils.c
> @@ -0,0 +1,173 @@
> +/*
> + * CXL Utility library for mailbox interface
> + *
> + * Copyright(C) 2020 Intel Corporation.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2. See the
> + * COPYING file in the top-level directory.
> + */
> +
> +#include "qemu/osdep.h"
> +#include "hw/cxl/cxl.h"
> +#include "hw/pci/pci.h"
> +#include "qemu/log.h"
> +#include "qemu/uuid.h"
> +
> +/*
> + * How to add a new command, example. The command set FOO, with cmd BAR.
> + *  1. Add the command set and cmd to the enum.
> + *     FOO    = 0x7f,
> + *          #define BAR 0
> + *  2. Forward declare the handler.
> + *     declare_mailbox_handler(FOO_BAR);
> + *  3. Add the command to the cxl_cmd_set[][]
> + *     CXL_CMD(FOO, BAR, 0, 0),
> + *  4. Implement your handler
> + *     define_mailbox_handler(FOO_BAR) { ... return CXL_MBOX_SUCCESS; }
> + *
> + *
> + *  Writing the handler:
> + *    The handler will provide the &struct cxl_cmd, the &CXLDeviceState, and the
> + *    in/out length of the payload. The handler is responsible for consuming the
> + *    payload from cmd->payload and operating upon it as necessary. It must then
> + *    fill the output data into cmd->payload (overwriting what was there),
> + *    setting the length, and returning a valid return code.
> + *
> + *  XXX: The handler need not worry about endianess. The payload is read out of
> + *  a register interface that already deals with it.
> + */
> +
> +/* 8.2.8.4.5.1 Command Return Codes */
> +typedef enum {
> +    CXL_MBOX_SUCCESS = 0x0,
> +    CXL_MBOX_BG_STARTED = 0x1,
> +    CXL_MBOX_INVALID_INPUT = 0x2,
> +    CXL_MBOX_UNSUPPORTED = 0x3,
> +    CXL_MBOX_INTERNAL_ERROR = 0x4,
> +    CXL_MBOX_RETRY_REQUIRED = 0x5,
> +    CXL_MBOX_BUSY = 0x6,
> +    CXL_MBOX_MEDIA_DISABLED = 0x7,
> +    CXL_MBOX_FW_XFER_IN_PROGRESS = 0x8,
> +    CXL_MBOX_FW_XFER_OUT_OF_ORDER = 0x9,
> +    CXL_MBOX_FW_AUTH_FAILED = 0xa,
> +    CXL_MBOX_FW_INVALID_SLOT = 0xb,
> +    CXL_MBOX_FW_ROLLEDBACK = 0xc,
> +    CXL_MBOX_FW_REST_REQD = 0xd,
> +    CXL_MBOX_INVALID_HANDLE = 0xe,
> +    CXL_MBOX_INVALID_PA = 0xf,
> +    CXL_MBOX_INJECT_POISON_LIMIT = 0x10,
> +    CXL_MBOX_PERMANENT_MEDIA_FAILURE = 0x11,
> +    CXL_MBOX_ABORTED = 0x12,
> +    CXL_MBOX_INVALID_SECURITY_STATE = 0x13,
> +    CXL_MBOX_INCORRECT_PASSPHRASE = 0x14,
> +    CXL_MBOX_UNSUPPORTED_MAILBOX = 0x15,
> +    CXL_MBOX_INVALID_PAYLOAD_LENGTH = 0x16,
> +    CXL_MBOX_MAX = 0x17
> +} ret_code;
> +
> +struct cxl_cmd;
> +typedef ret_code (*opcode_handler)(struct cxl_cmd *cmd,
> +                                   CXLDeviceState *cxl_dstate, uint16_t *len);
> +struct cxl_cmd {
> +    const char *name;
> +    opcode_handler handler;
> +    ssize_t in;
> +    uint16_t effect; /* Reported in CEL */
> +    uint8_t *payload;
> +};
> +
> +#define define_mailbox_handler(name)                \
> +    static ret_code cmd_##name(struct cxl_cmd *cmd, \
> +                               CXLDeviceState *cxl_dstate, uint16_t *len)
> +#define declare_mailbox_handler(name) define_mailbox_handler(name)
> +
> +#define CXL_CMD(s, c, in, cel_effect) \
> +    [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
> +
> +static struct cxl_cmd cxl_cmd_set[256][256] = {};
> +
> +#undef CXL_CMD
> +
> +QemuUUID cel_uuid;
> +
> +void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> +{
> +    uint16_t ret = CXL_MBOX_SUCCESS;
> +    struct cxl_cmd *cxl_cmd;
> +    uint64_t status_reg;
> +    opcode_handler h;
> +
> +    /*
> +     * current state of mailbox interface
> +     *  mbox_cap_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CAP];
> +     *  mbox_ctrl_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CTRL];
> +     *  status_reg = *(uint64_t *)&cxl_dstate->reg_state[A_CXL_DEV_MAILBOX_STS];
> +     */
> +    uint64_t command_reg =
> +        *(uint64_t *)&cxl_dstate->mbox_reg_state[A_CXL_DEV_MAILBOX_CMD];
> +
> +    /* Check if we have to do anything */
> +    if (!ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> +                          DOORBELL)) {
> +        qemu_log_mask(LOG_UNIMP, "Corrupt internal state for firmware\n");
> +        return;
> +    }
> +
> +    uint8_t set = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET);
> +    uint8_t cmd = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND);
> +    uint16_t len = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH);
> +    cxl_cmd = &cxl_cmd_set[set][cmd];
> +    h = cxl_cmd->handler;
> +    if (!h) {
> +        goto handled;
> +    }
> +
> +    if (len != cxl_cmd->in) {
> +        ret = CXL_MBOX_INVALID_PAYLOAD_LENGTH;
> +    }
> +
> +    cxl_cmd->payload = cxl_dstate->mbox_reg_state + A_CXL_DEV_CMD_PAYLOAD;
> +    ret = (*h)(cxl_cmd, cxl_dstate, &len);
> +    assert(len <= cxl_dstate->payload_size);
> +
> +handled:
> +    /*
> +     * Set the return code
> +     * XXX: it's a 64b register, but we're not setting the vendor, so we can get
> +     * away with this
> +     */
> +    status_reg = FIELD_DP64(0, CXL_DEV_MAILBOX_STS, ERRNO, ret);
> +
> +    /*
> +     * Set the return length
> +     */
> +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET, 0);
> +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND, 0);
> +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH, len);
> +
> +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_CMD, command_reg);
> +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_STS, status_reg);
> +
> +    /* Tell the host we're done */
> +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> +                     DOORBELL, 0);
> +}
> +
> +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate)
> +{
> +    const char *cel_uuidstr = "0da9c0b5-bf41-4b78-8f79-96b1623b3f17";
> +
> +    for (int i = 0; i < 256; i++) {
> +        for (int j = 0; j < 256; j++) {
> +            if (cxl_cmd_set[i][j].handler) {
> +                struct cxl_cmd *c = &cxl_cmd_set[i][j];
> +
> +                cxl_dstate->cel_log[cxl_dstate->cel_size].opcode = (i << 8) | j;
> +                cxl_dstate->cel_log[cxl_dstate->cel_size].effect = c->effect;
> +                cxl_dstate->cel_size++;
> +            }
> +        }
> +    }
> +
> +    return qemu_uuid_parse(cel_uuidstr, &cel_uuid);
> +}
> diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> index 47154d6850..0eca715d10 100644
> --- a/hw/cxl/meson.build
> +++ b/hw/cxl/meson.build
> @@ -1,4 +1,5 @@
>  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
>    'cxl-component-utils.c',
>    'cxl-device-utils.c',
> +  'cxl-mailbox-utils.c',
>  ))
> diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> index 23f52c4cf9..362cda40de 100644
> --- a/include/hw/cxl/cxl.h
> +++ b/include/hw/cxl/cxl.h
> @@ -14,5 +14,8 @@
>  #include "cxl_component.h"
>  #include "cxl_device.h"
>  
> +#define COMPONENT_REG_BAR_IDX 0
> +#define DEVICE_REG_BAR_IDX 2
> +
>  #endif
>  
> diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> index 3380fc9e7a..bdf990cec2 100644
> --- a/include/hw/cxl/cxl_device.h
> +++ b/include/hw/cxl/cxl_device.h
> @@ -70,9 +70,10 @@
>      (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
>  
>  typedef struct cxl_device_state {
> -    /* Boss container and caps registers */
> +    /* Main register container */
>      MemoryRegion device_registers;
>  
> +    MemoryRegion device;
>      struct {
>          MemoryRegion caps;
>          union {
> @@ -80,8 +81,19 @@ typedef struct cxl_device_state {
>              uint32_t caps_reg_state32[CXL_DEVICE_CAP_REG_SIZE >> 2];
>          };
>      };
> -    MemoryRegion device;
> -    MemoryRegion mailbox;
> +    struct {
> +        MemoryRegion mailbox;
> +        uint16_t payload_size;
> +        union {
> +            uint8_t mbox_reg_state[CXL_MAILBOX_REGISTERS_LENGTH];
> +            uint32_t mbox_reg_state32[CXL_MAILBOX_REGISTERS_LENGTH >> 2];
> +        };
> +        struct {
> +            uint16_t opcode;
> +            uint16_t effect;
> +        } cel_log[1 << 16];
> +        size_t cel_size;
> +    };
>  
>      MemoryRegion *pmem;
>      MemoryRegion *vmem;
> @@ -122,6 +134,9 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
>  CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
>                                                 CXL_DEVICE_CAP_REG_SIZE)
>  
> +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
> +void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
> +
>  #define cxl_device_cap_init(dstate, reg, cap_id)                                   \
>      do {                                                                           \
>          uint32_t *cap_hdrs = dstate->caps_reg_state32;                             \
> @@ -149,6 +164,12 @@ REG32(CXL_DEV_MAILBOX_CTRL, 4)
>      FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
>      FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
>  
> +/* XXX: actually a 64b register */
> +REG32(CXL_DEV_MAILBOX_CMD, 8)
> +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
> +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
> +    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
> +
>  /* XXX: actually a 64b register */
>  REG32(CXL_DEV_MAILBOX_STS, 0x10)
>      FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2)
  2021-01-05 16:52 ` [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2) Ben Widawsky
@ 2021-01-06 13:28   ` Jonathan Cameron
  2021-01-06 16:49     ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-06 13:28 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On Tue, 5 Jan 2021 08:52:56 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> This implements all device MMIO up to the first capability .That
> includes the CXL Device Capabilities Array Register, as well as all of
> the CXL Device Capability Header Registers. The latter are filled in as
> they are implemented in the following patches.
> 
> v2: Break out register alignment checks (Jonathan)
> 
> Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
Hi Ben,

One buglet / inconsistency inline that I spotted whilst chasing that issue
with size of reads.

Will get to a full review after messing around ('testing') this a bit more ;)

Jonathan

> ---
>  hw/cxl/cxl-device-utils.c | 72 +++++++++++++++++++++++++++++++++++++++
>  hw/cxl/meson.build        |  1 +
>  2 files changed, 73 insertions(+)
>  create mode 100644 hw/cxl/cxl-device-utils.c
> 
> diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> new file mode 100644
> index 0000000000..d1b1371e66
> --- /dev/null
> +++ b/hw/cxl/cxl-device-utils.c
> @@ -0,0 +1,72 @@
> +/*
> + * CXL Utility library for devices
> + *
> + * Copyright(C) 2020 Intel Corporation.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2. See the
> + * COPYING file in the top-level directory.
> + */
> +
> +#include "qemu/osdep.h"
> +#include "qemu/log.h"
> +#include "hw/cxl/cxl.h"
> +
> +static int cxl_device_check_register_alignment(hwaddr offset, unsigned size)
> +{
> +    if (unlikely(offset & (size - 1))) {
> +        return 1;
> +    }
> +
> +    return 0;
> +}
> +
> +static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
> +{
> +    CXLDeviceState *cxl_dstate = opaque;
> +
> +    if (cxl_device_check_register_alignment(offset, size)) {
> +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> +        return 0;
> +    }
> +
> +    return ldn_le_p(cxl_dstate->caps_reg_state + offset, size);
> +}
> +
> +static const MemoryRegionOps caps_ops = {
> +    .read = caps_reg_read,
> +    .write = NULL,
> +    .endianness = DEVICE_LITTLE_ENDIAN,
> +    .valid = {
> +        .min_access_size = 4,
> +        .max_access_size = 8,
> +    },
> +    .impl = {
> +        .min_access_size = 4,
> +        .max_access_size = 8,
> +    },
> +};
> +
> +void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> +{
> +    /* This will be a BAR, so needs to be rounded up to pow2 for PCI spec */
> +    memory_region_init(
> +        &cxl_dstate->device_registers, obj, "device-registers",
> +        pow2ceil(CXL_MAILBOX_REGISTERS_LENGTH + CXL_MAILBOX_REGISTERS_OFFSET));

I can see why you jumped directly to sizing this for the whole region, but the snag
is that means I think you missed the fact that patch 8 adds a region after the end
of the mailbox.   Doesn't result in an actual bug because the ceil above takes
you way past the space needed though (the memory device region is only 8 bytes long).


> +
> +    memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
> +                          "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> +
> +    memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> +                                &cxl_dstate->caps);
> +}
> +
> +void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> +{
> +    uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> +    const int cap_count = 0;
> +
> +    /* CXL Device Capabilities Array Register */
> +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
> +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY2, CAP_COUNT, cap_count);
> +}
> diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> index 00c3876a0f..47154d6850 100644
> --- a/hw/cxl/meson.build
> +++ b/hw/cxl/meson.build
> @@ -1,3 +1,4 @@
>  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
>    'cxl-component-utils.c',
> +  'cxl-device-utils.c',
>  ))



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-06 13:21   ` Jonathan Cameron
@ 2021-01-06 16:31     ` Ben Widawsky
  2021-01-06 17:40     ` [Linuxarm] " Jonathan Cameron
  1 sibling, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-06 16:31 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, linuxarm, Michael S. Tsirkin, Vishal Verma,
	Chris Browy, Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On 21-01-06 13:21:23, Jonathan Cameron wrote:
> On Tue, 5 Jan 2021 08:52:58 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > This is the beginning of implementing mailbox support for CXL 2.0
> > devices.
> > 
> > v2: Use register alignment helper (Ben)
> >     Minor cleanups (Jonathan)
> >     Rename error codes to match spec (Jonathan)
> >     Update cap count from 1 to 2 (Jonathan)
> >     Add infra to support CEL (Ben)
> >     Add more of the actual mailbox handling from later patch (Ben)
> > 
> > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> 
> Hi Ben,
> 
> I hacked support in for ARM64 to give this a spin and ran into an
> interesting problem around read sizes.

Cool!

> 
> The mailbox registers space allows 4 or 8 byte reads, but in the kernel
> driver (I think I have the right version from your github) you do
> the payload drain with
> memcpy_from_io()

https://gitlab.com/bwidawsk/linux/-/tree/cxl-2.0v3

Vishal in the Cc is finishing up some work on adding support to ndctl, after
which I'll be sending that (or something closely related) to the mailing list.

> 
> If the size of the payload is not a multiple of 8 bytes, on ARM64 that
> results in byte reads and an exception.  This happens with some of the
> existing calls which happen to have non multiple of 8 payload sizes.

Could you add the backtrace here? You caused me to go look at what x86 does. I
actually assumed on modern platforms it did movsb (byte moves), which are highly
optimized in microcode. In fact, it does some amount of word moves too, which
was new info to me.

So I'm not really sure yet why ARM64 is special here.

> 
> I hacked below to allow 1 byte reads from that region but that's probably
> not the right fix.  I found a statement in the CXL spec saying maximum read
> size from this register block was 8 bytes but couldn't immediately see a minimum.
> (I haven't looked that hard yet though!)

8.2
• A 32 bit register shall be accessed as a 1 Byte, 2 Bytes or 4 Bytes quantity.
• A 64 bit register shall be accessed as a 1 Byte, 2 Bytes, 4 Bytes or 8 Bytes quantity

Interesting though, I really thought QEMU should do the right thing here and you
wouldn't need to allow 1byte accesses. 

> 
> Various approaches in kernel could also be used:
> 1) Change the payload drain to have specific handling for the end few bytes.
> 2) Pad the various structures to ensure payloads are always 8 byte multiples
> in length (nasty).

I think the right solution is to support/allow byte and word access. I just
thought it should work in the existing code. I'd like to try to figure out
what's going on.

How much effort would it be for me to reproduce what you've done?

> 
> > ---
> >  hw/cxl/cxl-device-utils.c   | 122 ++++++++++++++++++++++++-
> >  hw/cxl/cxl-mailbox-utils.c  | 173 ++++++++++++++++++++++++++++++++++++
> >  hw/cxl/meson.build          |   1 +
> >  include/hw/cxl/cxl.h        |   3 +
> >  include/hw/cxl/cxl_device.h |  27 +++++-
> >  5 files changed, 322 insertions(+), 4 deletions(-)
> >  create mode 100644 hw/cxl/cxl-mailbox-utils.c
> > 
> > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > index b86e5466bd..642e3c2617 100644
> > --- a/hw/cxl/cxl-device-utils.c
> > +++ b/hw/cxl/cxl-device-utils.c
> > @@ -44,6 +44,108 @@ static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
> >      return ldn_le_p(&retval, size);
> >  }
> >  
> > +static uint64_t mailbox_reg_read(void *opaque, hwaddr offset, unsigned size)
> > +{
> > +    CXLDeviceState *cxl_dstate = opaque;
> > +
> > +    if (cxl_device_check_register_alignment(offset, size)) {
> > +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +        return 0;
> > +    }
> > +
> > +    return ldn_le_p(cxl_dstate->mbox_reg_state + offset, size);
> > +}
> > +
> > +static void mailbox_mem_writel(uint32_t *reg_state, hwaddr offset,
> > +                               uint64_t value)
> > +{
> > +    switch (offset) {
> > +    case A_CXL_DEV_MAILBOX_CTRL:
> > +        /* fallthrough */
> > +    case A_CXL_DEV_MAILBOX_CAP:
> > +        /* RO register */
> > +        break;
> > +    default:
> > +        qemu_log_mask(LOG_UNIMP,
> > +                      "%s Unexpected 32-bit access to 0x%" PRIx64 " (WI)\n",
> > +                      __func__, offset);
> > +        break;
> > +    }
> > +
> > +    stl_le_p((uint8_t *)reg_state + offset, value);
> > +}
> > +
> > +static void mailbox_mem_writeq(uint8_t *reg_state, hwaddr offset,
> > +                               uint64_t value)
> > +{
> > +    switch (offset) {
> > +    case A_CXL_DEV_MAILBOX_CMD:
> > +        break;
> > +    case A_CXL_DEV_BG_CMD_STS:
> > +        /* BG not supported */
> > +        /* fallthrough */
> > +    case A_CXL_DEV_MAILBOX_STS:
> > +        /* Read only register, will get updated by the state machine */
> > +        return;
> > +    default:
> > +        qemu_log_mask(LOG_UNIMP,
> > +                      "%s Unexpected 64-bit access to 0x%" PRIx64 " (WI)\n",
> > +                      __func__, offset);
> > +        return;
> > +    }
> > +
> > +    stq_le_p(reg_state + offset, value);
> > +}
> > +
> > +static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
> > +                              unsigned size)
> > +{
> > +    CXLDeviceState *cxl_dstate = opaque;
> > +
> > +    /*
> > +     * Lock is needed to prevent concurrent writes as well as to prevent writes
> > +     * coming in while the firmware is processing. Without background commands
> > +     * or the second mailbox implemented, this serves no purpose since the
> > +     * memory access is synchronized at a higher level (per memory region).
> > +     */
> > +    RCU_READ_LOCK_GUARD();
> > +
> > +    switch (size) {
> > +    case 4:
> > +        if (unlikely(offset & (sizeof(uint32_t) - 1))) {
> > +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +            return;
> > +        }
> > +        mailbox_mem_writel(cxl_dstate->mbox_reg_state32, offset, value);
> > +        break;
> > +    case 8:
> > +        if (unlikely(offset & (sizeof(uint64_t) - 1))) {
> > +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +            return;
> > +        }
> > +        mailbox_mem_writeq(cxl_dstate->mbox_reg_state, offset, value);
> > +        break;
> > +    }
> > +
> > +    if (ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > +                         DOORBELL))
> > +        cxl_process_mailbox(cxl_dstate);
> > +}
> > +
> > +static const MemoryRegionOps mailbox_ops = {
> > +    .read = mailbox_reg_read,
> > +    .write = mailbox_reg_write,
> > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > +    .valid = {
> > +        .min_access_size = 4,
> 
> Hack was to set this to 1 (and the impl below - no idea if both are needed).
> 
> > +        .max_access_size = 8,
> > +    },
> > +    .impl = {
> > +        .min_access_size = 4,
> > +        .max_access_size = 8,
> > +    },
> > +};
> > +
> >  static const MemoryRegionOps dev_ops = {
> >      .read = dev_reg_read,
> >      .write = NULL,
> > @@ -83,20 +185,33 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> >                            "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> >      memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
> >                            "device-status", CXL_DEVICE_REGISTERS_LENGTH);
> > +    memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
> > +                          "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
> >  
> >      memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> >                                  &cxl_dstate->caps);
> >      memory_region_add_subregion(&cxl_dstate->device_registers,
> >                                  CXL_DEVICE_REGISTERS_OFFSET,
> >                                  &cxl_dstate->device);
> > +    memory_region_add_subregion(&cxl_dstate->device_registers,
> > +                                CXL_MAILBOX_REGISTERS_OFFSET,
> > +                                &cxl_dstate->mailbox);
> >  }
> >  
> >  static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
> >  
> > +static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
> > +{
> > +    /* 2048 payload size, with no interrupt or background support */
> > +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CAP,
> > +                     PAYLOAD_SIZE, CXL_MAILBOX_PAYLOAD_SHIFT);
> > +    cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
> > +}
> > +
> >  void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> >  {
> >      uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> > -    const int cap_count = 1;
> > +    const int cap_count = 2;
> >  
> >      /* CXL Device Capabilities Array Register */
> >      ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> > @@ -105,4 +220,9 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> >  
> >      cxl_device_cap_init(cxl_dstate, DEVICE, 1);
> >      device_reg_init_common(cxl_dstate);
> > +
> > +    cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
> > +    mailbox_reg_init_common(cxl_dstate);
> > +
> > +    assert(cxl_initialize_mailbox(cxl_dstate) == 0);
> >  }
> > diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> > new file mode 100644
> > index 0000000000..3a39c936de
> > --- /dev/null
> > +++ b/hw/cxl/cxl-mailbox-utils.c
> > @@ -0,0 +1,173 @@
> > +/*
> > + * CXL Utility library for mailbox interface
> > + *
> > + * Copyright(C) 2020 Intel Corporation.
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2. See the
> > + * COPYING file in the top-level directory.
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +#include "hw/cxl/cxl.h"
> > +#include "hw/pci/pci.h"
> > +#include "qemu/log.h"
> > +#include "qemu/uuid.h"
> > +
> > +/*
> > + * How to add a new command, example. The command set FOO, with cmd BAR.
> > + *  1. Add the command set and cmd to the enum.
> > + *     FOO    = 0x7f,
> > + *          #define BAR 0
> > + *  2. Forward declare the handler.
> > + *     declare_mailbox_handler(FOO_BAR);
> > + *  3. Add the command to the cxl_cmd_set[][]
> > + *     CXL_CMD(FOO, BAR, 0, 0),
> > + *  4. Implement your handler
> > + *     define_mailbox_handler(FOO_BAR) { ... return CXL_MBOX_SUCCESS; }
> > + *
> > + *
> > + *  Writing the handler:
> > + *    The handler will provide the &struct cxl_cmd, the &CXLDeviceState, and the
> > + *    in/out length of the payload. The handler is responsible for consuming the
> > + *    payload from cmd->payload and operating upon it as necessary. It must then
> > + *    fill the output data into cmd->payload (overwriting what was there),
> > + *    setting the length, and returning a valid return code.
> > + *
> > + *  XXX: The handler need not worry about endianess. The payload is read out of
> > + *  a register interface that already deals with it.
> > + */
> > +
> > +/* 8.2.8.4.5.1 Command Return Codes */
> > +typedef enum {
> > +    CXL_MBOX_SUCCESS = 0x0,
> > +    CXL_MBOX_BG_STARTED = 0x1,
> > +    CXL_MBOX_INVALID_INPUT = 0x2,
> > +    CXL_MBOX_UNSUPPORTED = 0x3,
> > +    CXL_MBOX_INTERNAL_ERROR = 0x4,
> > +    CXL_MBOX_RETRY_REQUIRED = 0x5,
> > +    CXL_MBOX_BUSY = 0x6,
> > +    CXL_MBOX_MEDIA_DISABLED = 0x7,
> > +    CXL_MBOX_FW_XFER_IN_PROGRESS = 0x8,
> > +    CXL_MBOX_FW_XFER_OUT_OF_ORDER = 0x9,
> > +    CXL_MBOX_FW_AUTH_FAILED = 0xa,
> > +    CXL_MBOX_FW_INVALID_SLOT = 0xb,
> > +    CXL_MBOX_FW_ROLLEDBACK = 0xc,
> > +    CXL_MBOX_FW_REST_REQD = 0xd,
> > +    CXL_MBOX_INVALID_HANDLE = 0xe,
> > +    CXL_MBOX_INVALID_PA = 0xf,
> > +    CXL_MBOX_INJECT_POISON_LIMIT = 0x10,
> > +    CXL_MBOX_PERMANENT_MEDIA_FAILURE = 0x11,
> > +    CXL_MBOX_ABORTED = 0x12,
> > +    CXL_MBOX_INVALID_SECURITY_STATE = 0x13,
> > +    CXL_MBOX_INCORRECT_PASSPHRASE = 0x14,
> > +    CXL_MBOX_UNSUPPORTED_MAILBOX = 0x15,
> > +    CXL_MBOX_INVALID_PAYLOAD_LENGTH = 0x16,
> > +    CXL_MBOX_MAX = 0x17
> > +} ret_code;
> > +
> > +struct cxl_cmd;
> > +typedef ret_code (*opcode_handler)(struct cxl_cmd *cmd,
> > +                                   CXLDeviceState *cxl_dstate, uint16_t *len);
> > +struct cxl_cmd {
> > +    const char *name;
> > +    opcode_handler handler;
> > +    ssize_t in;
> > +    uint16_t effect; /* Reported in CEL */
> > +    uint8_t *payload;
> > +};
> > +
> > +#define define_mailbox_handler(name)                \
> > +    static ret_code cmd_##name(struct cxl_cmd *cmd, \
> > +                               CXLDeviceState *cxl_dstate, uint16_t *len)
> > +#define declare_mailbox_handler(name) define_mailbox_handler(name)
> > +
> > +#define CXL_CMD(s, c, in, cel_effect) \
> > +    [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
> > +
> > +static struct cxl_cmd cxl_cmd_set[256][256] = {};
> > +
> > +#undef CXL_CMD
> > +
> > +QemuUUID cel_uuid;
> > +
> > +void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> > +{
> > +    uint16_t ret = CXL_MBOX_SUCCESS;
> > +    struct cxl_cmd *cxl_cmd;
> > +    uint64_t status_reg;
> > +    opcode_handler h;
> > +
> > +    /*
> > +     * current state of mailbox interface
> > +     *  mbox_cap_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CAP];
> > +     *  mbox_ctrl_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CTRL];
> > +     *  status_reg = *(uint64_t *)&cxl_dstate->reg_state[A_CXL_DEV_MAILBOX_STS];
> > +     */
> > +    uint64_t command_reg =
> > +        *(uint64_t *)&cxl_dstate->mbox_reg_state[A_CXL_DEV_MAILBOX_CMD];
> > +
> > +    /* Check if we have to do anything */
> > +    if (!ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > +                          DOORBELL)) {
> > +        qemu_log_mask(LOG_UNIMP, "Corrupt internal state for firmware\n");
> > +        return;
> > +    }
> > +
> > +    uint8_t set = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET);
> > +    uint8_t cmd = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND);
> > +    uint16_t len = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH);
> > +    cxl_cmd = &cxl_cmd_set[set][cmd];
> > +    h = cxl_cmd->handler;
> > +    if (!h) {
> > +        goto handled;
> > +    }
> > +
> > +    if (len != cxl_cmd->in) {
> > +        ret = CXL_MBOX_INVALID_PAYLOAD_LENGTH;
> > +    }
> > +
> > +    cxl_cmd->payload = cxl_dstate->mbox_reg_state + A_CXL_DEV_CMD_PAYLOAD;
> > +    ret = (*h)(cxl_cmd, cxl_dstate, &len);
> > +    assert(len <= cxl_dstate->payload_size);
> > +
> > +handled:
> > +    /*
> > +     * Set the return code
> > +     * XXX: it's a 64b register, but we're not setting the vendor, so we can get
> > +     * away with this
> > +     */
> > +    status_reg = FIELD_DP64(0, CXL_DEV_MAILBOX_STS, ERRNO, ret);
> > +
> > +    /*
> > +     * Set the return length
> > +     */
> > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET, 0);
> > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND, 0);
> > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH, len);
> > +
> > +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_CMD, command_reg);
> > +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_STS, status_reg);
> > +
> > +    /* Tell the host we're done */
> > +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > +                     DOORBELL, 0);
> > +}
> > +
> > +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate)
> > +{
> > +    const char *cel_uuidstr = "0da9c0b5-bf41-4b78-8f79-96b1623b3f17";
> > +
> > +    for (int i = 0; i < 256; i++) {
> > +        for (int j = 0; j < 256; j++) {
> > +            if (cxl_cmd_set[i][j].handler) {
> > +                struct cxl_cmd *c = &cxl_cmd_set[i][j];
> > +
> > +                cxl_dstate->cel_log[cxl_dstate->cel_size].opcode = (i << 8) | j;
> > +                cxl_dstate->cel_log[cxl_dstate->cel_size].effect = c->effect;
> > +                cxl_dstate->cel_size++;
> > +            }
> > +        }
> > +    }
> > +
> > +    return qemu_uuid_parse(cel_uuidstr, &cel_uuid);
> > +}
> > diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> > index 47154d6850..0eca715d10 100644
> > --- a/hw/cxl/meson.build
> > +++ b/hw/cxl/meson.build
> > @@ -1,4 +1,5 @@
> >  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
> >    'cxl-component-utils.c',
> >    'cxl-device-utils.c',
> > +  'cxl-mailbox-utils.c',
> >  ))
> > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > index 23f52c4cf9..362cda40de 100644
> > --- a/include/hw/cxl/cxl.h
> > +++ b/include/hw/cxl/cxl.h
> > @@ -14,5 +14,8 @@
> >  #include "cxl_component.h"
> >  #include "cxl_device.h"
> >  
> > +#define COMPONENT_REG_BAR_IDX 0
> > +#define DEVICE_REG_BAR_IDX 2
> > +
> >  #endif
> >  
> > diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> > index 3380fc9e7a..bdf990cec2 100644
> > --- a/include/hw/cxl/cxl_device.h
> > +++ b/include/hw/cxl/cxl_device.h
> > @@ -70,9 +70,10 @@
> >      (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
> >  
> >  typedef struct cxl_device_state {
> > -    /* Boss container and caps registers */
> > +    /* Main register container */
> >      MemoryRegion device_registers;
> >  
> > +    MemoryRegion device;
> >      struct {
> >          MemoryRegion caps;
> >          union {
> > @@ -80,8 +81,19 @@ typedef struct cxl_device_state {
> >              uint32_t caps_reg_state32[CXL_DEVICE_CAP_REG_SIZE >> 2];
> >          };
> >      };
> > -    MemoryRegion device;
> > -    MemoryRegion mailbox;
> > +    struct {
> > +        MemoryRegion mailbox;
> > +        uint16_t payload_size;
> > +        union {
> > +            uint8_t mbox_reg_state[CXL_MAILBOX_REGISTERS_LENGTH];
> > +            uint32_t mbox_reg_state32[CXL_MAILBOX_REGISTERS_LENGTH >> 2];
> > +        };
> > +        struct {
> > +            uint16_t opcode;
> > +            uint16_t effect;
> > +        } cel_log[1 << 16];
> > +        size_t cel_size;
> > +    };
> >  
> >      MemoryRegion *pmem;
> >      MemoryRegion *vmem;
> > @@ -122,6 +134,9 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
> >  CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
> >                                                 CXL_DEVICE_CAP_REG_SIZE)
> >  
> > +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
> > +void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
> > +
> >  #define cxl_device_cap_init(dstate, reg, cap_id)                                   \
> >      do {                                                                           \
> >          uint32_t *cap_hdrs = dstate->caps_reg_state32;                             \
> > @@ -149,6 +164,12 @@ REG32(CXL_DEV_MAILBOX_CTRL, 4)
> >      FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
> >      FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
> >  
> > +/* XXX: actually a 64b register */
> > +REG32(CXL_DEV_MAILBOX_CMD, 8)
> > +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
> > +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
> > +    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
> > +
> >  /* XXX: actually a 64b register */
> >  REG32(CXL_DEV_MAILBOX_STS, 0x10)
> >      FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2)
  2021-01-06 13:28   ` Jonathan Cameron
@ 2021-01-06 16:49     ` Ben Widawsky
  2021-01-06 17:06       ` Jonathan Cameron
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-06 16:49 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On 21-01-06 13:28:05, Jonathan Cameron wrote:
> On Tue, 5 Jan 2021 08:52:56 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > This implements all device MMIO up to the first capability .That
> > includes the CXL Device Capabilities Array Register, as well as all of
> > the CXL Device Capability Header Registers. The latter are filled in as
> > they are implemented in the following patches.
> > 
> > v2: Break out register alignment checks (Jonathan)
> > 
> > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> Hi Ben,
> 
> One buglet / inconsistency inline that I spotted whilst chasing that issue
> with size of reads.
> 
> Will get to a full review after messing around ('testing') this a bit more ;)
> 
> Jonathan
> 
> > ---
> >  hw/cxl/cxl-device-utils.c | 72 +++++++++++++++++++++++++++++++++++++++
> >  hw/cxl/meson.build        |  1 +
> >  2 files changed, 73 insertions(+)
> >  create mode 100644 hw/cxl/cxl-device-utils.c
> > 
> > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > new file mode 100644
> > index 0000000000..d1b1371e66
> > --- /dev/null
> > +++ b/hw/cxl/cxl-device-utils.c
> > @@ -0,0 +1,72 @@
> > +/*
> > + * CXL Utility library for devices
> > + *
> > + * Copyright(C) 2020 Intel Corporation.
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2. See the
> > + * COPYING file in the top-level directory.
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +#include "qemu/log.h"
> > +#include "hw/cxl/cxl.h"
> > +
> > +static int cxl_device_check_register_alignment(hwaddr offset, unsigned size)
> > +{
> > +    if (unlikely(offset & (size - 1))) {
> > +        return 1;
> > +    }
> > +
> > +    return 0;
> > +}
> > +
> > +static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
> > +{
> > +    CXLDeviceState *cxl_dstate = opaque;
> > +
> > +    if (cxl_device_check_register_alignment(offset, size)) {
> > +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +        return 0;
> > +    }
> > +
> > +    return ldn_le_p(cxl_dstate->caps_reg_state + offset, size);
> > +}
> > +
> > +static const MemoryRegionOps caps_ops = {
> > +    .read = caps_reg_read,
> > +    .write = NULL,
> > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > +    .valid = {
> > +        .min_access_size = 4,
> > +        .max_access_size = 8,
> > +    },
> > +    .impl = {
> > +        .min_access_size = 4,
> > +        .max_access_size = 8,
> > +    },
> > +};
> > +
> > +void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> > +{
> > +    /* This will be a BAR, so needs to be rounded up to pow2 for PCI spec */
> > +    memory_region_init(
> > +        &cxl_dstate->device_registers, obj, "device-registers",
> > +        pow2ceil(CXL_MAILBOX_REGISTERS_LENGTH + CXL_MAILBOX_REGISTERS_OFFSET));
> 
> I can see why you jumped directly to sizing this for the whole region, but the snag
> is that means I think you missed the fact that patch 8 adds a region after the end
> of the mailbox.   Doesn't result in an actual bug because the ceil above takes
> you way past the space needed though (the memory device region is only 8 bytes long).
> 
> 

Maybe I misunderstand, but this is the intended behavior.
cxl_dstate->device_registers is the MemoryRegion container for all the
subregions that are the actual device MMIO.

                     +------------------+
                  ^  |                  |
                  |  |      unused      |
                  |  --------------------
                  |  |   memdev regs    |
                  |  --------------------
                  |  |                  |
                  |  | +--------------+ |
 cxl_dstate->     |  | |              | |
 device_registers |  | |              | |
                  |  | |payload       | |
                  |  | |(2k currently)| |
                  |  | |              | |
                  |  | |              | |
                  |  | +--------------+ |
                  |  |   mailbox regs   |
                  |  --------------------
                  |  |    device regs   |
                  v  --------------------
                     |     caps regs    |
         BAR ------> +------------------+


Perhaps I should add this as a comment in the code?

> > +
> > +    memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
> > +                          "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> > +
> > +    memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> > +                                &cxl_dstate->caps);
> > +}
> > +
> > +void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> > +{
> > +    uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> > +    const int cap_count = 0;
> > +
> > +    /* CXL Device Capabilities Array Register */
> > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
> > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY2, CAP_COUNT, cap_count);
> > +}
> > diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> > index 00c3876a0f..47154d6850 100644
> > --- a/hw/cxl/meson.build
> > +++ b/hw/cxl/meson.build
> > @@ -1,3 +1,4 @@
> >  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
> >    'cxl-component-utils.c',
> > +  'cxl-device-utils.c',
> >  ))
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2)
  2021-01-06 16:49     ` Ben Widawsky
@ 2021-01-06 17:06       ` Jonathan Cameron
  2021-01-06 17:09         ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-06 17:06 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On Wed, 6 Jan 2021 08:49:48 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> On 21-01-06 13:28:05, Jonathan Cameron wrote:
> > On Tue, 5 Jan 2021 08:52:56 -0800
> > Ben Widawsky <ben.widawsky@intel.com> wrote:
> >   
> > > This implements all device MMIO up to the first capability .That
> > > includes the CXL Device Capabilities Array Register, as well as all of
> > > the CXL Device Capability Header Registers. The latter are filled in as
> > > they are implemented in the following patches.
> > > 
> > > v2: Break out register alignment checks (Jonathan)
> > > 
> > > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>  
> > Hi Ben,
> > 
> > One buglet / inconsistency inline that I spotted whilst chasing that issue
> > with size of reads.
> > 
> > Will get to a full review after messing around ('testing') this a bit more ;)
> > 
> > Jonathan
> >   
> > > ---
> > >  hw/cxl/cxl-device-utils.c | 72 +++++++++++++++++++++++++++++++++++++++
> > >  hw/cxl/meson.build        |  1 +
> > >  2 files changed, 73 insertions(+)
> > >  create mode 100644 hw/cxl/cxl-device-utils.c
> > > 
> > > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > > new file mode 100644
> > > index 0000000000..d1b1371e66
> > > --- /dev/null
> > > +++ b/hw/cxl/cxl-device-utils.c
> > > @@ -0,0 +1,72 @@
> > > +/*
> > > + * CXL Utility library for devices
> > > + *
> > > + * Copyright(C) 2020 Intel Corporation.
> > > + *
> > > + * This work is licensed under the terms of the GNU GPL, version 2. See the
> > > + * COPYING file in the top-level directory.
> > > + */
> > > +
> > > +#include "qemu/osdep.h"
> > > +#include "qemu/log.h"
> > > +#include "hw/cxl/cxl.h"
> > > +
> > > +static int cxl_device_check_register_alignment(hwaddr offset, unsigned size)
> > > +{
> > > +    if (unlikely(offset & (size - 1))) {
> > > +        return 1;
> > > +    }
> > > +
> > > +    return 0;
> > > +}
> > > +
> > > +static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
> > > +{
> > > +    CXLDeviceState *cxl_dstate = opaque;
> > > +
> > > +    if (cxl_device_check_register_alignment(offset, size)) {
> > > +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > > +        return 0;
> > > +    }
> > > +
> > > +    return ldn_le_p(cxl_dstate->caps_reg_state + offset, size);
> > > +}
> > > +
> > > +static const MemoryRegionOps caps_ops = {
> > > +    .read = caps_reg_read,
> > > +    .write = NULL,
> > > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > > +    .valid = {
> > > +        .min_access_size = 4,
> > > +        .max_access_size = 8,
> > > +    },
> > > +    .impl = {
> > > +        .min_access_size = 4,
> > > +        .max_access_size = 8,
> > > +    },
> > > +};
> > > +
> > > +void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> > > +{
> > > +    /* This will be a BAR, so needs to be rounded up to pow2 for PCI spec */
> > > +    memory_region_init(
> > > +        &cxl_dstate->device_registers, obj, "device-registers",
> > > +        pow2ceil(CXL_MAILBOX_REGISTERS_LENGTH + CXL_MAILBOX_REGISTERS_OFFSET));  
> > 
> > I can see why you jumped directly to sizing this for the whole region, but the snag
> > is that means I think you missed the fact that patch 8 adds a region after the end
> > of the mailbox.   Doesn't result in an actual bug because the ceil above takes
> > you way past the space needed though (the memory device region is only 8 bytes long).
> > 
> >   
> 
> Maybe I misunderstand, but this is the intended behavior.
> cxl_dstate->device_registers is the MemoryRegion container for all the
> subregions that are the actual device MMIO.
> 
>                      +------------------+
>                   ^  |                  |
>                   |  |      unused      |
>                   |  --------------------
>                   |  |   memdev regs    |
>                   |  --------------------
>                   |  |                  |
>                   |  | +--------------+ |
>  cxl_dstate->     |  | |              | |
>  device_registers |  | |              | |
>                   |  | |payload       | |
>                   |  | |(2k currently)| |
>                   |  | |              | |
>                   |  | |              | |
>                   |  | +--------------+ |
>                   |  |   mailbox regs   |
>                   |  --------------------
>                   |  |    device regs   |
>                   v  --------------------
>                      |     caps regs    |
>          BAR ------> +------------------+
> 
> 
> Perhaps I should add this as a comment in the code?

I agree with intent but above it is setting limit to the top of
the mailbox regs, not the memdev regs which I'd expect to see.

> 
> > > +
> > > +    memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
> > > +                          "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> > > +
> > > +    memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> > > +                                &cxl_dstate->caps);
> > > +}
> > > +
> > > +void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> > > +{
> > > +    uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> > > +    const int cap_count = 0;
> > > +
> > > +    /* CXL Device Capabilities Array Register */
> > > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> > > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
> > > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY2, CAP_COUNT, cap_count);
> > > +}
> > > diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> > > index 00c3876a0f..47154d6850 100644
> > > --- a/hw/cxl/meson.build
> > > +++ b/hw/cxl/meson.build
> > > @@ -1,3 +1,4 @@
> > >  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
> > >    'cxl-component-utils.c',
> > > +  'cxl-device-utils.c',
> > >  ))  
> >   



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2)
  2021-01-06 17:06       ` Jonathan Cameron
@ 2021-01-06 17:09         ` Ben Widawsky
  0 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-06 17:09 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On 21-01-06 17:06:41, Jonathan Cameron wrote:
> On Wed, 6 Jan 2021 08:49:48 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > On 21-01-06 13:28:05, Jonathan Cameron wrote:
> > > On Tue, 5 Jan 2021 08:52:56 -0800
> > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > >   
> > > > This implements all device MMIO up to the first capability .That
> > > > includes the CXL Device Capabilities Array Register, as well as all of
> > > > the CXL Device Capability Header Registers. The latter are filled in as
> > > > they are implemented in the following patches.
> > > > 
> > > > v2: Break out register alignment checks (Jonathan)
> > > > 
> > > > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>  
> > > Hi Ben,
> > > 
> > > One buglet / inconsistency inline that I spotted whilst chasing that issue
> > > with size of reads.
> > > 
> > > Will get to a full review after messing around ('testing') this a bit more ;)
> > > 
> > > Jonathan
> > >   
> > > > ---
> > > >  hw/cxl/cxl-device-utils.c | 72 +++++++++++++++++++++++++++++++++++++++
> > > >  hw/cxl/meson.build        |  1 +
> > > >  2 files changed, 73 insertions(+)
> > > >  create mode 100644 hw/cxl/cxl-device-utils.c
> > > > 
> > > > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > > > new file mode 100644
> > > > index 0000000000..d1b1371e66
> > > > --- /dev/null
> > > > +++ b/hw/cxl/cxl-device-utils.c
> > > > @@ -0,0 +1,72 @@
> > > > +/*
> > > > + * CXL Utility library for devices
> > > > + *
> > > > + * Copyright(C) 2020 Intel Corporation.
> > > > + *
> > > > + * This work is licensed under the terms of the GNU GPL, version 2. See the
> > > > + * COPYING file in the top-level directory.
> > > > + */
> > > > +
> > > > +#include "qemu/osdep.h"
> > > > +#include "qemu/log.h"
> > > > +#include "hw/cxl/cxl.h"
> > > > +
> > > > +static int cxl_device_check_register_alignment(hwaddr offset, unsigned size)
> > > > +{
> > > > +    if (unlikely(offset & (size - 1))) {
> > > > +        return 1;
> > > > +    }
> > > > +
> > > > +    return 0;
> > > > +}
> > > > +
> > > > +static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
> > > > +{
> > > > +    CXLDeviceState *cxl_dstate = opaque;
> > > > +
> > > > +    if (cxl_device_check_register_alignment(offset, size)) {
> > > > +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > > > +        return 0;
> > > > +    }
> > > > +
> > > > +    return ldn_le_p(cxl_dstate->caps_reg_state + offset, size);
> > > > +}
> > > > +
> > > > +static const MemoryRegionOps caps_ops = {
> > > > +    .read = caps_reg_read,
> > > > +    .write = NULL,
> > > > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > > > +    .valid = {
> > > > +        .min_access_size = 4,
> > > > +        .max_access_size = 8,
> > > > +    },
> > > > +    .impl = {
> > > > +        .min_access_size = 4,
> > > > +        .max_access_size = 8,
> > > > +    },
> > > > +};
> > > > +
> > > > +void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> > > > +{
> > > > +    /* This will be a BAR, so needs to be rounded up to pow2 for PCI spec */
> > > > +    memory_region_init(
> > > > +        &cxl_dstate->device_registers, obj, "device-registers",
> > > > +        pow2ceil(CXL_MAILBOX_REGISTERS_LENGTH + CXL_MAILBOX_REGISTERS_OFFSET));  
> > > 
> > > I can see why you jumped directly to sizing this for the whole region, but the snag
> > > is that means I think you missed the fact that patch 8 adds a region after the end
> > > of the mailbox.   Doesn't result in an actual bug because the ceil above takes
> > > you way past the space needed though (the memory device region is only 8 bytes long).
> > > 
> > >   
> > 
> > Maybe I misunderstand, but this is the intended behavior.
> > cxl_dstate->device_registers is the MemoryRegion container for all the
> > subregions that are the actual device MMIO.
> > 
> >                      +------------------+
> >                   ^  |                  |
> >                   |  |      unused      |
> >                   |  --------------------
> >                   |  |   memdev regs    |
> >                   |  --------------------
> >                   |  |                  |
> >                   |  | +--------------+ |
> >  cxl_dstate->     |  | |              | |
> >  device_registers |  | |              | |
> >                   |  | |payload       | |
> >                   |  | |(2k currently)| |
> >                   |  | |              | |
> >                   |  | |              | |
> >                   |  | +--------------+ |
> >                   |  |   mailbox regs   |
> >                   |  --------------------
> >                   |  |    device regs   |
> >                   v  --------------------
> >                      |     caps regs    |
> >          BAR ------> +------------------+
> > 
> > 
> > Perhaps I should add this as a comment in the code?
> 
> I agree with intent but above it is setting limit to the top of
> the mailbox regs, not the memdev regs which I'd expect to see.
> 

oops :D - you're correct
 
> > 
> > > > +
> > > > +    memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
> > > > +                          "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> > > > +
> > > > +    memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> > > > +                                &cxl_dstate->caps);
> > > > +}
> > > > +
> > > > +void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> > > > +{
> > > > +    uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> > > > +    const int cap_count = 0;
> > > > +
> > > > +    /* CXL Device Capabilities Array Register */
> > > > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> > > > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
> > > > +    ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY2, CAP_COUNT, cap_count);
> > > > +}
> > > > diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> > > > index 00c3876a0f..47154d6850 100644
> > > > --- a/hw/cxl/meson.build
> > > > +++ b/hw/cxl/meson.build
> > > > @@ -1,3 +1,4 @@
> > > >  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
> > > >    'cxl-component-utils.c',
> > > > +  'cxl-device-utils.c',
> > > >  ))  
> > >   
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [Linuxarm]  Re: [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-06 13:21   ` Jonathan Cameron
  2021-01-06 16:31     ` Ben Widawsky
@ 2021-01-06 17:40     ` Jonathan Cameron
  2021-01-06 18:05       ` Ben Widawsky
  1 sibling, 1 reply; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-06 17:40 UTC (permalink / raw)
  To: Ben Widawsky, Thomas Huth
  Cc: Dan, Vishal Verma, qemu-devel,
	Prashant V Agarwal  <agpr123@gmail.com>,
	Chris Browy <cbrowy@avery-design.com>,
	 Michael S. Tsirkin  <mst@redhat.com>,
	Philippe Mathieu-Daudé

On Wed, 6 Jan 2021 13:21:23 +0000
Jonathan Cameron <Jonathan.Cameron@Huawei.com> wrote:

> On Tue, 5 Jan 2021 08:52:58 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > This is the beginning of implementing mailbox support for CXL 2.0
> > devices.
> > 
> > v2: Use register alignment helper (Ben)
> >     Minor cleanups (Jonathan)
> >     Rename error codes to match spec (Jonathan)
> >     Update cap count from 1 to 2 (Jonathan)
> >     Add infra to support CEL (Ben)
> >     Add more of the actual mailbox handling from later patch (Ben)
> > 
> > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>  
> 
> Hi Ben,
> 
> I hacked support in for ARM64 to give this a spin and ran into an
> interesting problem around read sizes.
> 
> The mailbox registers space allows 4 or 8 byte reads, but in the kernel
> driver (I think I have the right version from your github) you do
> the payload drain with
> memcpy_from_io()
> 
> If the size of the payload is not a multiple of 8 bytes, on ARM64 that
> results in byte reads and an exception.  This happens with some of the
> existing calls which happen to have non multiple of 8 payload sizes.
> 
> I hacked below to allow 1 byte reads from that region but that's probably
> not the right fix.  I found a statement in the CXL spec saying maximum read
> size from this register block was 8 bytes but couldn't immediately see a minimum.
> (I haven't looked that hard yet though!)
> 
> Various approaches in kernel could also be used:
> 1) Change the payload drain to have specific handling for the end few bytes.
> 2) Pad the various structures to ensure payloads are always 8 byte multiples
> in length (nasty).

Bit more testing an another little thing below.

J

> 
> > ---
> >  hw/cxl/cxl-device-utils.c   | 122 ++++++++++++++++++++++++-
> >  hw/cxl/cxl-mailbox-utils.c  | 173 ++++++++++++++++++++++++++++++++++++
> >  hw/cxl/meson.build          |   1 +
> >  include/hw/cxl/cxl.h        |   3 +
> >  include/hw/cxl/cxl_device.h |  27 +++++-
> >  5 files changed, 322 insertions(+), 4 deletions(-)
> >  create mode 100644 hw/cxl/cxl-mailbox-utils.c
> > 
> > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > index b86e5466bd..642e3c2617 100644
> > --- a/hw/cxl/cxl-device-utils.c
> > +++ b/hw/cxl/cxl-device-utils.c
> > @@ -44,6 +44,108 @@ static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
> >      return ldn_le_p(&retval, size);
> >  }
> >  
> > +static uint64_t mailbox_reg_read(void *opaque, hwaddr offset, unsigned size)
> > +{
> > +    CXLDeviceState *cxl_dstate = opaque;
> > +
> > +    if (cxl_device_check_register_alignment(offset, size)) {
> > +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +        return 0;
> > +    }
> > +
> > +    return ldn_le_p(cxl_dstate->mbox_reg_state + offset, size);
> > +}
> > +
> > +static void mailbox_mem_writel(uint32_t *reg_state, hwaddr offset,
> > +                               uint64_t value)
> > +{
> > +    switch (offset) {
> > +    case A_CXL_DEV_MAILBOX_CTRL:
> > +        /* fallthrough */
> > +    case A_CXL_DEV_MAILBOX_CAP:
> > +        /* RO register */
> > +        break;
> > +    default:
> > +        qemu_log_mask(LOG_UNIMP,
> > +                      "%s Unexpected 32-bit access to 0x%" PRIx64 " (WI)\n",
> > +                      __func__, offset);
> > +        break;
> > +    }
> > +
> > +    stl_le_p((uint8_t *)reg_state + offset, value);
> > +}
> > +
> > +static void mailbox_mem_writeq(uint8_t *reg_state, hwaddr offset,
> > +                               uint64_t value)
> > +{
> > +    switch (offset) {
> > +    case A_CXL_DEV_MAILBOX_CMD:
> > +        break;
> > +    case A_CXL_DEV_BG_CMD_STS:
> > +        /* BG not supported */
> > +        /* fallthrough */
> > +    case A_CXL_DEV_MAILBOX_STS:
> > +        /* Read only register, will get updated by the state machine */
> > +        return;
> > +    default:
> > +        qemu_log_mask(LOG_UNIMP,
> > +                      "%s Unexpected 64-bit access to 0x%" PRIx64 " (WI)\n",
> > +                      __func__, offset);

I've been debugging mail box issues and it seems we can hit this path if
a payload is written by the OS.  Result is it never gets written into the memory
and hence isn't available when we try to read it below.

> > +        return;
> > +    }
> > +
> > +    stq_le_p(reg_state + offset, value);
> > +}
> > +
> > +static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
> > +                              unsigned size)
> > +{
> > +    CXLDeviceState *cxl_dstate = opaque;
> > +
> > +    /*
> > +     * Lock is needed to prevent concurrent writes as well as to prevent writes
> > +     * coming in while the firmware is processing. Without background commands
> > +     * or the second mailbox implemented, this serves no purpose since the
> > +     * memory access is synchronized at a higher level (per memory region).
> > +     */
> > +    RCU_READ_LOCK_GUARD();
> > +
> > +    switch (size) {
> > +    case 4:
> > +        if (unlikely(offset & (sizeof(uint32_t) - 1))) {
> > +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +            return;
> > +        }
> > +        mailbox_mem_writel(cxl_dstate->mbox_reg_state32, offset, value);
> > +        break;
> > +    case 8:
> > +        if (unlikely(offset & (sizeof(uint64_t) - 1))) {
> > +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > +            return;
> > +        }
> > +        mailbox_mem_writeq(cxl_dstate->mbox_reg_state, offset, value);
> > +        break;
> > +    }
> > +
> > +    if (ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > +                         DOORBELL))
> > +        cxl_process_mailbox(cxl_dstate);
> > +}
> > +
> > +static const MemoryRegionOps mailbox_ops = {
> > +    .read = mailbox_reg_read,
> > +    .write = mailbox_reg_write,
> > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > +    .valid = {
> > +        .min_access_size = 4,  
> 
> Hack was to set this to 1 (and the impl below - no idea if both are needed).
> 
> > +        .max_access_size = 8,
> > +    },
> > +    .impl = {
> > +        .min_access_size = 4,
> > +        .max_access_size = 8,
> > +    },
> > +};
> > +
> >  static const MemoryRegionOps dev_ops = {
> >      .read = dev_reg_read,
> >      .write = NULL,
> > @@ -83,20 +185,33 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> >                            "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> >      memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
> >                            "device-status", CXL_DEVICE_REGISTERS_LENGTH);
> > +    memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
> > +                          "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
> >  
> >      memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> >                                  &cxl_dstate->caps);
> >      memory_region_add_subregion(&cxl_dstate->device_registers,
> >                                  CXL_DEVICE_REGISTERS_OFFSET,
> >                                  &cxl_dstate->device);
> > +    memory_region_add_subregion(&cxl_dstate->device_registers,
> > +                                CXL_MAILBOX_REGISTERS_OFFSET,
> > +                                &cxl_dstate->mailbox);
> >  }
> >  
> >  static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
> >  
> > +static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
> > +{
> > +    /* 2048 payload size, with no interrupt or background support */
> > +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CAP,
> > +                     PAYLOAD_SIZE, CXL_MAILBOX_PAYLOAD_SHIFT);
> > +    cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
> > +}
> > +
> >  void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> >  {
> >      uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> > -    const int cap_count = 1;
> > +    const int cap_count = 2;
> >  
> >      /* CXL Device Capabilities Array Register */
> >      ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> > @@ -105,4 +220,9 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> >  
> >      cxl_device_cap_init(cxl_dstate, DEVICE, 1);
> >      device_reg_init_common(cxl_dstate);
> > +
> > +    cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
> > +    mailbox_reg_init_common(cxl_dstate);
> > +
> > +    assert(cxl_initialize_mailbox(cxl_dstate) == 0);
> >  }
> > diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> > new file mode 100644
> > index 0000000000..3a39c936de
> > --- /dev/null
> > +++ b/hw/cxl/cxl-mailbox-utils.c
> > @@ -0,0 +1,173 @@
> > +/*
> > + * CXL Utility library for mailbox interface
> > + *
> > + * Copyright(C) 2020 Intel Corporation.
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2. See the
> > + * COPYING file in the top-level directory.
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +#include "hw/cxl/cxl.h"
> > +#include "hw/pci/pci.h"
> > +#include "qemu/log.h"
> > +#include "qemu/uuid.h"
> > +
> > +/*
> > + * How to add a new command, example. The command set FOO, with cmd BAR.
> > + *  1. Add the command set and cmd to the enum.
> > + *     FOO    = 0x7f,
> > + *          #define BAR 0
> > + *  2. Forward declare the handler.
> > + *     declare_mailbox_handler(FOO_BAR);
> > + *  3. Add the command to the cxl_cmd_set[][]
> > + *     CXL_CMD(FOO, BAR, 0, 0),
> > + *  4. Implement your handler
> > + *     define_mailbox_handler(FOO_BAR) { ... return CXL_MBOX_SUCCESS; }
> > + *
> > + *
> > + *  Writing the handler:
> > + *    The handler will provide the &struct cxl_cmd, the &CXLDeviceState, and the
> > + *    in/out length of the payload. The handler is responsible for consuming the
> > + *    payload from cmd->payload and operating upon it as necessary. It must then
> > + *    fill the output data into cmd->payload (overwriting what was there),
> > + *    setting the length, and returning a valid return code.
> > + *
> > + *  XXX: The handler need not worry about endianess. The payload is read out of
> > + *  a register interface that already deals with it.
> > + */
> > +
> > +/* 8.2.8.4.5.1 Command Return Codes */
> > +typedef enum {
> > +    CXL_MBOX_SUCCESS = 0x0,
> > +    CXL_MBOX_BG_STARTED = 0x1,
> > +    CXL_MBOX_INVALID_INPUT = 0x2,
> > +    CXL_MBOX_UNSUPPORTED = 0x3,
> > +    CXL_MBOX_INTERNAL_ERROR = 0x4,
> > +    CXL_MBOX_RETRY_REQUIRED = 0x5,
> > +    CXL_MBOX_BUSY = 0x6,
> > +    CXL_MBOX_MEDIA_DISABLED = 0x7,
> > +    CXL_MBOX_FW_XFER_IN_PROGRESS = 0x8,
> > +    CXL_MBOX_FW_XFER_OUT_OF_ORDER = 0x9,
> > +    CXL_MBOX_FW_AUTH_FAILED = 0xa,
> > +    CXL_MBOX_FW_INVALID_SLOT = 0xb,
> > +    CXL_MBOX_FW_ROLLEDBACK = 0xc,
> > +    CXL_MBOX_FW_REST_REQD = 0xd,
> > +    CXL_MBOX_INVALID_HANDLE = 0xe,
> > +    CXL_MBOX_INVALID_PA = 0xf,
> > +    CXL_MBOX_INJECT_POISON_LIMIT = 0x10,
> > +    CXL_MBOX_PERMANENT_MEDIA_FAILURE = 0x11,
> > +    CXL_MBOX_ABORTED = 0x12,
> > +    CXL_MBOX_INVALID_SECURITY_STATE = 0x13,
> > +    CXL_MBOX_INCORRECT_PASSPHRASE = 0x14,
> > +    CXL_MBOX_UNSUPPORTED_MAILBOX = 0x15,
> > +    CXL_MBOX_INVALID_PAYLOAD_LENGTH = 0x16,
> > +    CXL_MBOX_MAX = 0x17
> > +} ret_code;
> > +
> > +struct cxl_cmd;
> > +typedef ret_code (*opcode_handler)(struct cxl_cmd *cmd,
> > +                                   CXLDeviceState *cxl_dstate, uint16_t *len);
> > +struct cxl_cmd {
> > +    const char *name;
> > +    opcode_handler handler;
> > +    ssize_t in;
> > +    uint16_t effect; /* Reported in CEL */
> > +    uint8_t *payload;
> > +};
> > +
> > +#define define_mailbox_handler(name)                \
> > +    static ret_code cmd_##name(struct cxl_cmd *cmd, \
> > +                               CXLDeviceState *cxl_dstate, uint16_t *len)
> > +#define declare_mailbox_handler(name) define_mailbox_handler(name)
> > +
> > +#define CXL_CMD(s, c, in, cel_effect) \
> > +    [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
> > +
> > +static struct cxl_cmd cxl_cmd_set[256][256] = {};
> > +
> > +#undef CXL_CMD
> > +
> > +QemuUUID cel_uuid;
> > +
> > +void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> > +{
> > +    uint16_t ret = CXL_MBOX_SUCCESS;
> > +    struct cxl_cmd *cxl_cmd;
> > +    uint64_t status_reg;
> > +    opcode_handler h;
> > +
> > +    /*
> > +     * current state of mailbox interface
> > +     *  mbox_cap_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CAP];
> > +     *  mbox_ctrl_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CTRL];
> > +     *  status_reg = *(uint64_t *)&cxl_dstate->reg_state[A_CXL_DEV_MAILBOX_STS];
> > +     */
> > +    uint64_t command_reg =
> > +        *(uint64_t *)&cxl_dstate->mbox_reg_state[A_CXL_DEV_MAILBOX_CMD];
> > +
> > +    /* Check if we have to do anything */
> > +    if (!ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > +                          DOORBELL)) {
> > +        qemu_log_mask(LOG_UNIMP, "Corrupt internal state for firmware\n");
> > +        return;
> > +    }
> > +
> > +    uint8_t set = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET);
> > +    uint8_t cmd = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND);
> > +    uint16_t len = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH);
> > +    cxl_cmd = &cxl_cmd_set[set][cmd];
> > +    h = cxl_cmd->handler;
> > +    if (!h) {
> > +        goto handled;
> > +    }
> > +
> > +    if (len != cxl_cmd->in) {
> > +        ret = CXL_MBOX_INVALID_PAYLOAD_LENGTH;
> > +    }
> > +
> > +    cxl_cmd->payload = cxl_dstate->mbox_reg_state + A_CXL_DEV_CMD_PAYLOAD;

The return highlighted above on which registers are writable meant the payload
never has anything useful in it.  Remove that and this seems to work and I
can query implemented commands + enable them.

> > +    ret = (*h)(cxl_cmd, cxl_dstate, &len);
> > +    assert(len <= cxl_dstate->payload_size);
> > +
> > +handled:
> > +    /*
> > +     * Set the return code
> > +     * XXX: it's a 64b register, but we're not setting the vendor, so we can get
> > +     * away with this
> > +     */
> > +    status_reg = FIELD_DP64(0, CXL_DEV_MAILBOX_STS, ERRNO, ret);
> > +
> > +    /*
> > +     * Set the return length
> > +     */
> > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET, 0);
> > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND, 0);
> > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH, len);
> > +
> > +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_CMD, command_reg);
> > +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_STS, status_reg);
> > +
> > +    /* Tell the host we're done */
> > +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > +                     DOORBELL, 0);
> > +}
> > +
> > +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate)
> > +{
> > +    const char *cel_uuidstr = "0da9c0b5-bf41-4b78-8f79-96b1623b3f17";
> > +
> > +    for (int i = 0; i < 256; i++) {
> > +        for (int j = 0; j < 256; j++) {
> > +            if (cxl_cmd_set[i][j].handler) {
> > +                struct cxl_cmd *c = &cxl_cmd_set[i][j];
> > +
> > +                cxl_dstate->cel_log[cxl_dstate->cel_size].opcode = (i << 8) | j;
> > +                cxl_dstate->cel_log[cxl_dstate->cel_size].effect = c->effect;
> > +                cxl_dstate->cel_size++;
> > +            }
> > +        }
> > +    }
> > +
> > +    return qemu_uuid_parse(cel_uuidstr, &cel_uuid);
> > +}
> > diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> > index 47154d6850..0eca715d10 100644
> > --- a/hw/cxl/meson.build
> > +++ b/hw/cxl/meson.build
> > @@ -1,4 +1,5 @@
> >  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
> >    'cxl-component-utils.c',
> >    'cxl-device-utils.c',
> > +  'cxl-mailbox-utils.c',
> >  ))
> > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > index 23f52c4cf9..362cda40de 100644
> > --- a/include/hw/cxl/cxl.h
> > +++ b/include/hw/cxl/cxl.h
> > @@ -14,5 +14,8 @@
> >  #include "cxl_component.h"
> >  #include "cxl_device.h"
> >  
> > +#define COMPONENT_REG_BAR_IDX 0
> > +#define DEVICE_REG_BAR_IDX 2
> > +
> >  #endif
> >  
> > diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> > index 3380fc9e7a..bdf990cec2 100644
> > --- a/include/hw/cxl/cxl_device.h
> > +++ b/include/hw/cxl/cxl_device.h
> > @@ -70,9 +70,10 @@
> >      (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
> >  
> >  typedef struct cxl_device_state {
> > -    /* Boss container and caps registers */
> > +    /* Main register container */
> >      MemoryRegion device_registers;
> >  
> > +    MemoryRegion device;
> >      struct {
> >          MemoryRegion caps;
> >          union {
> > @@ -80,8 +81,19 @@ typedef struct cxl_device_state {
> >              uint32_t caps_reg_state32[CXL_DEVICE_CAP_REG_SIZE >> 2];
> >          };
> >      };
> > -    MemoryRegion device;
> > -    MemoryRegion mailbox;
> > +    struct {
> > +        MemoryRegion mailbox;
> > +        uint16_t payload_size;
> > +        union {
> > +            uint8_t mbox_reg_state[CXL_MAILBOX_REGISTERS_LENGTH];
> > +            uint32_t mbox_reg_state32[CXL_MAILBOX_REGISTERS_LENGTH >> 2];
> > +        };
> > +        struct {
> > +            uint16_t opcode;
> > +            uint16_t effect;
> > +        } cel_log[1 << 16];
> > +        size_t cel_size;
> > +    };
> >  
> >      MemoryRegion *pmem;
> >      MemoryRegion *vmem;
> > @@ -122,6 +134,9 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
> >  CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
> >                                                 CXL_DEVICE_CAP_REG_SIZE)
> >  
> > +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
> > +void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
> > +
> >  #define cxl_device_cap_init(dstate, reg, cap_id)                                   \
> >      do {                                                                           \
> >          uint32_t *cap_hdrs = dstate->caps_reg_state32;                             \
> > @@ -149,6 +164,12 @@ REG32(CXL_DEV_MAILBOX_CTRL, 4)
> >      FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
> >      FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
> >  
> > +/* XXX: actually a 64b register */
> > +REG32(CXL_DEV_MAILBOX_CMD, 8)
> > +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
> > +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
> > +    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
> > +
> >  /* XXX: actually a 64b register */
> >  REG32(CXL_DEV_MAILBOX_STS, 0x10)
> >      FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)  
> _______________________________________________
> Linuxarm mailing list -- linuxarm@openeuler.org
> To unsubscribe send an email to linuxarm-leave@openeuler.org



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [Linuxarm]  Re: [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-06 17:40     ` [Linuxarm] " Jonathan Cameron
@ 2021-01-06 18:05       ` Ben Widawsky
  2021-01-06 19:08         ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-06 18:05 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Dan, Vishal Verma, qemu-devel,
	Prashant V Agarwal <agpr123@gmail.com>,
	Chris Browy <cbrowy@avery-design.com>,
	 Michael S. Tsirkin  <mst@redhat.com>,
	Philippe Mathieu-Daudé

On 21-01-06 17:40:14, Jonathan Cameron wrote:
> On Wed, 6 Jan 2021 13:21:23 +0000
> Jonathan Cameron <Jonathan.Cameron@Huawei.com> wrote:
> 
> > On Tue, 5 Jan 2021 08:52:58 -0800
> > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > 
> > > This is the beginning of implementing mailbox support for CXL 2.0
> > > devices.
> > > 
> > > v2: Use register alignment helper (Ben)
> > >     Minor cleanups (Jonathan)
> > >     Rename error codes to match spec (Jonathan)
> > >     Update cap count from 1 to 2 (Jonathan)
> > >     Add infra to support CEL (Ben)
> > >     Add more of the actual mailbox handling from later patch (Ben)
> > > 
> > > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>  
> > 
> > Hi Ben,
> > 
> > I hacked support in for ARM64 to give this a spin and ran into an
> > interesting problem around read sizes.
> > 
> > The mailbox registers space allows 4 or 8 byte reads, but in the kernel
> > driver (I think I have the right version from your github) you do
> > the payload drain with
> > memcpy_from_io()
> > 
> > If the size of the payload is not a multiple of 8 bytes, on ARM64 that
> > results in byte reads and an exception.  This happens with some of the
> > existing calls which happen to have non multiple of 8 payload sizes.
> > 
> > I hacked below to allow 1 byte reads from that region but that's probably
> > not the right fix.  I found a statement in the CXL spec saying maximum read
> > size from this register block was 8 bytes but couldn't immediately see a minimum.
> > (I haven't looked that hard yet though!)
> > 
> > Various approaches in kernel could also be used:
> > 1) Change the payload drain to have specific handling for the end few bytes.
> > 2) Pad the various structures to ensure payloads are always 8 byte multiples
> > in length (nasty).
> 
> Bit more testing an another little thing below.
> 
> J
> 
> > 
> > > ---
> > >  hw/cxl/cxl-device-utils.c   | 122 ++++++++++++++++++++++++-
> > >  hw/cxl/cxl-mailbox-utils.c  | 173 ++++++++++++++++++++++++++++++++++++
> > >  hw/cxl/meson.build          |   1 +
> > >  include/hw/cxl/cxl.h        |   3 +
> > >  include/hw/cxl/cxl_device.h |  27 +++++-
> > >  5 files changed, 322 insertions(+), 4 deletions(-)
> > >  create mode 100644 hw/cxl/cxl-mailbox-utils.c
> > > 
> > > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > > index b86e5466bd..642e3c2617 100644
> > > --- a/hw/cxl/cxl-device-utils.c
> > > +++ b/hw/cxl/cxl-device-utils.c
> > > @@ -44,6 +44,108 @@ static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
> > >      return ldn_le_p(&retval, size);
> > >  }
> > >  
> > > +static uint64_t mailbox_reg_read(void *opaque, hwaddr offset, unsigned size)
> > > +{
> > > +    CXLDeviceState *cxl_dstate = opaque;
> > > +
> > > +    if (cxl_device_check_register_alignment(offset, size)) {
> > > +        qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > > +        return 0;
> > > +    }
> > > +
> > > +    return ldn_le_p(cxl_dstate->mbox_reg_state + offset, size);
> > > +}
> > > +
> > > +static void mailbox_mem_writel(uint32_t *reg_state, hwaddr offset,
> > > +                               uint64_t value)
> > > +{
> > > +    switch (offset) {
> > > +    case A_CXL_DEV_MAILBOX_CTRL:
> > > +        /* fallthrough */
> > > +    case A_CXL_DEV_MAILBOX_CAP:
> > > +        /* RO register */
> > > +        break;
> > > +    default:
> > > +        qemu_log_mask(LOG_UNIMP,
> > > +                      "%s Unexpected 32-bit access to 0x%" PRIx64 " (WI)\n",
> > > +                      __func__, offset);
> > > +        break;
> > > +    }
> > > +
> > > +    stl_le_p((uint8_t *)reg_state + offset, value);
> > > +}
> > > +
> > > +static void mailbox_mem_writeq(uint8_t *reg_state, hwaddr offset,
> > > +                               uint64_t value)
> > > +{
> > > +    switch (offset) {
> > > +    case A_CXL_DEV_MAILBOX_CMD:
> > > +        break;
> > > +    case A_CXL_DEV_BG_CMD_STS:
> > > +        /* BG not supported */
> > > +        /* fallthrough */
> > > +    case A_CXL_DEV_MAILBOX_STS:
> > > +        /* Read only register, will get updated by the state machine */
> > > +        return;
> > > +    default:
> > > +        qemu_log_mask(LOG_UNIMP,
> > > +                      "%s Unexpected 64-bit access to 0x%" PRIx64 " (WI)\n",
> > > +                      __func__, offset);
> 
> I've been debugging mail box issues and it seems we can hit this path if
> a payload is written by the OS.  Result is it never gets written into the memory
> and hence isn't available when we try to read it below.
> 

I'm sorry you had to debug this. I had fixed this previously and it got lost.
I'm currently between test applications, so my regression testing isn't great.

I think the fix should be something like this, but I can't easily test at the
moment:

diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
index c515d45d20..b38e9b4c17 100644
--- a/hw/cxl/cxl-device-utils.c
+++ b/hw/cxl/cxl-device-utils.c
@@ -102,6 +102,9 @@ static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
 {
     CXLDeviceState *cxl_dstate = opaque;

+    if (offset >= A_CXL_DEV_CMD_PAYLOAD)
+        stn_le_p(cxl_dstate->mbox_reg_state, size, value);
+
     /*
      * Lock is needed to prevent concurrent writes as well as to prevent writes
      * coming in while the firmware is processing. Without background commands



> > > +        return;
> > > +    }
> > > +
> > > +    stq_le_p(reg_state + offset, value);
> > > +}
> > > +
> > > +static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
> > > +                              unsigned size)
> > > +{
> > > +    CXLDeviceState *cxl_dstate = opaque;
> > > +
> > > +    /*
> > > +     * Lock is needed to prevent concurrent writes as well as to prevent writes
> > > +     * coming in while the firmware is processing. Without background commands
> > > +     * or the second mailbox implemented, this serves no purpose since the
> > > +     * memory access is synchronized at a higher level (per memory region).
> > > +     */
> > > +    RCU_READ_LOCK_GUARD();
> > > +
> > > +    switch (size) {
> > > +    case 4:
> > > +        if (unlikely(offset & (sizeof(uint32_t) - 1))) {
> > > +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > > +            return;
> > > +        }
> > > +        mailbox_mem_writel(cxl_dstate->mbox_reg_state32, offset, value);
> > > +        break;
> > > +    case 8:
> > > +        if (unlikely(offset & (sizeof(uint64_t) - 1))) {
> > > +            qemu_log_mask(LOG_UNIMP, "Unaligned register read\n");
> > > +            return;
> > > +        }
> > > +        mailbox_mem_writeq(cxl_dstate->mbox_reg_state, offset, value);
> > > +        break;
> > > +    }
> > > +
> > > +    if (ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > > +                         DOORBELL))
> > > +        cxl_process_mailbox(cxl_dstate);
> > > +}
> > > +
> > > +static const MemoryRegionOps mailbox_ops = {
> > > +    .read = mailbox_reg_read,
> > > +    .write = mailbox_reg_write,
> > > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > > +    .valid = {
> > > +        .min_access_size = 4,  
> > 
> > Hack was to set this to 1 (and the impl below - no idea if both are needed).
> > 
> > > +        .max_access_size = 8,
> > > +    },
> > > +    .impl = {
> > > +        .min_access_size = 4,
> > > +        .max_access_size = 8,
> > > +    },
> > > +};
> > > +
> > >  static const MemoryRegionOps dev_ops = {
> > >      .read = dev_reg_read,
> > >      .write = NULL,
> > > @@ -83,20 +185,33 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
> > >                            "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
> > >      memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
> > >                            "device-status", CXL_DEVICE_REGISTERS_LENGTH);
> > > +    memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
> > > +                          "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
> > >  
> > >      memory_region_add_subregion(&cxl_dstate->device_registers, 0,
> > >                                  &cxl_dstate->caps);
> > >      memory_region_add_subregion(&cxl_dstate->device_registers,
> > >                                  CXL_DEVICE_REGISTERS_OFFSET,
> > >                                  &cxl_dstate->device);
> > > +    memory_region_add_subregion(&cxl_dstate->device_registers,
> > > +                                CXL_MAILBOX_REGISTERS_OFFSET,
> > > +                                &cxl_dstate->mailbox);
> > >  }
> > >  
> > >  static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
> > >  
> > > +static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
> > > +{
> > > +    /* 2048 payload size, with no interrupt or background support */
> > > +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CAP,
> > > +                     PAYLOAD_SIZE, CXL_MAILBOX_PAYLOAD_SHIFT);
> > > +    cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
> > > +}
> > > +
> > >  void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> > >  {
> > >      uint32_t *cap_hdrs = cxl_dstate->caps_reg_state32;
> > > -    const int cap_count = 1;
> > > +    const int cap_count = 2;
> > >  
> > >      /* CXL Device Capabilities Array Register */
> > >      ARRAY_FIELD_DP32(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
> > > @@ -105,4 +220,9 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
> > >  
> > >      cxl_device_cap_init(cxl_dstate, DEVICE, 1);
> > >      device_reg_init_common(cxl_dstate);
> > > +
> > > +    cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
> > > +    mailbox_reg_init_common(cxl_dstate);
> > > +
> > > +    assert(cxl_initialize_mailbox(cxl_dstate) == 0);
> > >  }
> > > diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> > > new file mode 100644
> > > index 0000000000..3a39c936de
> > > --- /dev/null
> > > +++ b/hw/cxl/cxl-mailbox-utils.c
> > > @@ -0,0 +1,173 @@
> > > +/*
> > > + * CXL Utility library for mailbox interface
> > > + *
> > > + * Copyright(C) 2020 Intel Corporation.
> > > + *
> > > + * This work is licensed under the terms of the GNU GPL, version 2. See the
> > > + * COPYING file in the top-level directory.
> > > + */
> > > +
> > > +#include "qemu/osdep.h"
> > > +#include "hw/cxl/cxl.h"
> > > +#include "hw/pci/pci.h"
> > > +#include "qemu/log.h"
> > > +#include "qemu/uuid.h"
> > > +
> > > +/*
> > > + * How to add a new command, example. The command set FOO, with cmd BAR.
> > > + *  1. Add the command set and cmd to the enum.
> > > + *     FOO    = 0x7f,
> > > + *          #define BAR 0
> > > + *  2. Forward declare the handler.
> > > + *     declare_mailbox_handler(FOO_BAR);
> > > + *  3. Add the command to the cxl_cmd_set[][]
> > > + *     CXL_CMD(FOO, BAR, 0, 0),
> > > + *  4. Implement your handler
> > > + *     define_mailbox_handler(FOO_BAR) { ... return CXL_MBOX_SUCCESS; }
> > > + *
> > > + *
> > > + *  Writing the handler:
> > > + *    The handler will provide the &struct cxl_cmd, the &CXLDeviceState, and the
> > > + *    in/out length of the payload. The handler is responsible for consuming the
> > > + *    payload from cmd->payload and operating upon it as necessary. It must then
> > > + *    fill the output data into cmd->payload (overwriting what was there),
> > > + *    setting the length, and returning a valid return code.
> > > + *
> > > + *  XXX: The handler need not worry about endianess. The payload is read out of
> > > + *  a register interface that already deals with it.
> > > + */
> > > +
> > > +/* 8.2.8.4.5.1 Command Return Codes */
> > > +typedef enum {
> > > +    CXL_MBOX_SUCCESS = 0x0,
> > > +    CXL_MBOX_BG_STARTED = 0x1,
> > > +    CXL_MBOX_INVALID_INPUT = 0x2,
> > > +    CXL_MBOX_UNSUPPORTED = 0x3,
> > > +    CXL_MBOX_INTERNAL_ERROR = 0x4,
> > > +    CXL_MBOX_RETRY_REQUIRED = 0x5,
> > > +    CXL_MBOX_BUSY = 0x6,
> > > +    CXL_MBOX_MEDIA_DISABLED = 0x7,
> > > +    CXL_MBOX_FW_XFER_IN_PROGRESS = 0x8,
> > > +    CXL_MBOX_FW_XFER_OUT_OF_ORDER = 0x9,
> > > +    CXL_MBOX_FW_AUTH_FAILED = 0xa,
> > > +    CXL_MBOX_FW_INVALID_SLOT = 0xb,
> > > +    CXL_MBOX_FW_ROLLEDBACK = 0xc,
> > > +    CXL_MBOX_FW_REST_REQD = 0xd,
> > > +    CXL_MBOX_INVALID_HANDLE = 0xe,
> > > +    CXL_MBOX_INVALID_PA = 0xf,
> > > +    CXL_MBOX_INJECT_POISON_LIMIT = 0x10,
> > > +    CXL_MBOX_PERMANENT_MEDIA_FAILURE = 0x11,
> > > +    CXL_MBOX_ABORTED = 0x12,
> > > +    CXL_MBOX_INVALID_SECURITY_STATE = 0x13,
> > > +    CXL_MBOX_INCORRECT_PASSPHRASE = 0x14,
> > > +    CXL_MBOX_UNSUPPORTED_MAILBOX = 0x15,
> > > +    CXL_MBOX_INVALID_PAYLOAD_LENGTH = 0x16,
> > > +    CXL_MBOX_MAX = 0x17
> > > +} ret_code;
> > > +
> > > +struct cxl_cmd;
> > > +typedef ret_code (*opcode_handler)(struct cxl_cmd *cmd,
> > > +                                   CXLDeviceState *cxl_dstate, uint16_t *len);
> > > +struct cxl_cmd {
> > > +    const char *name;
> > > +    opcode_handler handler;
> > > +    ssize_t in;
> > > +    uint16_t effect; /* Reported in CEL */
> > > +    uint8_t *payload;
> > > +};
> > > +
> > > +#define define_mailbox_handler(name)                \
> > > +    static ret_code cmd_##name(struct cxl_cmd *cmd, \
> > > +                               CXLDeviceState *cxl_dstate, uint16_t *len)
> > > +#define declare_mailbox_handler(name) define_mailbox_handler(name)
> > > +
> > > +#define CXL_CMD(s, c, in, cel_effect) \
> > > +    [s][c] = { stringify(s##_##c), cmd_##s##_##c, in, cel_effect }
> > > +
> > > +static struct cxl_cmd cxl_cmd_set[256][256] = {};
> > > +
> > > +#undef CXL_CMD
> > > +
> > > +QemuUUID cel_uuid;
> > > +
> > > +void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> > > +{
> > > +    uint16_t ret = CXL_MBOX_SUCCESS;
> > > +    struct cxl_cmd *cxl_cmd;
> > > +    uint64_t status_reg;
> > > +    opcode_handler h;
> > > +
> > > +    /*
> > > +     * current state of mailbox interface
> > > +     *  mbox_cap_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CAP];
> > > +     *  mbox_ctrl_reg = cxl_dstate->reg_state32[R_CXL_DEV_MAILBOX_CTRL];
> > > +     *  status_reg = *(uint64_t *)&cxl_dstate->reg_state[A_CXL_DEV_MAILBOX_STS];
> > > +     */
> > > +    uint64_t command_reg =
> > > +        *(uint64_t *)&cxl_dstate->mbox_reg_state[A_CXL_DEV_MAILBOX_CMD];
> > > +
> > > +    /* Check if we have to do anything */
> > > +    if (!ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > > +                          DOORBELL)) {
> > > +        qemu_log_mask(LOG_UNIMP, "Corrupt internal state for firmware\n");
> > > +        return;
> > > +    }
> > > +
> > > +    uint8_t set = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET);
> > > +    uint8_t cmd = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND);
> > > +    uint16_t len = FIELD_EX64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH);
> > > +    cxl_cmd = &cxl_cmd_set[set][cmd];
> > > +    h = cxl_cmd->handler;
> > > +    if (!h) {
> > > +        goto handled;
> > > +    }
> > > +
> > > +    if (len != cxl_cmd->in) {
> > > +        ret = CXL_MBOX_INVALID_PAYLOAD_LENGTH;
> > > +    }
> > > +
> > > +    cxl_cmd->payload = cxl_dstate->mbox_reg_state + A_CXL_DEV_CMD_PAYLOAD;
> 
> The return highlighted above on which registers are writable meant the payload
> never has anything useful in it.  Remove that and this seems to work and I
> can query implemented commands + enable them.
> 
> > > +    ret = (*h)(cxl_cmd, cxl_dstate, &len);
> > > +    assert(len <= cxl_dstate->payload_size);
> > > +
> > > +handled:
> > > +    /*
> > > +     * Set the return code
> > > +     * XXX: it's a 64b register, but we're not setting the vendor, so we can get
> > > +     * away with this
> > > +     */
> > > +    status_reg = FIELD_DP64(0, CXL_DEV_MAILBOX_STS, ERRNO, ret);
> > > +
> > > +    /*
> > > +     * Set the return length
> > > +     */
> > > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND_SET, 0);
> > > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, COMMAND, 0);
> > > +    command_reg = FIELD_DP64(command_reg, CXL_DEV_MAILBOX_CMD, LENGTH, len);
> > > +
> > > +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_CMD, command_reg);
> > > +    stq_le_p(cxl_dstate->mbox_reg_state + A_CXL_DEV_MAILBOX_STS, status_reg);
> > > +
> > > +    /* Tell the host we're done */
> > > +    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
> > > +                     DOORBELL, 0);
> > > +}
> > > +
> > > +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate)
> > > +{
> > > +    const char *cel_uuidstr = "0da9c0b5-bf41-4b78-8f79-96b1623b3f17";
> > > +
> > > +    for (int i = 0; i < 256; i++) {
> > > +        for (int j = 0; j < 256; j++) {
> > > +            if (cxl_cmd_set[i][j].handler) {
> > > +                struct cxl_cmd *c = &cxl_cmd_set[i][j];
> > > +
> > > +                cxl_dstate->cel_log[cxl_dstate->cel_size].opcode = (i << 8) | j;
> > > +                cxl_dstate->cel_log[cxl_dstate->cel_size].effect = c->effect;
> > > +                cxl_dstate->cel_size++;
> > > +            }
> > > +        }
> > > +    }
> > > +
> > > +    return qemu_uuid_parse(cel_uuidstr, &cel_uuid);
> > > +}
> > > diff --git a/hw/cxl/meson.build b/hw/cxl/meson.build
> > > index 47154d6850..0eca715d10 100644
> > > --- a/hw/cxl/meson.build
> > > +++ b/hw/cxl/meson.build
> > > @@ -1,4 +1,5 @@
> > >  softmmu_ss.add(when: 'CONFIG_CXL', if_true: files(
> > >    'cxl-component-utils.c',
> > >    'cxl-device-utils.c',
> > > +  'cxl-mailbox-utils.c',
> > >  ))
> > > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > > index 23f52c4cf9..362cda40de 100644
> > > --- a/include/hw/cxl/cxl.h
> > > +++ b/include/hw/cxl/cxl.h
> > > @@ -14,5 +14,8 @@
> > >  #include "cxl_component.h"
> > >  #include "cxl_device.h"
> > >  
> > > +#define COMPONENT_REG_BAR_IDX 0
> > > +#define DEVICE_REG_BAR_IDX 2
> > > +
> > >  #endif
> > >  
> > > diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> > > index 3380fc9e7a..bdf990cec2 100644
> > > --- a/include/hw/cxl/cxl_device.h
> > > +++ b/include/hw/cxl/cxl_device.h
> > > @@ -70,9 +70,10 @@
> > >      (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
> > >  
> > >  typedef struct cxl_device_state {
> > > -    /* Boss container and caps registers */
> > > +    /* Main register container */
> > >      MemoryRegion device_registers;
> > >  
> > > +    MemoryRegion device;
> > >      struct {
> > >          MemoryRegion caps;
> > >          union {
> > > @@ -80,8 +81,19 @@ typedef struct cxl_device_state {
> > >              uint32_t caps_reg_state32[CXL_DEVICE_CAP_REG_SIZE >> 2];
> > >          };
> > >      };
> > > -    MemoryRegion device;
> > > -    MemoryRegion mailbox;
> > > +    struct {
> > > +        MemoryRegion mailbox;
> > > +        uint16_t payload_size;
> > > +        union {
> > > +            uint8_t mbox_reg_state[CXL_MAILBOX_REGISTERS_LENGTH];
> > > +            uint32_t mbox_reg_state32[CXL_MAILBOX_REGISTERS_LENGTH >> 2];
> > > +        };
> > > +        struct {
> > > +            uint16_t opcode;
> > > +            uint16_t effect;
> > > +        } cel_log[1 << 16];
> > > +        size_t cel_size;
> > > +    };
> > >  
> > >      MemoryRegion *pmem;
> > >      MemoryRegion *vmem;
> > > @@ -122,6 +134,9 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
> > >  CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
> > >                                                 CXL_DEVICE_CAP_REG_SIZE)
> > >  
> > > +int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
> > > +void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
> > > +
> > >  #define cxl_device_cap_init(dstate, reg, cap_id)                                   \
> > >      do {                                                                           \
> > >          uint32_t *cap_hdrs = dstate->caps_reg_state32;                             \
> > > @@ -149,6 +164,12 @@ REG32(CXL_DEV_MAILBOX_CTRL, 4)
> > >      FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
> > >      FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
> > >  
> > > +/* XXX: actually a 64b register */
> > > +REG32(CXL_DEV_MAILBOX_CMD, 8)
> > > +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
> > > +    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
> > > +    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
> > > +
> > >  /* XXX: actually a 64b register */
> > >  REG32(CXL_DEV_MAILBOX_STS, 0x10)
> > >      FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)  
> > _______________________________________________
> > Linuxarm mailing list -- linuxarm@openeuler.org
> > To unsubscribe send an email to linuxarm-leave@openeuler.org
> 


^ permalink raw reply related	[flat|nested] 57+ messages in thread

* Re: [Linuxarm]  Re: [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-06 18:05       ` Ben Widawsky
@ 2021-01-06 19:08         ` Ben Widawsky
  2021-01-08  5:36           ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-06 19:08 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Dan, Vishal Verma, qemu-devel,
	Prashant V Agarwal <agpr123@gmail.com>,
	Chris Browy <cbrowy@avery-design.com>,
	 Michael S. Tsirkin  <mst@redhat.com>,
	Philippe Mathieu-Daudé

On 21-01-06 10:05:57, Ben Widawsky wrote:
> On 21-01-06 17:40:14, Jonathan Cameron wrote:
> > On Wed, 6 Jan 2021 13:21:23 +0000
> > Jonathan Cameron <Jonathan.Cameron@Huawei.com> wrote:
> > 
> > > On Tue, 5 Jan 2021 08:52:58 -0800
> > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > 

[snip]

> 
> I'm sorry you had to debug this. I had fixed this previously and it got lost.
> I'm currently between test applications, so my regression testing isn't great.
> 
> I think the fix should be something like this, but I can't easily test at the
> moment:
> 
> diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> index c515d45d20..b38e9b4c17 100644
> --- a/hw/cxl/cxl-device-utils.c
> +++ b/hw/cxl/cxl-device-utils.c
> @@ -102,6 +102,9 @@ static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
>  {
>      CXLDeviceState *cxl_dstate = opaque;
> 
> +    if (offset >= A_CXL_DEV_CMD_PAYLOAD)
> +        stn_le_p(cxl_dstate->mbox_reg_state, size, value);
> +
>      /*
>       * Lock is needed to prevent concurrent writes as well as to prevent writes
>       * coming in while the firmware is processing. Without background commands
> 
> 
> 

+    if (offset >= A_CXL_DEV_CMD_PAYLOAD) {
+        stn_le_p(cxl_dstate->mbox_reg_state, size, value);
+        return;
+    }
+

[snip]

> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [Linuxarm]  Re: [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4)
  2021-01-06 19:08         ` Ben Widawsky
@ 2021-01-08  5:36           ` Ben Widawsky
  0 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-08  5:36 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Dan, Vishal Verma, qemu-devel,
	Prashant V Agarwal <agpr123@gmail.com>,
	Chris Browy <cbrowy@avery-design.com>,
	 Michael S. Tsirkin  <mst@redhat.com>,
	Philippe Mathieu-Daudé

On 21-01-06 11:08:28, Ben Widawsky wrote:
> On 21-01-06 10:05:57, Ben Widawsky wrote:
> > On 21-01-06 17:40:14, Jonathan Cameron wrote:
> > > On Wed, 6 Jan 2021 13:21:23 +0000
> > > Jonathan Cameron <Jonathan.Cameron@Huawei.com> wrote:
> > > 
> > > > On Tue, 5 Jan 2021 08:52:58 -0800
> > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > > 
> 
> [snip]
> 
> > 
> > I'm sorry you had to debug this. I had fixed this previously and it got lost.
> > I'm currently between test applications, so my regression testing isn't great.
> > 
> > I think the fix should be something like this, but I can't easily test at the
> > moment:
> > 
> > diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> > index c515d45d20..b38e9b4c17 100644
> > --- a/hw/cxl/cxl-device-utils.c
> > +++ b/hw/cxl/cxl-device-utils.c
> > @@ -102,6 +102,9 @@ static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
> >  {
> >      CXLDeviceState *cxl_dstate = opaque;
> > 
> > +    if (offset >= A_CXL_DEV_CMD_PAYLOAD)
> > +        stn_le_p(cxl_dstate->mbox_reg_state, size, value);
> > +
> >      /*
> >       * Lock is needed to prevent concurrent writes as well as to prevent writes
> >       * coming in while the firmware is processing. Without background commands
> > 
> > 
> > 
> 
> +    if (offset >= A_CXL_DEV_CMD_PAYLOAD) {
> +        stn_le_p(cxl_dstate->mbox_reg_state, size, value);
> +        return;
> +    }
> +

Last time's a charm
         stn_le_p(cxl_dstate->mbox_reg_state + offset, size, value);

> 
> [snip]
> 
> > 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 00/32] CXL 2.0 Support
  2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
                   ` (31 preceding siblings ...)
  2021-01-05 16:53 ` [RFC PATCH v2 32/32] qtest/cxl: Add very basic sanity tests Ben Widawsky
@ 2021-01-08 18:44 ` Jonathan Cameron
  2021-01-08 18:51   ` Ben Widawsky
  32 siblings, 1 reply; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-08 18:44 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On Tue, 5 Jan 2021 08:52:51 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> Fixes since v1 [1]:
>  * Defer introducing some commands/registers not yet used (Ben)
>  * Add stubbed device_reg_init_common() (Ben)
>  * Improve assertions in DVSEC creation (Jonathan)
>  * Use 'n' for HDM register offsets (Jonathan)
>  * Correct revision ID for extensions (Jonathan)
>  * Minor cleanups and clarifications (Jonathan)
>  * Remove error codes not yet used (Jonathan)
>  * Fix interrupt enable bit width (Jonathan)
>  * Add comment for weird register size (Jonathan)
>  * Break out register alignment checks (Jonathan)
>  * Use the reg alignment helper (Jonathan)
>  * Rename error codes to match spec
>  * Fix cap count mid series (Jonathan)
> 
> New since v1 [1]:
>  * Entirely reworked framework for firmware handling
>  * Implemented more device commands
>  * CEL support
> 
> (There are some new patches that I just named 'v2' for simplicity's sake)
> 
> Introduce emulation of Compute Express Link 2.0
> (https://www.computeexpresslink.org/).

Hi Ben,

In interests of avoid duplication, I thought I'd give a quick update of where
I am. I have what is here working on arm64 (enabling is fairly simple).
I'll have a bit more feedback next week and hopefully have time for an indepth review.

I did run into some issues around alignment for the persistent memory that
(I think) are down to the fact the memory-backend isn't quite tied up to the
device.

Flows wise, I initially started hacking in NFIT building changes in qemu but I'm not
yet sure how the hotplug flows are supposed to work and who is responsible for
setting up a hotplugged device (OS or firmware).

Plans wise, I was thinking I could look at DOE emulation as seems that we'll
be needing that fairly soon to get any useful memory usecases up and running.
Let me know if you already have that underway.  

Jonathan

> 
> The emulation has been critical to get the Linux enabling started [2], it would
> be an ideal place to land regression tests for different topology handling, and
> there may be applications for this emulation as a way for a guest to manipulate
> its address space relative to different performance memories.
> 
> Three of the five CXL component types are emulated with some level of functionality:
> host bridge, root port, and memory device. Upstream ports and downstream ports
> aren't implemented (the two components needed to make up a switch).
> 
> CXL 2.0 is built on top of PCIe (see spec for details). As a result, much of the
> implementation utilizes existing PCI paradigms. To implement the host bridge,
> I've chosen to use PXB (PCI Expander Bridge). It seemed to be the most natural
> fit even though it doesn't directly map to how hardware will work. For
> persistent capacity of the memory device, I utilized the memory subsystem
> (hw/mem).
> 
> We have 3 reasons why this work is valuable:
> 1. OS driver development and testing
> 2. OS driver regression testing
> 3. Possible guest support for HDMs
> 
> As mentioned above there are three benefits to carrying this enabling in
> upstream QEMU:
> 
> 1. Linux driver feature development benefits from emulation both due to
> a lack of initial hardware availability, but also, as is seen with
> NVDIMM/PMEM emulation, there is value in being able to share
> topologies with system-software developers even after hardware is
> available.
> 
> 2. The Linux kernel's unit test suite for NVDIMM/PMEM ended up injecting fake
> resources via custom modules (nfit_test). In retrospect a QEMU emulation of
> nfit_test capabilities would have made the test environment more portable, and
> allowed for easier community contributions of example configurations.
> 
> 3. This is still being fleshed out, but in short it provides a standardized
> mechanism for the guest to provide feedback to the host about size and placement
> needs of the memory. After the host gives the guest a physical window mapping to
> the CXL device, the emulated HDM decoders allow the guest a way to tell the host
> how much it wants and where. There are likely simpler ways to do this, but
> they'd require inventing a new interface and you'd need to have diverging driver
> code in the guest programming of the HDM decoder vs. the host. Since we've
> already done this work, why not use it?
> 
> There is quite a long list of work to do for full spec compliance, but I don't
> believe that any of it precludes merging. Off the top of my head:
> - Main host bridge support (WIP)
> - Interleaving
> - Better Tests
> - Huge swaths of firmware functionality
> - Hot plug support
> - Emulating volatile capacity
> 
> The flow of the patches in general is to define all the data structures and
> registers associated with the various components in a top down manner. Host
> bridge, component, ports, devices. Then, the actual implementation is done in
> the same order.
> 
> The summary is:
> 1-8: Put infrastructure in place for emulation of the components.
> 9-12: Implement device mailboxes
> 13-15: Create the concept of a CXL bus and plumb into PXB
> 16-22: Implement host bridges
> 23: Implement a root port
> 24: Implement a memory device
> 25: Implement HDM decoders
> 26-30: ACPI bits
> 31: Start working on enabling the main host bridge
> 32: Basic test case
> 
> [1]: https://lore.kernel.org/qemu-devel/20201111054724.794888-1-ben.widawsky@intel.com/
> [2]: https://lore.kernel.org/linux-cxl/20201209002418.1976362-1-ben.widawsky@intel.com/
> 
> Ben Widawsky (31):
>   hw/pci/cxl: Add a CXL component type (interface)
>   hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)
>   hw/cxl/device: Introduce a CXL device (8.2.8)
>   hw/cxl/device: Implement the CAP array (8.2.8.1-2)
>   hw/cxl/device: Add device status (8.2.8.3)
>   hw/cxl/device: Implement basic mailbox (8.2.8.4)
>   hw/cxl/device: Add memory devices (8.2.8.5)
>   hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1)
>   hw/cxl/device: Placeholder for firmware commands (8.2.9.2)
>   hw/cxl/device: Timestamp implementation (8.2.9.3)
>   hw/cxl/device: Add log commands (8.2.9.4) + CEL
>   hw/pxb: Use a type for realizing expanders
>   hw/pci/cxl: Create a CXL bus type
>   hw/pxb: Allow creation of a CXL PXB (host bridge)
>   qtest: allow DSDT acpi table changes
>   acpi/pci: Consolidate host bridge setup
>   tests/acpi: remove stale allowed tables
>   hw/pci: Plumb _UID through host bridges
>   hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)
>   acpi/pxb/cxl: Reserve host bridge MMIO
>   hw/pxb/cxl: Add "windows" for host bridges
>   hw/cxl/rp: Add a root port
>   hw/cxl/device: Add a memory device (8.2.8.5)
>   hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12)
>   acpi/cxl: Add _OSC implementation (9.14.2)
>   tests/acpi: allow CEDT table addition
>   acpi/cxl: Create the CEDT (9.14.1)
>   Temp: acpi/cxl: Add ACPI0017 (CEDT awareness)
>   tests/acpi: Add new CEDT files
>   WIP: i386/cxl: Initialize a host bridge
>   qtest/cxl: Add very basic sanity tests
> 
> Jonathan Cameron (1):
>   Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy.
> 
>  MAINTAINERS                               |   6 +
>  hw/Kconfig                                |   1 +
>  hw/acpi/Kconfig                           |   5 +
>  hw/acpi/cxl.c                             | 173 ++++++++++
>  hw/acpi/meson.build                       |   1 +
>  hw/arm/virt.c                             |   1 +
>  hw/core/machine.c                         |  26 ++
>  hw/core/numa.c                            |   3 +
>  hw/cxl/Kconfig                            |   3 +
>  hw/cxl/cxl-component-utils.c              | 193 +++++++++++
>  hw/cxl/cxl-device-utils.c                 | 278 ++++++++++++++++
>  hw/cxl/cxl-mailbox-utils.c                | 386 ++++++++++++++++++++++
>  hw/cxl/meson.build                        |   5 +
>  hw/i386/acpi-build.c                      |  87 ++++-
>  hw/i386/microvm.c                         |   1 +
>  hw/i386/pc.c                              |   2 +
>  hw/mem/Kconfig                            |   5 +
>  hw/mem/cxl_type3.c                        | 334 +++++++++++++++++++
>  hw/mem/meson.build                        |   1 +
>  hw/meson.build                            |   1 +
>  hw/pci-bridge/Kconfig                     |   5 +
>  hw/pci-bridge/cxl_root_port.c             | 231 +++++++++++++
>  hw/pci-bridge/meson.build                 |   1 +
>  hw/pci-bridge/pci_expander_bridge.c       | 209 +++++++++++-
>  hw/pci-bridge/pcie_root_port.c            |   6 +-
>  hw/pci/pci.c                              |  32 +-
>  hw/pci/pcie.c                             |  30 ++
>  hw/ppc/spapr.c                            |   2 +
>  include/hw/acpi/cxl.h                     |  27 ++
>  include/hw/boards.h                       |   2 +
>  include/hw/cxl/cxl.h                      |  30 ++
>  include/hw/cxl/cxl_component.h            | 187 +++++++++++
>  include/hw/cxl/cxl_device.h               | 210 ++++++++++++
>  include/hw/cxl/cxl_pci.h                  | 160 +++++++++
>  include/hw/pci/pci.h                      |  15 +
>  include/hw/pci/pci_bridge.h               |  25 ++
>  include/hw/pci/pci_bus.h                  |   8 +
>  include/hw/pci/pci_ids.h                  |   1 +
>  include/standard-headers/linux/pci_regs.h |   1 +
>  monitor/hmp-cmds.c                        |  15 +
>  qapi/machine.json                         |   1 +
>  tests/data/acpi/pc/CEDT                   | Bin 0 -> 36 bytes
>  tests/data/acpi/pc/DSDT                   | Bin 5065 -> 5065 bytes
>  tests/data/acpi/pc/DSDT.acpihmat          | Bin 6390 -> 6390 bytes
>  tests/data/acpi/pc/DSDT.bridge            | Bin 6924 -> 6924 bytes
>  tests/data/acpi/pc/DSDT.cphp              | Bin 5529 -> 5529 bytes
>  tests/data/acpi/pc/DSDT.dimmpxm           | Bin 6719 -> 6719 bytes
>  tests/data/acpi/pc/DSDT.hpbridge          | Bin 5026 -> 5026 bytes
>  tests/data/acpi/pc/DSDT.hpbrroot          | Bin 3084 -> 3084 bytes
>  tests/data/acpi/pc/DSDT.ipmikcs           | Bin 5137 -> 5137 bytes
>  tests/data/acpi/pc/DSDT.memhp             | Bin 6424 -> 6424 bytes
>  tests/data/acpi/pc/DSDT.numamem           | Bin 5071 -> 5071 bytes
>  tests/data/acpi/pc/DSDT.roothp            | Bin 5261 -> 5261 bytes
>  tests/data/acpi/q35/CEDT                  | Bin 0 -> 36 bytes
>  tests/data/acpi/q35/DSDT                  | Bin 7801 -> 7801 bytes
>  tests/data/acpi/q35/DSDT.acpihmat         | Bin 9126 -> 9126 bytes
>  tests/data/acpi/q35/DSDT.bridge           | Bin 7819 -> 7819 bytes
>  tests/data/acpi/q35/DSDT.cphp             | Bin 8265 -> 8265 bytes
>  tests/data/acpi/q35/DSDT.dimmpxm          | Bin 9455 -> 9455 bytes
>  tests/data/acpi/q35/DSDT.ipmibt           | Bin 7876 -> 7876 bytes
>  tests/data/acpi/q35/DSDT.memhp            | Bin 9160 -> 9160 bytes
>  tests/data/acpi/q35/DSDT.mmio64           | Bin 8932 -> 8932 bytes
>  tests/data/acpi/q35/DSDT.numamem          | Bin 7807 -> 7807 bytes
>  tests/qtest/cxl-test.c                    |  93 ++++++
>  tests/qtest/meson.build                   |   4 +
>  65 files changed, 2777 insertions(+), 30 deletions(-)
>  create mode 100644 hw/acpi/cxl.c
>  create mode 100644 hw/cxl/Kconfig
>  create mode 100644 hw/cxl/cxl-component-utils.c
>  create mode 100644 hw/cxl/cxl-device-utils.c
>  create mode 100644 hw/cxl/cxl-mailbox-utils.c
>  create mode 100644 hw/cxl/meson.build
>  create mode 100644 hw/mem/cxl_type3.c
>  create mode 100644 hw/pci-bridge/cxl_root_port.c
>  create mode 100644 include/hw/acpi/cxl.h
>  create mode 100644 include/hw/cxl/cxl.h
>  create mode 100644 include/hw/cxl/cxl_component.h
>  create mode 100644 include/hw/cxl/cxl_device.h
>  create mode 100644 include/hw/cxl/cxl_pci.h
>  create mode 100644 tests/data/acpi/pc/CEDT
>  create mode 100644 tests/data/acpi/q35/CEDT
>  create mode 100644 tests/qtest/cxl-test.c
> 



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 00/32] CXL 2.0 Support
  2021-01-08 18:44 ` [RFC PATCH v2 00/32] CXL 2.0 Support Jonathan Cameron
@ 2021-01-08 18:51   ` Ben Widawsky
  0 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-08 18:51 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Dan Williams

On 21-01-08 18:44:04, Jonathan Cameron wrote:
> On Tue, 5 Jan 2021 08:52:51 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > Fixes since v1 [1]:
> >  * Defer introducing some commands/registers not yet used (Ben)
> >  * Add stubbed device_reg_init_common() (Ben)
> >  * Improve assertions in DVSEC creation (Jonathan)
> >  * Use 'n' for HDM register offsets (Jonathan)
> >  * Correct revision ID for extensions (Jonathan)
> >  * Minor cleanups and clarifications (Jonathan)
> >  * Remove error codes not yet used (Jonathan)
> >  * Fix interrupt enable bit width (Jonathan)
> >  * Add comment for weird register size (Jonathan)
> >  * Break out register alignment checks (Jonathan)
> >  * Use the reg alignment helper (Jonathan)
> >  * Rename error codes to match spec
> >  * Fix cap count mid series (Jonathan)
> > 
> > New since v1 [1]:
> >  * Entirely reworked framework for firmware handling
> >  * Implemented more device commands
> >  * CEL support
> > 
> > (There are some new patches that I just named 'v2' for simplicity's sake)
> > 
> > Introduce emulation of Compute Express Link 2.0
> > (https://www.computeexpresslink.org/).
> 
> Hi Ben,
> 
> In interests of avoid duplication, I thought I'd give a quick update of where
> I am. I have what is here working on arm64 (enabling is fairly simple).
> I'll have a bit more feedback next week and hopefully have time for an indepth review.

Thanks for this. Dan recommended I send out an email with plans for what's
needed, what we intend to work on and so on. I'll try to get that out next week
also.

> 
> I did run into some issues around alignment for the persistent memory that
> (I think) are down to the fact the memory-backend isn't quite tied up to the
> device.

I did have plans to change this to support interleaving.

> 
> Flows wise, I initially started hacking in NFIT building changes in qemu but I'm not
> yet sure how the hotplug flows are supposed to work and who is responsible for
> setting up a hotplugged device (OS or firmware).

It won't be NFIT :-). I'd redirect this question to the linux-cxl mailing list,
as it's a good one. The spec is somewhat handwavy around this.

> 
> Plans wise, I was thinking I could look at DOE emulation as seems that we'll
> be needing that fairly soon to get any useful memory usecases up and running.
> Let me know if you already have that underway.  
> 
> Jonathan
> 

Chris in the Cc had some plans to enable DOE for their purposes. It'd be good to
sync with him. AFAIK though, nobody has plans to enable DOE in Linux, which
might be a good target next.

> > 
> > The emulation has been critical to get the Linux enabling started [2], it would
> > be an ideal place to land regression tests for different topology handling, and
> > there may be applications for this emulation as a way for a guest to manipulate
> > its address space relative to different performance memories.
> > 
> > Three of the five CXL component types are emulated with some level of functionality:
> > host bridge, root port, and memory device. Upstream ports and downstream ports
> > aren't implemented (the two components needed to make up a switch).
> > 
> > CXL 2.0 is built on top of PCIe (see spec for details). As a result, much of the
> > implementation utilizes existing PCI paradigms. To implement the host bridge,
> > I've chosen to use PXB (PCI Expander Bridge). It seemed to be the most natural
> > fit even though it doesn't directly map to how hardware will work. For
> > persistent capacity of the memory device, I utilized the memory subsystem
> > (hw/mem).
> > 
> > We have 3 reasons why this work is valuable:
> > 1. OS driver development and testing
> > 2. OS driver regression testing
> > 3. Possible guest support for HDMs
> > 
> > As mentioned above there are three benefits to carrying this enabling in
> > upstream QEMU:
> > 
> > 1. Linux driver feature development benefits from emulation both due to
> > a lack of initial hardware availability, but also, as is seen with
> > NVDIMM/PMEM emulation, there is value in being able to share
> > topologies with system-software developers even after hardware is
> > available.
> > 
> > 2. The Linux kernel's unit test suite for NVDIMM/PMEM ended up injecting fake
> > resources via custom modules (nfit_test). In retrospect a QEMU emulation of
> > nfit_test capabilities would have made the test environment more portable, and
> > allowed for easier community contributions of example configurations.
> > 
> > 3. This is still being fleshed out, but in short it provides a standardized
> > mechanism for the guest to provide feedback to the host about size and placement
> > needs of the memory. After the host gives the guest a physical window mapping to
> > the CXL device, the emulated HDM decoders allow the guest a way to tell the host
> > how much it wants and where. There are likely simpler ways to do this, but
> > they'd require inventing a new interface and you'd need to have diverging driver
> > code in the guest programming of the HDM decoder vs. the host. Since we've
> > already done this work, why not use it?
> > 
> > There is quite a long list of work to do for full spec compliance, but I don't
> > believe that any of it precludes merging. Off the top of my head:
> > - Main host bridge support (WIP)
> > - Interleaving
> > - Better Tests
> > - Huge swaths of firmware functionality
> > - Hot plug support
> > - Emulating volatile capacity
> > 
> > The flow of the patches in general is to define all the data structures and
> > registers associated with the various components in a top down manner. Host
> > bridge, component, ports, devices. Then, the actual implementation is done in
> > the same order.
> > 
> > The summary is:
> > 1-8: Put infrastructure in place for emulation of the components.
> > 9-12: Implement device mailboxes
> > 13-15: Create the concept of a CXL bus and plumb into PXB
> > 16-22: Implement host bridges
> > 23: Implement a root port
> > 24: Implement a memory device
> > 25: Implement HDM decoders
> > 26-30: ACPI bits
> > 31: Start working on enabling the main host bridge
> > 32: Basic test case
> > 
> > [1]: https://lore.kernel.org/qemu-devel/20201111054724.794888-1-ben.widawsky@intel.com/
> > [2]: https://lore.kernel.org/linux-cxl/20201209002418.1976362-1-ben.widawsky@intel.com/
> > 
> > Ben Widawsky (31):
> >   hw/pci/cxl: Add a CXL component type (interface)
> >   hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)
> >   hw/cxl/device: Introduce a CXL device (8.2.8)
> >   hw/cxl/device: Implement the CAP array (8.2.8.1-2)
> >   hw/cxl/device: Add device status (8.2.8.3)
> >   hw/cxl/device: Implement basic mailbox (8.2.8.4)
> >   hw/cxl/device: Add memory devices (8.2.8.5)
> >   hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1)
> >   hw/cxl/device: Placeholder for firmware commands (8.2.9.2)
> >   hw/cxl/device: Timestamp implementation (8.2.9.3)
> >   hw/cxl/device: Add log commands (8.2.9.4) + CEL
> >   hw/pxb: Use a type for realizing expanders
> >   hw/pci/cxl: Create a CXL bus type
> >   hw/pxb: Allow creation of a CXL PXB (host bridge)
> >   qtest: allow DSDT acpi table changes
> >   acpi/pci: Consolidate host bridge setup
> >   tests/acpi: remove stale allowed tables
> >   hw/pci: Plumb _UID through host bridges
> >   hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)
> >   acpi/pxb/cxl: Reserve host bridge MMIO
> >   hw/pxb/cxl: Add "windows" for host bridges
> >   hw/cxl/rp: Add a root port
> >   hw/cxl/device: Add a memory device (8.2.8.5)
> >   hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12)
> >   acpi/cxl: Add _OSC implementation (9.14.2)
> >   tests/acpi: allow CEDT table addition
> >   acpi/cxl: Create the CEDT (9.14.1)
> >   Temp: acpi/cxl: Add ACPI0017 (CEDT awareness)
> >   tests/acpi: Add new CEDT files
> >   WIP: i386/cxl: Initialize a host bridge
> >   qtest/cxl: Add very basic sanity tests
> > 
> > Jonathan Cameron (1):
> >   Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy.
> > 
> >  MAINTAINERS                               |   6 +
> >  hw/Kconfig                                |   1 +
> >  hw/acpi/Kconfig                           |   5 +
> >  hw/acpi/cxl.c                             | 173 ++++++++++
> >  hw/acpi/meson.build                       |   1 +
> >  hw/arm/virt.c                             |   1 +
> >  hw/core/machine.c                         |  26 ++
> >  hw/core/numa.c                            |   3 +
> >  hw/cxl/Kconfig                            |   3 +
> >  hw/cxl/cxl-component-utils.c              | 193 +++++++++++
> >  hw/cxl/cxl-device-utils.c                 | 278 ++++++++++++++++
> >  hw/cxl/cxl-mailbox-utils.c                | 386 ++++++++++++++++++++++
> >  hw/cxl/meson.build                        |   5 +
> >  hw/i386/acpi-build.c                      |  87 ++++-
> >  hw/i386/microvm.c                         |   1 +
> >  hw/i386/pc.c                              |   2 +
> >  hw/mem/Kconfig                            |   5 +
> >  hw/mem/cxl_type3.c                        | 334 +++++++++++++++++++
> >  hw/mem/meson.build                        |   1 +
> >  hw/meson.build                            |   1 +
> >  hw/pci-bridge/Kconfig                     |   5 +
> >  hw/pci-bridge/cxl_root_port.c             | 231 +++++++++++++
> >  hw/pci-bridge/meson.build                 |   1 +
> >  hw/pci-bridge/pci_expander_bridge.c       | 209 +++++++++++-
> >  hw/pci-bridge/pcie_root_port.c            |   6 +-
> >  hw/pci/pci.c                              |  32 +-
> >  hw/pci/pcie.c                             |  30 ++
> >  hw/ppc/spapr.c                            |   2 +
> >  include/hw/acpi/cxl.h                     |  27 ++
> >  include/hw/boards.h                       |   2 +
> >  include/hw/cxl/cxl.h                      |  30 ++
> >  include/hw/cxl/cxl_component.h            | 187 +++++++++++
> >  include/hw/cxl/cxl_device.h               | 210 ++++++++++++
> >  include/hw/cxl/cxl_pci.h                  | 160 +++++++++
> >  include/hw/pci/pci.h                      |  15 +
> >  include/hw/pci/pci_bridge.h               |  25 ++
> >  include/hw/pci/pci_bus.h                  |   8 +
> >  include/hw/pci/pci_ids.h                  |   1 +
> >  include/standard-headers/linux/pci_regs.h |   1 +
> >  monitor/hmp-cmds.c                        |  15 +
> >  qapi/machine.json                         |   1 +
> >  tests/data/acpi/pc/CEDT                   | Bin 0 -> 36 bytes
> >  tests/data/acpi/pc/DSDT                   | Bin 5065 -> 5065 bytes
> >  tests/data/acpi/pc/DSDT.acpihmat          | Bin 6390 -> 6390 bytes
> >  tests/data/acpi/pc/DSDT.bridge            | Bin 6924 -> 6924 bytes
> >  tests/data/acpi/pc/DSDT.cphp              | Bin 5529 -> 5529 bytes
> >  tests/data/acpi/pc/DSDT.dimmpxm           | Bin 6719 -> 6719 bytes
> >  tests/data/acpi/pc/DSDT.hpbridge          | Bin 5026 -> 5026 bytes
> >  tests/data/acpi/pc/DSDT.hpbrroot          | Bin 3084 -> 3084 bytes
> >  tests/data/acpi/pc/DSDT.ipmikcs           | Bin 5137 -> 5137 bytes
> >  tests/data/acpi/pc/DSDT.memhp             | Bin 6424 -> 6424 bytes
> >  tests/data/acpi/pc/DSDT.numamem           | Bin 5071 -> 5071 bytes
> >  tests/data/acpi/pc/DSDT.roothp            | Bin 5261 -> 5261 bytes
> >  tests/data/acpi/q35/CEDT                  | Bin 0 -> 36 bytes
> >  tests/data/acpi/q35/DSDT                  | Bin 7801 -> 7801 bytes
> >  tests/data/acpi/q35/DSDT.acpihmat         | Bin 9126 -> 9126 bytes
> >  tests/data/acpi/q35/DSDT.bridge           | Bin 7819 -> 7819 bytes
> >  tests/data/acpi/q35/DSDT.cphp             | Bin 8265 -> 8265 bytes
> >  tests/data/acpi/q35/DSDT.dimmpxm          | Bin 9455 -> 9455 bytes
> >  tests/data/acpi/q35/DSDT.ipmibt           | Bin 7876 -> 7876 bytes
> >  tests/data/acpi/q35/DSDT.memhp            | Bin 9160 -> 9160 bytes
> >  tests/data/acpi/q35/DSDT.mmio64           | Bin 8932 -> 8932 bytes
> >  tests/data/acpi/q35/DSDT.numamem          | Bin 7807 -> 7807 bytes
> >  tests/qtest/cxl-test.c                    |  93 ++++++
> >  tests/qtest/meson.build                   |   4 +
> >  65 files changed, 2777 insertions(+), 30 deletions(-)
> >  create mode 100644 hw/acpi/cxl.c
> >  create mode 100644 hw/cxl/Kconfig
> >  create mode 100644 hw/cxl/cxl-component-utils.c
> >  create mode 100644 hw/cxl/cxl-device-utils.c
> >  create mode 100644 hw/cxl/cxl-mailbox-utils.c
> >  create mode 100644 hw/cxl/meson.build
> >  create mode 100644 hw/mem/cxl_type3.c
> >  create mode 100644 hw/pci-bridge/cxl_root_port.c
> >  create mode 100644 include/hw/acpi/cxl.h
> >  create mode 100644 include/hw/cxl/cxl.h
> >  create mode 100644 include/hw/cxl/cxl_component.h
> >  create mode 100644 include/hw/cxl/cxl_device.h
> >  create mode 100644 include/hw/cxl/cxl_pci.h
> >  create mode 100644 tests/data/acpi/pc/CEDT
> >  create mode 100644 tests/data/acpi/q35/CEDT
> >  create mode 100644 tests/qtest/cxl-test.c
> > 
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-05 16:53 ` [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5) Ben Widawsky
@ 2021-01-27 21:03   ` Igor Mammedov
  2021-01-27 21:11     ` Ben Widawsky
  2021-01-27 21:26     ` Ben Widawsky
  0 siblings, 2 replies; 57+ messages in thread
From: Igor Mammedov @ 2021-01-27 21:03 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Jonathan Cameron, Prashant V Agarwal, Dan Williams

On Tue,  5 Jan 2021 08:53:15 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> A CXL memory device (AKA Type 3) is a CXL component that contains some
> combination of volatile and persistent memory. It also implements the
> previously defined mailbox interface as well as the memory device
> firmware interface.
> 
> The following example will create a 256M device in a 512M window:
> 
> -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"

I'd expect whole backend used by frontend, so one would not need "size" property
on frontend (like we do with memory devices).
So question is why it partially uses memdev?


> Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> ---
>  hw/core/numa.c             |   3 +
>  hw/cxl/cxl-mailbox-utils.c |  41 ++++++
>  hw/i386/pc.c               |   1 +
>  hw/mem/Kconfig             |   5 +
>  hw/mem/cxl_type3.c         | 262 +++++++++++++++++++++++++++++++++++++
>  hw/mem/meson.build         |   1 +
>  hw/pci/pcie.c              |  30 +++++
>  include/hw/cxl/cxl.h       |   2 +
>  include/hw/cxl/cxl_pci.h   |  22 ++++
>  include/hw/pci/pci_ids.h   |   1 +
>  monitor/hmp-cmds.c         |  15 +++
>  qapi/machine.json          |   1 +
>  12 files changed, 384 insertions(+)
>  create mode 100644 hw/mem/cxl_type3.c
> 
> diff --git a/hw/core/numa.c b/hw/core/numa.c
> index 68cee65f61..cd7df371e6 100644
> --- a/hw/core/numa.c
> +++ b/hw/core/numa.c
> @@ -770,6 +770,9 @@ static void numa_stat_memory_devices(NumaNodeMem node_mem[])
>                  node_mem[pcdimm_info->node].node_plugged_mem +=
>                      pcdimm_info->size;
>                  break;
> +            case MEMORY_DEVICE_INFO_KIND_CXL:
> +                /* FINISHME */
> +                break;
>              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
>                  vpi = value->u.virtio_pmem.data;
>                  /* TODO: once we support numa, assign to right node */
> diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> index f68ec5b5b9..eeb10b8943 100644
> --- a/hw/cxl/cxl-mailbox-utils.c
> +++ b/hw/cxl/cxl-mailbox-utils.c
> @@ -49,6 +49,8 @@ enum {
>      LOGS        = 0x04,
>          #define GET_SUPPORTED 0x0
>          #define GET_LOG       0x1
> +    IDENTIFY    = 0x40,
> +        #define MEMORY_DEVICE 0x0
>  };
>  
>  /* 8.2.8.4.5.1 Command Return Codes */
> @@ -127,6 +129,7 @@ declare_mailbox_handler(TIMESTAMP_GET);
>  declare_mailbox_handler(TIMESTAMP_SET);
>  declare_mailbox_handler(LOGS_GET_SUPPORTED);
>  declare_mailbox_handler(LOGS_GET_LOG);
> +declare_mailbox_handler(IDENTIFY_MEMORY_DEVICE);
>  
>  #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
>  #define IMMEDIATE_POLICY_CHANGE (1 << 3)
> @@ -144,6 +147,7 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
>      CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
>      CXL_CMD(LOGS, GET_SUPPORTED, 0, 0),
>      CXL_CMD(LOGS, GET_LOG, 0x18, 0),
> +    CXL_CMD(IDENTIFY, MEMORY_DEVICE, 0, 0),
>  };
>  
>  #undef CXL_CMD
> @@ -262,6 +266,43 @@ define_mailbox_handler(LOGS_GET_LOG)
>      return CXL_MBOX_SUCCESS;
>  }
>  
> +/* 8.2.9.5.1.1 */
> +define_mailbox_handler(IDENTIFY_MEMORY_DEVICE)
> +{
> +    struct {
> +        char fw_revision[0x10];
> +        uint64_t total_capacity;
> +        uint64_t volatile_capacity;
> +        uint64_t persistent_capacity;
> +        uint64_t partition_align;
> +        uint16_t info_event_log_size;
> +        uint16_t warning_event_log_size;
> +        uint16_t failure_event_log_size;
> +        uint16_t fatal_event_log_size;
> +        uint32_t lsa_size;
> +        uint8_t poison_list_max_mer[3];
> +        uint16_t inject_poison_limit;
> +        uint8_t poison_caps;
> +        uint8_t qos_telemetry_caps;
> +    } __attribute__((packed)) *id;
> +    _Static_assert(sizeof(*id) == 0x43, "Bad identify size");
> +
> +    if (memory_region_size(cxl_dstate->pmem) < (256 << 20)) {
> +        return CXL_MBOX_INTERNAL_ERROR;
> +    }
> +
> +    id = (void *)cmd->payload;
> +    memset(id, 0, sizeof(*id));
> +
> +    /* PMEM only */
> +    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
> +    id->total_capacity = memory_region_size(cxl_dstate->pmem);
> +    id->persistent_capacity = memory_region_size(cxl_dstate->pmem);
> +
> +    *len = sizeof(*id);
> +    return CXL_MBOX_SUCCESS;
> +}
> +
>  void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
>  {
>      uint16_t ret = CXL_MBOX_SUCCESS;
> diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> index 5458f61d10..5d41809b37 100644
> --- a/hw/i386/pc.c
> +++ b/hw/i386/pc.c
> @@ -79,6 +79,7 @@
>  #include "acpi-build.h"
>  #include "hw/mem/pc-dimm.h"
>  #include "hw/mem/nvdimm.h"
> +#include "hw/cxl/cxl.h"
>  #include "qapi/error.h"
>  #include "qapi/qapi-visit-common.h"
>  #include "qapi/visitor.h"
> diff --git a/hw/mem/Kconfig b/hw/mem/Kconfig
> index a0ef2cf648..7d9d1ced3e 100644
> --- a/hw/mem/Kconfig
> +++ b/hw/mem/Kconfig
> @@ -10,3 +10,8 @@ config NVDIMM
>      default y
>      depends on (PC || PSERIES || ARM_VIRT)
>      select MEM_DEVICE
> +
> +config CXL_MEM_DEVICE
> +    bool
> +    default y if CXL
> +    select MEM_DEVICE
> diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> new file mode 100644
> index 0000000000..3985bb8d0b
> --- /dev/null
> +++ b/hw/mem/cxl_type3.c
> @@ -0,0 +1,262 @@
> +#include "qemu/osdep.h"
> +#include "qemu/units.h"
> +#include "qemu/error-report.h"
> +#include "hw/mem/memory-device.h"
> +#include "hw/mem/pc-dimm.h"
> +#include "hw/pci/pci.h"
> +#include "hw/qdev-properties.h"
> +#include "qapi/error.h"
> +#include "qemu/log.h"
> +#include "qemu/module.h"
> +#include "qemu/range.h"
> +#include "qemu/rcu.h"
> +#include "sysemu/hostmem.h"
> +#include "hw/cxl/cxl.h"
> +
> +typedef struct cxl_type3_dev {
> +    /* Private */
> +    PCIDevice parent_obj;
> +
> +    /* Properties */
> +    uint64_t size;
> +    HostMemoryBackend *hostmem;
> +
> +    /* State */
> +    CXLComponentState cxl_cstate;
> +    CXLDeviceState cxl_dstate;
> +} CXLType3Dev;
> +
> +#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
> +
> +static void build_dvsecs(CXLType3Dev *ct3d)
> +{
> +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> +    uint8_t *dvsec;
> +
> +    dvsec = (uint8_t *)&(struct dvsec_device){
> +        .cap = 0x1e,
> +        .ctrl = 0x6,
> +        .status2 = 0x2,
> +        .range1_size_hi = 0,
> +        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
> +        .range1_base_hi = 0,
> +        .range1_base_lo = 0,
> +    };
> +    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
> +                               PCIE_CXL_DEVICE_DVSEC,
> +                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
> +
> +    dvsec = (uint8_t *)&(struct dvsec_register_locator){
> +        .rsvd         = 0,
> +        .reg0_base_lo = RBI_COMPONENT_REG | COMPONENT_REG_BAR_IDX,
> +        .reg0_base_hi = 0,
> +        .reg1_base_lo = RBI_CXL_DEVICE_REG | DEVICE_REG_BAR_IDX,
> +        .reg1_base_hi = 0,
> +    };
> +    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
> +                               REG_LOC_DVSEC_REVID, dvsec);
> +}
> +
> +static void ct3_instance_init(Object *obj)
> +{
> +    /* MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); */
> +}
> +
> +static void ct3_finalize(Object *obj)
> +{
> +    CXLType3Dev *ct3d = CT3(obj);
> +
> +    g_free(ct3d->cxl_dstate.pmem);
> +}
> +
> +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> +{
> +    MemoryRegionSection mrs;
> +    MemoryRegion *mr;
> +    uint64_t offset = 0;
> +    size_t remaining_size;
> +
> +    if (!ct3d->hostmem) {
> +        error_setg(errp, "memdev property must be set");
> +        return;
> +    }
> +
> +    /* FIXME: need to check mr is the host bridge's MR */
> +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> +
> +    /* Create our new subregion */
> +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> +
> +    /* Find the first free space in the window */
> +    WITH_RCU_READ_LOCK_GUARD()
> +    {
> +        mrs = memory_region_find(mr, offset, 1);
> +        while (mrs.mr && mrs.mr != mr) {
> +            offset += memory_region_size(mrs.mr);
> +            mrs = memory_region_find(mr, offset, 1);
> +        }
> +    }
> +
> +    remaining_size = memory_region_size(mr) - offset;
> +    if (remaining_size < ct3d->size) {
> +        g_free(ct3d->cxl_dstate.pmem);
> +        error_setg(errp,
> +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> +                   remaining_size, ct3d->size);
> +    }
> +
> +    /* Register our subregion as non-volatile */
> +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> +                           "cxl_type3-memory", ct3d->size, errp);
this allocates ct3d->size of anon RAM, was this an intention?
If yes, can you clarify why extra RAM is used instead of using what
backend provides?

> +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);

> +#ifdef SET_PMEM_PADDR
> +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> +#endif
What this hunk is supposed to do, why it's ifdef-ed?


> +}
> +
> +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> +                                              Error **errp)
> +{
> +    CXLType3Dev *ct3d = CT3(md);
> +
> +    if (!ct3d->cxl_dstate.pmem) {
> +        cxl_setup_memory(ct3d, errp);
> +    }
> +
> +    return ct3d->cxl_dstate.pmem;
> +}
> +
> +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> +{
> +    CXLType3Dev *ct3d = CT3(pci_dev);
> +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> +    ComponentRegisters *regs = &cxl_cstate->crb;
> +    MemoryRegion *mr = &regs->component_registers;
> +    uint8_t *pci_conf = pci_dev->config;
> +
> +    if (!ct3d->cxl_dstate.pmem) {
> +        cxl_setup_memory(ct3d, errp);
> +    }
> +
> +    pci_config_set_prog_interface(pci_conf, 0x10);
> +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> +
> +    pcie_endpoint_cap_init(pci_dev, 0x80);
> +    cxl_cstate->dvsec_offset = 0x100;
> +
> +    ct3d->cxl_cstate.pdev = pci_dev;
> +    build_dvsecs(ct3d);
> +
> +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> +                                      TYPE_CXL_TYPE3_DEV);
> +
> +    pci_register_bar(
> +        pci_dev, COMPONENT_REG_BAR_IDX,
> +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> +
> +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> +                     &ct3d->cxl_dstate.device_registers);
> +}
> +
> +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> +{
> +    CXLType3Dev *ct3d = CT3(md);
> +
> +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> +}
> +
> +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> +{
> +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> +}
> +
> +static void ct3d_reset(DeviceState *dev)
> +{
> +    CXLType3Dev *ct3d = CT3(dev);
> +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> +
> +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> +}
> +
> +static Property ct3_props[] = {
> +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> +                     HostMemoryBackend *),
> +    DEFINE_PROP_END_OF_LIST(),
> +};
> +
> +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> +                                        MemoryDeviceInfo *info)
> +{
> +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> +    const DeviceState *dev = DEVICE(md);
> +    CXLType3Dev *ct3d = CT3(md);
> +
> +    if (dev->id) {
> +        di->has_id = true;
> +        di->id = g_strdup(dev->id);
> +    }
> +    di->hotplugged = dev->hotplugged;
> +    di->hotpluggable = dc->hotpluggable;
> +    di->addr = cxl_md_get_addr(md);
> +    di->slot = 0;
> +    di->node = 0;
> +    di->size = memory_device_get_region_size(md, NULL);
> +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> +
> +
> +    info->u.cxl.data = di;
> +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> +}
> +
> +static void ct3_class_init(ObjectClass *oc, void *data)
> +{
> +    DeviceClass *dc = DEVICE_CLASS(oc);
> +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> +
> +    pc->realize = ct3_realize;
> +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> +    pc->device_id = 0xd93; /* LVF for now */
> +    pc->revision = 1;
> +
> +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> +    dc->desc = "CXL PMEM Device (Type 3)";
> +    dc->reset = ct3d_reset;
> +    device_class_set_props(dc, ct3_props);
> +
> +    mdc->get_memory_region = cxl_md_get_memory_region;
> +    mdc->get_addr = cxl_md_get_addr;
> +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> +    mdc->get_plugged_size = memory_device_get_region_size;
> +    mdc->set_addr = cxl_md_set_addr;
> +}
> +
> +static const TypeInfo ct3d_info = {
> +    .name = TYPE_CXL_TYPE3_DEV,
> +    .parent = TYPE_PCI_DEVICE,
> +    .class_init = ct3_class_init,
> +    .instance_size = sizeof(CXLType3Dev),
> +    .instance_init = ct3_instance_init,
> +    .instance_finalize = ct3_finalize,
> +    .interfaces = (InterfaceInfo[]) {
> +        { TYPE_MEMORY_DEVICE },
> +        { INTERFACE_CXL_DEVICE },
> +        { INTERFACE_PCIE_DEVICE },
> +        {}
> +    },
> +};
> +
> +static void ct3d_registers(void)
> +{
> +    type_register_static(&ct3d_info);
> +}
> +
> +type_init(ct3d_registers);
> diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> index 0d22f2b572..d13c3ed117 100644
> --- a/hw/mem/meson.build
> +++ b/hw/mem/meson.build
> @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
>  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
>  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
>  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
>  
>  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> index d4010cf8f3..1ecf6f6a55 100644
> --- a/hw/pci/pcie.c
> +++ b/hw/pci/pcie.c
> @@ -20,6 +20,7 @@
>  
>  #include "qemu/osdep.h"
>  #include "qapi/error.h"
> +#include "hw/mem/memory-device.h"
>  #include "hw/pci/pci_bridge.h"
>  #include "hw/pci/pcie.h"
>  #include "hw/pci/msix.h"
> @@ -27,6 +28,8 @@
>  #include "hw/pci/pci_bus.h"
>  #include "hw/pci/pcie_regs.h"
>  #include "hw/pci/pcie_port.h"
> +#include "hw/cxl/cxl.h"
> +#include "hw/boards.h"
>  #include "qemu/range.h"
>  
>  //#define DEBUG_PCIE
> @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
>      }
>  
>      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> +
> +#ifdef CXL_MEM_DEVICE
> +    /*
> +     * FIXME:
> +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> +     *    HotplugHandler *hotplug_ctrl;
> +     *   Error *local_err = NULL;
> +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> +     *  if (hotplug_ctrl) {
> +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> +     *      if (local_err) {
> +     *          error_propagate(errp, local_err);
> +     *          return;
> +     *      }
> +     *  }
> +     */
> +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> +                               NULL, errp);
> +    }

why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?

> +#endif
>  }
>  
>  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
>          pcie_cap_slot_event(hotplug_pdev,
>                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
>      }
> +
> +#ifdef CXL_MEM_DEVICE
> +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> +#endif
>  }
>  
>  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> index b1e5f4a8fa..809ed7de60 100644
> --- a/include/hw/cxl/cxl.h
> +++ b/include/hw/cxl/cxl.h
> @@ -17,6 +17,8 @@
>  #define COMPONENT_REG_BAR_IDX 0
>  #define DEVICE_REG_BAR_IDX 2
>  
> +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> +
>  #define CXL_HOST_BASE 0xD0000000
>  #define CXL_WINDOW_MAX 10
>  
> diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> index a53c2e5ae7..9ec28c9feb 100644
> --- a/include/hw/cxl/cxl_pci.h
> +++ b/include/hw/cxl/cxl_pci.h
> @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
>   * CXL 2.0 Downstream Port: 3, 4, 7, 8
>   */
>  
> +/* CXL 2.0 - 8.1.3 (ID 0001) */
> +struct dvsec_device {
> +    struct dvsec_header hdr;
> +    uint16_t cap;
> +    uint16_t ctrl;
> +    uint16_t status;
> +    uint16_t ctrl2;
> +    uint16_t status2;
> +    uint16_t lock;
> +    uint16_t cap2;
> +    uint32_t range1_size_hi;
> +    uint32_t range1_size_lo;
> +    uint32_t range1_base_hi;
> +    uint32_t range1_base_lo;
> +    uint32_t range2_size_hi;
> +    uint32_t range2_size_lo;
> +    uint32_t range2_base_hi;
> +    uint32_t range2_base_lo;
> +};
> +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> +               "dvsec device size incorrect");
> +
>  /* CXL 2.0 - 8.1.5 (ID 0003) */
>  struct extensions_dvsec_port {
>      struct dvsec_header hdr;
> diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> index 11f8ab7149..76bf3ed590 100644
> --- a/include/hw/pci/pci_ids.h
> +++ b/include/hw/pci/pci_ids.h
> @@ -53,6 +53,7 @@
>  #define PCI_BASE_CLASS_MEMORY            0x05
>  #define PCI_CLASS_MEMORY_RAM             0x0500
>  #define PCI_CLASS_MEMORY_FLASH           0x0501
> +#define PCI_CLASS_MEMORY_CXL             0x0502
>  #define PCI_CLASS_MEMORY_OTHER           0x0580
>  
>  #define PCI_BASE_CLASS_BRIDGE            0x06
> diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> index 0dd594f92b..0f67bc61ce 100644
> --- a/monitor/hmp-cmds.c
> +++ b/monitor/hmp-cmds.c
> @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
>                  monitor_printf(mon, "  hotpluggable: %s\n",
>                                 di->hotpluggable ? "true" : "false");
>                  break;
> +            case MEMORY_DEVICE_INFO_KIND_CXL:
> +                di = value->u.cxl.data;
> +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> +                               MemoryDeviceInfoKind_str(value->type),
> +                               di->id ? di->id : "");
> +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> +                monitor_printf(mon, "  hotplugged: %s\n",
> +                               di->hotplugged ? "true" : "false");
> +                monitor_printf(mon, "  hotpluggable: %s\n",
> +                               di->hotpluggable ? "true" : "false");
> +                break;
>              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
>                  vpi = value->u.virtio_pmem.data;
>                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> diff --git a/qapi/machine.json b/qapi/machine.json
> index 330189efe3..aa96d662bd 100644
> --- a/qapi/machine.json
> +++ b/qapi/machine.json
> @@ -1394,6 +1394,7 @@
>  { 'union': 'MemoryDeviceInfo',
>    'data': { 'dimm': 'PCDIMMDeviceInfo',
>              'nvdimm': 'PCDIMMDeviceInfo',
> +            'cxl': 'PCDIMMDeviceInfo',
>              'virtio-pmem': 'VirtioPMEMDeviceInfo',
>              'virtio-mem': 'VirtioMEMDeviceInfo'
>            }



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-27 21:03   ` Igor Mammedov
@ 2021-01-27 21:11     ` Ben Widawsky
  2021-01-27 21:21       ` Igor Mammedov
  2021-01-27 21:26     ` Ben Widawsky
  1 sibling, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-27 21:11 UTC (permalink / raw)
  To: Igor Mammedov
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Jonathan Cameron, Prashant V Agarwal, Dan Williams

On 21-01-27 22:03:12, Igor Mammedov wrote:
> On Tue,  5 Jan 2021 08:53:15 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > combination of volatile and persistent memory. It also implements the
> > previously defined mailbox interface as well as the memory device
> > firmware interface.
> > 
> > The following example will create a 256M device in a 512M window:
> > 
> > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"
> 
> I'd expect whole backend used by frontend, so one would not need "size" property
> on frontend (like we do with memory devices).
> So question is why it partially uses memdev?

A CXL memory device may participate in an interleave set. In such a case, it
would be < the total size of the memory window.

This isn't implemented in the code yet, but it is planned.

> 
> 
> > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> > ---
> >  hw/core/numa.c             |   3 +
> >  hw/cxl/cxl-mailbox-utils.c |  41 ++++++
> >  hw/i386/pc.c               |   1 +
> >  hw/mem/Kconfig             |   5 +
> >  hw/mem/cxl_type3.c         | 262 +++++++++++++++++++++++++++++++++++++
> >  hw/mem/meson.build         |   1 +
> >  hw/pci/pcie.c              |  30 +++++
> >  include/hw/cxl/cxl.h       |   2 +
> >  include/hw/cxl/cxl_pci.h   |  22 ++++
> >  include/hw/pci/pci_ids.h   |   1 +
> >  monitor/hmp-cmds.c         |  15 +++
> >  qapi/machine.json          |   1 +
> >  12 files changed, 384 insertions(+)
> >  create mode 100644 hw/mem/cxl_type3.c
> > 
> > diff --git a/hw/core/numa.c b/hw/core/numa.c
> > index 68cee65f61..cd7df371e6 100644
> > --- a/hw/core/numa.c
> > +++ b/hw/core/numa.c
> > @@ -770,6 +770,9 @@ static void numa_stat_memory_devices(NumaNodeMem node_mem[])
> >                  node_mem[pcdimm_info->node].node_plugged_mem +=
> >                      pcdimm_info->size;
> >                  break;
> > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > +                /* FINISHME */
> > +                break;
> >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> >                  vpi = value->u.virtio_pmem.data;
> >                  /* TODO: once we support numa, assign to right node */
> > diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> > index f68ec5b5b9..eeb10b8943 100644
> > --- a/hw/cxl/cxl-mailbox-utils.c
> > +++ b/hw/cxl/cxl-mailbox-utils.c
> > @@ -49,6 +49,8 @@ enum {
> >      LOGS        = 0x04,
> >          #define GET_SUPPORTED 0x0
> >          #define GET_LOG       0x1
> > +    IDENTIFY    = 0x40,
> > +        #define MEMORY_DEVICE 0x0
> >  };
> >  
> >  /* 8.2.8.4.5.1 Command Return Codes */
> > @@ -127,6 +129,7 @@ declare_mailbox_handler(TIMESTAMP_GET);
> >  declare_mailbox_handler(TIMESTAMP_SET);
> >  declare_mailbox_handler(LOGS_GET_SUPPORTED);
> >  declare_mailbox_handler(LOGS_GET_LOG);
> > +declare_mailbox_handler(IDENTIFY_MEMORY_DEVICE);
> >  
> >  #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
> >  #define IMMEDIATE_POLICY_CHANGE (1 << 3)
> > @@ -144,6 +147,7 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
> >      CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
> >      CXL_CMD(LOGS, GET_SUPPORTED, 0, 0),
> >      CXL_CMD(LOGS, GET_LOG, 0x18, 0),
> > +    CXL_CMD(IDENTIFY, MEMORY_DEVICE, 0, 0),
> >  };
> >  
> >  #undef CXL_CMD
> > @@ -262,6 +266,43 @@ define_mailbox_handler(LOGS_GET_LOG)
> >      return CXL_MBOX_SUCCESS;
> >  }
> >  
> > +/* 8.2.9.5.1.1 */
> > +define_mailbox_handler(IDENTIFY_MEMORY_DEVICE)
> > +{
> > +    struct {
> > +        char fw_revision[0x10];
> > +        uint64_t total_capacity;
> > +        uint64_t volatile_capacity;
> > +        uint64_t persistent_capacity;
> > +        uint64_t partition_align;
> > +        uint16_t info_event_log_size;
> > +        uint16_t warning_event_log_size;
> > +        uint16_t failure_event_log_size;
> > +        uint16_t fatal_event_log_size;
> > +        uint32_t lsa_size;
> > +        uint8_t poison_list_max_mer[3];
> > +        uint16_t inject_poison_limit;
> > +        uint8_t poison_caps;
> > +        uint8_t qos_telemetry_caps;
> > +    } __attribute__((packed)) *id;
> > +    _Static_assert(sizeof(*id) == 0x43, "Bad identify size");
> > +
> > +    if (memory_region_size(cxl_dstate->pmem) < (256 << 20)) {
> > +        return CXL_MBOX_INTERNAL_ERROR;
> > +    }
> > +
> > +    id = (void *)cmd->payload;
> > +    memset(id, 0, sizeof(*id));
> > +
> > +    /* PMEM only */
> > +    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
> > +    id->total_capacity = memory_region_size(cxl_dstate->pmem);
> > +    id->persistent_capacity = memory_region_size(cxl_dstate->pmem);
> > +
> > +    *len = sizeof(*id);
> > +    return CXL_MBOX_SUCCESS;
> > +}
> > +
> >  void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> >  {
> >      uint16_t ret = CXL_MBOX_SUCCESS;
> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> > index 5458f61d10..5d41809b37 100644
> > --- a/hw/i386/pc.c
> > +++ b/hw/i386/pc.c
> > @@ -79,6 +79,7 @@
> >  #include "acpi-build.h"
> >  #include "hw/mem/pc-dimm.h"
> >  #include "hw/mem/nvdimm.h"
> > +#include "hw/cxl/cxl.h"
> >  #include "qapi/error.h"
> >  #include "qapi/qapi-visit-common.h"
> >  #include "qapi/visitor.h"
> > diff --git a/hw/mem/Kconfig b/hw/mem/Kconfig
> > index a0ef2cf648..7d9d1ced3e 100644
> > --- a/hw/mem/Kconfig
> > +++ b/hw/mem/Kconfig
> > @@ -10,3 +10,8 @@ config NVDIMM
> >      default y
> >      depends on (PC || PSERIES || ARM_VIRT)
> >      select MEM_DEVICE
> > +
> > +config CXL_MEM_DEVICE
> > +    bool
> > +    default y if CXL
> > +    select MEM_DEVICE
> > diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> > new file mode 100644
> > index 0000000000..3985bb8d0b
> > --- /dev/null
> > +++ b/hw/mem/cxl_type3.c
> > @@ -0,0 +1,262 @@
> > +#include "qemu/osdep.h"
> > +#include "qemu/units.h"
> > +#include "qemu/error-report.h"
> > +#include "hw/mem/memory-device.h"
> > +#include "hw/mem/pc-dimm.h"
> > +#include "hw/pci/pci.h"
> > +#include "hw/qdev-properties.h"
> > +#include "qapi/error.h"
> > +#include "qemu/log.h"
> > +#include "qemu/module.h"
> > +#include "qemu/range.h"
> > +#include "qemu/rcu.h"
> > +#include "sysemu/hostmem.h"
> > +#include "hw/cxl/cxl.h"
> > +
> > +typedef struct cxl_type3_dev {
> > +    /* Private */
> > +    PCIDevice parent_obj;
> > +
> > +    /* Properties */
> > +    uint64_t size;
> > +    HostMemoryBackend *hostmem;
> > +
> > +    /* State */
> > +    CXLComponentState cxl_cstate;
> > +    CXLDeviceState cxl_dstate;
> > +} CXLType3Dev;
> > +
> > +#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
> > +
> > +static void build_dvsecs(CXLType3Dev *ct3d)
> > +{
> > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > +    uint8_t *dvsec;
> > +
> > +    dvsec = (uint8_t *)&(struct dvsec_device){
> > +        .cap = 0x1e,
> > +        .ctrl = 0x6,
> > +        .status2 = 0x2,
> > +        .range1_size_hi = 0,
> > +        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
> > +        .range1_base_hi = 0,
> > +        .range1_base_lo = 0,
> > +    };
> > +    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
> > +                               PCIE_CXL_DEVICE_DVSEC,
> > +                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
> > +
> > +    dvsec = (uint8_t *)&(struct dvsec_register_locator){
> > +        .rsvd         = 0,
> > +        .reg0_base_lo = RBI_COMPONENT_REG | COMPONENT_REG_BAR_IDX,
> > +        .reg0_base_hi = 0,
> > +        .reg1_base_lo = RBI_CXL_DEVICE_REG | DEVICE_REG_BAR_IDX,
> > +        .reg1_base_hi = 0,
> > +    };
> > +    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
> > +                               REG_LOC_DVSEC_REVID, dvsec);
> > +}
> > +
> > +static void ct3_instance_init(Object *obj)
> > +{
> > +    /* MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); */
> > +}
> > +
> > +static void ct3_finalize(Object *obj)
> > +{
> > +    CXLType3Dev *ct3d = CT3(obj);
> > +
> > +    g_free(ct3d->cxl_dstate.pmem);
> > +}
> > +
> > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > +{
> > +    MemoryRegionSection mrs;
> > +    MemoryRegion *mr;
> > +    uint64_t offset = 0;
> > +    size_t remaining_size;
> > +
> > +    if (!ct3d->hostmem) {
> > +        error_setg(errp, "memdev property must be set");
> > +        return;
> > +    }
> > +
> > +    /* FIXME: need to check mr is the host bridge's MR */
> > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > +
> > +    /* Create our new subregion */
> > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > +
> > +    /* Find the first free space in the window */
> > +    WITH_RCU_READ_LOCK_GUARD()
> > +    {
> > +        mrs = memory_region_find(mr, offset, 1);
> > +        while (mrs.mr && mrs.mr != mr) {
> > +            offset += memory_region_size(mrs.mr);
> > +            mrs = memory_region_find(mr, offset, 1);
> > +        }
> > +    }
> > +
> > +    remaining_size = memory_region_size(mr) - offset;
> > +    if (remaining_size < ct3d->size) {
> > +        g_free(ct3d->cxl_dstate.pmem);
> > +        error_setg(errp,
> > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > +                   remaining_size, ct3d->size);
> > +    }
> > +
> > +    /* Register our subregion as non-volatile */
> > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > +                           "cxl_type3-memory", ct3d->size, errp);
> this allocates ct3d->size of anon RAM, was this an intention?
> If yes, can you clarify why extra RAM is used instead of using what
> backend provides?
> 
> > +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
> 
> > +#ifdef SET_PMEM_PADDR
> > +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > +#endif
> What this hunk is supposed to do, why it's ifdef-ed?
> 
> 
> > +}
> > +
> > +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> > +                                              Error **errp)
> > +{
> > +    CXLType3Dev *ct3d = CT3(md);
> > +
> > +    if (!ct3d->cxl_dstate.pmem) {
> > +        cxl_setup_memory(ct3d, errp);
> > +    }
> > +
> > +    return ct3d->cxl_dstate.pmem;
> > +}
> > +
> > +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > +{
> > +    CXLType3Dev *ct3d = CT3(pci_dev);
> > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > +    ComponentRegisters *regs = &cxl_cstate->crb;
> > +    MemoryRegion *mr = &regs->component_registers;
> > +    uint8_t *pci_conf = pci_dev->config;
> > +
> > +    if (!ct3d->cxl_dstate.pmem) {
> > +        cxl_setup_memory(ct3d, errp);
> > +    }
> > +
> > +    pci_config_set_prog_interface(pci_conf, 0x10);
> > +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> > +
> > +    pcie_endpoint_cap_init(pci_dev, 0x80);
> > +    cxl_cstate->dvsec_offset = 0x100;
> > +
> > +    ct3d->cxl_cstate.pdev = pci_dev;
> > +    build_dvsecs(ct3d);
> > +
> > +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> > +                                      TYPE_CXL_TYPE3_DEV);
> > +
> > +    pci_register_bar(
> > +        pci_dev, COMPONENT_REG_BAR_IDX,
> > +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> > +
> > +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> > +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> > +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> > +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> > +                     &ct3d->cxl_dstate.device_registers);
> > +}
> > +
> > +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> > +{
> > +    CXLType3Dev *ct3d = CT3(md);
> > +
> > +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> > +}
> > +
> > +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> > +{
> > +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> > +}
> > +
> > +static void ct3d_reset(DeviceState *dev)
> > +{
> > +    CXLType3Dev *ct3d = CT3(dev);
> > +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> > +
> > +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> > +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> > +}
> > +
> > +static Property ct3_props[] = {
> > +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> > +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> > +                     HostMemoryBackend *),
> > +    DEFINE_PROP_END_OF_LIST(),
> > +};
> > +
> > +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> > +                                        MemoryDeviceInfo *info)
> > +{
> > +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> > +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> > +    const DeviceState *dev = DEVICE(md);
> > +    CXLType3Dev *ct3d = CT3(md);
> > +
> > +    if (dev->id) {
> > +        di->has_id = true;
> > +        di->id = g_strdup(dev->id);
> > +    }
> > +    di->hotplugged = dev->hotplugged;
> > +    di->hotpluggable = dc->hotpluggable;
> > +    di->addr = cxl_md_get_addr(md);
> > +    di->slot = 0;
> > +    di->node = 0;
> > +    di->size = memory_device_get_region_size(md, NULL);
> > +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> > +
> > +
> > +    info->u.cxl.data = di;
> > +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> > +}
> > +
> > +static void ct3_class_init(ObjectClass *oc, void *data)
> > +{
> > +    DeviceClass *dc = DEVICE_CLASS(oc);
> > +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> > +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> > +
> > +    pc->realize = ct3_realize;
> > +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> > +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> > +    pc->device_id = 0xd93; /* LVF for now */
> > +    pc->revision = 1;
> > +
> > +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> > +    dc->desc = "CXL PMEM Device (Type 3)";
> > +    dc->reset = ct3d_reset;
> > +    device_class_set_props(dc, ct3_props);
> > +
> > +    mdc->get_memory_region = cxl_md_get_memory_region;
> > +    mdc->get_addr = cxl_md_get_addr;
> > +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> > +    mdc->get_plugged_size = memory_device_get_region_size;
> > +    mdc->set_addr = cxl_md_set_addr;
> > +}
> > +
> > +static const TypeInfo ct3d_info = {
> > +    .name = TYPE_CXL_TYPE3_DEV,
> > +    .parent = TYPE_PCI_DEVICE,
> > +    .class_init = ct3_class_init,
> > +    .instance_size = sizeof(CXLType3Dev),
> > +    .instance_init = ct3_instance_init,
> > +    .instance_finalize = ct3_finalize,
> > +    .interfaces = (InterfaceInfo[]) {
> > +        { TYPE_MEMORY_DEVICE },
> > +        { INTERFACE_CXL_DEVICE },
> > +        { INTERFACE_PCIE_DEVICE },
> > +        {}
> > +    },
> > +};
> > +
> > +static void ct3d_registers(void)
> > +{
> > +    type_register_static(&ct3d_info);
> > +}
> > +
> > +type_init(ct3d_registers);
> > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > index 0d22f2b572..d13c3ed117 100644
> > --- a/hw/mem/meson.build
> > +++ b/hw/mem/meson.build
> > @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
> >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
> >  
> >  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> > diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> > index d4010cf8f3..1ecf6f6a55 100644
> > --- a/hw/pci/pcie.c
> > +++ b/hw/pci/pcie.c
> > @@ -20,6 +20,7 @@
> >  
> >  #include "qemu/osdep.h"
> >  #include "qapi/error.h"
> > +#include "hw/mem/memory-device.h"
> >  #include "hw/pci/pci_bridge.h"
> >  #include "hw/pci/pcie.h"
> >  #include "hw/pci/msix.h"
> > @@ -27,6 +28,8 @@
> >  #include "hw/pci/pci_bus.h"
> >  #include "hw/pci/pcie_regs.h"
> >  #include "hw/pci/pcie_port.h"
> > +#include "hw/cxl/cxl.h"
> > +#include "hw/boards.h"
> >  #include "qemu/range.h"
> >  
> >  //#define DEBUG_PCIE
> > @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> >      }
> >  
> >      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> > +
> > +#ifdef CXL_MEM_DEVICE
> > +    /*
> > +     * FIXME:
> > +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > +     *    HotplugHandler *hotplug_ctrl;
> > +     *   Error *local_err = NULL;
> > +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> > +     *  if (hotplug_ctrl) {
> > +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> > +     *      if (local_err) {
> > +     *          error_propagate(errp, local_err);
> > +     *          return;
> > +     *      }
> > +     *  }
> > +     */
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> > +                               NULL, errp);
> > +    }
> 
> why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?
> 
> > +#endif
> >  }
> >  
> >  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> >          pcie_cap_slot_event(hotplug_pdev,
> >                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
> >      }
> > +
> > +#ifdef CXL_MEM_DEVICE
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> > +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> > +#endif
> >  }
> >  
> >  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > index b1e5f4a8fa..809ed7de60 100644
> > --- a/include/hw/cxl/cxl.h
> > +++ b/include/hw/cxl/cxl.h
> > @@ -17,6 +17,8 @@
> >  #define COMPONENT_REG_BAR_IDX 0
> >  #define DEVICE_REG_BAR_IDX 2
> >  
> > +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> > +
> >  #define CXL_HOST_BASE 0xD0000000
> >  #define CXL_WINDOW_MAX 10
> >  
> > diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> > index a53c2e5ae7..9ec28c9feb 100644
> > --- a/include/hw/cxl/cxl_pci.h
> > +++ b/include/hw/cxl/cxl_pci.h
> > @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
> >   * CXL 2.0 Downstream Port: 3, 4, 7, 8
> >   */
> >  
> > +/* CXL 2.0 - 8.1.3 (ID 0001) */
> > +struct dvsec_device {
> > +    struct dvsec_header hdr;
> > +    uint16_t cap;
> > +    uint16_t ctrl;
> > +    uint16_t status;
> > +    uint16_t ctrl2;
> > +    uint16_t status2;
> > +    uint16_t lock;
> > +    uint16_t cap2;
> > +    uint32_t range1_size_hi;
> > +    uint32_t range1_size_lo;
> > +    uint32_t range1_base_hi;
> > +    uint32_t range1_base_lo;
> > +    uint32_t range2_size_hi;
> > +    uint32_t range2_size_lo;
> > +    uint32_t range2_base_hi;
> > +    uint32_t range2_base_lo;
> > +};
> > +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> > +               "dvsec device size incorrect");
> > +
> >  /* CXL 2.0 - 8.1.5 (ID 0003) */
> >  struct extensions_dvsec_port {
> >      struct dvsec_header hdr;
> > diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> > index 11f8ab7149..76bf3ed590 100644
> > --- a/include/hw/pci/pci_ids.h
> > +++ b/include/hw/pci/pci_ids.h
> > @@ -53,6 +53,7 @@
> >  #define PCI_BASE_CLASS_MEMORY            0x05
> >  #define PCI_CLASS_MEMORY_RAM             0x0500
> >  #define PCI_CLASS_MEMORY_FLASH           0x0501
> > +#define PCI_CLASS_MEMORY_CXL             0x0502
> >  #define PCI_CLASS_MEMORY_OTHER           0x0580
> >  
> >  #define PCI_BASE_CLASS_BRIDGE            0x06
> > diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> > index 0dd594f92b..0f67bc61ce 100644
> > --- a/monitor/hmp-cmds.c
> > +++ b/monitor/hmp-cmds.c
> > @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
> >                  monitor_printf(mon, "  hotpluggable: %s\n",
> >                                 di->hotpluggable ? "true" : "false");
> >                  break;
> > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > +                di = value->u.cxl.data;
> > +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > +                               MemoryDeviceInfoKind_str(value->type),
> > +                               di->id ? di->id : "");
> > +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> > +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> > +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> > +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> > +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> > +                monitor_printf(mon, "  hotplugged: %s\n",
> > +                               di->hotplugged ? "true" : "false");
> > +                monitor_printf(mon, "  hotpluggable: %s\n",
> > +                               di->hotpluggable ? "true" : "false");
> > +                break;
> >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> >                  vpi = value->u.virtio_pmem.data;
> >                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > diff --git a/qapi/machine.json b/qapi/machine.json
> > index 330189efe3..aa96d662bd 100644
> > --- a/qapi/machine.json
> > +++ b/qapi/machine.json
> > @@ -1394,6 +1394,7 @@
> >  { 'union': 'MemoryDeviceInfo',
> >    'data': { 'dimm': 'PCDIMMDeviceInfo',
> >              'nvdimm': 'PCDIMMDeviceInfo',
> > +            'cxl': 'PCDIMMDeviceInfo',
> >              'virtio-pmem': 'VirtioPMEMDeviceInfo',
> >              'virtio-mem': 'VirtioMEMDeviceInfo'
> >            }
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-27 21:11     ` Ben Widawsky
@ 2021-01-27 21:21       ` Igor Mammedov
  2021-01-27 21:30         ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Igor Mammedov @ 2021-01-27 21:21 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Jonathan Cameron, Prashant V Agarwal, Dan Williams

On Wed, 27 Jan 2021 13:11:16 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> On 21-01-27 22:03:12, Igor Mammedov wrote:
> > On Tue,  5 Jan 2021 08:53:15 -0800
> > Ben Widawsky <ben.widawsky@intel.com> wrote:
> >   
> > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > combination of volatile and persistent memory. It also implements the
> > > previously defined mailbox interface as well as the memory device
> > > firmware interface.
> > > 
> > > The following example will create a 256M device in a 512M window:
> > > 
> > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > 
> > I'd expect whole backend used by frontend, so one would not need "size" property
> > on frontend (like we do with memory devices).
> > So question is why it partially uses memdev?  
> 
> A CXL memory device may participate in an interleave set. In such a case, it
> would be < the total size of the memory window.
> 
> This isn't implemented in the code yet, but it is planned.

could you add here how it supposed to look like CLI interface wise? 

also see other questions below.

> > 
> >   
> > > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> > > ---
> > >  hw/core/numa.c             |   3 +
> > >  hw/cxl/cxl-mailbox-utils.c |  41 ++++++
> > >  hw/i386/pc.c               |   1 +
> > >  hw/mem/Kconfig             |   5 +
> > >  hw/mem/cxl_type3.c         | 262 +++++++++++++++++++++++++++++++++++++
> > >  hw/mem/meson.build         |   1 +
> > >  hw/pci/pcie.c              |  30 +++++
> > >  include/hw/cxl/cxl.h       |   2 +
> > >  include/hw/cxl/cxl_pci.h   |  22 ++++
> > >  include/hw/pci/pci_ids.h   |   1 +
> > >  monitor/hmp-cmds.c         |  15 +++
> > >  qapi/machine.json          |   1 +
> > >  12 files changed, 384 insertions(+)
> > >  create mode 100644 hw/mem/cxl_type3.c
> > > 
> > > diff --git a/hw/core/numa.c b/hw/core/numa.c
> > > index 68cee65f61..cd7df371e6 100644
> > > --- a/hw/core/numa.c
> > > +++ b/hw/core/numa.c
> > > @@ -770,6 +770,9 @@ static void numa_stat_memory_devices(NumaNodeMem node_mem[])
> > >                  node_mem[pcdimm_info->node].node_plugged_mem +=
> > >                      pcdimm_info->size;
> > >                  break;
> > > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > > +                /* FINISHME */
> > > +                break;
> > >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> > >                  vpi = value->u.virtio_pmem.data;
> > >                  /* TODO: once we support numa, assign to right node */
> > > diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> > > index f68ec5b5b9..eeb10b8943 100644
> > > --- a/hw/cxl/cxl-mailbox-utils.c
> > > +++ b/hw/cxl/cxl-mailbox-utils.c
> > > @@ -49,6 +49,8 @@ enum {
> > >      LOGS        = 0x04,
> > >          #define GET_SUPPORTED 0x0
> > >          #define GET_LOG       0x1
> > > +    IDENTIFY    = 0x40,
> > > +        #define MEMORY_DEVICE 0x0
> > >  };
> > >  
> > >  /* 8.2.8.4.5.1 Command Return Codes */
> > > @@ -127,6 +129,7 @@ declare_mailbox_handler(TIMESTAMP_GET);
> > >  declare_mailbox_handler(TIMESTAMP_SET);
> > >  declare_mailbox_handler(LOGS_GET_SUPPORTED);
> > >  declare_mailbox_handler(LOGS_GET_LOG);
> > > +declare_mailbox_handler(IDENTIFY_MEMORY_DEVICE);
> > >  
> > >  #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
> > >  #define IMMEDIATE_POLICY_CHANGE (1 << 3)
> > > @@ -144,6 +147,7 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
> > >      CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
> > >      CXL_CMD(LOGS, GET_SUPPORTED, 0, 0),
> > >      CXL_CMD(LOGS, GET_LOG, 0x18, 0),
> > > +    CXL_CMD(IDENTIFY, MEMORY_DEVICE, 0, 0),
> > >  };
> > >  
> > >  #undef CXL_CMD
> > > @@ -262,6 +266,43 @@ define_mailbox_handler(LOGS_GET_LOG)
> > >      return CXL_MBOX_SUCCESS;
> > >  }
> > >  
> > > +/* 8.2.9.5.1.1 */
> > > +define_mailbox_handler(IDENTIFY_MEMORY_DEVICE)
> > > +{
> > > +    struct {
> > > +        char fw_revision[0x10];
> > > +        uint64_t total_capacity;
> > > +        uint64_t volatile_capacity;
> > > +        uint64_t persistent_capacity;
> > > +        uint64_t partition_align;
> > > +        uint16_t info_event_log_size;
> > > +        uint16_t warning_event_log_size;
> > > +        uint16_t failure_event_log_size;
> > > +        uint16_t fatal_event_log_size;
> > > +        uint32_t lsa_size;
> > > +        uint8_t poison_list_max_mer[3];
> > > +        uint16_t inject_poison_limit;
> > > +        uint8_t poison_caps;
> > > +        uint8_t qos_telemetry_caps;
> > > +    } __attribute__((packed)) *id;
> > > +    _Static_assert(sizeof(*id) == 0x43, "Bad identify size");
> > > +
> > > +    if (memory_region_size(cxl_dstate->pmem) < (256 << 20)) {
> > > +        return CXL_MBOX_INTERNAL_ERROR;
> > > +    }
> > > +
> > > +    id = (void *)cmd->payload;
> > > +    memset(id, 0, sizeof(*id));
> > > +
> > > +    /* PMEM only */
> > > +    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
> > > +    id->total_capacity = memory_region_size(cxl_dstate->pmem);
> > > +    id->persistent_capacity = memory_region_size(cxl_dstate->pmem);
> > > +
> > > +    *len = sizeof(*id);
> > > +    return CXL_MBOX_SUCCESS;
> > > +}
> > > +
> > >  void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> > >  {
> > >      uint16_t ret = CXL_MBOX_SUCCESS;
> > > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> > > index 5458f61d10..5d41809b37 100644
> > > --- a/hw/i386/pc.c
> > > +++ b/hw/i386/pc.c
> > > @@ -79,6 +79,7 @@
> > >  #include "acpi-build.h"
> > >  #include "hw/mem/pc-dimm.h"
> > >  #include "hw/mem/nvdimm.h"
> > > +#include "hw/cxl/cxl.h"
> > >  #include "qapi/error.h"
> > >  #include "qapi/qapi-visit-common.h"
> > >  #include "qapi/visitor.h"
> > > diff --git a/hw/mem/Kconfig b/hw/mem/Kconfig
> > > index a0ef2cf648..7d9d1ced3e 100644
> > > --- a/hw/mem/Kconfig
> > > +++ b/hw/mem/Kconfig
> > > @@ -10,3 +10,8 @@ config NVDIMM
> > >      default y
> > >      depends on (PC || PSERIES || ARM_VIRT)
> > >      select MEM_DEVICE
> > > +
> > > +config CXL_MEM_DEVICE
> > > +    bool
> > > +    default y if CXL
> > > +    select MEM_DEVICE
> > > diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> > > new file mode 100644
> > > index 0000000000..3985bb8d0b
> > > --- /dev/null
> > > +++ b/hw/mem/cxl_type3.c
> > > @@ -0,0 +1,262 @@
> > > +#include "qemu/osdep.h"
> > > +#include "qemu/units.h"
> > > +#include "qemu/error-report.h"
> > > +#include "hw/mem/memory-device.h"
> > > +#include "hw/mem/pc-dimm.h"
> > > +#include "hw/pci/pci.h"
> > > +#include "hw/qdev-properties.h"
> > > +#include "qapi/error.h"
> > > +#include "qemu/log.h"
> > > +#include "qemu/module.h"
> > > +#include "qemu/range.h"
> > > +#include "qemu/rcu.h"
> > > +#include "sysemu/hostmem.h"
> > > +#include "hw/cxl/cxl.h"
> > > +
> > > +typedef struct cxl_type3_dev {
> > > +    /* Private */
> > > +    PCIDevice parent_obj;
> > > +
> > > +    /* Properties */
> > > +    uint64_t size;
> > > +    HostMemoryBackend *hostmem;
> > > +
> > > +    /* State */
> > > +    CXLComponentState cxl_cstate;
> > > +    CXLDeviceState cxl_dstate;
> > > +} CXLType3Dev;
> > > +
> > > +#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
> > > +
> > > +static void build_dvsecs(CXLType3Dev *ct3d)
> > > +{
> > > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > > +    uint8_t *dvsec;
> > > +
> > > +    dvsec = (uint8_t *)&(struct dvsec_device){
> > > +        .cap = 0x1e,
> > > +        .ctrl = 0x6,
> > > +        .status2 = 0x2,
> > > +        .range1_size_hi = 0,
> > > +        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
> > > +        .range1_base_hi = 0,
> > > +        .range1_base_lo = 0,
> > > +    };
> > > +    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
> > > +                               PCIE_CXL_DEVICE_DVSEC,
> > > +                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
> > > +
> > > +    dvsec = (uint8_t *)&(struct dvsec_register_locator){
> > > +        .rsvd         = 0,
> > > +        .reg0_base_lo = RBI_COMPONENT_REG | COMPONENT_REG_BAR_IDX,
> > > +        .reg0_base_hi = 0,
> > > +        .reg1_base_lo = RBI_CXL_DEVICE_REG | DEVICE_REG_BAR_IDX,
> > > +        .reg1_base_hi = 0,
> > > +    };
> > > +    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
> > > +                               REG_LOC_DVSEC_REVID, dvsec);
> > > +}
> > > +
> > > +static void ct3_instance_init(Object *obj)
> > > +{
> > > +    /* MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); */
> > > +}
> > > +
> > > +static void ct3_finalize(Object *obj)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(obj);
> > > +
> > > +    g_free(ct3d->cxl_dstate.pmem);
> > > +}
> > > +
> > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > +{
> > > +    MemoryRegionSection mrs;
> > > +    MemoryRegion *mr;
> > > +    uint64_t offset = 0;
> > > +    size_t remaining_size;
> > > +
> > > +    if (!ct3d->hostmem) {
> > > +        error_setg(errp, "memdev property must be set");
> > > +        return;
> > > +    }
> > > +
> > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > +
> > > +    /* Create our new subregion */
> > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > +
> > > +    /* Find the first free space in the window */
> > > +    WITH_RCU_READ_LOCK_GUARD()
> > > +    {
> > > +        mrs = memory_region_find(mr, offset, 1);
> > > +        while (mrs.mr && mrs.mr != mr) {
> > > +            offset += memory_region_size(mrs.mr);
> > > +            mrs = memory_region_find(mr, offset, 1);
> > > +        }
> > > +    }
> > > +
> > > +    remaining_size = memory_region_size(mr) - offset;
> > > +    if (remaining_size < ct3d->size) {
> > > +        g_free(ct3d->cxl_dstate.pmem);
> > > +        error_setg(errp,
> > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > +                   remaining_size, ct3d->size);
> > > +    }
> > > +
> > > +    /* Register our subregion as non-volatile */
> > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > +                           "cxl_type3-memory", ct3d->size, errp);  
> > this allocates ct3d->size of anon RAM, was this an intention?
> > If yes, can you clarify why extra RAM is used instead of using what
> > backend provides?
> >   
> > > +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);  
> >   
> > > +#ifdef SET_PMEM_PADDR
> > > +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > > +#endif  
> > What this hunk is supposed to do, why it's ifdef-ed?
> > 
> >   
> > > +}
> > > +
> > > +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> > > +                                              Error **errp)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(md);
> > > +
> > > +    if (!ct3d->cxl_dstate.pmem) {
> > > +        cxl_setup_memory(ct3d, errp);
> > > +    }
> > > +
> > > +    return ct3d->cxl_dstate.pmem;
> > > +}
> > > +
> > > +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(pci_dev);
> > > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > > +    ComponentRegisters *regs = &cxl_cstate->crb;
> > > +    MemoryRegion *mr = &regs->component_registers;
> > > +    uint8_t *pci_conf = pci_dev->config;
> > > +
> > > +    if (!ct3d->cxl_dstate.pmem) {
> > > +        cxl_setup_memory(ct3d, errp);
> > > +    }
> > > +
> > > +    pci_config_set_prog_interface(pci_conf, 0x10);
> > > +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> > > +
> > > +    pcie_endpoint_cap_init(pci_dev, 0x80);
> > > +    cxl_cstate->dvsec_offset = 0x100;
> > > +
> > > +    ct3d->cxl_cstate.pdev = pci_dev;
> > > +    build_dvsecs(ct3d);
> > > +
> > > +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> > > +                                      TYPE_CXL_TYPE3_DEV);
> > > +
> > > +    pci_register_bar(
> > > +        pci_dev, COMPONENT_REG_BAR_IDX,
> > > +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> > > +
> > > +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> > > +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> > > +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> > > +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> > > +                     &ct3d->cxl_dstate.device_registers);
> > > +}
> > > +
> > > +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(md);
> > > +
> > > +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> > > +}
> > > +
> > > +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> > > +{
> > > +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> > > +}
> > > +
> > > +static void ct3d_reset(DeviceState *dev)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(dev);
> > > +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> > > +
> > > +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> > > +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> > > +}
> > > +
> > > +static Property ct3_props[] = {
> > > +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> > > +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> > > +                     HostMemoryBackend *),
> > > +    DEFINE_PROP_END_OF_LIST(),
> > > +};
> > > +
> > > +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> > > +                                        MemoryDeviceInfo *info)
> > > +{
> > > +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> > > +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> > > +    const DeviceState *dev = DEVICE(md);
> > > +    CXLType3Dev *ct3d = CT3(md);
> > > +
> > > +    if (dev->id) {
> > > +        di->has_id = true;
> > > +        di->id = g_strdup(dev->id);
> > > +    }
> > > +    di->hotplugged = dev->hotplugged;
> > > +    di->hotpluggable = dc->hotpluggable;
> > > +    di->addr = cxl_md_get_addr(md);
> > > +    di->slot = 0;
> > > +    di->node = 0;
> > > +    di->size = memory_device_get_region_size(md, NULL);
> > > +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> > > +
> > > +
> > > +    info->u.cxl.data = di;
> > > +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> > > +}
> > > +
> > > +static void ct3_class_init(ObjectClass *oc, void *data)
> > > +{
> > > +    DeviceClass *dc = DEVICE_CLASS(oc);
> > > +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> > > +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> > > +
> > > +    pc->realize = ct3_realize;
> > > +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> > > +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> > > +    pc->device_id = 0xd93; /* LVF for now */
> > > +    pc->revision = 1;
> > > +
> > > +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> > > +    dc->desc = "CXL PMEM Device (Type 3)";
> > > +    dc->reset = ct3d_reset;
> > > +    device_class_set_props(dc, ct3_props);
> > > +
> > > +    mdc->get_memory_region = cxl_md_get_memory_region;
> > > +    mdc->get_addr = cxl_md_get_addr;
> > > +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> > > +    mdc->get_plugged_size = memory_device_get_region_size;
> > > +    mdc->set_addr = cxl_md_set_addr;
> > > +}
> > > +
> > > +static const TypeInfo ct3d_info = {
> > > +    .name = TYPE_CXL_TYPE3_DEV,
> > > +    .parent = TYPE_PCI_DEVICE,
> > > +    .class_init = ct3_class_init,
> > > +    .instance_size = sizeof(CXLType3Dev),
> > > +    .instance_init = ct3_instance_init,
> > > +    .instance_finalize = ct3_finalize,
> > > +    .interfaces = (InterfaceInfo[]) {
> > > +        { TYPE_MEMORY_DEVICE },
> > > +        { INTERFACE_CXL_DEVICE },
> > > +        { INTERFACE_PCIE_DEVICE },
> > > +        {}
> > > +    },
> > > +};
> > > +
> > > +static void ct3d_registers(void)
> > > +{
> > > +    type_register_static(&ct3d_info);
> > > +}
> > > +
> > > +type_init(ct3d_registers);
> > > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > > index 0d22f2b572..d13c3ed117 100644
> > > --- a/hw/mem/meson.build
> > > +++ b/hw/mem/meson.build
> > > @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
> > >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> > >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> > >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > > +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
> > >  
> > >  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> > > diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> > > index d4010cf8f3..1ecf6f6a55 100644
> > > --- a/hw/pci/pcie.c
> > > +++ b/hw/pci/pcie.c
> > > @@ -20,6 +20,7 @@
> > >  
> > >  #include "qemu/osdep.h"
> > >  #include "qapi/error.h"
> > > +#include "hw/mem/memory-device.h"
> > >  #include "hw/pci/pci_bridge.h"
> > >  #include "hw/pci/pcie.h"
> > >  #include "hw/pci/msix.h"
> > > @@ -27,6 +28,8 @@
> > >  #include "hw/pci/pci_bus.h"
> > >  #include "hw/pci/pcie_regs.h"
> > >  #include "hw/pci/pcie_port.h"
> > > +#include "hw/cxl/cxl.h"
> > > +#include "hw/boards.h"
> > >  #include "qemu/range.h"
> > >  
> > >  //#define DEBUG_PCIE
> > > @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > >      }
> > >  
> > >      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> > > +
> > > +#ifdef CXL_MEM_DEVICE
> > > +    /*
> > > +     * FIXME:
> > > +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > +     *    HotplugHandler *hotplug_ctrl;
> > > +     *   Error *local_err = NULL;
> > > +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> > > +     *  if (hotplug_ctrl) {
> > > +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> > > +     *      if (local_err) {
> > > +     *          error_propagate(errp, local_err);
> > > +     *          return;
> > > +     *      }
> > > +     *  }
> > > +     */
> > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> > > +                               NULL, errp);
> > > +    }  
> > 
> > why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?
> >   
> > > +#endif
> > >  }
> > >  
> > >  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > >          pcie_cap_slot_event(hotplug_pdev,
> > >                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
> > >      }
> > > +
> > > +#ifdef CXL_MEM_DEVICE
> > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> > > +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> > > +#endif
> > >  }
> > >  
> > >  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > > index b1e5f4a8fa..809ed7de60 100644
> > > --- a/include/hw/cxl/cxl.h
> > > +++ b/include/hw/cxl/cxl.h
> > > @@ -17,6 +17,8 @@
> > >  #define COMPONENT_REG_BAR_IDX 0
> > >  #define DEVICE_REG_BAR_IDX 2
> > >  
> > > +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> > > +
> > >  #define CXL_HOST_BASE 0xD0000000
> > >  #define CXL_WINDOW_MAX 10
> > >  
> > > diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> > > index a53c2e5ae7..9ec28c9feb 100644
> > > --- a/include/hw/cxl/cxl_pci.h
> > > +++ b/include/hw/cxl/cxl_pci.h
> > > @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
> > >   * CXL 2.0 Downstream Port: 3, 4, 7, 8
> > >   */
> > >  
> > > +/* CXL 2.0 - 8.1.3 (ID 0001) */
> > > +struct dvsec_device {
> > > +    struct dvsec_header hdr;
> > > +    uint16_t cap;
> > > +    uint16_t ctrl;
> > > +    uint16_t status;
> > > +    uint16_t ctrl2;
> > > +    uint16_t status2;
> > > +    uint16_t lock;
> > > +    uint16_t cap2;
> > > +    uint32_t range1_size_hi;
> > > +    uint32_t range1_size_lo;
> > > +    uint32_t range1_base_hi;
> > > +    uint32_t range1_base_lo;
> > > +    uint32_t range2_size_hi;
> > > +    uint32_t range2_size_lo;
> > > +    uint32_t range2_base_hi;
> > > +    uint32_t range2_base_lo;
> > > +};
> > > +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> > > +               "dvsec device size incorrect");
> > > +
> > >  /* CXL 2.0 - 8.1.5 (ID 0003) */
> > >  struct extensions_dvsec_port {
> > >      struct dvsec_header hdr;
> > > diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> > > index 11f8ab7149..76bf3ed590 100644
> > > --- a/include/hw/pci/pci_ids.h
> > > +++ b/include/hw/pci/pci_ids.h
> > > @@ -53,6 +53,7 @@
> > >  #define PCI_BASE_CLASS_MEMORY            0x05
> > >  #define PCI_CLASS_MEMORY_RAM             0x0500
> > >  #define PCI_CLASS_MEMORY_FLASH           0x0501
> > > +#define PCI_CLASS_MEMORY_CXL             0x0502
> > >  #define PCI_CLASS_MEMORY_OTHER           0x0580
> > >  
> > >  #define PCI_BASE_CLASS_BRIDGE            0x06
> > > diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> > > index 0dd594f92b..0f67bc61ce 100644
> > > --- a/monitor/hmp-cmds.c
> > > +++ b/monitor/hmp-cmds.c
> > > @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
> > >                  monitor_printf(mon, "  hotpluggable: %s\n",
> > >                                 di->hotpluggable ? "true" : "false");
> > >                  break;
> > > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > > +                di = value->u.cxl.data;
> > > +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > +                               MemoryDeviceInfoKind_str(value->type),
> > > +                               di->id ? di->id : "");
> > > +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> > > +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> > > +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> > > +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> > > +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> > > +                monitor_printf(mon, "  hotplugged: %s\n",
> > > +                               di->hotplugged ? "true" : "false");
> > > +                monitor_printf(mon, "  hotpluggable: %s\n",
> > > +                               di->hotpluggable ? "true" : "false");
> > > +                break;
> > >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> > >                  vpi = value->u.virtio_pmem.data;
> > >                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > diff --git a/qapi/machine.json b/qapi/machine.json
> > > index 330189efe3..aa96d662bd 100644
> > > --- a/qapi/machine.json
> > > +++ b/qapi/machine.json
> > > @@ -1394,6 +1394,7 @@
> > >  { 'union': 'MemoryDeviceInfo',
> > >    'data': { 'dimm': 'PCDIMMDeviceInfo',
> > >              'nvdimm': 'PCDIMMDeviceInfo',
> > > +            'cxl': 'PCDIMMDeviceInfo',
> > >              'virtio-pmem': 'VirtioPMEMDeviceInfo',
> > >              'virtio-mem': 'VirtioMEMDeviceInfo'
> > >            }  
> >   
> 



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-27 21:03   ` Igor Mammedov
  2021-01-27 21:11     ` Ben Widawsky
@ 2021-01-27 21:26     ` Ben Widawsky
  2021-01-28 10:25       ` Jonathan Cameron
  1 sibling, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-27 21:26 UTC (permalink / raw)
  To: Igor Mammedov
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Jonathan Cameron, Prashant V Agarwal, Dan Williams

On 21-01-27 22:03:12, Igor Mammedov wrote:
> On Tue,  5 Jan 2021 08:53:15 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > combination of volatile and persistent memory. It also implements the
> > previously defined mailbox interface as well as the memory device
> > firmware interface.
> > 
> > The following example will create a 256M device in a 512M window:
> > 
> > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"
> 
> I'd expect whole backend used by frontend, so one would not need "size" property
> on frontend (like we do with memory devices).
> So question is why it partially uses memdev?

Answered in a separate thread...

> 
> 
> > Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
> > ---
> >  hw/core/numa.c             |   3 +
> >  hw/cxl/cxl-mailbox-utils.c |  41 ++++++
> >  hw/i386/pc.c               |   1 +
> >  hw/mem/Kconfig             |   5 +
> >  hw/mem/cxl_type3.c         | 262 +++++++++++++++++++++++++++++++++++++
> >  hw/mem/meson.build         |   1 +
> >  hw/pci/pcie.c              |  30 +++++
> >  include/hw/cxl/cxl.h       |   2 +
> >  include/hw/cxl/cxl_pci.h   |  22 ++++
> >  include/hw/pci/pci_ids.h   |   1 +
> >  monitor/hmp-cmds.c         |  15 +++
> >  qapi/machine.json          |   1 +
> >  12 files changed, 384 insertions(+)
> >  create mode 100644 hw/mem/cxl_type3.c
> > 
> > diff --git a/hw/core/numa.c b/hw/core/numa.c
> > index 68cee65f61..cd7df371e6 100644
> > --- a/hw/core/numa.c
> > +++ b/hw/core/numa.c
> > @@ -770,6 +770,9 @@ static void numa_stat_memory_devices(NumaNodeMem node_mem[])
> >                  node_mem[pcdimm_info->node].node_plugged_mem +=
> >                      pcdimm_info->size;
> >                  break;
> > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > +                /* FINISHME */
> > +                break;
> >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> >                  vpi = value->u.virtio_pmem.data;
> >                  /* TODO: once we support numa, assign to right node */
> > diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
> > index f68ec5b5b9..eeb10b8943 100644
> > --- a/hw/cxl/cxl-mailbox-utils.c
> > +++ b/hw/cxl/cxl-mailbox-utils.c
> > @@ -49,6 +49,8 @@ enum {
> >      LOGS        = 0x04,
> >          #define GET_SUPPORTED 0x0
> >          #define GET_LOG       0x1
> > +    IDENTIFY    = 0x40,
> > +        #define MEMORY_DEVICE 0x0
> >  };
> >  
> >  /* 8.2.8.4.5.1 Command Return Codes */
> > @@ -127,6 +129,7 @@ declare_mailbox_handler(TIMESTAMP_GET);
> >  declare_mailbox_handler(TIMESTAMP_SET);
> >  declare_mailbox_handler(LOGS_GET_SUPPORTED);
> >  declare_mailbox_handler(LOGS_GET_LOG);
> > +declare_mailbox_handler(IDENTIFY_MEMORY_DEVICE);
> >  
> >  #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
> >  #define IMMEDIATE_POLICY_CHANGE (1 << 3)
> > @@ -144,6 +147,7 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
> >      CXL_CMD(TIMESTAMP, SET, 8, IMMEDIATE_POLICY_CHANGE),
> >      CXL_CMD(LOGS, GET_SUPPORTED, 0, 0),
> >      CXL_CMD(LOGS, GET_LOG, 0x18, 0),
> > +    CXL_CMD(IDENTIFY, MEMORY_DEVICE, 0, 0),
> >  };
> >  
> >  #undef CXL_CMD
> > @@ -262,6 +266,43 @@ define_mailbox_handler(LOGS_GET_LOG)
> >      return CXL_MBOX_SUCCESS;
> >  }
> >  
> > +/* 8.2.9.5.1.1 */
> > +define_mailbox_handler(IDENTIFY_MEMORY_DEVICE)
> > +{
> > +    struct {
> > +        char fw_revision[0x10];
> > +        uint64_t total_capacity;
> > +        uint64_t volatile_capacity;
> > +        uint64_t persistent_capacity;
> > +        uint64_t partition_align;
> > +        uint16_t info_event_log_size;
> > +        uint16_t warning_event_log_size;
> > +        uint16_t failure_event_log_size;
> > +        uint16_t fatal_event_log_size;
> > +        uint32_t lsa_size;
> > +        uint8_t poison_list_max_mer[3];
> > +        uint16_t inject_poison_limit;
> > +        uint8_t poison_caps;
> > +        uint8_t qos_telemetry_caps;
> > +    } __attribute__((packed)) *id;
> > +    _Static_assert(sizeof(*id) == 0x43, "Bad identify size");
> > +
> > +    if (memory_region_size(cxl_dstate->pmem) < (256 << 20)) {
> > +        return CXL_MBOX_INTERNAL_ERROR;
> > +    }
> > +
> > +    id = (void *)cmd->payload;
> > +    memset(id, 0, sizeof(*id));
> > +
> > +    /* PMEM only */
> > +    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
> > +    id->total_capacity = memory_region_size(cxl_dstate->pmem);
> > +    id->persistent_capacity = memory_region_size(cxl_dstate->pmem);
> > +
> > +    *len = sizeof(*id);
> > +    return CXL_MBOX_SUCCESS;
> > +}
> > +
> >  void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
> >  {
> >      uint16_t ret = CXL_MBOX_SUCCESS;
> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> > index 5458f61d10..5d41809b37 100644
> > --- a/hw/i386/pc.c
> > +++ b/hw/i386/pc.c
> > @@ -79,6 +79,7 @@
> >  #include "acpi-build.h"
> >  #include "hw/mem/pc-dimm.h"
> >  #include "hw/mem/nvdimm.h"
> > +#include "hw/cxl/cxl.h"
> >  #include "qapi/error.h"
> >  #include "qapi/qapi-visit-common.h"
> >  #include "qapi/visitor.h"
> > diff --git a/hw/mem/Kconfig b/hw/mem/Kconfig
> > index a0ef2cf648..7d9d1ced3e 100644
> > --- a/hw/mem/Kconfig
> > +++ b/hw/mem/Kconfig
> > @@ -10,3 +10,8 @@ config NVDIMM
> >      default y
> >      depends on (PC || PSERIES || ARM_VIRT)
> >      select MEM_DEVICE
> > +
> > +config CXL_MEM_DEVICE
> > +    bool
> > +    default y if CXL
> > +    select MEM_DEVICE
> > diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> > new file mode 100644
> > index 0000000000..3985bb8d0b
> > --- /dev/null
> > +++ b/hw/mem/cxl_type3.c
> > @@ -0,0 +1,262 @@
> > +#include "qemu/osdep.h"
> > +#include "qemu/units.h"
> > +#include "qemu/error-report.h"
> > +#include "hw/mem/memory-device.h"
> > +#include "hw/mem/pc-dimm.h"
> > +#include "hw/pci/pci.h"
> > +#include "hw/qdev-properties.h"
> > +#include "qapi/error.h"
> > +#include "qemu/log.h"
> > +#include "qemu/module.h"
> > +#include "qemu/range.h"
> > +#include "qemu/rcu.h"
> > +#include "sysemu/hostmem.h"
> > +#include "hw/cxl/cxl.h"
> > +
> > +typedef struct cxl_type3_dev {
> > +    /* Private */
> > +    PCIDevice parent_obj;
> > +
> > +    /* Properties */
> > +    uint64_t size;
> > +    HostMemoryBackend *hostmem;
> > +
> > +    /* State */
> > +    CXLComponentState cxl_cstate;
> > +    CXLDeviceState cxl_dstate;
> > +} CXLType3Dev;
> > +
> > +#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
> > +
> > +static void build_dvsecs(CXLType3Dev *ct3d)
> > +{
> > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > +    uint8_t *dvsec;
> > +
> > +    dvsec = (uint8_t *)&(struct dvsec_device){
> > +        .cap = 0x1e,
> > +        .ctrl = 0x6,
> > +        .status2 = 0x2,
> > +        .range1_size_hi = 0,
> > +        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
> > +        .range1_base_hi = 0,
> > +        .range1_base_lo = 0,
> > +    };
> > +    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
> > +                               PCIE_CXL_DEVICE_DVSEC,
> > +                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
> > +
> > +    dvsec = (uint8_t *)&(struct dvsec_register_locator){
> > +        .rsvd         = 0,
> > +        .reg0_base_lo = RBI_COMPONENT_REG | COMPONENT_REG_BAR_IDX,
> > +        .reg0_base_hi = 0,
> > +        .reg1_base_lo = RBI_CXL_DEVICE_REG | DEVICE_REG_BAR_IDX,
> > +        .reg1_base_hi = 0,
> > +    };
> > +    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
> > +                               REG_LOC_DVSEC_REVID, dvsec);
> > +}
> > +
> > +static void ct3_instance_init(Object *obj)
> > +{
> > +    /* MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); */
> > +}
> > +
> > +static void ct3_finalize(Object *obj)
> > +{
> > +    CXLType3Dev *ct3d = CT3(obj);
> > +
> > +    g_free(ct3d->cxl_dstate.pmem);
> > +}
> > +
> > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > +{
> > +    MemoryRegionSection mrs;
> > +    MemoryRegion *mr;
> > +    uint64_t offset = 0;
> > +    size_t remaining_size;
> > +
> > +    if (!ct3d->hostmem) {
> > +        error_setg(errp, "memdev property must be set");
> > +        return;
> > +    }
> > +
> > +    /* FIXME: need to check mr is the host bridge's MR */
> > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > +
> > +    /* Create our new subregion */
> > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > +
> > +    /* Find the first free space in the window */
> > +    WITH_RCU_READ_LOCK_GUARD()
> > +    {
> > +        mrs = memory_region_find(mr, offset, 1);
> > +        while (mrs.mr && mrs.mr != mr) {
> > +            offset += memory_region_size(mrs.mr);
> > +            mrs = memory_region_find(mr, offset, 1);
> > +        }
> > +    }
> > +
> > +    remaining_size = memory_region_size(mr) - offset;
> > +    if (remaining_size < ct3d->size) {
> > +        g_free(ct3d->cxl_dstate.pmem);
> > +        error_setg(errp,
> > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > +                   remaining_size, ct3d->size);
> > +    }
> > +
> > +    /* Register our subregion as non-volatile */
> > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > +                           "cxl_type3-memory", ct3d->size, errp);
> this allocates ct3d->size of anon RAM, was this an intention?
> If yes, can you clarify why extra RAM is used instead of using what
> backend provides?

It sounds like I'm doing the wrong thing then. There should be one chunk of
memory which is a subset of the full memory backend object. Could you please
advise on what I should be doing instead? Is add_subregion() sufficient?


> 
> > +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
> 
> > +#ifdef SET_PMEM_PADDR
> > +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > +#endif
> What this hunk is supposed to do, why it's ifdef-ed?

In general, the BIOS or OS should program the physical address of the device via
a piece of hardware called the HDM (host defined memory) decoder. That is hooked
up in the code in a later patch. For debug, I set the address internally in
QEMU, mimicking what might be done by BIOS.

> 
> 
> > +}
> > +
> > +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> > +                                              Error **errp)
> > +{
> > +    CXLType3Dev *ct3d = CT3(md);
> > +
> > +    if (!ct3d->cxl_dstate.pmem) {
> > +        cxl_setup_memory(ct3d, errp);
> > +    }
> > +
> > +    return ct3d->cxl_dstate.pmem;
> > +}
> > +
> > +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > +{
> > +    CXLType3Dev *ct3d = CT3(pci_dev);
> > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > +    ComponentRegisters *regs = &cxl_cstate->crb;
> > +    MemoryRegion *mr = &regs->component_registers;
> > +    uint8_t *pci_conf = pci_dev->config;
> > +
> > +    if (!ct3d->cxl_dstate.pmem) {
> > +        cxl_setup_memory(ct3d, errp);
> > +    }
> > +
> > +    pci_config_set_prog_interface(pci_conf, 0x10);
> > +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> > +
> > +    pcie_endpoint_cap_init(pci_dev, 0x80);
> > +    cxl_cstate->dvsec_offset = 0x100;
> > +
> > +    ct3d->cxl_cstate.pdev = pci_dev;
> > +    build_dvsecs(ct3d);
> > +
> > +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> > +                                      TYPE_CXL_TYPE3_DEV);
> > +
> > +    pci_register_bar(
> > +        pci_dev, COMPONENT_REG_BAR_IDX,
> > +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> > +
> > +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> > +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> > +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> > +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> > +                     &ct3d->cxl_dstate.device_registers);
> > +}
> > +
> > +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> > +{
> > +    CXLType3Dev *ct3d = CT3(md);
> > +
> > +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> > +}
> > +
> > +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> > +{
> > +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> > +}
> > +
> > +static void ct3d_reset(DeviceState *dev)
> > +{
> > +    CXLType3Dev *ct3d = CT3(dev);
> > +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> > +
> > +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> > +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> > +}
> > +
> > +static Property ct3_props[] = {
> > +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> > +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> > +                     HostMemoryBackend *),
> > +    DEFINE_PROP_END_OF_LIST(),
> > +};
> > +
> > +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> > +                                        MemoryDeviceInfo *info)
> > +{
> > +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> > +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> > +    const DeviceState *dev = DEVICE(md);
> > +    CXLType3Dev *ct3d = CT3(md);
> > +
> > +    if (dev->id) {
> > +        di->has_id = true;
> > +        di->id = g_strdup(dev->id);
> > +    }
> > +    di->hotplugged = dev->hotplugged;
> > +    di->hotpluggable = dc->hotpluggable;
> > +    di->addr = cxl_md_get_addr(md);
> > +    di->slot = 0;
> > +    di->node = 0;
> > +    di->size = memory_device_get_region_size(md, NULL);
> > +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> > +
> > +
> > +    info->u.cxl.data = di;
> > +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> > +}
> > +
> > +static void ct3_class_init(ObjectClass *oc, void *data)
> > +{
> > +    DeviceClass *dc = DEVICE_CLASS(oc);
> > +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> > +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> > +
> > +    pc->realize = ct3_realize;
> > +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> > +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> > +    pc->device_id = 0xd93; /* LVF for now */
> > +    pc->revision = 1;
> > +
> > +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> > +    dc->desc = "CXL PMEM Device (Type 3)";
> > +    dc->reset = ct3d_reset;
> > +    device_class_set_props(dc, ct3_props);
> > +
> > +    mdc->get_memory_region = cxl_md_get_memory_region;
> > +    mdc->get_addr = cxl_md_get_addr;
> > +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> > +    mdc->get_plugged_size = memory_device_get_region_size;
> > +    mdc->set_addr = cxl_md_set_addr;
> > +}
> > +
> > +static const TypeInfo ct3d_info = {
> > +    .name = TYPE_CXL_TYPE3_DEV,
> > +    .parent = TYPE_PCI_DEVICE,
> > +    .class_init = ct3_class_init,
> > +    .instance_size = sizeof(CXLType3Dev),
> > +    .instance_init = ct3_instance_init,
> > +    .instance_finalize = ct3_finalize,
> > +    .interfaces = (InterfaceInfo[]) {
> > +        { TYPE_MEMORY_DEVICE },
> > +        { INTERFACE_CXL_DEVICE },
> > +        { INTERFACE_PCIE_DEVICE },
> > +        {}
> > +    },
> > +};
> > +
> > +static void ct3d_registers(void)
> > +{
> > +    type_register_static(&ct3d_info);
> > +}
> > +
> > +type_init(ct3d_registers);
> > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > index 0d22f2b572..d13c3ed117 100644
> > --- a/hw/mem/meson.build
> > +++ b/hw/mem/meson.build
> > @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
> >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
> >  
> >  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> > diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> > index d4010cf8f3..1ecf6f6a55 100644
> > --- a/hw/pci/pcie.c
> > +++ b/hw/pci/pcie.c
> > @@ -20,6 +20,7 @@
> >  
> >  #include "qemu/osdep.h"
> >  #include "qapi/error.h"
> > +#include "hw/mem/memory-device.h"
> >  #include "hw/pci/pci_bridge.h"
> >  #include "hw/pci/pcie.h"
> >  #include "hw/pci/msix.h"
> > @@ -27,6 +28,8 @@
> >  #include "hw/pci/pci_bus.h"
> >  #include "hw/pci/pcie_regs.h"
> >  #include "hw/pci/pcie_port.h"
> > +#include "hw/cxl/cxl.h"
> > +#include "hw/boards.h"
> >  #include "qemu/range.h"
> >  
> >  //#define DEBUG_PCIE
> > @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> >      }
> >  
> >      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> > +
> > +#ifdef CXL_MEM_DEVICE
> > +    /*
> > +     * FIXME:
> > +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > +     *    HotplugHandler *hotplug_ctrl;
> > +     *   Error *local_err = NULL;
> > +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> > +     *  if (hotplug_ctrl) {
> > +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> > +     *      if (local_err) {
> > +     *          error_propagate(errp, local_err);
> > +     *          return;
> > +     *      }
> > +     *  }
> > +     */
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> > +                               NULL, errp);
> > +    }
> 
> why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?
> 

A CXL memory device adds memory to a system that is physically addressable in
the same way an NVDIMM device would be.

> > +#endif
> >  }
> >  
> >  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> >          pcie_cap_slot_event(hotplug_pdev,
> >                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
> >      }
> > +
> > +#ifdef CXL_MEM_DEVICE
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> > +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> > +#endif
> >  }
> >  
> >  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > index b1e5f4a8fa..809ed7de60 100644
> > --- a/include/hw/cxl/cxl.h
> > +++ b/include/hw/cxl/cxl.h
> > @@ -17,6 +17,8 @@
> >  #define COMPONENT_REG_BAR_IDX 0
> >  #define DEVICE_REG_BAR_IDX 2
> >  
> > +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> > +
> >  #define CXL_HOST_BASE 0xD0000000
> >  #define CXL_WINDOW_MAX 10
> >  
> > diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> > index a53c2e5ae7..9ec28c9feb 100644
> > --- a/include/hw/cxl/cxl_pci.h
> > +++ b/include/hw/cxl/cxl_pci.h
> > @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
> >   * CXL 2.0 Downstream Port: 3, 4, 7, 8
> >   */
> >  
> > +/* CXL 2.0 - 8.1.3 (ID 0001) */
> > +struct dvsec_device {
> > +    struct dvsec_header hdr;
> > +    uint16_t cap;
> > +    uint16_t ctrl;
> > +    uint16_t status;
> > +    uint16_t ctrl2;
> > +    uint16_t status2;
> > +    uint16_t lock;
> > +    uint16_t cap2;
> > +    uint32_t range1_size_hi;
> > +    uint32_t range1_size_lo;
> > +    uint32_t range1_base_hi;
> > +    uint32_t range1_base_lo;
> > +    uint32_t range2_size_hi;
> > +    uint32_t range2_size_lo;
> > +    uint32_t range2_base_hi;
> > +    uint32_t range2_base_lo;
> > +};
> > +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> > +               "dvsec device size incorrect");
> > +
> >  /* CXL 2.0 - 8.1.5 (ID 0003) */
> >  struct extensions_dvsec_port {
> >      struct dvsec_header hdr;
> > diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> > index 11f8ab7149..76bf3ed590 100644
> > --- a/include/hw/pci/pci_ids.h
> > +++ b/include/hw/pci/pci_ids.h
> > @@ -53,6 +53,7 @@
> >  #define PCI_BASE_CLASS_MEMORY            0x05
> >  #define PCI_CLASS_MEMORY_RAM             0x0500
> >  #define PCI_CLASS_MEMORY_FLASH           0x0501
> > +#define PCI_CLASS_MEMORY_CXL             0x0502
> >  #define PCI_CLASS_MEMORY_OTHER           0x0580
> >  
> >  #define PCI_BASE_CLASS_BRIDGE            0x06
> > diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> > index 0dd594f92b..0f67bc61ce 100644
> > --- a/monitor/hmp-cmds.c
> > +++ b/monitor/hmp-cmds.c
> > @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
> >                  monitor_printf(mon, "  hotpluggable: %s\n",
> >                                 di->hotpluggable ? "true" : "false");
> >                  break;
> > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > +                di = value->u.cxl.data;
> > +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > +                               MemoryDeviceInfoKind_str(value->type),
> > +                               di->id ? di->id : "");
> > +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> > +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> > +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> > +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> > +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> > +                monitor_printf(mon, "  hotplugged: %s\n",
> > +                               di->hotplugged ? "true" : "false");
> > +                monitor_printf(mon, "  hotpluggable: %s\n",
> > +                               di->hotpluggable ? "true" : "false");
> > +                break;
> >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> >                  vpi = value->u.virtio_pmem.data;
> >                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > diff --git a/qapi/machine.json b/qapi/machine.json
> > index 330189efe3..aa96d662bd 100644
> > --- a/qapi/machine.json
> > +++ b/qapi/machine.json
> > @@ -1394,6 +1394,7 @@
> >  { 'union': 'MemoryDeviceInfo',
> >    'data': { 'dimm': 'PCDIMMDeviceInfo',
> >              'nvdimm': 'PCDIMMDeviceInfo',
> > +            'cxl': 'PCDIMMDeviceInfo',
> >              'virtio-pmem': 'VirtioPMEMDeviceInfo',
> >              'virtio-mem': 'VirtioMEMDeviceInfo'
> >            }
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-27 21:21       ` Igor Mammedov
@ 2021-01-27 21:30         ` Ben Widawsky
  0 siblings, 0 replies; 57+ messages in thread
From: Ben Widawsky @ 2021-01-27 21:30 UTC (permalink / raw)
  To: Igor Mammedov
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Jonathan Cameron, Prashant V Agarwal, Dan Williams

On 21-01-27 22:21:04, Igor Mammedov wrote:
> On Wed, 27 Jan 2021 13:11:16 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > On 21-01-27 22:03:12, Igor Mammedov wrote:
> > > On Tue,  5 Jan 2021 08:53:15 -0800
> > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > >   
> > > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > > combination of volatile and persistent memory. It also implements the
> > > > previously defined mailbox interface as well as the memory device
> > > > firmware interface.
> > > > 
> > > > The following example will create a 256M device in a 512M window:
> > > > 
> > > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > > 
> > > I'd expect whole backend used by frontend, so one would not need "size" property
> > > on frontend (like we do with memory devices).
> > > So question is why it partially uses memdev?  
> > 
> > A CXL memory device may participate in an interleave set. In such a case, it
> > would be < the total size of the memory window.
> > 
> > This isn't implemented in the code yet, but it is planned.
> 
> could you add here how it supposed to look like CLI interface wise? 
> 
> also see other questions below.
> 

My mistake on the other questions. I forked another thread for those.

Interleave is still being fleshed out. But generally to set up a 512M address
range interleaves across 2 devices, each 256M, and each connected to a root port
on the host bridge:

# Memory backend
-object memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M

# Host Bridge
-device pxb-cxl id=cxl.0,bus=pcie.0,bus_nr=52,uid=0 len-window-base=1,window-base[0]=0x4c0000000 memdev[0]=cxl-mem1

# 2 root ports
-device cxl-rp,id=rp0,bus=cxl.0,addr=0.0,chassis=0,slot=0,memdev=cxl-mem1
-device cxl-rp,id=rp1,bus=cxl.0,addr=0.1,chassis=0,slot=1,memdev=cxl-mem1

# 2 PMEM devices
-device cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M
-device cxl-type3,bus=rp1,memdev=cxl-mem1,id=cxl-pmem1,size=256M


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-27 21:26     ` Ben Widawsky
@ 2021-01-28 10:25       ` Jonathan Cameron
  2021-01-28 15:03         ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-28 10:25 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	qemu-devel, Philippe Mathieu-Daudé,
	Prashant V Agarwal, Igor Mammedov, Dan Williams

On Wed, 27 Jan 2021 13:26:45 -0800
Ben Widawsky <ben.widawsky@intel.com> wrote:

> On 21-01-27 22:03:12, Igor Mammedov wrote:
> > On Tue,  5 Jan 2021 08:53:15 -0800
> > Ben Widawsky <ben.widawsky@intel.com> wrote:
> >   
> > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > combination of volatile and persistent memory. It also implements the
> > > previously defined mailbox interface as well as the memory device
> > > firmware interface.
> > > 
> > > The following example will create a 256M device in a 512M window:
> > > 
> > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > 
> > I'd expect whole backend used by frontend, so one would not need "size" property
> > on frontend (like we do with memory devices).
> > So question is why it partially uses memdev?  
> 
> Answered in a separate thread...

One possible suggestion inline.

> > > +
> > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > +{
> > > +    MemoryRegionSection mrs;
> > > +    MemoryRegion *mr;
> > > +    uint64_t offset = 0;
> > > +    size_t remaining_size;
> > > +
> > > +    if (!ct3d->hostmem) {
> > > +        error_setg(errp, "memdev property must be set");
> > > +        return;
> > > +    }
> > > +
> > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > +
> > > +    /* Create our new subregion */
> > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > +
> > > +    /* Find the first free space in the window */
> > > +    WITH_RCU_READ_LOCK_GUARD()
> > > +    {
> > > +        mrs = memory_region_find(mr, offset, 1);
> > > +        while (mrs.mr && mrs.mr != mr) {
> > > +            offset += memory_region_size(mrs.mr);
> > > +            mrs = memory_region_find(mr, offset, 1);
> > > +        }
> > > +    }
> > > +
> > > +    remaining_size = memory_region_size(mr) - offset;
> > > +    if (remaining_size < ct3d->size) {
> > > +        g_free(ct3d->cxl_dstate.pmem);
> > > +        error_setg(errp,
> > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > +                   remaining_size, ct3d->size);
> > > +    }
> > > +
> > > +    /* Register our subregion as non-volatile */
> > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > +                           "cxl_type3-memory", ct3d->size, errp);  
> > this allocates ct3d->size of anon RAM, was this an intention?
> > If yes, can you clarify why extra RAM is used instead of using what
> > backend provides?  
> 
> It sounds like I'm doing the wrong thing then. There should be one chunk of
> memory which is a subset of the full memory backend object. Could you please
> advise on what I should be doing instead? Is add_subregion() sufficient?

Taking inspiration from nvdimm I'm carrying a patch that uses
memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(qct3d)q,
			 "cxl_type3-memory", mr, offset, ct3d->size);

I 'think' that's doing the right thing, but haven't fully tested it yet
so may be completely wrong :)

Then for the pmem addr, call memory_region_set_address() to put it
in a particular location.

> 
> 
> >   
> > > +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);  
> >   
> > > +#ifdef SET_PMEM_PADDR
> > > +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > > +#endif  
> > What this hunk is supposed to do, why it's ifdef-ed?  
> 
> In general, the BIOS or OS should program the physical address of the device via
> a piece of hardware called the HDM (host defined memory) decoder. That is hooked
> up in the code in a later patch. For debug, I set the address internally in
> QEMU, mimicking what might be done by BIOS.
> 
> > 
> >   
> > > +}
> > > +
> > > +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> > > +                                              Error **errp)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(md);
> > > +
> > > +    if (!ct3d->cxl_dstate.pmem) {
> > > +        cxl_setup_memory(ct3d, errp);
> > > +    }
> > > +
> > > +    return ct3d->cxl_dstate.pmem;
> > > +}
> > > +
> > > +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(pci_dev);
> > > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > > +    ComponentRegisters *regs = &cxl_cstate->crb;
> > > +    MemoryRegion *mr = &regs->component_registers;
> > > +    uint8_t *pci_conf = pci_dev->config;
> > > +
> > > +    if (!ct3d->cxl_dstate.pmem) {
> > > +        cxl_setup_memory(ct3d, errp);
> > > +    }
> > > +
> > > +    pci_config_set_prog_interface(pci_conf, 0x10);
> > > +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> > > +
> > > +    pcie_endpoint_cap_init(pci_dev, 0x80);
> > > +    cxl_cstate->dvsec_offset = 0x100;
> > > +
> > > +    ct3d->cxl_cstate.pdev = pci_dev;
> > > +    build_dvsecs(ct3d);
> > > +
> > > +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> > > +                                      TYPE_CXL_TYPE3_DEV);
> > > +
> > > +    pci_register_bar(
> > > +        pci_dev, COMPONENT_REG_BAR_IDX,
> > > +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> > > +
> > > +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> > > +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> > > +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> > > +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> > > +                     &ct3d->cxl_dstate.device_registers);
> > > +}
> > > +
> > > +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(md);
> > > +
> > > +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> > > +}
> > > +
> > > +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> > > +{
> > > +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> > > +}
> > > +
> > > +static void ct3d_reset(DeviceState *dev)
> > > +{
> > > +    CXLType3Dev *ct3d = CT3(dev);
> > > +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> > > +
> > > +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> > > +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> > > +}
> > > +
> > > +static Property ct3_props[] = {
> > > +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> > > +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> > > +                     HostMemoryBackend *),
> > > +    DEFINE_PROP_END_OF_LIST(),
> > > +};
> > > +
> > > +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> > > +                                        MemoryDeviceInfo *info)
> > > +{
> > > +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> > > +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> > > +    const DeviceState *dev = DEVICE(md);
> > > +    CXLType3Dev *ct3d = CT3(md);
> > > +
> > > +    if (dev->id) {
> > > +        di->has_id = true;
> > > +        di->id = g_strdup(dev->id);
> > > +    }
> > > +    di->hotplugged = dev->hotplugged;
> > > +    di->hotpluggable = dc->hotpluggable;
> > > +    di->addr = cxl_md_get_addr(md);
> > > +    di->slot = 0;
> > > +    di->node = 0;
> > > +    di->size = memory_device_get_region_size(md, NULL);
> > > +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> > > +
> > > +
> > > +    info->u.cxl.data = di;
> > > +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> > > +}
> > > +
> > > +static void ct3_class_init(ObjectClass *oc, void *data)
> > > +{
> > > +    DeviceClass *dc = DEVICE_CLASS(oc);
> > > +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> > > +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> > > +
> > > +    pc->realize = ct3_realize;
> > > +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> > > +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> > > +    pc->device_id = 0xd93; /* LVF for now */
> > > +    pc->revision = 1;
> > > +
> > > +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> > > +    dc->desc = "CXL PMEM Device (Type 3)";
> > > +    dc->reset = ct3d_reset;
> > > +    device_class_set_props(dc, ct3_props);
> > > +
> > > +    mdc->get_memory_region = cxl_md_get_memory_region;
> > > +    mdc->get_addr = cxl_md_get_addr;
> > > +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> > > +    mdc->get_plugged_size = memory_device_get_region_size;
> > > +    mdc->set_addr = cxl_md_set_addr;
> > > +}
> > > +
> > > +static const TypeInfo ct3d_info = {
> > > +    .name = TYPE_CXL_TYPE3_DEV,
> > > +    .parent = TYPE_PCI_DEVICE,
> > > +    .class_init = ct3_class_init,
> > > +    .instance_size = sizeof(CXLType3Dev),
> > > +    .instance_init = ct3_instance_init,
> > > +    .instance_finalize = ct3_finalize,
> > > +    .interfaces = (InterfaceInfo[]) {
> > > +        { TYPE_MEMORY_DEVICE },
> > > +        { INTERFACE_CXL_DEVICE },
> > > +        { INTERFACE_PCIE_DEVICE },
> > > +        {}
> > > +    },
> > > +};
> > > +
> > > +static void ct3d_registers(void)
> > > +{
> > > +    type_register_static(&ct3d_info);
> > > +}
> > > +
> > > +type_init(ct3d_registers);
> > > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > > index 0d22f2b572..d13c3ed117 100644
> > > --- a/hw/mem/meson.build
> > > +++ b/hw/mem/meson.build
> > > @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
> > >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> > >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> > >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > > +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
> > >  
> > >  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> > > diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> > > index d4010cf8f3..1ecf6f6a55 100644
> > > --- a/hw/pci/pcie.c
> > > +++ b/hw/pci/pcie.c
> > > @@ -20,6 +20,7 @@
> > >  
> > >  #include "qemu/osdep.h"
> > >  #include "qapi/error.h"
> > > +#include "hw/mem/memory-device.h"
> > >  #include "hw/pci/pci_bridge.h"
> > >  #include "hw/pci/pcie.h"
> > >  #include "hw/pci/msix.h"
> > > @@ -27,6 +28,8 @@
> > >  #include "hw/pci/pci_bus.h"
> > >  #include "hw/pci/pcie_regs.h"
> > >  #include "hw/pci/pcie_port.h"
> > > +#include "hw/cxl/cxl.h"
> > > +#include "hw/boards.h"
> > >  #include "qemu/range.h"
> > >  
> > >  //#define DEBUG_PCIE
> > > @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > >      }
> > >  
> > >      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> > > +
> > > +#ifdef CXL_MEM_DEVICE
> > > +    /*
> > > +     * FIXME:
> > > +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > +     *    HotplugHandler *hotplug_ctrl;
> > > +     *   Error *local_err = NULL;
> > > +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> > > +     *  if (hotplug_ctrl) {
> > > +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> > > +     *      if (local_err) {
> > > +     *          error_propagate(errp, local_err);
> > > +     *          return;
> > > +     *      }
> > > +     *  }
> > > +     */
> > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> > > +                               NULL, errp);
> > > +    }  
> > 
> > why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?
> >   
> 
> A CXL memory device adds memory to a system that is physically addressable in
> the same way an NVDIMM device would be.
> 
> > > +#endif
> > >  }
> > >  
> > >  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > >          pcie_cap_slot_event(hotplug_pdev,
> > >                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
> > >      }
> > > +
> > > +#ifdef CXL_MEM_DEVICE
> > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> > > +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> > > +#endif
> > >  }
> > >  
> > >  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > > index b1e5f4a8fa..809ed7de60 100644
> > > --- a/include/hw/cxl/cxl.h
> > > +++ b/include/hw/cxl/cxl.h
> > > @@ -17,6 +17,8 @@
> > >  #define COMPONENT_REG_BAR_IDX 0
> > >  #define DEVICE_REG_BAR_IDX 2
> > >  
> > > +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> > > +
> > >  #define CXL_HOST_BASE 0xD0000000
> > >  #define CXL_WINDOW_MAX 10
> > >  
> > > diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> > > index a53c2e5ae7..9ec28c9feb 100644
> > > --- a/include/hw/cxl/cxl_pci.h
> > > +++ b/include/hw/cxl/cxl_pci.h
> > > @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
> > >   * CXL 2.0 Downstream Port: 3, 4, 7, 8
> > >   */
> > >  
> > > +/* CXL 2.0 - 8.1.3 (ID 0001) */
> > > +struct dvsec_device {
> > > +    struct dvsec_header hdr;
> > > +    uint16_t cap;
> > > +    uint16_t ctrl;
> > > +    uint16_t status;
> > > +    uint16_t ctrl2;
> > > +    uint16_t status2;
> > > +    uint16_t lock;
> > > +    uint16_t cap2;
> > > +    uint32_t range1_size_hi;
> > > +    uint32_t range1_size_lo;
> > > +    uint32_t range1_base_hi;
> > > +    uint32_t range1_base_lo;
> > > +    uint32_t range2_size_hi;
> > > +    uint32_t range2_size_lo;
> > > +    uint32_t range2_base_hi;
> > > +    uint32_t range2_base_lo;
> > > +};
> > > +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> > > +               "dvsec device size incorrect");
> > > +
> > >  /* CXL 2.0 - 8.1.5 (ID 0003) */
> > >  struct extensions_dvsec_port {
> > >      struct dvsec_header hdr;
> > > diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> > > index 11f8ab7149..76bf3ed590 100644
> > > --- a/include/hw/pci/pci_ids.h
> > > +++ b/include/hw/pci/pci_ids.h
> > > @@ -53,6 +53,7 @@
> > >  #define PCI_BASE_CLASS_MEMORY            0x05
> > >  #define PCI_CLASS_MEMORY_RAM             0x0500
> > >  #define PCI_CLASS_MEMORY_FLASH           0x0501
> > > +#define PCI_CLASS_MEMORY_CXL             0x0502
> > >  #define PCI_CLASS_MEMORY_OTHER           0x0580
> > >  
> > >  #define PCI_BASE_CLASS_BRIDGE            0x06
> > > diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> > > index 0dd594f92b..0f67bc61ce 100644
> > > --- a/monitor/hmp-cmds.c
> > > +++ b/monitor/hmp-cmds.c
> > > @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
> > >                  monitor_printf(mon, "  hotpluggable: %s\n",
> > >                                 di->hotpluggable ? "true" : "false");
> > >                  break;
> > > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > > +                di = value->u.cxl.data;
> > > +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > +                               MemoryDeviceInfoKind_str(value->type),
> > > +                               di->id ? di->id : "");
> > > +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> > > +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> > > +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> > > +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> > > +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> > > +                monitor_printf(mon, "  hotplugged: %s\n",
> > > +                               di->hotplugged ? "true" : "false");
> > > +                monitor_printf(mon, "  hotpluggable: %s\n",
> > > +                               di->hotpluggable ? "true" : "false");
> > > +                break;
> > >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> > >                  vpi = value->u.virtio_pmem.data;
> > >                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > diff --git a/qapi/machine.json b/qapi/machine.json
> > > index 330189efe3..aa96d662bd 100644
> > > --- a/qapi/machine.json
> > > +++ b/qapi/machine.json
> > > @@ -1394,6 +1394,7 @@
> > >  { 'union': 'MemoryDeviceInfo',
> > >    'data': { 'dimm': 'PCDIMMDeviceInfo',
> > >              'nvdimm': 'PCDIMMDeviceInfo',
> > > +            'cxl': 'PCDIMMDeviceInfo',
> > >              'virtio-pmem': 'VirtioPMEMDeviceInfo',
> > >              'virtio-mem': 'VirtioMEMDeviceInfo'
> > >            }  
> >   



^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-28 10:25       ` Jonathan Cameron
@ 2021-01-28 15:03         ` Ben Widawsky
  2021-01-28 15:14           ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-28 15:03 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Igor Mammedov, Dan Williams

On 21-01-28 10:25:38, Jonathan Cameron wrote:
> On Wed, 27 Jan 2021 13:26:45 -0800
> Ben Widawsky <ben.widawsky@intel.com> wrote:
> 
> > On 21-01-27 22:03:12, Igor Mammedov wrote:
> > > On Tue,  5 Jan 2021 08:53:15 -0800
> > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > >   
> > > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > > combination of volatile and persistent memory. It also implements the
> > > > previously defined mailbox interface as well as the memory device
> > > > firmware interface.
> > > > 
> > > > The following example will create a 256M device in a 512M window:
> > > > 
> > > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > > 
> > > I'd expect whole backend used by frontend, so one would not need "size" property
> > > on frontend (like we do with memory devices).
> > > So question is why it partially uses memdev?  
> > 
> > Answered in a separate thread...
> 
> One possible suggestion inline.
> 
> > > > +
> > > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > > +{
> > > > +    MemoryRegionSection mrs;
> > > > +    MemoryRegion *mr;
> > > > +    uint64_t offset = 0;
> > > > +    size_t remaining_size;
> > > > +
> > > > +    if (!ct3d->hostmem) {
> > > > +        error_setg(errp, "memdev property must be set");
> > > > +        return;
> > > > +    }
> > > > +
> > > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > > +
> > > > +    /* Create our new subregion */
> > > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > > +
> > > > +    /* Find the first free space in the window */
> > > > +    WITH_RCU_READ_LOCK_GUARD()
> > > > +    {
> > > > +        mrs = memory_region_find(mr, offset, 1);
> > > > +        while (mrs.mr && mrs.mr != mr) {
> > > > +            offset += memory_region_size(mrs.mr);
> > > > +            mrs = memory_region_find(mr, offset, 1);
> > > > +        }
> > > > +    }
> > > > +
> > > > +    remaining_size = memory_region_size(mr) - offset;
> > > > +    if (remaining_size < ct3d->size) {
> > > > +        g_free(ct3d->cxl_dstate.pmem);
> > > > +        error_setg(errp,
> > > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > > +                   remaining_size, ct3d->size);
> > > > +    }
> > > > +
> > > > +    /* Register our subregion as non-volatile */
> > > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > > +                           "cxl_type3-memory", ct3d->size, errp);  
> > > this allocates ct3d->size of anon RAM, was this an intention?
> > > If yes, can you clarify why extra RAM is used instead of using what
> > > backend provides?  
> > 
> > It sounds like I'm doing the wrong thing then. There should be one chunk of
> > memory which is a subset of the full memory backend object. Could you please
> > advise on what I should be doing instead? Is add_subregion() sufficient?
> 
> Taking inspiration from nvdimm I'm carrying a patch that uses
> memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(qct3d)q,
> 			 "cxl_type3-memory", mr, offset, ct3d->size);
> 
> I 'think' that's doing the right thing, but haven't fully tested it yet
> so may be completely wrong :)
> 
> Then for the pmem addr, call memory_region_set_address() to put it
> in a particular location.
> 

Yes - this is what I'd like to do and what I initially tried, and I also believe
it's right, but it doesn't work.

range_invariant: Assertion `range->lob <= range->upb || range->lob == range->upb + 1' failed.

I was digging into this yesterday, but opted to start a new thread on the
matter.

> > 
> > 
> > >   
> > > > +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);  
> > >   
> > > > +#ifdef SET_PMEM_PADDR
> > > > +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > > > +#endif  
> > > What this hunk is supposed to do, why it's ifdef-ed?  
> > 
> > In general, the BIOS or OS should program the physical address of the device via
> > a piece of hardware called the HDM (host defined memory) decoder. That is hooked
> > up in the code in a later patch. For debug, I set the address internally in
> > QEMU, mimicking what might be done by BIOS.
> > 
> > > 
> > >   
> > > > +}
> > > > +
> > > > +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> > > > +                                              Error **errp)
> > > > +{
> > > > +    CXLType3Dev *ct3d = CT3(md);
> > > > +
> > > > +    if (!ct3d->cxl_dstate.pmem) {
> > > > +        cxl_setup_memory(ct3d, errp);
> > > > +    }
> > > > +
> > > > +    return ct3d->cxl_dstate.pmem;
> > > > +}
> > > > +
> > > > +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > > > +{
> > > > +    CXLType3Dev *ct3d = CT3(pci_dev);
> > > > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > > > +    ComponentRegisters *regs = &cxl_cstate->crb;
> > > > +    MemoryRegion *mr = &regs->component_registers;
> > > > +    uint8_t *pci_conf = pci_dev->config;
> > > > +
> > > > +    if (!ct3d->cxl_dstate.pmem) {
> > > > +        cxl_setup_memory(ct3d, errp);
> > > > +    }
> > > > +
> > > > +    pci_config_set_prog_interface(pci_conf, 0x10);
> > > > +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> > > > +
> > > > +    pcie_endpoint_cap_init(pci_dev, 0x80);
> > > > +    cxl_cstate->dvsec_offset = 0x100;
> > > > +
> > > > +    ct3d->cxl_cstate.pdev = pci_dev;
> > > > +    build_dvsecs(ct3d);
> > > > +
> > > > +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> > > > +                                      TYPE_CXL_TYPE3_DEV);
> > > > +
> > > > +    pci_register_bar(
> > > > +        pci_dev, COMPONENT_REG_BAR_IDX,
> > > > +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> > > > +
> > > > +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> > > > +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> > > > +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> > > > +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> > > > +                     &ct3d->cxl_dstate.device_registers);
> > > > +}
> > > > +
> > > > +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> > > > +{
> > > > +    CXLType3Dev *ct3d = CT3(md);
> > > > +
> > > > +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> > > > +}
> > > > +
> > > > +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> > > > +{
> > > > +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> > > > +}
> > > > +
> > > > +static void ct3d_reset(DeviceState *dev)
> > > > +{
> > > > +    CXLType3Dev *ct3d = CT3(dev);
> > > > +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> > > > +
> > > > +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> > > > +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> > > > +}
> > > > +
> > > > +static Property ct3_props[] = {
> > > > +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> > > > +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> > > > +                     HostMemoryBackend *),
> > > > +    DEFINE_PROP_END_OF_LIST(),
> > > > +};
> > > > +
> > > > +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> > > > +                                        MemoryDeviceInfo *info)
> > > > +{
> > > > +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> > > > +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> > > > +    const DeviceState *dev = DEVICE(md);
> > > > +    CXLType3Dev *ct3d = CT3(md);
> > > > +
> > > > +    if (dev->id) {
> > > > +        di->has_id = true;
> > > > +        di->id = g_strdup(dev->id);
> > > > +    }
> > > > +    di->hotplugged = dev->hotplugged;
> > > > +    di->hotpluggable = dc->hotpluggable;
> > > > +    di->addr = cxl_md_get_addr(md);
> > > > +    di->slot = 0;
> > > > +    di->node = 0;
> > > > +    di->size = memory_device_get_region_size(md, NULL);
> > > > +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> > > > +
> > > > +
> > > > +    info->u.cxl.data = di;
> > > > +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> > > > +}
> > > > +
> > > > +static void ct3_class_init(ObjectClass *oc, void *data)
> > > > +{
> > > > +    DeviceClass *dc = DEVICE_CLASS(oc);
> > > > +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> > > > +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> > > > +
> > > > +    pc->realize = ct3_realize;
> > > > +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> > > > +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> > > > +    pc->device_id = 0xd93; /* LVF for now */
> > > > +    pc->revision = 1;
> > > > +
> > > > +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> > > > +    dc->desc = "CXL PMEM Device (Type 3)";
> > > > +    dc->reset = ct3d_reset;
> > > > +    device_class_set_props(dc, ct3_props);
> > > > +
> > > > +    mdc->get_memory_region = cxl_md_get_memory_region;
> > > > +    mdc->get_addr = cxl_md_get_addr;
> > > > +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> > > > +    mdc->get_plugged_size = memory_device_get_region_size;
> > > > +    mdc->set_addr = cxl_md_set_addr;
> > > > +}
> > > > +
> > > > +static const TypeInfo ct3d_info = {
> > > > +    .name = TYPE_CXL_TYPE3_DEV,
> > > > +    .parent = TYPE_PCI_DEVICE,
> > > > +    .class_init = ct3_class_init,
> > > > +    .instance_size = sizeof(CXLType3Dev),
> > > > +    .instance_init = ct3_instance_init,
> > > > +    .instance_finalize = ct3_finalize,
> > > > +    .interfaces = (InterfaceInfo[]) {
> > > > +        { TYPE_MEMORY_DEVICE },
> > > > +        { INTERFACE_CXL_DEVICE },
> > > > +        { INTERFACE_PCIE_DEVICE },
> > > > +        {}
> > > > +    },
> > > > +};
> > > > +
> > > > +static void ct3d_registers(void)
> > > > +{
> > > > +    type_register_static(&ct3d_info);
> > > > +}
> > > > +
> > > > +type_init(ct3d_registers);
> > > > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > > > index 0d22f2b572..d13c3ed117 100644
> > > > --- a/hw/mem/meson.build
> > > > +++ b/hw/mem/meson.build
> > > > @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
> > > >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> > > >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> > > >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > > > +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
> > > >  
> > > >  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> > > > diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> > > > index d4010cf8f3..1ecf6f6a55 100644
> > > > --- a/hw/pci/pcie.c
> > > > +++ b/hw/pci/pcie.c
> > > > @@ -20,6 +20,7 @@
> > > >  
> > > >  #include "qemu/osdep.h"
> > > >  #include "qapi/error.h"
> > > > +#include "hw/mem/memory-device.h"
> > > >  #include "hw/pci/pci_bridge.h"
> > > >  #include "hw/pci/pcie.h"
> > > >  #include "hw/pci/msix.h"
> > > > @@ -27,6 +28,8 @@
> > > >  #include "hw/pci/pci_bus.h"
> > > >  #include "hw/pci/pcie_regs.h"
> > > >  #include "hw/pci/pcie_port.h"
> > > > +#include "hw/cxl/cxl.h"
> > > > +#include "hw/boards.h"
> > > >  #include "qemu/range.h"
> > > >  
> > > >  //#define DEBUG_PCIE
> > > > @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > >      }
> > > >  
> > > >      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> > > > +
> > > > +#ifdef CXL_MEM_DEVICE
> > > > +    /*
> > > > +     * FIXME:
> > > > +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > > +     *    HotplugHandler *hotplug_ctrl;
> > > > +     *   Error *local_err = NULL;
> > > > +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> > > > +     *  if (hotplug_ctrl) {
> > > > +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> > > > +     *      if (local_err) {
> > > > +     *          error_propagate(errp, local_err);
> > > > +     *          return;
> > > > +     *      }
> > > > +     *  }
> > > > +     */
> > > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > > +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> > > > +                               NULL, errp);
> > > > +    }  
> > > 
> > > why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?
> > >   
> > 
> > A CXL memory device adds memory to a system that is physically addressable in
> > the same way an NVDIMM device would be.
> > 
> > > > +#endif
> > > >  }
> > > >  
> > > >  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > > @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > >          pcie_cap_slot_event(hotplug_pdev,
> > > >                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
> > > >      }
> > > > +
> > > > +#ifdef CXL_MEM_DEVICE
> > > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> > > > +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> > > > +#endif
> > > >  }
> > > >  
> > > >  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > > > index b1e5f4a8fa..809ed7de60 100644
> > > > --- a/include/hw/cxl/cxl.h
> > > > +++ b/include/hw/cxl/cxl.h
> > > > @@ -17,6 +17,8 @@
> > > >  #define COMPONENT_REG_BAR_IDX 0
> > > >  #define DEVICE_REG_BAR_IDX 2
> > > >  
> > > > +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> > > > +
> > > >  #define CXL_HOST_BASE 0xD0000000
> > > >  #define CXL_WINDOW_MAX 10
> > > >  
> > > > diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> > > > index a53c2e5ae7..9ec28c9feb 100644
> > > > --- a/include/hw/cxl/cxl_pci.h
> > > > +++ b/include/hw/cxl/cxl_pci.h
> > > > @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
> > > >   * CXL 2.0 Downstream Port: 3, 4, 7, 8
> > > >   */
> > > >  
> > > > +/* CXL 2.0 - 8.1.3 (ID 0001) */
> > > > +struct dvsec_device {
> > > > +    struct dvsec_header hdr;
> > > > +    uint16_t cap;
> > > > +    uint16_t ctrl;
> > > > +    uint16_t status;
> > > > +    uint16_t ctrl2;
> > > > +    uint16_t status2;
> > > > +    uint16_t lock;
> > > > +    uint16_t cap2;
> > > > +    uint32_t range1_size_hi;
> > > > +    uint32_t range1_size_lo;
> > > > +    uint32_t range1_base_hi;
> > > > +    uint32_t range1_base_lo;
> > > > +    uint32_t range2_size_hi;
> > > > +    uint32_t range2_size_lo;
> > > > +    uint32_t range2_base_hi;
> > > > +    uint32_t range2_base_lo;
> > > > +};
> > > > +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> > > > +               "dvsec device size incorrect");
> > > > +
> > > >  /* CXL 2.0 - 8.1.5 (ID 0003) */
> > > >  struct extensions_dvsec_port {
> > > >      struct dvsec_header hdr;
> > > > diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> > > > index 11f8ab7149..76bf3ed590 100644
> > > > --- a/include/hw/pci/pci_ids.h
> > > > +++ b/include/hw/pci/pci_ids.h
> > > > @@ -53,6 +53,7 @@
> > > >  #define PCI_BASE_CLASS_MEMORY            0x05
> > > >  #define PCI_CLASS_MEMORY_RAM             0x0500
> > > >  #define PCI_CLASS_MEMORY_FLASH           0x0501
> > > > +#define PCI_CLASS_MEMORY_CXL             0x0502
> > > >  #define PCI_CLASS_MEMORY_OTHER           0x0580
> > > >  
> > > >  #define PCI_BASE_CLASS_BRIDGE            0x06
> > > > diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> > > > index 0dd594f92b..0f67bc61ce 100644
> > > > --- a/monitor/hmp-cmds.c
> > > > +++ b/monitor/hmp-cmds.c
> > > > @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
> > > >                  monitor_printf(mon, "  hotpluggable: %s\n",
> > > >                                 di->hotpluggable ? "true" : "false");
> > > >                  break;
> > > > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > > > +                di = value->u.cxl.data;
> > > > +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > > +                               MemoryDeviceInfoKind_str(value->type),
> > > > +                               di->id ? di->id : "");
> > > > +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> > > > +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> > > > +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> > > > +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> > > > +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> > > > +                monitor_printf(mon, "  hotplugged: %s\n",
> > > > +                               di->hotplugged ? "true" : "false");
> > > > +                monitor_printf(mon, "  hotpluggable: %s\n",
> > > > +                               di->hotpluggable ? "true" : "false");
> > > > +                break;
> > > >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> > > >                  vpi = value->u.virtio_pmem.data;
> > > >                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > > diff --git a/qapi/machine.json b/qapi/machine.json
> > > > index 330189efe3..aa96d662bd 100644
> > > > --- a/qapi/machine.json
> > > > +++ b/qapi/machine.json
> > > > @@ -1394,6 +1394,7 @@
> > > >  { 'union': 'MemoryDeviceInfo',
> > > >    'data': { 'dimm': 'PCDIMMDeviceInfo',
> > > >              'nvdimm': 'PCDIMMDeviceInfo',
> > > > +            'cxl': 'PCDIMMDeviceInfo',
> > > >              'virtio-pmem': 'VirtioPMEMDeviceInfo',
> > > >              'virtio-mem': 'VirtioMEMDeviceInfo'
> > > >            }  
> > >   
> 
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-28 15:03         ` Ben Widawsky
@ 2021-01-28 15:14           ` Ben Widawsky
  2021-01-28 16:51             ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-28 15:14 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	qemu-devel, Philippe Mathieu-Daudé,
	Prashant V Agarwal, Igor Mammedov, Dan Williams

On 21-01-28 07:03:18, Ben Widawsky wrote:
> On 21-01-28 10:25:38, Jonathan Cameron wrote:
> > On Wed, 27 Jan 2021 13:26:45 -0800
> > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > 
> > > On 21-01-27 22:03:12, Igor Mammedov wrote:
> > > > On Tue,  5 Jan 2021 08:53:15 -0800
> > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > >   
> > > > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > > > combination of volatile and persistent memory. It also implements the
> > > > > previously defined mailbox interface as well as the memory device
> > > > > firmware interface.
> > > > > 
> > > > > The following example will create a 256M device in a 512M window:
> > > > > 
> > > > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > > > 
> > > > I'd expect whole backend used by frontend, so one would not need "size" property
> > > > on frontend (like we do with memory devices).
> > > > So question is why it partially uses memdev?  
> > > 
> > > Answered in a separate thread...
> > 
> > One possible suggestion inline.
> > 
> > > > > +
> > > > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > > > +{
> > > > > +    MemoryRegionSection mrs;
> > > > > +    MemoryRegion *mr;
> > > > > +    uint64_t offset = 0;
> > > > > +    size_t remaining_size;
> > > > > +
> > > > > +    if (!ct3d->hostmem) {
> > > > > +        error_setg(errp, "memdev property must be set");
> > > > > +        return;
> > > > > +    }
> > > > > +
> > > > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > > > +
> > > > > +    /* Create our new subregion */
> > > > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > > > +
> > > > > +    /* Find the first free space in the window */
> > > > > +    WITH_RCU_READ_LOCK_GUARD()
> > > > > +    {
> > > > > +        mrs = memory_region_find(mr, offset, 1);
> > > > > +        while (mrs.mr && mrs.mr != mr) {
> > > > > +            offset += memory_region_size(mrs.mr);
> > > > > +            mrs = memory_region_find(mr, offset, 1);
> > > > > +        }
> > > > > +    }
> > > > > +
> > > > > +    remaining_size = memory_region_size(mr) - offset;
> > > > > +    if (remaining_size < ct3d->size) {
> > > > > +        g_free(ct3d->cxl_dstate.pmem);
> > > > > +        error_setg(errp,
> > > > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > > > +                   remaining_size, ct3d->size);
> > > > > +    }
> > > > > +
> > > > > +    /* Register our subregion as non-volatile */
> > > > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > > > +                           "cxl_type3-memory", ct3d->size, errp);  
> > > > this allocates ct3d->size of anon RAM, was this an intention?
> > > > If yes, can you clarify why extra RAM is used instead of using what
> > > > backend provides?  
> > > 
> > > It sounds like I'm doing the wrong thing then. There should be one chunk of
> > > memory which is a subset of the full memory backend object. Could you please
> > > advise on what I should be doing instead? Is add_subregion() sufficient?
> > 
> > Taking inspiration from nvdimm I'm carrying a patch that uses
> > memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(qct3d)q,
> > 			 "cxl_type3-memory", mr, offset, ct3d->size);
> > 
> > I 'think' that's doing the right thing, but haven't fully tested it yet
> > so may be completely wrong :)
> > 
> > Then for the pmem addr, call memory_region_set_address() to put it
> > in a particular location.
> > 
> 
> Yes - this is what I'd like to do and what I initially tried, and I also believe
> it's right, but it doesn't work.
> 
> range_invariant: Assertion `range->lob <= range->upb || range->lob == range->upb + 1' failed.
> 
> I was digging into this yesterday, but opted to start a new thread on the
> matter.
> 

Hmm. I think I need to figure out the right add_subregion after this and it
might work. I'll keep digging, but if you have ideas, let me know.

> > > 
> > > 
> > > >   
> > > > > +    memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);  
> > > >   
> > > > > +#ifdef SET_PMEM_PADDR
> > > > > +    memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > > > > +#endif  
> > > > What this hunk is supposed to do, why it's ifdef-ed?  
> > > 
> > > In general, the BIOS or OS should program the physical address of the device via
> > > a piece of hardware called the HDM (host defined memory) decoder. That is hooked
> > > up in the code in a later patch. For debug, I set the address internally in
> > > QEMU, mimicking what might be done by BIOS.
> > > 
> > > > 
> > > >   
> > > > > +}
> > > > > +
> > > > > +static MemoryRegion *cxl_md_get_memory_region(MemoryDeviceState *md,
> > > > > +                                              Error **errp)
> > > > > +{
> > > > > +    CXLType3Dev *ct3d = CT3(md);
> > > > > +
> > > > > +    if (!ct3d->cxl_dstate.pmem) {
> > > > > +        cxl_setup_memory(ct3d, errp);
> > > > > +    }
> > > > > +
> > > > > +    return ct3d->cxl_dstate.pmem;
> > > > > +}
> > > > > +
> > > > > +static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > > > > +{
> > > > > +    CXLType3Dev *ct3d = CT3(pci_dev);
> > > > > +    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
> > > > > +    ComponentRegisters *regs = &cxl_cstate->crb;
> > > > > +    MemoryRegion *mr = &regs->component_registers;
> > > > > +    uint8_t *pci_conf = pci_dev->config;
> > > > > +
> > > > > +    if (!ct3d->cxl_dstate.pmem) {
> > > > > +        cxl_setup_memory(ct3d, errp);
> > > > > +    }
> > > > > +
> > > > > +    pci_config_set_prog_interface(pci_conf, 0x10);
> > > > > +    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
> > > > > +
> > > > > +    pcie_endpoint_cap_init(pci_dev, 0x80);
> > > > > +    cxl_cstate->dvsec_offset = 0x100;
> > > > > +
> > > > > +    ct3d->cxl_cstate.pdev = pci_dev;
> > > > > +    build_dvsecs(ct3d);
> > > > > +
> > > > > +    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
> > > > > +                                      TYPE_CXL_TYPE3_DEV);
> > > > > +
> > > > > +    pci_register_bar(
> > > > > +        pci_dev, COMPONENT_REG_BAR_IDX,
> > > > > +        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
> > > > > +
> > > > > +    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
> > > > > +    pci_register_bar(pci_dev, DEVICE_REG_BAR_IDX,
> > > > > +                     PCI_BASE_ADDRESS_SPACE_MEMORY |
> > > > > +                         PCI_BASE_ADDRESS_MEM_TYPE_64,
> > > > > +                     &ct3d->cxl_dstate.device_registers);
> > > > > +}
> > > > > +
> > > > > +static uint64_t cxl_md_get_addr(const MemoryDeviceState *md)
> > > > > +{
> > > > > +    CXLType3Dev *ct3d = CT3(md);
> > > > > +
> > > > > +    return memory_region_get_ram_addr(ct3d->cxl_dstate.pmem);
> > > > > +}
> > > > > +
> > > > > +static void cxl_md_set_addr(MemoryDeviceState *md, uint64_t addr, Error **errp)
> > > > > +{
> > > > > +    object_property_set_uint(OBJECT(md), "paddr", addr, errp);
> > > > > +}
> > > > > +
> > > > > +static void ct3d_reset(DeviceState *dev)
> > > > > +{
> > > > > +    CXLType3Dev *ct3d = CT3(dev);
> > > > > +    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
> > > > > +
> > > > > +    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
> > > > > +    cxl_device_register_init_common(&ct3d->cxl_dstate);
> > > > > +}
> > > > > +
> > > > > +static Property ct3_props[] = {
> > > > > +    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
> > > > > +    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
> > > > > +                     HostMemoryBackend *),
> > > > > +    DEFINE_PROP_END_OF_LIST(),
> > > > > +};
> > > > > +
> > > > > +static void pc_dimm_md_fill_device_info(const MemoryDeviceState *md,
> > > > > +                                        MemoryDeviceInfo *info)
> > > > > +{
> > > > > +    PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
> > > > > +    const DeviceClass *dc = DEVICE_GET_CLASS(md);
> > > > > +    const DeviceState *dev = DEVICE(md);
> > > > > +    CXLType3Dev *ct3d = CT3(md);
> > > > > +
> > > > > +    if (dev->id) {
> > > > > +        di->has_id = true;
> > > > > +        di->id = g_strdup(dev->id);
> > > > > +    }
> > > > > +    di->hotplugged = dev->hotplugged;
> > > > > +    di->hotpluggable = dc->hotpluggable;
> > > > > +    di->addr = cxl_md_get_addr(md);
> > > > > +    di->slot = 0;
> > > > > +    di->node = 0;
> > > > > +    di->size = memory_device_get_region_size(md, NULL);
> > > > > +    di->memdev = object_get_canonical_path(OBJECT(ct3d->hostmem));
> > > > > +
> > > > > +
> > > > > +    info->u.cxl.data = di;
> > > > > +    info->type = MEMORY_DEVICE_INFO_KIND_CXL;
> > > > > +}
> > > > > +
> > > > > +static void ct3_class_init(ObjectClass *oc, void *data)
> > > > > +{
> > > > > +    DeviceClass *dc = DEVICE_CLASS(oc);
> > > > > +    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
> > > > > +    MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(oc);
> > > > > +
> > > > > +    pc->realize = ct3_realize;
> > > > > +    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
> > > > > +    pc->vendor_id = PCI_VENDOR_ID_INTEL;
> > > > > +    pc->device_id = 0xd93; /* LVF for now */
> > > > > +    pc->revision = 1;
> > > > > +
> > > > > +    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> > > > > +    dc->desc = "CXL PMEM Device (Type 3)";
> > > > > +    dc->reset = ct3d_reset;
> > > > > +    device_class_set_props(dc, ct3_props);
> > > > > +
> > > > > +    mdc->get_memory_region = cxl_md_get_memory_region;
> > > > > +    mdc->get_addr = cxl_md_get_addr;
> > > > > +    mdc->fill_device_info = pc_dimm_md_fill_device_info;
> > > > > +    mdc->get_plugged_size = memory_device_get_region_size;
> > > > > +    mdc->set_addr = cxl_md_set_addr;
> > > > > +}
> > > > > +
> > > > > +static const TypeInfo ct3d_info = {
> > > > > +    .name = TYPE_CXL_TYPE3_DEV,
> > > > > +    .parent = TYPE_PCI_DEVICE,
> > > > > +    .class_init = ct3_class_init,
> > > > > +    .instance_size = sizeof(CXLType3Dev),
> > > > > +    .instance_init = ct3_instance_init,
> > > > > +    .instance_finalize = ct3_finalize,
> > > > > +    .interfaces = (InterfaceInfo[]) {
> > > > > +        { TYPE_MEMORY_DEVICE },
> > > > > +        { INTERFACE_CXL_DEVICE },
> > > > > +        { INTERFACE_PCIE_DEVICE },
> > > > > +        {}
> > > > > +    },
> > > > > +};
> > > > > +
> > > > > +static void ct3d_registers(void)
> > > > > +{
> > > > > +    type_register_static(&ct3d_info);
> > > > > +}
> > > > > +
> > > > > +type_init(ct3d_registers);
> > > > > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > > > > index 0d22f2b572..d13c3ed117 100644
> > > > > --- a/hw/mem/meson.build
> > > > > +++ b/hw/mem/meson.build
> > > > > @@ -3,5 +3,6 @@ mem_ss.add(files('memory-device.c'))
> > > > >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> > > > >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> > > > >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > > > > +mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
> > > > >  
> > > > >  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
> > > > > diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> > > > > index d4010cf8f3..1ecf6f6a55 100644
> > > > > --- a/hw/pci/pcie.c
> > > > > +++ b/hw/pci/pcie.c
> > > > > @@ -20,6 +20,7 @@
> > > > >  
> > > > >  #include "qemu/osdep.h"
> > > > >  #include "qapi/error.h"
> > > > > +#include "hw/mem/memory-device.h"
> > > > >  #include "hw/pci/pci_bridge.h"
> > > > >  #include "hw/pci/pcie.h"
> > > > >  #include "hw/pci/msix.h"
> > > > > @@ -27,6 +28,8 @@
> > > > >  #include "hw/pci/pci_bus.h"
> > > > >  #include "hw/pci/pcie_regs.h"
> > > > >  #include "hw/pci/pcie_port.h"
> > > > > +#include "hw/cxl/cxl.h"
> > > > > +#include "hw/boards.h"
> > > > >  #include "qemu/range.h"
> > > > >  
> > > > >  //#define DEBUG_PCIE
> > > > > @@ -419,6 +422,28 @@ void pcie_cap_slot_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > > >      }
> > > > >  
> > > > >      pcie_cap_slot_plug_common(PCI_DEVICE(hotplug_dev), dev, errp);
> > > > > +
> > > > > +#ifdef CXL_MEM_DEVICE
> > > > > +    /*
> > > > > +     * FIXME:
> > > > > +     * if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > > > +     *    HotplugHandler *hotplug_ctrl;
> > > > > +     *   Error *local_err = NULL;
> > > > > +     *  hotplug_ctrl = qdev_get_hotplug_handler(dev);
> > > > > +     *  if (hotplug_ctrl) {
> > > > > +     *      hotplug_handler_pre_plug(hotplug_ctrl, dev, &local_err);
> > > > > +     *      if (local_err) {
> > > > > +     *          error_propagate(errp, local_err);
> > > > > +     *          return;
> > > > > +     *      }
> > > > > +     *  }
> > > > > +     */
> > > > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV)) {
> > > > > +        memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()),
> > > > > +                               NULL, errp);
> > > > > +    }  
> > > > 
> > > > why use MEMORY_DEVICE interface instead of exposing memory as PCI BAR?
> > > >   
> > > 
> > > A CXL memory device adds memory to a system that is physically addressable in
> > > the same way an NVDIMM device would be.
> > > 
> > > > > +#endif
> > > > >  }
> > > > >  
> > > > >  void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > > > @@ -455,6 +480,11 @@ void pcie_cap_slot_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > > >          pcie_cap_slot_event(hotplug_pdev,
> > > > >                              PCI_EXP_HP_EV_PDC | PCI_EXP_HP_EV_ABP);
> > > > >      }
> > > > > +
> > > > > +#ifdef CXL_MEM_DEVICE
> > > > > +    if (object_dynamic_cast(OBJECT(dev), TYPE_CXL_TYPE3_DEV))
> > > > > +        memory_device_plug(MEMORY_DEVICE(dev), MACHINE(qdev_get_machine()));
> > > > > +#endif
> > > > >  }
> > > > >  
> > > > >  void pcie_cap_slot_unplug_cb(HotplugHandler *hotplug_dev, DeviceState *dev,
> > > > > diff --git a/include/hw/cxl/cxl.h b/include/hw/cxl/cxl.h
> > > > > index b1e5f4a8fa..809ed7de60 100644
> > > > > --- a/include/hw/cxl/cxl.h
> > > > > +++ b/include/hw/cxl/cxl.h
> > > > > @@ -17,6 +17,8 @@
> > > > >  #define COMPONENT_REG_BAR_IDX 0
> > > > >  #define DEVICE_REG_BAR_IDX 2
> > > > >  
> > > > > +#define TYPE_CXL_TYPE3_DEV "cxl-type3"
> > > > > +
> > > > >  #define CXL_HOST_BASE 0xD0000000
> > > > >  #define CXL_WINDOW_MAX 10
> > > > >  
> > > > > diff --git a/include/hw/cxl/cxl_pci.h b/include/hw/cxl/cxl_pci.h
> > > > > index a53c2e5ae7..9ec28c9feb 100644
> > > > > --- a/include/hw/cxl/cxl_pci.h
> > > > > +++ b/include/hw/cxl/cxl_pci.h
> > > > > @@ -64,6 +64,28 @@ _Static_assert(sizeof(struct dvsec_header) == 10,
> > > > >   * CXL 2.0 Downstream Port: 3, 4, 7, 8
> > > > >   */
> > > > >  
> > > > > +/* CXL 2.0 - 8.1.3 (ID 0001) */
> > > > > +struct dvsec_device {
> > > > > +    struct dvsec_header hdr;
> > > > > +    uint16_t cap;
> > > > > +    uint16_t ctrl;
> > > > > +    uint16_t status;
> > > > > +    uint16_t ctrl2;
> > > > > +    uint16_t status2;
> > > > > +    uint16_t lock;
> > > > > +    uint16_t cap2;
> > > > > +    uint32_t range1_size_hi;
> > > > > +    uint32_t range1_size_lo;
> > > > > +    uint32_t range1_base_hi;
> > > > > +    uint32_t range1_base_lo;
> > > > > +    uint32_t range2_size_hi;
> > > > > +    uint32_t range2_size_lo;
> > > > > +    uint32_t range2_base_hi;
> > > > > +    uint32_t range2_base_lo;
> > > > > +};
> > > > > +_Static_assert(sizeof(struct dvsec_device) == 0x38,
> > > > > +               "dvsec device size incorrect");
> > > > > +
> > > > >  /* CXL 2.0 - 8.1.5 (ID 0003) */
> > > > >  struct extensions_dvsec_port {
> > > > >      struct dvsec_header hdr;
> > > > > diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h
> > > > > index 11f8ab7149..76bf3ed590 100644
> > > > > --- a/include/hw/pci/pci_ids.h
> > > > > +++ b/include/hw/pci/pci_ids.h
> > > > > @@ -53,6 +53,7 @@
> > > > >  #define PCI_BASE_CLASS_MEMORY            0x05
> > > > >  #define PCI_CLASS_MEMORY_RAM             0x0500
> > > > >  #define PCI_CLASS_MEMORY_FLASH           0x0501
> > > > > +#define PCI_CLASS_MEMORY_CXL             0x0502
> > > > >  #define PCI_CLASS_MEMORY_OTHER           0x0580
> > > > >  
> > > > >  #define PCI_BASE_CLASS_BRIDGE            0x06
> > > > > diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> > > > > index 0dd594f92b..0f67bc61ce 100644
> > > > > --- a/monitor/hmp-cmds.c
> > > > > +++ b/monitor/hmp-cmds.c
> > > > > @@ -1887,6 +1887,21 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
> > > > >                  monitor_printf(mon, "  hotpluggable: %s\n",
> > > > >                                 di->hotpluggable ? "true" : "false");
> > > > >                  break;
> > > > > +            case MEMORY_DEVICE_INFO_KIND_CXL:
> > > > > +                di = value->u.cxl.data;
> > > > > +                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > > > +                               MemoryDeviceInfoKind_str(value->type),
> > > > > +                               di->id ? di->id : "");
> > > > > +                monitor_printf(mon, "  addr: 0x%" PRIx64 "\n", di->addr);
> > > > > +                monitor_printf(mon, "  slot: %" PRId64 "\n", di->slot);
> > > > > +                monitor_printf(mon, "  node: %" PRId64 "\n", di->node);
> > > > > +                monitor_printf(mon, "  size: %" PRIu64 "\n", di->size);
> > > > > +                monitor_printf(mon, "  memdev: %s\n", di->memdev);
> > > > > +                monitor_printf(mon, "  hotplugged: %s\n",
> > > > > +                               di->hotplugged ? "true" : "false");
> > > > > +                monitor_printf(mon, "  hotpluggable: %s\n",
> > > > > +                               di->hotpluggable ? "true" : "false");
> > > > > +                break;
> > > > >              case MEMORY_DEVICE_INFO_KIND_VIRTIO_PMEM:
> > > > >                  vpi = value->u.virtio_pmem.data;
> > > > >                  monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
> > > > > diff --git a/qapi/machine.json b/qapi/machine.json
> > > > > index 330189efe3..aa96d662bd 100644
> > > > > --- a/qapi/machine.json
> > > > > +++ b/qapi/machine.json
> > > > > @@ -1394,6 +1394,7 @@
> > > > >  { 'union': 'MemoryDeviceInfo',
> > > > >    'data': { 'dimm': 'PCDIMMDeviceInfo',
> > > > >              'nvdimm': 'PCDIMMDeviceInfo',
> > > > > +            'cxl': 'PCDIMMDeviceInfo',
> > > > >              'virtio-pmem': 'VirtioPMEMDeviceInfo',
> > > > >              'virtio-mem': 'VirtioMEMDeviceInfo'
> > > > >            }  
> > > >   
> > 
> > 
> 


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-28 15:14           ` Ben Widawsky
@ 2021-01-28 16:51             ` Ben Widawsky
  2021-01-28 16:58               ` Ben Widawsky
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-28 16:51 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	Philippe Mathieu-Daudé,
	qemu-devel, Prashant V Agarwal, Igor Mammedov, Dan Williams

On 21-01-28 07:14:44, Ben Widawsky wrote:
> On 21-01-28 07:03:18, Ben Widawsky wrote:
> > On 21-01-28 10:25:38, Jonathan Cameron wrote:
> > > On Wed, 27 Jan 2021 13:26:45 -0800
> > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > 
> > > > On 21-01-27 22:03:12, Igor Mammedov wrote:
> > > > > On Tue,  5 Jan 2021 08:53:15 -0800
> > > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > > >   
> > > > > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > > > > combination of volatile and persistent memory. It also implements the
> > > > > > previously defined mailbox interface as well as the memory device
> > > > > > firmware interface.
> > > > > > 
> > > > > > The following example will create a 256M device in a 512M window:
> > > > > > 
> > > > > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > > > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > > > > 
> > > > > I'd expect whole backend used by frontend, so one would not need "size" property
> > > > > on frontend (like we do with memory devices).
> > > > > So question is why it partially uses memdev?  
> > > > 
> > > > Answered in a separate thread...
> > > 
> > > One possible suggestion inline.
> > > 
> > > > > > +
> > > > > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > > > > +{
> > > > > > +    MemoryRegionSection mrs;
> > > > > > +    MemoryRegion *mr;
> > > > > > +    uint64_t offset = 0;
> > > > > > +    size_t remaining_size;
> > > > > > +
> > > > > > +    if (!ct3d->hostmem) {
> > > > > > +        error_setg(errp, "memdev property must be set");
> > > > > > +        return;
> > > > > > +    }
> > > > > > +
> > > > > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > > > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > > > > +
> > > > > > +    /* Create our new subregion */
> > > > > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > > > > +
> > > > > > +    /* Find the first free space in the window */
> > > > > > +    WITH_RCU_READ_LOCK_GUARD()
> > > > > > +    {
> > > > > > +        mrs = memory_region_find(mr, offset, 1);
> > > > > > +        while (mrs.mr && mrs.mr != mr) {
> > > > > > +            offset += memory_region_size(mrs.mr);
> > > > > > +            mrs = memory_region_find(mr, offset, 1);
> > > > > > +        }
> > > > > > +    }
> > > > > > +
> > > > > > +    remaining_size = memory_region_size(mr) - offset;
> > > > > > +    if (remaining_size < ct3d->size) {
> > > > > > +        g_free(ct3d->cxl_dstate.pmem);
> > > > > > +        error_setg(errp,
> > > > > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > > > > +                   remaining_size, ct3d->size);
> > > > > > +    }
> > > > > > +
> > > > > > +    /* Register our subregion as non-volatile */
> > > > > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > > > > +                           "cxl_type3-memory", ct3d->size, errp);  
> > > > > this allocates ct3d->size of anon RAM, was this an intention?
> > > > > If yes, can you clarify why extra RAM is used instead of using what
> > > > > backend provides?  
> > > > 
> > > > It sounds like I'm doing the wrong thing then. There should be one chunk of
> > > > memory which is a subset of the full memory backend object. Could you please
> > > > advise on what I should be doing instead? Is add_subregion() sufficient?
> > > 
> > > Taking inspiration from nvdimm I'm carrying a patch that uses
> > > memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(qct3d)q,
> > > 			 "cxl_type3-memory", mr, offset, ct3d->size);
> > > 
> > > I 'think' that's doing the right thing, but haven't fully tested it yet
> > > so may be completely wrong :)
> > > 
> > > Then for the pmem addr, call memory_region_set_address() to put it
> > > in a particular location.
> > > 
> > 
> > Yes - this is what I'd like to do and what I initially tried, and I also believe
> > it's right, but it doesn't work.
> > 
> > range_invariant: Assertion `range->lob <= range->upb || range->lob == range->upb + 1' failed.
> > 
> > I was digging into this yesterday, but opted to start a new thread on the
> > matter.
> > 
> 
> Hmm. I think I need to figure out the right add_subregion after this and it
> might work. I'll keep digging, but if you have ideas, let me know.

[snip]

I managed to get a bit further. With the following, I start getting complaints
about fragmented memory when adding devices later.

     memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
             "cxl_type3-memory", mr, mr->addr + offset, ct3d->size);
     memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
     memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);

-device nvdimm,memdev=nvmem1,id=nv1,label-size=2M,node=5: could not find position in guest address space for memory device - memory fragmented due to alignments


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-28 16:51             ` Ben Widawsky
@ 2021-01-28 16:58               ` Ben Widawsky
  2021-01-28 17:40                 ` Jonathan Cameron
  0 siblings, 1 reply; 57+ messages in thread
From: Ben Widawsky @ 2021-01-28 16:58 UTC (permalink / raw)
  To: Jonathan Cameron
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	qemu-devel, Philippe Mathieu-Daudé,
	Prashant V Agarwal, Igor Mammedov, Dan Williams

On 21-01-28 08:51:51, Ben Widawsky wrote:
> On 21-01-28 07:14:44, Ben Widawsky wrote:
> > On 21-01-28 07:03:18, Ben Widawsky wrote:
> > > On 21-01-28 10:25:38, Jonathan Cameron wrote:
> > > > On Wed, 27 Jan 2021 13:26:45 -0800
> > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > > 
> > > > > On 21-01-27 22:03:12, Igor Mammedov wrote:
> > > > > > On Tue,  5 Jan 2021 08:53:15 -0800
> > > > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > > > >   
> > > > > > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > > > > > combination of volatile and persistent memory. It also implements the
> > > > > > > previously defined mailbox interface as well as the memory device
> > > > > > > firmware interface.
> > > > > > > 
> > > > > > > The following example will create a 256M device in a 512M window:
> > > > > > > 
> > > > > > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > > > > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"  
> > > > > > 
> > > > > > I'd expect whole backend used by frontend, so one would not need "size" property
> > > > > > on frontend (like we do with memory devices).
> > > > > > So question is why it partially uses memdev?  
> > > > > 
> > > > > Answered in a separate thread...
> > > > 
> > > > One possible suggestion inline.
> > > > 
> > > > > > > +
> > > > > > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > > > > > +{
> > > > > > > +    MemoryRegionSection mrs;
> > > > > > > +    MemoryRegion *mr;
> > > > > > > +    uint64_t offset = 0;
> > > > > > > +    size_t remaining_size;
> > > > > > > +
> > > > > > > +    if (!ct3d->hostmem) {
> > > > > > > +        error_setg(errp, "memdev property must be set");
> > > > > > > +        return;
> > > > > > > +    }
> > > > > > > +
> > > > > > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > > > > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > > > > > +
> > > > > > > +    /* Create our new subregion */
> > > > > > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > > > > > +
> > > > > > > +    /* Find the first free space in the window */
> > > > > > > +    WITH_RCU_READ_LOCK_GUARD()
> > > > > > > +    {
> > > > > > > +        mrs = memory_region_find(mr, offset, 1);
> > > > > > > +        while (mrs.mr && mrs.mr != mr) {
> > > > > > > +            offset += memory_region_size(mrs.mr);
> > > > > > > +            mrs = memory_region_find(mr, offset, 1);
> > > > > > > +        }
> > > > > > > +    }
> > > > > > > +
> > > > > > > +    remaining_size = memory_region_size(mr) - offset;
> > > > > > > +    if (remaining_size < ct3d->size) {
> > > > > > > +        g_free(ct3d->cxl_dstate.pmem);
> > > > > > > +        error_setg(errp,
> > > > > > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > > > > > +                   remaining_size, ct3d->size);
> > > > > > > +    }
> > > > > > > +
> > > > > > > +    /* Register our subregion as non-volatile */
> > > > > > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > > > > > +                           "cxl_type3-memory", ct3d->size, errp);  
> > > > > > this allocates ct3d->size of anon RAM, was this an intention?
> > > > > > If yes, can you clarify why extra RAM is used instead of using what
> > > > > > backend provides?  
> > > > > 
> > > > > It sounds like I'm doing the wrong thing then. There should be one chunk of
> > > > > memory which is a subset of the full memory backend object. Could you please
> > > > > advise on what I should be doing instead? Is add_subregion() sufficient?
> > > > 
> > > > Taking inspiration from nvdimm I'm carrying a patch that uses
> > > > memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(qct3d)q,
> > > > 			 "cxl_type3-memory", mr, offset, ct3d->size);
> > > > 
> > > > I 'think' that's doing the right thing, but haven't fully tested it yet
> > > > so may be completely wrong :)
> > > > 
> > > > Then for the pmem addr, call memory_region_set_address() to put it
> > > > in a particular location.
> > > > 
> > > 
> > > Yes - this is what I'd like to do and what I initially tried, and I also believe
> > > it's right, but it doesn't work.
> > > 
> > > range_invariant: Assertion `range->lob <= range->upb || range->lob == range->upb + 1' failed.
> > > 
> > > I was digging into this yesterday, but opted to start a new thread on the
> > > matter.
> > > 
> > 
> > Hmm. I think I need to figure out the right add_subregion after this and it
> > might work. I'll keep digging, but if you have ideas, let me know.
> 
> [snip]
> 
> I managed to get a bit further. With the following, I start getting complaints
> about fragmented memory when adding devices later.
> 
>      memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
>              "cxl_type3-memory", mr, mr->addr + offset, ct3d->size);
>      memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
>      memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> 
> -device nvdimm,memdev=nvmem1,id=nv1,label-size=2M,node=5: could not find position in guest address space for memory device - memory fragmented due to alignments
> 

Ignore this. It was a problem with my commandline.

I think I have something limping along now.


^ permalink raw reply	[flat|nested] 57+ messages in thread

* Re: [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5)
  2021-01-28 16:58               ` Ben Widawsky
@ 2021-01-28 17:40                 ` Jonathan Cameron
  0 siblings, 0 replies; 57+ messages in thread
From: Jonathan Cameron @ 2021-01-28 17:40 UTC (permalink / raw)
  To: Ben Widawsky
  Cc: Thomas Huth, Michael S. Tsirkin, Vishal Verma, Chris Browy,
	qemu-devel, Philippe Mathieu-Daudé,
	Prashant V Agarwal, Igor Mammedov, Dan Williams

On Thu, 28 Jan 2021 08:58:01 -0800
Ben Widawsky <ben@bwidawsk.net> wrote:

> On 21-01-28 08:51:51, Ben Widawsky wrote:
> > On 21-01-28 07:14:44, Ben Widawsky wrote:  
> > > On 21-01-28 07:03:18, Ben Widawsky wrote:  
> > > > On 21-01-28 10:25:38, Jonathan Cameron wrote:  
> > > > > On Wed, 27 Jan 2021 13:26:45 -0800
> > > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > > >   
> > > > > > On 21-01-27 22:03:12, Igor Mammedov wrote:  
> > > > > > > On Tue,  5 Jan 2021 08:53:15 -0800
> > > > > > > Ben Widawsky <ben.widawsky@intel.com> wrote:
> > > > > > >     
> > > > > > > > A CXL memory device (AKA Type 3) is a CXL component that contains some
> > > > > > > > combination of volatile and persistent memory. It also implements the
> > > > > > > > previously defined mailbox interface as well as the memory device
> > > > > > > > firmware interface.
> > > > > > > > 
> > > > > > > > The following example will create a 256M device in a 512M window:
> > > > > > > > 
> > > > > > > > -object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
> > > > > > > > -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0,size=256M"    
> > > > > > > 
> > > > > > > I'd expect whole backend used by frontend, so one would not need "size" property
> > > > > > > on frontend (like we do with memory devices).
> > > > > > > So question is why it partially uses memdev?    
> > > > > > 
> > > > > > Answered in a separate thread...  
> > > > > 
> > > > > One possible suggestion inline.
> > > > >   
> > > > > > > > +
> > > > > > > > +static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > > > > > > > +{
> > > > > > > > +    MemoryRegionSection mrs;
> > > > > > > > +    MemoryRegion *mr;
> > > > > > > > +    uint64_t offset = 0;
> > > > > > > > +    size_t remaining_size;
> > > > > > > > +
> > > > > > > > +    if (!ct3d->hostmem) {
> > > > > > > > +        error_setg(errp, "memdev property must be set");
> > > > > > > > +        return;
> > > > > > > > +    }
> > > > > > > > +
> > > > > > > > +    /* FIXME: need to check mr is the host bridge's MR */
> > > > > > > > +    mr = host_memory_backend_get_memory(ct3d->hostmem);
> > > > > > > > +
> > > > > > > > +    /* Create our new subregion */
> > > > > > > > +    ct3d->cxl_dstate.pmem = g_new(MemoryRegion, 1);
> > > > > > > > +
> > > > > > > > +    /* Find the first free space in the window */
> > > > > > > > +    WITH_RCU_READ_LOCK_GUARD()
> > > > > > > > +    {
> > > > > > > > +        mrs = memory_region_find(mr, offset, 1);
> > > > > > > > +        while (mrs.mr && mrs.mr != mr) {
> > > > > > > > +            offset += memory_region_size(mrs.mr);
> > > > > > > > +            mrs = memory_region_find(mr, offset, 1);
> > > > > > > > +        }
> > > > > > > > +    }
> > > > > > > > +
> > > > > > > > +    remaining_size = memory_region_size(mr) - offset;
> > > > > > > > +    if (remaining_size < ct3d->size) {
> > > > > > > > +        g_free(ct3d->cxl_dstate.pmem);
> > > > > > > > +        error_setg(errp,
> > > > > > > > +                   "Not enough free space (%zd) required for device (%" PRId64  ")",
> > > > > > > > +                   remaining_size, ct3d->size);
> > > > > > > > +    }
> > > > > > > > +
> > > > > > > > +    /* Register our subregion as non-volatile */
> > > > > > > > +    memory_region_init_ram(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> > > > > > > > +                           "cxl_type3-memory", ct3d->size, errp);    
> > > > > > > this allocates ct3d->size of anon RAM, was this an intention?
> > > > > > > If yes, can you clarify why extra RAM is used instead of using what
> > > > > > > backend provides?    
> > > > > > 
> > > > > > It sounds like I'm doing the wrong thing then. There should be one chunk of
> > > > > > memory which is a subset of the full memory backend object. Could you please
> > > > > > advise on what I should be doing instead? Is add_subregion() sufficient?  
> > > > > 
> > > > > Taking inspiration from nvdimm I'm carrying a patch that uses
> > > > > memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(qct3d)q,
> > > > > 			 "cxl_type3-memory", mr, offset, ct3d->size);
> > > > > 
> > > > > I 'think' that's doing the right thing, but haven't fully tested it yet
> > > > > so may be completely wrong :)
> > > > > 
> > > > > Then for the pmem addr, call memory_region_set_address() to put it
> > > > > in a particular location.
> > > > >   
> > > > 
> > > > Yes - this is what I'd like to do and what I initially tried, and I also believe
> > > > it's right, but it doesn't work.
> > > > 
> > > > range_invariant: Assertion `range->lob <= range->upb || range->lob == range->upb + 1' failed.
> > > > 
> > > > I was digging into this yesterday, but opted to start a new thread on the
> > > > matter.
> > > >   
> > > 
> > > Hmm. I think I need to figure out the right add_subregion after this and it
> > > might work. I'll keep digging, but if you have ideas, let me know.  
> > 
> > [snip]
> > 
> > I managed to get a bit further. With the following, I start getting complaints
> > about fragmented memory when adding devices later.
> > 
> >      memory_region_init_alias(ct3d->cxl_dstate.pmem, OBJECT(ct3d),
> >              "cxl_type3-memory", mr, mr->addr + offset, ct3d->size);
> >      memory_region_set_nonvolatile(ct3d->cxl_dstate.pmem, true);
> >      memory_region_add_subregion(mr, offset, ct3d->cxl_dstate.pmem);
> > 
> > -device nvdimm,memdev=nvmem1,id=nv1,label-size=2M,node=5: could not find position in guest address space for memory device - memory fragmented due to alignments
> >   
> 
> Ignore this. It was a problem with my commandline.
> 
> I think I have something limping along now.

Great.  It's going to get more interesting to do this 'right' though.

We'll need to define a memory window similar to device_mem which is used for
normal memory hotplug then export that window via the relevant ACPI tables
(once defined somewhere public).  Finally we'll then want to actually
do the memory_region_add_subregion() only on the OS / firmware having
configured the type3 device by setting it's PA base.

Going to be fiddly - but should be doable.  With it basically working
should be a series of sensible (ish) steps to get there.

Jonathan







^ permalink raw reply	[flat|nested] 57+ messages in thread

end of thread, other threads:[~2021-01-28 17:53 UTC | newest]

Thread overview: 57+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-05 16:52 [RFC PATCH v2 00/32] CXL 2.0 Support Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 01/32] Temp: Add the PCI_EXT_ID_DVSEC definition to the qemu pci_regs.h copy Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 02/32] hw/pci/cxl: Add a CXL component type (interface) Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 03/32] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5) Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 04/32] hw/cxl/device: Introduce a CXL device (8.2.8) Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 05/32] hw/cxl/device: Implement the CAP array (8.2.8.1-2) Ben Widawsky
2021-01-06 13:28   ` Jonathan Cameron
2021-01-06 16:49     ` Ben Widawsky
2021-01-06 17:06       ` Jonathan Cameron
2021-01-06 17:09         ` Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 06/32] hw/cxl/device: Add device status (8.2.8.3) Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 07/32] hw/cxl/device: Implement basic mailbox (8.2.8.4) Ben Widawsky
2021-01-06 13:21   ` Jonathan Cameron
2021-01-06 16:31     ` Ben Widawsky
2021-01-06 17:40     ` [Linuxarm] " Jonathan Cameron
2021-01-06 18:05       ` Ben Widawsky
2021-01-06 19:08         ` Ben Widawsky
2021-01-08  5:36           ` Ben Widawsky
2021-01-05 16:52 ` [RFC PATCH v2 08/32] hw/cxl/device: Add memory devices (8.2.8.5) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 09/32] hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 10/32] hw/cxl/device: Placeholder for firmware commands Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 11/32] hw/cxl/device: Timestamp implementation (8.2.9.3) Ben Widawsky
2021-01-05 17:12   ` Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 12/32] hw/cxl/device: Add log commands (8.2.9.4) + CEL Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 13/32] hw/pxb: Use a type for realizing expanders Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 14/32] hw/pci/cxl: Create a CXL bus type Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 15/32] hw/pxb: Allow creation of a CXL PXB (host bridge) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 16/32] qtest: allow DSDT acpi table changes Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 17/32] acpi/pci: Consolidate host bridge setup Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 18/32] tests/acpi: remove stale allowed tables Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 19/32] hw/pci: Plumb _UID through host bridges Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 20/32] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 21/32] acpi/pxb/cxl: Reserve host bridge MMIO Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 22/32] hw/pxb/cxl: Add "windows" for host bridges Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 23/32] hw/cxl/rp: Add a root port Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 24/32] hw/cxl/device: Add a memory device (8.2.8.5) Ben Widawsky
2021-01-27 21:03   ` Igor Mammedov
2021-01-27 21:11     ` Ben Widawsky
2021-01-27 21:21       ` Igor Mammedov
2021-01-27 21:30         ` Ben Widawsky
2021-01-27 21:26     ` Ben Widawsky
2021-01-28 10:25       ` Jonathan Cameron
2021-01-28 15:03         ` Ben Widawsky
2021-01-28 15:14           ` Ben Widawsky
2021-01-28 16:51             ` Ben Widawsky
2021-01-28 16:58               ` Ben Widawsky
2021-01-28 17:40                 ` Jonathan Cameron
2021-01-05 16:53 ` [RFC PATCH v2 25/32] hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 26/32] acpi/cxl: Add _OSC implementation (9.14.2) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 27/32] tests/acpi: allow CEDT table addition Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 28/32] acpi/cxl: Create the CEDT (9.14.1) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 29/32] Temp: acpi/cxl: Add ACPI0017 (CEDT awareness) Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 30/32] tests/acpi: Add new CEDT files Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 31/32] WIP: i386/cxl: Initialize a host bridge Ben Widawsky
2021-01-05 16:53 ` [RFC PATCH v2 32/32] qtest/cxl: Add very basic sanity tests Ben Widawsky
2021-01-08 18:44 ` [RFC PATCH v2 00/32] CXL 2.0 Support Jonathan Cameron
2021-01-08 18:51   ` Ben Widawsky

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).