[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RFC PATCH v3 04/31] hw/cxl/device: Implement the CAP array (8.2.8.1
From: |
Jonathan Cameron |
Subject: |
Re: [RFC PATCH v3 04/31] hw/cxl/device: Implement the CAP array (8.2.8.1-2) |
Date: |
Tue, 2 Feb 2021 12:23:50 +0000 |
On Mon, 1 Feb 2021 16:59:21 -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.
>
> Endianness and alignment are managed by softmmu memory core.
>
> Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
A few trivials
> ---
> hw/cxl/cxl-device-utils.c | 105 ++++++++++++++++++++++++++++++++++++
> hw/cxl/meson.build | 1 +
> include/hw/cxl/cxl_device.h | 27 +++++++++-
> 3 files changed, 132 insertions(+), 1 deletion(-)
> 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..bb15ad9a0f
> --- /dev/null
> +++ b/hw/cxl/cxl-device-utils.c
> @@ -0,0 +1,105 @@
> +/*
> + * 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"
> +
> +/*
> + * Device registers have no restrictions per the spec, and so fall back to
> the
> + * default memory mapped register rules in 8.2:
> + * Software shall use CXL.io Memory Read and Write to access memory mapped
> + * register defined in this section. Unless otherwise specified, software
> + * shall restrict the accesses width based on the following:
> + * • A 32 bit register shall be accessed as a 1 Byte, 2 Bytes or 4 Bytes
odd spacing
> + * quantity.
> + * • A 64 bit register shall be accessed as a 1 Byte, 2 Bytes, 4 Bytes or 8
> + * Bytes
> + * • The address shall be a multiple of the access width, e.g. when
> + * accessing a register as a 4 Byte quantity, the address shall be
> + * multiple of 4.
> + * • The accesses shall map to contiguous bytes.If these rules are not
> + * followed, the behavior is undefined
> + */
> +
> +static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
> +{
> + CXLDeviceState *cxl_dstate = opaque;
> +
> + return cxl_dstate->caps_reg_state32[offset / 4];
> +}
> +
> +static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
> +{
> + return 0;
> +}
> +
> +static const MemoryRegionOps dev_ops = {
> + .read = dev_reg_read,
> + .write = NULL, /* status register is read only */
> + .endianness = DEVICE_LITTLE_ENDIAN,
> + .valid = {
> + .min_access_size = 1,
> + .max_access_size = 8,
> + .unaligned = false,
> + },
> + .impl = {
> + .min_access_size = 1,
> + .max_access_size = 8,
> + },
> +};
> +
> +static const MemoryRegionOps caps_ops = {
> + .read = caps_reg_read,
> + .write = NULL, /* caps registers are read only */
> + .endianness = DEVICE_LITTLE_ENDIAN,
> + .valid = {
> + .min_access_size = 1,
> + .max_access_size = 8,
> + .unaligned = false,
> + },
> + .impl = {
> + .min_access_size = 4,
> + .max_access_size = 4,
> + },
> +};
> +
> +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_MMIO_SIZE));
> +
> + memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
> + "cap-array", CXL_DEVICE_REGISTERS_OFFSET - 0);
Specifying a size in terms of the offset of another region isn't exactly
intuitive so perhaps a comment on why or better yet actually use a size
parameter covering what is there rather than simply the region below
the CXL_DEVICE_REGISTERS_OFFSET.
> + 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 = 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/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',
> ))
> diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> index a85f250503..f3bcf19410 100644
> --- a/include/hw/cxl/cxl_device.h
> +++ b/include/hw/cxl/cxl_device.h
> @@ -58,6 +58,8 @@
> #define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
> #define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
> #define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
> +#define CXL_CAPS_SIZE \
> + (CXL_DEVICE_CAP_REG_SIZE * CXL_DEVICE_CAPS_MAX + 1) /* +1 for header */
>
> #define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */
> #define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */
> @@ -70,11 +72,18 @@
> #define CXL_MAILBOX_REGISTERS_LENGTH \
> (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
>
> +#define CXL_MMIO_SIZE \
> + CXL_DEVICE_CAP_REG_SIZE + CXL_DEVICE_REGISTERS_LENGTH + \
> + CXL_MAILBOX_REGISTERS_LENGTH
> +
> typedef struct cxl_device_state {
> MemoryRegion device_registers;
>
> /* mmio for device capabilities array - 8.2.8.2 */
> - MemoryRegion caps;
> + struct {
> + MemoryRegion caps;
> + uint32_t caps_reg_state32[CXL_CAPS_SIZE / 4];
> + };
With this unnamed,w hat is the benefit of having these two in a
struct? The naming makes it clear they are related anyway.
>
> /* mmio for the device status registers 8.2.8.3 */
> MemoryRegion device;
> @@ -126,6 +135,22 @@ 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)
- [RFC PATCH v3 00/31] CXL 2.0 Support, Ben Widawsky, 2021/02/01
- [RFC PATCH v3 01/31] hw/pci/cxl: Add a CXL component type (interface), Ben Widawsky, 2021/02/01
- [RFC PATCH v3 02/31] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5), Ben Widawsky, 2021/02/01
- [RFC PATCH v3 03/31] hw/cxl/device: Introduce a CXL device (8.2.8), Ben Widawsky, 2021/02/01
- [RFC PATCH v3 04/31] hw/cxl/device: Implement the CAP array (8.2.8.1-2), Ben Widawsky, 2021/02/01
- Re: [RFC PATCH v3 04/31] hw/cxl/device: Implement the CAP array (8.2.8.1-2),
Jonathan Cameron <=
- [RFC PATCH v3 05/31] hw/cxl/device: Implement basic mailbox (8.2.8.4), Ben Widawsky, 2021/02/01
- Re: [RFC PATCH v3 05/31] hw/cxl/device: Implement basic mailbox (8.2.8.4), Jonathan Cameron, 2021/02/11
[RFC PATCH v3 07/31] hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1), Ben Widawsky, 2021/02/01