2022-02-25 04:15:01

by Zhang, Tianfei

[permalink] [raw]
Subject: [PATCH v2 0/5] Add Intel OFS support for DFL driver

This is v2 patchset adding IOFS (Intel Open FPGA stack) support for
DFL driver. IOFS is Intel's version of a common core set of
RTL to allow customers to easily interface to logic and IP
on the FPGA. IOFS leverages the DFL for the implementation of
the FPGA RTL design.

Patch 1, allows for ports without specific bar space.
Patch 2, introduces features in dfl_fpga_cdev after DFL enumeration.
On IOFS, we will add more extensions or features in DFL in
future, so adding a new member "features"in dfl_fpga_cdev.
Patch 3, fixs VF creation in "Multiple VFs per PR slot" and legacy model.
Patch 4, handles dfl's starting with AFU and allows for VFs to be created.
Patch 5, adds architecture description about IOFS support for DFL
in documentation.

Changelog v1 -> v2:
- Introducing a new member "features" in dfl_fpga_cdev for feature
control.
- Adding new flag DFL_FEAT_PORT_CONNECTED_AFU for IOFS legacy model.
- Updates the documentation for the access models about AFU in IOFS.
- Drop the PCI PID patch and will send it later.

Matthew Gerlach (2):
fpga: dfl: Allow for ports without specific bar space.
fpga: dfl: Handle dfl's starting with AFU

Tianfei zhang (3):
fpga: dfl: add features in dfl_fpga_cdev
fpga: dfl: fix VF creation in IOFS
Documentation: fpga: dfl: add description of IOFS

Documentation/fpga/dfl.rst | 113 +++++++++++++++++++++++++++++++++++++
drivers/fpga/dfl-pci.c | 13 ++++-
drivers/fpga/dfl.c | 38 ++++++++-----
drivers/fpga/dfl.h | 6 ++
4 files changed, 155 insertions(+), 15 deletions(-)

--
2.26.2


2022-02-25 06:44:14

by Zhang, Tianfei

[permalink] [raw]
Subject: [PATCH v2 2/5] fpga: dfl: add features in dfl_fpga_cdev

Introducing features in dfl_fpga_cdev during DFL enumeration.
On IOFS, we will add more extensions or features in DFL in
future, so adding a new member "features"in dfl_fpga_cdev.
For example, in the legacy model, the AFU was connected to
Port device, but in "multiple VFs per PR slot" model, the
AFU or PR slot without connected to Port device directly,
so in this model, we only can access the resource of AFU
or PR slot via VFs. In this patch, we introducing a new
flags DFL_FEAT_PORT_CONNECTED_AFU to distinguish them.

Signed-off-by: Tianfei zhang <[email protected]>
---
drivers/fpga/dfl.c | 6 +++++-
drivers/fpga/dfl.h | 5 +++++
2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
index 6bff39ff21a0..9b7a01a4af04 100644
--- a/drivers/fpga/dfl.c
+++ b/drivers/fpga/dfl.c
@@ -1130,6 +1130,7 @@ static void build_info_complete(struct build_feature_devs_info *binfo)
static int parse_feature_fiu(struct build_feature_devs_info *binfo,
resource_size_t ofst)
{
+ struct dfl_fpga_cdev *cdev = binfo->cdev;
int ret = 0;
u32 offset;
u16 id;
@@ -1166,8 +1167,11 @@ static int parse_feature_fiu(struct build_feature_devs_info *binfo,
v = readq(binfo->ioaddr + NEXT_AFU);

offset = FIELD_GET(NEXT_AFU_NEXT_DFH_OFST, v);
- if (offset)
+ if (offset) {
+ if (dfh_id_to_type(id) == PORT_ID)
+ cdev->features |= DFL_FEAT_PORT_CONNECTED_AFU;
return parse_feature_afu(binfo, offset);
+ }

dev_dbg(binfo->dev, "No AFUs detected on FIU %d\n", id);

diff --git a/drivers/fpga/dfl.h b/drivers/fpga/dfl.h
index 1fd493e82dd8..6171bcdcb3c5 100644
--- a/drivers/fpga/dfl.h
+++ b/drivers/fpga/dfl.h
@@ -461,6 +461,9 @@ int dfl_fpga_enum_info_add_irq(struct dfl_fpga_enum_info *info,
unsigned int nr_irqs, int *irq_table);
void dfl_fpga_enum_info_free(struct dfl_fpga_enum_info *info);

+/* in legacy model, the AFU was connected to Port device */
+#define DFL_FEAT_PORT_CONNECTED_AFU BIT_ULL(0)
+
/**
* struct dfl_fpga_cdev - container device of DFL based FPGA
*
@@ -470,6 +473,7 @@ void dfl_fpga_enum_info_free(struct dfl_fpga_enum_info *info);
* @lock: mutex lock to protect the port device list.
* @port_dev_list: list of all port feature devices under this container device.
* @released_port_num: released port number under this container device.
+ * @features: features discovered during DFL enumeration.
*/
struct dfl_fpga_cdev {
struct device *parent;
@@ -478,6 +482,7 @@ struct dfl_fpga_cdev {
struct mutex lock;
struct list_head port_dev_list;
int released_port_num;
+ u64 features;
};

struct dfl_fpga_cdev *
--
2.26.2

2022-02-25 08:16:13

by Zhang, Tianfei

[permalink] [raw]
Subject: [PATCH v2 5/5] Documentation: fpga: dfl: add description of IOFS

This patch adds description about IOFS support for DFL.

---
v2:
* Fixs some typos.
* Adds more detail description about the models of AFU access which supported in IOFS.

Signed-off-by: Tianfei zhang <[email protected]>
---
Documentation/fpga/dfl.rst | 113 +++++++++++++++++++++++++++++++++++++
1 file changed, 113 insertions(+)

diff --git a/Documentation/fpga/dfl.rst b/Documentation/fpga/dfl.rst
index ef9eec71f6f3..4c3ac6f452bc 100644
--- a/Documentation/fpga/dfl.rst
+++ b/Documentation/fpga/dfl.rst
@@ -556,6 +556,119 @@ new DFL feature via UIO direct access, its feature id should be added to the
driver's id_table.


+Intel Open FPGA stack
+=====================
+
+Intel Open FPGA stack aka IOFS, Intel's version of a common core set of
+RTL to allow customers to easily interface to logic and IP on the FPGA.
+IOFS leverages the DFL for the implementation of the FPGA RTL design.
+
+IOFS designs allow for the arrangement of software interfaces across multiple
+PCIe endpoints. Some of these interfaces may be PFs defined in the static region
+that connect to interfaces in an IP that is loaded via Partial Reconfiguration (PR).
+And some of these interfaces may be VFs defined in the PR region that can be
+reconfigured by the end-user. Furthermore, these PFs/VFs may also be arranged
+using a DFL such that features may be discovered and accessed in user space
+(with the aid of a generic kernel driver like vfio-pci). The diagram below depicts
+an example design with two PFs and two VFs. In this example, it will export the
+management functions via PF0, PF1 will bind with virtio-net driver presenting itself
+as a network interface to the OS. The other functions, VF0 and VF1, leverage VFIO
+to export the MMIO space to an application or assign to a VM.
+::
+
+ +-----------------+ +--------------+ +-------------+ +------------+
+ | FPGA Management | | VirtIO | | User App | | Virtual |
+ | App | | App | | | | Machine |
+ +--------+--------+ +------+-------+ +------+------+ +-----+------+
+ | | | |
+ +--------+--------+ +------+-------+ +------+------+ |
+ | DFL Driver | |VirtIO driver | | VFIO | |
+ +--------+--------+--+------+-------+ +------+------+ |
+ | | | |
+ | | | |
+ +--------+--------+ +------+-------+ +------+------+ +----+------+
+ | PF0 | | PF1 | | PF0_VF0 | | PF0_VF1 |
+ +-----------------+ +--------------+ +-------------+ +-----------+
+
+As accelerators are specialized hardware, they are typically limited in the
+number installed in a given system. Many use cases require them to be shared
+across multiple software contexts or threads of software execution, either
+through partitioning of individual dedicated resources, or virtualization of
+shared resources. On IOFS, it provides several models to share the AFU
+resources via PR mechanism and hardware-based virtualization schemes.
+
+1. Legacy model.
+ In legacy FPGA card platforms (like Intel PAC N3000 or N5000 Card),there is
+ a notion that the boundary between the AFU and the shell is also the unit of
+ PR for those FPGA platforms. In this model, it can only able to handle a
+ single context, because it only has one PR engine, and one PR region which
+ has an associated Port device.
+2. Multiple VFs per PR slot.
+ In this model, available AFU resources may allow instantiation of many of VFs
+ which has a dedicated PCIe function with their own dedicated MMIO space, or
+ partition a region of MMIO space on a single PCIe function.
+ In this model, the Port device would not connected to AFU/PR slot, so we don't
+ need to release the Port device before creating the VFs. For DFL's view, the AFU
+ will not connect to Port device, so the Next_AFU pointer in FIU feature header
+ of port device points to NULL in this model. On the other hand, each VF can start
+ with an AFU feature header without connected to a FIU Port feature header.
+3. Micro-Personas in AFU.
+ IOFS introducing a new concept to extend the FPGA usage, Micro-Personas in
+ AFU. It finds some downsides of the legacy model to be unacceptable, because
+ this may be desirable by a customer who intends to switch out one accelerator
+ for another accelerator without having to reconfigure the entire FPGA.
+ Micro-Personas allow the developer to designate their own AFU-defined PR
+ regions. In this model the unit of PR is not the entire AFU, instead
+ the unit of PR can be any size block or blocks inside the AFU.
+ In this model, it has PR capability includes one PR engine and multiple PR regions,
+ and each PR region has an associated port gasket. A PR region may also be
+ referred to as a PR slot. Port gasket is similar with port device in legacy
+ model which include the port control, port user clock control and port errors.
+
+IOFS provides the diversity for access the AFU resource to RTL developer.
+An IP designer may choose to add more than one PF for interfacing with IP
+on the FPGA and choose different model to access the AFU resource.
+
+There is one reference architecture design using the "Multiple VFs per PR slot"
+model for IOFS as illustrated below. In this reference design, it exports the
+FPGA management functions via PF0. PF1 will bind with virtio-net driver
+presenting itself as a network interface to the OS. PF2 will bound to the
+vfio-pci driver allowing the user space software to discover and interface
+with the specific workload like diagnostic test. To access the AFU resource,
+it uses SR-IOV to partition workload interfaces across various VFs.
+::
+
+ +----------------------+
+ | PF/VF mux/demux |
+ +--+--+-----+------+-+-+
+ | | | | |
+ +------------------------+ | | | |
+ PF0 | +---------+ +-+ | |
+ +---+---+ | +---+----+ | |
+ | DFH | | | DFH | | |
+ +-------+ +-----+----+ +--------+ | |
+ | FME | | VirtIO | | Test | | |
+ +---+---+ +----------+ +--------+ | |
+ | PF1 PF2 | |
+ | | |
+ | +----------+ |
+ | | ++
+ | | |
+ | | PF0_VF0 | PF0_VF1
+ | +-----------------+-----------+------------+
+ | | +-----+-----------+--------+ |
+ | | | | | | |
+ | | +------+ | +--+ -+ +--+---+ | |
+ | | | Port | | | DFH | | DFH | | |
+ +-----------+ +------+ | +-----+ +------+ | |
+ | | | DEV | | DEV | | |
+ | | +-----+ +------+ | |
+ | | PR Slot | |
+ | +--------------------------+ |
+ | Port Gasket |
+ +------------------------------------------+
+
+
Open discussion
===============
FME driver exports one ioctl (DFL_FPGA_FME_PORT_PR) for partial reconfiguration
--
2.26.2

2022-02-25 09:03:24

by Zhang, Tianfei

[permalink] [raw]
Subject: [PATCH v2 4/5] fpga: dfl: Handle dfl's starting with AFU

From: Matthew Gerlach <[email protected]>

Allow for a Device Feature List (DFL) to start with
a Device Feature Header (DFH) of type Accelerator Function Unit (AFU)
by doing nothing. This allows for PCIe VFs to be created.

Signed-off-by: Matthew Gerlach <[email protected]>
Signed-off-by: Tianfei Zhang <[email protected]>
---
drivers/fpga/dfl-pci.c | 7 ++++++-
drivers/fpga/dfl.c | 22 +++++++++++++---------
2 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/drivers/fpga/dfl-pci.c b/drivers/fpga/dfl-pci.c
index 33545c999c06..e7d58e7b1bbd 100644
--- a/drivers/fpga/dfl-pci.c
+++ b/drivers/fpga/dfl-pci.c
@@ -275,7 +275,12 @@ static int find_dfls_by_default(struct pci_dev *pcidev,

dfl_fpga_enum_info_add_dfl(info, start, len);
} else {
- ret = -ENODEV;
+ v = readq(base + DFH);
+ if (FIELD_GET(DFH_TYPE, v) != DFH_TYPE_AFU) {
+ dev_info(&pcidev->dev, "Unknown feature type 0x%llx id 0x%llx\n",
+ FIELD_GET(DFH_TYPE, v), FIELD_GET(DFH_ID, v));
+ ret = -ENODEV;
+ }
}

/* release I/O mappings for next step enumeration */
diff --git a/drivers/fpga/dfl.c b/drivers/fpga/dfl.c
index afa9311c30d5..8a71ff6c3e2f 100644
--- a/drivers/fpga/dfl.c
+++ b/drivers/fpga/dfl.c
@@ -900,9 +900,11 @@ static void build_info_free(struct build_feature_devs_info *binfo)
dfl_id_free(feature_dev_id_type(binfo->feature_dev),
binfo->feature_dev->id);

- list_for_each_entry_safe(finfo, p, &binfo->sub_features, node) {
- list_del(&finfo->node);
- kfree(finfo);
+ if (!list_empty(&binfo->sub_features)) {
+ list_for_each_entry_safe(finfo, p, &binfo->sub_features, node) {
+ list_del(&finfo->node);
+ kfree(finfo);
+ }
}
}

@@ -1445,12 +1447,14 @@ dfl_fpga_feature_devs_enumerate(struct dfl_fpga_enum_info *info)
* start enumeration for all feature devices based on Device Feature
* Lists.
*/
- list_for_each_entry(dfl, &info->dfls, node) {
- ret = parse_feature_list(binfo, dfl->start, dfl->len);
- if (ret) {
- remove_feature_devs(cdev);
- build_info_free(binfo);
- goto unregister_region_exit;
+ if (!list_empty(&info->dfls)) {
+ list_for_each_entry(dfl, &info->dfls, node) {
+ ret = parse_feature_list(binfo, dfl->start, dfl->len);
+ if (ret) {
+ remove_feature_devs(cdev);
+ build_info_free(binfo);
+ goto unregister_region_exit;
+ }
}
}

--
2.26.2

2022-02-25 13:04:26

by Zhang, Tianfei

[permalink] [raw]
Subject: [PATCH v2 1/5] fpga: dfl: Allow for ports without specific bar space.

From: Matthew Gerlach <[email protected]>

In IOFS, there is a Port device for each PR slot, like Port
control, Port user clock control and Port errors, those feature
devices are linked with DFL. The DFL of Port device was located
in PCIe Bar 0 MMIO space by default, but it also can put into any
PCIe Bar space. If the BarID (3bits field) in PORTn_OFFSET register
set to invalid means that DFL of Port device is located in the Bar 0
by default, in this case, it don't need add the Bar 0 into dfl list
twice.

---
v2: use FME_HDR_NO_PORT_BAR instead of PCI_STD_NUM_BARS.

Signed-off-by: Matthew Gerlach <[email protected]>
Signed-off-by: Tianfei Zhang <[email protected]>
---
drivers/fpga/dfl-pci.c | 6 ++++++
drivers/fpga/dfl.h | 1 +
2 files changed, 7 insertions(+)

diff --git a/drivers/fpga/dfl-pci.c b/drivers/fpga/dfl-pci.c
index 4d68719e608f..33545c999c06 100644
--- a/drivers/fpga/dfl-pci.c
+++ b/drivers/fpga/dfl-pci.c
@@ -258,6 +258,12 @@ static int find_dfls_by_default(struct pci_dev *pcidev,
*/
bar = FIELD_GET(FME_PORT_OFST_BAR_ID, v);
offset = FIELD_GET(FME_PORT_OFST_DFH_OFST, v);
+ if (bar >= FME_HDR_NO_PORT_BAR) {
+ dev_dbg(&pcidev->dev, "skipping port without specific BAR space %d\n",
+ bar);
+ continue;
+ }
+
start = pci_resource_start(pcidev, bar) + offset;
len = pci_resource_len(pcidev, bar) - offset;

diff --git a/drivers/fpga/dfl.h b/drivers/fpga/dfl.h
index 53572c7aced0..1fd493e82dd8 100644
--- a/drivers/fpga/dfl.h
+++ b/drivers/fpga/dfl.h
@@ -91,6 +91,7 @@
#define FME_HDR_PORT_OFST(n) (0x38 + ((n) * 0x8))
#define FME_HDR_BITSTREAM_ID 0x60
#define FME_HDR_BITSTREAM_MD 0x68
+#define FME_HDR_NO_PORT_BAR 7

/* FME Fab Capability Register Bitfield */
#define FME_CAP_FABRIC_VERID GENMASK_ULL(7, 0) /* Fabric version ID */
--
2.26.2