[PATCH v2 00/28] CXL PMEM Region Provisioning

[PATCH v2 00/28] CXL PMEM Region Provisioning

[Dan Williams]

Changes since v1 [1]:
- Move 19 patches that have received a Reviewed-by to the 'pending'
  branch in cxl.git (Thanks Alison, Adam, and Jonathan!)
- Improve the changelog and add more Cc's to "cxl/acpi: Track CXL
  resources in iomem_resource" and highlight the new export of
  insert_resource_expand_to_fit()
- Switch all occurrences of the pattern "rc = -ECODE; if (condition)
  goto err;" to "if (condition) { rc = -ECODE; goto err; }" (Jonathan)
- Re-organize all the cxl_{root,switch,endpoint}_decoder() patches to
  move the decoder-type-specific setup into the decoder-type-specific
  allocation routines (Jonathan)
- Add kdoc to clarify the behavior of add_cxl_resources() (Jonathan)
- Add IORES_DESC_CXL for kernel components like EDAC to determine when
  they might be dealing with a CXL address range (Tony)
- Drop usage of dev_set_drvdata() for passing @cxl_res (Jonathan)
- Drop @remove_action argument to __cxl_dpa_release(), make it behave
  like any other devm_<free> helper (Jonathan)
- Clarify 'skip' vs 'skipped' in DPA handling helpers (Jonathan)
- Clarify why port teardown no proceeds under the lock with the
  conversion from list to xarray (Jonathan)
- Revert rename of cxl_find_dport_by_dev() (Jonathan)
- Fold down_read() / up_write() mismatch fix to the patch that
  introduced the problem (Jonathan)
- Fix description of interleave_ways and interleave_granularity in the
  sysfs ABI document
- Clarify tangential cleanups in "resource: Introduce
  alloc_free_mem_region()" (Jonathan)
- Clarify rationale for the region creation / naming ABI (Jonathan)
- Add SET_CXL_REGION_ATTR() to supplement CXL_REGION_ATTR() the former
  is used to optionally added region attributes to an attribute list
  (position independent) and the latter is used to retrieve a pointer to
  the attribute in code.  (Jonathan)
- For writes to region attributes allow the same value to be written
  multiple times without error (Jonathan)
- Clarify the actions performed by cxl_port_attach_region() (Jonathan)
- Commit message spelling fixes (Alison and Jonathan)
- Rename cxl_dpa_resource() => cxl_dpa_resource_start() (Jonathan)
- Reword error message in cxl_parse_cfmws() (Adam)
- Keep @expected_len signed in cxl_acpi_cfmws_verify() (Jonathan)
- Miscellaneous formatting and doc fixes (Jonathan)
- Rename port->dpa_end port->hdm_end (Jonathan)
- Rename unregister_region() => unregister_nvdimm_region() (Jonathan)

[1]: https://lore.kernel.org/linux-cxl/165603869943.551046.3498980330327696732.stgit@dwillia2-xfh

---

Until the CXL 2.0 definition arrived there was little reason for OS
drivers to care about CXL memory expanders. Similar to DDR they just
implemented a physical address range that was described to the OS by
platform firmware (EFI Memory Map + ACPI SRAT/SLIT/HMAT etc). The CXL
2.0 definition adds support for PMEM, hotplug, switch topologies, and
device-interleaving which exceeds the limits of what can be reasonably
abstracted by EFI + ACPI mechanisms. As a result, Linux needs a native
capability to provision new CXL regions.

The term "region" is the same term that originated in the LIBNVDIMM
implementation to describe a host physical / system physical address
range. For PMEM a region is a persistent memory range that can be
further sub-divided into namespaces. For CXL there are three
classifications of regions:
- PMEM: set up by CXL native tooling and persisted in CXL region labels

- RAM: set up dynamically by CXL native tooling after hotplug events, or
  leftover capacity not mapped by platform firmware. Any persistent
  configuration would come from set up scripts / configuration files in
  userspace.

- System RAM: set up by platform firmware and described by EFI + ACPI
  metadata, these regions are static.

For now, these patches implement just PMEM regions without region label
support. Note though that the infrastructure routines like
cxl_region_attach() and cxl_region_setup_targets() are building blocks
for region-label support, provisioning RAM regions, and enumerating
System RAM regions.

The general flow for provisioning a CXL region is to:
- Find a device or set of devices with available device-physical-address
  (DPA) capacity

- Find a platform CXL window that has free capacity to map a new region
  and that is able to target the devices in the previous step.

- Allocate DPA according to the CXL specification rules of sequential
  enabling of decoders by id and when a device hosts multiple decoders
  make sure that lower-id decoders map lower HPA and higher-id decoders
  map higher HPA.

- Assign endpoint decoders to a region and validate that the switching
  topology supports the requested configuration. Recall that
  interleaving is governed by modulo or xormap math that constrains which
  device can support which positions in a given region interleave.

- Program all the decoders an all endpoints and participating switches
  to bring the new address range online.

Once the range is online then existing drivers like LIBNVDIMM or
device-dax can manage the memory range as if the ACPI BIOS had conveyed
its parameters at boot.

This patch kit is the result of significant amounts of path finding work
[2] and long discussions with Ben. Thank you Ben for all that work!
Where the patches in this kit go in a different design direction than
the RFC, the authorship is changed and a Co-developed-by is added mainly
so I get blamed for the bad decisions and not Ben. The major updates
from that last posting are:

- all CXL resources are reflected in full in iomem_resource

- host-physical-address (HPA) range allocation moves to a
  devm_request_free_mem_region() derivative

- locking moves to two global rwsems, one for DPA / endpoint decoders
  and one for HPA / regions.

- the existing port scanning path is augmented to cache more topology
  information rather than recreate it at region creation time

[2]: https://lore.kernel.org/r/20220413183720.2444089-1-ben.widawsky@intel.com

---

Ben Widawsky (4):
      cxl/hdm: Add sysfs attributes for interleave ways + granularity
      cxl/region: Add region creation support
      cxl/region: Add a 'uuid' attribute
      cxl/region: Add interleave geometry attributes

Dan Williams (24):
      Documentation/cxl: Use a double line break between entries
      cxl/core: Define a 'struct cxl_switch_decoder'
      cxl/acpi: Track CXL resources in iomem_resource
      cxl/core: Define a 'struct cxl_root_decoder'
      cxl/core: Define a 'struct cxl_endpoint_decoder'
      cxl/hdm: Enumerate allocated DPA
      cxl/hdm: Add 'mode' attribute to decoder objects
      cxl/hdm: Track next decoder to allocate
      cxl/hdm: Add support for allocating DPA to an endpoint decoder
      cxl/port: Record dport in endpoint references
      cxl/port: Record parent dport when adding ports
      cxl/port: Move 'cxl_ep' references to an xarray per port
      cxl/port: Move dport tracking to an xarray
      cxl/mem: Enumerate port targets before adding endpoints
      resource: Introduce alloc_free_mem_region()
      cxl/region: Allocate HPA capacity to regions
      cxl/region: Enable the assignment of endpoint decoders to regions
      cxl/acpi: Add a host-bridge index lookup mechanism
      cxl/region: Attach endpoint decoders
      cxl/region: Program target lists
      cxl/hdm: Commit decoder state to hardware
      cxl/region: Add region driver boiler plate
      cxl/pmem: Fix offline_nvdimm_bus() to offline by bridge
      cxl/region: Introduce cxl_pmem_region objects


 Documentation/ABI/testing/sysfs-bus-cxl         |  213 +++
 Documentation/driver-api/cxl/memory-devices.rst |   11 
 drivers/cxl/Kconfig                             |    8 
 drivers/cxl/acpi.c                              |  185 ++
 drivers/cxl/core/Makefile                       |    1 
 drivers/cxl/core/core.h                         |   49 +
 drivers/cxl/core/hdm.c                          |  623 +++++++-
 drivers/cxl/core/pmem.c                         |    4 
 drivers/cxl/core/port.c                         |  669 ++++++--
 drivers/cxl/core/region.c                       | 1830 +++++++++++++++++++++++
 drivers/cxl/cxl.h                               |  263 +++
 drivers/cxl/cxlmem.h                            |   18 
 drivers/cxl/mem.c                               |   32 
 drivers/cxl/pmem.c                              |  259 +++
 drivers/nvdimm/region_devs.c                    |   28 
 include/linux/ioport.h                          |    3 
 include/linux/libnvdimm.h                       |    5 
 kernel/resource.c                               |  185 ++
 mm/Kconfig                                      |    5 
 tools/testing/cxl/Kbuild                        |    1 
 tools/testing/cxl/test/cxl.c                    |   75 +
 21 files changed, 4156 insertions(+), 311 deletions(-)
 create mode 100644 drivers/cxl/core/region.c

base-commit: b060edfd8cdd52bc8648392500bf152a8dd6d4c5

[PATCH v2 01/28] Documentation/cxl: Use a double line break between entries

[Dan Williams]

Make it easier to read delineations between the "Description" line
break, new paragraph line breaks, and new entries.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 1fd5984b6158..16d9ffa94bbd 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -7,6 +7,7 @@ Description:
 		all descendant memdevs for unbind. Writing '1' to this attribute
 		flushes that work.
 
+
 What:		/sys/bus/cxl/devices/memX/firmware_version
 Date:		December, 2020
 KernelVersion:	v5.12
@@ -16,6 +17,7 @@ Description:
 		Memory Device Output Payload in the CXL-2.0
 		specification.
 
+
 What:		/sys/bus/cxl/devices/memX/ram/size
 Date:		December, 2020
 KernelVersion:	v5.12
@@ -25,6 +27,7 @@ Description:
 		identically named field in the Identify Memory Device Output
 		Payload in the CXL-2.0 specification.
 
+
 What:		/sys/bus/cxl/devices/memX/pmem/size
 Date:		December, 2020
 KernelVersion:	v5.12
@@ -34,6 +37,7 @@ Description:
 		identically named field in the Identify Memory Device Output
 		Payload in the CXL-2.0 specification.
 
+
 What:		/sys/bus/cxl/devices/memX/serial
 Date:		January, 2022
 KernelVersion:	v5.18
@@ -43,6 +47,7 @@ Description:
 		capability. Mandatory for CXL devices, see CXL 2.0 8.1.12.2
 		Memory Device PCIe Capabilities and Extended Capabilities.
 
+
 What:		/sys/bus/cxl/devices/memX/numa_node
 Date:		January, 2022
 KernelVersion:	v5.18
@@ -52,6 +57,7 @@ Description:
 		host PCI device for this memory device, emit the CPU node
 		affinity for this device.
 
+
 What:		/sys/bus/cxl/devices/*/devtype
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -61,6 +67,7 @@ Description:
 		mirrors the same value communicated in the DEVTYPE environment
 		variable for uevents for devices on the "cxl" bus.
 
+
 What:		/sys/bus/cxl/devices/*/modalias
 Date:		December, 2021
 KernelVersion:	v5.18
@@ -70,6 +77,7 @@ Description:
 		mirrors the same value communicated in the MODALIAS environment
 		variable for uevents for devices on the "cxl" bus.
 
+
 What:		/sys/bus/cxl/devices/portX/uport
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -81,6 +89,7 @@ Description:
 		the CXL portX object to the device that published the CXL port
 		capability.
 
+
 What:		/sys/bus/cxl/devices/portX/dportY
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -94,6 +103,7 @@ Description:
 		integer reflects the hardware port unique-id used in the
 		hardware decoder target list.
 
+
 What:		/sys/bus/cxl/devices/decoderX.Y
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -106,6 +116,7 @@ Description:
 		cxl_port container of this decoder, and 'Y' represents the
 		instance id of a given decoder resource.
 
+
 What:		/sys/bus/cxl/devices/decoderX.Y/{start,size}
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -120,6 +131,7 @@ Description:
 		and dynamically updates based on the active memory regions in
 		that address space.
 
+
 What:		/sys/bus/cxl/devices/decoderX.Y/locked
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -132,6 +144,7 @@ Description:
 		secondary bus reset, of the PCIe bridge that provides the bus
 		for this decoders uport, unlocks / resets the decoder.
 
+
 What:		/sys/bus/cxl/devices/decoderX.Y/target_list
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -142,6 +155,7 @@ Description:
 		configured interleave order of the decoder's dport instances.
 		Each entry in the list is a dport id.
 
+
 What:		/sys/bus/cxl/devices/decoderX.Y/cap_{pmem,ram,type2,type3}
 Date:		June, 2021
 KernelVersion:	v5.14
@@ -154,6 +168,7 @@ Description:
 		memory, volatile memory, accelerator memory, and / or expander
 		memory may be mapped behind this decoder's memory window.
 
+
 What:		/sys/bus/cxl/devices/decoderX.Y/target_type
 Date:		June, 2021
 KernelVersion:	v5.14

[PATCH v2 02/28] cxl/core: Define a 'struct cxl_switch_decoder'

[Dan Williams]

Currently 'struct cxl_decoder' contains the superset of attributes
needed for all decoder types. Before more type-specific attributes are
added to the common definition, reorganize 'struct cxl_decoder' into type
specific objects.

This patch, the first of three, factors out a cxl_switch_decoder type.
See the new kdoc for what a 'struct cxl_switch_decoder' represents in a
CXL topology.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/acpi.c           |    4 +
 drivers/cxl/core/hdm.c       |   33 +++++--
 drivers/cxl/core/port.c      |  188 ++++++++++++++++++++++++++++--------------
 drivers/cxl/cxl.h            |   30 +++++--
 tools/testing/cxl/test/cxl.c |   23 ++++-
 5 files changed, 189 insertions(+), 89 deletions(-)

diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
index 541fc0b28b8f..62bf22ffb7aa 100644
--- a/drivers/cxl/acpi.c
+++ b/drivers/cxl/acpi.c
@@ -81,6 +81,7 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	int target_map[CXL_DECODER_MAX_INTERLEAVE];
 	struct cxl_cfmws_context *ctx = arg;
 	struct cxl_port *root_port = ctx->root_port;
+	struct cxl_switch_decoder *cxlsd;
 	struct device *dev = ctx->dev;
 	struct acpi_cedt_cfmws *cfmws;
 	struct cxl_decoder *cxld;
@@ -106,10 +107,11 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	for (i = 0; i < ways; i++)
 		target_map[i] = cfmws->interleave_targets[i];
 
-	cxld = cxl_root_decoder_alloc(root_port, ways);
+	cxlsd = cxl_root_decoder_alloc(root_port, ways);
 	if (IS_ERR(cxld))
 		return 0;
 
+	cxld = &cxlsd->cxld;
 	cxld->flags = cfmws_to_decoder_flags(cfmws->restrictions);
 	cxld->target_type = CXL_DECODER_EXPANDER;
 	cxld->hpa_range = (struct range) {
diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index c524e772fdae..2f10d42798de 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -49,20 +49,20 @@ static int add_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
  */
 int devm_cxl_add_passthrough_decoder(struct cxl_port *port)
 {
-	struct cxl_decoder *cxld;
+	struct cxl_switch_decoder *cxlsd;
 	struct cxl_dport *dport;
 	int single_port_map[1];
 
-	cxld = cxl_switch_decoder_alloc(port, 1);
-	if (IS_ERR(cxld))
-		return PTR_ERR(cxld);
+	cxlsd = cxl_switch_decoder_alloc(port, 1);
+	if (IS_ERR(cxlsd))
+		return PTR_ERR(cxlsd);
 
 	device_lock_assert(&port->dev);
 
 	dport = list_first_entry(&port->dports, typeof(*dport), list);
 	single_port_map[0] = dport->port_id;
 
-	return add_hdm_decoder(port, cxld, single_port_map);
+	return add_hdm_decoder(port, &cxlsd->cxld, single_port_map);
 }
 EXPORT_SYMBOL_NS_GPL(devm_cxl_add_passthrough_decoder, CXL);
 
@@ -255,14 +255,23 @@ int devm_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 		int rc, target_count = cxlhdm->target_count;
 		struct cxl_decoder *cxld;
 
-		if (is_cxl_endpoint(port))
+		if (is_cxl_endpoint(port)) {
 			cxld = cxl_endpoint_decoder_alloc(port);
-		else
-			cxld = cxl_switch_decoder_alloc(port, target_count);
-		if (IS_ERR(cxld)) {
-			dev_warn(&port->dev,
-				 "Failed to allocate the decoder\n");
-			return PTR_ERR(cxld);
+			if (IS_ERR(cxld)) {
+				dev_warn(&port->dev,
+					 "Failed to allocate the decoder\n");
+				return PTR_ERR(cxld);
+			}
+		} else {
+			struct cxl_switch_decoder *cxlsd;
+
+			cxlsd = cxl_switch_decoder_alloc(port, target_count);
+			if (IS_ERR(cxlsd)) {
+				dev_warn(&port->dev,
+					 "Failed to allocate the decoder\n");
+				return PTR_ERR(cxlsd);
+			}
+			cxld = &cxlsd->cxld;
 		}
 
 		rc = init_hdm_decoder(port, cxld, target_map, hdm, i);
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index f62c0a6e17ea..27a2a6b839aa 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -120,20 +120,21 @@ static ssize_t target_type_show(struct device *dev,
 }
 static DEVICE_ATTR_RO(target_type);
 
-static ssize_t emit_target_list(struct cxl_decoder *cxld, char *buf)
+static ssize_t emit_target_list(struct cxl_switch_decoder *cxlsd, char *buf)
 {
+	struct cxl_decoder *cxld = &cxlsd->cxld;
 	ssize_t offset = 0;
 	int i, rc = 0;
 
 	for (i = 0; i < cxld->interleave_ways; i++) {
-		struct cxl_dport *dport = cxld->target[i];
+		struct cxl_dport *dport = cxlsd->target[i];
 		struct cxl_dport *next = NULL;
 
 		if (!dport)
 			break;
 
 		if (i + 1 < cxld->interleave_ways)
-			next = cxld->target[i + 1];
+			next = cxlsd->target[i + 1];
 		rc = sysfs_emit_at(buf, offset, "%d%s", dport->port_id,
 				   next ? "," : "");
 		if (rc < 0)
@@ -144,18 +145,20 @@ static ssize_t emit_target_list(struct cxl_decoder *cxld, char *buf)
 	return offset;
 }
 
+static struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev);
+
 static ssize_t target_list_show(struct device *dev,
 				struct device_attribute *attr, char *buf)
 {
-	struct cxl_decoder *cxld = to_cxl_decoder(dev);
+	struct cxl_switch_decoder *cxlsd = to_cxl_switch_decoder(dev);
 	ssize_t offset;
 	unsigned int seq;
 	int rc;
 
 	do {
-		seq = read_seqbegin(&cxld->target_lock);
-		rc = emit_target_list(cxld, buf);
-	} while (read_seqretry(&cxld->target_lock, seq));
+		seq = read_seqbegin(&cxlsd->target_lock);
+		rc = emit_target_list(cxlsd, buf);
+	} while (read_seqretry(&cxlsd->target_lock, seq));
 
 	if (rc < 0)
 		return rc;
@@ -233,14 +236,28 @@ static const struct attribute_group *cxl_decoder_endpoint_attribute_groups[] = {
 	NULL,
 };
 
+static void __cxl_decoder_release(struct cxl_decoder *cxld)
+{
+	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
+
+	ida_free(&port->decoder_ida, cxld->id);
+	put_device(&port->dev);
+}
+
 static void cxl_decoder_release(struct device *dev)
 {
 	struct cxl_decoder *cxld = to_cxl_decoder(dev);
-	struct cxl_port *port = to_cxl_port(dev->parent);
 
-	ida_free(&port->decoder_ida, cxld->id);
+	__cxl_decoder_release(cxld);
 	kfree(cxld);
-	put_device(&port->dev);
+}
+
+static void cxl_switch_decoder_release(struct device *dev)
+{
+	struct cxl_switch_decoder *cxlsd = to_cxl_switch_decoder(dev);
+
+	__cxl_decoder_release(&cxlsd->cxld);
+	kfree(cxlsd);
 }
 
 static const struct device_type cxl_decoder_endpoint_type = {
@@ -251,13 +268,13 @@ static const struct device_type cxl_decoder_endpoint_type = {
 
 static const struct device_type cxl_decoder_switch_type = {
 	.name = "cxl_decoder_switch",
-	.release = cxl_decoder_release,
+	.release = cxl_switch_decoder_release,
 	.groups = cxl_decoder_switch_attribute_groups,
 };
 
 static const struct device_type cxl_decoder_root_type = {
 	.name = "cxl_decoder_root",
-	.release = cxl_decoder_release,
+	.release = cxl_switch_decoder_release,
 	.groups = cxl_decoder_root_attribute_groups,
 };
 
@@ -272,15 +289,29 @@ bool is_root_decoder(struct device *dev)
 }
 EXPORT_SYMBOL_NS_GPL(is_root_decoder, CXL);
 
+static bool is_switch_decoder(struct device *dev)
+{
+	return is_root_decoder(dev) || dev->type == &cxl_decoder_switch_type;
+}
+
 struct cxl_decoder *to_cxl_decoder(struct device *dev)
 {
-	if (dev_WARN_ONCE(dev, dev->type->release != cxl_decoder_release,
+	if (dev_WARN_ONCE(dev,
+			  !is_switch_decoder(dev) && !is_endpoint_decoder(dev),
 			  "not a cxl_decoder device\n"))
 		return NULL;
 	return container_of(dev, struct cxl_decoder, dev);
 }
 EXPORT_SYMBOL_NS_GPL(to_cxl_decoder, CXL);
 
+static struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev)
+{
+	if (dev_WARN_ONCE(dev, !is_switch_decoder(dev),
+			  "not a cxl_switch_decoder device\n"))
+		return NULL;
+	return container_of(dev, struct cxl_switch_decoder, cxld.dev);
+}
+
 static void cxl_ep_release(struct cxl_ep *ep)
 {
 	if (!ep)
@@ -1146,7 +1177,7 @@ struct cxl_dport *cxl_find_dport_by_dev(struct cxl_port *port,
 }
 EXPORT_SYMBOL_NS_GPL(cxl_find_dport_by_dev, CXL);
 
-static int decoder_populate_targets(struct cxl_decoder *cxld,
+static int decoder_populate_targets(struct cxl_switch_decoder *cxlsd,
 				    struct cxl_port *port, int *target_map)
 {
 	int i, rc = 0;
@@ -1159,17 +1190,17 @@ static int decoder_populate_targets(struct cxl_decoder *cxld,
 	if (list_empty(&port->dports))
 		return -EINVAL;
 
-	write_seqlock(&cxld->target_lock);
-	for (i = 0; i < cxld->nr_targets; i++) {
+	write_seqlock(&cxlsd->target_lock);
+	for (i = 0; i < cxlsd->nr_targets; i++) {
 		struct cxl_dport *dport = find_dport(port, target_map[i]);
 
 		if (!dport) {
 			rc = -ENXIO;
 			break;
 		}
-		cxld->target[i] = dport;
+		cxlsd->target[i] = dport;
 	}
-	write_sequnlock(&cxld->target_lock);
+	write_sequnlock(&cxlsd->target_lock);
 
 	return rc;
 }
@@ -1177,56 +1208,34 @@ static int decoder_populate_targets(struct cxl_decoder *cxld,
 static struct lock_class_key cxl_decoder_key;
 
 /**
- * cxl_decoder_alloc - Allocate a new CXL decoder
+ * cxl_decoder_alloc - Common decoder setup / initialization
  * @port: owning port of this decoder
- * @nr_targets: downstream targets accessible by this decoder. All upstream
- *		ports and root ports must have at least 1 target. Endpoint
- *		devices will have 0 targets. Callers wishing to register an
- *		endpoint device should specify 0.
- *
- * A port should contain one or more decoders. Each of those decoders enable
- * some address space for CXL.mem utilization. A decoder is expected to be
- * configured by the caller before registering.
+ * @cxld: common decoder properties to initialize
  *
- * Return: A new cxl decoder to be registered by cxl_decoder_add(). The decoder
- *	   is initialized to be a "passthrough" decoder.
+ * A port may contain one or more decoders. Each of those decoders
+ * enable some address space for CXL.mem utilization. A decoder is
+ * expected to be configured by the caller before registering via
+ * cxl_decoder_add()
  */
-static struct cxl_decoder *cxl_decoder_alloc(struct cxl_port *port,
-					     unsigned int nr_targets)
+static int cxl_decoder_init(struct cxl_port *port, struct cxl_decoder *cxld)
 {
-	struct cxl_decoder *cxld;
 	struct device *dev;
-	int rc = 0;
-
-	if (nr_targets > CXL_DECODER_MAX_INTERLEAVE)
-		return ERR_PTR(-EINVAL);
-
-	cxld = kzalloc(struct_size(cxld, target, nr_targets), GFP_KERNEL);
-	if (!cxld)
-		return ERR_PTR(-ENOMEM);
+	int rc;
 
 	rc = ida_alloc(&port->decoder_ida, GFP_KERNEL);
 	if (rc < 0)
-		goto err;
+		return rc;
 
 	/* need parent to stick around to release the id */
 	get_device(&port->dev);
 	cxld->id = rc;
 
-	cxld->nr_targets = nr_targets;
-	seqlock_init(&cxld->target_lock);
 	dev = &cxld->dev;
 	device_initialize(dev);
 	lockdep_set_class(&dev->mutex, &cxl_decoder_key);
 	device_set_pm_not_required(dev);
 	dev->parent = &port->dev;
 	dev->bus = &cxl_bus_type;
-	if (is_cxl_root(port))
-		cxld->dev.type = &cxl_decoder_root_type;
-	else if (is_cxl_endpoint(port))
-		cxld->dev.type = &cxl_decoder_endpoint_type;
-	else
-		cxld->dev.type = &cxl_decoder_switch_type;
 
 	/* Pre initialize an "empty" decoder */
 	cxld->interleave_ways = 1;
@@ -1237,10 +1246,19 @@ static struct cxl_decoder *cxl_decoder_alloc(struct cxl_port *port,
 		.end = -1,
 	};
 
-	return cxld;
-err:
-	kfree(cxld);
-	return ERR_PTR(rc);
+	return 0;
+}
+
+static int cxl_switch_decoder_init(struct cxl_port *port,
+				   struct cxl_switch_decoder *cxlsd,
+				   int nr_targets)
+{
+	if (nr_targets > CXL_DECODER_MAX_INTERLEAVE)
+		return -EINVAL;
+
+	cxlsd->nr_targets = nr_targets;
+	seqlock_init(&cxlsd->target_lock);
+	return cxl_decoder_init(port, &cxlsd->cxld);
 }
 
 /**
@@ -1253,13 +1271,29 @@ static struct cxl_decoder *cxl_decoder_alloc(struct cxl_port *port,
  * firmware description of CXL resources into a CXL standard decode
  * topology.
  */
-struct cxl_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
-					   unsigned int nr_targets)
+struct cxl_switch_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
+						  unsigned int nr_targets)
 {
+	struct cxl_switch_decoder *cxlsd;
+	struct cxl_decoder *cxld;
+	int rc;
+
 	if (!is_cxl_root(port))
 		return ERR_PTR(-EINVAL);
 
-	return cxl_decoder_alloc(port, nr_targets);
+	cxlsd = kzalloc(struct_size(cxlsd, target, nr_targets), GFP_KERNEL);
+	if (!cxlsd)
+		return ERR_PTR(-ENOMEM);
+
+	rc = cxl_switch_decoder_init(port, cxlsd, nr_targets);
+	if (rc) {
+		kfree(cxlsd);
+		return ERR_PTR(rc);
+	}
+
+	cxld = &cxlsd->cxld;
+	cxld->dev.type = &cxl_decoder_root_type;
+	return cxlsd;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_root_decoder_alloc, CXL);
 
@@ -1274,13 +1308,29 @@ EXPORT_SYMBOL_NS_GPL(cxl_root_decoder_alloc, CXL);
  * that sit between Switch Upstream Ports / Switch Downstream Ports and
  * Host Bridges / Root Ports.
  */
-struct cxl_decoder *cxl_switch_decoder_alloc(struct cxl_port *port,
-					     unsigned int nr_targets)
+struct cxl_switch_decoder *cxl_switch_decoder_alloc(struct cxl_port *port,
+						    unsigned int nr_targets)
 {
+	struct cxl_switch_decoder *cxlsd;
+	struct cxl_decoder *cxld;
+	int rc;
+
 	if (is_cxl_root(port) || is_cxl_endpoint(port))
 		return ERR_PTR(-EINVAL);
 
-	return cxl_decoder_alloc(port, nr_targets);
+	cxlsd = kzalloc(struct_size(cxlsd, target, nr_targets), GFP_KERNEL);
+	if (!cxlsd)
+		return ERR_PTR(-ENOMEM);
+
+	rc = cxl_switch_decoder_init(port, cxlsd, nr_targets);
+	if (rc) {
+		kfree(cxlsd);
+		return ERR_PTR(rc);
+	}
+
+	cxld = &cxlsd->cxld;
+	cxld->dev.type = &cxl_decoder_switch_type;
+	return cxlsd;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_switch_decoder_alloc, CXL);
 
@@ -1292,10 +1342,24 @@ EXPORT_SYMBOL_NS_GPL(cxl_switch_decoder_alloc, CXL);
  */
 struct cxl_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port)
 {
+	struct cxl_decoder *cxld;
+	int rc;
+
 	if (!is_cxl_endpoint(port))
 		return ERR_PTR(-EINVAL);
 
-	return cxl_decoder_alloc(port, 0);
+	cxld = kzalloc(sizeof(*cxld), GFP_KERNEL);
+	if (!cxld)
+		return ERR_PTR(-ENOMEM);
+
+	rc = cxl_decoder_init(port, cxld);
+	if (rc)	 {
+		kfree(cxld);
+		return ERR_PTR(rc);
+	}
+
+	cxld->dev.type = &cxl_decoder_endpoint_type;
+	return cxld;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_endpoint_decoder_alloc, CXL);
 
@@ -1337,7 +1401,9 @@ int cxl_decoder_add_locked(struct cxl_decoder *cxld, int *target_map)
 
 	port = to_cxl_port(cxld->dev.parent);
 	if (!is_endpoint_decoder(dev)) {
-		rc = decoder_populate_targets(cxld, port, target_map);
+		struct cxl_switch_decoder *cxlsd = to_cxl_switch_decoder(dev);
+
+		rc = decoder_populate_targets(cxlsd, port, target_map);
 		if (rc && (cxld->flags & CXL_DECODER_F_ENABLE)) {
 			dev_err(&port->dev,
 				"Failed to populate active decoder targets\n");
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 570bd9f8141b..0289c06ec72c 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -220,7 +220,7 @@ enum cxl_decoder_type {
 #define CXL_DECODER_MAX_INTERLEAVE 16
 
 /**
- * struct cxl_decoder - CXL address range decode configuration
+ * struct cxl_decoder - Common CXL HDM Decoder Attributes
  * @dev: this decoder's device
  * @id: kernel device name id
  * @hpa_range: Host physical address range mapped by this decoder
@@ -228,9 +228,6 @@ enum cxl_decoder_type {
  * @interleave_granularity: data stride per dport
  * @target_type: accelerator vs expander (type2 vs type3) selector
  * @flags: memory type capabilities and locking
- * @target_lock: coordinate coherent reads of the target list
- * @nr_targets: number of elements in @target
- * @target: active ordered target list in current decoder configuration
  */
 struct cxl_decoder {
 	struct device dev;
@@ -240,6 +237,23 @@ struct cxl_decoder {
 	int interleave_granularity;
 	enum cxl_decoder_type target_type;
 	unsigned long flags;
+};
+
+/**
+ * struct cxl_switch_decoder - Switch specific CXL HDM Decoder
+ * @cxld: base cxl_decoder object
+ * @target_lock: coordinate coherent reads of the target list
+ * @nr_targets: number of elements in @target
+ * @target: active ordered target list in current decoder configuration
+ *
+ * The 'switch' decoder type represents the decoder instances of cxl_port's that
+ * route from the root of a CXL memory decode topology to the endpoints. They
+ * come in two flavors, root-level decoders, statically defined by platform
+ * firmware, and mid-level decoders, where interleave-granularity,
+ * interleave-width, and the target list are mutable.
+ */
+struct cxl_switch_decoder {
+	struct cxl_decoder cxld;
 	seqlock_t target_lock;
 	int nr_targets;
 	struct cxl_dport *target[];
@@ -364,10 +378,10 @@ struct cxl_dport *cxl_find_dport_by_dev(struct cxl_port *port,
 struct cxl_decoder *to_cxl_decoder(struct device *dev);
 bool is_root_decoder(struct device *dev);
 bool is_endpoint_decoder(struct device *dev);
-struct cxl_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
-					   unsigned int nr_targets);
-struct cxl_decoder *cxl_switch_decoder_alloc(struct cxl_port *port,
-					     unsigned int nr_targets);
+struct cxl_switch_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
+						  unsigned int nr_targets);
+struct cxl_switch_decoder *cxl_switch_decoder_alloc(struct cxl_port *port,
+						    unsigned int nr_targets);
 int cxl_decoder_add(struct cxl_decoder *cxld, int *target_map);
 struct cxl_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port);
 int cxl_decoder_add_locked(struct cxl_decoder *cxld, int *target_map);
diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
index 6e086fbc5c5b..7991ddc6e562 100644
--- a/tools/testing/cxl/test/cxl.c
+++ b/tools/testing/cxl/test/cxl.c
@@ -451,14 +451,23 @@ static int mock_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 		struct cxl_decoder *cxld;
 		int rc;
 
-		if (target_count)
-			cxld = cxl_switch_decoder_alloc(port, target_count);
-		else
+		if (target_count) {
+			struct cxl_switch_decoder *cxlsd;
+
+			cxlsd = cxl_switch_decoder_alloc(port, target_count);
+			if (IS_ERR(cxlsd)) {
+				dev_warn(&port->dev,
+					 "Failed to allocate the decoder\n");
+				return PTR_ERR(cxlsd);
+			}
+			cxld = &cxlsd->cxld;
+		} else {
 			cxld = cxl_endpoint_decoder_alloc(port);
-		if (IS_ERR(cxld)) {
-			dev_warn(&port->dev,
-				 "Failed to allocate the decoder\n");
-			return PTR_ERR(cxld);
+			if (IS_ERR(cxld)) {
+				dev_warn(&port->dev,
+					 "Failed to allocate the decoder\n");
+				return PTR_ERR(cxld);
+			}
 		}
 
 		cxld->hpa_range = (struct range) {

[PATCH v2 03/28] cxl/acpi: Track CXL resources in iomem_resource

[Dan Williams]

Recall that CXL capable address ranges, on ACPI platforms, are published
in the CEDT.CFMWS (CXL Early Discovery Table: CXL Fixed Memory Window
Structures). These windows represent both the actively mapped capacity
and the potential address space that can be dynamically assigned to a
new CXL decode configuration (region / interleave-set).

CXL endpoints like DDR DIMMs can be mapped at any physical address
including 0 and legacy ranges.

There is an expectation and requirement that the /proc/iomem interface
and the iomem_resource tree in the kernel reflect the full set of
platform address ranges. I.e. that every address range that platform
firmware and bus drivers enumerate be reflected as an iomem_resource
entry. The hard requirement to do this for CXL arises from the fact that
facilities like CONFIG_DEVICE_PRIVATE expect to be able to treat empty
iomem_resource ranges as free for software to use as proxy address
space. Without CXL publishing its potential address ranges in
iomem_resource, the CONFIG_DEVICE_PRIVATE mechanism may inadvertently
steal capacity reserved for runtime provisioning of new CXL regions.

So, iomem_resource needs to know about both active and potential CXL
resource ranges. The active CXL resources might already be reflected in
iomem_resource as "System RAM". insert_resource_expand_to_fit() handles
re-parenting "System RAM" underneath a CXL window.

The "_expand_to_fit()" behavior handles cases where a CXL window is not
a strict superset of an existing entry in the iomem_resource tree. The
"_expand_to_fit()" behavior is acceptable from the perspective of
resource allocation. The expansion happens because a conflicting
resource range is already populated, which means the resource boundary
expansion does not result in any additional free CXL address space being
made available. CXL address space allocation is always bounded by the
orginal unexpanded address range.

However, the potential for expansion does mean that something like
walk_iomem_res_desc(IORES_DESC_CXL...) can only return fuzzy answers on
corner case platforms that cause the resource tree to expand a CXL
window resource over a range that is not decoded by CXL. This would be
an odd platform configuration, but if it becomes a problem in practice
the CXL subsytem could just publish an API that returns definitive
answers.

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/acpi.c     |  144 +++++++++++++++++++++++++++++++++++++++++++++++-
 include/linux/ioport.h |    1 
 kernel/resource.c      |    7 ++
 3 files changed, 149 insertions(+), 3 deletions(-)

diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
index 62bf22ffb7aa..e2b6cbd04846 100644
--- a/drivers/cxl/acpi.c
+++ b/drivers/cxl/acpi.c
@@ -73,6 +73,8 @@ static int cxl_acpi_cfmws_verify(struct device *dev,
 struct cxl_cfmws_context {
 	struct device *dev;
 	struct cxl_port *root_port;
+	struct resource *cxl_res;
+	int id;
 };
 
 static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
@@ -81,11 +83,13 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	int target_map[CXL_DECODER_MAX_INTERLEAVE];
 	struct cxl_cfmws_context *ctx = arg;
 	struct cxl_port *root_port = ctx->root_port;
+	struct resource *cxl_res = ctx->cxl_res;
 	struct cxl_switch_decoder *cxlsd;
 	struct device *dev = ctx->dev;
 	struct acpi_cedt_cfmws *cfmws;
 	struct cxl_decoder *cxld;
 	unsigned int ways, i, ig;
+	struct resource *res;
 	int rc;
 
 	cfmws = (struct acpi_cedt_cfmws *) header;
@@ -107,6 +111,23 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	for (i = 0; i < ways; i++)
 		target_map[i] = cfmws->interleave_targets[i];
 
+	res = kzalloc(sizeof(*res), GFP_KERNEL);
+	if (!res)
+		return -ENOMEM;
+
+	res->name = kasprintf(GFP_KERNEL, "CXL Window %d", ctx->id++);
+	if (!res->name)
+		goto err_name;
+
+	res->start = cfmws->base_hpa;
+	res->end = cfmws->base_hpa + cfmws->window_size - 1;
+	res->flags = IORESOURCE_MEM;
+
+	/* add to the local resource tracking to establish a sort order */
+	rc = insert_resource(cxl_res, res);
+	if (rc)
+		goto err_insert;
+
 	cxlsd = cxl_root_decoder_alloc(root_port, ways);
 	if (IS_ERR(cxld))
 		return 0;
@@ -115,8 +136,8 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	cxld->flags = cfmws_to_decoder_flags(cfmws->restrictions);
 	cxld->target_type = CXL_DECODER_EXPANDER;
 	cxld->hpa_range = (struct range) {
-		.start = cfmws->base_hpa,
-		.end = cfmws->base_hpa + cfmws->window_size - 1,
+		.start = res->start,
+		.end = res->end,
 	};
 	cxld->interleave_ways = ways;
 	cxld->interleave_granularity = ig;
@@ -137,6 +158,12 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 		cxld->hpa_range.start, cxld->hpa_range.end);
 
 	return 0;
+
+err_insert:
+	kfree(res->name);
+err_name:
+	kfree(res);
+	return -ENOMEM;
 }
 
 __mock struct acpi_device *to_cxl_host_bridge(struct device *host,
@@ -291,9 +318,101 @@ static void cxl_acpi_lock_reset_class(void *dev)
 	device_lock_reset_class(dev);
 }
 
+static void del_cxl_resource(struct resource *res)
+{
+	kfree(res->name);
+	kfree(res);
+}
+
+static void cxl_set_public_resource(struct resource *priv, struct resource *pub)
+{
+	priv->desc = (unsigned long) pub;
+}
+
+static struct resource *cxl_get_public_resource(struct resource *priv)
+{
+	return (struct resource *) priv->desc;
+}
+
+static void remove_cxl_resources(void *data)
+{
+	struct resource *res, *next, *cxl = data;
+
+	for (res = cxl->child; res; res = next) {
+		struct resource *victim = cxl_get_public_resource(res);
+
+		next = res->sibling;
+		remove_resource(res);
+
+		if (victim) {
+			remove_resource(victim);
+			kfree(victim);
+		}
+
+		del_cxl_resource(res);
+	}
+}
+
+/**
+ * add_cxl_resources() - reflect CXL fixed memory windows in iomem_resource
+ * @cxl_res: A standalone resource tree where each CXL window is a sibling
+ *
+ * Walk each CXL window in @cxl_res and add it to iomem_resource potentially
+ * expanding its boundaries to ensure that any conflicting resources become
+ * children. If a window is expanded it may then conflict with a another window
+ * entry and require the window to be truncated or trimmed. Consider this
+ * situation:
+ *
+ * |-- "CXL Window 0" --||----- "CXL Window 1" -----|
+ * |--------------- "System RAM" -------------|
+ *
+ * ...where platform firmware has established as System RAM resource across 2
+ * windows, but has left some portion of window 1 for dynamic CXL region
+ * provisioning. In this case "Window 0" will span the entirety of the "System
+ * RAM" span, and "CXL Window 1" is truncated to the remaining tail past the end
+ * of that "System RAM" resource.
+ */
+static int add_cxl_resources(struct resource *cxl_res)
+{
+	struct resource *res, *new, *next;
+
+	for (res = cxl_res->child; res; res = next) {
+		new = kzalloc(sizeof(*new), GFP_KERNEL);
+		if (!new)
+			return -ENOMEM;
+		new->name = res->name;
+		new->start = res->start;
+		new->end = res->end;
+		new->flags = IORESOURCE_MEM;
+		new->desc = IORES_DESC_CXL;
+
+		/*
+		 * Record the public resource in the private cxl_res tree for
+		 * later removal.
+		 */
+		cxl_set_public_resource(res, new);
+
+		insert_resource_expand_to_fit(&iomem_resource, new);
+
+		next = res->sibling;
+		while (next && resource_overlaps(new, next)) {
+			if (resource_contains(new, next)) {
+				struct resource *_next = next->sibling;
+
+				remove_resource(next);
+				del_cxl_resource(next);
+				next = _next;
+			} else
+				next->start = new->end + 1;
+		}
+	}
+	return 0;
+}
+
 static int cxl_acpi_probe(struct platform_device *pdev)
 {
 	int rc;
+	struct resource *cxl_res;
 	struct cxl_port *root_port;
 	struct device *host = &pdev->dev;
 	struct acpi_device *adev = ACPI_COMPANION(host);
@@ -305,6 +424,14 @@ static int cxl_acpi_probe(struct platform_device *pdev)
 	if (rc)
 		return rc;
 
+	cxl_res = devm_kzalloc(host, sizeof(*cxl_res), GFP_KERNEL);
+	if (!cxl_res)
+		return -ENOMEM;
+	cxl_res->name = "CXL mem";
+	cxl_res->start = 0;
+	cxl_res->end = -1;
+	cxl_res->flags = IORESOURCE_MEM;
+
 	root_port = devm_cxl_add_port(host, host, CXL_RESOURCE_NONE, NULL);
 	if (IS_ERR(root_port))
 		return PTR_ERR(root_port);
@@ -315,11 +442,22 @@ static int cxl_acpi_probe(struct platform_device *pdev)
 	if (rc < 0)
 		return rc;
 
+	rc = devm_add_action_or_reset(host, remove_cxl_resources, cxl_res);
+	if (rc)
+		return rc;
+
 	ctx = (struct cxl_cfmws_context) {
 		.dev = host,
 		.root_port = root_port,
+		.cxl_res = cxl_res,
 	};
-	acpi_table_parse_cedt(ACPI_CEDT_TYPE_CFMWS, cxl_parse_cfmws, &ctx);
+	rc = acpi_table_parse_cedt(ACPI_CEDT_TYPE_CFMWS, cxl_parse_cfmws, &ctx);
+	if (rc < 0)
+		return -ENXIO;
+
+	rc = add_cxl_resources(cxl_res);
+	if (rc)
+		return rc;
 
 	/*
 	 * Root level scanned with host-bridge as dports, now scan host-bridges
diff --git a/include/linux/ioport.h b/include/linux/ioport.h
index ec5f71f7135b..79d1ad6d6275 100644
--- a/include/linux/ioport.h
+++ b/include/linux/ioport.h
@@ -141,6 +141,7 @@ enum {
 	IORES_DESC_DEVICE_PRIVATE_MEMORY	= 6,
 	IORES_DESC_RESERVED			= 7,
 	IORES_DESC_SOFT_RESERVED		= 8,
+	IORES_DESC_CXL				= 9,
 };
 
 /*
diff --git a/kernel/resource.c b/kernel/resource.c
index 34eaee179689..53a534db350e 100644
--- a/kernel/resource.c
+++ b/kernel/resource.c
@@ -891,6 +891,13 @@ void insert_resource_expand_to_fit(struct resource *root, struct resource *new)
 	}
 	write_unlock(&resource_lock);
 }
+/*
+ * Not for general consumption, only early boot memory map parsing, PCI
+ * resource discovery, and late discovery of CXL resources are expected
+ * to use this interface. The former are built-in and only the latter,
+ * CXL, is a module.
+ */
+EXPORT_SYMBOL_NS_GPL(insert_resource_expand_to_fit, CXL);
 
 /**
  * remove_resource - Remove a resource in the resource tree

[PATCH v2 04/28] cxl/core: Define a 'struct cxl_root_decoder'

[Dan Williams]

Previously the target routing specifics of switch decoders were factored
out of 'struct cxl_decoder' into 'struct cxl_switch_decoder'.

This patch, 2 of 3, adds a 'struct cxl_root_decoder' as a superset of a
switch decoder that also track the associated CXL window platform
resource.

Note that the reason the resource for a given root decoder needs to be
looked up after the fact (i.e. after cxl_parse_cfmws() and
add_cxl_resource()) is because add_cxl_resource() may have merged CXL
windows in order to keep them at the top of the resource tree / decode
hierarchy.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/acpi.c      |   40 ++++++++++++++++++++++++++++++++++++----
 drivers/cxl/core/port.c |   34 +++++++++++++++++++++++++++-------
 drivers/cxl/cxl.h       |   15 +++++++++++++--
 3 files changed, 76 insertions(+), 13 deletions(-)

diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
index e2b6cbd04846..8f021241699f 100644
--- a/drivers/cxl/acpi.c
+++ b/drivers/cxl/acpi.c
@@ -84,7 +84,7 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	struct cxl_cfmws_context *ctx = arg;
 	struct cxl_port *root_port = ctx->root_port;
 	struct resource *cxl_res = ctx->cxl_res;
-	struct cxl_switch_decoder *cxlsd;
+	struct cxl_root_decoder *cxlrd;
 	struct device *dev = ctx->dev;
 	struct acpi_cedt_cfmws *cfmws;
 	struct cxl_decoder *cxld;
@@ -128,11 +128,11 @@ static int cxl_parse_cfmws(union acpi_subtable_headers *header, void *arg,
 	if (rc)
 		goto err_insert;
 
-	cxlsd = cxl_root_decoder_alloc(root_port, ways);
-	if (IS_ERR(cxld))
+	cxlrd = cxl_root_decoder_alloc(root_port, ways);
+	if (IS_ERR(cxlrd))
 		return 0;
 
-	cxld = &cxlsd->cxld;
+	cxld = &cxlrd->cxlsd.cxld;
 	cxld->flags = cfmws_to_decoder_flags(cfmws->restrictions);
 	cxld->target_type = CXL_DECODER_EXPANDER;
 	cxld->hpa_range = (struct range) {
@@ -409,6 +409,32 @@ static int add_cxl_resources(struct resource *cxl_res)
 	return 0;
 }
 
+static int pair_cxl_resource(struct device *dev, void *data)
+{
+	struct resource *cxl_res = data;
+	struct resource *p;
+
+	if (!is_root_decoder(dev))
+		return 0;
+
+	for (p = cxl_res->child; p; p = p->sibling) {
+		struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
+		struct cxl_decoder *cxld = &cxlrd->cxlsd.cxld;
+		struct resource res = {
+			.start = cxld->hpa_range.start,
+			.end = cxld->hpa_range.end,
+			.flags = IORESOURCE_MEM,
+		};
+
+		if (resource_contains(p, &res)) {
+			cxlrd->res = cxl_get_public_resource(p);
+			break;
+		}
+	}
+
+	return 0;
+}
+
 static int cxl_acpi_probe(struct platform_device *pdev)
 {
 	int rc;
@@ -459,6 +485,12 @@ static int cxl_acpi_probe(struct platform_device *pdev)
 	if (rc)
 		return rc;
 
+	/*
+	 * Populate the root decoders with their related iomem resource,
+	 * if present
+	 */
+	device_for_each_child(&root_port->dev, cxl_res, pair_cxl_resource);
+
 	/*
 	 * Root level scanned with host-bridge as dports, now scan host-bridges
 	 * for their role as CXL uports to their CXL-capable PCIe Root Ports.
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 27a2a6b839aa..4953a1c7b245 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -260,6 +260,23 @@ static void cxl_switch_decoder_release(struct device *dev)
 	kfree(cxlsd);
 }
 
+struct cxl_root_decoder *to_cxl_root_decoder(struct device *dev)
+{
+	if (dev_WARN_ONCE(dev, !is_root_decoder(dev),
+			  "not a cxl_root_decoder device\n"))
+		return NULL;
+	return container_of(dev, struct cxl_root_decoder, cxlsd.cxld.dev);
+}
+EXPORT_SYMBOL_NS_GPL(to_cxl_root_decoder, CXL);
+
+static void cxl_root_decoder_release(struct device *dev)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
+
+	__cxl_decoder_release(&cxlrd->cxlsd.cxld);
+	kfree(cxlrd);
+}
+
 static const struct device_type cxl_decoder_endpoint_type = {
 	.name = "cxl_decoder_endpoint",
 	.release = cxl_decoder_release,
@@ -274,7 +291,7 @@ static const struct device_type cxl_decoder_switch_type = {
 
 static const struct device_type cxl_decoder_root_type = {
 	.name = "cxl_decoder_root",
-	.release = cxl_switch_decoder_release,
+	.release = cxl_root_decoder_release,
 	.groups = cxl_decoder_root_attribute_groups,
 };
 
@@ -1271,9 +1288,10 @@ static int cxl_switch_decoder_init(struct cxl_port *port,
  * firmware description of CXL resources into a CXL standard decode
  * topology.
  */
-struct cxl_switch_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
-						  unsigned int nr_targets)
+struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
+						unsigned int nr_targets)
 {
+	struct cxl_root_decoder *cxlrd;
 	struct cxl_switch_decoder *cxlsd;
 	struct cxl_decoder *cxld;
 	int rc;
@@ -1281,19 +1299,21 @@ struct cxl_switch_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
 	if (!is_cxl_root(port))
 		return ERR_PTR(-EINVAL);
 
-	cxlsd = kzalloc(struct_size(cxlsd, target, nr_targets), GFP_KERNEL);
-	if (!cxlsd)
+	cxlrd = kzalloc(struct_size(cxlrd, cxlsd.target, nr_targets),
+			GFP_KERNEL);
+	if (!cxlrd)
 		return ERR_PTR(-ENOMEM);
 
+	cxlsd = &cxlrd->cxlsd;
 	rc = cxl_switch_decoder_init(port, cxlsd, nr_targets);
 	if (rc) {
-		kfree(cxlsd);
+		kfree(cxlrd);
 		return ERR_PTR(rc);
 	}
 
 	cxld = &cxlsd->cxld;
 	cxld->dev.type = &cxl_decoder_root_type;
-	return cxlsd;
+	return cxlrd;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_root_decoder_alloc, CXL);
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 0289c06ec72c..ebdac8e7d181 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -260,6 +260,16 @@ struct cxl_switch_decoder {
 };
 
 
+/**
+ * struct cxl_root_decoder - Static platform CXL address decoder
+ * @res: host / parent resource for region allocations
+ * @cxlsd: base cxl switch decoder
+ */
+struct cxl_root_decoder {
+	struct resource *res;
+	struct cxl_switch_decoder cxlsd;
+};
+
 /**
  * enum cxl_nvdimm_brige_state - state machine for managing bus rescans
  * @CXL_NVB_NEW: Set at bridge create and after cxl_pmem_wq is destroyed
@@ -376,10 +386,11 @@ struct cxl_dport *cxl_find_dport_by_dev(struct cxl_port *port,
 					const struct device *dev);
 
 struct cxl_decoder *to_cxl_decoder(struct device *dev);
+struct cxl_root_decoder *to_cxl_root_decoder(struct device *dev);
 bool is_root_decoder(struct device *dev);
 bool is_endpoint_decoder(struct device *dev);
-struct cxl_switch_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
-						  unsigned int nr_targets);
+struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
+						unsigned int nr_targets);
 struct cxl_switch_decoder *cxl_switch_decoder_alloc(struct cxl_port *port,
 						    unsigned int nr_targets);
 int cxl_decoder_add(struct cxl_decoder *cxld, int *target_map);

[PATCH v2 05/28] cxl/core: Define a 'struct cxl_endpoint_decoder'

[Dan Williams]

Previously the target routing specifics of switch decoders and platform
CXL window resource tracking of root decoders were factored out of
'struct cxl_decoder'. While switch decoders translate from SPA to
downstream ports, endpoint decoders translate from SPA to DPA.

This patch, 3 of 3, adds a 'struct cxl_endpoint_decoder' that tracks an
endpoint-specific Device Physical Address (DPA) resource. For now this
just defines ->dpa_res, a follow-on patch will handle requesting DPA
resource ranges from a device-DPA resource tree.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/hdm.c       |    9 ++++++---
 drivers/cxl/core/port.c      |   31 +++++++++++++++++++++----------
 drivers/cxl/cxl.h            |   15 ++++++++++++++-
 tools/testing/cxl/test/cxl.c |   10 +++++++---
 4 files changed, 48 insertions(+), 17 deletions(-)

diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 2f10d42798de..650363d5272f 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -256,12 +256,15 @@ int devm_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 		struct cxl_decoder *cxld;
 
 		if (is_cxl_endpoint(port)) {
-			cxld = cxl_endpoint_decoder_alloc(port);
-			if (IS_ERR(cxld)) {
+			struct cxl_endpoint_decoder *cxled;
+
+			cxled = cxl_endpoint_decoder_alloc(port);
+			if (IS_ERR(cxled)) {
 				dev_warn(&port->dev,
 					 "Failed to allocate the decoder\n");
-				return PTR_ERR(cxld);
+				return PTR_ERR(cxled);
 			}
+			cxld = &cxled->cxld;
 		} else {
 			struct cxl_switch_decoder *cxlsd;
 
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 4953a1c7b245..ca4f23204e5c 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -244,12 +244,12 @@ static void __cxl_decoder_release(struct cxl_decoder *cxld)
 	put_device(&port->dev);
 }
 
-static void cxl_decoder_release(struct device *dev)
+static void cxl_endpoint_decoder_release(struct device *dev)
 {
-	struct cxl_decoder *cxld = to_cxl_decoder(dev);
+	struct cxl_endpoint_decoder *cxled = to_cxl_endpoint_decoder(dev);
 
-	__cxl_decoder_release(cxld);
-	kfree(cxld);
+	__cxl_decoder_release(&cxled->cxld);
+	kfree(cxled);
 }
 
 static void cxl_switch_decoder_release(struct device *dev)
@@ -279,7 +279,7 @@ static void cxl_root_decoder_release(struct device *dev)
 
 static const struct device_type cxl_decoder_endpoint_type = {
 	.name = "cxl_decoder_endpoint",
-	.release = cxl_decoder_release,
+	.release = cxl_endpoint_decoder_release,
 	.groups = cxl_decoder_endpoint_attribute_groups,
 };
 
@@ -321,6 +321,15 @@ struct cxl_decoder *to_cxl_decoder(struct device *dev)
 }
 EXPORT_SYMBOL_NS_GPL(to_cxl_decoder, CXL);
 
+struct cxl_endpoint_decoder *to_cxl_endpoint_decoder(struct device *dev)
+{
+	if (dev_WARN_ONCE(dev, !is_endpoint_decoder(dev),
+			  "not a cxl_endpoint_decoder device\n"))
+		return NULL;
+	return container_of(dev, struct cxl_endpoint_decoder, cxld.dev);
+}
+EXPORT_SYMBOL_NS_GPL(to_cxl_endpoint_decoder, CXL);
+
 static struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev)
 {
 	if (dev_WARN_ONCE(dev, !is_switch_decoder(dev),
@@ -1360,26 +1369,28 @@ EXPORT_SYMBOL_NS_GPL(cxl_switch_decoder_alloc, CXL);
  *
  * Return: A new cxl decoder to be registered by cxl_decoder_add()
  */
-struct cxl_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port)
+struct cxl_endpoint_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port)
 {
+	struct cxl_endpoint_decoder *cxled;
 	struct cxl_decoder *cxld;
 	int rc;
 
 	if (!is_cxl_endpoint(port))
 		return ERR_PTR(-EINVAL);
 
-	cxld = kzalloc(sizeof(*cxld), GFP_KERNEL);
-	if (!cxld)
+	cxled = kzalloc(sizeof(*cxled), GFP_KERNEL);
+	if (!cxled)
 		return ERR_PTR(-ENOMEM);
 
+	cxld = &cxled->cxld;
 	rc = cxl_decoder_init(port, cxld);
 	if (rc)	 {
-		kfree(cxld);
+		kfree(cxled);
 		return ERR_PTR(rc);
 	}
 
 	cxld->dev.type = &cxl_decoder_endpoint_type;
-	return cxld;
+	return cxled;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_endpoint_decoder_alloc, CXL);
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index ebdac8e7d181..7e1460d89296 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -239,6 +239,18 @@ struct cxl_decoder {
 	unsigned long flags;
 };
 
+/**
+ * struct cxl_endpoint_decoder - Endpoint  / SPA to DPA decoder
+ * @cxld: base cxl_decoder_object
+ * @dpa_res: actively claimed DPA span of this decoder
+ * @skip: offset into @dpa_res where @cxld.hpa_range maps
+ */
+struct cxl_endpoint_decoder {
+	struct cxl_decoder cxld;
+	struct resource *dpa_res;
+	resource_size_t skip;
+};
+
 /**
  * struct cxl_switch_decoder - Switch specific CXL HDM Decoder
  * @cxld: base cxl_decoder object
@@ -387,6 +399,7 @@ struct cxl_dport *cxl_find_dport_by_dev(struct cxl_port *port,
 
 struct cxl_decoder *to_cxl_decoder(struct device *dev);
 struct cxl_root_decoder *to_cxl_root_decoder(struct device *dev);
+struct cxl_endpoint_decoder *to_cxl_endpoint_decoder(struct device *dev);
 bool is_root_decoder(struct device *dev);
 bool is_endpoint_decoder(struct device *dev);
 struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
@@ -394,7 +407,7 @@ struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
 struct cxl_switch_decoder *cxl_switch_decoder_alloc(struct cxl_port *port,
 						    unsigned int nr_targets);
 int cxl_decoder_add(struct cxl_decoder *cxld, int *target_map);
-struct cxl_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port);
+struct cxl_endpoint_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port);
 int cxl_decoder_add_locked(struct cxl_decoder *cxld, int *target_map);
 int cxl_decoder_autoremove(struct device *host, struct cxl_decoder *cxld);
 int cxl_endpoint_autoremove(struct cxl_memdev *cxlmd, struct cxl_port *endpoint);
diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
index 7991ddc6e562..4dad0fa7ac4c 100644
--- a/tools/testing/cxl/test/cxl.c
+++ b/tools/testing/cxl/test/cxl.c
@@ -462,12 +462,16 @@ static int mock_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 			}
 			cxld = &cxlsd->cxld;
 		} else {
-			cxld = cxl_endpoint_decoder_alloc(port);
-			if (IS_ERR(cxld)) {
+			struct cxl_endpoint_decoder *cxled;
+
+			cxled = cxl_endpoint_decoder_alloc(port);
+
+			if (IS_ERR(cxled)) {
 				dev_warn(&port->dev,
 					 "Failed to allocate the decoder\n");
-				return PTR_ERR(cxld);
+				return PTR_ERR(cxled);
 			}
+			cxld = &cxled->cxld;
 		}
 
 		cxld->hpa_range = (struct range) {

[PATCH v2 06/28] cxl/hdm: Enumerate allocated DPA

[Dan Williams]

In preparation for provisioning CXL regions, add accounting for the DPA
space consumed by existing regions / decoders. Recall, a CXL region is a
memory range comprised from one or more endpoint devices contributing a
mapping of their DPA into HPA space through a decoder.

Record the DPA ranges covered by committed decoders at initial probe of
endpoint ports relative to a per-device resource tree of the DPA type
(pmem or volatile-ram).

The cxl_dpa_rwsem semaphore is introduced to globally synchronize DPA
state across all endpoints and their decoders at once. The vast majority
of DPA operations are reads as region creation is expected to be as rare
as disk partitioning and volume creation. The device_lock() for this
synchronization is specifically avoided for concern of entangling with
sysfs attribute removal.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/hdm.c |  143 ++++++++++++++++++++++++++++++++++++++++++++----
 drivers/cxl/cxl.h      |    2 +
 drivers/cxl/cxlmem.h   |   13 ++++
 3 files changed, 147 insertions(+), 11 deletions(-)

diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 650363d5272f..d4c17325001b 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -153,10 +153,105 @@ void cxl_dpa_debug(struct seq_file *file, struct cxl_dev_state *cxlds)
 }
 EXPORT_SYMBOL_NS_GPL(cxl_dpa_debug, CXL);
 
+/*
+ * Must be called in a context that synchronizes against this decoder's
+ * port ->remove() callback (like an endpoint decoder sysfs attribute)
+ */
+static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_dev_state *cxlds = cxlmd->cxlds;
+	struct resource *res = cxled->dpa_res;
+
+	lockdep_assert_held_write(&cxl_dpa_rwsem);
+
+	if (cxled->skip)
+		__release_region(&cxlds->dpa_res, res->start - cxled->skip,
+				 cxled->skip);
+	cxled->skip = 0;
+	__release_region(&cxlds->dpa_res, res->start, resource_size(res));
+	cxled->dpa_res = NULL;
+}
+
+static void cxl_dpa_release(void *cxled)
+{
+	down_write(&cxl_dpa_rwsem);
+	__cxl_dpa_release(cxled);
+	up_write(&cxl_dpa_rwsem);
+}
+
+static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
+			     resource_size_t base, resource_size_t len,
+			     resource_size_t skipped)
+{
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_port *port = cxled_to_port(cxled);
+	struct cxl_dev_state *cxlds = cxlmd->cxlds;
+	struct device *dev = &port->dev;
+	struct resource *res;
+
+	lockdep_assert_held_write(&cxl_dpa_rwsem);
+
+	if (!len)
+		return 0;
+
+	if (cxled->dpa_res) {
+		dev_dbg(dev, "decoder%d.%d: existing allocation %pr assigned\n",
+			port->id, cxled->cxld.id, cxled->dpa_res);
+		return -EBUSY;
+	}
+
+	if (skipped) {
+		res = __request_region(&cxlds->dpa_res, base - skipped, skipped,
+				       dev_name(&cxled->cxld.dev), 0);
+		if (!res) {
+			dev_dbg(dev,
+				"decoder%d.%d: failed to reserve skipped space\n",
+				port->id, cxled->cxld.id);
+			return -EBUSY;
+		}
+	}
+	res = __request_region(&cxlds->dpa_res, base, len,
+			       dev_name(&cxled->cxld.dev), 0);
+	if (!res) {
+		dev_dbg(dev, "decoder%d.%d: failed to reserve allocation\n",
+			port->id, cxled->cxld.id);
+		if (skipped)
+			__release_region(&cxlds->dpa_res, base - skipped,
+					 skipped);
+		return -EBUSY;
+	}
+	cxled->dpa_res = res;
+	cxled->skip = skipped;
+
+	return 0;
+}
+
+static int devm_cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
+				resource_size_t base, resource_size_t len,
+				resource_size_t skipped)
+{
+	struct cxl_port *port = cxled_to_port(cxled);
+	int rc;
+
+	down_write(&cxl_dpa_rwsem);
+	rc = __cxl_dpa_reserve(cxled, base, len, skipped);
+	up_write(&cxl_dpa_rwsem);
+
+	if (rc)
+		return rc;
+
+	return devm_add_action_or_reset(&port->dev, cxl_dpa_release, cxled);
+}
+
 static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
-			    int *target_map, void __iomem *hdm, int which)
+			    int *target_map, void __iomem *hdm, int which,
+			    u64 *dpa_base)
 {
-	u64 size, base;
+	struct cxl_endpoint_decoder *cxled = NULL;
+	u64 size, base, skip, dpa_size;
+	bool committed;
+	u32 remainder;
 	int i, rc;
 	u32 ctrl;
 	union {
@@ -164,11 +259,15 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 		unsigned char target_id[8];
 	} target_list;
 
+	if (is_endpoint_decoder(&cxld->dev))
+		cxled = to_cxl_endpoint_decoder(&cxld->dev);
+
 	ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(which));
 	base = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_BASE_LOW_OFFSET(which));
 	size = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(which));
+	committed = !!(ctrl & CXL_HDM_DECODER0_CTRL_COMMITTED);
 
-	if (!(ctrl & CXL_HDM_DECODER0_CTRL_COMMITTED))
+	if (!committed)
 		size = 0;
 	if (base == U64_MAX || size == U64_MAX) {
 		dev_warn(&port->dev, "decoder%d.%d: Invalid resource range\n",
@@ -181,8 +280,8 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 		.end = base + size - 1,
 	};
 
-	/* switch decoders are always enabled if committed */
-	if (ctrl & CXL_HDM_DECODER0_CTRL_COMMITTED) {
+	/* decoders are enabled if committed */
+	if (committed) {
 		cxld->flags |= CXL_DECODER_F_ENABLE;
 		if (ctrl & CXL_HDM_DECODER0_CTRL_LOCK)
 			cxld->flags |= CXL_DECODER_F_LOCK;
@@ -211,14 +310,35 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 	if (rc)
 		return rc;
 
-	if (is_endpoint_decoder(&cxld->dev))
+	if (!cxled) {
+		target_list.value =
+			ioread64_hi_lo(hdm + CXL_HDM_DECODER0_TL_LOW(which));
+		for (i = 0; i < cxld->interleave_ways; i++)
+			target_map[i] = target_list.target_id[i];
+
 		return 0;
+	}
 
-	target_list.value =
-		ioread64_hi_lo(hdm + CXL_HDM_DECODER0_TL_LOW(which));
-	for (i = 0; i < cxld->interleave_ways; i++)
-		target_map[i] = target_list.target_id[i];
+	if (!committed)
+		return 0;
 
+	dpa_size = div_u64_rem(size, cxld->interleave_ways, &remainder);
+	if (remainder) {
+		dev_err(&port->dev,
+			"decoder%d.%d: invalid committed configuration size: %#llx ways: %d\n",
+			port->id, cxld->id, size, cxld->interleave_ways);
+		return -ENXIO;
+	}
+	skip = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_SKIP_LOW(which));
+	rc = devm_cxl_dpa_reserve(cxled, *dpa_base + skip, dpa_size, skip);
+	if (rc) {
+		dev_err(&port->dev,
+			"decoder%d.%d: Failed to reserve DPA range %#llx - %#llx\n (%d)",
+			port->id, cxld->id, *dpa_base,
+			*dpa_base + dpa_size + skip - 1, rc);
+		return rc;
+	}
+	*dpa_base += dpa_size + skip;
 	return 0;
 }
 
@@ -231,6 +351,7 @@ int devm_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 	void __iomem *hdm = cxlhdm->regs.hdm_decoder;
 	struct cxl_port *port = cxlhdm->port;
 	int i, committed;
+	u64 dpa_base = 0;
 	u32 ctrl;
 
 	/*
@@ -277,7 +398,7 @@ int devm_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 			cxld = &cxlsd->cxld;
 		}
 
-		rc = init_hdm_decoder(port, cxld, target_map, hdm, i);
+		rc = init_hdm_decoder(port, cxld, target_map, hdm, i, &dpa_base);
 		if (rc) {
 			put_device(&cxld->dev);
 			return rc;
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 7e1460d89296..d5e4cfac35ea 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -56,6 +56,8 @@
 #define   CXL_HDM_DECODER0_CTRL_TYPE BIT(12)
 #define CXL_HDM_DECODER0_TL_LOW(i) (0x20 * (i) + 0x24)
 #define CXL_HDM_DECODER0_TL_HIGH(i) (0x20 * (i) + 0x28)
+#define CXL_HDM_DECODER0_SKIP_LOW(i) CXL_HDM_DECODER0_TL_LOW(i)
+#define CXL_HDM_DECODER0_SKIP_HIGH(i) CXL_HDM_DECODER0_TL_HIGH(i)
 
 static inline int cxl_hdm_decoder_count(u32 cap_hdr)
 {
diff --git a/drivers/cxl/cxlmem.h b/drivers/cxl/cxlmem.h
index c6d6f57856cc..eee96016c3c7 100644
--- a/drivers/cxl/cxlmem.h
+++ b/drivers/cxl/cxlmem.h
@@ -50,6 +50,19 @@ static inline struct cxl_memdev *to_cxl_memdev(struct device *dev)
 	return container_of(dev, struct cxl_memdev, dev);
 }
 
+static inline struct cxl_port *cxled_to_port(struct cxl_endpoint_decoder *cxled)
+{
+	return to_cxl_port(cxled->cxld.dev.parent);
+}
+
+static inline struct cxl_memdev *
+cxled_to_memdev(struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_port *port = to_cxl_port(cxled->cxld.dev.parent);
+
+	return to_cxl_memdev(port->uport);
+}
+
 bool is_cxl_memdev(struct device *dev);
 static inline bool is_cxl_endpoint(struct cxl_port *port)
 {

[PATCH v2 07/28] cxl/hdm: Add 'mode' attribute to decoder objects

[Dan Williams]

Recall that the Device Physical Address (DPA) space of a CXL Memory
Expander is potentially partitioned into a volatile and persistent
portion. A decoder maps a Host Physical Address (HPA) range to a DPA
range and that translation depends on the value of all previous (lower
instance number) decoders before the current one.

In preparation for allowing dynamic provisioning of regions, decoders
need an ABI to indicate which DPA partition a decoder targets. This ABI
needs to be prepared for the possibility that some other agent committed
and locked a decoder that spans the partition boundary.

Add 'decoderX.Y/mode' to endpoint decoders that indicates which
partition 'ram' / 'pmem' the decoder targets, or 'mixed' if the decoder
currently spans the partition boundary.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603881967.551046.6007594190951596439.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   16 ++++++++++++++++
 drivers/cxl/core/hdm.c                  |   10 ++++++++++
 drivers/cxl/core/port.c                 |   20 ++++++++++++++++++++
 drivers/cxl/cxl.h                       |    9 +++++++++
 4 files changed, 55 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 16d9ffa94bbd..b8ef8aedaf39 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -179,3 +179,19 @@ Description:
 		expander memory (type-3). The 'target_type' attribute indicates
 		the current setting which may dynamically change based on what
 		memory regions are activated in this decode hierarchy.
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/mode
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RO) When a CXL decoder is of devtype "cxl_decoder_endpoint" it
+		translates from a host physical address range, to a device local
+		address range. Device-local address ranges are further split
+		into a 'ram' (volatile memory) range and 'pmem' (persistent
+		memory) range. The 'mode' attribute emits one of 'ram', 'pmem',
+		'mixed', or 'none'. The 'mixed' indication is for error cases
+		when a decoder straddles the volatile/persistent partition
+		boundary, and 'none' indicates the decoder is not actively
+		decoding, or no DPA allocation policy has been set.
diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index d4c17325001b..acd46b0d69c6 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -224,6 +224,16 @@ static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
 	cxled->dpa_res = res;
 	cxled->skip = skipped;
 
+	if (resource_contains(&cxlds->pmem_res, res))
+		cxled->mode = CXL_DECODER_PMEM;
+	else if (resource_contains(&cxlds->ram_res, res))
+		cxled->mode = CXL_DECODER_RAM;
+	else {
+		dev_dbg(dev, "decoder%d.%d: %pr mixed\n", port->id,
+			cxled->cxld.id, cxled->dpa_res);
+		cxled->mode = CXL_DECODER_MIXED;
+	}
+
 	return 0;
 }
 
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index ca4f23204e5c..0ac5dcd612e0 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -172,6 +172,25 @@ static ssize_t target_list_show(struct device *dev,
 }
 static DEVICE_ATTR_RO(target_list);
 
+static ssize_t mode_show(struct device *dev, struct device_attribute *attr,
+			 char *buf)
+{
+	struct cxl_endpoint_decoder *cxled = to_cxl_endpoint_decoder(dev);
+
+	switch (cxled->mode) {
+	case CXL_DECODER_RAM:
+		return sysfs_emit(buf, "ram\n");
+	case CXL_DECODER_PMEM:
+		return sysfs_emit(buf, "pmem\n");
+	case CXL_DECODER_NONE:
+		return sysfs_emit(buf, "none\n");
+	case CXL_DECODER_MIXED:
+	default:
+		return sysfs_emit(buf, "mixed\n");
+	}
+}
+static DEVICE_ATTR_RO(mode);
+
 static struct attribute *cxl_decoder_base_attrs[] = {
 	&dev_attr_start.attr,
 	&dev_attr_size.attr,
@@ -222,6 +241,7 @@ static const struct attribute_group *cxl_decoder_switch_attribute_groups[] = {
 
 static struct attribute *cxl_decoder_endpoint_attrs[] = {
 	&dev_attr_target_type.attr,
+	&dev_attr_mode.attr,
 	NULL,
 };
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index d5e4cfac35ea..3e7363dde80f 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -241,16 +241,25 @@ struct cxl_decoder {
 	unsigned long flags;
 };
 
+enum cxl_decoder_mode {
+	CXL_DECODER_NONE,
+	CXL_DECODER_RAM,
+	CXL_DECODER_PMEM,
+	CXL_DECODER_MIXED,
+};
+
 /**
  * struct cxl_endpoint_decoder - Endpoint  / SPA to DPA decoder
  * @cxld: base cxl_decoder_object
  * @dpa_res: actively claimed DPA span of this decoder
  * @skip: offset into @dpa_res where @cxld.hpa_range maps
+ * @mode: which memory type / access-mode-partition this decoder targets
  */
 struct cxl_endpoint_decoder {
 	struct cxl_decoder cxld;
 	struct resource *dpa_res;
 	resource_size_t skip;
+	enum cxl_decoder_mode mode;
 };
 
 /**

[PATCH v2 08/28] cxl/hdm: Track next decoder to allocate

[Dan Williams]

The CXL specification enforces that endpoint decoders are committed in
hw instance id order. In preparation for adding dynamic DPA allocation,
record the hw instance id in endpoint decoders, and enforce allocations
to occur in hw instance id order.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/hdm.c  |   15 +++++++++++++++
 drivers/cxl/core/port.c |    1 +
 drivers/cxl/cxl.h       |    2 ++
 3 files changed, 18 insertions(+)

diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index acd46b0d69c6..582f48141767 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -160,6 +160,7 @@ EXPORT_SYMBOL_NS_GPL(cxl_dpa_debug, CXL);
 static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
 {
 	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_port *port = cxled_to_port(cxled);
 	struct cxl_dev_state *cxlds = cxlmd->cxlds;
 	struct resource *res = cxled->dpa_res;
 
@@ -171,6 +172,7 @@ static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
 	cxled->skip = 0;
 	__release_region(&cxlds->dpa_res, res->start, resource_size(res));
 	cxled->dpa_res = NULL;
+	port->hdm_end--;
 }
 
 static void cxl_dpa_release(void *cxled)
@@ -201,6 +203,18 @@ static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
 		return -EBUSY;
 	}
 
+	if (port->hdm_end + 1 != cxled->cxld.id) {
+		/*
+		 * Assumes alloc and commit order is always in hardware instance
+		 * order per expectations from 8.2.5.12.20 Committing Decoder
+		 * Programming that enforce decoder[m] committed before
+		 * decoder[m+1] commit start.
+		 */
+		dev_dbg(dev, "decoder%d.%d: expected decoder%d.%d\n", port->id,
+			cxled->cxld.id, port->id, port->hdm_end + 1);
+		return -EBUSY;
+	}
+
 	if (skipped) {
 		res = __request_region(&cxlds->dpa_res, base - skipped, skipped,
 				       dev_name(&cxled->cxld.dev), 0);
@@ -233,6 +247,7 @@ static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
 			cxled->cxld.id, cxled->dpa_res);
 		cxled->mode = CXL_DECODER_MIXED;
 	}
+	port->hdm_end++;
 
 	return 0;
 }
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 0ac5dcd612e0..109611318760 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -502,6 +502,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
 
 	port->component_reg_phys = component_reg_phys;
 	ida_init(&port->decoder_ida);
+	port->hdm_end = -1;
 	INIT_LIST_HEAD(&port->dports);
 	INIT_LIST_HEAD(&port->endpoints);
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 3e7363dde80f..70cd24e4f3ce 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -333,6 +333,7 @@ struct cxl_nvdimm {
  * @dports: cxl_dport instances referenced by decoders
  * @endpoints: cxl_ep instances, endpoints that are a descendant of this port
  * @decoder_ida: allocator for decoder ids
+ * @hdm_end: track last allocated HDM decoder instance for allocation ordering
  * @component_reg_phys: component register capability base address (optional)
  * @dead: last ep has been removed, force port re-creation
  * @depth: How deep this port is relative to the root. depth 0 is the root.
@@ -345,6 +346,7 @@ struct cxl_port {
 	struct list_head dports;
 	struct list_head endpoints;
 	struct ida decoder_ida;
+	int hdm_end;
 	resource_size_t component_reg_phys;
 	bool dead;
 	unsigned int depth;

[PATCH v2 09/28] cxl/hdm: Add support for allocating DPA to an endpoint decoder

[Dan Williams]

The region provisioning flow will roughly follow a sequence of:

1/ Allocate DPA to a set of decoders

2/ Allocate HPA to a region

3/ Associate decoders with a region and validate that the DPA allocations
   and topologies match the parameters of the region.

For now, this change (step 1) arranges for DPA capacity to be allocated
and deleted from non-committed decoders based on the decoder's mode /
partition selection. Capacity is allocated from the lowest DPA in the
partition and any 'pmem' allocation blocks out all remaining ram
capacity in its 'skip' setting. DPA allocations are enforced in decoder
instance order. I.e. decoder N + 1 always starts at a higher DPA than
instance N, and deleting allocations must proceed from the
highest-instance allocated decoder to the lowest.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   37 ++++++
 drivers/cxl/core/core.h                 |    7 +
 drivers/cxl/core/hdm.c                  |  180 +++++++++++++++++++++++++++++++
 drivers/cxl/core/port.c                 |   73 ++++++++++++-
 4 files changed, 295 insertions(+), 2 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index b8ef8aedaf39..9b6cc7cdc73b 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -186,7 +186,7 @@ Date:		May, 2022
 KernelVersion:	v5.20
 Contact:	linux-cxl@vger.kernel.org
 Description:
-		(RO) When a CXL decoder is of devtype "cxl_decoder_endpoint" it
+		(RW) When a CXL decoder is of devtype "cxl_decoder_endpoint" it
 		translates from a host physical address range, to a device local
 		address range. Device-local address ranges are further split
 		into a 'ram' (volatile memory) range and 'pmem' (persistent
@@ -195,3 +195,38 @@ Description:
 		when a decoder straddles the volatile/persistent partition
 		boundary, and 'none' indicates the decoder is not actively
 		decoding, or no DPA allocation policy has been set.
+
+		'mode' can be written, when the decoder is in the 'disabled'
+		state, with either 'ram' or 'pmem' to set the boundaries for the
+		next allocation.
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/dpa_resource
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RO) When a CXL decoder is of devtype "cxl_decoder_endpoint",
+		and its 'dpa_size' attribute is non-zero, this attribute
+		indicates the device physical address (DPA) base address of the
+		allocation.
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/dpa_size
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) When a CXL decoder is of devtype "cxl_decoder_endpoint" it
+		translates from a host physical address range, to a device local
+		address range. The range, base address plus length in bytes, of
+		DPA allocated to this decoder is conveyed in these 2 attributes.
+		Allocations can be mutated as long as the decoder is in the
+		disabled state. A write to 'dpa_size' releases the previous DPA
+		allocation and then attempts to allocate from the free capacity
+		in the device partition referred to by 'decoderX.Y/mode'.
+		Allocate and free requests can only be performed on the highest
+		instance number disabled decoder with non-zero size. I.e.
+		allocations are enforced to occur in increasing 'decoderX.Y/id'
+		order and frees are enforced to occur in decreasing
+		'decoderX.Y/id' order.
diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index a0808cdaffba..5551b82b2da0 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -18,6 +18,13 @@ void __iomem *devm_cxl_iomap_block(struct device *dev, resource_size_t addr,
 				   resource_size_t length);
 
 struct dentry *cxl_debugfs_create_dir(const char *dir);
+int cxl_dpa_set_mode(struct cxl_endpoint_decoder *cxled,
+		     enum cxl_decoder_mode mode);
+int cxl_dpa_alloc(struct cxl_endpoint_decoder *cxled, unsigned long long size);
+int cxl_dpa_free(struct cxl_endpoint_decoder *cxled);
+resource_size_t cxl_dpa_size(struct cxl_endpoint_decoder *cxled);
+resource_size_t cxl_dpa_resource_start(struct cxl_endpoint_decoder *cxled);
+
 int cxl_memdev_init(void);
 void cxl_memdev_exit(void);
 void cxl_mbox_init(void);
diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 582f48141767..596b57fb60df 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -182,6 +182,19 @@ static void cxl_dpa_release(void *cxled)
 	up_write(&cxl_dpa_rwsem);
 }
 
+/*
+ * Must be called from context that will not race port device
+ * unregistration, like decoder sysfs attribute methods
+ */
+static void devm_cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_port *port = cxled_to_port(cxled);
+
+	lockdep_assert_held_write(&cxl_dpa_rwsem);
+	devm_remove_action(&port->dev, cxl_dpa_release, cxled);
+	__cxl_dpa_release(cxled);
+}
+
 static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
 			     resource_size_t base, resource_size_t len,
 			     resource_size_t skipped)
@@ -269,6 +282,173 @@ static int devm_cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
 	return devm_add_action_or_reset(&port->dev, cxl_dpa_release, cxled);
 }
 
+resource_size_t cxl_dpa_size(struct cxl_endpoint_decoder *cxled)
+{
+	resource_size_t size = 0;
+
+	down_read(&cxl_dpa_rwsem);
+	if (cxled->dpa_res)
+		size = resource_size(cxled->dpa_res);
+	up_read(&cxl_dpa_rwsem);
+
+	return size;
+}
+
+resource_size_t cxl_dpa_resource_start(struct cxl_endpoint_decoder *cxled)
+{
+	resource_size_t base = -1;
+
+	down_read(&cxl_dpa_rwsem);
+	if (cxled->dpa_res)
+		base = cxled->dpa_res->start;
+	up_read(&cxl_dpa_rwsem);
+
+	return base;
+}
+
+int cxl_dpa_free(struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_port *port = cxled_to_port(cxled);
+	struct device *dev = &cxled->cxld.dev;
+	int rc;
+
+	down_write(&cxl_dpa_rwsem);
+	if (!cxled->dpa_res) {
+		rc = 0;
+		goto out;
+	}
+	if (cxled->cxld.flags & CXL_DECODER_F_ENABLE) {
+		dev_dbg(dev, "decoder enabled\n");
+		rc = -EBUSY;
+		goto out;
+	}
+	if (cxled->cxld.id != port->hdm_end) {
+		dev_dbg(dev, "expected decoder%d.%d\n", port->id,
+			port->hdm_end);
+		rc = -EBUSY;
+		goto out;
+	}
+	devm_cxl_dpa_release(cxled);
+	rc = 0;
+out:
+	up_write(&cxl_dpa_rwsem);
+	return rc;
+}
+
+int cxl_dpa_set_mode(struct cxl_endpoint_decoder *cxled,
+		     enum cxl_decoder_mode mode)
+{
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_dev_state *cxlds = cxlmd->cxlds;
+	struct device *dev = &cxled->cxld.dev;
+	int rc;
+
+	switch (mode) {
+	case CXL_DECODER_RAM:
+	case CXL_DECODER_PMEM:
+		break;
+	default:
+		dev_dbg(dev, "unsupported mode: %d\n", mode);
+		return -EINVAL;
+	}
+
+	down_write(&cxl_dpa_rwsem);
+	if (cxled->cxld.flags & CXL_DECODER_F_ENABLE) {
+		rc = -EBUSY;
+		goto out;
+	}
+
+	/*
+	 * Only allow modes that are supported by the current partition
+	 * configuration
+	 */
+	if (mode == CXL_DECODER_PMEM && !resource_size(&cxlds->pmem_res)) {
+		dev_dbg(dev, "no available pmem capacity\n");
+		rc = -ENXIO;
+		goto out;
+	}
+	if (mode == CXL_DECODER_RAM && !resource_size(&cxlds->ram_res)) {
+		dev_dbg(dev, "no available ram capacity\n");
+		rc = -ENXIO;
+		goto out;
+	}
+
+	cxled->mode = mode;
+	rc = 0;
+out:
+	up_write(&cxl_dpa_rwsem);
+
+	return rc;
+}
+
+int cxl_dpa_alloc(struct cxl_endpoint_decoder *cxled, unsigned long long size)
+{
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	resource_size_t free_ram_start, free_pmem_start;
+	struct cxl_port *port = cxled_to_port(cxled);
+	struct cxl_dev_state *cxlds = cxlmd->cxlds;
+	struct device *dev = &cxled->cxld.dev;
+	resource_size_t start, avail, skip;
+	struct resource *p, *last;
+	int rc;
+
+	down_write(&cxl_dpa_rwsem);
+	if (cxled->cxld.flags & CXL_DECODER_F_ENABLE) {
+		dev_dbg(dev, "decoder enabled\n");
+		rc = -EBUSY;
+		goto out;
+	}
+
+	for (p = cxlds->ram_res.child, last = NULL; p; p = p->sibling)
+		last = p;
+	if (last)
+		free_ram_start = last->end + 1;
+	else
+		free_ram_start = cxlds->ram_res.start;
+
+	for (p = cxlds->pmem_res.child, last = NULL; p; p = p->sibling)
+		last = p;
+	if (last)
+		free_pmem_start = last->end + 1;
+	else
+		free_pmem_start = cxlds->pmem_res.start;
+
+	if (cxled->mode == CXL_DECODER_RAM) {
+		start = free_ram_start;
+		avail = cxlds->ram_res.end - start + 1;
+		skip = 0;
+	} else if (cxled->mode == CXL_DECODER_PMEM) {
+		resource_size_t skip_start, skip_end;
+
+		start = free_pmem_start;
+		avail = cxlds->pmem_res.end - start + 1;
+		skip_start = free_ram_start;
+		skip_end = start - 1;
+		skip = skip_end - skip_start + 1;
+	} else {
+		dev_dbg(dev, "mode not set\n");
+		rc = -EINVAL;
+		goto out;
+	}
+
+	if (size > avail) {
+		dev_dbg(dev, "%pa exceeds available %s capacity: %pa\n", &size,
+			cxled->mode == CXL_DECODER_RAM ? "ram" : "pmem",
+			&avail);
+		rc = -ENOSPC;
+		goto out;
+	}
+
+	rc = __cxl_dpa_reserve(cxled, start, size, skip);
+out:
+	up_write(&cxl_dpa_rwsem);
+
+	if (rc)
+		return rc;
+
+	return devm_add_action_or_reset(&port->dev, cxl_dpa_release, cxled);
+}
+
 static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 			    int *target_map, void __iomem *hdm, int which,
 			    u64 *dpa_base)
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 109611318760..fdc1be7db917 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -189,7 +189,76 @@ static ssize_t mode_show(struct device *dev, struct device_attribute *attr,
 		return sysfs_emit(buf, "mixed\n");
 	}
 }
-static DEVICE_ATTR_RO(mode);
+
+static ssize_t mode_store(struct device *dev, struct device_attribute *attr,
+			  const char *buf, size_t len)
+{
+	struct cxl_endpoint_decoder *cxled = to_cxl_endpoint_decoder(dev);
+	enum cxl_decoder_mode mode;
+	ssize_t rc;
+
+	if (sysfs_streq(buf, "pmem"))
+		mode = CXL_DECODER_PMEM;
+	else if (sysfs_streq(buf, "ram"))
+		mode = CXL_DECODER_RAM;
+	else
+		return -EINVAL;
+
+	rc = cxl_dpa_set_mode(cxled, mode);
+	if (rc)
+		return rc;
+
+	return len;
+}
+static DEVICE_ATTR_RW(mode);
+
+static ssize_t dpa_resource_show(struct device *dev, struct device_attribute *attr,
+			    char *buf)
+{
+	struct cxl_endpoint_decoder *cxled = to_cxl_endpoint_decoder(dev);
+	u64 base = cxl_dpa_resource_start(cxled);
+
+	return sysfs_emit(buf, "%#llx\n", base);
+}
+static DEVICE_ATTR_RO(dpa_resource);
+
+static ssize_t dpa_size_show(struct device *dev, struct device_attribute *attr,
+			     char *buf)
+{
+	struct cxl_endpoint_decoder *cxled = to_cxl_endpoint_decoder(dev);
+	resource_size_t size = cxl_dpa_size(cxled);
+
+	return sysfs_emit(buf, "%pa\n", &size);
+}
+
+static ssize_t dpa_size_store(struct device *dev, struct device_attribute *attr,
+			      const char *buf, size_t len)
+{
+	struct cxl_endpoint_decoder *cxled = to_cxl_endpoint_decoder(dev);
+	unsigned long long size;
+	ssize_t rc;
+
+	rc = kstrtoull(buf, 0, &size);
+	if (rc)
+		return rc;
+
+	if (!IS_ALIGNED(size, SZ_256M))
+		return -EINVAL;
+
+	rc = cxl_dpa_free(cxled);
+	if (rc)
+		return rc;
+
+	if (size == 0)
+		return len;
+
+	rc = cxl_dpa_alloc(cxled, size);
+	if (rc)
+		return rc;
+
+	return len;
+}
+static DEVICE_ATTR_RW(dpa_size);
 
 static struct attribute *cxl_decoder_base_attrs[] = {
 	&dev_attr_start.attr,
@@ -242,6 +311,8 @@ static const struct attribute_group *cxl_decoder_switch_attribute_groups[] = {
 static struct attribute *cxl_decoder_endpoint_attrs[] = {
 	&dev_attr_target_type.attr,
 	&dev_attr_mode.attr,
+	&dev_attr_dpa_size.attr,
+	&dev_attr_dpa_resource.attr,
 	NULL,
 };
 

[PATCH v2 10/28] cxl/port: Record dport in endpoint references

[Dan Williams]

Recall that the primary role of the cxl_mem driver is to probe if the
given endpoint is connected to a CXL port topology. In that process it
walks its device ancestry to its PCI root port. If that root port is
also a CXL root port then the probe process adds cxl_port object
instances at switch in the path between to the root and the endpoint. As
those cxl_port instances are added, or if a previous enumeration
attempt already created the port, a 'struct cxl_ep' instance is
registered with that port to track the endpoints interested in that
port.

At the time the cxl_ep is registered the downstream egress path from the
port to the endpoint is known. Take the opportunity to record that
information as it will be needed for dynamic programming of decoder
targets during region provisioning.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/port.c |   52 ++++++++++++++++++++++++++++++++---------------
 drivers/cxl/cxl.h       |    2 ++
 2 files changed, 37 insertions(+), 17 deletions(-)

diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index fdc1be7db917..a8d350361548 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -882,8 +882,9 @@ static struct cxl_ep *find_ep(struct cxl_port *port, struct device *ep_dev)
 	return NULL;
 }
 
-static int add_ep(struct cxl_port *port, struct cxl_ep *new)
+static int add_ep(struct cxl_ep *new)
 {
+	struct cxl_port *port = new->dport->port;
 	struct cxl_ep *dup;
 
 	device_lock(&port->dev);
@@ -901,14 +902,14 @@ static int add_ep(struct cxl_port *port, struct cxl_ep *new)
 
 /**
  * cxl_add_ep - register an endpoint's interest in a port
- * @port: a port in the endpoint's topology ancestry
+ * @dport: the dport that routes to @ep_dev
  * @ep_dev: device representing the endpoint
  *
  * Intermediate CXL ports are scanned based on the arrival of endpoints.
  * When those endpoints depart the port can be destroyed once all
  * endpoints that care about that port have been removed.
  */
-static int cxl_add_ep(struct cxl_port *port, struct device *ep_dev)
+static int cxl_add_ep(struct cxl_dport *dport, struct device *ep_dev)
 {
 	struct cxl_ep *ep;
 	int rc;
@@ -919,8 +920,9 @@ static int cxl_add_ep(struct cxl_port *port, struct device *ep_dev)
 
 	INIT_LIST_HEAD(&ep->list);
 	ep->ep = get_device(ep_dev);
+	ep->dport = dport;
 
-	rc = add_ep(port, ep);
+	rc = add_ep(ep);
 	if (rc)
 		cxl_ep_release(ep);
 	return rc;
@@ -929,11 +931,13 @@ static int cxl_add_ep(struct cxl_port *port, struct device *ep_dev)
 struct cxl_find_port_ctx {
 	const struct device *dport_dev;
 	const struct cxl_port *parent_port;
+	struct cxl_dport **dport;
 };
 
 static int match_port_by_dport(struct device *dev, const void *data)
 {
 	const struct cxl_find_port_ctx *ctx = data;
+	struct cxl_dport *dport;
 	struct cxl_port *port;
 
 	if (!is_cxl_port(dev))
@@ -942,7 +946,10 @@ static int match_port_by_dport(struct device *dev, const void *data)
 		return 0;
 
 	port = to_cxl_port(dev);
-	return cxl_find_dport_by_dev(port, ctx->dport_dev) != NULL;
+	dport = cxl_find_dport_by_dev(port, ctx->dport_dev);
+	if (ctx->dport)
+		*ctx->dport = dport;
+	return dport != NULL;
 }
 
 static struct cxl_port *__find_cxl_port(struct cxl_find_port_ctx *ctx)
@@ -958,24 +965,32 @@ static struct cxl_port *__find_cxl_port(struct cxl_find_port_ctx *ctx)
 	return NULL;
 }
 
-static struct cxl_port *find_cxl_port(struct device *dport_dev)
+static struct cxl_port *find_cxl_port(struct device *dport_dev,
+				      struct cxl_dport **dport)
 {
 	struct cxl_find_port_ctx ctx = {
 		.dport_dev = dport_dev,
+		.dport = dport,
 	};
+	struct cxl_port *port;
 
-	return __find_cxl_port(&ctx);
+	port = __find_cxl_port(&ctx);
+	return port;
 }
 
 static struct cxl_port *find_cxl_port_at(struct cxl_port *parent_port,
-					 struct device *dport_dev)
+					 struct device *dport_dev,
+					 struct cxl_dport **dport)
 {
 	struct cxl_find_port_ctx ctx = {
 		.dport_dev = dport_dev,
 		.parent_port = parent_port,
+		.dport = dport,
 	};
+	struct cxl_port *port;
 
-	return __find_cxl_port(&ctx);
+	port = __find_cxl_port(&ctx);
+	return port;
 }
 
 /*
@@ -1060,7 +1075,7 @@ static void cxl_detach_ep(void *data)
 		if (!dport_dev)
 			break;
 
-		port = find_cxl_port(dport_dev);
+		port = find_cxl_port(dport_dev, NULL);
 		if (!port)
 			continue;
 
@@ -1135,6 +1150,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 	struct device *dparent = grandparent(dport_dev);
 	struct cxl_port *port, *parent_port = NULL;
 	resource_size_t component_reg_phys;
+	struct cxl_dport *dport;
 	int rc;
 
 	if (!dparent) {
@@ -1148,7 +1164,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 		return -ENXIO;
 	}
 
-	parent_port = find_cxl_port(dparent);
+	parent_port = find_cxl_port(dparent, NULL);
 	if (!parent_port) {
 		/* iterate to create this parent_port */
 		return -EAGAIN;
@@ -1163,13 +1179,14 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 		goto out;
 	}
 
-	port = find_cxl_port_at(parent_port, dport_dev);
+	port = find_cxl_port_at(parent_port, dport_dev, &dport);
 	if (!port) {
 		component_reg_phys = find_component_registers(uport_dev);
 		port = devm_cxl_add_port(&parent_port->dev, uport_dev,
 					 component_reg_phys, parent_port);
+		/* retry find to pick up the new dport information */
 		if (!IS_ERR(port))
-			get_device(&port->dev);
+			port = find_cxl_port_at(parent_port, dport_dev, &dport);
 	}
 out:
 	device_unlock(&parent_port->dev);
@@ -1179,7 +1196,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 	else {
 		dev_dbg(&cxlmd->dev, "add to new port %s:%s\n",
 			dev_name(&port->dev), dev_name(port->uport));
-		rc = cxl_add_ep(port, &cxlmd->dev);
+		rc = cxl_add_ep(dport, &cxlmd->dev);
 		if (rc == -EEXIST) {
 			/*
 			 * "can't" happen, but this error code means
@@ -1213,6 +1230,7 @@ int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd)
 	for (iter = dev; iter; iter = grandparent(iter)) {
 		struct device *dport_dev = grandparent(iter);
 		struct device *uport_dev;
+		struct cxl_dport *dport;
 		struct cxl_port *port;
 
 		if (!dport_dev)
@@ -1228,12 +1246,12 @@ int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd)
 		dev_dbg(dev, "scan: iter: %s dport_dev: %s parent: %s\n",
 			dev_name(iter), dev_name(dport_dev),
 			dev_name(uport_dev));
-		port = find_cxl_port(dport_dev);
+		port = find_cxl_port(dport_dev, &dport);
 		if (port) {
 			dev_dbg(&cxlmd->dev,
 				"found already registered port %s:%s\n",
 				dev_name(&port->dev), dev_name(port->uport));
-			rc = cxl_add_ep(port, &cxlmd->dev);
+			rc = cxl_add_ep(dport, &cxlmd->dev);
 
 			/*
 			 * If the endpoint already exists in the port's list,
@@ -1274,7 +1292,7 @@ EXPORT_SYMBOL_NS_GPL(devm_cxl_enumerate_ports, CXL);
 
 struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd)
 {
-	return find_cxl_port(grandparent(&cxlmd->dev));
+	return find_cxl_port(grandparent(&cxlmd->dev), NULL);
 }
 EXPORT_SYMBOL_NS_GPL(cxl_mem_find_port, CXL);
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 70cd24e4f3ce..31f33844279a 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -371,10 +371,12 @@ struct cxl_dport {
 /**
  * struct cxl_ep - track an endpoint's interest in a port
  * @ep: device that hosts a generic CXL endpoint (expander or accelerator)
+ * @dport: which dport routes to this endpoint on @port
  * @list: node on port->endpoints list
  */
 struct cxl_ep {
 	struct device *ep;
+	struct cxl_dport *dport;
 	struct list_head list;
 };
 

[PATCH v2 11/28] cxl/port: Record parent dport when adding ports

[Dan Williams]

At the time that cxl_port instances are being created, cache the dport
from the parent port that points to this new child port. This will be
useful for region provisioning when walking the tree to calculate
decoder targets, and saves rewalking the dport list after the fact to
build this information.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-1-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/acpi.c      |    3 +--
 drivers/cxl/core/port.c |   27 +++++++++++++++------------
 drivers/cxl/cxl.h       |    7 +++++--
 drivers/cxl/mem.c       |   10 ++++++----
 4 files changed, 27 insertions(+), 20 deletions(-)

diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
index 8f021241699f..64004eb672d0 100644
--- a/drivers/cxl/acpi.c
+++ b/drivers/cxl/acpi.c
@@ -211,8 +211,7 @@ static int add_host_bridge_uport(struct device *match, void *arg)
 	if (rc)
 		return rc;
 
-	port = devm_cxl_add_port(host, match, dport->component_reg_phys,
-				 root_port);
+	port = devm_cxl_add_port(host, match, dport->component_reg_phys, dport);
 	if (IS_ERR(port))
 		return PTR_ERR(port);
 	dev_dbg(host, "%s: add: %s\n", dev_name(match), dev_name(&port->dev));
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index a8d350361548..6d2846404ab8 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -526,7 +526,7 @@ static struct lock_class_key cxl_port_key;
 
 static struct cxl_port *cxl_port_alloc(struct device *uport,
 				       resource_size_t component_reg_phys,
-				       struct cxl_port *parent_port)
+				       struct cxl_dport *parent_dport)
 {
 	struct cxl_port *port;
 	struct device *dev;
@@ -549,11 +549,13 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
 	 * description.
 	 */
 	dev = &port->dev;
-	if (parent_port) {
+	if (parent_dport) {
+		struct cxl_port *parent_port = parent_dport->port;
 		struct cxl_port *iter;
 
 		dev->parent = &parent_port->dev;
 		port->depth = parent_port->depth + 1;
+		port->parent_dport = parent_dport;
 
 		/*
 		 * walk to the host bridge, or the first ancestor that knows
@@ -595,24 +597,24 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
  * @host: host device for devm operations
  * @uport: "physical" device implementing this upstream port
  * @component_reg_phys: (optional) for configurable cxl_port instances
- * @parent_port: next hop up in the CXL memory decode hierarchy
+ * @parent_dport: next hop up in the CXL memory decode hierarchy
  */
 struct cxl_port *devm_cxl_add_port(struct device *host, struct device *uport,
 				   resource_size_t component_reg_phys,
-				   struct cxl_port *parent_port)
+				   struct cxl_dport *parent_dport)
 {
 	struct cxl_port *port;
 	struct device *dev;
 	int rc;
 
-	port = cxl_port_alloc(uport, component_reg_phys, parent_port);
+	port = cxl_port_alloc(uport, component_reg_phys, parent_dport);
 	if (IS_ERR(port))
 		return port;
 
 	dev = &port->dev;
 	if (is_cxl_memdev(uport))
 		rc = dev_set_name(dev, "endpoint%d", port->id);
-	else if (parent_port)
+	else if (parent_dport)
 		rc = dev_set_name(dev, "port%d", port->id);
 	else
 		rc = dev_set_name(dev, "root%d", port->id);
@@ -1014,7 +1016,7 @@ static void delete_endpoint(void *data)
 	struct cxl_port *parent_port;
 	struct device *parent;
 
-	parent_port = cxl_mem_find_port(cxlmd);
+	parent_port = cxl_mem_find_port(cxlmd, NULL);
 	if (!parent_port)
 		goto out;
 	parent = &parent_port->dev;
@@ -1149,8 +1151,8 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 {
 	struct device *dparent = grandparent(dport_dev);
 	struct cxl_port *port, *parent_port = NULL;
+	struct cxl_dport *dport, *parent_dport;
 	resource_size_t component_reg_phys;
-	struct cxl_dport *dport;
 	int rc;
 
 	if (!dparent) {
@@ -1164,7 +1166,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 		return -ENXIO;
 	}
 
-	parent_port = find_cxl_port(dparent, NULL);
+	parent_port = find_cxl_port(dparent, &parent_dport);
 	if (!parent_port) {
 		/* iterate to create this parent_port */
 		return -EAGAIN;
@@ -1183,7 +1185,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 	if (!port) {
 		component_reg_phys = find_component_registers(uport_dev);
 		port = devm_cxl_add_port(&parent_port->dev, uport_dev,
-					 component_reg_phys, parent_port);
+					 component_reg_phys, parent_dport);
 		/* retry find to pick up the new dport information */
 		if (!IS_ERR(port))
 			port = find_cxl_port_at(parent_port, dport_dev, &dport);
@@ -1290,9 +1292,10 @@ int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd)
 }
 EXPORT_SYMBOL_NS_GPL(devm_cxl_enumerate_ports, CXL);
 
-struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd)
+struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd,
+				   struct cxl_dport **dport)
 {
-	return find_cxl_port(grandparent(&cxlmd->dev), NULL);
+	return find_cxl_port(grandparent(&cxlmd->dev), dport);
 }
 EXPORT_SYMBOL_NS_GPL(cxl_mem_find_port, CXL);
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 31f33844279a..973e0efe4bd4 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -332,6 +332,7 @@ struct cxl_nvdimm {
  * @id: id for port device-name
  * @dports: cxl_dport instances referenced by decoders
  * @endpoints: cxl_ep instances, endpoints that are a descendant of this port
+ * @parent_dport: dport that points to this port in the parent
  * @decoder_ida: allocator for decoder ids
  * @hdm_end: track last allocated HDM decoder instance for allocation ordering
  * @component_reg_phys: component register capability base address (optional)
@@ -345,6 +346,7 @@ struct cxl_port {
 	int id;
 	struct list_head dports;
 	struct list_head endpoints;
+	struct cxl_dport *parent_dport;
 	struct ida decoder_ida;
 	int hdm_end;
 	resource_size_t component_reg_phys;
@@ -399,11 +401,12 @@ int devm_cxl_register_pci_bus(struct device *host, struct device *uport,
 struct pci_bus *cxl_port_to_pci_bus(struct cxl_port *port);
 struct cxl_port *devm_cxl_add_port(struct device *host, struct device *uport,
 				   resource_size_t component_reg_phys,
-				   struct cxl_port *parent_port);
+				   struct cxl_dport *parent_dport);
 struct cxl_port *find_cxl_root(struct device *dev);
 int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd);
 int cxl_bus_rescan(void);
-struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd);
+struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd,
+				   struct cxl_dport **dport);
 bool schedule_cxl_memdev_detach(struct cxl_memdev *cxlmd);
 
 struct cxl_dport *devm_cxl_add_dport(struct cxl_port *port,
diff --git a/drivers/cxl/mem.c b/drivers/cxl/mem.c
index 7513bea55145..2786d3402c9e 100644
--- a/drivers/cxl/mem.c
+++ b/drivers/cxl/mem.c
@@ -26,14 +26,15 @@
  */
 
 static int create_endpoint(struct cxl_memdev *cxlmd,
-			   struct cxl_port *parent_port)
+			   struct cxl_dport *parent_dport)
 {
+	struct cxl_port *parent_port = parent_dport->port;
 	struct cxl_dev_state *cxlds = cxlmd->cxlds;
 	struct cxl_port *endpoint;
 	int rc;
 
 	endpoint = devm_cxl_add_port(&parent_port->dev, &cxlmd->dev,
-				     cxlds->component_reg_phys, parent_port);
+				     cxlds->component_reg_phys, parent_dport);
 	if (IS_ERR(endpoint))
 		return PTR_ERR(endpoint);
 
@@ -76,6 +77,7 @@ static int cxl_mem_probe(struct device *dev)
 {
 	struct cxl_memdev *cxlmd = to_cxl_memdev(dev);
 	struct cxl_port *parent_port;
+	struct cxl_dport *dport;
 	struct dentry *dentry;
 	int rc;
 
@@ -100,7 +102,7 @@ static int cxl_mem_probe(struct device *dev)
 	if (rc)
 		return rc;
 
-	parent_port = cxl_mem_find_port(cxlmd);
+	parent_port = cxl_mem_find_port(cxlmd, &dport);
 	if (!parent_port) {
 		dev_err(dev, "CXL port topology not found\n");
 		return -ENXIO;
@@ -114,7 +116,7 @@ static int cxl_mem_probe(struct device *dev)
 		goto unlock;
 	}
 
-	rc = create_endpoint(cxlmd, parent_port);
+	rc = create_endpoint(cxlmd, dport);
 unlock:
 	device_unlock(&parent_port->dev);
 	put_device(&parent_port->dev);

[PATCH v2 12/28] cxl/port: Move 'cxl_ep' references to an xarray per port

[Dan Williams]

In preparation for region provisioning that needs to walk the topology
by endpoints, use an xarray to record endpoint interest in a given port.
In addition to being more space and time efficient it also reduces the
complexity of the implementation by moving locking internal to the
xarray implementation. It also allows for a single cxl_ep reference to
be recorded in multiple xarrays.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-2-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/port.c |   60 +++++++++++++++++++++++------------------------
 drivers/cxl/cxl.h       |    4 +--
 2 files changed, 30 insertions(+), 34 deletions(-)

diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 6d2846404ab8..727d861e21db 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -431,22 +431,27 @@ static struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev)
 
 static void cxl_ep_release(struct cxl_ep *ep)
 {
-	if (!ep)
-		return;
-	list_del(&ep->list);
 	put_device(ep->ep);
 	kfree(ep);
 }
 
+static void cxl_ep_remove(struct cxl_port *port, struct cxl_ep *ep)
+{
+	if (!ep)
+		return;
+	xa_erase(&port->endpoints, (unsigned long) ep->ep);
+	cxl_ep_release(ep);
+}
+
 static void cxl_port_release(struct device *dev)
 {
 	struct cxl_port *port = to_cxl_port(dev);
-	struct cxl_ep *ep, *_e;
+	unsigned long index;
+	struct cxl_ep *ep;
 
-	device_lock(dev);
-	list_for_each_entry_safe(ep, _e, &port->endpoints, list)
-		cxl_ep_release(ep);
-	device_unlock(dev);
+	xa_for_each(&port->endpoints, index, ep)
+		cxl_ep_remove(port, ep);
+	xa_destroy(&port->endpoints);
 	ida_free(&cxl_port_ida, port->id);
 	kfree(port);
 }
@@ -577,7 +582,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
 	ida_init(&port->decoder_ida);
 	port->hdm_end = -1;
 	INIT_LIST_HEAD(&port->dports);
-	INIT_LIST_HEAD(&port->endpoints);
+	xa_init(&port->endpoints);
 
 	device_initialize(dev);
 	lockdep_set_class_and_subclass(&dev->mutex, &cxl_port_key, port->depth);
@@ -873,33 +878,21 @@ struct cxl_dport *devm_cxl_add_dport(struct cxl_port *port,
 }
 EXPORT_SYMBOL_NS_GPL(devm_cxl_add_dport, CXL);
 
-static struct cxl_ep *find_ep(struct cxl_port *port, struct device *ep_dev)
-{
-	struct cxl_ep *ep;
-
-	device_lock_assert(&port->dev);
-	list_for_each_entry(ep, &port->endpoints, list)
-		if (ep->ep == ep_dev)
-			return ep;
-	return NULL;
-}
-
 static int add_ep(struct cxl_ep *new)
 {
 	struct cxl_port *port = new->dport->port;
-	struct cxl_ep *dup;
+	int rc;
 
 	device_lock(&port->dev);
 	if (port->dead) {
 		device_unlock(&port->dev);
 		return -ENXIO;
 	}
-	dup = find_ep(port, new->ep);
-	if (!dup)
-		list_add_tail(&new->list, &port->endpoints);
+	rc = xa_insert(&port->endpoints, (unsigned long)new->ep, new,
+		       GFP_KERNEL);
 	device_unlock(&port->dev);
 
-	return dup ? -EEXIST : 0;
+	return rc;
 }
 
 /**
@@ -920,7 +913,6 @@ static int cxl_add_ep(struct cxl_dport *dport, struct device *ep_dev)
 	if (!ep)
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&ep->list);
 	ep->ep = get_device(ep_dev);
 	ep->dport = dport;
 
@@ -1063,6 +1055,12 @@ static void delete_switch_port(struct cxl_port *port, struct list_head *dports)
 	devm_release_action(port->dev.parent, unregister_port, port);
 }
 
+static struct cxl_ep *cxl_ep_load(struct cxl_port *port,
+				  struct cxl_memdev *cxlmd)
+{
+	return xa_load(&port->endpoints, (unsigned long)&cxlmd->dev);
+}
+
 static void cxl_detach_ep(void *data)
 {
 	struct cxl_memdev *cxlmd = data;
@@ -1101,11 +1099,11 @@ static void cxl_detach_ep(void *data)
 		}
 
 		device_lock(&port->dev);
-		ep = find_ep(port, &cxlmd->dev);
+		ep = cxl_ep_load(port, cxlmd);
 		dev_dbg(&cxlmd->dev, "disconnect %s from %s\n",
 			ep ? dev_name(ep->ep) : "", dev_name(&port->dev));
-		cxl_ep_release(ep);
-		if (ep && !port->dead && list_empty(&port->endpoints) &&
+		cxl_ep_remove(port, ep);
+		if (ep && !port->dead && xa_empty(&port->endpoints) &&
 		    !is_cxl_root(parent_port)) {
 			/*
 			 * This was the last ep attached to a dynamically
@@ -1199,7 +1197,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
 		dev_dbg(&cxlmd->dev, "add to new port %s:%s\n",
 			dev_name(&port->dev), dev_name(port->uport));
 		rc = cxl_add_ep(dport, &cxlmd->dev);
-		if (rc == -EEXIST) {
+		if (rc == -EBUSY) {
 			/*
 			 * "can't" happen, but this error code means
 			 * something to the caller, so translate it.
@@ -1262,7 +1260,7 @@ int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd)
 			 * the parent_port lock as the current port may be being
 			 * reaped.
 			 */
-			if (rc && rc != -EEXIST) {
+			if (rc && rc != -EBUSY) {
 				put_device(&port->dev);
 				return rc;
 			}
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 973e0efe4bd4..de5cb8288cd4 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -345,7 +345,7 @@ struct cxl_port {
 	struct device *host_bridge;
 	int id;
 	struct list_head dports;
-	struct list_head endpoints;
+	struct xarray endpoints;
 	struct cxl_dport *parent_dport;
 	struct ida decoder_ida;
 	int hdm_end;
@@ -374,12 +374,10 @@ struct cxl_dport {
  * struct cxl_ep - track an endpoint's interest in a port
  * @ep: device that hosts a generic CXL endpoint (expander or accelerator)
  * @dport: which dport routes to this endpoint on @port
- * @list: node on port->endpoints list
  */
 struct cxl_ep {
 	struct device *ep;
 	struct cxl_dport *dport;
-	struct list_head list;
 };
 
 /*

[PATCH v2 13/28] cxl/port: Move dport tracking to an xarray

[Dan Williams]

Reduce the complexity and the overhead of walking the topology to
determine endpoint connectivity to root decoder interleave
configurations.

Note that cxl_detach_ep(), after it determines that the last @ep has
departed and decides to delete the port, now needs to walk the dport
array with the device_lock() held to remove entries. Previously
list_splice_init() could be used atomically delete all dport entries at
once and then perform entry tear down outside the lock. There is no
list_splice_init() equivalent for the xarray.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/hdm.c  |    6 ++-
 drivers/cxl/core/port.c |   85 ++++++++++++++++++++---------------------------
 drivers/cxl/cxl.h       |   12 ++++---
 3 files changed, 47 insertions(+), 56 deletions(-)

diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 596b57fb60df..4a0325b02ca4 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -50,8 +50,9 @@ static int add_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 int devm_cxl_add_passthrough_decoder(struct cxl_port *port)
 {
 	struct cxl_switch_decoder *cxlsd;
-	struct cxl_dport *dport;
+	struct cxl_dport *dport = NULL;
 	int single_port_map[1];
+	unsigned long index;
 
 	cxlsd = cxl_switch_decoder_alloc(port, 1);
 	if (IS_ERR(cxlsd))
@@ -59,7 +60,8 @@ int devm_cxl_add_passthrough_decoder(struct cxl_port *port)
 
 	device_lock_assert(&port->dev);
 
-	dport = list_first_entry(&port->dports, typeof(*dport), list);
+	xa_for_each(&port->dports, index, dport)
+		break;
 	single_port_map[0] = dport->port_id;
 
 	return add_hdm_decoder(port, &cxlsd->cxld, single_port_map);
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 727d861e21db..b2c44e7ef6a8 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -452,6 +452,7 @@ static void cxl_port_release(struct device *dev)
 	xa_for_each(&port->endpoints, index, ep)
 		cxl_ep_remove(port, ep);
 	xa_destroy(&port->endpoints);
+	xa_destroy(&port->dports);
 	ida_free(&cxl_port_ida, port->id);
 	kfree(port);
 }
@@ -581,7 +582,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
 	port->component_reg_phys = component_reg_phys;
 	ida_init(&port->decoder_ida);
 	port->hdm_end = -1;
-	INIT_LIST_HEAD(&port->dports);
+	xa_init(&port->dports);
 	xa_init(&port->endpoints);
 
 	device_initialize(dev);
@@ -711,17 +712,13 @@ static int match_root_child(struct device *dev, const void *match)
 		return 0;
 
 	port = to_cxl_port(dev);
-	device_lock(dev);
-	list_for_each_entry(dport, &port->dports, list) {
-		iter = match;
-		while (iter) {
-			if (iter == dport->dport)
-				goto out;
-			iter = iter->parent;
-		}
+	iter = match;
+	while (iter) {
+		dport = cxl_find_dport_by_dev(port, iter);
+		if (dport)
+			break;
+		iter = iter->parent;
 	}
-out:
-	device_unlock(dev);
 
 	return !!iter;
 }
@@ -745,9 +742,10 @@ EXPORT_SYMBOL_NS_GPL(find_cxl_root, CXL);
 static struct cxl_dport *find_dport(struct cxl_port *port, int id)
 {
 	struct cxl_dport *dport;
+	unsigned long index;
 
 	device_lock_assert(&port->dev);
-	list_for_each_entry (dport, &port->dports, list)
+	xa_for_each(&port->dports, index, dport)
 		if (dport->port_id == id)
 			return dport;
 	return NULL;
@@ -759,15 +757,15 @@ static int add_dport(struct cxl_port *port, struct cxl_dport *new)
 
 	device_lock_assert(&port->dev);
 	dup = find_dport(port, new->port_id);
-	if (dup)
+	if (dup) {
 		dev_err(&port->dev,
 			"unable to add dport%d-%s non-unique port id (%s)\n",
 			new->port_id, dev_name(new->dport),
 			dev_name(dup->dport));
-	else
-		list_add_tail(&new->list, &port->dports);
-
-	return dup ? -EEXIST : 0;
+		return -EBUSY;
+	}
+	return xa_insert(&port->dports, (unsigned long)new->dport, new,
+			 GFP_KERNEL);
 }
 
 /*
@@ -794,10 +792,8 @@ static void cxl_dport_remove(void *data)
 	struct cxl_dport *dport = data;
 	struct cxl_port *port = dport->port;
 
+	xa_erase(&port->dports, (unsigned long) dport->dport);
 	put_device(dport->dport);
-	cond_cxl_root_lock(port);
-	list_del(&dport->list);
-	cond_cxl_root_unlock(port);
 }
 
 static void cxl_dport_unlink(void *data)
@@ -849,7 +845,6 @@ struct cxl_dport *devm_cxl_add_dport(struct cxl_port *port,
 	if (!dport)
 		return ERR_PTR(-ENOMEM);
 
-	INIT_LIST_HEAD(&dport->list);
 	dport->dport = dport_dev;
 	dport->port_id = port_id;
 	dport->component_reg_phys = component_reg_phys;
@@ -1040,19 +1035,27 @@ EXPORT_SYMBOL_NS_GPL(cxl_endpoint_autoremove, CXL);
  * for a port to be unregistered is when all memdevs beneath that port have gone
  * through ->remove(). This "bottom-up" removal selectively removes individual
  * child ports manually. This depends on devm_cxl_add_port() to not change is
- * devm action registration order.
+ * devm action registration order, and for dports to have already been
+ * destroyed by reap_dports().
  */
-static void delete_switch_port(struct cxl_port *port, struct list_head *dports)
+static void delete_switch_port(struct cxl_port *port)
+{
+	devm_release_action(port->dev.parent, cxl_unlink_uport, port);
+	devm_release_action(port->dev.parent, unregister_port, port);
+}
+
+static void reap_dports(struct cxl_port *port)
 {
-	struct cxl_dport *dport, *_d;
+	struct cxl_dport *dport;
+	unsigned long index;
 
-	list_for_each_entry_safe(dport, _d, dports, list) {
+	device_lock_assert(&port->dev);
+
+	xa_for_each(&port->dports, index, dport) {
 		devm_release_action(&port->dev, cxl_dport_unlink, dport);
 		devm_release_action(&port->dev, cxl_dport_remove, dport);
 		devm_kfree(&port->dev, dport);
 	}
-	devm_release_action(port->dev.parent, cxl_unlink_uport, port);
-	devm_release_action(port->dev.parent, unregister_port, port);
 }
 
 static struct cxl_ep *cxl_ep_load(struct cxl_port *port,
@@ -1069,8 +1072,8 @@ static void cxl_detach_ep(void *data)
 	for (iter = &cxlmd->dev; iter; iter = grandparent(iter)) {
 		struct device *dport_dev = grandparent(iter);
 		struct cxl_port *port, *parent_port;
-		LIST_HEAD(reap_dports);
 		struct cxl_ep *ep;
+		bool died = false;
 
 		if (!dport_dev)
 			break;
@@ -1110,15 +1113,16 @@ static void cxl_detach_ep(void *data)
 			 * enumerated port. Block new cxl_add_ep() and garbage
 			 * collect the port.
 			 */
+			died = true;
 			port->dead = true;
-			list_splice_init(&port->dports, &reap_dports);
+			reap_dports(port);
 		}
 		device_unlock(&port->dev);
 
-		if (!list_empty(&reap_dports)) {
+		if (died) {
 			dev_dbg(&cxlmd->dev, "delete %s\n",
 				dev_name(&port->dev));
-			delete_switch_port(port, &reap_dports);
+			delete_switch_port(port);
 		}
 		put_device(&port->dev);
 		device_unlock(&parent_port->dev);
@@ -1297,23 +1301,6 @@ struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd,
 }
 EXPORT_SYMBOL_NS_GPL(cxl_mem_find_port, CXL);
 
-struct cxl_dport *cxl_find_dport_by_dev(struct cxl_port *port,
-					const struct device *dev)
-{
-	struct cxl_dport *dport;
-
-	device_lock(&port->dev);
-	list_for_each_entry(dport, &port->dports, list)
-		if (dport->dport == dev) {
-			device_unlock(&port->dev);
-			return dport;
-		}
-
-	device_unlock(&port->dev);
-	return NULL;
-}
-EXPORT_SYMBOL_NS_GPL(cxl_find_dport_by_dev, CXL);
-
 static int decoder_populate_targets(struct cxl_switch_decoder *cxlsd,
 				    struct cxl_port *port, int *target_map)
 {
@@ -1324,7 +1311,7 @@ static int decoder_populate_targets(struct cxl_switch_decoder *cxlsd,
 
 	device_lock_assert(&port->dev);
 
-	if (list_empty(&port->dports))
+	if (xa_empty(&port->dports))
 		return -EINVAL;
 
 	write_seqlock(&cxlsd->target_lock);
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index de5cb8288cd4..bf5f0c305115 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -344,7 +344,7 @@ struct cxl_port {
 	struct device *uport;
 	struct device *host_bridge;
 	int id;
-	struct list_head dports;
+	struct xarray dports;
 	struct xarray endpoints;
 	struct cxl_dport *parent_dport;
 	struct ida decoder_ida;
@@ -354,20 +354,24 @@ struct cxl_port {
 	unsigned int depth;
 };
 
+static inline struct cxl_dport *
+cxl_find_dport_by_dev(struct cxl_port *port, const struct device *dport_dev)
+{
+	return xa_load(&port->dports, (unsigned long)dport_dev);
+}
+
 /**
  * struct cxl_dport - CXL downstream port
  * @dport: PCI bridge or firmware device representing the downstream link
  * @port_id: unique hardware identifier for dport in decoder target list
  * @component_reg_phys: downstream port component registers
  * @port: reference to cxl_port that contains this downstream port
- * @list: node for a cxl_port's list of cxl_dport instances
  */
 struct cxl_dport {
 	struct device *dport;
 	int port_id;
 	resource_size_t component_reg_phys;
 	struct cxl_port *port;
-	struct list_head list;
 };
 
 /**
@@ -410,8 +414,6 @@ bool schedule_cxl_memdev_detach(struct cxl_memdev *cxlmd);
 struct cxl_dport *devm_cxl_add_dport(struct cxl_port *port,
 				     struct device *dport, int port_id,
 				     resource_size_t component_reg_phys);
-struct cxl_dport *cxl_find_dport_by_dev(struct cxl_port *port,
-					const struct device *dev);
 
 struct cxl_decoder *to_cxl_decoder(struct device *dev);
 struct cxl_root_decoder *to_cxl_root_decoder(struct device *dev);

[PATCH v2 14/28] cxl/hdm: Add sysfs attributes for interleave ways + granularity

[Dan Williams]

From: Ben Widawsky <bwidawsk@kernel.org>

The region provisioning flow involves selecting interleave ways +
granularity settings for a region, and then programming the decoder
topology to meet those constraints, if possible. For example, root
decoders set the minimum interleave ways + granularity for any hosted
regions.

Given decoder programming is not atomic and collisions can occur between
multiple requesting regions userspace will be responsible for conflict
resolution and it needs these attributes to make those decisions.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: reword changelog, make read-only, add sysfs ABI documentaion]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   27 +++++++++++++++++++++++++++
 drivers/cxl/core/port.c                 |   23 +++++++++++++++++++++++
 2 files changed, 50 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 9b6cc7cdc73b..0362ae98218e 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -230,3 +230,30 @@ Description:
 		allocations are enforced to occur in increasing 'decoderX.Y/id'
 		order and frees are enforced to occur in decreasing
 		'decoderX.Y/id' order.
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/interleave_ways
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RO) The number of targets across which this decoder's host
+		physical address (HPA) memory range is interleaved. The device
+		maps every Nth block of HPA (of size ==
+		'interleave_granularity') to consecutive DPA addresses. The
+		decoder's position in the interleave is determined by the
+		device's (endpoint or switch) switch ancestry. For root
+		decoders their interleave is specified by platform firmware and
+		they only specify a downstream target order for host bridges.
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/interleave_granularity
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RO) The number of consecutive bytes of host physical address
+		space this decoder claims at address N before the decode rotates
+		to the next target in the interleave at address N +
+		interleave_granularity (assuming N is aligned to
+		interleave_granularity).
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index b2c44e7ef6a8..a43735f349d6 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -260,10 +260,33 @@ static ssize_t dpa_size_store(struct device *dev, struct device_attribute *attr,
 }
 static DEVICE_ATTR_RW(dpa_size);
 
+static ssize_t interleave_granularity_show(struct device *dev,
+					   struct device_attribute *attr,
+					   char *buf)
+{
+	struct cxl_decoder *cxld = to_cxl_decoder(dev);
+
+	return sysfs_emit(buf, "%d\n", cxld->interleave_granularity);
+}
+
+static DEVICE_ATTR_RO(interleave_granularity);
+
+static ssize_t interleave_ways_show(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	struct cxl_decoder *cxld = to_cxl_decoder(dev);
+
+	return sysfs_emit(buf, "%d\n", cxld->interleave_ways);
+}
+
+static DEVICE_ATTR_RO(interleave_ways);
+
 static struct attribute *cxl_decoder_base_attrs[] = {
 	&dev_attr_start.attr,
 	&dev_attr_size.attr,
 	&dev_attr_locked.attr,
+	&dev_attr_interleave_granularity.attr,
+	&dev_attr_interleave_ways.attr,
 	NULL,
 };
 

[PATCH v2 15/28] cxl/mem: Enumerate port targets before adding endpoints

[Dan Williams]

The port scanning algorithm in devm_cxl_enumerate_ports() walks up the
topology and adds cxl_port objects starting from the root down to the
endpoint. When those ports are initially created they know all their
dports, but they do not know the downstream cxl_port instance that
represents the next descendant in the topology. Rework create_endpoint()
into devm_cxl_add_endpoint() that enumerates the downstream cxl_port
topology into each port's 'struct cxl_ep' record for each endpoint it
that the port is an ancestor.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-7-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/port.c |   41 +++++++++++++++++++++++++++++++++++++++++
 drivers/cxl/cxl.h       |    5 +++++
 drivers/cxl/mem.c       |   30 +-----------------------------
 3 files changed, 47 insertions(+), 29 deletions(-)

diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index a43735f349d6..4907db798ad9 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -1087,6 +1087,47 @@ static struct cxl_ep *cxl_ep_load(struct cxl_port *port,
 	return xa_load(&port->endpoints, (unsigned long)&cxlmd->dev);
 }
 
+int devm_cxl_add_endpoint(struct cxl_memdev *cxlmd,
+			  struct cxl_dport *parent_dport)
+{
+	struct cxl_port *parent_port = parent_dport->port;
+	struct cxl_dev_state *cxlds = cxlmd->cxlds;
+	struct cxl_port *endpoint, *iter, *down;
+	int rc;
+
+	/*
+	 * Now that the path to the root is established record all the
+	 * intervening ports in the chain.
+	 */
+	for (iter = parent_port, down = NULL; !is_cxl_root(iter);
+	     down = iter, iter = to_cxl_port(iter->dev.parent)) {
+		struct cxl_ep *ep;
+
+		ep = cxl_ep_load(iter, cxlmd);
+		ep->next = down;
+	}
+
+	endpoint = devm_cxl_add_port(&parent_port->dev, &cxlmd->dev,
+				     cxlds->component_reg_phys, parent_dport);
+	if (IS_ERR(endpoint))
+		return PTR_ERR(endpoint);
+
+	dev_dbg(&cxlmd->dev, "add: %s\n", dev_name(&endpoint->dev));
+
+	rc = cxl_endpoint_autoremove(cxlmd, endpoint);
+	if (rc)
+		return rc;
+
+	if (!endpoint->dev.driver) {
+		dev_err(&cxlmd->dev, "%s failed probe\n",
+			dev_name(&endpoint->dev));
+		return -ENXIO;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_NS_GPL(devm_cxl_add_endpoint, CXL);
+
 static void cxl_detach_ep(void *data)
 {
 	struct cxl_memdev *cxlmd = data;
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index bf5f0c305115..a108d5c288ca 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -378,10 +378,13 @@ struct cxl_dport {
  * struct cxl_ep - track an endpoint's interest in a port
  * @ep: device that hosts a generic CXL endpoint (expander or accelerator)
  * @dport: which dport routes to this endpoint on @port
+ * @next: cxl switch port across the link attached to @dport NULL if
+ *	  attached to an endpoint
  */
 struct cxl_ep {
 	struct device *ep;
 	struct cxl_dport *dport;
+	struct cxl_port *next;
 };
 
 /*
@@ -404,6 +407,8 @@ struct pci_bus *cxl_port_to_pci_bus(struct cxl_port *port);
 struct cxl_port *devm_cxl_add_port(struct device *host, struct device *uport,
 				   resource_size_t component_reg_phys,
 				   struct cxl_dport *parent_dport);
+int devm_cxl_add_endpoint(struct cxl_memdev *cxlmd,
+			  struct cxl_dport *parent_dport);
 struct cxl_port *find_cxl_root(struct device *dev);
 int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd);
 int cxl_bus_rescan(void);
diff --git a/drivers/cxl/mem.c b/drivers/cxl/mem.c
index 2786d3402c9e..64ccf053d32c 100644
--- a/drivers/cxl/mem.c
+++ b/drivers/cxl/mem.c
@@ -25,34 +25,6 @@
  * in higher level operations.
  */
 
-static int create_endpoint(struct cxl_memdev *cxlmd,
-			   struct cxl_dport *parent_dport)
-{
-	struct cxl_port *parent_port = parent_dport->port;
-	struct cxl_dev_state *cxlds = cxlmd->cxlds;
-	struct cxl_port *endpoint;
-	int rc;
-
-	endpoint = devm_cxl_add_port(&parent_port->dev, &cxlmd->dev,
-				     cxlds->component_reg_phys, parent_dport);
-	if (IS_ERR(endpoint))
-		return PTR_ERR(endpoint);
-
-	dev_dbg(&cxlmd->dev, "add: %s\n", dev_name(&endpoint->dev));
-
-	rc = cxl_endpoint_autoremove(cxlmd, endpoint);
-	if (rc)
-		return rc;
-
-	if (!endpoint->dev.driver) {
-		dev_err(&cxlmd->dev, "%s failed probe\n",
-			dev_name(&endpoint->dev));
-		return -ENXIO;
-	}
-
-	return 0;
-}
-
 static void enable_suspend(void *data)
 {
 	cxl_mem_active_dec();
@@ -116,7 +88,7 @@ static int cxl_mem_probe(struct device *dev)
 		goto unlock;
 	}
 
-	rc = create_endpoint(cxlmd, dport);
+	rc = devm_cxl_add_endpoint(cxlmd, dport);
 unlock:
 	device_unlock(&parent_port->dev);
 	put_device(&parent_port->dev);

[PATCH v2 16/28] resource: Introduce alloc_free_mem_region()

[Dan Williams]

The core of devm_request_free_mem_region() is a helper that searches for
free space in iomem_resource and performs __request_region_locked() on
the result of that search. The policy choices of the implementation
conform to what CONFIG_DEVICE_PRIVATE users want which is memory that is
immediately marked busy, and a preference to search for the first-fit
free range in descending order from the top of the physical address
space.

CXL has a need for a similar allocator, but with the following tweaks:

1/ Search for free space in ascending order

2/ Search for free space relative to a given CXL window

3/ 'insert' rather than 'request' the new resource given downstream
   drivers from the CXL Region driver (like the pmem or dax drivers) are
   responsible for request_mem_region() when they activate the memory
   range.

Rework __request_free_mem_region() into get_free_mem_region() which
takes a set of GFR_* (Get Free Region) flags to control the allocation
policy (ascending vs descending), and "busy" policy (insert_resource()
vs request_region()).

As part of the consolidation of the legacy GFR_REQUEST_REGION case with
the new default of just inserting a new resource into the free space
some minor cleanups like not checking for NULL before calling
devres_free() (which does its own check) is included.

Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/linux-cxl/20220420143406.GY2120790@nvidia.com/
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 include/linux/ioport.h |    2 +
 kernel/resource.c      |  178 +++++++++++++++++++++++++++++++++++++++---------
 mm/Kconfig             |    5 +
 3 files changed, 150 insertions(+), 35 deletions(-)

diff --git a/include/linux/ioport.h b/include/linux/ioport.h
index 79d1ad6d6275..616b683563a9 100644
--- a/include/linux/ioport.h
+++ b/include/linux/ioport.h
@@ -330,6 +330,8 @@ struct resource *devm_request_free_mem_region(struct device *dev,
 		struct resource *base, unsigned long size);
 struct resource *request_free_mem_region(struct resource *base,
 		unsigned long size, const char *name);
+struct resource *alloc_free_mem_region(struct resource *base,
+		unsigned long size, unsigned long align, const char *name);
 
 static inline void irqresource_disabled(struct resource *res, u32 irq)
 {
diff --git a/kernel/resource.c b/kernel/resource.c
index 53a534db350e..4c5e80b92f2f 100644
--- a/kernel/resource.c
+++ b/kernel/resource.c
@@ -489,8 +489,9 @@ int __weak page_is_ram(unsigned long pfn)
 }
 EXPORT_SYMBOL_GPL(page_is_ram);
 
-static int __region_intersects(resource_size_t start, size_t size,
-			unsigned long flags, unsigned long desc)
+static int __region_intersects(struct resource *parent, resource_size_t start,
+			       size_t size, unsigned long flags,
+			       unsigned long desc)
 {
 	struct resource res;
 	int type = 0; int other = 0;
@@ -499,7 +500,7 @@ static int __region_intersects(resource_size_t start, size_t size,
 	res.start = start;
 	res.end = start + size - 1;
 
-	for (p = iomem_resource.child; p ; p = p->sibling) {
+	for (p = parent->child; p ; p = p->sibling) {
 		bool is_type = (((p->flags & flags) == flags) &&
 				((desc == IORES_DESC_NONE) ||
 				 (desc == p->desc)));
@@ -543,7 +544,7 @@ int region_intersects(resource_size_t start, size_t size, unsigned long flags,
 	int ret;
 
 	read_lock(&resource_lock);
-	ret = __region_intersects(start, size, flags, desc);
+	ret = __region_intersects(&iomem_resource, start, size, flags, desc);
 	read_unlock(&resource_lock);
 
 	return ret;
@@ -1780,62 +1781,139 @@ void resource_list_free(struct list_head *head)
 }
 EXPORT_SYMBOL(resource_list_free);
 
-#ifdef CONFIG_DEVICE_PRIVATE
-static struct resource *__request_free_mem_region(struct device *dev,
-		struct resource *base, unsigned long size, const char *name)
+#ifdef CONFIG_GET_FREE_REGION
+#define GFR_DESCENDING		(1UL << 0)
+#define GFR_REQUEST_REGION	(1UL << 1)
+#define GFR_DEFAULT_ALIGN (1UL << PA_SECTION_SHIFT)
+
+static resource_size_t gfr_start(struct resource *base, resource_size_t size,
+				 resource_size_t align, unsigned long flags)
+{
+	if (flags & GFR_DESCENDING) {
+		resource_size_t end;
+
+		end = min_t(resource_size_t, base->end,
+			    (1ULL << MAX_PHYSMEM_BITS) - 1);
+		return end - size + 1;
+	}
+
+	return ALIGN(base->start, align);
+}
+
+static bool gfr_continue(struct resource *base, resource_size_t addr,
+			 resource_size_t size, unsigned long flags)
+{
+	if (flags & GFR_DESCENDING)
+		return addr > size && addr >= base->start;
+	/*
+	 * In the ascend case be careful that the last increment by
+	 * @size did not wrap 0.
+	 */
+	return addr > addr - size &&
+	       addr <= min_t(resource_size_t, base->end,
+			     (1ULL << MAX_PHYSMEM_BITS) - 1);
+}
+
+static resource_size_t gfr_next(resource_size_t addr, resource_size_t size,
+				unsigned long flags)
+{
+	if (flags & GFR_DESCENDING)
+		return addr - size;
+	return addr + size;
+}
+
+static void remove_free_mem_region(void *_res)
+{
+	struct resource *res = _res;
+
+	if (res->parent)
+		remove_resource(res);
+	free_resource(res);
+}
+
+static struct resource *
+get_free_mem_region(struct device *dev, struct resource *base,
+		    resource_size_t size, const unsigned long align,
+		    const char *name, const unsigned long desc,
+		    const unsigned long flags)
 {
-	resource_size_t end, addr;
+	resource_size_t addr;
 	struct resource *res;
 	struct region_devres *dr = NULL;
 
-	size = ALIGN(size, 1UL << PA_SECTION_SHIFT);
-	end = min_t(unsigned long, base->end, (1UL << MAX_PHYSMEM_BITS) - 1);
-	addr = end - size + 1UL;
+	size = ALIGN(size, align);
 
 	res = alloc_resource(GFP_KERNEL);
 	if (!res)
 		return ERR_PTR(-ENOMEM);
 
-	if (dev) {
+	if (dev && (flags & GFR_REQUEST_REGION)) {
 		dr = devres_alloc(devm_region_release,
 				sizeof(struct region_devres), GFP_KERNEL);
 		if (!dr) {
 			free_resource(res);
 			return ERR_PTR(-ENOMEM);
 		}
+	} else if (dev) {
+		if (devm_add_action_or_reset(dev, remove_free_mem_region, res))
+			return ERR_PTR(-ENOMEM);
 	}
 
 	write_lock(&resource_lock);
-	for (; addr > size && addr >= base->start; addr -= size) {
-		if (__region_intersects(addr, size, 0, IORES_DESC_NONE) !=
-				REGION_DISJOINT)
+	for (addr = gfr_start(base, size, align, flags);
+	     gfr_continue(base, addr, size, flags);
+	     addr = gfr_next(addr, size, flags)) {
+		if (__region_intersects(base, addr, size, 0, IORES_DESC_NONE) !=
+		    REGION_DISJOINT)
 			continue;
 
-		if (__request_region_locked(res, &iomem_resource, addr, size,
-						name, 0))
-			break;
+		if (flags & GFR_REQUEST_REGION) {
+			if (__request_region_locked(res, &iomem_resource, addr,
+						    size, name, 0))
+				break;
 
-		if (dev) {
-			dr->parent = &iomem_resource;
-			dr->start = addr;
-			dr->n = size;
-			devres_add(dev, dr);
-		}
+			if (dev) {
+				dr->parent = &iomem_resource;
+				dr->start = addr;
+				dr->n = size;
+				devres_add(dev, dr);
+			}
 
-		res->desc = IORES_DESC_DEVICE_PRIVATE_MEMORY;
-		write_unlock(&resource_lock);
+			res->desc = desc;
+			write_unlock(&resource_lock);
+
+
+			/*
+			 * A driver is claiming this region so revoke any
+			 * mappings.
+			 */
+			revoke_iomem(res);
+		} else {
+			res->start = addr;
+			res->end = addr + size - 1;
+			res->name = name;
+			res->desc = desc;
+			res->flags = IORESOURCE_MEM;
+
+			/*
+			 * Only succeed if the resource hosts an exclusive
+			 * range after the insert
+			 */
+			if (__insert_resource(base, res) || res->child)
+				break;
+
+			write_unlock(&resource_lock);
+		}
 
-		/*
-		 * A driver is claiming this region so revoke any mappings.
-		 */
-		revoke_iomem(res);
 		return res;
 	}
 	write_unlock(&resource_lock);
 
-	free_resource(res);
-	if (dr)
+	if (flags & GFR_REQUEST_REGION) {
+		free_resource(res);
 		devres_free(dr);
+	} else if (dev)
+		devm_release_action(dev, remove_free_mem_region, res);
 
 	return ERR_PTR(-ERANGE);
 }
@@ -1854,18 +1932,48 @@ static struct resource *__request_free_mem_region(struct device *dev,
 struct resource *devm_request_free_mem_region(struct device *dev,
 		struct resource *base, unsigned long size)
 {
-	return __request_free_mem_region(dev, base, size, dev_name(dev));
+	unsigned long flags = GFR_DESCENDING | GFR_REQUEST_REGION;
+
+	return get_free_mem_region(dev, base, size, GFR_DEFAULT_ALIGN,
+				   dev_name(dev),
+				   IORES_DESC_DEVICE_PRIVATE_MEMORY, flags);
 }
 EXPORT_SYMBOL_GPL(devm_request_free_mem_region);
 
 struct resource *request_free_mem_region(struct resource *base,
 		unsigned long size, const char *name)
 {
-	return __request_free_mem_region(NULL, base, size, name);
+	unsigned long flags = GFR_DESCENDING | GFR_REQUEST_REGION;
+
+	return get_free_mem_region(NULL, base, size, GFR_DEFAULT_ALIGN, name,
+				   IORES_DESC_DEVICE_PRIVATE_MEMORY, flags);
 }
 EXPORT_SYMBOL_GPL(request_free_mem_region);
 
-#endif /* CONFIG_DEVICE_PRIVATE */
+/**
+ * alloc_free_mem_region - find a free region relative to @base
+ * @base: resource that will parent the new resource
+ * @size: size in bytes of memory to allocate from @base
+ * @align: alignment requirements for the allocation
+ * @name: resource name
+ *
+ * Buses like CXL, that can dynamically instantiate new memory regions,
+ * need a method to allocate physical address space for those regions.
+ * Allocate and insert a new resource to cover a free, unclaimed by a
+ * descendant of @base, range in the span of @base.
+ */
+struct resource *alloc_free_mem_region(struct resource *base,
+				       unsigned long size, unsigned long align,
+				       const char *name)
+{
+	/* Default of ascending direction and insert resource */
+	unsigned long flags = 0;
+
+	return get_free_mem_region(NULL, base, size, align, name,
+				   IORES_DESC_NONE, flags);
+}
+EXPORT_SYMBOL_NS_GPL(alloc_free_mem_region, CXL);
+#endif /* CONFIG_GET_FREE_REGION */
 
 static int __init strict_iomem(char *str)
 {
diff --git a/mm/Kconfig b/mm/Kconfig
index 169e64192e48..a5b4fee2e3fd 100644
--- a/mm/Kconfig
+++ b/mm/Kconfig
@@ -994,9 +994,14 @@ config HMM_MIRROR
 	bool
 	depends on MMU
 
+config GET_FREE_REGION
+	depends on SPARSEMEM
+	bool
+
 config DEVICE_PRIVATE
 	bool "Unaddressable device memory (GPU memory, ...)"
 	depends on ZONE_DEVICE
+	select GET_FREE_REGION
 
 	help
 	  Allows creation of struct pages to represent unaddressable device

[PATCH v2 17/28] cxl/region: Add region creation support

[Dan Williams]

From: Ben Widawsky <bwidawsk@kernel.org>

CXL 2.0 allows for dynamic provisioning of new memory regions (system
physical address resources like "System RAM" and "Persistent Memory").
Whereas DDR and PMEM resources are conveyed statically at boot, CXL
allows for assembling and instantiating new regions from the available
capacity of CXL memory expanders in the system.

Sysfs with an "echo $region_name > $create_region_attribute" interface
is chosen as the mechanism to initiate the provisioning process. This
was chosen over ioctl() and netlink() to keep the configuration
interface entirely in a pseudo-fs interface, and it was chosen over
configfs since, aside from this one creation event, the interface is
read-mostly. I.e. configfs supports cases where an object is designed to
be provisioned each boot, like an iSCSI storage target, and CXL region
creation is mostly for PMEM regions which are created usually once
per-lifetime of a server instance. This is an improvement over nvdimm
that pre-created "seed" devices that tended to confuse users looking to
determine which devices are active and which are idle.

Recall that the major change that CXL brings over previous persistent
memory architectures is the ability to dynamically define new regions.
Compare that to drivers like 'nfit' where the region configuration is
statically defined by platform firmware.

Regions are created as a child of a root decoder that encompasses an
address space with constraints. When created through sysfs, the root
decoder is explicit. When created from an LSA's region structure a root
decoder will possibly need to be inferred by the driver.

Upon region creation through sysfs, a vacant region is created with a
unique name. Regions have a number of attributes that must be configured
before the region can be bound to the driver where HDM decoder program
is completed.

An example of creating a new region:

- Allocate a new region name:
region=$(cat /sys/bus/cxl/devices/decoder0.0/create_pmem_region)

- Create a new region by name:
while
region=$(cat /sys/bus/cxl/devices/decoder0.0/create_pmem_region)
! echo $region > /sys/bus/cxl/devices/decoder0.0/create_pmem_region
do true; done

- Region now exists in sysfs:
stat -t /sys/bus/cxl/devices/decoder0.0/$region

- Delete the region, and name:
echo $region > /sys/bus/cxl/devices/decoder0.0/delete_region

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: simplify locking, reword changelog]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl         |   25 +++
 Documentation/driver-api/cxl/memory-devices.rst |   11 +
 drivers/cxl/Kconfig                             |    5 +
 drivers/cxl/core/Makefile                       |    1 
 drivers/cxl/core/core.h                         |   10 +
 drivers/cxl/core/port.c                         |   39 ++++
 drivers/cxl/core/region.c                       |  201 +++++++++++++++++++++++
 drivers/cxl/cxl.h                               |   18 ++
 tools/testing/cxl/Kbuild                        |    1 
 9 files changed, 311 insertions(+)
 create mode 100644 drivers/cxl/core/region.c

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 0362ae98218e..b6156a499a5a 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -257,3 +257,28 @@ Description:
 		to the next target in the interleave at address N +
 		interleave_granularity (assuming N is aligned to
 		interleave_granularity).
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/create_pmem_region
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) Write a string in the form 'regionZ' to start the process
+		of defining a new persistent memory region (interleave-set)
+		within the decode range bounded by root decoder 'decoderX.Y'.
+		The value written must match the current value returned from
+		reading this attribute. An atomic compare exchange operation is
+		done on write to assign the requested id to a region and
+		allocate the region-id for the next creation attempt. EBUSY is
+		returned if the region name written does not match the current
+		cached value.
+
+
+What:		/sys/bus/cxl/devices/decoderX.Y/delete_region
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(WO) Write a string in the form 'regionZ' to delete that region,
+		provided it is currently idle / not bound to a driver.
diff --git a/Documentation/driver-api/cxl/memory-devices.rst b/Documentation/driver-api/cxl/memory-devices.rst
index db476bb170b6..66ddc58a21b1 100644
--- a/Documentation/driver-api/cxl/memory-devices.rst
+++ b/Documentation/driver-api/cxl/memory-devices.rst
@@ -362,6 +362,17 @@ CXL Core
 .. kernel-doc:: drivers/cxl/core/mbox.c
    :doc: cxl mbox
 
+CXL Regions
+-----------
+.. kernel-doc:: drivers/cxl/region.h
+   :identifiers:
+
+.. kernel-doc:: drivers/cxl/core/region.c
+   :doc: cxl core region
+
+.. kernel-doc:: drivers/cxl/core/region.c
+   :identifiers:
+
 External Interfaces
 ===================
 
diff --git a/drivers/cxl/Kconfig b/drivers/cxl/Kconfig
index f64e3984689f..aa2728de419e 100644
--- a/drivers/cxl/Kconfig
+++ b/drivers/cxl/Kconfig
@@ -102,4 +102,9 @@ config CXL_SUSPEND
 	def_bool y
 	depends on SUSPEND && CXL_MEM
 
+config CXL_REGION
+	bool
+	default CXL_BUS
+	select MEMREGION
+
 endif
diff --git a/drivers/cxl/core/Makefile b/drivers/cxl/core/Makefile
index 9d35085d25af..79c7257f4107 100644
--- a/drivers/cxl/core/Makefile
+++ b/drivers/cxl/core/Makefile
@@ -10,3 +10,4 @@ cxl_core-y += memdev.o
 cxl_core-y += mbox.o
 cxl_core-y += pci.o
 cxl_core-y += hdm.o
+cxl_core-$(CONFIG_CXL_REGION) += region.o
diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index 5551b82b2da0..29272df7e212 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -9,6 +9,16 @@ extern const struct device_type cxl_nvdimm_type;
 
 extern struct attribute_group cxl_base_attribute_group;
 
+#ifdef CONFIG_CXL_REGION
+extern struct device_attribute dev_attr_create_pmem_region;
+extern struct device_attribute dev_attr_delete_region;
+#define CXL_REGION_ATTR(x) (&dev_attr_##x.attr)
+#define SET_CXL_REGION_ATTR(x) (&dev_attr_##x.attr),
+#else
+#define CXL_REGION_ATTR(x) NULL
+#define SET_CXL_REGION_ATTR(x)
+#endif
+
 struct cxl_send_command;
 struct cxl_mem_query_commands;
 int cxl_query_cmd(struct cxl_memdev *cxlmd,
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 4907db798ad9..89672b126b30 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-only
 /* Copyright(c) 2020 Intel Corporation. All rights reserved. */
 #include <linux/io-64-nonatomic-lo-hi.h>
+#include <linux/memregion.h>
 #include <linux/workqueue.h>
 #include <linux/debugfs.h>
 #include <linux/device.h>
@@ -300,11 +301,35 @@ static struct attribute *cxl_decoder_root_attrs[] = {
 	&dev_attr_cap_type2.attr,
 	&dev_attr_cap_type3.attr,
 	&dev_attr_target_list.attr,
+	SET_CXL_REGION_ATTR(create_pmem_region)
+	SET_CXL_REGION_ATTR(delete_region)
 	NULL,
 };
 
+static bool can_create_pmem(struct cxl_root_decoder *cxlrd)
+{
+	unsigned long flags = CXL_DECODER_F_TYPE3 | CXL_DECODER_F_PMEM;
+
+	return (cxlrd->cxlsd.cxld.flags & flags) == flags;
+}
+
+static umode_t cxl_root_decoder_visible(struct kobject *kobj, struct attribute *a, int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
+
+	if (a == CXL_REGION_ATTR(create_pmem_region) && !can_create_pmem(cxlrd))
+		return 0;
+
+	if (a == CXL_REGION_ATTR(delete_region) && !can_create_pmem(cxlrd))
+		return 0;
+
+	return a->mode;
+}
+
 static struct attribute_group cxl_decoder_root_attribute_group = {
 	.attrs = cxl_decoder_root_attrs,
+	.is_visible = cxl_root_decoder_visible,
 };
 
 static const struct attribute_group *cxl_decoder_root_attribute_groups[] = {
@@ -387,6 +412,8 @@ static void cxl_root_decoder_release(struct device *dev)
 {
 	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
 
+	if (atomic_read(&cxlrd->region_id) >= 0)
+		memregion_free(atomic_read(&cxlrd->region_id));
 	__cxl_decoder_release(&cxlrd->cxlsd.cxld);
 	kfree(cxlrd);
 }
@@ -1484,6 +1511,18 @@ struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
 
 	cxld = &cxlsd->cxld;
 	cxld->dev.type = &cxl_decoder_root_type;
+	/*
+	 * cxl_root_decoder_release() special cases negative ids to
+	 * detect memregion_alloc() failures.
+	 */
+	atomic_set(&cxlrd->region_id, -1);
+	rc = memregion_alloc(GFP_KERNEL);
+	if (rc < 0) {
+		put_device(&cxld->dev);
+		return ERR_PTR(rc);
+	}
+
+	atomic_set(&cxlrd->region_id, rc);
 	return cxlrd;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_root_decoder_alloc, CXL);
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
new file mode 100644
index 000000000000..6d2a7aa53379
--- /dev/null
+++ b/drivers/cxl/core/region.c
@@ -0,0 +1,201 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright(c) 2022 Intel Corporation. All rights reserved. */
+#include <linux/memregion.h>
+#include <linux/genalloc.h>
+#include <linux/device.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/idr.h>
+#include <cxl.h>
+#include "core.h"
+
+/**
+ * DOC: cxl core region
+ *
+ * CXL Regions represent mapped memory capacity in system physical address
+ * space. Whereas the CXL Root Decoders identify the bounds of potential CXL
+ * Memory ranges, Regions represent the active mapped capacity by the HDM
+ * Decoder Capability structures throughout the Host Bridges, Switches, and
+ * Endpoints in the topology.
+ */
+
+static struct cxl_region *to_cxl_region(struct device *dev);
+
+static void cxl_region_release(struct device *dev)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+
+	memregion_free(cxlr->id);
+	kfree(cxlr);
+}
+
+static const struct device_type cxl_region_type = {
+	.name = "cxl_region",
+	.release = cxl_region_release,
+};
+
+bool is_cxl_region(struct device *dev)
+{
+	return dev->type == &cxl_region_type;
+}
+EXPORT_SYMBOL_NS_GPL(is_cxl_region, CXL);
+
+static struct cxl_region *to_cxl_region(struct device *dev)
+{
+	if (dev_WARN_ONCE(dev, dev->type != &cxl_region_type,
+			  "not a cxl_region device\n"))
+		return NULL;
+
+	return container_of(dev, struct cxl_region, dev);
+}
+
+static void unregister_region(void *dev)
+{
+	device_unregister(dev);
+}
+
+static struct lock_class_key cxl_region_key;
+
+static struct cxl_region *cxl_region_alloc(struct cxl_root_decoder *cxlrd, int id)
+{
+	struct cxl_region *cxlr;
+	struct device *dev;
+
+	cxlr = kzalloc(sizeof(*cxlr), GFP_KERNEL);
+	if (!cxlr) {
+		memregion_free(id);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	dev = &cxlr->dev;
+	device_initialize(dev);
+	lockdep_set_class(&dev->mutex, &cxl_region_key);
+	dev->parent = &cxlrd->cxlsd.cxld.dev;
+	device_set_pm_not_required(dev);
+	dev->bus = &cxl_bus_type;
+	dev->type = &cxl_region_type;
+	cxlr->id = id;
+
+	return cxlr;
+}
+
+/**
+ * devm_cxl_add_region - Adds a region to a decoder
+ * @cxlrd: root decoder
+ * @id: memregion id to create, or memregion_free() on failure
+ * @mode: mode for the endpoint decoders of this region
+ * @type: select whether this is an expander or accelerator (type-2 or type-3)
+ *
+ * This is the second step of region initialization. Regions exist within an
+ * address space which is mapped by a @cxlrd.
+ *
+ * Return: 0 if the region was added to the @cxlrd, else returns negative error
+ * code. The region will be named "regionZ" where Z is the unique region number.
+ */
+static struct cxl_region *devm_cxl_add_region(struct cxl_root_decoder *cxlrd,
+					      int id,
+					      enum cxl_decoder_mode mode,
+					      enum cxl_decoder_type type)
+{
+	struct cxl_port *port = to_cxl_port(cxlrd->cxlsd.cxld.dev.parent);
+	struct cxl_region *cxlr;
+	struct device *dev;
+	int rc;
+
+	cxlr = cxl_region_alloc(cxlrd, id);
+	if (IS_ERR(cxlr))
+		return cxlr;
+	cxlr->mode = mode;
+	cxlr->type = type;
+
+	dev = &cxlr->dev;
+	rc = dev_set_name(dev, "region%d", id);
+	if (rc)
+		goto err;
+
+	rc = device_add(dev);
+	if (rc)
+		goto err;
+
+	rc = devm_add_action_or_reset(port->uport, unregister_region, cxlr);
+	if (rc)
+		return ERR_PTR(rc);
+
+	dev_dbg(port->uport, "%s: created %s\n",
+		dev_name(&cxlrd->cxlsd.cxld.dev), dev_name(dev));
+	return cxlr;
+
+err:
+	put_device(dev);
+	return ERR_PTR(rc);
+}
+
+static ssize_t create_pmem_region_show(struct device *dev,
+				       struct device_attribute *attr, char *buf)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
+
+	return sysfs_emit(buf, "region%u\n", atomic_read(&cxlrd->region_id));
+}
+
+static ssize_t create_pmem_region_store(struct device *dev,
+					struct device_attribute *attr,
+					const char *buf, size_t len)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
+	struct cxl_region *cxlr;
+	unsigned int id, rc;
+
+	rc = sscanf(buf, "region%u\n", &id);
+	if (rc != 1)
+		return -EINVAL;
+
+	rc = memregion_alloc(GFP_KERNEL);
+	if (rc < 0)
+		return rc;
+
+	if (atomic_cmpxchg(&cxlrd->region_id, id, rc) != id) {
+		memregion_free(rc);
+		return -EBUSY;
+	}
+
+	cxlr = devm_cxl_add_region(cxlrd, id, CXL_DECODER_PMEM,
+				   CXL_DECODER_EXPANDER);
+	if (IS_ERR(cxlr))
+		return PTR_ERR(cxlr);
+
+	return len;
+}
+DEVICE_ATTR_RW(create_pmem_region);
+
+static struct cxl_region *
+cxl_find_region_by_name(struct cxl_root_decoder *cxlrd, const char *name)
+{
+	struct cxl_decoder *cxld = &cxlrd->cxlsd.cxld;
+	struct device *region_dev;
+
+	region_dev = device_find_child_by_name(&cxld->dev, name);
+	if (!region_dev)
+		return ERR_PTR(-ENODEV);
+
+	return to_cxl_region(region_dev);
+}
+
+static ssize_t delete_region_store(struct device *dev,
+				   struct device_attribute *attr,
+				   const char *buf, size_t len)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev);
+	struct cxl_port *port = to_cxl_port(dev->parent);
+	struct cxl_region *cxlr;
+
+	cxlr = cxl_find_region_by_name(cxlrd, buf);
+	if (IS_ERR(cxlr))
+		return PTR_ERR(cxlr);
+
+	devm_release_action(port->uport, unregister_region, cxlr);
+	put_device(&cxlr->dev);
+
+	return len;
+}
+DEVICE_ATTR_WO(delete_region);
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index a108d5c288ca..c3696e76306a 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -286,13 +286,29 @@ struct cxl_switch_decoder {
 /**
  * struct cxl_root_decoder - Static platform CXL address decoder
  * @res: host / parent resource for region allocations
+ * @region_id: region id for next region provisioning event
  * @cxlsd: base cxl switch decoder
  */
 struct cxl_root_decoder {
 	struct resource *res;
+	atomic_t region_id;
 	struct cxl_switch_decoder cxlsd;
 };
 
+/**
+ * struct cxl_region - CXL region
+ * @dev: This region's device
+ * @id: This region's id. Id is globally unique across all regions
+ * @mode: Endpoint decoder allocation / access mode
+ * @type: Endpoint decoder target type
+ */
+struct cxl_region {
+	struct device dev;
+	int id;
+	enum cxl_decoder_mode mode;
+	enum cxl_decoder_type type;
+};
+
 /**
  * enum cxl_nvdimm_brige_state - state machine for managing bus rescans
  * @CXL_NVB_NEW: Set at bridge create and after cxl_pmem_wq is destroyed
@@ -440,6 +456,8 @@ struct cxl_hdm *devm_cxl_setup_hdm(struct cxl_port *port);
 int devm_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm);
 int devm_cxl_add_passthrough_decoder(struct cxl_port *port);
 
+bool is_cxl_region(struct device *dev);
+
 extern struct bus_type cxl_bus_type;
 
 struct cxl_driver {
diff --git a/tools/testing/cxl/Kbuild b/tools/testing/cxl/Kbuild
index 33543231d453..500be85729cc 100644
--- a/tools/testing/cxl/Kbuild
+++ b/tools/testing/cxl/Kbuild
@@ -47,6 +47,7 @@ cxl_core-y += $(CXL_CORE_SRC)/memdev.o
 cxl_core-y += $(CXL_CORE_SRC)/mbox.o
 cxl_core-y += $(CXL_CORE_SRC)/pci.o
 cxl_core-y += $(CXL_CORE_SRC)/hdm.o
+cxl_core-$(CONFIG_CXL_REGION) += $(CXL_CORE_SRC)/region.o
 cxl_core-y += config_check.o
 
 obj-m += test/

[PATCH v2 18/28] cxl/region: Add a 'uuid' attribute

[Dan Williams]

From: Ben Widawsky <bwidawsk@kernel.org>

The process of provisioning a region involves triggering the creation of
a new region object, pouring in the configuration, and then binding that
configured object to the region driver to start its operation. For
persistent memory regions the CXL specification mandates that it
identified by a uuid. Add an ABI for userspace to specify a region's
uuid.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: simplify locking]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   10 +++
 drivers/cxl/core/region.c               |  118 +++++++++++++++++++++++++++++++
 drivers/cxl/cxl.h                       |   25 +++++++
 3 files changed, 153 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index b6156a499a5a..0760b8402c23 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -282,3 +282,13 @@ Contact:	linux-cxl@vger.kernel.org
 Description:
 		(WO) Write a string in the form 'regionZ' to delete that region,
 		provided it is currently idle / not bound to a driver.
+
+
+What:		/sys/bus/cxl/devices/regionZ/uuid
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) Write a unique identifier for the region. This field must
+		be set for persistent regions and it must not conflict with the
+		UUID of another region.
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 6d2a7aa53379..22dccb4702e5 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -5,6 +5,7 @@
 #include <linux/device.h>
 #include <linux/module.h>
 #include <linux/slab.h>
+#include <linux/uuid.h>
 #include <linux/idr.h>
 #include <cxl.h>
 #include "core.h"
@@ -17,10 +18,126 @@
  * Memory ranges, Regions represent the active mapped capacity by the HDM
  * Decoder Capability structures throughout the Host Bridges, Switches, and
  * Endpoints in the topology.
+ *
+ * Region configuration has ordering constraints. UUID may be set at any time
+ * but is only visible for persistent regions.
+ */
+
+/*
+ * All changes to the interleave configuration occur with this lock held
+ * for write.
  */
+static DECLARE_RWSEM(cxl_region_rwsem);
 
 static struct cxl_region *to_cxl_region(struct device *dev);
 
+static ssize_t uuid_show(struct device *dev, struct device_attribute *attr,
+			 char *buf)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	rc = sysfs_emit(buf, "%pUb\n", &p->uuid);
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+
+static int is_dup(struct device *match, void *data)
+{
+	struct cxl_region_params *p;
+	struct cxl_region *cxlr;
+	uuid_t *uuid = data;
+
+	if (!is_cxl_region(match))
+		return 0;
+
+	lockdep_assert_held(&cxl_region_rwsem);
+	cxlr = to_cxl_region(match);
+	p = &cxlr->params;
+
+	if (uuid_equal(&p->uuid, uuid)) {
+		dev_dbg(match, "already has uuid: %pUb\n", uuid);
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+static ssize_t uuid_store(struct device *dev, struct device_attribute *attr,
+			  const char *buf, size_t len)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	uuid_t temp;
+	ssize_t rc;
+
+	if (len != UUID_STRING_LEN + 1)
+		return -EINVAL;
+
+	rc = uuid_parse(buf, &temp);
+	if (rc)
+		return rc;
+
+	if (uuid_is_null(&temp))
+		return -EINVAL;
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+
+	if (uuid_equal(&p->uuid, &temp))
+		goto out;
+
+	rc = -EBUSY;
+	if (p->state >= CXL_CONFIG_ACTIVE)
+		goto out;
+
+	rc = bus_for_each_dev(&cxl_bus_type, NULL, &temp, is_dup);
+	if (rc < 0)
+		goto out;
+
+	uuid_copy(&p->uuid, &temp);
+out:
+	up_write(&cxl_region_rwsem);
+
+	if (rc)
+		return rc;
+	return len;
+}
+static DEVICE_ATTR_RW(uuid);
+
+static umode_t cxl_region_visible(struct kobject *kobj, struct attribute *a,
+				  int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct cxl_region *cxlr = to_cxl_region(dev);
+
+	if (a == &dev_attr_uuid.attr && cxlr->mode != CXL_DECODER_PMEM)
+		return 0;
+	return a->mode;
+}
+
+static struct attribute *cxl_region_attrs[] = {
+	&dev_attr_uuid.attr,
+	NULL,
+};
+
+static const struct attribute_group cxl_region_group = {
+	.attrs = cxl_region_attrs,
+	.is_visible = cxl_region_visible,
+};
+
+static const struct attribute_group *region_groups[] = {
+	&cxl_base_attribute_group,
+	&cxl_region_group,
+	NULL,
+};
+
 static void cxl_region_release(struct device *dev)
 {
 	struct cxl_region *cxlr = to_cxl_region(dev);
@@ -32,6 +149,7 @@ static void cxl_region_release(struct device *dev)
 static const struct device_type cxl_region_type = {
 	.name = "cxl_region",
 	.release = cxl_region_release,
+	.groups = region_groups
 };
 
 bool is_cxl_region(struct device *dev)
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index c3696e76306a..cf2ece363015 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -295,18 +295,43 @@ struct cxl_root_decoder {
 	struct cxl_switch_decoder cxlsd;
 };
 
+/*
+ * enum cxl_config_state - State machine for region configuration
+ * @CXL_CONFIG_IDLE: Any sysfs attribute can be written freely
+ * @CXL_CONFIG_ACTIVE: All targets have been added the region is now
+ * active
+ */
+enum cxl_config_state {
+	CXL_CONFIG_IDLE,
+	CXL_CONFIG_ACTIVE,
+};
+
+/**
+ * struct cxl_region_params - region settings
+ * @state: allow the driver to lockdown further parameter changes
+ * @uuid: unique id for persistent regions
+ *
+ * State transitions are protected by the cxl_region_rwsem
+ */
+struct cxl_region_params {
+	enum cxl_config_state state;
+	uuid_t uuid;
+};
+
 /**
  * struct cxl_region - CXL region
  * @dev: This region's device
  * @id: This region's id. Id is globally unique across all regions
  * @mode: Endpoint decoder allocation / access mode
  * @type: Endpoint decoder target type
+ * @params: active + config params for the region
  */
 struct cxl_region {
 	struct device dev;
 	int id;
 	enum cxl_decoder_mode mode;
 	enum cxl_decoder_type type;
+	struct cxl_region_params params;
 };
 
 /**

[PATCH v2 19/28] cxl/region: Add interleave geometry attributes

[Dan Williams]

From: Ben Widawsky <bwidawsk@kernel.org>

Add ABI to allow the number of devices that comprise a region to be
set as well as the interleave granularity for the region.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: reword changelog]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-11-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   21 +++++
 drivers/cxl/core/region.c               |  128 +++++++++++++++++++++++++++++++
 drivers/cxl/cxl.h                       |   33 ++++++++
 3 files changed, 182 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 0760b8402c23..bfa42bcc8383 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -292,3 +292,24 @@ Description:
 		(RW) Write a unique identifier for the region. This field must
 		be set for persistent regions and it must not conflict with the
 		UUID of another region.
+
+
+What:		/sys/bus/cxl/devices/regionZ/interleave_granularity
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) Set the number of consecutive bytes each device in the
+		interleave set will claim. The possible interleave granularity
+		values are determined by the CXL spec and the participating
+		devices.
+
+
+What:		/sys/bus/cxl/devices/regionZ/interleave_ways
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) Configures the number of devices participating in the
+		region is set by writing this value. Each device will provide
+		1/interleave_ways of storage for the region.
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 22dccb4702e5..3289caa5d882 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -7,6 +7,7 @@
 #include <linux/slab.h>
 #include <linux/uuid.h>
 #include <linux/idr.h>
+#include <cxlmem.h>
 #include <cxl.h>
 #include "core.h"
 
@@ -21,6 +22,8 @@
  *
  * Region configuration has ordering constraints. UUID may be set at any time
  * but is only visible for persistent regions.
+ * 1. Interleave granularity
+ * 2. Interleave size
  */
 
 /*
@@ -122,8 +125,129 @@ static umode_t cxl_region_visible(struct kobject *kobj, struct attribute *a,
 	return a->mode;
 }
 
+static ssize_t interleave_ways_show(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	rc = sysfs_emit(buf, "%d\n", p->interleave_ways);
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+
+static ssize_t interleave_ways_store(struct device *dev,
+				     struct device_attribute *attr,
+				     const char *buf, size_t len)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev->parent);
+	struct cxl_decoder *cxld = &cxlrd->cxlsd.cxld;
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	int rc, val;
+	u8 iw;
+
+	rc = kstrtoint(buf, 0, &val);
+	if (rc)
+		return rc;
+
+	rc = ways_to_cxl(val, &iw);
+	if (rc)
+		return rc;
+
+	/*
+	 * Even for x3, x9, and x12 interleaves the region interleave must be a
+	 * power of 2 multiple of the host bridge interleave.
+	 */
+	if (!is_power_of_2(val / cxld->interleave_ways) ||
+	    (val % cxld->interleave_ways)) {
+		dev_dbg(&cxlr->dev, "invalid interleave: %d\n", val);
+		return -EINVAL;
+	}
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	if (p->state >= CXL_CONFIG_INTERLEAVE_ACTIVE) {
+		rc = -EBUSY;
+		goto out;
+	}
+
+	p->interleave_ways = val;
+out:
+	up_write(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	return len;
+}
+static DEVICE_ATTR_RW(interleave_ways);
+
+static ssize_t interleave_granularity_show(struct device *dev,
+					   struct device_attribute *attr,
+					   char *buf)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	rc = sysfs_emit(buf, "%d\n", p->interleave_granularity);
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+
+static ssize_t interleave_granularity_store(struct device *dev,
+					    struct device_attribute *attr,
+					    const char *buf, size_t len)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(dev->parent);
+	struct cxl_decoder *cxld = &cxlrd->cxlsd.cxld;
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	int rc, val;
+	u16 ig;
+
+	rc = kstrtoint(buf, 0, &val);
+	if (rc)
+		return rc;
+
+	rc = granularity_to_cxl(val, &ig);
+	if (rc)
+		return rc;
+
+	/* region granularity must be >= root granularity */
+	if (val < cxld->interleave_granularity)
+		return -EINVAL;
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	if (p->state >= CXL_CONFIG_INTERLEAVE_ACTIVE) {
+		rc = -EBUSY;
+		goto out;
+	}
+
+	p->interleave_granularity = val;
+out:
+	up_write(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	return len;
+}
+static DEVICE_ATTR_RW(interleave_granularity);
+
 static struct attribute *cxl_region_attrs[] = {
 	&dev_attr_uuid.attr,
+	&dev_attr_interleave_ways.attr,
+	&dev_attr_interleave_granularity.attr,
 	NULL,
 };
 
@@ -216,6 +340,8 @@ static struct cxl_region *devm_cxl_add_region(struct cxl_root_decoder *cxlrd,
 					      enum cxl_decoder_type type)
 {
 	struct cxl_port *port = to_cxl_port(cxlrd->cxlsd.cxld.dev.parent);
+	struct cxl_decoder *cxld = &cxlrd->cxlsd.cxld;
+	struct cxl_region_params *p;
 	struct cxl_region *cxlr;
 	struct device *dev;
 	int rc;
@@ -223,8 +349,10 @@ static struct cxl_region *devm_cxl_add_region(struct cxl_root_decoder *cxlrd,
 	cxlr = cxl_region_alloc(cxlrd, id);
 	if (IS_ERR(cxlr))
 		return cxlr;
+	p = &cxlr->params;
 	cxlr->mode = mode;
 	cxlr->type = type;
+	p->interleave_granularity = cxld->interleave_granularity;
 
 	dev = &cxlr->dev;
 	rc = dev_set_name(dev, "region%d", id);
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index cf2ece363015..a4e65c102bed 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -7,6 +7,7 @@
 #include <linux/libnvdimm.h>
 #include <linux/bitfield.h>
 #include <linux/bitops.h>
+#include <linux/log2.h>
 #include <linux/io.h>
 
 /**
@@ -92,6 +93,31 @@ static inline int cxl_to_ways(u8 eniw, unsigned int *val)
 	return 0;
 }
 
+static inline int granularity_to_cxl(int g, u16 *ig)
+{
+	if (g > SZ_16K || g < 256 || !is_power_of_2(g))
+		return -EINVAL;
+	*ig = ilog2(g) - 8;
+	return 0;
+}
+
+static inline int ways_to_cxl(int ways, u8 *iw)
+{
+	if (ways > 16)
+		return -EINVAL;
+	if (is_power_of_2(ways)) {
+		*iw = ilog2(ways);
+		return 0;
+	}
+	if (ways % 3)
+		return -EINVAL;
+	ways /= 3;
+	if (!is_power_of_2(ways))
+		return -EINVAL;
+	*iw = ilog2(ways) + 8;
+	return 0;
+}
+
 /* CXL 2.0 8.2.8.1 Device Capabilities Array Register */
 #define CXLDEV_CAP_ARRAY_OFFSET 0x0
 #define   CXLDEV_CAP_ARRAY_CAP_ID 0
@@ -298,11 +324,14 @@ struct cxl_root_decoder {
 /*
  * enum cxl_config_state - State machine for region configuration
  * @CXL_CONFIG_IDLE: Any sysfs attribute can be written freely
+ * @CXL_CONFIG_INTERLEAVE_ACTIVE: region size has been set, no more
+ * changes to interleave_ways or interleave_granularity
  * @CXL_CONFIG_ACTIVE: All targets have been added the region is now
  * active
  */
 enum cxl_config_state {
 	CXL_CONFIG_IDLE,
+	CXL_CONFIG_INTERLEAVE_ACTIVE,
 	CXL_CONFIG_ACTIVE,
 };
 
@@ -310,12 +339,16 @@ enum cxl_config_state {
  * struct cxl_region_params - region settings
  * @state: allow the driver to lockdown further parameter changes
  * @uuid: unique id for persistent regions
+ * @interleave_ways: number of endpoints in the region
+ * @interleave_granularity: capacity each endpoint contributes to a stripe
  *
  * State transitions are protected by the cxl_region_rwsem
  */
 struct cxl_region_params {
 	enum cxl_config_state state;
 	uuid_t uuid;
+	int interleave_ways;
+	int interleave_granularity;
 };
 
 /**

[PATCH v2 20/28] cxl/region: Allocate HPA capacity to regions

[Dan Williams]

After a region's interleave parameters (ways and granularity) are set,
add a way for regions to allocate HPA (host physical address space) from
the free capacity in their parent root-decoder. The allocator for this
capacity reuses the 'struct resource' based allocator used for
CONFIG_DEVICE_PRIVATE.

Once the tuple of "ways, granularity, and size" is set the
region configuration transitions to the CXL_CONFIG_INTERLEAVE_ACTIVE
state which is a precursor to allowing endpoint decoders to be added to
a region.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   29 ++++++
 drivers/cxl/Kconfig                     |    3 +
 drivers/cxl/core/region.c               |  150 +++++++++++++++++++++++++++++++
 drivers/cxl/cxl.h                       |    2 
 4 files changed, 183 insertions(+), 1 deletion(-)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index bfa42bcc8383..0c6c3da4da5a 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -313,3 +313,32 @@ Description:
 		(RW) Configures the number of devices participating in the
 		region is set by writing this value. Each device will provide
 		1/interleave_ways of storage for the region.
+
+
+What:		/sys/bus/cxl/devices/regionZ/size
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) System physical address space to be consumed by the region.
+		When written trigger the driver to allocate space out of the
+		parent root decoder's address space. When read the size of the
+		address space is reported and should match the span of the
+		region's resource attribute. Size shall be set after the
+		interleave configuration parameters. Once set it cannot be
+		changed, only freed by writing 0. The kernel makes no guarantees
+		that data is maintained over an address space freeing event, and
+		there is no guarantee that a free followed by an allocate
+		results in the same address being allocated.
+
+
+What:		/sys/bus/cxl/devices/regionZ/resource
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RO) A region is a contiguous partition of a CXL root decoder
+		address space. Region capacity is allocated by writing to the
+		size attribute, the resulting physical address space determined
+		by the driver is reflected here. It is therefore not useful to
+		read this before writing a value to the size attribute.
diff --git a/drivers/cxl/Kconfig b/drivers/cxl/Kconfig
index aa2728de419e..74c2cd069d9d 100644
--- a/drivers/cxl/Kconfig
+++ b/drivers/cxl/Kconfig
@@ -105,6 +105,9 @@ config CXL_SUSPEND
 config CXL_REGION
 	bool
 	default CXL_BUS
+	# For MAX_PHYSMEM_BITS
+	depends on SPARSEMEM
 	select MEMREGION
+	select GET_FREE_REGION
 
 endif
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 3289caa5d882..b1e847827c6b 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -244,10 +244,152 @@ static ssize_t interleave_granularity_store(struct device *dev,
 }
 static DEVICE_ATTR_RW(interleave_granularity);
 
+static ssize_t resource_show(struct device *dev, struct device_attribute *attr,
+			     char *buf)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	u64 resource = -1ULL;
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	if (p->res)
+		resource = p->res->start;
+	rc = sysfs_emit(buf, "%#llx\n", resource);
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+static DEVICE_ATTR_RO(resource);
+
+static int alloc_hpa(struct cxl_region *cxlr, resource_size_t size)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(cxlr->dev.parent);
+	struct cxl_region_params *p = &cxlr->params;
+	struct resource *res;
+	u32 remainder = 0;
+
+	lockdep_assert_held_write(&cxl_region_rwsem);
+
+	/* Nothing to do... */
+	if (p->res && resource_size(res) == size)
+		return 0;
+
+	/* To change size the old size must be freed first */
+	if (p->res)
+		return -EBUSY;
+
+	if (p->state >= CXL_CONFIG_INTERLEAVE_ACTIVE)
+		return -EBUSY;
+
+	/* ways, granularity and uuid (if PMEM) need to be set before HPA */
+	if (!p->interleave_ways || !p->interleave_granularity ||
+	    (cxlr->mode == CXL_DECODER_PMEM && uuid_is_null(&p->uuid)))
+		return -ENXIO;
+
+	div_u64_rem(size, SZ_256M * p->interleave_ways, &remainder);
+	if (remainder)
+		return -EINVAL;
+
+	res = alloc_free_mem_region(cxlrd->res, size, SZ_256M,
+				    dev_name(&cxlr->dev));
+	if (IS_ERR(res)) {
+		dev_dbg(&cxlr->dev, "failed to allocate HPA: %ld\n",
+			PTR_ERR(res));
+		return PTR_ERR(res);
+	}
+
+	p->res = res;
+	p->state = CXL_CONFIG_INTERLEAVE_ACTIVE;
+
+	return 0;
+}
+
+static void cxl_region_iomem_release(struct cxl_region *cxlr)
+{
+	struct cxl_region_params *p = &cxlr->params;
+
+	if (device_is_registered(&cxlr->dev))
+		lockdep_assert_held_write(&cxl_region_rwsem);
+	if (p->res) {
+		remove_resource(p->res);
+		kfree(p->res);
+		p->res = NULL;
+	}
+}
+
+static int free_hpa(struct cxl_region *cxlr)
+{
+	struct cxl_region_params *p = &cxlr->params;
+
+	lockdep_assert_held_write(&cxl_region_rwsem);
+
+	if (!p->res)
+		return 0;
+
+	if (p->state >= CXL_CONFIG_ACTIVE)
+		return -EBUSY;
+
+	cxl_region_iomem_release(cxlr);
+	p->state = CXL_CONFIG_IDLE;
+	return 0;
+}
+
+static ssize_t size_store(struct device *dev, struct device_attribute *attr,
+			  const char *buf, size_t len)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	u64 val;
+	int rc;
+
+	rc = kstrtou64(buf, 0, &val);
+	if (rc)
+		return rc;
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+
+	if (val)
+		rc = alloc_hpa(cxlr, val);
+	else
+		rc = free_hpa(cxlr);
+	up_write(&cxl_region_rwsem);
+
+	if (rc)
+		return rc;
+
+	return len;
+}
+
+static ssize_t size_show(struct device *dev, struct device_attribute *attr,
+			 char *buf)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	u64 size = 0;
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	if (p->res)
+		size = resource_size(p->res);
+	rc = sysfs_emit(buf, "%#llx\n", size);
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+static DEVICE_ATTR_RW(size);
+
 static struct attribute *cxl_region_attrs[] = {
 	&dev_attr_uuid.attr,
 	&dev_attr_interleave_ways.attr,
 	&dev_attr_interleave_granularity.attr,
+	&dev_attr_resource.attr,
+	&dev_attr_size.attr,
 	NULL,
 };
 
@@ -293,7 +435,11 @@ static struct cxl_region *to_cxl_region(struct device *dev)
 
 static void unregister_region(void *dev)
 {
-	device_unregister(dev);
+	struct cxl_region *cxlr = to_cxl_region(dev);
+
+	device_del(dev);
+	cxl_region_iomem_release(cxlr);
+	put_device(dev);
 }
 
 static struct lock_class_key cxl_region_key;
@@ -445,3 +591,5 @@ static ssize_t delete_region_store(struct device *dev,
 	return len;
 }
 DEVICE_ATTR_WO(delete_region);
+
+MODULE_IMPORT_NS(CXL);
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index a4e65c102bed..837bfa67f469 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -341,6 +341,7 @@ enum cxl_config_state {
  * @uuid: unique id for persistent regions
  * @interleave_ways: number of endpoints in the region
  * @interleave_granularity: capacity each endpoint contributes to a stripe
+ * @res: allocated iomem capacity for this region
  *
  * State transitions are protected by the cxl_region_rwsem
  */
@@ -349,6 +350,7 @@ struct cxl_region_params {
 	uuid_t uuid;
 	int interleave_ways;
 	int interleave_granularity;
+	struct resource *res;
 };
 
 /**

[PATCH v2 21/28] cxl/region: Enable the assignment of endpoint decoders to regions

[Dan Williams]

The region provisioning process involves allocating DPA to a set of
endpoint decoders, and HPA plus the region geometry to a region device.
Then the decoder is assigned to the region. At this point several
validation steps can be performed to validate that the decoder is
suitable to participate in the region.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   19 ++
 drivers/cxl/core/core.h                 |    6 +
 drivers/cxl/core/hdm.c                  |   13 +
 drivers/cxl/core/port.c                 |    9 +
 drivers/cxl/core/region.c               |  282 +++++++++++++++++++++++++++++++
 drivers/cxl/cxl.h                       |   11 +
 6 files changed, 338 insertions(+), 2 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 0c6c3da4da5a..94e19e24de8d 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -342,3 +342,22 @@ Description:
 		size attribute, the resulting physical address space determined
 		by the driver is reflected here. It is therefore not useful to
 		read this before writing a value to the size attribute.
+
+
+What:		/sys/bus/cxl/devices/regionZ/target[0..N]
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) Write an endpoint decoder object name to 'targetX' where X
+		is the intended position of the endpoint device in the region
+		interleave and N is the 'interleave_ways' setting for the
+		region. ENXIO is returned if the write results in an impossible
+		to map decode scenario, like the endpoint is unreachable at that
+		position relative to the root decoder interleave. EBUSY is
+		returned if the position in the region is already occupied, or
+		if the region is not in a state to accept interleave
+		configuration changes. EINVAL is returned if the object name is
+		not an endpoint decoder. Once all positions have been
+		successfully written a final validation for decode conflicts is
+		performed before activating the region.
diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index 29272df7e212..a60ad9f656fd 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -12,9 +12,14 @@ extern struct attribute_group cxl_base_attribute_group;
 #ifdef CONFIG_CXL_REGION
 extern struct device_attribute dev_attr_create_pmem_region;
 extern struct device_attribute dev_attr_delete_region;
+extern struct device_attribute dev_attr_region;
+void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled);
 #define CXL_REGION_ATTR(x) (&dev_attr_##x.attr)
 #define SET_CXL_REGION_ATTR(x) (&dev_attr_##x.attr),
 #else
+static inline void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled)
+{
+}
 #define CXL_REGION_ATTR(x) NULL
 #define SET_CXL_REGION_ATTR(x)
 #endif
@@ -34,6 +39,7 @@ int cxl_dpa_alloc(struct cxl_endpoint_decoder *cxled, unsigned long long size);
 int cxl_dpa_free(struct cxl_endpoint_decoder *cxled);
 resource_size_t cxl_dpa_size(struct cxl_endpoint_decoder *cxled);
 resource_size_t cxl_dpa_resource_start(struct cxl_endpoint_decoder *cxled);
+extern struct rw_semaphore cxl_dpa_rwsem;
 
 int cxl_memdev_init(void);
 void cxl_memdev_exit(void);
diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 4a0325b02ca4..81645de1064f 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -17,7 +17,7 @@
  * for enumerating these registers and capabilities.
  */
 
-static DECLARE_RWSEM(cxl_dpa_rwsem);
+DECLARE_RWSEM(cxl_dpa_rwsem);
 
 static int add_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 			   int *target_map)
@@ -319,6 +319,11 @@ int cxl_dpa_free(struct cxl_endpoint_decoder *cxled)
 		rc = 0;
 		goto out;
 	}
+	if (cxled->cxld.region) {
+		dev_dbg(dev, "decoder assigned to: %s\n",
+			dev_name(&cxled->cxld.region->dev));
+		goto out;
+	}
 	if (cxled->cxld.flags & CXL_DECODER_F_ENABLE) {
 		dev_dbg(dev, "decoder enabled\n");
 		rc = -EBUSY;
@@ -395,6 +400,12 @@ int cxl_dpa_alloc(struct cxl_endpoint_decoder *cxled, unsigned long long size)
 	int rc;
 
 	down_write(&cxl_dpa_rwsem);
+	if (cxled->cxld.region) {
+		dev_dbg(dev, "decoder attached to %s\n",
+			dev_name(&cxled->cxld.region->dev));
+		goto out;
+	}
+
 	if (cxled->cxld.flags & CXL_DECODER_F_ENABLE) {
 		dev_dbg(dev, "decoder enabled\n");
 		rc = -EBUSY;
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 89672b126b30..bd0673821d28 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -288,6 +288,7 @@ static struct attribute *cxl_decoder_base_attrs[] = {
 	&dev_attr_locked.attr,
 	&dev_attr_interleave_granularity.attr,
 	&dev_attr_interleave_ways.attr,
+	SET_CXL_REGION_ATTR(region)
 	NULL,
 };
 
@@ -1583,6 +1584,7 @@ struct cxl_endpoint_decoder *cxl_endpoint_decoder_alloc(struct cxl_port *port)
 	if (!cxled)
 		return ERR_PTR(-ENOMEM);
 
+	cxled->pos = -1;
 	cxld = &cxled->cxld;
 	rc = cxl_decoder_init(port, cxld);
 	if (rc)	 {
@@ -1687,6 +1689,13 @@ EXPORT_SYMBOL_NS_GPL(cxl_decoder_add, CXL);
 
 static void cxld_unregister(void *dev)
 {
+	struct cxl_endpoint_decoder *cxled;
+
+	if (is_endpoint_decoder(dev)) {
+		cxled = to_cxl_endpoint_decoder(dev);
+		cxl_decoder_kill_region(cxled);
+	}
+
 	device_unregister(dev);
 }
 
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index b1e847827c6b..871bfdbb9bc8 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -24,6 +24,7 @@
  * but is only visible for persistent regions.
  * 1. Interleave granularity
  * 2. Interleave size
+ * 3. Decoder targets
  */
 
 /*
@@ -141,6 +142,8 @@ static ssize_t interleave_ways_show(struct device *dev,
 	return rc;
 }
 
+static const struct attribute_group *get_cxl_region_target_group(void);
+
 static ssize_t interleave_ways_store(struct device *dev,
 				     struct device_attribute *attr,
 				     const char *buf, size_t len)
@@ -149,7 +152,7 @@ static ssize_t interleave_ways_store(struct device *dev,
 	struct cxl_decoder *cxld = &cxlrd->cxlsd.cxld;
 	struct cxl_region *cxlr = to_cxl_region(dev);
 	struct cxl_region_params *p = &cxlr->params;
-	int rc, val;
+	int rc, val, save;
 	u8 iw;
 
 	rc = kstrtoint(buf, 0, &val);
@@ -178,7 +181,11 @@ static ssize_t interleave_ways_store(struct device *dev,
 		goto out;
 	}
 
+	save = p->interleave_ways;
 	p->interleave_ways = val;
+	rc = sysfs_update_group(&cxlr->dev.kobj, get_cxl_region_target_group());
+	if (rc)
+		p->interleave_ways = save;
 out:
 	up_write(&cxl_region_rwsem);
 	if (rc)
@@ -398,9 +405,262 @@ static const struct attribute_group cxl_region_group = {
 	.is_visible = cxl_region_visible,
 };
 
+static size_t show_targetN(struct cxl_region *cxlr, char *buf, int pos)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_endpoint_decoder *cxled;
+	int rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+
+	if (pos >= p->interleave_ways) {
+		dev_dbg(&cxlr->dev, "position %d out of range %d\n", pos,
+			p->interleave_ways);
+		rc = -ENXIO;
+		goto out;
+	}
+
+	cxled = p->targets[pos];
+	if (!cxled)
+		rc = sysfs_emit(buf, "\n");
+	else
+		rc = sysfs_emit(buf, "%s\n", dev_name(&cxled->cxld.dev));
+out:
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+
+/*
+ * - Check that the given endpoint is attached to a host-bridge identified
+ *   in the root interleave.
+ */
+static int cxl_region_attach(struct cxl_region *cxlr,
+			     struct cxl_endpoint_decoder *cxled, int pos)
+{
+	struct cxl_region_params *p = &cxlr->params;
+
+	if (cxled->mode == CXL_DECODER_DEAD) {
+		dev_dbg(&cxlr->dev, "%s dead\n", dev_name(&cxled->cxld.dev));
+		return -ENODEV;
+	}
+
+	if (pos >= p->interleave_ways) {
+		dev_dbg(&cxlr->dev, "position %d out of range %d\n", pos,
+			p->interleave_ways);
+		return -ENXIO;
+	}
+
+	if (p->targets[pos] == cxled)
+		return 0;
+
+	if (p->targets[pos]) {
+		struct cxl_endpoint_decoder *cxled_target = p->targets[pos];
+		struct cxl_memdev *cxlmd_target = cxled_to_memdev(cxled_target);
+
+		dev_dbg(&cxlr->dev, "position %d already assigned to %s:%s\n",
+			pos, dev_name(&cxlmd_target->dev),
+			dev_name(&cxled_target->cxld.dev));
+		return -EBUSY;
+	}
+
+	p->targets[pos] = cxled;
+	cxled->pos = pos;
+	p->nr_targets++;
+
+	return 0;
+}
+
+static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_region *cxlr = cxled->cxld.region;
+	struct cxl_region_params *p;
+
+	lockdep_assert_held_write(&cxl_region_rwsem);
+
+	if (!cxlr)
+		return;
+
+	p = &cxlr->params;
+	get_device(&cxlr->dev);
+
+	if (cxled->pos < 0 || cxled->pos >= p->interleave_ways ||
+	    p->targets[cxled->pos] != cxled) {
+		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+
+		dev_WARN_ONCE(&cxlr->dev, 1, "expected %s:%s at position %d\n",
+			      dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev),
+			      cxled->pos);
+		goto out;
+	}
+
+	p->targets[cxled->pos] = NULL;
+	p->nr_targets--;
+
+	/* notify the region driver that one of its targets has deparated */
+	up_write(&cxl_region_rwsem);
+	device_release_driver(&cxlr->dev);
+	down_write(&cxl_region_rwsem);
+out:
+	put_device(&cxlr->dev);
+}
+
+void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled)
+{
+	down_write(&cxl_region_rwsem);
+	cxled->mode = CXL_DECODER_DEAD;
+	cxl_region_detach(cxled);
+	up_write(&cxl_region_rwsem);
+}
+
+static int attach_target(struct cxl_region *cxlr, const char *decoder, int pos)
+{
+	struct device *dev;
+	int rc;
+
+	dev = bus_find_device_by_name(&cxl_bus_type, NULL, decoder);
+	if (!dev)
+		return -ENODEV;
+
+	if (!is_endpoint_decoder(dev)) {
+		put_device(dev);
+		return -EINVAL;
+	}
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		goto out;
+	down_read(&cxl_dpa_rwsem);
+	rc = cxl_region_attach(cxlr, to_cxl_endpoint_decoder(dev), pos);
+	up_read(&cxl_dpa_rwsem);
+	up_write(&cxl_region_rwsem);
+out:
+	put_device(dev);
+	return rc;
+}
+
+static int detach_target(struct cxl_region *cxlr, int pos)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	int rc;
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+
+	if (pos >= p->interleave_ways) {
+		dev_dbg(&cxlr->dev, "position %d out of range %d\n", pos,
+			p->interleave_ways);
+		rc = -ENXIO;
+		goto out;
+	}
+
+	if (!p->targets[pos]) {
+		rc = 0;
+		goto out;
+	}
+
+	cxl_region_detach(p->targets[pos]);
+	rc = 0;
+out:
+	up_write(&cxl_region_rwsem);
+	return rc;
+}
+
+static size_t store_targetN(struct cxl_region *cxlr, const char *buf, int pos,
+			    size_t len)
+{
+	int rc;
+
+	if (sysfs_streq(buf, "\n"))
+		rc = detach_target(cxlr, pos);
+	else
+		rc = attach_target(cxlr, buf, pos);
+
+	if (rc < 0)
+		return rc;
+	return len;
+}
+
+#define TARGET_ATTR_RW(n)                                              \
+static ssize_t target##n##_show(                                       \
+	struct device *dev, struct device_attribute *attr, char *buf)  \
+{                                                                      \
+	return show_targetN(to_cxl_region(dev), buf, (n));             \
+}                                                                      \
+static ssize_t target##n##_store(struct device *dev,                   \
+				 struct device_attribute *attr,        \
+				 const char *buf, size_t len)          \
+{                                                                      \
+	return store_targetN(to_cxl_region(dev), buf, (n), len);       \
+}                                                                      \
+static DEVICE_ATTR_RW(target##n)
+
+TARGET_ATTR_RW(0);
+TARGET_ATTR_RW(1);
+TARGET_ATTR_RW(2);
+TARGET_ATTR_RW(3);
+TARGET_ATTR_RW(4);
+TARGET_ATTR_RW(5);
+TARGET_ATTR_RW(6);
+TARGET_ATTR_RW(7);
+TARGET_ATTR_RW(8);
+TARGET_ATTR_RW(9);
+TARGET_ATTR_RW(10);
+TARGET_ATTR_RW(11);
+TARGET_ATTR_RW(12);
+TARGET_ATTR_RW(13);
+TARGET_ATTR_RW(14);
+TARGET_ATTR_RW(15);
+
+static struct attribute *target_attrs[] = {
+	&dev_attr_target0.attr,
+	&dev_attr_target1.attr,
+	&dev_attr_target2.attr,
+	&dev_attr_target3.attr,
+	&dev_attr_target4.attr,
+	&dev_attr_target5.attr,
+	&dev_attr_target6.attr,
+	&dev_attr_target7.attr,
+	&dev_attr_target8.attr,
+	&dev_attr_target9.attr,
+	&dev_attr_target10.attr,
+	&dev_attr_target11.attr,
+	&dev_attr_target12.attr,
+	&dev_attr_target13.attr,
+	&dev_attr_target14.attr,
+	&dev_attr_target15.attr,
+	NULL,
+};
+
+static umode_t cxl_region_target_visible(struct kobject *kobj,
+					 struct attribute *a, int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+
+	if (n < p->interleave_ways)
+		return a->mode;
+	return 0;
+}
+
+static const struct attribute_group cxl_region_target_group = {
+	.attrs = target_attrs,
+	.is_visible = cxl_region_target_visible,
+};
+
+static const struct attribute_group *get_cxl_region_target_group(void)
+{
+	return &cxl_region_target_group;
+}
+
 static const struct attribute_group *region_groups[] = {
 	&cxl_base_attribute_group,
 	&cxl_region_group,
+	&cxl_region_target_group,
 	NULL,
 };
 
@@ -560,6 +820,26 @@ static ssize_t create_pmem_region_store(struct device *dev,
 }
 DEVICE_ATTR_RW(create_pmem_region);
 
+static ssize_t region_show(struct device *dev, struct device_attribute *attr,
+			   char *buf)
+{
+	struct cxl_decoder *cxld = to_cxl_decoder(dev);
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+
+	if (cxld->region)
+		rc = sysfs_emit(buf, "%s\n", dev_name(&cxld->region->dev));
+	else
+		rc = sysfs_emit(buf, "\n");
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+DEVICE_ATTR_RO(region);
+
 static struct cxl_region *
 cxl_find_region_by_name(struct cxl_root_decoder *cxlrd, const char *name)
 {
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 837bfa67f469..95d74cf425a4 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -255,6 +255,7 @@ enum cxl_decoder_type {
  * @interleave_ways: number of cxl_dports in this decode
  * @interleave_granularity: data stride per dport
  * @target_type: accelerator vs expander (type2 vs type3) selector
+ * @region: currently assigned region for this decoder
  * @flags: memory type capabilities and locking
  */
 struct cxl_decoder {
@@ -264,14 +265,20 @@ struct cxl_decoder {
 	int interleave_ways;
 	int interleave_granularity;
 	enum cxl_decoder_type target_type;
+	struct cxl_region *region;
 	unsigned long flags;
 };
 
+/*
+ * CXL_DECODER_DEAD prevents endpoints from being reattached to regions
+ * while cxld_unregister() is running
+ */
 enum cxl_decoder_mode {
 	CXL_DECODER_NONE,
 	CXL_DECODER_RAM,
 	CXL_DECODER_PMEM,
 	CXL_DECODER_MIXED,
+	CXL_DECODER_DEAD,
 };
 
 /**
@@ -280,12 +287,14 @@ enum cxl_decoder_mode {
  * @dpa_res: actively claimed DPA span of this decoder
  * @skip: offset into @dpa_res where @cxld.hpa_range maps
  * @mode: which memory type / access-mode-partition this decoder targets
+ * @pos: interleave position in @cxld.region
  */
 struct cxl_endpoint_decoder {
 	struct cxl_decoder cxld;
 	struct resource *dpa_res;
 	resource_size_t skip;
 	enum cxl_decoder_mode mode;
+	int pos;
 };
 
 /**
@@ -351,6 +360,8 @@ struct cxl_region_params {
 	int interleave_ways;
 	int interleave_granularity;
 	struct resource *res;
+	struct cxl_endpoint_decoder *targets[CXL_DECODER_MAX_INTERLEAVE];
+	int nr_targets;
 };
 
 /**

[PATCH v2 22/28] cxl/acpi: Add a host-bridge index lookup mechanism

[Dan Williams]

The ACPI CXL Fixed Memory Window Structure (CFMWS) defines multiple
methods to determine which host bridge provides access to a given
endpoint relative to that device's position in the interleave. The
"Interleave Arithmetic" defines either a "standard modulo" /
round-random algorithm, or "xormap" based algorithm which can be defined
as a non-linear transform. Given that there are already more options
beyond "standard modulo" and that "xormap" may turn out to be ACPI CXL
specific, provide a callback for the region provisioning code to map
endpoint positions back to expected host bridge id (cxl_dport target).

For now just support the simple modulo math case and save the xormap for
a follow-on change.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-14-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/port.c |   16 ++++++++++++++++
 drivers/cxl/cxl.h       |    2 ++
 2 files changed, 18 insertions(+)

diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index bd0673821d28..2f0b47db53da 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -1421,6 +1421,20 @@ static int decoder_populate_targets(struct cxl_switch_decoder *cxlsd,
 	return rc;
 }
 
+static struct cxl_dport *cxl_hb_modulo(struct cxl_root_decoder *cxlrd, int pos)
+{
+	struct cxl_switch_decoder *cxlsd = &cxlrd->cxlsd;
+	struct cxl_decoder *cxld = &cxlsd->cxld;
+	int iw;
+
+	iw = cxld->interleave_ways;
+	if (dev_WARN_ONCE(&cxld->dev, iw != cxlsd->nr_targets,
+			  "misconfigured root decoder\n"))
+		return NULL;
+
+	return cxlrd->cxlsd.target[pos % iw];
+}
+
 static struct lock_class_key cxl_decoder_key;
 
 /**
@@ -1510,6 +1524,8 @@ struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
 		return ERR_PTR(rc);
 	}
 
+	cxlrd->calc_hb = cxl_hb_modulo;
+
 	cxld = &cxlsd->cxld;
 	cxld->dev.type = &cxl_decoder_root_type;
 	/*
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 95d74cf425a4..cd81e642e900 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -322,11 +322,13 @@ struct cxl_switch_decoder {
  * struct cxl_root_decoder - Static platform CXL address decoder
  * @res: host / parent resource for region allocations
  * @region_id: region id for next region provisioning event
+ * @calc_hb: which host bridge covers the n'th position by granularity
  * @cxlsd: base cxl switch decoder
  */
 struct cxl_root_decoder {
 	struct resource *res;
 	atomic_t region_id;
+	struct cxl_dport *(*calc_hb)(struct cxl_root_decoder *cxlrd, int pos);
 	struct cxl_switch_decoder cxlsd;
 };
 

[PATCH v2 23/28] cxl/region: Attach endpoint decoders

[Dan Williams]

CXL regions (interleave sets) are made up of a set of memory devices
where each device maps a portion of the interleave with one of its
decoders (see CXL 2.0 8.2.5.12 CXL HDM Decoder Capability Structure).
As endpoint decoders are identified by a provisioning tool they can be
added to a region provided the region interleave properties are set
(way, granularity, HPA) and DPA has been assigned to the decoder.

The attach event triggers several validation checks, for example:
- is the DPA sized appropriately for the region
- is the decoder reachable via the host-bridges identified by the
  region's root decoder
- is the device already active in a different region position slot
- are there already regions with a higher HPA active on a given port
  (per CXL 2.0 8.2.5.12.20 Committing Decoder Programming)

...and the attach event affords an opportunity to collect data and
resources relevant to later programming the target lists in switch
decoders, for example:
- allocate a decoder at each cxl_port in the decode chain
- for a given switch port, how many the region's endpoints are hosted
  through the port
- how many unique targets (next hops) does a port need to map to reach
  those endpoints

The act of reconciling this information and deploying it to the decoder
configuration is saved for a follow-on patch.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/core.h   |    7 +
 drivers/cxl/core/port.c   |   10 -
 drivers/cxl/core/region.c |  364 ++++++++++++++++++++++++++++++++++++++++++++-
 drivers/cxl/cxl.h         |   20 ++
 drivers/cxl/cxlmem.h      |    5 +
 5 files changed, 394 insertions(+), 12 deletions(-)

diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index a60ad9f656fd..6edd8174c2b5 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -41,6 +41,13 @@ resource_size_t cxl_dpa_size(struct cxl_endpoint_decoder *cxled);
 resource_size_t cxl_dpa_resource_start(struct cxl_endpoint_decoder *cxled);
 extern struct rw_semaphore cxl_dpa_rwsem;
 
+bool is_switch_decoder(struct device *dev);
+static inline struct cxl_ep *cxl_ep_load(struct cxl_port *port,
+					 struct cxl_memdev *cxlmd)
+{
+	return xa_load(&port->endpoints, (unsigned long)&cxlmd->dev);
+}
+
 int cxl_memdev_init(void);
 void cxl_memdev_exit(void);
 void cxl_mbox_init(void);
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 2f0b47db53da..d234afc47e89 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -448,7 +448,7 @@ bool is_root_decoder(struct device *dev)
 }
 EXPORT_SYMBOL_NS_GPL(is_root_decoder, CXL);
 
-static bool is_switch_decoder(struct device *dev)
+bool is_switch_decoder(struct device *dev)
 {
 	return is_root_decoder(dev) || dev->type == &cxl_decoder_switch_type;
 }
@@ -504,6 +504,7 @@ static void cxl_port_release(struct device *dev)
 		cxl_ep_remove(port, ep);
 	xa_destroy(&port->endpoints);
 	xa_destroy(&port->dports);
+	xa_destroy(&port->regions);
 	ida_free(&cxl_port_ida, port->id);
 	kfree(port);
 }
@@ -635,6 +636,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
 	port->hdm_end = -1;
 	xa_init(&port->dports);
 	xa_init(&port->endpoints);
+	xa_init(&port->regions);
 
 	device_initialize(dev);
 	lockdep_set_class_and_subclass(&dev->mutex, &cxl_port_key, port->depth);
@@ -1109,12 +1111,6 @@ static void reap_dports(struct cxl_port *port)
 	}
 }
 
-static struct cxl_ep *cxl_ep_load(struct cxl_port *port,
-				  struct cxl_memdev *cxlmd)
-{
-	return xa_load(&port->endpoints, (unsigned long)&cxlmd->dev);
-}
-
 int devm_cxl_add_endpoint(struct cxl_memdev *cxlmd,
 			  struct cxl_dport *parent_dport)
 {
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 871bfdbb9bc8..8ac0c557f6aa 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -433,21 +433,294 @@ static size_t show_targetN(struct cxl_region *cxlr, char *buf, int pos)
 	return rc;
 }
 
-/*
- * - Check that the given endpoint is attached to a host-bridge identified
- *   in the root interleave.
+static int match_free_decoder(struct device *dev, void *data)
+{
+	struct cxl_decoder *cxld;
+	int *id = data;
+
+	if (!is_switch_decoder(dev))
+		return 0;
+
+	cxld = to_cxl_decoder(dev);
+
+	/* enforce ordered allocation */
+	if (cxld->id != *id)
+		return 0;
+
+	if (!cxld->region)
+		return 1;
+
+	(*id)++;
+
+	return 0;
+}
+
+static struct cxl_decoder *cxl_region_find_decoder(struct cxl_port *port,
+						   struct cxl_region *cxlr)
+{
+	struct device *dev;
+	int id = 0;
+
+	dev = device_find_child(&port->dev, &id, match_free_decoder);
+	if (!dev)
+		return NULL;
+	/*
+	 * This decoder is pinned registered as long as the endpoint decoder is
+	 * registered, and endpoint decoder unregistration holds the
+	 * cxl_region_rwsem over unregister events, so no need to hold on to
+	 * this extra reference.
+	 */
+	put_device(dev);
+	return to_cxl_decoder(dev);
+}
+
+static struct cxl_region_ref *alloc_region_ref(struct cxl_port *port,
+					       struct cxl_region *cxlr)
+{
+	struct cxl_region_ref *cxl_rr;
+	int rc;
+
+	cxl_rr = kzalloc(sizeof(*cxl_rr), GFP_KERNEL);
+	if (!cxl_rr)
+		return NULL;
+	cxl_rr->port = port;
+	cxl_rr->region = cxlr;
+	xa_init(&cxl_rr->endpoints);
+
+	rc = xa_insert(&port->regions, (unsigned long)cxlr, cxl_rr, GFP_KERNEL);
+	if (rc) {
+		dev_dbg(&cxlr->dev,
+			"%s: failed to track region reference: %d\n",
+			dev_name(&port->dev), rc);
+		kfree(cxl_rr);
+		return NULL;
+	}
+
+	return cxl_rr;
+}
+
+static void free_region_ref(struct cxl_region_ref *cxl_rr)
+{
+	struct cxl_port *port = cxl_rr->port;
+	struct cxl_region *cxlr = cxl_rr->region;
+	struct cxl_decoder *cxld = cxl_rr->decoder;
+
+	dev_WARN_ONCE(&cxlr->dev, cxld->region != cxlr, "region mismatch\n");
+	if (cxld->region == cxlr) {
+		cxld->region = NULL;
+		put_device(&cxlr->dev);
+	}
+
+	xa_erase(&port->regions, (unsigned long)cxlr);
+	xa_destroy(&cxl_rr->endpoints);
+	kfree(cxl_rr);
+}
+
+static int cxl_rr_ep_add(struct cxl_region_ref *cxl_rr,
+			 struct cxl_endpoint_decoder *cxled)
+{
+	int rc;
+	struct cxl_port *port = cxl_rr->port;
+	struct cxl_region *cxlr = cxl_rr->region;
+	struct cxl_decoder *cxld = cxl_rr->decoder;
+	struct cxl_ep *ep = cxl_ep_load(port, cxled_to_memdev(cxled));
+
+	rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
+			 GFP_KERNEL);
+	if (rc)
+		return rc;
+	cxl_rr->nr_eps++;
+
+	if (!cxld->region) {
+		cxld->region = cxlr;
+		get_device(&cxlr->dev);
+	}
+
+	return 0;
+}
+
+/**
+ * cxl_port_attach_region() - track a region's interest in a port by endpoint
+ * @port: port to add a new region reference 'struct cxl_region_ref'
+ * @cxlr: region to attach to @port
+ * @cxled: endpoint decoder used to create or further pin a region reference
+ * @pos: interleave position of @cxled in @cxlr
+ *
+ * The attach event is an opportunity to validate CXL decode setup
+ * constraints and record metadata needed for programming HDM decoders,
+ * in particular decoder target lists.
+ *
+ * The steps are:
+ * - validate that there are no other regions with a higher HPA already
+ *   associated with @port
+ * - establish a region reference if one is not already present
+ *   - additionally allocate a decoder instance that will host @cxlr on
+ *     @port
+ * - pin the region reference by the endpoint
+ * - account for how many entries in @port's target list are needed to
+ *   cover all of the added endpoints.
  */
+static int cxl_port_attach_region(struct cxl_port *port,
+				  struct cxl_region *cxlr,
+				  struct cxl_endpoint_decoder *cxled, int pos)
+{
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
+	struct cxl_region_ref *cxl_rr = NULL, *iter;
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_decoder *cxld = NULL;
+	unsigned long index;
+	int rc = -EBUSY;
+
+	lockdep_assert_held_write(&cxl_region_rwsem);
+
+	xa_for_each(&port->regions, index, iter) {
+		struct cxl_region_params *ip = &iter->region->params;
+
+		if (iter->region == cxlr)
+			cxl_rr = iter;
+		if (ip->res->start > p->res->start) {
+			dev_dbg(&cxlr->dev,
+				"%s: HPA order violation %s:%pr vs %pr\n",
+				dev_name(&port->dev),
+				dev_name(&iter->region->dev), ip->res, p->res);
+			return -EBUSY;
+		}
+	}
+
+	if (cxl_rr) {
+		struct cxl_ep *ep_iter;
+		int found = 0;
+
+		cxld = cxl_rr->decoder;
+		xa_for_each(&cxl_rr->endpoints, index, ep_iter) {
+			if (ep_iter == ep)
+				continue;
+			if (ep_iter->next == ep->next) {
+				found++;
+				break;
+			}
+		}
+
+		/*
+		 * If this is a new target or if this port is direct connected
+		 * to this endpoint then add to the target count.
+		 */
+		if (!found || !ep->next)
+			cxl_rr->nr_targets++;
+	} else {
+		cxl_rr = alloc_region_ref(port, cxlr);
+		if (!cxl_rr) {
+			dev_dbg(&cxlr->dev,
+				"%s: failed to allocate region reference\n",
+				dev_name(&port->dev));
+			return -ENOMEM;
+		}
+	}
+
+	if (!cxld) {
+		if (port == cxled_to_port(cxled))
+			cxld = &cxled->cxld;
+		else
+			cxld = cxl_region_find_decoder(port, cxlr);
+		if (!cxld) {
+			dev_dbg(&cxlr->dev, "%s: no decoder available\n",
+				dev_name(&port->dev));
+			goto out_erase;
+		}
+
+		if (cxld->region) {
+			dev_dbg(&cxlr->dev, "%s: %s already attached to %s\n",
+				dev_name(&port->dev), dev_name(&cxld->dev),
+				dev_name(&cxld->region->dev));
+			rc = -EBUSY;
+			goto out_erase;
+		}
+
+		cxl_rr->decoder = cxld;
+	}
+
+	rc = cxl_rr_ep_add(cxl_rr, cxled);
+	if (rc) {
+		dev_dbg(&cxlr->dev,
+			"%s: failed to track endpoint %s:%s reference\n",
+			dev_name(&port->dev), dev_name(&cxlmd->dev),
+			dev_name(&cxld->dev));
+		goto out_erase;
+	}
+
+	return 0;
+out_erase:
+	if (cxl_rr->nr_eps == 0)
+		free_region_ref(cxl_rr);
+	return rc;
+}
+
+static struct cxl_region_ref *cxl_rr_load(struct cxl_port *port,
+					  struct cxl_region *cxlr)
+{
+	return xa_load(&port->regions, (unsigned long)cxlr);
+}
+
+static void cxl_port_detach_region(struct cxl_port *port,
+				   struct cxl_region *cxlr,
+				   struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_region_ref *cxl_rr;
+	struct cxl_ep *ep;
+
+	lockdep_assert_held_write(&cxl_region_rwsem);
+
+	cxl_rr = cxl_rr_load(port, cxlr);
+	if (!cxl_rr)
+		return;
+
+	ep = xa_erase(&cxl_rr->endpoints, (unsigned long)cxled);
+	if (ep) {
+		struct cxl_ep *ep_iter;
+		unsigned long index;
+		int found = 0;
+
+		cxl_rr->nr_eps--;
+		xa_for_each(&cxl_rr->endpoints, index, ep_iter) {
+			if (ep_iter->next == ep->next) {
+				found++;
+				break;
+			}
+		}
+		if (!found)
+			cxl_rr->nr_targets--;
+	}
+
+	if (cxl_rr->nr_eps == 0)
+		free_region_ref(cxl_rr);
+}
+
 static int cxl_region_attach(struct cxl_region *cxlr,
 			     struct cxl_endpoint_decoder *cxled, int pos)
 {
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(cxlr->dev.parent);
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_port *ep_port, *root_port, *iter;
 	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_dport *dport;
+	int i, rc = -ENXIO;
 
 	if (cxled->mode == CXL_DECODER_DEAD) {
 		dev_dbg(&cxlr->dev, "%s dead\n", dev_name(&cxled->cxld.dev));
 		return -ENODEV;
 	}
 
-	if (pos >= p->interleave_ways) {
+	/* all full of members, or interleave config not established? */
+	if (p->state > CXL_CONFIG_INTERLEAVE_ACTIVE) {
+		dev_dbg(&cxlr->dev, "region already active\n");
+		return -EBUSY;
+	} else if (p->state < CXL_CONFIG_INTERLEAVE_ACTIVE) {
+		dev_dbg(&cxlr->dev, "interleave config missing\n");
+		return -ENXIO;
+	}
+
+	if (pos < 0 || pos >= p->interleave_ways) {
 		dev_dbg(&cxlr->dev, "position %d out of range %d\n", pos,
 			p->interleave_ways);
 		return -ENXIO;
@@ -466,15 +739,90 @@ static int cxl_region_attach(struct cxl_region *cxlr,
 		return -EBUSY;
 	}
 
+	for (i = 0; i < p->interleave_ways; i++) {
+		struct cxl_endpoint_decoder *cxled_target;
+		struct cxl_memdev *cxlmd_target;
+
+		cxled_target = p->targets[pos];
+		if (!cxled_target)
+			continue;
+
+		cxlmd_target = cxled_to_memdev(cxled_target);
+		if (cxlmd_target == cxlmd) {
+			dev_dbg(&cxlr->dev,
+				"%s already specified at position %d via: %s\n",
+				dev_name(&cxlmd->dev), pos,
+				dev_name(&cxled_target->cxld.dev));
+			return -EBUSY;
+		}
+	}
+
+	ep_port = cxled_to_port(cxled);
+	root_port = cxlrd_to_port(cxlrd);
+	dport = cxl_find_dport_by_dev(root_port, ep_port->host_bridge);
+	if (!dport) {
+		dev_dbg(&cxlr->dev, "%s:%s invalid target for %s\n",
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev),
+			dev_name(cxlr->dev.parent));
+		return -ENXIO;
+	}
+
+	if (cxlrd->calc_hb(cxlrd, pos) != dport) {
+		dev_dbg(&cxlr->dev, "%s:%s invalid target position for %s\n",
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev),
+			dev_name(&cxlrd->cxlsd.cxld.dev));
+		return -ENXIO;
+	}
+
+	if (cxled->cxld.target_type != cxlr->type) {
+		dev_dbg(&cxlr->dev, "%s:%s type mismatch: %d vs %d\n",
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev),
+			cxled->cxld.target_type, cxlr->type);
+		return -ENXIO;
+	}
+
+	if (!cxled->dpa_res) {
+		dev_dbg(&cxlr->dev, "%s:%s: missing DPA allocation.\n",
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev));
+		return -ENXIO;
+	}
+
+	if (resource_size(cxled->dpa_res) * p->interleave_ways !=
+	    resource_size(p->res)) {
+		dev_dbg(&cxlr->dev,
+			"%s:%s: decoder-size-%#llx * ways-%d != region-size-%#llx\n",
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev),
+			(u64)resource_size(cxled->dpa_res), p->interleave_ways,
+			(u64)resource_size(p->res));
+		return -EINVAL;
+	}
+
+	for (iter = ep_port; !is_cxl_root(iter);
+	     iter = to_cxl_port(iter->dev.parent)) {
+		rc = cxl_port_attach_region(iter, cxlr, cxled, pos);
+		if (rc)
+			goto err;
+	}
+
 	p->targets[pos] = cxled;
 	cxled->pos = pos;
 	p->nr_targets++;
 
+	if (p->nr_targets == p->interleave_ways)
+		p->state = CXL_CONFIG_ACTIVE;
+
 	return 0;
+
+err:
+	for (iter = ep_port; !is_cxl_root(iter);
+	     iter = to_cxl_port(iter->dev.parent))
+		cxl_port_detach_region(iter, cxlr, cxled);
+	return rc;
 }
 
 static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
 {
+	struct cxl_port *iter, *ep_port = cxled_to_port(cxled);
 	struct cxl_region *cxlr = cxled->cxld.region;
 	struct cxl_region_params *p;
 
@@ -486,6 +834,10 @@ static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
 	p = &cxlr->params;
 	get_device(&cxlr->dev);
 
+	for (iter = ep_port; !is_cxl_root(iter);
+	     iter = to_cxl_port(iter->dev.parent))
+		cxl_port_detach_region(iter, cxlr, cxled);
+
 	if (cxled->pos < 0 || cxled->pos >= p->interleave_ways ||
 	    p->targets[cxled->pos] != cxled) {
 		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
@@ -496,10 +848,12 @@ static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
 		goto out;
 	}
 
+	if (p->state == CXL_CONFIG_ACTIVE)
+		p->state = CXL_CONFIG_INTERLEAVE_ACTIVE;
 	p->targets[cxled->pos] = NULL;
 	p->nr_targets--;
 
-	/* notify the region driver that one of its targets has deparated */
+	/* notify the region driver that one of its targets has departed */
 	up_write(&cxl_region_rwsem);
 	device_release_driver(&cxlr->dev);
 	down_write(&cxl_region_rwsem);
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index cd81e642e900..637768609a75 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -421,6 +421,7 @@ struct cxl_nvdimm {
  * @id: id for port device-name
  * @dports: cxl_dport instances referenced by decoders
  * @endpoints: cxl_ep instances, endpoints that are a descendant of this port
+ * @regions: cxl_region_ref instances, regions mapped by this port
  * @parent_dport: dport that points to this port in the parent
  * @decoder_ida: allocator for decoder ids
  * @hdm_end: track last allocated HDM decoder instance for allocation ordering
@@ -435,6 +436,7 @@ struct cxl_port {
 	int id;
 	struct xarray dports;
 	struct xarray endpoints;
+	struct xarray regions;
 	struct cxl_dport *parent_dport;
 	struct ida decoder_ida;
 	int hdm_end;
@@ -476,6 +478,24 @@ struct cxl_ep {
 	struct cxl_port *next;
 };
 
+/**
+ * struct cxl_region_ref - track a region's interest in a port
+ * @port: point in topology to install this reference
+ * @decoder: decoder assigned for @region in @port
+ * @region: region for this reference
+ * @endpoints: cxl_ep references for region members beneath @port
+ * @nr_eps: number of endpoints beneath @port
+ * @nr_targets: number of distinct targets needed to reach @nr_eps
+ */
+struct cxl_region_ref {
+	struct cxl_port *port;
+	struct cxl_decoder *decoder;
+	struct cxl_region *region;
+	struct xarray endpoints;
+	int nr_eps;
+	int nr_targets;
+};
+
 /*
  * The platform firmware device hosting the root is also the top of the
  * CXL port topology. All other CXL ports have another CXL port as their
diff --git a/drivers/cxl/cxlmem.h b/drivers/cxl/cxlmem.h
index eee96016c3c7..a83bb6782d23 100644
--- a/drivers/cxl/cxlmem.h
+++ b/drivers/cxl/cxlmem.h
@@ -55,6 +55,11 @@ static inline struct cxl_port *cxled_to_port(struct cxl_endpoint_decoder *cxled)
 	return to_cxl_port(cxled->cxld.dev.parent);
 }
 
+static inline struct cxl_port *cxlrd_to_port(struct cxl_root_decoder *cxlrd)
+{
+	return to_cxl_port(cxlrd->cxlsd.cxld.dev.parent);
+}
+
 static inline struct cxl_memdev *
 cxled_to_memdev(struct cxl_endpoint_decoder *cxled)
 {

[PATCH v2 24/28] cxl/region: Program target lists

[Dan Williams]

Once the region's interleave geometry (ways, granularity, size) is
established and all the endpoint decoder targets are assigned, the next
phase is to program all the intermediate decoders. Specifically, each
CXL switch in the path between the endpoint and its CXL host-bridge
(including the logical switch internal to the host-bridge) needs to have
its decoders programmed and the target list order assigned.

The difficulty in this implementation lies in determining which endpoint
decoder ordering combinations are valid. Consider the cxl_test case of 2
host bridges, each of those host-bridges attached to 2 switches, and
each of those switches attached to 2 endpoints for a potential 8-way
interleave. The x2 interleave at the host-bridge level requires that all
even numbered endpoint decoder positions be located on the "left" hand
side of the topology tree, and the odd numbered positions on the other.
The endpoints that are peers on the same switch need to have a position
that can be routed with a dedicated address bit per-endpoint. See
check_last_peer() for the details.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/core.h   |    4 +
 drivers/cxl/core/port.c   |    4 -
 drivers/cxl/core/region.c |  262 +++++++++++++++++++++++++++++++++++++++++++--
 drivers/cxl/cxl.h         |    2 
 4 files changed, 260 insertions(+), 12 deletions(-)

diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index 6edd8174c2b5..fcf14b8a3c87 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -42,9 +42,13 @@ resource_size_t cxl_dpa_resource_start(struct cxl_endpoint_decoder *cxled);
 extern struct rw_semaphore cxl_dpa_rwsem;
 
 bool is_switch_decoder(struct device *dev);
+struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev);
 static inline struct cxl_ep *cxl_ep_load(struct cxl_port *port,
 					 struct cxl_memdev *cxlmd)
 {
+	if (!port)
+		return NULL;
+
 	return xa_load(&port->endpoints, (unsigned long)&cxlmd->dev);
 }
 
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index d234afc47e89..215ce5e16986 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -146,8 +146,6 @@ static ssize_t emit_target_list(struct cxl_switch_decoder *cxlsd, char *buf)
 	return offset;
 }
 
-static struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev);
-
 static ssize_t target_list_show(struct device *dev,
 				struct device_attribute *attr, char *buf)
 {
@@ -472,7 +470,7 @@ struct cxl_endpoint_decoder *to_cxl_endpoint_decoder(struct device *dev)
 }
 EXPORT_SYMBOL_NS_GPL(to_cxl_endpoint_decoder, CXL);
 
-static struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev)
+struct cxl_switch_decoder *to_cxl_switch_decoder(struct device *dev)
 {
 	if (dev_WARN_ONCE(dev, !is_switch_decoder(dev),
 			  "not a cxl_switch_decoder device\n"))
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 8ac0c557f6aa..225340529fc3 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -485,6 +485,7 @@ static struct cxl_region_ref *alloc_region_ref(struct cxl_port *port,
 		return NULL;
 	cxl_rr->port = port;
 	cxl_rr->region = cxlr;
+	cxl_rr->nr_targets = 1;
 	xa_init(&cxl_rr->endpoints);
 
 	rc = xa_insert(&port->regions, (unsigned long)cxlr, cxl_rr, GFP_KERNEL);
@@ -525,10 +526,12 @@ static int cxl_rr_ep_add(struct cxl_region_ref *cxl_rr,
 	struct cxl_decoder *cxld = cxl_rr->decoder;
 	struct cxl_ep *ep = cxl_ep_load(port, cxled_to_memdev(cxled));
 
-	rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
-			 GFP_KERNEL);
-	if (rc)
-		return rc;
+	if (ep) {
+		rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
+			       GFP_KERNEL);
+		if (rc)
+			return rc;
+	}
 	cxl_rr->nr_eps++;
 
 	if (!cxld->region) {
@@ -565,7 +568,7 @@ static int cxl_port_attach_region(struct cxl_port *port,
 				  struct cxl_endpoint_decoder *cxled, int pos)
 {
 	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
-	struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
+	const struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
 	struct cxl_region_ref *cxl_rr = NULL, *iter;
 	struct cxl_region_params *p = &cxlr->params;
 	struct cxl_decoder *cxld = NULL;
@@ -649,6 +652,16 @@ static int cxl_port_attach_region(struct cxl_port *port,
 		goto out_erase;
 	}
 
+	dev_dbg(&cxlr->dev,
+		"%s:%s %s add: %s:%s @ %d next: %s nr_eps: %d nr_targets: %d\n",
+		dev_name(port->uport), dev_name(&port->dev),
+		dev_name(&cxld->dev), dev_name(&cxlmd->dev),
+		dev_name(&cxled->cxld.dev), pos,
+		ep ? ep->next ? dev_name(ep->next->uport) :
+				      dev_name(&cxlmd->dev) :
+			   "none",
+		cxl_rr->nr_eps, cxl_rr->nr_targets);
+
 	return 0;
 out_erase:
 	if (cxl_rr->nr_eps == 0)
@@ -667,7 +680,7 @@ static void cxl_port_detach_region(struct cxl_port *port,
 				   struct cxl_endpoint_decoder *cxled)
 {
 	struct cxl_region_ref *cxl_rr;
-	struct cxl_ep *ep;
+	struct cxl_ep *ep = NULL;
 
 	lockdep_assert_held_write(&cxl_region_rwsem);
 
@@ -675,7 +688,14 @@ static void cxl_port_detach_region(struct cxl_port *port,
 	if (!cxl_rr)
 		return;
 
-	ep = xa_erase(&cxl_rr->endpoints, (unsigned long)cxled);
+	/*
+	 * Endpoint ports do not carry cxl_ep references, and they
+	 * never target more than one endpoint by definition
+	 */
+	if (cxl_rr->decoder == &cxled->cxld)
+		cxl_rr->nr_eps--;
+	else
+		ep = xa_erase(&cxl_rr->endpoints, (unsigned long)cxled);
 	if (ep) {
 		struct cxl_ep *ep_iter;
 		unsigned long index;
@@ -696,6 +716,224 @@ static void cxl_port_detach_region(struct cxl_port *port,
 		free_region_ref(cxl_rr);
 }
 
+static int check_last_peer(struct cxl_endpoint_decoder *cxled,
+			   struct cxl_ep *ep, struct cxl_region_ref *cxl_rr,
+			   int distance)
+{
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_region *cxlr = cxl_rr->region;
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_endpoint_decoder *cxled_peer;
+	struct cxl_port *port = cxl_rr->port;
+	struct cxl_memdev *cxlmd_peer;
+	struct cxl_ep *ep_peer;
+	int pos = cxled->pos;
+
+	/*
+	 * If this position wants to share a dport with the last endpoint mapped
+	 * then that endpoint, at index 'position - distance', must also be
+	 * mapped by this dport.
+	 */
+	if (pos < distance) {
+		dev_dbg(&cxlr->dev, "%s:%s: cannot host %s:%s at %d\n",
+			dev_name(port->uport), dev_name(&port->dev),
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev), pos);
+		return -ENXIO;
+	}
+	cxled_peer = p->targets[pos - distance];
+	cxlmd_peer = cxled_to_memdev(cxled_peer);
+	ep_peer = cxl_ep_load(port, cxlmd_peer);
+	if (ep->dport != ep_peer->dport) {
+		dev_dbg(&cxlr->dev,
+			"%s:%s: %s:%s pos %d mismatched peer %s:%s\n",
+			dev_name(port->uport), dev_name(&port->dev),
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev), pos,
+			dev_name(&cxlmd_peer->dev),
+			dev_name(&cxled_peer->cxld.dev));
+		return -ENXIO;
+	}
+
+	return 0;
+}
+
+static int cxl_port_setup_targets(struct cxl_port *port,
+				  struct cxl_region *cxlr,
+				  struct cxl_endpoint_decoder *cxled)
+{
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(cxlr->dev.parent);
+	int parent_iw, parent_ig, ig, iw, rc, inc = 0, pos = cxled->pos;
+	struct cxl_port *parent_port = to_cxl_port(port->dev.parent);
+	struct cxl_region_ref *cxl_rr = cxl_rr_load(port, cxlr);
+	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+	struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_decoder *cxld = cxl_rr->decoder;
+	struct cxl_switch_decoder *cxlsd;
+	u16 eig, peig;
+	u8 eiw, peiw;
+
+	/*
+	 * While root level decoders support x3, x6, x12, switch level
+	 * decoders only support powers of 2 up to x16.
+	 */
+	if (!is_power_of_2(cxl_rr->nr_targets)) {
+		dev_dbg(&cxlr->dev, "%s:%s: invalid target count %d\n",
+			dev_name(port->uport), dev_name(&port->dev),
+			cxl_rr->nr_targets);
+		return -EINVAL;
+	}
+
+	cxlsd = to_cxl_switch_decoder(&cxld->dev);
+	if (cxl_rr->nr_targets_set) {
+		int i, distance;
+
+		distance = p->nr_targets / cxl_rr->nr_targets;
+		for (i = 0; i < cxl_rr->nr_targets_set; i++)
+			if (ep->dport == cxlsd->target[i]) {
+				rc = check_last_peer(cxled, ep, cxl_rr,
+						     distance);
+				if (rc)
+					return rc;
+				goto out_target_set;
+			}
+		goto add_target;
+	}
+
+	if (is_cxl_root(parent_port)) {
+		parent_ig = cxlrd->cxlsd.cxld.interleave_granularity;
+		parent_iw = cxlrd->cxlsd.cxld.interleave_ways;
+		/*
+		 * For purposes of address bit routing, use power-of-2 math for
+		 * switch ports.
+		 */
+		if (!is_power_of_2(parent_iw))
+			parent_iw /= 3;
+	} else {
+		struct cxl_region_ref *parent_rr;
+		struct cxl_decoder *parent_cxld;
+
+		parent_rr = cxl_rr_load(parent_port, cxlr);
+		parent_cxld = parent_rr->decoder;
+		parent_ig = parent_cxld->interleave_granularity;
+		parent_iw = parent_cxld->interleave_ways;
+	}
+
+	granularity_to_cxl(parent_ig, &peig);
+	ways_to_cxl(parent_iw, &peiw);
+
+	iw = cxl_rr->nr_targets;
+	ways_to_cxl(iw, &eiw);
+	if (cxl_rr->nr_targets > 1) {
+		u32 address_bit = max(peig + peiw, eiw + peig);
+
+		eig = address_bit - eiw + 1;
+	} else {
+		eiw = peiw;
+		eig = peig;
+	}
+
+	rc = cxl_to_granularity(eig, &ig);
+	if (rc) {
+		dev_dbg(&cxlr->dev, "%s:%s: invalid interleave: %d\n",
+			dev_name(port->uport), dev_name(&port->dev),
+			256 << eig);
+		return rc;
+	}
+
+	cxld->interleave_ways = iw;
+	cxld->interleave_granularity = ig;
+	dev_dbg(&cxlr->dev, "%s:%s iw: %d ig: %d\n", dev_name(port->uport),
+		dev_name(&port->dev), iw, ig);
+add_target:
+	if (cxl_rr->nr_targets_set == cxl_rr->nr_targets) {
+		dev_dbg(&cxlr->dev,
+			"%s:%s: targets full trying to add %s:%s at %d\n",
+			dev_name(port->uport), dev_name(&port->dev),
+			dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev), pos);
+		return -ENXIO;
+	}
+	cxlsd->target[cxl_rr->nr_targets_set] = ep->dport;
+	inc = 1;
+out_target_set:
+	cxl_rr->nr_targets_set += inc;
+	dev_dbg(&cxlr->dev, "%s:%s target[%d] = %s for %s:%s @ %d\n",
+		dev_name(port->uport), dev_name(&port->dev),
+		cxl_rr->nr_targets_set - 1, dev_name(ep->dport->dport),
+		dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev), pos);
+
+	return 0;
+}
+
+static void cxl_port_reset_targets(struct cxl_port *port,
+				   struct cxl_region *cxlr)
+{
+	struct cxl_region_ref *cxl_rr = cxl_rr_load(port, cxlr);
+
+	/*
+	 * After the last endpoint has been detached the entire cxl_rr may now
+	 * be gone.
+	 */
+	if (cxl_rr)
+		cxl_rr->nr_targets_set = 0;
+}
+
+static void cxl_region_teardown_targets(struct cxl_region *cxlr)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_endpoint_decoder *cxled;
+	struct cxl_memdev *cxlmd;
+	struct cxl_port *iter;
+	struct cxl_ep *ep;
+	int i;
+
+	for (i = 0; i < p->nr_targets; i++) {
+		cxled = p->targets[i];
+		cxlmd = cxled_to_memdev(cxled);
+
+		iter = cxled_to_port(cxled);
+		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
+			iter = to_cxl_port(iter->dev.parent);
+
+		for (ep = cxl_ep_load(iter, cxlmd); iter;
+		     iter = ep->next, ep = cxl_ep_load(iter, cxlmd))
+			cxl_port_reset_targets(iter, cxlr);
+	}
+}
+
+static int cxl_region_setup_targets(struct cxl_region *cxlr)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_endpoint_decoder *cxled;
+	struct cxl_memdev *cxlmd;
+	struct cxl_port *iter;
+	struct cxl_ep *ep;
+	int i, rc;
+
+	for (i = 0; i < p->nr_targets; i++) {
+		cxled = p->targets[i];
+		cxlmd = cxled_to_memdev(cxled);
+
+		iter = cxled_to_port(cxled);
+		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
+			iter = to_cxl_port(iter->dev.parent);
+
+		/*
+		 * Descend the topology tree programming targets while
+		 * looking for conflicts.
+		 */
+		for (ep = cxl_ep_load(iter, cxlmd); iter;
+		     iter = ep->next, ep = cxl_ep_load(iter, cxlmd)) {
+			rc = cxl_port_setup_targets(iter, cxlr, cxled);
+			if (rc) {
+				cxl_region_teardown_targets(cxlr);
+				return rc;
+			}
+		}
+	}
+
+	return 0;
+}
+
 static int cxl_region_attach(struct cxl_region *cxlr,
 			     struct cxl_endpoint_decoder *cxled, int pos)
 {
@@ -808,8 +1046,12 @@ static int cxl_region_attach(struct cxl_region *cxlr,
 	cxled->pos = pos;
 	p->nr_targets++;
 
-	if (p->nr_targets == p->interleave_ways)
+	if (p->nr_targets == p->interleave_ways) {
+		rc = cxl_region_setup_targets(cxlr);
+		if (rc)
+			goto err;
 		p->state = CXL_CONFIG_ACTIVE;
+	}
 
 	return 0;
 
@@ -848,8 +1090,10 @@ static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
 		goto out;
 	}
 
-	if (p->state == CXL_CONFIG_ACTIVE)
+	if (p->state == CXL_CONFIG_ACTIVE) {
 		p->state = CXL_CONFIG_INTERLEAVE_ACTIVE;
+		cxl_region_teardown_targets(cxlr);
+	}
 	p->targets[cxled->pos] = NULL;
 	p->nr_targets--;
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 637768609a75..70862141209b 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -484,6 +484,7 @@ struct cxl_ep {
  * @decoder: decoder assigned for @region in @port
  * @region: region for this reference
  * @endpoints: cxl_ep references for region members beneath @port
+ * @nr_targets_set: track how many targets have been programmed during setup
  * @nr_eps: number of endpoints beneath @port
  * @nr_targets: number of distinct targets needed to reach @nr_eps
  */
@@ -492,6 +493,7 @@ struct cxl_region_ref {
 	struct cxl_decoder *decoder;
 	struct cxl_region *region;
 	struct xarray endpoints;
+	int nr_targets_set;
 	int nr_eps;
 	int nr_targets;
 };

[PATCH v2 25/28] cxl/hdm: Commit decoder state to hardware

[Dan Williams]

After all the soft validation of the region has completed, convey the
region configuration to hardware while being careful to commit decoders
in specification mandated order. In addition to programming the endpoint
decoder base-address, interleave ways and granularity, the switch
decoder target lists are also established.

While the kernel can enforce spec-mandated commit order, it can not
enforce spec-mandated reset order. For example, the kernel can't stop
someone from removing an endpoint device that is occupying decoderN in a
switch decoder where decoderN+1 is also committed. To reset decoderN,
decoderN+1 must be torn down first. That "tear down the world"
implementation is saved for a follow-on patch.

Callback operations are provided for the 'commit' and 'reset'
operations. While those callbacks may prove useful for CXL accelerators
(Type-2 devices with memory) the primary motivation is to enable a
simple way for cxl_test to intercept those operations.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 Documentation/ABI/testing/sysfs-bus-cxl |   16 ++
 drivers/cxl/core/hdm.c                  |  218 +++++++++++++++++++++++++++++++
 drivers/cxl/core/port.c                 |    1 
 drivers/cxl/core/region.c               |  194 ++++++++++++++++++++++++++--
 drivers/cxl/cxl.h                       |   13 ++
 tools/testing/cxl/test/cxl.c            |   46 +++++++
 6 files changed, 477 insertions(+), 11 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 94e19e24de8d..2c42888a3df0 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -361,3 +361,19 @@ Description:
 		not an endpoint decoder. Once all positions have been
 		successfully written a final validation for decode conflicts is
 		performed before activating the region.
+
+
+What:		/sys/bus/cxl/devices/regionZ/commit
+Date:		May, 2022
+KernelVersion:	v5.20
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RW) Write a boolean 'true' string value to this attribute to
+		trigger the region to transition from the software programmed
+		state to the actively decoding in hardware state. The commit
+		operation in addition to validating that the region is in proper
+		configured state, validates that the decoders are being
+		committed in spec mandated order (last committed decoder id +
+		1), and checks that the hardware accepts the commit request.
+		Reading this value indicates whether the region is committed or
+		not.
diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 81645de1064f..88edb8391fbd 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -129,6 +129,8 @@ struct cxl_hdm *devm_cxl_setup_hdm(struct cxl_port *port)
 		return ERR_PTR(-ENXIO);
 	}
 
+	dev_set_drvdata(dev, cxlhdm);
+
 	return cxlhdm;
 }
 EXPORT_SYMBOL_NS_GPL(devm_cxl_setup_hdm, CXL);
@@ -462,6 +464,213 @@ int cxl_dpa_alloc(struct cxl_endpoint_decoder *cxled, unsigned long long size)
 	return devm_add_action_or_reset(&port->dev, cxl_dpa_release, cxled);
 }
 
+static void cxld_set_interleave(struct cxl_decoder *cxld, u32 *ctrl)
+{
+	u16 eig;
+	u8 eiw;
+
+	ways_to_cxl(cxld->interleave_ways, &eiw);
+	granularity_to_cxl(cxld->interleave_granularity, &eig);
+
+	u32p_replace_bits(ctrl, eig, CXL_HDM_DECODER0_CTRL_IG_MASK);
+	u32p_replace_bits(ctrl, eiw, CXL_HDM_DECODER0_CTRL_IW_MASK);
+	*ctrl |= CXL_HDM_DECODER0_CTRL_COMMIT;
+}
+
+static void cxld_set_type(struct cxl_decoder *cxld, u32 *ctrl)
+{
+	u32p_replace_bits(ctrl, !!(cxld->target_type == 3),
+			  CXL_HDM_DECODER0_CTRL_TYPE);
+}
+
+static void cxld_set_hpa(struct cxl_decoder *cxld, u64 *base, u64 *size)
+{
+	struct cxl_region *cxlr = cxld->region;
+	struct cxl_region_params *p = &cxlr->params;
+
+	cxld->hpa_range = (struct range) {
+		.start = p->res->start,
+		.end = p->res->end,
+	};
+
+	*base = p->res->start;
+	*size = resource_size(p->res);
+}
+
+static void cxld_clear_hpa(struct cxl_decoder *cxld)
+{
+	cxld->hpa_range = (struct range) {
+		.start = 0,
+		.end = -1,
+	};
+}
+
+static int cxlsd_set_targets(struct cxl_switch_decoder *cxlsd, u64 *tgt)
+{
+	struct cxl_dport **t = &cxlsd->target[0];
+	int ways = cxlsd->cxld.interleave_ways;
+
+	if (dev_WARN_ONCE(&cxlsd->cxld.dev,
+			  ways > 8 || ways > cxlsd->nr_targets,
+			  "ways: %d overflows targets: %d\n", ways,
+			  cxlsd->nr_targets))
+		return -ENXIO;
+
+	*tgt = FIELD_PREP(GENMASK(7, 0), t[0]->port_id);
+	if (ways > 1)
+		*tgt |= FIELD_PREP(GENMASK(15, 8), t[1]->port_id);
+	if (ways > 2)
+		*tgt |= FIELD_PREP(GENMASK(23, 16), t[2]->port_id);
+	if (ways > 3)
+		*tgt |= FIELD_PREP(GENMASK(31, 24), t[3]->port_id);
+	if (ways > 4)
+		*tgt |= FIELD_PREP(GENMASK_ULL(39, 32), t[4]->port_id);
+	if (ways > 5)
+		*tgt |= FIELD_PREP(GENMASK_ULL(47, 40), t[5]->port_id);
+	if (ways > 6)
+		*tgt |= FIELD_PREP(GENMASK_ULL(55, 48), t[6]->port_id);
+	if (ways > 7)
+		*tgt |= FIELD_PREP(GENMASK_ULL(63, 56), t[7]->port_id);
+
+	return 0;
+}
+
+/*
+ * Per CXL 2.0 8.2.5.12.20 Committing Decoder Programming, hardware must set
+ * committed or error within 10ms, but just be generous with 20ms to account for
+ * clock skew and other marginal behavior
+ */
+#define COMMIT_TIMEOUT_MS 20
+static int cxld_await_commit(void __iomem *hdm, int id)
+{
+	u32 ctrl;
+	int i;
+
+	for (i = 0; i < COMMIT_TIMEOUT_MS; i++) {
+		ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(id));
+		if (FIELD_GET(CXL_HDM_DECODER0_CTRL_COMMIT_ERROR, ctrl)) {
+			ctrl &= ~CXL_HDM_DECODER0_CTRL_COMMIT;
+			writel(ctrl, hdm + CXL_HDM_DECODER0_CTRL_OFFSET(id));
+			return -EIO;
+		}
+		if (FIELD_GET(CXL_HDM_DECODER0_CTRL_COMMITTED, ctrl))
+			return 0;
+		fsleep(1000);
+	}
+
+	return -ETIMEDOUT;
+}
+
+static int cxl_decoder_commit(struct cxl_decoder *cxld)
+{
+	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
+	struct cxl_hdm *cxlhdm = dev_get_drvdata(&port->dev);
+	void __iomem *hdm = cxlhdm->regs.hdm_decoder;
+	int id = cxld->id, rc;
+	u64 base, size;
+	u32 ctrl;
+
+	if (cxld->flags & CXL_DECODER_F_ENABLE)
+		return 0;
+
+	if (port->commit_end + 1 != id) {
+		dev_dbg(&port->dev,
+			"%s: out of order commit, expected decoder%d.%d\n",
+			dev_name(&cxld->dev), port->id, port->commit_end + 1);
+		return -EBUSY;
+	}
+
+	down_read(&cxl_dpa_rwsem);
+	/* common decoder settings */
+	ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(cxld->id));
+	cxld_set_interleave(cxld, &ctrl);
+	cxld_set_type(cxld, &ctrl);
+	cxld_set_hpa(cxld, &base, &size);
+
+	writel(upper_32_bits(base), hdm + CXL_HDM_DECODER0_BASE_HIGH_OFFSET(id));
+	writel(lower_32_bits(base), hdm + CXL_HDM_DECODER0_BASE_LOW_OFFSET(id));
+	writel(upper_32_bits(size), hdm + CXL_HDM_DECODER0_SIZE_HIGH_OFFSET(id));
+	writel(lower_32_bits(size), hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(id));
+
+	if (is_switch_decoder(&cxld->dev)) {
+		struct cxl_switch_decoder *cxlsd =
+			to_cxl_switch_decoder(&cxld->dev);
+		void __iomem *tl_hi = hdm + CXL_HDM_DECODER0_TL_HIGH(id);
+		void __iomem *tl_lo = hdm + CXL_HDM_DECODER0_TL_LOW(id);
+		u64 targets;
+
+		rc = cxlsd_set_targets(cxlsd, &targets);
+		if (rc) {
+			dev_dbg(&port->dev, "%s: target configuration error\n",
+				dev_name(&cxld->dev));
+			goto err;
+		}
+
+		writel(upper_32_bits(targets), tl_hi);
+		writel(lower_32_bits(targets), tl_lo);
+	} else {
+		struct cxl_endpoint_decoder *cxled =
+			to_cxl_endpoint_decoder(&cxld->dev);
+		void __iomem *sk_hi = hdm + CXL_HDM_DECODER0_SKIP_HIGH(id);
+		void __iomem *sk_lo = hdm + CXL_HDM_DECODER0_SKIP_LOW(id);
+
+		writel(upper_32_bits(cxled->skip), sk_hi);
+		writel(lower_32_bits(cxled->skip), sk_lo);
+	}
+
+	writel(ctrl, hdm + CXL_HDM_DECODER0_CTRL_OFFSET(id));
+	up_read(&cxl_dpa_rwsem);
+
+	port->commit_end++;
+	rc = cxld_await_commit(hdm, cxld->id);
+err:
+	if (rc) {
+		dev_dbg(&port->dev, "%s: error %d committing decoder\n",
+			dev_name(&cxld->dev), rc);
+		cxld->reset(cxld);
+		return rc;
+	}
+	cxld->flags |= CXL_DECODER_F_ENABLE;
+
+	return 0;
+}
+
+static int cxl_decoder_reset(struct cxl_decoder *cxld)
+{
+	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
+	struct cxl_hdm *cxlhdm = dev_get_drvdata(&port->dev);
+	void __iomem *hdm = cxlhdm->regs.hdm_decoder;
+	int id = cxld->id;
+	u32 ctrl;
+
+	if ((cxld->flags & CXL_DECODER_F_ENABLE) == 0)
+		return 0;
+
+	if (port->commit_end != id) {
+		dev_dbg(&port->dev,
+			"%s: out of order reset, expected decoder%d.%d\n",
+			dev_name(&cxld->dev), port->id, port->commit_end);
+		return -EBUSY;
+	}
+
+	down_read(&cxl_dpa_rwsem);
+	ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(id));
+	ctrl &= ~CXL_HDM_DECODER0_CTRL_COMMIT;
+	writel(ctrl, hdm + CXL_HDM_DECODER0_CTRL_OFFSET(id));
+
+	cxld_clear_hpa(cxld);
+	writel(0, hdm + CXL_HDM_DECODER0_SIZE_HIGH_OFFSET(id));
+	writel(0, hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(id));
+	writel(0, hdm + CXL_HDM_DECODER0_BASE_HIGH_OFFSET(id));
+	writel(0, hdm + CXL_HDM_DECODER0_BASE_LOW_OFFSET(id));
+	up_read(&cxl_dpa_rwsem);
+
+	port->commit_end--;
+	cxld->flags &= ~CXL_DECODER_F_ENABLE;
+
+	return 0;
+}
+
 static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 			    int *target_map, void __iomem *hdm, int which,
 			    u64 *dpa_base)
@@ -484,6 +693,8 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 	base = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_BASE_LOW_OFFSET(which));
 	size = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(which));
 	committed = !!(ctrl & CXL_HDM_DECODER0_CTRL_COMMITTED);
+	cxld->commit = cxl_decoder_commit;
+	cxld->reset = cxl_decoder_reset;
 
 	if (!committed)
 		size = 0;
@@ -507,6 +718,13 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 			cxld->target_type = CXL_DECODER_EXPANDER;
 		else
 			cxld->target_type = CXL_DECODER_ACCELERATOR;
+		if (cxld->id != port->commit_end + 1) {
+			dev_warn(&port->dev,
+				 "decoder%d.%d: Committed out of order\n",
+				 port->id, cxld->id);
+			return -ENXIO;
+		}
+		port->commit_end = cxld->id;
 	} else {
 		/* unless / until type-2 drivers arrive, assume type-3 */
 		if (FIELD_GET(CXL_HDM_DECODER0_CTRL_TYPE, ctrl) == 0) {
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 215ce5e16986..7ab9a98c5d4f 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -632,6 +632,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
 	port->component_reg_phys = component_reg_phys;
 	ida_init(&port->decoder_ida);
 	port->hdm_end = -1;
+	port->commit_end = -1;
 	xa_init(&port->dports);
 	xa_init(&port->endpoints);
 	xa_init(&port->regions);
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 225340529fc3..de794344d964 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -115,6 +115,173 @@ static ssize_t uuid_store(struct device *dev, struct device_attribute *attr,
 }
 static DEVICE_ATTR_RW(uuid);
 
+static struct cxl_region_ref *cxl_rr_load(struct cxl_port *port,
+					  struct cxl_region *cxlr)
+{
+	return xa_load(&port->regions, (unsigned long)cxlr);
+}
+
+static int cxl_region_decode_reset(struct cxl_region *cxlr, int count)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	int i;
+
+	for (i = count - 1; i >= 0; i--) {
+		struct cxl_endpoint_decoder *cxled = p->targets[i];
+		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+		struct cxl_port *iter = cxled_to_port(cxled);
+		struct cxl_ep *ep;
+		int rc;
+
+		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
+			iter = to_cxl_port(iter->dev.parent);
+
+		for (ep = cxl_ep_load(iter, cxlmd); iter;
+		     iter = ep->next, ep = cxl_ep_load(iter, cxlmd)) {
+			struct cxl_region_ref *cxl_rr;
+			struct cxl_decoder *cxld;
+
+			cxl_rr = cxl_rr_load(iter, cxlr);
+			cxld = cxl_rr->decoder;
+			rc = cxld->reset(cxld);
+			if (rc)
+				return rc;
+		}
+
+		rc = cxled->cxld.reset(&cxled->cxld);
+		if (rc)
+			return rc;
+	}
+
+	return 0;
+}
+
+static int cxl_region_decode_commit(struct cxl_region *cxlr)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	int i, rc;
+
+	for (i = 0; i < p->nr_targets; i++) {
+		struct cxl_endpoint_decoder *cxled = p->targets[i];
+		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+		struct cxl_region_ref *cxl_rr;
+		struct cxl_decoder *cxld;
+		struct cxl_port *iter;
+		struct cxl_ep *ep;
+
+		/* commit bottom up */
+		for (iter = cxled_to_port(cxled); !is_cxl_root(iter);
+		     iter = to_cxl_port(iter->dev.parent)) {
+			cxl_rr = cxl_rr_load(iter, cxlr);
+			cxld = cxl_rr->decoder;
+			rc = cxld->commit(cxld);
+			if (rc)
+				break;
+		}
+
+		/* success, all decoders up to the root are programmed */
+		if (is_cxl_root(iter))
+			continue;
+
+		/* programming @iter failed, teardown */
+		for (ep = cxl_ep_load(iter, cxlmd); ep && iter;
+		     iter = ep->next, ep = cxl_ep_load(iter, cxlmd)) {
+			cxl_rr = cxl_rr_load(iter, cxlr);
+			cxld = cxl_rr->decoder;
+			cxld->reset(cxld);
+		}
+
+		cxled->cxld.reset(&cxled->cxld);
+		if (i == 0)
+			return rc;
+		break;
+	}
+
+	if (i >= p->nr_targets)
+		return 0;
+
+	/* undo the targets that were successfully committed */
+	cxl_region_decode_reset(cxlr, i);
+	return rc;
+}
+
+static ssize_t commit_store(struct device *dev, struct device_attribute *attr,
+			    const char *buf, size_t len)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	bool commit;
+	ssize_t rc;
+
+	rc = kstrtobool(buf, &commit);
+	if (rc)
+		return rc;
+
+	rc = down_write_killable(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+
+	/* Already in the requested state? */
+	if (commit && p->state >= CXL_CONFIG_COMMIT)
+		goto out;
+	if (!commit && p->state < CXL_CONFIG_COMMIT)
+		goto out;
+
+	/* Not ready to commit? */
+	if (commit && p->state < CXL_CONFIG_ACTIVE) {
+		rc = -ENXIO;
+		goto out;
+	}
+
+	if (commit)
+		rc = cxl_region_decode_commit(cxlr);
+	else {
+		p->state = CXL_CONFIG_RESET_PENDING;
+		up_write(&cxl_region_rwsem);
+		device_release_driver(&cxlr->dev);
+		down_write(&cxl_region_rwsem);
+
+		/*
+		 * The lock was dropped, so need to revalidate that the reset is
+		 * still pending.
+		 */
+		if (p->state == CXL_CONFIG_RESET_PENDING)
+			rc = cxl_region_decode_reset(cxlr, p->interleave_ways);
+	}
+
+	if (rc)
+		goto out;
+
+	if (commit)
+		p->state = CXL_CONFIG_COMMIT;
+	else if (p->state == CXL_CONFIG_RESET_PENDING)
+		p->state = CXL_CONFIG_ACTIVE;
+
+out:
+	up_write(&cxl_region_rwsem);
+
+	if (rc)
+		return rc;
+	return len;
+}
+
+static ssize_t commit_show(struct device *dev, struct device_attribute *attr,
+			   char *buf)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	ssize_t rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc)
+		return rc;
+	rc = sysfs_emit(buf, "%d\n", p->state >= CXL_CONFIG_COMMIT);
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+static DEVICE_ATTR_RW(commit);
+
 static umode_t cxl_region_visible(struct kobject *kobj, struct attribute *a,
 				  int n)
 {
@@ -393,6 +560,7 @@ static DEVICE_ATTR_RW(size);
 
 static struct attribute *cxl_region_attrs[] = {
 	&dev_attr_uuid.attr,
+	&dev_attr_commit.attr,
 	&dev_attr_interleave_ways.attr,
 	&dev_attr_interleave_granularity.attr,
 	&dev_attr_resource.attr,
@@ -669,12 +837,6 @@ static int cxl_port_attach_region(struct cxl_port *port,
 	return rc;
 }
 
-static struct cxl_region_ref *cxl_rr_load(struct cxl_port *port,
-					  struct cxl_region *cxlr)
-{
-	return xa_load(&port->regions, (unsigned long)cxlr);
-}
-
 static void cxl_port_detach_region(struct cxl_port *port,
 				   struct cxl_region *cxlr,
 				   struct cxl_endpoint_decoder *cxled)
@@ -1062,20 +1224,32 @@ static int cxl_region_attach(struct cxl_region *cxlr,
 	return rc;
 }
 
-static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
+static int cxl_region_detach(struct cxl_endpoint_decoder *cxled)
 {
 	struct cxl_port *iter, *ep_port = cxled_to_port(cxled);
 	struct cxl_region *cxlr = cxled->cxld.region;
 	struct cxl_region_params *p;
+	int rc = 0;
 
 	lockdep_assert_held_write(&cxl_region_rwsem);
 
 	if (!cxlr)
-		return;
+		return 0;
 
 	p = &cxlr->params;
 	get_device(&cxlr->dev);
 
+	if (p->state > CXL_CONFIG_ACTIVE) {
+		/*
+		 * TODO: tear down all impacted regions if a device is
+		 * removed out of order
+		 */
+		rc = cxl_region_decode_reset(cxlr, p->interleave_ways);
+		if (rc)
+			goto out;
+		p->state = CXL_CONFIG_ACTIVE;
+	}
+
 	for (iter = ep_port; !is_cxl_root(iter);
 	     iter = to_cxl_port(iter->dev.parent))
 		cxl_port_detach_region(iter, cxlr, cxled);
@@ -1103,6 +1277,7 @@ static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
 	down_write(&cxl_region_rwsem);
 out:
 	put_device(&cxlr->dev);
+	return rc;
 }
 
 void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled)
@@ -1160,8 +1335,7 @@ static int detach_target(struct cxl_region *cxlr, int pos)
 		goto out;
 	}
 
-	cxl_region_detach(p->targets[pos]);
-	rc = 0;
+	rc = cxl_region_detach(p->targets[pos]);
 out:
 	up_write(&cxl_region_rwsem);
 	return rc;
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 70862141209b..a51709613c43 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -54,6 +54,7 @@
 #define   CXL_HDM_DECODER0_CTRL_LOCK BIT(8)
 #define   CXL_HDM_DECODER0_CTRL_COMMIT BIT(9)
 #define   CXL_HDM_DECODER0_CTRL_COMMITTED BIT(10)
+#define   CXL_HDM_DECODER0_CTRL_COMMIT_ERROR BIT(11)
 #define   CXL_HDM_DECODER0_CTRL_TYPE BIT(12)
 #define CXL_HDM_DECODER0_TL_LOW(i) (0x20 * (i) + 0x24)
 #define CXL_HDM_DECODER0_TL_HIGH(i) (0x20 * (i) + 0x28)
@@ -257,7 +258,9 @@ enum cxl_decoder_type {
  * @target_type: accelerator vs expander (type2 vs type3) selector
  * @region: currently assigned region for this decoder
  * @flags: memory type capabilities and locking
- */
+ * @commit: device/decoder-type specific callback to commit settings to hw
+ * @reset: device/decoder-type specific callback to reset hw settings
+*/
 struct cxl_decoder {
 	struct device dev;
 	int id;
@@ -267,6 +270,8 @@ struct cxl_decoder {
 	enum cxl_decoder_type target_type;
 	struct cxl_region *region;
 	unsigned long flags;
+	int (*commit)(struct cxl_decoder *cxld);
+	int (*reset)(struct cxl_decoder *cxld);
 };
 
 /*
@@ -339,11 +344,15 @@ struct cxl_root_decoder {
  * changes to interleave_ways or interleave_granularity
  * @CXL_CONFIG_ACTIVE: All targets have been added the region is now
  * active
+ * @CXL_CONFIG_RESET_PENDING: see commit_store()
+ * @CXL_CONFIG_COMMIT: Soft-config has been committed to hardware
  */
 enum cxl_config_state {
 	CXL_CONFIG_IDLE,
 	CXL_CONFIG_INTERLEAVE_ACTIVE,
 	CXL_CONFIG_ACTIVE,
+	CXL_CONFIG_RESET_PENDING,
+	CXL_CONFIG_COMMIT,
 };
 
 /**
@@ -425,6 +434,7 @@ struct cxl_nvdimm {
  * @parent_dport: dport that points to this port in the parent
  * @decoder_ida: allocator for decoder ids
  * @hdm_end: track last allocated HDM decoder instance for allocation ordering
+ * @commit_end: cursor to track highest committed decoder for commit ordering
  * @component_reg_phys: component register capability base address (optional)
  * @dead: last ep has been removed, force port re-creation
  * @depth: How deep this port is relative to the root. depth 0 is the root.
@@ -440,6 +450,7 @@ struct cxl_port {
 	struct cxl_dport *parent_dport;
 	struct ida decoder_ida;
 	int hdm_end;
+	int commit_end;
 	resource_size_t component_reg_phys;
 	bool dead;
 	unsigned int depth;
diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
index 4dad0fa7ac4c..a072b2d3e726 100644
--- a/tools/testing/cxl/test/cxl.c
+++ b/tools/testing/cxl/test/cxl.c
@@ -429,6 +429,50 @@ static int map_targets(struct device *dev, void *data)
 	return 0;
 }
 
+static int mock_decoder_commit(struct cxl_decoder *cxld)
+{
+	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
+	int id = cxld->id;
+
+	if (cxld->flags & CXL_DECODER_F_ENABLE)
+		return 0;
+
+	dev_dbg(&port->dev, "%s commit\n", dev_name(&cxld->dev));
+	if (port->commit_end + 1 != id) {
+		dev_dbg(&port->dev,
+			"%s: out of order commit, expected decoder%d.%d\n",
+			dev_name(&cxld->dev), port->id, port->commit_end + 1);
+		return -EBUSY;
+	}
+
+	port->commit_end++;
+	cxld->flags |= CXL_DECODER_F_ENABLE;
+
+	return 0;
+}
+
+static int mock_decoder_reset(struct cxl_decoder *cxld)
+{
+	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
+	int id = cxld->id;
+
+	if ((cxld->flags & CXL_DECODER_F_ENABLE) == 0)
+		return 0;
+
+	dev_dbg(&port->dev, "%s reset\n", dev_name(&cxld->dev));
+	if (port->commit_end != id) {
+		dev_dbg(&port->dev,
+			"%s: out of order reset, expected decoder%d.%d\n",
+			dev_name(&cxld->dev), port->id, port->commit_end);
+		return -EBUSY;
+	}
+
+	port->commit_end--;
+	cxld->flags &= ~CXL_DECODER_F_ENABLE;
+
+	return 0;
+}
+
 static int mock_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 {
 	struct cxl_port *port = cxlhdm->port;
@@ -482,6 +526,8 @@ static int mock_cxl_enumerate_decoders(struct cxl_hdm *cxlhdm)
 		cxld->interleave_ways = min_not_zero(target_count, 1);
 		cxld->interleave_granularity = SZ_4K;
 		cxld->target_type = CXL_DECODER_EXPANDER;
+		cxld->commit = mock_decoder_commit;
+		cxld->reset = mock_decoder_reset;
 
 		if (target_count) {
 			rc = device_for_each_child(port->uport, &ctx,

[PATCH v2 26/28] cxl/region: Add region driver boiler plate

[Dan Williams]

The CXL region driver is responsible for routing fully formed CXL
regions to one of libnvdimm, for persistent memory regions, device-dax
for volatile memory regions, or just act as an enumeration placeholder
if the region was setup and configuration locked by platform firmware.
In the platform-firmware-setup case the expectation is that region is
already accounted in the system memory map, i.e. already enabled as
"System RAM".

For now, just attach to CXL regions in the CXL_CONFIG_COMMIT state, and
take no further action.

Given this driver is just a small / simple router, include it in the
core rather than its own module.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-18-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/core.h   |   12 ++++++++++++
 drivers/cxl/core/port.c   |    9 +++++++++
 drivers/cxl/core/region.c |   45 ++++++++++++++++++++++++++++++++++++++++++++-
 drivers/cxl/cxl.h         |    1 +
 4 files changed, 66 insertions(+), 1 deletion(-)

diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index fcf14b8a3c87..391aadf9e7fa 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -13,14 +13,26 @@ extern struct attribute_group cxl_base_attribute_group;
 extern struct device_attribute dev_attr_create_pmem_region;
 extern struct device_attribute dev_attr_delete_region;
 extern struct device_attribute dev_attr_region;
+extern const struct device_type cxl_region_type;
 void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled);
 #define CXL_REGION_ATTR(x) (&dev_attr_##x.attr)
+#define CXL_REGION_TYPE(x) (&cxl_region_type)
 #define SET_CXL_REGION_ATTR(x) (&dev_attr_##x.attr),
+int cxl_region_init(void);
+void cxl_region_exit(void);
 #else
 static inline void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled)
 {
 }
+static inline int cxl_region_init(void)
+{
+	return 0;
+}
+static inline void cxl_region_exit(void)
+{
+}
 #define CXL_REGION_ATTR(x) NULL
+#define CXL_REGION_TYPE(x) NULL
 #define SET_CXL_REGION_ATTR(x)
 #endif
 
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 7ab9a98c5d4f..194003525397 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -51,6 +51,8 @@ static int cxl_device_id(struct device *dev)
 	}
 	if (is_cxl_memdev(dev))
 		return CXL_DEVICE_MEMORY_EXPANDER;
+	if (dev->type == CXL_REGION_TYPE())
+		return CXL_DEVICE_REGION;
 	return 0;
 }
 
@@ -1864,8 +1866,14 @@ static __init int cxl_core_init(void)
 	if (rc)
 		goto err_bus;
 
+	rc = cxl_region_init();
+	if (rc)
+		goto err_region;
+
 	return 0;
 
+err_region:
+	bus_unregister(&cxl_bus_type);
 err_bus:
 	destroy_workqueue(cxl_bus_wq);
 err_wq:
@@ -1875,6 +1883,7 @@ static __init int cxl_core_init(void)
 
 static void cxl_core_exit(void)
 {
+	cxl_region_exit();
 	bus_unregister(&cxl_bus_type);
 	destroy_workqueue(cxl_bus_wq);
 	cxl_memdev_exit();
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index de794344d964..20871bdb6858 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -1444,7 +1444,7 @@ static void cxl_region_release(struct device *dev)
 	kfree(cxlr);
 }
 
-static const struct device_type cxl_region_type = {
+const struct device_type cxl_region_type = {
 	.name = "cxl_region",
 	.release = cxl_region_release,
 	.groups = region_groups
@@ -1644,4 +1644,47 @@ static ssize_t delete_region_store(struct device *dev,
 }
 DEVICE_ATTR_WO(delete_region);
 
+static int cxl_region_probe(struct device *dev)
+{
+	struct cxl_region *cxlr = to_cxl_region(dev);
+	struct cxl_region_params *p = &cxlr->params;
+	int rc;
+
+	rc = down_read_interruptible(&cxl_region_rwsem);
+	if (rc) {
+		dev_dbg(&cxlr->dev, "probe interrupted\n");
+		return rc;
+	}
+
+	if (p->state < CXL_CONFIG_COMMIT) {
+		dev_dbg(&cxlr->dev, "config state: %d\n", p->state);
+		rc = -ENXIO;
+	}
+
+	/*
+	 * From this point on any path that changes the region's state away from
+	 * CXL_CONFIG_COMMIT is also responsible for releasing the driver.
+	 */
+	up_read(&cxl_region_rwsem);
+
+	return rc;
+}
+
+static struct cxl_driver cxl_region_driver = {
+	.name = "cxl_region",
+	.probe = cxl_region_probe,
+	.id = CXL_DEVICE_REGION,
+};
+
+int cxl_region_init(void)
+{
+	return cxl_driver_register(&cxl_region_driver);
+}
+
+void cxl_region_exit(void)
+{
+	cxl_driver_unregister(&cxl_region_driver);
+}
+
 MODULE_IMPORT_NS(CXL);
+MODULE_ALIAS_CXL(CXL_DEVICE_REGION);
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index a51709613c43..9aedd471193a 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -592,6 +592,7 @@ void cxl_driver_unregister(struct cxl_driver *cxl_drv);
 #define CXL_DEVICE_PORT			3
 #define CXL_DEVICE_ROOT			4
 #define CXL_DEVICE_MEMORY_EXPANDER	5
+#define CXL_DEVICE_REGION		6
 
 #define MODULE_ALIAS_CXL(type) MODULE_ALIAS("cxl:t" __stringify(type) "*")
 #define CXL_MODALIAS_FMT "cxl:t%d"

[PATCH v2 27/28] cxl/pmem: Fix offline_nvdimm_bus() to offline by bridge

[Dan Williams]

Be careful to only disable cxl_pmem objects related to a given
cxl_nvdimm_bridge. Otherwise, offline_nvdimm_bus() reaches across CXL
domains and disables more than is expected.

Fixes: 21083f51521f ("cxl/pmem: Register 'pmem' / cxl_nvdimm devices")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/cxl.h  |    1 +
 drivers/cxl/pmem.c |   21 +++++++++++++++++----
 2 files changed, 18 insertions(+), 4 deletions(-)

diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 9aedd471193a..a32093602df9 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -418,6 +418,7 @@ struct cxl_nvdimm_bridge {
 struct cxl_nvdimm {
 	struct device dev;
 	struct cxl_memdev *cxlmd;
+	struct cxl_nvdimm_bridge *bridge;
 };
 
 /**
diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c
index 0aaa70b4e0f7..b271f6e90b91 100644
--- a/drivers/cxl/pmem.c
+++ b/drivers/cxl/pmem.c
@@ -26,7 +26,10 @@ static void clear_exclusive(void *cxlds)
 
 static void unregister_nvdimm(void *nvdimm)
 {
+	struct cxl_nvdimm *cxl_nvd = nvdimm_provider_data(nvdimm);
+
 	nvdimm_delete(nvdimm);
+	cxl_nvd->bridge = NULL;
 }
 
 static int cxl_nvdimm_probe(struct device *dev)
@@ -66,6 +69,7 @@ static int cxl_nvdimm_probe(struct device *dev)
 	}
 
 	dev_set_drvdata(dev, nvdimm);
+	cxl_nvd->bridge = cxl_nvb;
 	rc = devm_add_action_or_reset(dev, unregister_nvdimm, nvdimm);
 out:
 	device_unlock(&cxl_nvb->dev);
@@ -204,15 +208,23 @@ static bool online_nvdimm_bus(struct cxl_nvdimm_bridge *cxl_nvb)
 	return cxl_nvb->nvdimm_bus != NULL;
 }
 
-static int cxl_nvdimm_release_driver(struct device *dev, void *data)
+static int cxl_nvdimm_release_driver(struct device *dev, void *cxl_nvb)
 {
+	struct cxl_nvdimm *cxl_nvd;
+
 	if (!is_cxl_nvdimm(dev))
 		return 0;
+
+	cxl_nvd = to_cxl_nvdimm(dev);
+	if (cxl_nvd->bridge != cxl_nvb)
+		return 0;
+
 	device_release_driver(dev);
 	return 0;
 }
 
-static void offline_nvdimm_bus(struct nvdimm_bus *nvdimm_bus)
+static void offline_nvdimm_bus(struct cxl_nvdimm_bridge *cxl_nvb,
+			       struct nvdimm_bus *nvdimm_bus)
 {
 	if (!nvdimm_bus)
 		return;
@@ -222,7 +234,8 @@ static void offline_nvdimm_bus(struct nvdimm_bus *nvdimm_bus)
 	 * nvdimm_bus_unregister() rips the nvdimm objects out from
 	 * underneath them.
 	 */
-	bus_for_each_dev(&cxl_bus_type, NULL, NULL, cxl_nvdimm_release_driver);
+	bus_for_each_dev(&cxl_bus_type, NULL, cxl_nvb,
+			 cxl_nvdimm_release_driver);
 	nvdimm_bus_unregister(nvdimm_bus);
 }
 
@@ -260,7 +273,7 @@ static void cxl_nvb_update_state(struct work_struct *work)
 
 		dev_dbg(&cxl_nvb->dev, "rescan: %d\n", rc);
 	}
-	offline_nvdimm_bus(victim_bus);
+	offline_nvdimm_bus(cxl_nvb, victim_bus);
 
 	put_device(&cxl_nvb->dev);
 }

[PATCH v2 28/28] cxl/region: Introduce cxl_pmem_region objects

[Dan Williams]

The LIBNVDIMM subsystem is a platform agnostic representation of system
NVDIMM / persistent memory resources. To date, the CXL subsystem's
interaction with LIBNVDIMM has been to register an nvdimm-bridge device
and cxl_nvdimm objects to proxy CXL capabilities into existing LIBNVDIMM
subsystem mechanics.

With regions the approach is the same. Create a new cxl_pmem_region
object to proxy CXL region details into a LIBNVDIMM definition. With
this enabling LIBNVDIMM can partition CXL persistent memory regions with
legacy namespace labels. A follow-on patch will add CXL region label and
CXL namespace label support to persist region configurations across
driver reload / system-reset events.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/cxl/core/core.h      |    3 +
 drivers/cxl/core/pmem.c      |    4 -
 drivers/cxl/core/port.c      |    2 
 drivers/cxl/core/region.c    |  142 +++++++++++++++++++++++++
 drivers/cxl/cxl.h            |   36 ++++++
 drivers/cxl/pmem.c           |  238 ++++++++++++++++++++++++++++++++++++++++++
 drivers/nvdimm/region_devs.c |   28 ++++-
 include/linux/libnvdimm.h    |    5 +
 8 files changed, 446 insertions(+), 12 deletions(-)

diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index 391aadf9e7fa..1d8f87be283f 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -13,11 +13,13 @@ extern struct attribute_group cxl_base_attribute_group;
 extern struct device_attribute dev_attr_create_pmem_region;
 extern struct device_attribute dev_attr_delete_region;
 extern struct device_attribute dev_attr_region;
+extern const struct device_type cxl_pmem_region_type;
 extern const struct device_type cxl_region_type;
 void cxl_decoder_kill_region(struct cxl_endpoint_decoder *cxled);
 #define CXL_REGION_ATTR(x) (&dev_attr_##x.attr)
 #define CXL_REGION_TYPE(x) (&cxl_region_type)
 #define SET_CXL_REGION_ATTR(x) (&dev_attr_##x.attr),
+#define CXL_PMEM_REGION_TYPE(x) (&cxl_pmem_region_type)
 int cxl_region_init(void);
 void cxl_region_exit(void);
 #else
@@ -34,6 +36,7 @@ static inline void cxl_region_exit(void)
 #define CXL_REGION_ATTR(x) NULL
 #define CXL_REGION_TYPE(x) NULL
 #define SET_CXL_REGION_ATTR(x)
+#define CXL_PMEM_REGION_TYPE(x) NULL
 #endif
 
 struct cxl_send_command;
diff --git a/drivers/cxl/core/pmem.c b/drivers/cxl/core/pmem.c
index bec7cfb54ebf..1d12a8206444 100644
--- a/drivers/cxl/core/pmem.c
+++ b/drivers/cxl/core/pmem.c
@@ -62,9 +62,9 @@ static int match_nvdimm_bridge(struct device *dev, void *data)
 	return is_cxl_nvdimm_bridge(dev);
 }
 
-struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct cxl_nvdimm *cxl_nvd)
+struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct device *start)
 {
-	struct cxl_port *port = find_cxl_root(&cxl_nvd->dev);
+	struct cxl_port *port = find_cxl_root(start);
 	struct device *dev;
 
 	if (!port)
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 194003525397..af65491d878b 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -44,6 +44,8 @@ static int cxl_device_id(struct device *dev)
 		return CXL_DEVICE_NVDIMM_BRIDGE;
 	if (dev->type == &cxl_nvdimm_type)
 		return CXL_DEVICE_NVDIMM;
+	if (dev->type == CXL_PMEM_REGION_TYPE())
+		return CXL_DEVICE_PMEM_REGION;
 	if (is_cxl_port(dev)) {
 		if (is_cxl_root(to_cxl_port(dev)))
 			return CXL_DEVICE_ROOT;
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index 20871bdb6858..23d1a34077be 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -1644,6 +1644,139 @@ static ssize_t delete_region_store(struct device *dev,
 }
 DEVICE_ATTR_WO(delete_region);
 
+static void cxl_pmem_region_release(struct device *dev)
+{
+	struct cxl_pmem_region *cxlr_pmem = to_cxl_pmem_region(dev);
+	int i;
+
+	for (i = 0; i < cxlr_pmem->nr_mappings; i++) {
+		struct cxl_memdev *cxlmd = cxlr_pmem->mapping[i].cxlmd;
+
+		put_device(&cxlmd->dev);
+	}
+
+	kfree(cxlr_pmem);
+}
+
+static const struct attribute_group *cxl_pmem_region_attribute_groups[] = {
+	&cxl_base_attribute_group,
+	NULL,
+};
+
+const struct device_type cxl_pmem_region_type = {
+	.name = "cxl_pmem_region",
+	.release = cxl_pmem_region_release,
+	.groups = cxl_pmem_region_attribute_groups,
+};
+
+bool is_cxl_pmem_region(struct device *dev)
+{
+	return dev->type == &cxl_pmem_region_type;
+}
+EXPORT_SYMBOL_NS_GPL(is_cxl_pmem_region, CXL);
+
+struct cxl_pmem_region *to_cxl_pmem_region(struct device *dev)
+{
+	if (dev_WARN_ONCE(dev, !is_cxl_pmem_region(dev),
+			  "not a cxl_pmem_region device\n"))
+		return NULL;
+	return container_of(dev, struct cxl_pmem_region, dev);
+}
+EXPORT_SYMBOL_NS_GPL(to_cxl_pmem_region, CXL);
+
+static struct lock_class_key cxl_pmem_region_key;
+
+static struct cxl_pmem_region *cxl_pmem_region_alloc(struct cxl_region *cxlr)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_pmem_region *cxlr_pmem;
+	struct device *dev;
+	int i;
+
+	down_read(&cxl_region_rwsem);
+	if (p->state != CXL_CONFIG_COMMIT) {
+		cxlr_pmem = ERR_PTR(-ENXIO);
+		goto out;
+	}
+
+	cxlr_pmem = kzalloc(struct_size(cxlr_pmem, mapping, p->nr_targets),
+			    GFP_KERNEL);
+	if (!cxlr_pmem) {
+		cxlr_pmem = ERR_PTR(-ENOMEM);
+		goto out;
+	}
+
+	cxlr_pmem->hpa_range.start = p->res->start;
+	cxlr_pmem->hpa_range.end = p->res->end;
+
+	/* Snapshot the region configuration underneath the cxl_region_rwsem */
+	cxlr_pmem->nr_mappings = p->nr_targets;
+	for (i = 0; i < p->nr_targets; i++) {
+		struct cxl_endpoint_decoder *cxled = p->targets[i];
+		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
+		struct cxl_pmem_region_mapping *m = &cxlr_pmem->mapping[i];
+
+		m->cxlmd = cxlmd;
+		get_device(&cxlmd->dev);
+		m->start = cxled->dpa_res->start;
+		m->size = resource_size(cxled->dpa_res);
+		m->position = i;
+	}
+
+	dev = &cxlr_pmem->dev;
+	cxlr_pmem->cxlr = cxlr;
+	device_initialize(dev);
+	lockdep_set_class(&dev->mutex, &cxl_pmem_region_key);
+	device_set_pm_not_required(dev);
+	dev->parent = &cxlr->dev;
+	dev->bus = &cxl_bus_type;
+	dev->type = &cxl_pmem_region_type;
+out:
+	up_read(&cxl_region_rwsem);
+
+	return cxlr_pmem;
+}
+
+static void cxlr_pmem_unregister(void *dev)
+{
+	device_unregister(dev);
+}
+
+/**
+ * devm_cxl_add_pmem_region() - add a cxl_region-to-nd_region bridge
+ * @cxlr: parent CXL region for this pmem region bridge device
+ *
+ * Return: 0 on success negative error code on failure.
+ */
+static int devm_cxl_add_pmem_region(struct cxl_region *cxlr)
+{
+	struct cxl_pmem_region *cxlr_pmem;
+	struct device *dev;
+	int rc;
+
+	cxlr_pmem = cxl_pmem_region_alloc(cxlr);
+	if (IS_ERR(cxlr_pmem))
+		return PTR_ERR(cxlr_pmem);
+
+	dev = &cxlr_pmem->dev;
+	rc = dev_set_name(dev, "pmem_region%d", cxlr->id);
+	if (rc)
+		goto err;
+
+	rc = device_add(dev);
+	if (rc)
+		goto err;
+
+	dev_dbg(&cxlr->dev, "%s: register %s\n", dev_name(dev->parent),
+		dev_name(dev));
+
+	return devm_add_action_or_reset(&cxlr->dev, cxlr_pmem_unregister, dev);
+
+err:
+	put_device(dev);
+	return rc;
+}
+
 static int cxl_region_probe(struct device *dev)
 {
 	struct cxl_region *cxlr = to_cxl_region(dev);
@@ -1667,7 +1800,14 @@ static int cxl_region_probe(struct device *dev)
 	 */
 	up_read(&cxl_region_rwsem);
 
-	return rc;
+	switch (cxlr->mode) {
+	case CXL_DECODER_PMEM:
+		return devm_cxl_add_pmem_region(cxlr);
+	default:
+		dev_dbg(&cxlr->dev, "unsupported region mode: %d\n",
+			cxlr->mode);
+		return -ENXIO;
+	}
 }
 
 static struct cxl_driver cxl_region_driver = {
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index a32093602df9..8a484acbb32d 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -419,6 +419,25 @@ struct cxl_nvdimm {
 	struct device dev;
 	struct cxl_memdev *cxlmd;
 	struct cxl_nvdimm_bridge *bridge;
+	struct cxl_pmem_region *region;
+};
+
+struct cxl_pmem_region_mapping {
+	struct cxl_memdev *cxlmd;
+	struct cxl_nvdimm *cxl_nvd;
+	u64 start;
+	u64 size;
+	int position;
+};
+
+struct cxl_pmem_region {
+	struct device dev;
+	struct cxl_region *cxlr;
+	struct nd_region *nd_region;
+	struct cxl_nvdimm_bridge *bridge;
+	struct range hpa_range;
+	int nr_mappings;
+	struct cxl_pmem_region_mapping mapping[];
 };
 
 /**
@@ -594,6 +613,7 @@ void cxl_driver_unregister(struct cxl_driver *cxl_drv);
 #define CXL_DEVICE_ROOT			4
 #define CXL_DEVICE_MEMORY_EXPANDER	5
 #define CXL_DEVICE_REGION		6
+#define CXL_DEVICE_PMEM_REGION		7
 
 #define MODULE_ALIAS_CXL(type) MODULE_ALIAS("cxl:t" __stringify(type) "*")
 #define CXL_MODALIAS_FMT "cxl:t%d"
@@ -605,7 +625,21 @@ struct cxl_nvdimm *to_cxl_nvdimm(struct device *dev);
 bool is_cxl_nvdimm(struct device *dev);
 bool is_cxl_nvdimm_bridge(struct device *dev);
 int devm_cxl_add_nvdimm(struct device *host, struct cxl_memdev *cxlmd);
-struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct cxl_nvdimm *cxl_nvd);
+struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct device *dev);
+
+#ifdef CONFIG_CXL_REGION
+bool is_cxl_pmem_region(struct device *dev);
+struct cxl_pmem_region *to_cxl_pmem_region(struct device *dev);
+#else
+static inline bool is_cxl_pmem_region(struct device *dev)
+{
+	return false;
+}
+static inline struct cxl_pmem_region *to_cxl_pmem_region(struct device *dev)
+{
+	return NULL;
+}
+#endif
 
 /*
  * Unit test builds overrides this to __weak, find the 'strong' version
diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c
index b271f6e90b91..e69f99a0747d 100644
--- a/drivers/cxl/pmem.c
+++ b/drivers/cxl/pmem.c
@@ -7,6 +7,7 @@
 #include <linux/ndctl.h>
 #include <linux/async.h>
 #include <linux/slab.h>
+#include <linux/nd.h>
 #include "cxlmem.h"
 #include "cxl.h"
 
@@ -27,6 +28,19 @@ static void clear_exclusive(void *cxlds)
 static void unregister_nvdimm(void *nvdimm)
 {
 	struct cxl_nvdimm *cxl_nvd = nvdimm_provider_data(nvdimm);
+	struct cxl_nvdimm_bridge *cxl_nvb = cxl_nvd->bridge;
+	struct cxl_pmem_region *cxlr_pmem;
+
+	device_lock(&cxl_nvb->dev);
+	cxlr_pmem = cxl_nvd->region;
+	dev_set_drvdata(&cxl_nvd->dev, NULL);
+	cxl_nvd->region = NULL;
+	device_unlock(&cxl_nvb->dev);
+
+	if (cxlr_pmem) {
+		device_release_driver(&cxlr_pmem->dev);
+		put_device(&cxlr_pmem->dev);
+	}
 
 	nvdimm_delete(nvdimm);
 	cxl_nvd->bridge = NULL;
@@ -42,7 +56,7 @@ static int cxl_nvdimm_probe(struct device *dev)
 	struct nvdimm *nvdimm;
 	int rc;
 
-	cxl_nvb = cxl_find_nvdimm_bridge(cxl_nvd);
+	cxl_nvb = cxl_find_nvdimm_bridge(dev);
 	if (!cxl_nvb)
 		return -ENXIO;
 
@@ -223,6 +237,21 @@ static int cxl_nvdimm_release_driver(struct device *dev, void *cxl_nvb)
 	return 0;
 }
 
+static int cxl_pmem_region_release_driver(struct device *dev, void *cxl_nvb)
+{
+	struct cxl_pmem_region *cxlr_pmem;
+
+	if (!is_cxl_pmem_region(dev))
+		return 0;
+
+	cxlr_pmem = to_cxl_pmem_region(dev);
+	if (cxlr_pmem->bridge != cxl_nvb)
+		return 0;
+
+	device_release_driver(dev);
+	return 0;
+}
+
 static void offline_nvdimm_bus(struct cxl_nvdimm_bridge *cxl_nvb,
 			       struct nvdimm_bus *nvdimm_bus)
 {
@@ -234,6 +263,8 @@ static void offline_nvdimm_bus(struct cxl_nvdimm_bridge *cxl_nvb,
 	 * nvdimm_bus_unregister() rips the nvdimm objects out from
 	 * underneath them.
 	 */
+	bus_for_each_dev(&cxl_bus_type, NULL, cxl_nvb,
+			 cxl_pmem_region_release_driver);
 	bus_for_each_dev(&cxl_bus_type, NULL, cxl_nvb,
 			 cxl_nvdimm_release_driver);
 	nvdimm_bus_unregister(nvdimm_bus);
@@ -328,6 +359,203 @@ static struct cxl_driver cxl_nvdimm_bridge_driver = {
 	.id = CXL_DEVICE_NVDIMM_BRIDGE,
 };
 
+static int match_cxl_nvdimm(struct device *dev, void *data)
+{
+	return is_cxl_nvdimm(dev);
+}
+
+static void unregister_nvdimm_region(void *nd_region)
+{
+	struct cxl_nvdimm_bridge *cxl_nvb;
+	struct cxl_pmem_region *cxlr_pmem;
+	int i;
+
+	cxlr_pmem = nd_region_provider_data(nd_region);
+	cxl_nvb = cxlr_pmem->bridge;
+	device_lock(&cxl_nvb->dev);
+	for (i = 0; i < cxlr_pmem->nr_mappings; i++) {
+		struct cxl_pmem_region_mapping *m = &cxlr_pmem->mapping[i];
+		struct cxl_nvdimm *cxl_nvd = m->cxl_nvd;
+
+		if (cxl_nvd->region) {
+			put_device(&cxlr_pmem->dev);
+			cxl_nvd->region = NULL;
+		}
+	}
+	device_unlock(&cxl_nvb->dev);
+
+	nvdimm_region_delete(nd_region);
+}
+
+static void cxlr_pmem_remove_resource(void *res)
+{
+	remove_resource(res);
+}
+
+struct cxl_pmem_region_info {
+	u64 offset;
+	u64 serial;
+};
+
+static int cxl_pmem_region_probe(struct device *dev)
+{
+	struct nd_mapping_desc mappings[CXL_DECODER_MAX_INTERLEAVE];
+	struct cxl_pmem_region *cxlr_pmem = to_cxl_pmem_region(dev);
+	struct cxl_region *cxlr = cxlr_pmem->cxlr;
+	struct cxl_pmem_region_info *info = NULL;
+	struct cxl_nvdimm_bridge *cxl_nvb;
+	struct nd_interleave_set *nd_set;
+	struct nd_region_desc ndr_desc;
+	struct cxl_nvdimm *cxl_nvd;
+	struct nvdimm *nvdimm;
+	struct resource *res;
+	int rc, i = 0;
+
+	cxl_nvb = cxl_find_nvdimm_bridge(&cxlr_pmem->mapping[0].cxlmd->dev);
+	if (!cxl_nvb) {
+		dev_dbg(dev, "bridge not found\n");
+		return -ENXIO;
+	}
+	cxlr_pmem->bridge = cxl_nvb;
+
+	device_lock(&cxl_nvb->dev);
+	if (!cxl_nvb->nvdimm_bus) {
+		dev_dbg(dev, "nvdimm bus not found\n");
+		rc = -ENXIO;
+		goto err;
+	}
+
+	memset(&mappings, 0, sizeof(mappings));
+	memset(&ndr_desc, 0, sizeof(ndr_desc));
+
+	res = devm_kzalloc(dev, sizeof(*res), GFP_KERNEL);
+	if (!res) {
+		rc = -ENOMEM;
+		goto err;
+	}
+
+	res->name = "Persistent Memory";
+	res->start = cxlr_pmem->hpa_range.start;
+	res->end = cxlr_pmem->hpa_range.end;
+	res->flags = IORESOURCE_MEM;
+	res->desc = IORES_DESC_PERSISTENT_MEMORY;
+
+	rc = insert_resource(&iomem_resource, res);
+	if (rc)
+		goto err;
+
+	rc = devm_add_action_or_reset(dev, cxlr_pmem_remove_resource, res);
+	if (rc)
+		goto err;
+
+	ndr_desc.res = res;
+	ndr_desc.provider_data = cxlr_pmem;
+
+	ndr_desc.numa_node = memory_add_physaddr_to_nid(res->start);
+	ndr_desc.target_node = phys_to_target_node(res->start);
+	if (ndr_desc.target_node == NUMA_NO_NODE) {
+		ndr_desc.target_node = ndr_desc.numa_node;
+		dev_dbg(&cxlr->dev, "changing target node from %d to %d",
+			NUMA_NO_NODE, ndr_desc.target_node);
+	}
+
+	nd_set = devm_kzalloc(dev, sizeof(*nd_set), GFP_KERNEL);
+	if (!nd_set) {
+		rc = -ENOMEM;
+		goto err;
+	}
+
+	ndr_desc.memregion = cxlr->id;
+	set_bit(ND_REGION_CXL, &ndr_desc.flags);
+	set_bit(ND_REGION_PERSIST_MEMCTRL, &ndr_desc.flags);
+
+	info = kmalloc_array(cxlr_pmem->nr_mappings, sizeof(*info), GFP_KERNEL);
+	if (!info) {
+		rc = -ENOMEM;
+		goto err;
+	}
+
+	for (i = 0; i < cxlr_pmem->nr_mappings; i++) {
+		struct cxl_pmem_region_mapping *m = &cxlr_pmem->mapping[i];
+		struct cxl_memdev *cxlmd = m->cxlmd;
+		struct cxl_dev_state *cxlds = cxlmd->cxlds;
+		struct device *d;
+
+		d = device_find_child(&cxlmd->dev, NULL, match_cxl_nvdimm);
+		if (!d) {
+			dev_dbg(dev, "[%d]: %s: no cxl_nvdimm found\n", i,
+				dev_name(&cxlmd->dev));
+			rc = -ENODEV;
+			goto err;
+		}
+
+		/* safe to drop ref now with bridge lock held */
+		put_device(d);
+
+		cxl_nvd = to_cxl_nvdimm(d);
+		nvdimm = dev_get_drvdata(&cxl_nvd->dev);
+		if (!nvdimm) {
+			dev_dbg(dev, "[%d]: %s: no nvdimm found\n", i,
+				dev_name(&cxlmd->dev));
+			rc = -ENODEV;
+			goto err;
+		}
+		cxl_nvd->region = cxlr_pmem;
+		get_device(&cxlr_pmem->dev);
+		m->cxl_nvd = cxl_nvd;
+		mappings[i] = (struct nd_mapping_desc) {
+			.nvdimm = nvdimm,
+			.start = m->start,
+			.size = m->size,
+			.position = i,
+		};
+		info[i].offset = m->start;
+		info[i].serial = cxlds->serial;
+	}
+	ndr_desc.num_mappings = cxlr_pmem->nr_mappings;
+	ndr_desc.mapping = mappings;
+
+	/*
+	 * TODO enable CXL labels which skip the need for 'interleave-set cookie'
+	 */
+	nd_set->cookie1 =
+		nd_fletcher64(info, sizeof(*info) * cxlr_pmem->nr_mappings, 0);
+	nd_set->cookie2 = nd_set->cookie1;
+	ndr_desc.nd_set = nd_set;
+
+	cxlr_pmem->nd_region =
+		nvdimm_pmem_region_create(cxl_nvb->nvdimm_bus, &ndr_desc);
+	if (IS_ERR(cxlr_pmem->nd_region)) {
+		rc = PTR_ERR(cxlr_pmem->nd_region);
+		goto err;
+	}
+
+	rc = devm_add_action_or_reset(dev, unregister_nvdimm_region,
+				      cxlr_pmem->nd_region);
+out:
+	kfree(info);
+	device_unlock(&cxl_nvb->dev);
+	put_device(&cxl_nvb->dev);
+
+	return rc;
+
+err:
+	dev_dbg(dev, "failed to create nvdimm region\n");
+	for (i--; i >= 0; i--) {
+		nvdimm = mappings[i].nvdimm;
+		cxl_nvd = nvdimm_provider_data(nvdimm);
+		put_device(&cxl_nvd->region->dev);
+		cxl_nvd->region = NULL;
+	}
+	goto out;
+}
+
+static struct cxl_driver cxl_pmem_region_driver = {
+	.name = "cxl_pmem_region",
+	.probe = cxl_pmem_region_probe,
+	.id = CXL_DEVICE_PMEM_REGION,
+};
+
 /*
  * Return all bridges to the CXL_NVB_NEW state to invalidate any
  * ->state_work referring to the now destroyed cxl_pmem_wq.
@@ -372,8 +600,14 @@ static __init int cxl_pmem_init(void)
 	if (rc)
 		goto err_nvdimm;
 
+	rc = cxl_driver_register(&cxl_pmem_region_driver);
+	if (rc)
+		goto err_region;
+
 	return 0;
 
+err_region:
+	cxl_driver_unregister(&cxl_nvdimm_driver);
 err_nvdimm:
 	cxl_driver_unregister(&cxl_nvdimm_bridge_driver);
 err_bridge:
@@ -383,6 +617,7 @@ static __init int cxl_pmem_init(void)
 
 static __exit void cxl_pmem_exit(void)
 {
+	cxl_driver_unregister(&cxl_pmem_region_driver);
 	cxl_driver_unregister(&cxl_nvdimm_driver);
 	cxl_driver_unregister(&cxl_nvdimm_bridge_driver);
 	destroy_cxl_pmem_wq();
@@ -394,3 +629,4 @@ module_exit(cxl_pmem_exit);
 MODULE_IMPORT_NS(CXL);
 MODULE_ALIAS_CXL(CXL_DEVICE_NVDIMM_BRIDGE);
 MODULE_ALIAS_CXL(CXL_DEVICE_NVDIMM);
+MODULE_ALIAS_CXL(CXL_DEVICE_PMEM_REGION);
diff --git a/drivers/nvdimm/region_devs.c b/drivers/nvdimm/region_devs.c
index d976260eca7a..473a71bbd9c9 100644
--- a/drivers/nvdimm/region_devs.c
+++ b/drivers/nvdimm/region_devs.c
@@ -133,7 +133,8 @@ static void nd_region_release(struct device *dev)
 		put_device(&nvdimm->dev);
 	}
 	free_percpu(nd_region->lane);
-	memregion_free(nd_region->id);
+	if (!test_bit(ND_REGION_CXL, &nd_region->flags))
+		memregion_free(nd_region->id);
 	kfree(nd_region);
 }
 
@@ -982,9 +983,14 @@ static struct nd_region *nd_region_create(struct nvdimm_bus *nvdimm_bus,
 
 	if (!nd_region)
 		return NULL;
-	nd_region->id = memregion_alloc(GFP_KERNEL);
-	if (nd_region->id < 0)
-		goto err_id;
+	/* CXL pre-assigns memregion ids before creating nvdimm regions */
+	if (test_bit(ND_REGION_CXL, &ndr_desc->flags)) {
+		nd_region->id = ndr_desc->memregion;
+	} else {
+		nd_region->id = memregion_alloc(GFP_KERNEL);
+		if (nd_region->id < 0)
+			goto err_id;
+	}
 
 	nd_region->lane = alloc_percpu(struct nd_percpu_lane);
 	if (!nd_region->lane)
@@ -1043,9 +1049,10 @@ static struct nd_region *nd_region_create(struct nvdimm_bus *nvdimm_bus,
 
 	return nd_region;
 
- err_percpu:
-	memregion_free(nd_region->id);
- err_id:
+err_percpu:
+	if (!test_bit(ND_REGION_CXL, &ndr_desc->flags))
+		memregion_free(nd_region->id);
+err_id:
 	kfree(nd_region);
 	return NULL;
 }
@@ -1068,6 +1075,13 @@ struct nd_region *nvdimm_volatile_region_create(struct nvdimm_bus *nvdimm_bus,
 }
 EXPORT_SYMBOL_GPL(nvdimm_volatile_region_create);
 
+void nvdimm_region_delete(struct nd_region *nd_region)
+{
+	if (nd_region)
+		nd_device_unregister(&nd_region->dev, ND_SYNC);
+}
+EXPORT_SYMBOL_GPL(nvdimm_region_delete);
+
 int nvdimm_flush(struct nd_region *nd_region, struct bio *bio)
 {
 	int rc = 0;
diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h
index 0d61e07b6827..c74acfa1a3fe 100644
--- a/include/linux/libnvdimm.h
+++ b/include/linux/libnvdimm.h
@@ -59,6 +59,9 @@ enum {
 	/* Platform provides asynchronous flush mechanism */
 	ND_REGION_ASYNC = 3,
 
+	/* Region was created by CXL subsystem */
+	ND_REGION_CXL = 4,
+
 	/* mark newly adjusted resources as requiring a label update */
 	DPA_RESOURCE_ADJUSTED = 1 << 0,
 };
@@ -122,6 +125,7 @@ struct nd_region_desc {
 	int numa_node;
 	int target_node;
 	unsigned long flags;
+	int memregion;
 	struct device_node *of_node;
 	int (*flush)(struct nd_region *nd_region, struct bio *bio);
 };
@@ -259,6 +263,7 @@ static inline struct nvdimm *nvdimm_create(struct nvdimm_bus *nvdimm_bus,
 			cmd_mask, num_flush, flush_wpq, NULL, NULL, NULL);
 }
 void nvdimm_delete(struct nvdimm *nvdimm);
+void nvdimm_region_delete(struct nd_region *nd_region);
 
 const struct nd_cmd_desc *nd_cmd_dimm_desc(int cmd);
 const struct nd_cmd_desc *nd_cmd_bus_desc(int cmd);

Re: [PATCH v2 03/28] cxl/acpi: Track CXL resources in iomem_resource

[Greg Kroah-Hartman]

On Thu, Jul 14, 2022 at 05:00:59PM -0700, Dan Williams wrote:
> Recall that CXL capable address ranges, on ACPI platforms, are published
> in the CEDT.CFMWS (CXL Early Discovery Table: CXL Fixed Memory Window
> Structures). These windows represent both the actively mapped capacity
> and the potential address space that can be dynamically assigned to a
> new CXL decode configuration (region / interleave-set).
> 
> CXL endpoints like DDR DIMMs can be mapped at any physical address
> including 0 and legacy ranges.
> 
> There is an expectation and requirement that the /proc/iomem interface
> and the iomem_resource tree in the kernel reflect the full set of
> platform address ranges. I.e. that every address range that platform
> firmware and bus drivers enumerate be reflected as an iomem_resource
> entry. The hard requirement to do this for CXL arises from the fact that
> facilities like CONFIG_DEVICE_PRIVATE expect to be able to treat empty
> iomem_resource ranges as free for software to use as proxy address
> space. Without CXL publishing its potential address ranges in
> iomem_resource, the CONFIG_DEVICE_PRIVATE mechanism may inadvertently
> steal capacity reserved for runtime provisioning of new CXL regions.
> 
> So, iomem_resource needs to know about both active and potential CXL
> resource ranges. The active CXL resources might already be reflected in
> iomem_resource as "System RAM". insert_resource_expand_to_fit() handles
> re-parenting "System RAM" underneath a CXL window.
> 
> The "_expand_to_fit()" behavior handles cases where a CXL window is not
> a strict superset of an existing entry in the iomem_resource tree. The
> "_expand_to_fit()" behavior is acceptable from the perspective of
> resource allocation. The expansion happens because a conflicting
> resource range is already populated, which means the resource boundary
> expansion does not result in any additional free CXL address space being
> made available. CXL address space allocation is always bounded by the
> orginal unexpanded address range.
> 
> However, the potential for expansion does mean that something like
> walk_iomem_res_desc(IORES_DESC_CXL...) can only return fuzzy answers on
> corner case platforms that cause the resource tree to expand a CXL
> window resource over a range that is not decoded by CXL. This would be
> an odd platform configuration, but if it becomes a problem in practice
> the CXL subsytem could just publish an API that returns definitive
> answers.
> 
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: Jason Gunthorpe <jgg@nvidia.com>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>


Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Re: [PATCH v2 01/28] Documentation/cxl: Use a double line break between entries

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:00:47 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Make it easier to read delineations between the "Description" line
> break, new paragraph line breaks, and new entries.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

I'm not that fussed either way on this (indentation was enough for my
brain), but this is at least consistent and I can't see it breaking
the docs build or similar.  Just hope no one decides this is a 'fix' they
want to propagate to all the other ABI docs!

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> ---
>  Documentation/ABI/testing/sysfs-bus-cxl |   15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 
> diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
> index 1fd5984b6158..16d9ffa94bbd 100644
> --- a/Documentation/ABI/testing/sysfs-bus-cxl
> +++ b/Documentation/ABI/testing/sysfs-bus-cxl
> @@ -7,6 +7,7 @@ Description:
>  		all descendant memdevs for unbind. Writing '1' to this attribute
>  		flushes that work.
>  
> +
>  What:		/sys/bus/cxl/devices/memX/firmware_version
>  Date:		December, 2020
>  KernelVersion:	v5.12
> @@ -16,6 +17,7 @@ Description:
>  		Memory Device Output Payload in the CXL-2.0
>  		specification.
>  
> +
>  What:		/sys/bus/cxl/devices/memX/ram/size
>  Date:		December, 2020
>  KernelVersion:	v5.12
> @@ -25,6 +27,7 @@ Description:
>  		identically named field in the Identify Memory Device Output
>  		Payload in the CXL-2.0 specification.
>  
> +
>  What:		/sys/bus/cxl/devices/memX/pmem/size
>  Date:		December, 2020
>  KernelVersion:	v5.12
> @@ -34,6 +37,7 @@ Description:
>  		identically named field in the Identify Memory Device Output
>  		Payload in the CXL-2.0 specification.
>  
> +
>  What:		/sys/bus/cxl/devices/memX/serial
>  Date:		January, 2022
>  KernelVersion:	v5.18
> @@ -43,6 +47,7 @@ Description:
>  		capability. Mandatory for CXL devices, see CXL 2.0 8.1.12.2
>  		Memory Device PCIe Capabilities and Extended Capabilities.
>  
> +
>  What:		/sys/bus/cxl/devices/memX/numa_node
>  Date:		January, 2022
>  KernelVersion:	v5.18
> @@ -52,6 +57,7 @@ Description:
>  		host PCI device for this memory device, emit the CPU node
>  		affinity for this device.
>  
> +
>  What:		/sys/bus/cxl/devices/*/devtype
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -61,6 +67,7 @@ Description:
>  		mirrors the same value communicated in the DEVTYPE environment
>  		variable for uevents for devices on the "cxl" bus.
>  
> +
>  What:		/sys/bus/cxl/devices/*/modalias
>  Date:		December, 2021
>  KernelVersion:	v5.18
> @@ -70,6 +77,7 @@ Description:
>  		mirrors the same value communicated in the MODALIAS environment
>  		variable for uevents for devices on the "cxl" bus.
>  
> +
>  What:		/sys/bus/cxl/devices/portX/uport
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -81,6 +89,7 @@ Description:
>  		the CXL portX object to the device that published the CXL port
>  		capability.
>  
> +
>  What:		/sys/bus/cxl/devices/portX/dportY
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -94,6 +103,7 @@ Description:
>  		integer reflects the hardware port unique-id used in the
>  		hardware decoder target list.
>  
> +
>  What:		/sys/bus/cxl/devices/decoderX.Y
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -106,6 +116,7 @@ Description:
>  		cxl_port container of this decoder, and 'Y' represents the
>  		instance id of a given decoder resource.
>  
> +
>  What:		/sys/bus/cxl/devices/decoderX.Y/{start,size}
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -120,6 +131,7 @@ Description:
>  		and dynamically updates based on the active memory regions in
>  		that address space.
>  
> +
>  What:		/sys/bus/cxl/devices/decoderX.Y/locked
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -132,6 +144,7 @@ Description:
>  		secondary bus reset, of the PCIe bridge that provides the bus
>  		for this decoders uport, unlocks / resets the decoder.
>  
> +
>  What:		/sys/bus/cxl/devices/decoderX.Y/target_list
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -142,6 +155,7 @@ Description:
>  		configured interleave order of the decoder's dport instances.
>  		Each entry in the list is a dport id.
>  
> +
>  What:		/sys/bus/cxl/devices/decoderX.Y/cap_{pmem,ram,type2,type3}
>  Date:		June, 2021
>  KernelVersion:	v5.14
> @@ -154,6 +168,7 @@ Description:
>  		memory, volatile memory, accelerator memory, and / or expander
>  		memory may be mapped behind this decoder's memory window.
>  
> +
>  What:		/sys/bus/cxl/devices/decoderX.Y/target_type
>  Date:		June, 2021
>  KernelVersion:	v5.14
> 

Re: [PATCH v2 02/28] cxl/core: Define a 'struct cxl_switch_decoder'

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:00:53 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Currently 'struct cxl_decoder' contains the superset of attributes
> needed for all decoder types. Before more type-specific attributes are
> added to the common definition, reorganize 'struct cxl_decoder' into type
> specific objects.
> 
> This patch, the first of three, factors out a cxl_switch_decoder type.
> See the new kdoc for what a 'struct cxl_switch_decoder' represents in a
> CXL topology.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 03/28] cxl/acpi: Track CXL resources in iomem_resource

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:00:59 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Recall that CXL capable address ranges, on ACPI platforms, are published
> in the CEDT.CFMWS (CXL Early Discovery Table: CXL Fixed Memory Window
> Structures). These windows represent both the actively mapped capacity
> and the potential address space that can be dynamically assigned to a
> new CXL decode configuration (region / interleave-set).
> 
> CXL endpoints like DDR DIMMs can be mapped at any physical address
> including 0 and legacy ranges.
> 
> There is an expectation and requirement that the /proc/iomem interface
> and the iomem_resource tree in the kernel reflect the full set of
> platform address ranges. I.e. that every address range that platform
> firmware and bus drivers enumerate be reflected as an iomem_resource
> entry. The hard requirement to do this for CXL arises from the fact that
> facilities like CONFIG_DEVICE_PRIVATE expect to be able to treat empty
> iomem_resource ranges as free for software to use as proxy address
> space. Without CXL publishing its potential address ranges in
> iomem_resource, the CONFIG_DEVICE_PRIVATE mechanism may inadvertently
> steal capacity reserved for runtime provisioning of new CXL regions.
> 
> So, iomem_resource needs to know about both active and potential CXL
> resource ranges. The active CXL resources might already be reflected in
> iomem_resource as "System RAM". insert_resource_expand_to_fit() handles
> re-parenting "System RAM" underneath a CXL window.
> 
> The "_expand_to_fit()" behavior handles cases where a CXL window is not
> a strict superset of an existing entry in the iomem_resource tree. The
> "_expand_to_fit()" behavior is acceptable from the perspective of
> resource allocation. The expansion happens because a conflicting
> resource range is already populated, which means the resource boundary
> expansion does not result in any additional free CXL address space being
> made available. CXL address space allocation is always bounded by the
> orginal unexpanded address range.
> 
> However, the potential for expansion does mean that something like
> walk_iomem_res_desc(IORES_DESC_CXL...) can only return fuzzy answers on
> corner case platforms that cause the resource tree to expand a CXL
> window resource over a range that is not decoded by CXL. This would be
> an odd platform configuration, but if it becomes a problem in practice
> the CXL subsytem could just publish an API that returns definitive
> answers.
> 
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: Jason Gunthorpe <jgg@nvidia.com>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

I wish this was a bit simpler, (particularly the complex relationship
between the places res is added vs single cleanup location) but having
stared at it for a while I can't figure out a way that works.. *defeated*
:)
 

Re: [PATCH v2 04/28] cxl/core: Define a 'struct cxl_root_decoder'

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:05 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Previously the target routing specifics of switch decoders were factored
> out of 'struct cxl_decoder' into 'struct cxl_switch_decoder'.
> 
> This patch, 2 of 3, adds a 'struct cxl_root_decoder' as a superset of a
> switch decoder that also track the associated CXL window platform
> resource.
> 
> Note that the reason the resource for a given root decoder needs to be
> looked up after the fact (i.e. after cxl_parse_cfmws() and
> add_cxl_resource()) is because add_cxl_resource() may have merged CXL
> windows in order to keep them at the top of the resource tree / decode
> hierarchy.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 05/28] cxl/core: Define a 'struct cxl_endpoint_decoder'

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:10 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Previously the target routing specifics of switch decoders and platform
> CXL window resource tracking of root decoders were factored out of
> 'struct cxl_decoder'. While switch decoders translate from SPA to
> downstream ports, endpoint decoders translate from SPA to DPA.
> 
> This patch, 3 of 3, adds a 'struct cxl_endpoint_decoder' that tracks an
> endpoint-specific Device Physical Address (DPA) resource. For now this
> just defines ->dpa_res, a follow-on patch will handle requesting DPA
> resource ranges from a device-DPA resource tree.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 06/28] cxl/hdm: Enumerate allocated DPA

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:16 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> In preparation for provisioning CXL regions, add accounting for the DPA
> space consumed by existing regions / decoders. Recall, a CXL region is a
> memory range comprised from one or more endpoint devices contributing a
> mapping of their DPA into HPA space through a decoder.
> 
> Record the DPA ranges covered by committed decoders at initial probe of
> endpoint ports relative to a per-device resource tree of the DPA type
> (pmem or volatile-ram).
> 
> The cxl_dpa_rwsem semaphore is introduced to globally synchronize DPA
> state across all endpoints and their decoders at once. The vast majority
> of DPA operations are reads as region creation is expected to be as rare
> as disk partitioning and volume creation. The device_lock() for this
> synchronization is specifically avoided for concern of entangling with
> sysfs attribute removal.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com> 

One trivial ordering question inline. I'm not that bothered whether you
do anything about it though as it's all very local.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> ---
>  drivers/cxl/core/hdm.c |  143 ++++++++++++++++++++++++++++++++++++++++++++----
>  drivers/cxl/cxl.h      |    2 +
>  drivers/cxl/cxlmem.h   |   13 ++++
>  3 files changed, 147 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> index 650363d5272f..d4c17325001b 100644
> --- a/drivers/cxl/core/hdm.c
> +++ b/drivers/cxl/core/hdm.c
> @@ -153,10 +153,105 @@ void cxl_dpa_debug(struct seq_file *file, struct cxl_dev_state *cxlds)
>  }
>  EXPORT_SYMBOL_NS_GPL(cxl_dpa_debug, CXL);
>  
> +/*
> + * Must be called in a context that synchronizes against this decoder's
> + * port ->remove() callback (like an endpoint decoder sysfs attribute)
> + */
> +static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
> +{
> +	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> +	struct cxl_dev_state *cxlds = cxlmd->cxlds;
> +	struct resource *res = cxled->dpa_res;
> +
> +	lockdep_assert_held_write(&cxl_dpa_rwsem);
> +
> +	if (cxled->skip)
> +		__release_region(&cxlds->dpa_res, res->start - cxled->skip,
> +				 cxled->skip);
> +	cxled->skip = 0;
> +	__release_region(&cxlds->dpa_res, res->start, resource_size(res));

Minor but I think the ordering in here is unnecessarily not the opposite
of what is going on in __cxl_dpa_reserve()  Should be releasing the
actual rs first, then releasing the skip.

> +	cxled->dpa_res = NULL;
> +}
> +
> +static void cxl_dpa_release(void *cxled)
> +{
> +	down_write(&cxl_dpa_rwsem);
> +	__cxl_dpa_release(cxled);
> +	up_write(&cxl_dpa_rwsem);
> +}
> +
> +static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
> +			     resource_size_t base, resource_size_t len,
> +			     resource_size_t skipped)
> +{
> +	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> +	struct cxl_port *port = cxled_to_port(cxled);
> +	struct cxl_dev_state *cxlds = cxlmd->cxlds;
> +	struct device *dev = &port->dev;
> +	struct resource *res;
> +
> +	lockdep_assert_held_write(&cxl_dpa_rwsem);
> +
> +	if (!len)
> +		return 0;
> +
> +	if (cxled->dpa_res) {
> +		dev_dbg(dev, "decoder%d.%d: existing allocation %pr assigned\n",
> +			port->id, cxled->cxld.id, cxled->dpa_res);
> +		return -EBUSY;
> +	}
> +
> +	if (skipped) {
> +		res = __request_region(&cxlds->dpa_res, base - skipped, skipped,
> +				       dev_name(&cxled->cxld.dev), 0);
> +		if (!res) {
> +			dev_dbg(dev,
> +				"decoder%d.%d: failed to reserve skipped space\n",
> +				port->id, cxled->cxld.id);
> +			return -EBUSY;
> +		}
> +	}
> +	res = __request_region(&cxlds->dpa_res, base, len,
> +			       dev_name(&cxled->cxld.dev), 0);
> +	if (!res) {
> +		dev_dbg(dev, "decoder%d.%d: failed to reserve allocation\n",
> +			port->id, cxled->cxld.id);
> +		if (skipped)
> +			__release_region(&cxlds->dpa_res, base - skipped,
> +					 skipped);
> +		return -EBUSY;
> +	}
> +	cxled->dpa_res = res;
> +	cxled->skip = skipped;
> +
> +	return 0;
> +}
> +

...

Re: [PATCH v2 08/28] cxl/hdm: Track next decoder to allocate

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:28 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> The CXL specification enforces that endpoint decoders are committed in
> hw instance id order. In preparation for adding dynamic DPA allocation,
> record the hw instance id in endpoint decoders, and enforce allocations
> to occur in hw instance id order.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Lifting from v1 thread to ease picking it up.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> ---
>  drivers/cxl/core/hdm.c  |   15 +++++++++++++++
>  drivers/cxl/core/port.c |    1 +
>  drivers/cxl/cxl.h       |    2 ++
>  3 files changed, 18 insertions(+)
> 
> diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> index acd46b0d69c6..582f48141767 100644
> --- a/drivers/cxl/core/hdm.c
> +++ b/drivers/cxl/core/hdm.c
> @@ -160,6 +160,7 @@ EXPORT_SYMBOL_NS_GPL(cxl_dpa_debug, CXL);
>  static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
>  {
>  	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> +	struct cxl_port *port = cxled_to_port(cxled);
>  	struct cxl_dev_state *cxlds = cxlmd->cxlds;
>  	struct resource *res = cxled->dpa_res;
>  
> @@ -171,6 +172,7 @@ static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
>  	cxled->skip = 0;
>  	__release_region(&cxlds->dpa_res, res->start, resource_size(res));
>  	cxled->dpa_res = NULL;
> +	port->hdm_end--;
>  }
>  
>  static void cxl_dpa_release(void *cxled)
> @@ -201,6 +203,18 @@ static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
>  		return -EBUSY;
>  	}
>  
> +	if (port->hdm_end + 1 != cxled->cxld.id) {
> +		/*
> +		 * Assumes alloc and commit order is always in hardware instance
> +		 * order per expectations from 8.2.5.12.20 Committing Decoder
> +		 * Programming that enforce decoder[m] committed before
> +		 * decoder[m+1] commit start.
> +		 */
> +		dev_dbg(dev, "decoder%d.%d: expected decoder%d.%d\n", port->id,
> +			cxled->cxld.id, port->id, port->hdm_end + 1);
> +		return -EBUSY;
> +	}
> +
>  	if (skipped) {
>  		res = __request_region(&cxlds->dpa_res, base - skipped, skipped,
>  				       dev_name(&cxled->cxld.dev), 0);
> @@ -233,6 +247,7 @@ static int __cxl_dpa_reserve(struct cxl_endpoint_decoder *cxled,
>  			cxled->cxld.id, cxled->dpa_res);
>  		cxled->mode = CXL_DECODER_MIXED;
>  	}
> +	port->hdm_end++;
>  
>  	return 0;
>  }
> diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
> index 0ac5dcd612e0..109611318760 100644
> --- a/drivers/cxl/core/port.c
> +++ b/drivers/cxl/core/port.c
> @@ -502,6 +502,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
>  
>  	port->component_reg_phys = component_reg_phys;
>  	ida_init(&port->decoder_ida);
> +	port->hdm_end = -1;
>  	INIT_LIST_HEAD(&port->dports);
>  	INIT_LIST_HEAD(&port->endpoints);
>  
> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
> index 3e7363dde80f..70cd24e4f3ce 100644
> --- a/drivers/cxl/cxl.h
> +++ b/drivers/cxl/cxl.h
> @@ -333,6 +333,7 @@ struct cxl_nvdimm {
>   * @dports: cxl_dport instances referenced by decoders
>   * @endpoints: cxl_ep instances, endpoints that are a descendant of this port
>   * @decoder_ida: allocator for decoder ids
> + * @hdm_end: track last allocated HDM decoder instance for allocation ordering
>   * @component_reg_phys: component register capability base address (optional)
>   * @dead: last ep has been removed, force port re-creation
>   * @depth: How deep this port is relative to the root. depth 0 is the root.
> @@ -345,6 +346,7 @@ struct cxl_port {
>  	struct list_head dports;
>  	struct list_head endpoints;
>  	struct ida decoder_ida;
> +	int hdm_end;
>  	resource_size_t component_reg_phys;
>  	bool dead;
>  	unsigned int depth;
> 

Re: [PATCH v2 09/28] cxl/hdm: Add support for allocating DPA to an endpoint decoder

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:34 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> The region provisioning flow will roughly follow a sequence of:
> 
> 1/ Allocate DPA to a set of decoders
> 
> 2/ Allocate HPA to a region
> 
> 3/ Associate decoders with a region and validate that the DPA allocations
>    and topologies match the parameters of the region.
> 
> For now, this change (step 1) arranges for DPA capacity to be allocated
> and deleted from non-committed decoders based on the decoder's mode /
> partition selection. Capacity is allocated from the lowest DPA in the
> partition and any 'pmem' allocation blocks out all remaining ram
> capacity in its 'skip' setting. DPA allocations are enforced in decoder
> instance order. I.e. decoder N + 1 always starts at a higher DPA than
> instance N, and deleting allocations must proceed from the
> highest-instance allocated decoder to the lowest.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 10/28] cxl/port: Record dport in endpoint references

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:39 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Recall that the primary role of the cxl_mem driver is to probe if the
> given endpoint is connected to a CXL port topology. In that process it
> walks its device ancestry to its PCI root port. If that root port is
> also a CXL root port then the probe process adds cxl_port object
> instances at switch in the path between to the root and the endpoint. As
> those cxl_port instances are added, or if a previous enumeration
> attempt already created the port, a 'struct cxl_ep' instance is
> registered with that port to track the endpoints interested in that
> port.
> 
> At the time the cxl_ep is registered the downstream egress path from the
> port to the endpoint is known. Take the opportunity to record that
> information as it will be needed for dynamic programming of decoder
> targets during region provisioning.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
You answered my queries on previous version, so I'm happy with this.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> ---
>  drivers/cxl/core/port.c |   52 ++++++++++++++++++++++++++++++++---------------
>  drivers/cxl/cxl.h       |    2 ++
>  2 files changed, 37 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
> index fdc1be7db917..a8d350361548 100644
> --- a/drivers/cxl/core/port.c
> +++ b/drivers/cxl/core/port.c
> @@ -882,8 +882,9 @@ static struct cxl_ep *find_ep(struct cxl_port *port, struct device *ep_dev)
>  	return NULL;
>  }
>  
> -static int add_ep(struct cxl_port *port, struct cxl_ep *new)
> +static int add_ep(struct cxl_ep *new)
>  {
> +	struct cxl_port *port = new->dport->port;
>  	struct cxl_ep *dup;
>  
>  	device_lock(&port->dev);
> @@ -901,14 +902,14 @@ static int add_ep(struct cxl_port *port, struct cxl_ep *new)
>  
>  /**
>   * cxl_add_ep - register an endpoint's interest in a port
> - * @port: a port in the endpoint's topology ancestry
> + * @dport: the dport that routes to @ep_dev
>   * @ep_dev: device representing the endpoint
>   *
>   * Intermediate CXL ports are scanned based on the arrival of endpoints.
>   * When those endpoints depart the port can be destroyed once all
>   * endpoints that care about that port have been removed.
>   */
> -static int cxl_add_ep(struct cxl_port *port, struct device *ep_dev)
> +static int cxl_add_ep(struct cxl_dport *dport, struct device *ep_dev)
>  {
>  	struct cxl_ep *ep;
>  	int rc;
> @@ -919,8 +920,9 @@ static int cxl_add_ep(struct cxl_port *port, struct device *ep_dev)
>  
>  	INIT_LIST_HEAD(&ep->list);
>  	ep->ep = get_device(ep_dev);
> +	ep->dport = dport;
>  
> -	rc = add_ep(port, ep);
> +	rc = add_ep(ep);
>  	if (rc)
>  		cxl_ep_release(ep);
>  	return rc;
> @@ -929,11 +931,13 @@ static int cxl_add_ep(struct cxl_port *port, struct device *ep_dev)
>  struct cxl_find_port_ctx {
>  	const struct device *dport_dev;
>  	const struct cxl_port *parent_port;
> +	struct cxl_dport **dport;
>  };
>  
>  static int match_port_by_dport(struct device *dev, const void *data)
>  {
>  	const struct cxl_find_port_ctx *ctx = data;
> +	struct cxl_dport *dport;
>  	struct cxl_port *port;
>  
>  	if (!is_cxl_port(dev))
> @@ -942,7 +946,10 @@ static int match_port_by_dport(struct device *dev, const void *data)
>  		return 0;
>  
>  	port = to_cxl_port(dev);
> -	return cxl_find_dport_by_dev(port, ctx->dport_dev) != NULL;
> +	dport = cxl_find_dport_by_dev(port, ctx->dport_dev);
> +	if (ctx->dport)
> +		*ctx->dport = dport;
> +	return dport != NULL;
>  }
>  
>  static struct cxl_port *__find_cxl_port(struct cxl_find_port_ctx *ctx)
> @@ -958,24 +965,32 @@ static struct cxl_port *__find_cxl_port(struct cxl_find_port_ctx *ctx)
>  	return NULL;
>  }
>  
> -static struct cxl_port *find_cxl_port(struct device *dport_dev)
> +static struct cxl_port *find_cxl_port(struct device *dport_dev,
> +				      struct cxl_dport **dport)
>  {
>  	struct cxl_find_port_ctx ctx = {
>  		.dport_dev = dport_dev,
> +		.dport = dport,
>  	};
> +	struct cxl_port *port;
>  
> -	return __find_cxl_port(&ctx);
> +	port = __find_cxl_port(&ctx);
> +	return port;
>  }
>  
>  static struct cxl_port *find_cxl_port_at(struct cxl_port *parent_port,
> -					 struct device *dport_dev)
> +					 struct device *dport_dev,
> +					 struct cxl_dport **dport)
>  {
>  	struct cxl_find_port_ctx ctx = {
>  		.dport_dev = dport_dev,
>  		.parent_port = parent_port,
> +		.dport = dport,
>  	};
> +	struct cxl_port *port;
>  
> -	return __find_cxl_port(&ctx);
> +	port = __find_cxl_port(&ctx);
> +	return port;
>  }
>  
>  /*
> @@ -1060,7 +1075,7 @@ static void cxl_detach_ep(void *data)
>  		if (!dport_dev)
>  			break;
>  
> -		port = find_cxl_port(dport_dev);
> +		port = find_cxl_port(dport_dev, NULL);
>  		if (!port)
>  			continue;
>  
> @@ -1135,6 +1150,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
>  	struct device *dparent = grandparent(dport_dev);
>  	struct cxl_port *port, *parent_port = NULL;
>  	resource_size_t component_reg_phys;
> +	struct cxl_dport *dport;
>  	int rc;
>  
>  	if (!dparent) {
> @@ -1148,7 +1164,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
>  		return -ENXIO;
>  	}
>  
> -	parent_port = find_cxl_port(dparent);
> +	parent_port = find_cxl_port(dparent, NULL);
>  	if (!parent_port) {
>  		/* iterate to create this parent_port */
>  		return -EAGAIN;
> @@ -1163,13 +1179,14 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
>  		goto out;
>  	}
>  
> -	port = find_cxl_port_at(parent_port, dport_dev);
> +	port = find_cxl_port_at(parent_port, dport_dev, &dport);
>  	if (!port) {
>  		component_reg_phys = find_component_registers(uport_dev);
>  		port = devm_cxl_add_port(&parent_port->dev, uport_dev,
>  					 component_reg_phys, parent_port);
> +		/* retry find to pick up the new dport information */
>  		if (!IS_ERR(port))
> -			get_device(&port->dev);
> +			port = find_cxl_port_at(parent_port, dport_dev, &dport);
>  	}
>  out:
>  	device_unlock(&parent_port->dev);
> @@ -1179,7 +1196,7 @@ static int add_port_attach_ep(struct cxl_memdev *cxlmd,
>  	else {
>  		dev_dbg(&cxlmd->dev, "add to new port %s:%s\n",
>  			dev_name(&port->dev), dev_name(port->uport));
> -		rc = cxl_add_ep(port, &cxlmd->dev);
> +		rc = cxl_add_ep(dport, &cxlmd->dev);
>  		if (rc == -EEXIST) {
>  			/*
>  			 * "can't" happen, but this error code means
> @@ -1213,6 +1230,7 @@ int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd)
>  	for (iter = dev; iter; iter = grandparent(iter)) {
>  		struct device *dport_dev = grandparent(iter);
>  		struct device *uport_dev;
> +		struct cxl_dport *dport;
>  		struct cxl_port *port;
>  
>  		if (!dport_dev)
> @@ -1228,12 +1246,12 @@ int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd)
>  		dev_dbg(dev, "scan: iter: %s dport_dev: %s parent: %s\n",
>  			dev_name(iter), dev_name(dport_dev),
>  			dev_name(uport_dev));
> -		port = find_cxl_port(dport_dev);
> +		port = find_cxl_port(dport_dev, &dport);
>  		if (port) {
>  			dev_dbg(&cxlmd->dev,
>  				"found already registered port %s:%s\n",
>  				dev_name(&port->dev), dev_name(port->uport));
> -			rc = cxl_add_ep(port, &cxlmd->dev);
> +			rc = cxl_add_ep(dport, &cxlmd->dev);
>  
>  			/*
>  			 * If the endpoint already exists in the port's list,
> @@ -1274,7 +1292,7 @@ EXPORT_SYMBOL_NS_GPL(devm_cxl_enumerate_ports, CXL);
>  
>  struct cxl_port *cxl_mem_find_port(struct cxl_memdev *cxlmd)
>  {
> -	return find_cxl_port(grandparent(&cxlmd->dev));
> +	return find_cxl_port(grandparent(&cxlmd->dev), NULL);
>  }
>  EXPORT_SYMBOL_NS_GPL(cxl_mem_find_port, CXL);
>  
> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
> index 70cd24e4f3ce..31f33844279a 100644
> --- a/drivers/cxl/cxl.h
> +++ b/drivers/cxl/cxl.h
> @@ -371,10 +371,12 @@ struct cxl_dport {
>  /**
>   * struct cxl_ep - track an endpoint's interest in a port
>   * @ep: device that hosts a generic CXL endpoint (expander or accelerator)
> + * @dport: which dport routes to this endpoint on @port
>   * @list: node on port->endpoints list
>   */
>  struct cxl_ep {
>  	struct device *ep;
> +	struct cxl_dport *dport;
>  	struct list_head list;
>  };
>  
> 

Re: [PATCH v2 13/28] cxl/port: Move dport tracking to an xarray

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:01:56 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Reduce the complexity and the overhead of walking the topology to
> determine endpoint connectivity to root decoder interleave
> configurations.
> 
> Note that cxl_detach_ep(), after it determines that the last @ep has
> departed and decides to delete the port, now needs to walk the dport
> array with the device_lock() held to remove entries. Previously
> list_splice_init() could be used atomically delete all dport entries at
> once and then perform entry tear down outside the lock. There is no
> list_splice_init() equivalent for the xarray.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 14/28] cxl/hdm: Add sysfs attributes for interleave ways + granularity

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:02 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> From: Ben Widawsky <bwidawsk@kernel.org>
> 
> The region provisioning flow involves selecting interleave ways +
> granularity settings for a region, and then programming the decoder
> topology to meet those constraints, if possible. For example, root
> decoders set the minimum interleave ways + granularity for any hosted
> regions.
> 
> Given decoder programming is not atomic and collisions can occur between
> multiple requesting regions userspace will be responsible for conflict
> resolution and it needs these attributes to make those decisions.
> 
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> [djbw: reword changelog, make read-only, add sysfs ABI documentaion]
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 16/28] resource: Introduce alloc_free_mem_region()

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:13 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> The core of devm_request_free_mem_region() is a helper that searches for
> free space in iomem_resource and performs __request_region_locked() on
> the result of that search. The policy choices of the implementation
> conform to what CONFIG_DEVICE_PRIVATE users want which is memory that is
> immediately marked busy, and a preference to search for the first-fit
> free range in descending order from the top of the physical address
> space.
> 
> CXL has a need for a similar allocator, but with the following tweaks:
> 
> 1/ Search for free space in ascending order
> 
> 2/ Search for free space relative to a given CXL window
> 
> 3/ 'insert' rather than 'request' the new resource given downstream
>    drivers from the CXL Region driver (like the pmem or dax drivers) are
>    responsible for request_mem_region() when they activate the memory
>    range.
> 
> Rework __request_free_mem_region() into get_free_mem_region() which
> takes a set of GFR_* (Get Free Region) flags to control the allocation
> policy (ascending vs descending), and "busy" policy (insert_resource()
> vs request_region()).
> 
> As part of the consolidation of the legacy GFR_REQUEST_REGION case with
> the new default of just inserting a new resource into the free space
> some minor cleanups like not checking for NULL before calling
> devres_free() (which does its own check) is included.
> 
> Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
> Link: https://lore.kernel.org/linux-cxl/20220420143406.GY2120790@nvidia.com/
> Cc: Matthew Wilcox <willy@infradead.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Given you answered all my questions on v1, I'm fine with this.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 17/28] cxl/region: Add region creation support

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:19 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> From: Ben Widawsky <bwidawsk@kernel.org>
> 
> CXL 2.0 allows for dynamic provisioning of new memory regions (system
> physical address resources like "System RAM" and "Persistent Memory").
> Whereas DDR and PMEM resources are conveyed statically at boot, CXL
> allows for assembling and instantiating new regions from the available
> capacity of CXL memory expanders in the system.
> 
> Sysfs with an "echo $region_name > $create_region_attribute" interface
> is chosen as the mechanism to initiate the provisioning process. This
> was chosen over ioctl() and netlink() to keep the configuration
> interface entirely in a pseudo-fs interface, and it was chosen over
> configfs since, aside from this one creation event, the interface is
> read-mostly. I.e. configfs supports cases where an object is designed to
> be provisioned each boot, like an iSCSI storage target, and CXL region
> creation is mostly for PMEM regions which are created usually once
> per-lifetime of a server instance. This is an improvement over nvdimm
> that pre-created "seed" devices that tended to confuse users looking to
> determine which devices are active and which are idle.
> 
> Recall that the major change that CXL brings over previous persistent
> memory architectures is the ability to dynamically define new regions.
> Compare that to drivers like 'nfit' where the region configuration is
> statically defined by platform firmware.
> 
> Regions are created as a child of a root decoder that encompasses an
> address space with constraints. When created through sysfs, the root
> decoder is explicit. When created from an LSA's region structure a root
> decoder will possibly need to be inferred by the driver.
> 
> Upon region creation through sysfs, a vacant region is created with a
> unique name. Regions have a number of attributes that must be configured
> before the region can be bound to the driver where HDM decoder program
> is completed.
> 
> An example of creating a new region:
> 
> - Allocate a new region name:
> region=$(cat /sys/bus/cxl/devices/decoder0.0/create_pmem_region)
> 
> - Create a new region by name:
> while
> region=$(cat /sys/bus/cxl/devices/decoder0.0/create_pmem_region)
> ! echo $region > /sys/bus/cxl/devices/decoder0.0/create_pmem_region
> do true; done
> 
> - Region now exists in sysfs:
> stat -t /sys/bus/cxl/devices/decoder0.0/$region
> 
> - Delete the region, and name:
> echo $region > /sys/bus/cxl/devices/decoder0.0/delete_region
> 
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> [djbw: simplify locking, reword changelog]
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
V1 comments all addressed so
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 18/28] cxl/region: Add a 'uuid' attribute

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:24 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> From: Ben Widawsky <bwidawsk@kernel.org>
> 
> The process of provisioning a region involves triggering the creation of
> a new region object, pouring in the configuration, and then binding that
> configured object to the region driver to start its operation. For
> persistent memory regions the CXL specification mandates that it
> identified by a uuid. Add an ABI for userspace to specify a region's
> uuid.
> 
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> [djbw: simplify locking]
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 20/28] cxl/region: Allocate HPA capacity to regions

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:36 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> After a region's interleave parameters (ways and granularity) are set,
> add a way for regions to allocate HPA (host physical address space) from
> the free capacity in their parent root-decoder. The allocator for this
> capacity reuses the 'struct resource' based allocator used for
> CONFIG_DEVICE_PRIVATE.
> 
> Once the tuple of "ways, granularity, and size" is set the
> region configuration transitions to the CXL_CONFIG_INTERLEAVE_ACTIVE
> state which is a precursor to allowing endpoint decoders to be added to
> a region.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Comments all look to be addressed either in reply to v1 review or here so LGTM

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> ---
>  Documentation/ABI/testing/sysfs-bus-cxl |   29 ++++++
>  drivers/cxl/Kconfig                     |    3 +
>  drivers/cxl/core/region.c               |  150 +++++++++++++++++++++++++++++++
>  drivers/cxl/cxl.h                       |    2 
>  4 files changed, 183 insertions(+), 1 deletion(-)
> 
> diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
> index bfa42bcc8383..0c6c3da4da5a 100644
> --- a/Documentation/ABI/testing/sysfs-bus-cxl
> +++ b/Documentation/ABI/testing/sysfs-bus-cxl
> @@ -313,3 +313,32 @@ Description:
>  		(RW) Configures the number of devices participating in the
>  		region is set by writing this value. Each device will provide
>  		1/interleave_ways of storage for the region.
> +
> +
> +What:		/sys/bus/cxl/devices/regionZ/size
> +Date:		May, 2022
> +KernelVersion:	v5.20
> +Contact:	linux-cxl@vger.kernel.org
> +Description:
> +		(RW) System physical address space to be consumed by the region.
> +		When written trigger the driver to allocate space out of the
> +		parent root decoder's address space. When read the size of the
> +		address space is reported and should match the span of the
> +		region's resource attribute. Size shall be set after the
> +		interleave configuration parameters. Once set it cannot be
> +		changed, only freed by writing 0. The kernel makes no guarantees
> +		that data is maintained over an address space freeing event, and
> +		there is no guarantee that a free followed by an allocate
> +		results in the same address being allocated.
> +
> +
> +What:		/sys/bus/cxl/devices/regionZ/resource
> +Date:		May, 2022
> +KernelVersion:	v5.20
> +Contact:	linux-cxl@vger.kernel.org
> +Description:
> +		(RO) A region is a contiguous partition of a CXL root decoder
> +		address space. Region capacity is allocated by writing to the
> +		size attribute, the resulting physical address space determined
> +		by the driver is reflected here. It is therefore not useful to
> +		read this before writing a value to the size attribute.
> diff --git a/drivers/cxl/Kconfig b/drivers/cxl/Kconfig
> index aa2728de419e..74c2cd069d9d 100644
> --- a/drivers/cxl/Kconfig
> +++ b/drivers/cxl/Kconfig
> @@ -105,6 +105,9 @@ config CXL_SUSPEND
>  config CXL_REGION
>  	bool
>  	default CXL_BUS
> +	# For MAX_PHYSMEM_BITS
> +	depends on SPARSEMEM
>  	select MEMREGION
> +	select GET_FREE_REGION
>  
>  endif
> diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> index 3289caa5d882..b1e847827c6b 100644
> --- a/drivers/cxl/core/region.c
> +++ b/drivers/cxl/core/region.c
> @@ -244,10 +244,152 @@ static ssize_t interleave_granularity_store(struct device *dev,
>  }
>  static DEVICE_ATTR_RW(interleave_granularity);
>  
> +static ssize_t resource_show(struct device *dev, struct device_attribute *attr,
> +			     char *buf)
> +{
> +	struct cxl_region *cxlr = to_cxl_region(dev);
> +	struct cxl_region_params *p = &cxlr->params;
> +	u64 resource = -1ULL;
> +	ssize_t rc;
> +
> +	rc = down_read_interruptible(&cxl_region_rwsem);
> +	if (rc)
> +		return rc;
> +	if (p->res)
> +		resource = p->res->start;
> +	rc = sysfs_emit(buf, "%#llx\n", resource);
> +	up_read(&cxl_region_rwsem);
> +
> +	return rc;
> +}
> +static DEVICE_ATTR_RO(resource);
> +
> +static int alloc_hpa(struct cxl_region *cxlr, resource_size_t size)
> +{
> +	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(cxlr->dev.parent);
> +	struct cxl_region_params *p = &cxlr->params;
> +	struct resource *res;
> +	u32 remainder = 0;
> +
> +	lockdep_assert_held_write(&cxl_region_rwsem);
> +
> +	/* Nothing to do... */
> +	if (p->res && resource_size(res) == size)
> +		return 0;
> +
> +	/* To change size the old size must be freed first */
> +	if (p->res)
> +		return -EBUSY;
> +
> +	if (p->state >= CXL_CONFIG_INTERLEAVE_ACTIVE)
> +		return -EBUSY;
> +
> +	/* ways, granularity and uuid (if PMEM) need to be set before HPA */
> +	if (!p->interleave_ways || !p->interleave_granularity ||
> +	    (cxlr->mode == CXL_DECODER_PMEM && uuid_is_null(&p->uuid)))
> +		return -ENXIO;
> +
> +	div_u64_rem(size, SZ_256M * p->interleave_ways, &remainder);
> +	if (remainder)
> +		return -EINVAL;
> +
> +	res = alloc_free_mem_region(cxlrd->res, size, SZ_256M,
> +				    dev_name(&cxlr->dev));
> +	if (IS_ERR(res)) {
> +		dev_dbg(&cxlr->dev, "failed to allocate HPA: %ld\n",
> +			PTR_ERR(res));
> +		return PTR_ERR(res);
> +	}
> +
> +	p->res = res;
> +	p->state = CXL_CONFIG_INTERLEAVE_ACTIVE;
> +
> +	return 0;
> +}
> +
> +static void cxl_region_iomem_release(struct cxl_region *cxlr)
> +{
> +	struct cxl_region_params *p = &cxlr->params;
> +
> +	if (device_is_registered(&cxlr->dev))
> +		lockdep_assert_held_write(&cxl_region_rwsem);
> +	if (p->res) {
> +		remove_resource(p->res);
> +		kfree(p->res);
> +		p->res = NULL;
> +	}
> +}
> +
> +static int free_hpa(struct cxl_region *cxlr)
> +{
> +	struct cxl_region_params *p = &cxlr->params;
> +
> +	lockdep_assert_held_write(&cxl_region_rwsem);
> +
> +	if (!p->res)
> +		return 0;
> +
> +	if (p->state >= CXL_CONFIG_ACTIVE)
> +		return -EBUSY;
> +
> +	cxl_region_iomem_release(cxlr);
> +	p->state = CXL_CONFIG_IDLE;
> +	return 0;
> +}
> +
> +static ssize_t size_store(struct device *dev, struct device_attribute *attr,
> +			  const char *buf, size_t len)
> +{
> +	struct cxl_region *cxlr = to_cxl_region(dev);
> +	u64 val;
> +	int rc;
> +
> +	rc = kstrtou64(buf, 0, &val);
> +	if (rc)
> +		return rc;
> +
> +	rc = down_write_killable(&cxl_region_rwsem);
> +	if (rc)
> +		return rc;
> +
> +	if (val)
> +		rc = alloc_hpa(cxlr, val);
> +	else
> +		rc = free_hpa(cxlr);
> +	up_write(&cxl_region_rwsem);
> +
> +	if (rc)
> +		return rc;
> +
> +	return len;
> +}
> +
> +static ssize_t size_show(struct device *dev, struct device_attribute *attr,
> +			 char *buf)
> +{
> +	struct cxl_region *cxlr = to_cxl_region(dev);
> +	struct cxl_region_params *p = &cxlr->params;
> +	u64 size = 0;
> +	ssize_t rc;
> +
> +	rc = down_read_interruptible(&cxl_region_rwsem);
> +	if (rc)
> +		return rc;
> +	if (p->res)
> +		size = resource_size(p->res);
> +	rc = sysfs_emit(buf, "%#llx\n", size);
> +	up_read(&cxl_region_rwsem);
> +
> +	return rc;
> +}
> +static DEVICE_ATTR_RW(size);
> +
>  static struct attribute *cxl_region_attrs[] = {
>  	&dev_attr_uuid.attr,
>  	&dev_attr_interleave_ways.attr,
>  	&dev_attr_interleave_granularity.attr,
> +	&dev_attr_resource.attr,
> +	&dev_attr_size.attr,
>  	NULL,
>  };
>  
> @@ -293,7 +435,11 @@ static struct cxl_region *to_cxl_region(struct device *dev)
>  
>  static void unregister_region(void *dev)
>  {
> -	device_unregister(dev);
> +	struct cxl_region *cxlr = to_cxl_region(dev);
> +
> +	device_del(dev);
> +	cxl_region_iomem_release(cxlr);
> +	put_device(dev);
>  }
>  
>  static struct lock_class_key cxl_region_key;
> @@ -445,3 +591,5 @@ static ssize_t delete_region_store(struct device *dev,
>  	return len;
>  }
>  DEVICE_ATTR_WO(delete_region);
> +
> +MODULE_IMPORT_NS(CXL);
> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
> index a4e65c102bed..837bfa67f469 100644
> --- a/drivers/cxl/cxl.h
> +++ b/drivers/cxl/cxl.h
> @@ -341,6 +341,7 @@ enum cxl_config_state {
>   * @uuid: unique id for persistent regions
>   * @interleave_ways: number of endpoints in the region
>   * @interleave_granularity: capacity each endpoint contributes to a stripe
> + * @res: allocated iomem capacity for this region
>   *
>   * State transitions are protected by the cxl_region_rwsem
>   */
> @@ -349,6 +350,7 @@ struct cxl_region_params {
>  	uuid_t uuid;
>  	int interleave_ways;
>  	int interleave_granularity;
> +	struct resource *res;
>  };
>  
>  /**
> 

Re: [PATCH v2 21/28] cxl/region: Enable the assignment of endpoint decoders to regions

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:41 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> The region provisioning process involves allocating DPA to a set of
> endpoint decoders, and HPA plus the region geometry to a region device.
> Then the decoder is assigned to the region. At this point several
> validation steps can be performed to validate that the decoder is
> suitable to participate in the region.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

I think you've lost some planned changes here as typos from v1 review
still here as is the stale comment.

With those fixed
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>


 
> diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> index b1e847827c6b..871bfdbb9bc8 100644
> --- a/drivers/cxl/core/region.c
> +++ b/drivers/cxl/core/region.c

> +/*
> + * - Check that the given endpoint is attached to a host-bridge identified
> + *   in the root interleave.

In reply to v1 review I think you said you had dropped this comment as stale?

> + */
> +static int cxl_region_attach(struct cxl_region *cxlr,
> +			     struct cxl_endpoint_decoder *cxled, int pos)
> +{
> +	struct cxl_region_params *p = &cxlr->params;
> +
> +	if (cxled->mode == CXL_DECODER_DEAD) {
> +		dev_dbg(&cxlr->dev, "%s dead\n", dev_name(&cxled->cxld.dev));
> +		return -ENODEV;
> +	}
> +
> +	if (pos >= p->interleave_ways) {
> +		dev_dbg(&cxlr->dev, "position %d out of range %d\n", pos,
> +			p->interleave_ways);
> +		return -ENXIO;
> +	}
> +
> +	if (p->targets[pos] == cxled)
> +		return 0;
> +
> +	if (p->targets[pos]) {
> +		struct cxl_endpoint_decoder *cxled_target = p->targets[pos];
> +		struct cxl_memdev *cxlmd_target = cxled_to_memdev(cxled_target);
> +
> +		dev_dbg(&cxlr->dev, "position %d already assigned to %s:%s\n",
> +			pos, dev_name(&cxlmd_target->dev),
> +			dev_name(&cxled_target->cxld.dev));
> +		return -EBUSY;
> +	}
> +
> +	p->targets[pos] = cxled;
> +	cxled->pos = pos;
> +	p->nr_targets++;
> +
> +	return 0;
> +}
> +
> +static void cxl_region_detach(struct cxl_endpoint_decoder *cxled)
> +{
> +	struct cxl_region *cxlr = cxled->cxld.region;
> +	struct cxl_region_params *p;
> +
> +	lockdep_assert_held_write(&cxl_region_rwsem);
> +
> +	if (!cxlr)
> +		return;
> +
> +	p = &cxlr->params;
> +	get_device(&cxlr->dev);
> +
> +	if (cxled->pos < 0 || cxled->pos >= p->interleave_ways ||
> +	    p->targets[cxled->pos] != cxled) {
> +		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> +
> +		dev_WARN_ONCE(&cxlr->dev, 1, "expected %s:%s at position %d\n",
> +			      dev_name(&cxlmd->dev), dev_name(&cxled->cxld.dev),
> +			      cxled->pos);
> +		goto out;
> +	}
> +
> +	p->targets[cxled->pos] = NULL;
> +	p->nr_targets--;
> +
> +	/* notify the region driver that one of its targets has deparated */
typo still here.

> +	up_write(&cxl_region_rwsem);
> +	device_release_driver(&cxlr->dev);
> +	down_write(&cxl_region_rwsem);
> +out:
> +	put_device(&cxlr->dev);
> +}

Re: [PATCH v2 23/28] cxl/region: Attach endpoint decoders

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:52 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> CXL regions (interleave sets) are made up of a set of memory devices
> where each device maps a portion of the interleave with one of its
> decoders (see CXL 2.0 8.2.5.12 CXL HDM Decoder Capability Structure).
> As endpoint decoders are identified by a provisioning tool they can be
> added to a region provided the region interleave properties are set
> (way, granularity, HPA) and DPA has been assigned to the decoder.
> 
> The attach event triggers several validation checks, for example:
> - is the DPA sized appropriately for the region
> - is the decoder reachable via the host-bridges identified by the
>   region's root decoder
> - is the device already active in a different region position slot
> - are there already regions with a higher HPA active on a given port
>   (per CXL 2.0 8.2.5.12.20 Committing Decoder Programming)
> 
> ...and the attach event affords an opportunity to collect data and
> resources relevant to later programming the target lists in switch
> decoders, for example:
> - allocate a decoder at each cxl_port in the decode chain
> - for a given switch port, how many the region's endpoints are hosted
>   through the port
> - how many unique targets (next hops) does a port need to map to reach
>   those endpoints
> 
> The act of reconciling this information and deploying it to the decoder
> configuration is saved for a follow-on patch.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
All the review comment resolutions from v1 look good to me.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Thanks,

Jonathan

Re: [PATCH v2 24/28] cxl/region: Program target lists

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:02:58 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Once the region's interleave geometry (ways, granularity, size) is
> established and all the endpoint decoder targets are assigned, the next
> phase is to program all the intermediate decoders. Specifically, each
> CXL switch in the path between the endpoint and its CXL host-bridge
> (including the logical switch internal to the host-bridge) needs to have
> its decoders programmed and the target list order assigned.
> 
> The difficulty in this implementation lies in determining which endpoint
> decoder ordering combinations are valid. Consider the cxl_test case of 2
> host bridges, each of those host-bridges attached to 2 switches, and
> each of those switches attached to 2 endpoints for a potential 8-way
> interleave. The x2 interleave at the host-bridge level requires that all
> even numbered endpoint decoder positions be located on the "left" hand
> side of the topology tree, and the odd numbered positions on the other.
> The endpoints that are peers on the same switch need to have a position
> that can be routed with a dedicated address bit per-endpoint. See
> check_last_peer() for the details.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

I'm less confident on this one than most the other patches (and I see I skipped
reviewing it in v1) as I haven't closely checked the verification logic
but except for one trivial comment inline it looks fine to me.
I want to hit the whole series with a wide range of test cases (I'm sure you
already have) to build that confidence, but won't have time to do that till early
August. However, if there are gremlins hiding, I'd expect them to be minor.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> index 8ac0c557f6aa..225340529fc3 100644
> --- a/drivers/cxl/core/region.c
> +++ b/drivers/cxl/core/region.c
> @@ -485,6 +485,7 @@ static struct cxl_region_ref *alloc_region_ref(struct cxl_port *port,
>  		return NULL;
>  	cxl_rr->port = port;
>  	cxl_rr->region = cxlr;
> +	cxl_rr->nr_targets = 1;
>  	xa_init(&cxl_rr->endpoints);
>  
>  	rc = xa_insert(&port->regions, (unsigned long)cxlr, cxl_rr, GFP_KERNEL);
> @@ -525,10 +526,12 @@ static int cxl_rr_ep_add(struct cxl_region_ref *cxl_rr,
>  	struct cxl_decoder *cxld = cxl_rr->decoder;
>  	struct cxl_ep *ep = cxl_ep_load(port, cxled_to_memdev(cxled));
>  
> -	rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
> -			 GFP_KERNEL);
> -	if (rc)
> -		return rc;
> +	if (ep) {
> +		rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
> +			       GFP_KERNEL);
> +		if (rc)
> +			return rc;
> +	}
>  	cxl_rr->nr_eps++;
>  
>  	if (!cxld->region) {
> @@ -565,7 +568,7 @@ static int cxl_port_attach_region(struct cxl_port *port,
>  				  struct cxl_endpoint_decoder *cxled, int pos)
>  {
>  	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> -	struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
> +	const struct cxl_ep *ep = cxl_ep_load(port, cxlmd);

Why const now and not previously?  Feels like this should be in an earlier patch.
Maybe I'm missing something though.

>  	struct cxl_region_ref *cxl_rr = NULL, *iter;
>  	struct cxl_region_params *p = &cxlr->params;
>  	struct cxl_decoder *cxld = NULL;
> @@ -649,6 +652,16 @@ static int cxl_port_attach_region(struct cxl_port *port,
>  		goto out_erase;
>  	}
>  
> +	dev_dbg(&cxlr->dev,
> +		"%s:%s %s add: %s:%s @ %d next: %s nr_eps: %d nr_targets: %d\n",
> +		dev_name(port->uport), dev_name(&port->dev),
> +		dev_name(&cxld->dev), dev_name(&cxlmd->dev),
> +		dev_name(&cxled->cxld.dev), pos,
> +		ep ? ep->next ? dev_name(ep->next->uport) :
> +				      dev_name(&cxlmd->dev) :
> +			   "none",
> +		cxl_rr->nr_eps, cxl_rr->nr_targets);
> +
>  	return 0;
>  out_erase:
>  	if (cxl_rr->nr_eps == 0)

Re: [PATCH v2 25/28] cxl/hdm: Commit decoder state to hardware

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:03:04 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> After all the soft validation of the region has completed, convey the
> region configuration to hardware while being careful to commit decoders
> in specification mandated order. In addition to programming the endpoint
> decoder base-address, interleave ways and granularity, the switch
> decoder target lists are also established.
> 
> While the kernel can enforce spec-mandated commit order, it can not
> enforce spec-mandated reset order. For example, the kernel can't stop
> someone from removing an endpoint device that is occupying decoderN in a
> switch decoder where decoderN+1 is also committed. To reset decoderN,
> decoderN+1 must be torn down first. That "tear down the world"
> implementation is saved for a follow-on patch.
> 
> Callback operations are provided for the 'commit' and 'reset'
> operations. While those callbacks may prove useful for CXL accelerators
> (Type-2 devices with memory) the primary motivation is to enable a
> simple way for cxl_test to intercept those operations.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

LGTM
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 27/28] cxl/pmem: Fix offline_nvdimm_bus() to offline by bridge

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:03:15 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Be careful to only disable cxl_pmem objects related to a given
> cxl_nvdimm_bridge. Otherwise, offline_nvdimm_bus() reaches across CXL
> domains and disables more than is expected.
> 
> Fixes: 21083f51521f ("cxl/pmem: Register 'pmem' / cxl_nvdimm devices")
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

All comments on v1 were addressed in your reply so LGTM

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Re: [PATCH v2 28/28] cxl/region: Introduce cxl_pmem_region objects

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:03:21 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> The LIBNVDIMM subsystem is a platform agnostic representation of system
> NVDIMM / persistent memory resources. To date, the CXL subsystem's
> interaction with LIBNVDIMM has been to register an nvdimm-bridge device
> and cxl_nvdimm objects to proxy CXL capabilities into existing LIBNVDIMM
> subsystem mechanics.
> 
> With regions the approach is the same. Create a new cxl_pmem_region
> object to proxy CXL region details into a LIBNVDIMM definition. With
> this enabling LIBNVDIMM can partition CXL persistent memory regions with
> legacy namespace labels. A follow-on patch will add CXL region label and
> CXL namespace label support to persist region configurations across
> driver reload / system-reset events.
> 
> Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
On trivial query below.  Either way I think this looks good.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

> +
> +static void unregister_nvdimm_region(void *nd_region)
> +{
> +	struct cxl_nvdimm_bridge *cxl_nvb;
> +	struct cxl_pmem_region *cxlr_pmem;
> +	int i;
> +
> +	cxlr_pmem = nd_region_provider_data(nd_region);
> +	cxl_nvb = cxlr_pmem->bridge;
> +	device_lock(&cxl_nvb->dev);
> +	for (i = 0; i < cxlr_pmem->nr_mappings; i++) {
> +		struct cxl_pmem_region_mapping *m = &cxlr_pmem->mapping[i];
> +		struct cxl_nvdimm *cxl_nvd = m->cxl_nvd;
> +
> +		if (cxl_nvd->region) {
> +			put_device(&cxlr_pmem->dev);
> +			cxl_nvd->region = NULL;
> +		}
> +	}
> +	device_unlock(&cxl_nvb->dev);
> +
> +	nvdimm_region_delete(nd_region);

I'm not convinced by the ordering in here. Can we do the nvdimm_region_delete() before
taking the device_lock()?  That would make this a nice mirror image of what is
going on in probe().  I may well be missing a reason that doesn't work though.

> +}
> +

Re: [PATCH v2 00/28] CXL PMEM Region Provisioning

[Jonathan Cameron]

Hi Dan,

As I mentioned in one of my reviews I'd love to run a bunch of test
cases against this, but won't get to that until sometime in August.
For some of those tests on QEMU I'll need to add some minor features
(multiple HDM decoder support and handling of skip for example).

However, my limited testing of v1 was looking good and I doesn't seem
like there were any fundamental changes.

So personally I'd be happy with this going in this cycle and getting
additional testing later if you and anyone else who comments feels
that's the way to go.

Thanks,

Jonathan

Re: [PATCH v2 21/28] cxl/region: Enable the assignment of endpoint decoders to regions

[Dan Williams]

Jonathan Cameron wrote:
> On Thu, 14 Jul 2022 17:02:41 -0700
> Dan Williams <dan.j.williams@intel.com> wrote:
> 
> > The region provisioning process involves allocating DPA to a set of
> > endpoint decoders, and HPA plus the region geometry to a region device.
> > Then the decoder is assigned to the region. At this point several
> > validation steps can be performed to validate that the decoder is
> > suitable to participate in the region.
> > 
> > Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> > Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> > Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> 
> I think you've lost some planned changes here as typos from v1 review
> still here as is the stale comment.
> 
> With those fixed
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> 
> 
>  
> > diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> > index b1e847827c6b..871bfdbb9bc8 100644
> > --- a/drivers/cxl/core/region.c
> > +++ b/drivers/cxl/core/region.c
> 
> > +/*
> > + * - Check that the given endpoint is attached to a host-bridge identified
> > + *   in the root interleave.
> 
> In reply to v1 review I think you said you had dropped this comment as stale?

Correct, its gone now along with some other 0day robot reported fixups:

https://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl.git/tree/drivers/cxl/core/region.c?h=preview#n1099

Re: [PATCH v2 00/28] CXL PMEM Region Provisioning

[Jonathan Cameron]

On Thu, 14 Jul 2022 17:00:41 -0700
Dan Williams <dan.j.williams@intel.com> wrote:


Hi Dan,

I'm low on time unfortunately and will be OoO for next week,
But whilst fixing a bug in QEMU, I set up a test to exercise
the high port target register on the hb with

CFMWS interleave ways = 1
hb with 8 rp with a type3 device connected to each.

The resulting interleave granularity isn't what I'd expect to see.
Setting region interleave to 1k (which happens to match the CFMWS)
I'm getting 1k for the CFMWS, 2k for the hb and 256 bytes for the type3
devices.  Which is crazy...  Now there may be another bug lurking
in QEMU so this might not be a kernel issue at all.

For this special case we should be ignoring the CFMWS IG
as it's irrelevant if we aren't interleaving at that level.
We also know we don't have any address bits used for interleave
decoding until the HB.

Thanks,

Jonathan


> Changes since v1 [1]:
> - Move 19 patches that have received a Reviewed-by to the 'pending'
>   branch in cxl.git (Thanks Alison, Adam, and Jonathan!)
> - Improve the changelog and add more Cc's to "cxl/acpi: Track CXL
>   resources in iomem_resource" and highlight the new export of
>   insert_resource_expand_to_fit()
> - Switch all occurrences of the pattern "rc = -ECODE; if (condition)
>   goto err;" to "if (condition) { rc = -ECODE; goto err; }" (Jonathan)
> - Re-organize all the cxl_{root,switch,endpoint}_decoder() patches to
>   move the decoder-type-specific setup into the decoder-type-specific
>   allocation routines (Jonathan)
> - Add kdoc to clarify the behavior of add_cxl_resources() (Jonathan)
> - Add IORES_DESC_CXL for kernel components like EDAC to determine when
>   they might be dealing with a CXL address range (Tony)
> - Drop usage of dev_set_drvdata() for passing @cxl_res (Jonathan)
> - Drop @remove_action argument to __cxl_dpa_release(), make it behave
>   like any other devm_<free> helper (Jonathan)
> - Clarify 'skip' vs 'skipped' in DPA handling helpers (Jonathan)
> - Clarify why port teardown no proceeds under the lock with the
>   conversion from list to xarray (Jonathan)
> - Revert rename of cxl_find_dport_by_dev() (Jonathan)
> - Fold down_read() / up_write() mismatch fix to the patch that
>   introduced the problem (Jonathan)
> - Fix description of interleave_ways and interleave_granularity in the
>   sysfs ABI document
> - Clarify tangential cleanups in "resource: Introduce
>   alloc_free_mem_region()" (Jonathan)
> - Clarify rationale for the region creation / naming ABI (Jonathan)
> - Add SET_CXL_REGION_ATTR() to supplement CXL_REGION_ATTR() the former
>   is used to optionally added region attributes to an attribute list
>   (position independent) and the latter is used to retrieve a pointer to
>   the attribute in code.  (Jonathan)
> - For writes to region attributes allow the same value to be written
>   multiple times without error (Jonathan)
> - Clarify the actions performed by cxl_port_attach_region() (Jonathan)
> - Commit message spelling fixes (Alison and Jonathan)
> - Rename cxl_dpa_resource() => cxl_dpa_resource_start() (Jonathan)
> - Reword error message in cxl_parse_cfmws() (Adam)
> - Keep @expected_len signed in cxl_acpi_cfmws_verify() (Jonathan)
> - Miscellaneous formatting and doc fixes (Jonathan)
> - Rename port->dpa_end port->hdm_end (Jonathan)
> - Rename unregister_region() => unregister_nvdimm_region() (Jonathan)
> 
> [1]: https://lore.kernel.org/linux-cxl/165603869943.551046.3498980330327696732.stgit@dwillia2-xfh
> 
> ---
> 
> Until the CXL 2.0 definition arrived there was little reason for OS
> drivers to care about CXL memory expanders. Similar to DDR they just
> implemented a physical address range that was described to the OS by
> platform firmware (EFI Memory Map + ACPI SRAT/SLIT/HMAT etc). The CXL
> 2.0 definition adds support for PMEM, hotplug, switch topologies, and
> device-interleaving which exceeds the limits of what can be reasonably
> abstracted by EFI + ACPI mechanisms. As a result, Linux needs a native
> capability to provision new CXL regions.
> 
> The term "region" is the same term that originated in the LIBNVDIMM
> implementation to describe a host physical / system physical address
> range. For PMEM a region is a persistent memory range that can be
> further sub-divided into namespaces. For CXL there are three
> classifications of regions:
> - PMEM: set up by CXL native tooling and persisted in CXL region labels
> 
> - RAM: set up dynamically by CXL native tooling after hotplug events, or
>   leftover capacity not mapped by platform firmware. Any persistent
>   configuration would come from set up scripts / configuration files in
>   userspace.
> 
> - System RAM: set up by platform firmware and described by EFI + ACPI
>   metadata, these regions are static.
> 
> For now, these patches implement just PMEM regions without region label
> support. Note though that the infrastructure routines like
> cxl_region_attach() and cxl_region_setup_targets() are building blocks
> for region-label support, provisioning RAM regions, and enumerating
> System RAM regions.
> 
> The general flow for provisioning a CXL region is to:
> - Find a device or set of devices with available device-physical-address
>   (DPA) capacity
> 
> - Find a platform CXL window that has free capacity to map a new region
>   and that is able to target the devices in the previous step.
> 
> - Allocate DPA according to the CXL specification rules of sequential
>   enabling of decoders by id and when a device hosts multiple decoders
>   make sure that lower-id decoders map lower HPA and higher-id decoders
>   map higher HPA.
> 
> - Assign endpoint decoders to a region and validate that the switching
>   topology supports the requested configuration. Recall that
>   interleaving is governed by modulo or xormap math that constrains which
>   device can support which positions in a given region interleave.
> 
> - Program all the decoders an all endpoints and participating switches
>   to bring the new address range online.
> 
> Once the range is online then existing drivers like LIBNVDIMM or
> device-dax can manage the memory range as if the ACPI BIOS had conveyed
> its parameters at boot.
> 
> This patch kit is the result of significant amounts of path finding work
> [2] and long discussions with Ben. Thank you Ben for all that work!
> Where the patches in this kit go in a different design direction than
> the RFC, the authorship is changed and a Co-developed-by is added mainly
> so I get blamed for the bad decisions and not Ben. The major updates
> from that last posting are:
> 
> - all CXL resources are reflected in full in iomem_resource
> 
> - host-physical-address (HPA) range allocation moves to a
>   devm_request_free_mem_region() derivative
> 
> - locking moves to two global rwsems, one for DPA / endpoint decoders
>   and one for HPA / regions.
> 
> - the existing port scanning path is augmented to cache more topology
>   information rather than recreate it at region creation time
> 
> [2]: https://lore.kernel.org/r/20220413183720.2444089-1-ben.widawsky@intel.com
> 
> ---
> 
> Ben Widawsky (4):
>       cxl/hdm: Add sysfs attributes for interleave ways + granularity
>       cxl/region: Add region creation support
>       cxl/region: Add a 'uuid' attribute
>       cxl/region: Add interleave geometry attributes
> 
> Dan Williams (24):
>       Documentation/cxl: Use a double line break between entries
>       cxl/core: Define a 'struct cxl_switch_decoder'
>       cxl/acpi: Track CXL resources in iomem_resource
>       cxl/core: Define a 'struct cxl_root_decoder'
>       cxl/core: Define a 'struct cxl_endpoint_decoder'
>       cxl/hdm: Enumerate allocated DPA
>       cxl/hdm: Add 'mode' attribute to decoder objects
>       cxl/hdm: Track next decoder to allocate
>       cxl/hdm: Add support for allocating DPA to an endpoint decoder
>       cxl/port: Record dport in endpoint references
>       cxl/port: Record parent dport when adding ports
>       cxl/port: Move 'cxl_ep' references to an xarray per port
>       cxl/port: Move dport tracking to an xarray
>       cxl/mem: Enumerate port targets before adding endpoints
>       resource: Introduce alloc_free_mem_region()
>       cxl/region: Allocate HPA capacity to regions
>       cxl/region: Enable the assignment of endpoint decoders to regions
>       cxl/acpi: Add a host-bridge index lookup mechanism
>       cxl/region: Attach endpoint decoders
>       cxl/region: Program target lists
>       cxl/hdm: Commit decoder state to hardware
>       cxl/region: Add region driver boiler plate
>       cxl/pmem: Fix offline_nvdimm_bus() to offline by bridge
>       cxl/region: Introduce cxl_pmem_region objects
> 
> 
>  Documentation/ABI/testing/sysfs-bus-cxl         |  213 +++
>  Documentation/driver-api/cxl/memory-devices.rst |   11 
>  drivers/cxl/Kconfig                             |    8 
>  drivers/cxl/acpi.c                              |  185 ++
>  drivers/cxl/core/Makefile                       |    1 
>  drivers/cxl/core/core.h                         |   49 +
>  drivers/cxl/core/hdm.c                          |  623 +++++++-
>  drivers/cxl/core/pmem.c                         |    4 
>  drivers/cxl/core/port.c                         |  669 ++++++--
>  drivers/cxl/core/region.c                       | 1830 +++++++++++++++++++++++
>  drivers/cxl/cxl.h                               |  263 +++
>  drivers/cxl/cxlmem.h                            |   18 
>  drivers/cxl/mem.c                               |   32 
>  drivers/cxl/pmem.c                              |  259 +++
>  drivers/nvdimm/region_devs.c                    |   28 
>  include/linux/ioport.h                          |    3 
>  include/linux/libnvdimm.h                       |    5 
>  kernel/resource.c                               |  185 ++
>  mm/Kconfig                                      |    5 
>  tools/testing/cxl/Kbuild                        |    1 
>  tools/testing/cxl/test/cxl.c                    |   75 +
>  21 files changed, 4156 insertions(+), 311 deletions(-)
>  create mode 100644 drivers/cxl/core/region.c
> 
> base-commit: b060edfd8cdd52bc8648392500bf152a8dd6d4c5

Re: [PATCH v2 06/28] cxl/hdm: Enumerate allocated DPA

[Dan Williams]

Jonathan Cameron wrote:
> On Thu, 14 Jul 2022 17:01:16 -0700
> Dan Williams <dan.j.williams@intel.com> wrote:
> 
> > In preparation for provisioning CXL regions, add accounting for the DPA
> > space consumed by existing regions / decoders. Recall, a CXL region is a
> > memory range comprised from one or more endpoint devices contributing a
> > mapping of their DPA into HPA space through a decoder.
> > 
> > Record the DPA ranges covered by committed decoders at initial probe of
> > endpoint ports relative to a per-device resource tree of the DPA type
> > (pmem or volatile-ram).
> > 
> > The cxl_dpa_rwsem semaphore is introduced to globally synchronize DPA
> > state across all endpoints and their decoders at once. The vast majority
> > of DPA operations are reads as region creation is expected to be as rare
> > as disk partitioning and volume creation. The device_lock() for this
> > synchronization is specifically avoided for concern of entangling with
> > sysfs attribute removal.
> > 
> > Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
> > Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> 
> 
> One trivial ordering question inline. I'm not that bothered whether you
> do anything about it though as it's all very local.
> 
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> 
> > ---
> >  drivers/cxl/core/hdm.c |  143 ++++++++++++++++++++++++++++++++++++++++++++----
> >  drivers/cxl/cxl.h      |    2 +
> >  drivers/cxl/cxlmem.h   |   13 ++++
> >  3 files changed, 147 insertions(+), 11 deletions(-)
> > 
> > diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> > index 650363d5272f..d4c17325001b 100644
> > --- a/drivers/cxl/core/hdm.c
> > +++ b/drivers/cxl/core/hdm.c
> > @@ -153,10 +153,105 @@ void cxl_dpa_debug(struct seq_file *file, struct cxl_dev_state *cxlds)
> >  }
> >  EXPORT_SYMBOL_NS_GPL(cxl_dpa_debug, CXL);
> >  
> > +/*
> > + * Must be called in a context that synchronizes against this decoder's
> > + * port ->remove() callback (like an endpoint decoder sysfs attribute)
> > + */
> > +static void __cxl_dpa_release(struct cxl_endpoint_decoder *cxled)
> > +{
> > +	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> > +	struct cxl_dev_state *cxlds = cxlmd->cxlds;
> > +	struct resource *res = cxled->dpa_res;
> > +
> > +	lockdep_assert_held_write(&cxl_dpa_rwsem);
> > +
> > +	if (cxled->skip)
> > +		__release_region(&cxlds->dpa_res, res->start - cxled->skip,
> > +				 cxled->skip);
> > +	cxled->skip = 0;
> > +	__release_region(&cxlds->dpa_res, res->start, resource_size(res));
> 
> Minor but I think the ordering in here is unnecessarily not the opposite
> of what is going on in __cxl_dpa_reserve()  Should be releasing the
> actual rs first, then releasing the skip.

Done.

RE: [PATCH v2 16/28] resource: Introduce alloc_free_mem_region()

[Dan Williams]

[ add dri-devel and nouveau ]

Dan Williams wrote:
> The core of devm_request_free_mem_region() is a helper that searches for
> free space in iomem_resource and performs __request_region_locked() on
> the result of that search. The policy choices of the implementation
> conform to what CONFIG_DEVICE_PRIVATE users want which is memory that is
> immediately marked busy, and a preference to search for the first-fit
> free range in descending order from the top of the physical address
> space.
> 
> CXL has a need for a similar allocator, but with the following tweaks:
> 
> 1/ Search for free space in ascending order
> 
> 2/ Search for free space relative to a given CXL window
> 
> 3/ 'insert' rather than 'request' the new resource given downstream
>    drivers from the CXL Region driver (like the pmem or dax drivers) are
>    responsible for request_mem_region() when they activate the memory
>    range.
> 
> Rework __request_free_mem_region() into get_free_mem_region() which
> takes a set of GFR_* (Get Free Region) flags to control the allocation
> policy (ascending vs descending), and "busy" policy (insert_resource()
> vs request_region()).
> 
> As part of the consolidation of the legacy GFR_REQUEST_REGION case with
> the new default of just inserting a new resource into the free space
> some minor cleanups like not checking for NULL before calling
> devres_free() (which does its own check) is included.
> 
> Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
> Link: https://lore.kernel.org/linux-cxl/20220420143406.GY2120790@nvidia.com/
> Cc: Matthew Wilcox <willy@infradead.org>
> Cc: Christoph Hellwig <hch@lst.de>

Jason, Christoph, anyone else that depends on CONFIG_DEVICE_PRIVATE,

Just a heads up that with Jonathan's review I am going to proceed with
pushing this change to linux-next. Please holler if
CONFIG_DEVICE_PRIVATE starts misbehaving, or if you have other feedback,
and I will get it fixed up.

> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
>  include/linux/ioport.h |    2 +
>  kernel/resource.c      |  178 +++++++++++++++++++++++++++++++++++++++---------
>  mm/Kconfig             |    5 +
>  3 files changed, 150 insertions(+), 35 deletions(-)
> 
> diff --git a/include/linux/ioport.h b/include/linux/ioport.h
> index 79d1ad6d6275..616b683563a9 100644
> --- a/include/linux/ioport.h
> +++ b/include/linux/ioport.h
> @@ -330,6 +330,8 @@ struct resource *devm_request_free_mem_region(struct device *dev,
>  		struct resource *base, unsigned long size);
>  struct resource *request_free_mem_region(struct resource *base,
>  		unsigned long size, const char *name);
> +struct resource *alloc_free_mem_region(struct resource *base,
> +		unsigned long size, unsigned long align, const char *name);
>  
>  static inline void irqresource_disabled(struct resource *res, u32 irq)
>  {
> diff --git a/kernel/resource.c b/kernel/resource.c
> index 53a534db350e..4c5e80b92f2f 100644
> --- a/kernel/resource.c
> +++ b/kernel/resource.c
> @@ -489,8 +489,9 @@ int __weak page_is_ram(unsigned long pfn)
>  }
>  EXPORT_SYMBOL_GPL(page_is_ram);
>  
> -static int __region_intersects(resource_size_t start, size_t size,
> -			unsigned long flags, unsigned long desc)
> +static int __region_intersects(struct resource *parent, resource_size_t start,
> +			       size_t size, unsigned long flags,
> +			       unsigned long desc)
>  {
>  	struct resource res;
>  	int type = 0; int other = 0;
> @@ -499,7 +500,7 @@ static int __region_intersects(resource_size_t start, size_t size,
>  	res.start = start;
>  	res.end = start + size - 1;
>  
> -	for (p = iomem_resource.child; p ; p = p->sibling) {
> +	for (p = parent->child; p ; p = p->sibling) {
>  		bool is_type = (((p->flags & flags) == flags) &&
>  				((desc == IORES_DESC_NONE) ||
>  				 (desc == p->desc)));
> @@ -543,7 +544,7 @@ int region_intersects(resource_size_t start, size_t size, unsigned long flags,
>  	int ret;
>  
>  	read_lock(&resource_lock);
> -	ret = __region_intersects(start, size, flags, desc);
> +	ret = __region_intersects(&iomem_resource, start, size, flags, desc);
>  	read_unlock(&resource_lock);
>  
>  	return ret;
> @@ -1780,62 +1781,139 @@ void resource_list_free(struct list_head *head)
>  }
>  EXPORT_SYMBOL(resource_list_free);
>  
> -#ifdef CONFIG_DEVICE_PRIVATE
> -static struct resource *__request_free_mem_region(struct device *dev,
> -		struct resource *base, unsigned long size, const char *name)
> +#ifdef CONFIG_GET_FREE_REGION
> +#define GFR_DESCENDING		(1UL << 0)
> +#define GFR_REQUEST_REGION	(1UL << 1)
> +#define GFR_DEFAULT_ALIGN (1UL << PA_SECTION_SHIFT)
> +
> +static resource_size_t gfr_start(struct resource *base, resource_size_t size,
> +				 resource_size_t align, unsigned long flags)
> +{
> +	if (flags & GFR_DESCENDING) {
> +		resource_size_t end;
> +
> +		end = min_t(resource_size_t, base->end,
> +			    (1ULL << MAX_PHYSMEM_BITS) - 1);
> +		return end - size + 1;
> +	}
> +
> +	return ALIGN(base->start, align);
> +}
> +
> +static bool gfr_continue(struct resource *base, resource_size_t addr,
> +			 resource_size_t size, unsigned long flags)
> +{
> +	if (flags & GFR_DESCENDING)
> +		return addr > size && addr >= base->start;
> +	/*
> +	 * In the ascend case be careful that the last increment by
> +	 * @size did not wrap 0.
> +	 */
> +	return addr > addr - size &&
> +	       addr <= min_t(resource_size_t, base->end,
> +			     (1ULL << MAX_PHYSMEM_BITS) - 1);
> +}
> +
> +static resource_size_t gfr_next(resource_size_t addr, resource_size_t size,
> +				unsigned long flags)
> +{
> +	if (flags & GFR_DESCENDING)
> +		return addr - size;
> +	return addr + size;
> +}
> +
> +static void remove_free_mem_region(void *_res)
> +{
> +	struct resource *res = _res;
> +
> +	if (res->parent)
> +		remove_resource(res);
> +	free_resource(res);
> +}
> +
> +static struct resource *
> +get_free_mem_region(struct device *dev, struct resource *base,
> +		    resource_size_t size, const unsigned long align,
> +		    const char *name, const unsigned long desc,
> +		    const unsigned long flags)
>  {
> -	resource_size_t end, addr;
> +	resource_size_t addr;
>  	struct resource *res;
>  	struct region_devres *dr = NULL;
>  
> -	size = ALIGN(size, 1UL << PA_SECTION_SHIFT);
> -	end = min_t(unsigned long, base->end, (1UL << MAX_PHYSMEM_BITS) - 1);
> -	addr = end - size + 1UL;
> +	size = ALIGN(size, align);
>  
>  	res = alloc_resource(GFP_KERNEL);
>  	if (!res)
>  		return ERR_PTR(-ENOMEM);
>  
> -	if (dev) {
> +	if (dev && (flags & GFR_REQUEST_REGION)) {
>  		dr = devres_alloc(devm_region_release,
>  				sizeof(struct region_devres), GFP_KERNEL);
>  		if (!dr) {
>  			free_resource(res);
>  			return ERR_PTR(-ENOMEM);
>  		}
> +	} else if (dev) {
> +		if (devm_add_action_or_reset(dev, remove_free_mem_region, res))
> +			return ERR_PTR(-ENOMEM);
>  	}
>  
>  	write_lock(&resource_lock);
> -	for (; addr > size && addr >= base->start; addr -= size) {
> -		if (__region_intersects(addr, size, 0, IORES_DESC_NONE) !=
> -				REGION_DISJOINT)
> +	for (addr = gfr_start(base, size, align, flags);
> +	     gfr_continue(base, addr, size, flags);
> +	     addr = gfr_next(addr, size, flags)) {
> +		if (__region_intersects(base, addr, size, 0, IORES_DESC_NONE) !=
> +		    REGION_DISJOINT)
>  			continue;
>  
> -		if (__request_region_locked(res, &iomem_resource, addr, size,
> -						name, 0))
> -			break;
> +		if (flags & GFR_REQUEST_REGION) {
> +			if (__request_region_locked(res, &iomem_resource, addr,
> +						    size, name, 0))
> +				break;
>  
> -		if (dev) {
> -			dr->parent = &iomem_resource;
> -			dr->start = addr;
> -			dr->n = size;
> -			devres_add(dev, dr);
> -		}
> +			if (dev) {
> +				dr->parent = &iomem_resource;
> +				dr->start = addr;
> +				dr->n = size;
> +				devres_add(dev, dr);
> +			}
>  
> -		res->desc = IORES_DESC_DEVICE_PRIVATE_MEMORY;
> -		write_unlock(&resource_lock);
> +			res->desc = desc;
> +			write_unlock(&resource_lock);
> +
> +
> +			/*
> +			 * A driver is claiming this region so revoke any
> +			 * mappings.
> +			 */
> +			revoke_iomem(res);
> +		} else {
> +			res->start = addr;
> +			res->end = addr + size - 1;
> +			res->name = name;
> +			res->desc = desc;
> +			res->flags = IORESOURCE_MEM;
> +
> +			/*
> +			 * Only succeed if the resource hosts an exclusive
> +			 * range after the insert
> +			 */
> +			if (__insert_resource(base, res) || res->child)
> +				break;
> +
> +			write_unlock(&resource_lock);
> +		}
>  
> -		/*
> -		 * A driver is claiming this region so revoke any mappings.
> -		 */
> -		revoke_iomem(res);
>  		return res;
>  	}
>  	write_unlock(&resource_lock);
>  
> -	free_resource(res);
> -	if (dr)
> +	if (flags & GFR_REQUEST_REGION) {
> +		free_resource(res);
>  		devres_free(dr);
> +	} else if (dev)
> +		devm_release_action(dev, remove_free_mem_region, res);
>  
>  	return ERR_PTR(-ERANGE);
>  }
> @@ -1854,18 +1932,48 @@ static struct resource *__request_free_mem_region(struct device *dev,
>  struct resource *devm_request_free_mem_region(struct device *dev,
>  		struct resource *base, unsigned long size)
>  {
> -	return __request_free_mem_region(dev, base, size, dev_name(dev));
> +	unsigned long flags = GFR_DESCENDING | GFR_REQUEST_REGION;
> +
> +	return get_free_mem_region(dev, base, size, GFR_DEFAULT_ALIGN,
> +				   dev_name(dev),
> +				   IORES_DESC_DEVICE_PRIVATE_MEMORY, flags);
>  }
>  EXPORT_SYMBOL_GPL(devm_request_free_mem_region);
>  
>  struct resource *request_free_mem_region(struct resource *base,
>  		unsigned long size, const char *name)
>  {
> -	return __request_free_mem_region(NULL, base, size, name);
> +	unsigned long flags = GFR_DESCENDING | GFR_REQUEST_REGION;
> +
> +	return get_free_mem_region(NULL, base, size, GFR_DEFAULT_ALIGN, name,
> +				   IORES_DESC_DEVICE_PRIVATE_MEMORY, flags);
>  }
>  EXPORT_SYMBOL_GPL(request_free_mem_region);
>  
> -#endif /* CONFIG_DEVICE_PRIVATE */
> +/**
> + * alloc_free_mem_region - find a free region relative to @base
> + * @base: resource that will parent the new resource
> + * @size: size in bytes of memory to allocate from @base
> + * @align: alignment requirements for the allocation
> + * @name: resource name
> + *
> + * Buses like CXL, that can dynamically instantiate new memory regions,
> + * need a method to allocate physical address space for those regions.
> + * Allocate and insert a new resource to cover a free, unclaimed by a
> + * descendant of @base, range in the span of @base.
> + */
> +struct resource *alloc_free_mem_region(struct resource *base,
> +				       unsigned long size, unsigned long align,
> +				       const char *name)
> +{
> +	/* Default of ascending direction and insert resource */
> +	unsigned long flags = 0;
> +
> +	return get_free_mem_region(NULL, base, size, align, name,
> +				   IORES_DESC_NONE, flags);
> +}
> +EXPORT_SYMBOL_NS_GPL(alloc_free_mem_region, CXL);
> +#endif /* CONFIG_GET_FREE_REGION */
>  
>  static int __init strict_iomem(char *str)
>  {
> diff --git a/mm/Kconfig b/mm/Kconfig
> index 169e64192e48..a5b4fee2e3fd 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -994,9 +994,14 @@ config HMM_MIRROR
>  	bool
>  	depends on MMU
>  
> +config GET_FREE_REGION
> +	depends on SPARSEMEM
> +	bool
> +
>  config DEVICE_PRIVATE
>  	bool "Unaddressable device memory (GPU memory, ...)"
>  	depends on ZONE_DEVICE
> +	select GET_FREE_REGION
>  
>  	help
>  	  Allows creation of struct pages to represent unaddressable device
> 

Re: [PATCH v2 00/28] CXL PMEM Region Provisioning

[Dan Williams]

Jonathan Cameron wrote:
> On Thu, 14 Jul 2022 17:00:41 -0700
> Dan Williams <dan.j.williams@intel.com> wrote:
> 
> 
> Hi Dan,
> 
> I'm low on time unfortunately and will be OoO for next week,
> But whilst fixing a bug in QEMU, I set up a test to exercise
> the high port target register on the hb with
> 
> CFMWS interleave ways = 1
> hb with 8 rp with a type3 device connected to each.
> 
> The resulting interleave granularity isn't what I'd expect to see.
> Setting region interleave to 1k (which happens to match the CFMWS)
> I'm getting 1k for the CFMWS, 2k for the hb and 256 bytes for the type3
> devices.  Which is crazy...  Now there may be another bug lurking
> in QEMU so this might not be a kernel issue at all.

Potentially, I will note that there seems to be a QEMU issue, that I
have not had time to dig into, that is preventing region creation
compared to other environments, maybe this is it...

> For this special case we should be ignoring the CFMWS IG
> as it's irrelevant if we aren't interleaving at that level.
> We also know we don't have any address bits used for interleave
> decoding until the HB.

...but I am certain that the drvier implementation is not accounting for
the freedom to specify any granularity when the CFMWS interleave is x1.
Will craft an incremental fix.

Re: [PATCH v2 24/28] cxl/region: Program target lists

[Dan Williams]

Jonathan Cameron wrote:
> On Thu, 14 Jul 2022 17:02:58 -0700
> Dan Williams <dan.j.williams@intel.com> wrote:
> 
> > Once the region's interleave geometry (ways, granularity, size) is
> > established and all the endpoint decoder targets are assigned, the next
> > phase is to program all the intermediate decoders. Specifically, each
> > CXL switch in the path between the endpoint and its CXL host-bridge
> > (including the logical switch internal to the host-bridge) needs to have
> > its decoders programmed and the target list order assigned.
> > 
> > The difficulty in this implementation lies in determining which endpoint
> > decoder ordering combinations are valid. Consider the cxl_test case of 2
> > host bridges, each of those host-bridges attached to 2 switches, and
> > each of those switches attached to 2 endpoints for a potential 8-way
> > interleave. The x2 interleave at the host-bridge level requires that all
> > even numbered endpoint decoder positions be located on the "left" hand
> > side of the topology tree, and the odd numbered positions on the other.
> > The endpoints that are peers on the same switch need to have a position
> > that can be routed with a dedicated address bit per-endpoint. See
> > check_last_peer() for the details.
> > 
> > Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> 
> I'm less confident on this one than most the other patches (and I see I skipped
> reviewing it in v1) as I haven't closely checked the verification logic
> but except for one trivial comment inline it looks fine to me.
> I want to hit the whole series with a wide range of test cases (I'm sure you
> already have) to build that confidence, but won't have time to do that till early
> August. However, if there are gremlins hiding, I'd expect them to be minor.
> 
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> 
> > diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> > index 8ac0c557f6aa..225340529fc3 100644
> > --- a/drivers/cxl/core/region.c
> > +++ b/drivers/cxl/core/region.c
> > @@ -485,6 +485,7 @@ static struct cxl_region_ref *alloc_region_ref(struct cxl_port *port,
> >  		return NULL;
> >  	cxl_rr->port = port;
> >  	cxl_rr->region = cxlr;
> > +	cxl_rr->nr_targets = 1;
> >  	xa_init(&cxl_rr->endpoints);
> >  
> >  	rc = xa_insert(&port->regions, (unsigned long)cxlr, cxl_rr, GFP_KERNEL);
> > @@ -525,10 +526,12 @@ static int cxl_rr_ep_add(struct cxl_region_ref *cxl_rr,
> >  	struct cxl_decoder *cxld = cxl_rr->decoder;
> >  	struct cxl_ep *ep = cxl_ep_load(port, cxled_to_memdev(cxled));
> >  
> > -	rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
> > -			 GFP_KERNEL);
> > -	if (rc)
> > -		return rc;
> > +	if (ep) {
> > +		rc = xa_insert(&cxl_rr->endpoints, (unsigned long)cxled, ep,
> > +			       GFP_KERNEL);
> > +		if (rc)
> > +			return rc;
> > +	}
> >  	cxl_rr->nr_eps++;
> >  
> >  	if (!cxld->region) {
> > @@ -565,7 +568,7 @@ static int cxl_port_attach_region(struct cxl_port *port,
> >  				  struct cxl_endpoint_decoder *cxled, int pos)
> >  {
> >  	struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> > -	struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
> > +	const struct cxl_ep *ep = cxl_ep_load(port, cxlmd);
> 
> Why const now and not previously?

Good question.

> Feels like this should be in an earlier patch.  Maybe I'm missing
> something though.

...or just dropped for now, I can not recall a justification at this
point.

Re: [PATCH v2 00/28] CXL PMEM Region Provisioning

[Jonathan Cameron]

On Thu, 21 Jul 2022 09:29:46 -0700
Dan Williams <dan.j.williams@intel.com> wrote:

> Jonathan Cameron wrote:
> > On Thu, 14 Jul 2022 17:00:41 -0700
> > Dan Williams <dan.j.williams@intel.com> wrote:
> > 
> > 
> > Hi Dan,
> > 
> > I'm low on time unfortunately and will be OoO for next week,
> > But whilst fixing a bug in QEMU, I set up a test to exercise
> > the high port target register on the hb with
> > 
> > CFMWS interleave ways = 1
> > hb with 8 rp with a type3 device connected to each.
> > 
> > The resulting interleave granularity isn't what I'd expect to see.
> > Setting region interleave to 1k (which happens to match the CFMWS)
> > I'm getting 1k for the CFMWS, 2k for the hb and 256 bytes for the type3
> > devices.  Which is crazy...  Now there may be another bug lurking
> > in QEMU so this might not be a kernel issue at all.  
> 
> Potentially, I will note that there seems to be a QEMU issue, that I
> have not had time to dig into, that is preventing region creation
> compared to other environments, maybe this is it...
> 

Fwiw, proposed QEMU fix is this.  One of those where I'm not sure how
it gave the impression of working previously.

From 64c6566601c782e91eafe48eb28711e4bb85e8d0 Mon Sep 17 00:00:00 2001
From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Date: Thu, 21 Jul 2022 13:29:59 +0100
Subject: [PATCH] hw/cxl: Fix wrong query of target ports.

Two issues in this code:
1) Check on which register to look in was inverted.
2) Both branches read the _LO register.

Whilst here moved to extract32() rather than hand rolling
the field extraction as simpler and hopefully less error prone.

Fixes Coverity CID: 1488873

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
---
 hw/cxl/cxl-host.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/hw/cxl/cxl-host.c b/hw/cxl/cxl-host.c
index faa68ef038..1adf61231a 100644
--- a/hw/cxl/cxl-host.c
+++ b/hw/cxl/cxl-host.c
@@ -104,7 +104,6 @@ static bool cxl_hdm_find_target(uint32_t *cache_mem, hwaddr addr,
     uint32_t ctrl;
     uint32_t ig_enc;
     uint32_t iw_enc;
-    uint32_t target_reg;
     uint32_t target_idx;

     ctrl = cache_mem[R_CXL_HDM_DECODER0_CTRL];
@@ -116,14 +115,13 @@ static bool cxl_hdm_find_target(uint32_t *cache_mem, hwaddr addr,
     iw_enc = FIELD_EX32(ctrl, CXL_HDM_DECODER0_CTRL, IW);
     target_idx = (addr / cxl_decode_ig(ig_enc)) % (1 << iw_enc);

-    if (target_idx > 4) {
-        target_reg = cache_mem[R_CXL_HDM_DECODER0_TARGET_LIST_LO];
-        target_reg >>= target_idx * 8;
+    if (target_idx < 4) {
+        *target = extract32(cache_mem[R_CXL_HDM_DECODER0_TARGET_LIST_LO],
+                            target_idx * 8, 8);
     } else {
-        target_reg = cache_mem[R_CXL_HDM_DECODER0_TARGET_LIST_LO];
-        target_reg >>= (target_idx - 4) * 8;
+        *target = extract32(cache_mem[R_CXL_HDM_DECODER0_TARGET_LIST_HI],
+                            (target_idx - 4) * 8, 8);
     }
-    *target = target_reg & 0xff;

     return true;
 }
--
2.32.0



> > For this special case we should be ignoring the CFMWS IG
> > as it's irrelevant if we aren't interleaving at that level.
> > We also know we don't have any address bits used for interleave
> > decoding until the HB.  
> 
> ...but I am certain that the drvier implementation is not accounting for
> the freedom to specify any granularity when the CFMWS interleave is x1.
> Will craft an incremental fix.

Re: [PATCH v2 00/28] CXL PMEM Region Provisioning

[Dan Williams]

Jonathan Cameron wrote:
> 
> 
> Hi Dan,
> 
> As I mentioned in one of my reviews I'd love to run a bunch of test
> cases against this,

Yes, Vishal and I are also looking to have a create-region permutation
test in cxl_test to go beyond:

https://lore.kernel.org/r/165781817516.1555691.3557156570639615515.stgit@dwillia2-xfh.jf.intel.com/

> but won't get to that until sometime in August.
> For some of those tests on QEMU I'll need to add some minor features
> (multiple HDM decoder support and handling of skip for example).
> 
> However, my limited testing of v1 was looking good and I doesn't seem
> like there were any fundamental changes.
> 
> So personally I'd be happy with this going in this cycle and getting
> additional testing later if you and anyone else who comments feels
> that's the way to go.

Thank you for being such a reliable partner on the review and picking up
the torch on the QEMU work. It has significantly accelerated the
development, and I appreciate it.

Re: [PATCH v2 16/28] resource: Introduce alloc_free_mem_region()

[Rogerio Alves]

On 7/14/22 21:02, Dan Williams wrote:
> The core of devm_request_free_mem_region() is a helper that searches for
> free space in iomem_resource and performs __request_region_locked() on
> the result of that search. The policy choices of the implementation
> conform to what CONFIG_DEVICE_PRIVATE users want which is memory that is
> immediately marked busy, and a preference to search for the first-fit
> free range in descending order from the top of the physical address
> space.
> 
> CXL has a need for a similar allocator, but with the following tweaks:
> 
> 1/ Search for free space in ascending order
> 
> 2/ Search for free space relative to a given CXL window
> 
> 3/ 'insert' rather than 'request' the new resource given downstream
>     drivers from the CXL Region driver (like the pmem or dax drivers) are
>     responsible for request_mem_region() when they activate the memory
>     range.
> 
> Rework __request_free_mem_region() into get_free_mem_region() which
> takes a set of GFR_* (Get Free Region) flags to control the allocation
> policy (ascending vs descending), and "busy" policy (insert_resource()
> vs request_region()).
> 
> As part of the consolidation of the legacy GFR_REQUEST_REGION case with
> the new default of just inserting a new resource into the free space
> some minor cleanups like not checking for NULL before calling
> devres_free() (which does its own check) is included.
> 
> Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
> Link: https://lore.kernel.org/linux-cxl/20220420143406.GY2120790@nvidia.com/
> Cc: Matthew Wilcox <willy@infradead.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
>   include/linux/ioport.h |    2 +
>   kernel/resource.c      |  178 +++++++++++++++++++++++++++++++++++++++---------
>   mm/Kconfig             |    5 +
>   3 files changed, 150 insertions(+), 35 deletions(-)
> 
> diff --git a/include/linux/ioport.h b/include/linux/ioport.h
> index 79d1ad6d6275..616b683563a9 100644
> --- a/include/linux/ioport.h
> +++ b/include/linux/ioport.h
> @@ -330,6 +330,8 @@ struct resource *devm_request_free_mem_region(struct device *dev,
>   		struct resource *base, unsigned long size);
>   struct resource *request_free_mem_region(struct resource *base,
>   		unsigned long size, const char *name);
> +struct resource *alloc_free_mem_region(struct resource *base,
> +		unsigned long size, unsigned long align, const char *name);
>   
>   static inline void irqresource_disabled(struct resource *res, u32 irq)
>   {
> diff --git a/kernel/resource.c b/kernel/resource.c
> index 53a534db350e..4c5e80b92f2f 100644
> --- a/kernel/resource.c
> +++ b/kernel/resource.c
> @@ -489,8 +489,9 @@ int __weak page_is_ram(unsigned long pfn)
>   }
>   EXPORT_SYMBOL_GPL(page_is_ram);
>   
> -static int __region_intersects(resource_size_t start, size_t size,
> -			unsigned long flags, unsigned long desc)
> +static int __region_intersects(struct resource *parent, resource_size_t start,
> +			       size_t size, unsigned long flags,
> +			       unsigned long desc)
>   {
>   	struct resource res;
>   	int type = 0; int other = 0;
> @@ -499,7 +500,7 @@ static int __region_intersects(resource_size_t start, size_t size,
>   	res.start = start;
>   	res.end = start + size - 1;
>   
> -	for (p = iomem_resource.child; p ; p = p->sibling) {
> +	for (p = parent->child; p ; p = p->sibling) {
>   		bool is_type = (((p->flags & flags) == flags) &&
>   				((desc == IORES_DESC_NONE) ||
>   				 (desc == p->desc)));
> @@ -543,7 +544,7 @@ int region_intersects(resource_size_t start, size_t size, unsigned long flags,
>   	int ret;
>   
>   	read_lock(&resource_lock);
> -	ret = __region_intersects(start, size, flags, desc);
> +	ret = __region_intersects(&iomem_resource, start, size, flags, desc);
>   	read_unlock(&resource_lock);
>   
>   	return ret;
> @@ -1780,62 +1781,139 @@ void resource_list_free(struct list_head *head)
>   }
>   EXPORT_SYMBOL(resource_list_free);
>   
> -#ifdef CONFIG_DEVICE_PRIVATE
> -static struct resource *__request_free_mem_region(struct device *dev,
> -		struct resource *base, unsigned long size, const char *name)
> +#ifdef CONFIG_GET_FREE_REGION
> +#define GFR_DESCENDING		(1UL << 0)
> +#define GFR_REQUEST_REGION	(1UL << 1)
> +#define GFR_DEFAULT_ALIGN (1UL << PA_SECTION_SHIFT)
> +
> +static resource_size_t gfr_start(struct resource *base, resource_size_t size,
> +				 resource_size_t align, unsigned long flags)
> +{
> +	if (flags & GFR_DESCENDING) {
> +		resource_size_t end;
> +
> +		end = min_t(resource_size_t, base->end,
> +			    (1ULL << MAX_PHYSMEM_BITS) - 1);
> +		return end - size + 1;
> +	}
> +
> +	return ALIGN(base->start, align);
> +}
> +
> +static bool gfr_continue(struct resource *base, resource_size_t addr,
> +			 resource_size_t size, unsigned long flags)
> +{
> +	if (flags & GFR_DESCENDING)
> +		return addr > size && addr >= base->start;
> +	/*
> +	 * In the ascend case be careful that the last increment by
> +	 * @size did not wrap 0.
> +	 */
> +	return addr > addr - size &&
> +	       addr <= min_t(resource_size_t, base->end,
> +			     (1ULL << MAX_PHYSMEM_BITS) - 1);
> +}
> +
> +static resource_size_t gfr_next(resource_size_t addr, resource_size_t size,
> +				unsigned long flags)
> +{
> +	if (flags & GFR_DESCENDING)
> +		return addr - size;
> +	return addr + size;
> +}
> +
> +static void remove_free_mem_region(void *_res)
> +{
> +	struct resource *res = _res;
> +
> +	if (res->parent)
> +		remove_resource(res);
> +	free_resource(res);
> +}
> +
> +static struct resource *
> +get_free_mem_region(struct device *dev, struct resource *base,
> +		    resource_size_t size, const unsigned long align,
> +		    const char *name, const unsigned long desc,
> +		    const unsigned long flags)
>   {
> -	resource_size_t end, addr;
> +	resource_size_t addr;
>   	struct resource *res;
>   	struct region_devres *dr = NULL;
>   
> -	size = ALIGN(size, 1UL << PA_SECTION_SHIFT);
> -	end = min_t(unsigned long, base->end, (1UL << MAX_PHYSMEM_BITS) - 1);
> -	addr = end - size + 1UL;
> +	size = ALIGN(size, align);
>   
>   	res = alloc_resource(GFP_KERNEL);
>   	if (!res)
>   		return ERR_PTR(-ENOMEM);
>   
> -	if (dev) {
> +	if (dev && (flags & GFR_REQUEST_REGION)) {
>   		dr = devres_alloc(devm_region_release,
>   				sizeof(struct region_devres), GFP_KERNEL);
>   		if (!dr) {
>   			free_resource(res);
>   			return ERR_PTR(-ENOMEM);
>   		}
> +	} else if (dev) {
> +		if (devm_add_action_or_reset(dev, remove_free_mem_region, res))
> +			return ERR_PTR(-ENOMEM);
>   	}
>   
>   	write_lock(&resource_lock);
> -	for (; addr > size && addr >= base->start; addr -= size) {
> -		if (__region_intersects(addr, size, 0, IORES_DESC_NONE) !=
> -				REGION_DISJOINT)
> +	for (addr = gfr_start(base, size, align, flags);
> +	     gfr_continue(base, addr, size, flags);
> +	     addr = gfr_next(addr, size, flags)) {
> +		if (__region_intersects(base, addr, size, 0, IORES_DESC_NONE) !=
> +		    REGION_DISJOINT)
>   			continue;
>   
> -		if (__request_region_locked(res, &iomem_resource, addr, size,
> -						name, 0))
> -			break;
> +		if (flags & GFR_REQUEST_REGION) {
> +			if (__request_region_locked(res, &iomem_resource, addr,
> +						    size, name, 0))
> +				break;
>   
> -		if (dev) {
> -			dr->parent = &iomem_resource;
> -			dr->start = addr;
> -			dr->n = size;
> -			devres_add(dev, dr);
> -		}
> +			if (dev) {
> +				dr->parent = &iomem_resource;
> +				dr->start = addr;
> +				dr->n = size;
> +				devres_add(dev, dr);
> +			}
>   
> -		res->desc = IORES_DESC_DEVICE_PRIVATE_MEMORY;
> -		write_unlock(&resource_lock);
> +			res->desc = desc;
> +			write_unlock(&resource_lock);
> +
> +
> +			/*
> +			 * A driver is claiming this region so revoke any
> +			 * mappings.
> +			 */
> +			revoke_iomem(res);
> +		} else {
> +			res->start = addr;
> +			res->end = addr + size - 1;
> +			res->name = name;
> +			res->desc = desc;
> +			res->flags = IORESOURCE_MEM;
> +
> +			/*
> +			 * Only succeed if the resource hosts an exclusive
> +			 * range after the insert
> +			 */
> +			if (__insert_resource(base, res) || res->child)
> +				break;
> +
> +			write_unlock(&resource_lock);
> +		}
>   
> -		/*
> -		 * A driver is claiming this region so revoke any mappings.
> -		 */
> -		revoke_iomem(res);
>   		return res;
>   	}
>   	write_unlock(&resource_lock);
>   
> -	free_resource(res);
> -	if (dr)
> +	if (flags & GFR_REQUEST_REGION) {
> +		free_resource(res);
>   		devres_free(dr);
> +	} else if (dev)
> +		devm_release_action(dev, remove_free_mem_region, res);
>

Coverity is reporting a resource leak here. This may happens if we have 
dev=NULL and flags=0 which seems it is the case when we call this 
function using alloc_free_mem. However, I think this may be a false 
positve since flags=0 means the default behavior aka GFR_ASCENDING | 
GFR_INSERT_RESOURCE which seems fine. I just want to confirm if it is 
possible that we have a corner case such as that.


>   	return ERR_PTR(-ERANGE);
>   }
> @@ -1854,18 +1932,48 @@ static struct resource *__request_free_mem_region(struct device *dev,
>   struct resource *devm_request_free_mem_region(struct device *dev,
>   		struct resource *base, unsigned long size)
>   {
> -	return __request_free_mem_region(dev, base, size, dev_name(dev));
> +	unsigned long flags = GFR_DESCENDING | GFR_REQUEST_REGION;
> +
> +	return get_free_mem_region(dev, base, size, GFR_DEFAULT_ALIGN,
> +				   dev_name(dev),
> +				   IORES_DESC_DEVICE_PRIVATE_MEMORY, flags);
>   }
>   EXPORT_SYMBOL_GPL(devm_request_free_mem_region);
>   
>   struct resource *request_free_mem_region(struct resource *base,
>   		unsigned long size, const char *name)
>   {
> -	return __request_free_mem_region(NULL, base, size, name);
> +	unsigned long flags = GFR_DESCENDING | GFR_REQUEST_REGION;
> +
> +	return get_free_mem_region(NULL, base, size, GFR_DEFAULT_ALIGN, name,
> +				   IORES_DESC_DEVICE_PRIVATE_MEMORY, flags);
>   }
>   EXPORT_SYMBOL_GPL(request_free_mem_region);
>   
> -#endif /* CONFIG_DEVICE_PRIVATE */
> +/**
> + * alloc_free_mem_region - find a free region relative to @base
> + * @base: resource that will parent the new resource
> + * @size: size in bytes of memory to allocate from @base
> + * @align: alignment requirements for the allocation
> + * @name: resource name
> + *
> + * Buses like CXL, that can dynamically instantiate new memory regions,
> + * need a method to allocate physical address space for those regions.
> + * Allocate and insert a new resource to cover a free, unclaimed by a
> + * descendant of @base, range in the span of @base.
> + */
> +struct resource *alloc_free_mem_region(struct resource *base,
> +				       unsigned long size, unsigned long align,
> +				       const char *name)
> +{
> +	/* Default of ascending direction and insert resource */
> +	unsigned long flags = 0;
> +
> +	return get_free_mem_region(NULL, base, size, align, name,
> +				   IORES_DESC_NONE, flags);
> +}
> +EXPORT_SYMBOL_NS_GPL(alloc_free_mem_region, CXL);
> +#endif /* CONFIG_GET_FREE_REGION */
>   
>   static int __init strict_iomem(char *str)
>   {
> diff --git a/mm/Kconfig b/mm/Kconfig
> index 169e64192e48..a5b4fee2e3fd 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -994,9 +994,14 @@ config HMM_MIRROR
>   	bool
>   	depends on MMU
>   
> +config GET_FREE_REGION
> +	depends on SPARSEMEM
> +	bool
> +
>   config DEVICE_PRIVATE
>   	bool "Unaddressable device memory (GPU memory, ...)"
>   	depends on ZONE_DEVICE
> +	select GET_FREE_REGION
>   
>   	help
>   	  Allows creation of struct pages to represent unaddressable device
>