Merge tag 'cxl-fixes-6.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl

Pull compute express link (cxl) fixes from Dan Williams:
 "Several fixes for driver startup regressions that landed during the
  merge window as well as some older bugs.

  The regressions were due to a lack of testing with what the CXL
  specification calls Restricted CXL Host (RCH) topologies compared to
  the testing with Virtual Host (VH) CXL topologies. A VH topology is
  typical PCIe while RCH topologies map CXL endpoints as Root Complex
  Integrated endpoints. The impact is some driver crashes on startup.

  This merge window also added compatibility for range registers (the
  mechanism that CXL 1.1 defined for mapping memory) to treat them like
  HDM decoders (the mechanism that CXL 2.0 defined for mapping
  Host-managed Device Memory). That work collided with the new region
  enumeration code that was tested with CXL 2.0 setups, and fails with
  crashes at startup.

  Lastly, the DOE (Data Object Exchange) implementation for retrieving
  an ACPI-like data table from CXL devices is being reworked for v6.4.
  Several fixes fell out of that work that are suitable for v6.3.

  All of this has been in linux-next for a while, and all reported
  issues [1] have been addressed.

  Summary:

   - Fix several issues with region enumeration in RCH topologies that
     can trigger crashes on driver startup or shutdown.

   - Fix CXL DVSEC range register compatibility versus region
     enumeration that leads to startup crashes

   - Fix CDAT endiannes handling

   - Fix multiple buffer handling boundary conditions

   - Fix Data Object Exchange (DOE) workqueue usage vs
     CONFIG_DEBUG_OBJECTS warn splats"

Link: http://lore.kernel.org/r/20230405075704.33de8121@canb.auug.org.au [1]

* tag 'cxl-fixes-6.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl:
  cxl/hdm: Extend DVSEC range register emulation for region enumeration
  cxl/hdm: Limit emulation to the number of range registers
  cxl/region: Move coherence tracking into cxl_region_attach()
  cxl/region: Fix region setup/teardown for RCDs
  cxl/port: Fix find_cxl_root() for RCDs and simplify it
  cxl/hdm: Skip emulation when driver manages mem_enable
  cxl/hdm: Fix double allocation of @cxlhdm
  PCI/DOE: Fix memory leak with CONFIG_DEBUG_OBJECTS=y
  PCI/DOE: Silence WARN splat with CONFIG_DEBUG_OBJECTS=y
  cxl/pci: Handle excessive CDAT length
  cxl/pci: Handle truncated CDAT entries
  cxl/pci: Handle truncated CDAT header
  cxl/pci: Fix CDAT retrieval on big endian

10 files changed, 175 insertions, 130 deletions

diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index 45deda18ed32..02cc2c38b44b 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -101,25 +101,40 @@ static int map_hdm_decoder_regs(struct cxl_port *port, void __iomem *crb,
 				      BIT(CXL_CM_CAP_CAP_ID_HDM));
 }
 
-static struct cxl_hdm *devm_cxl_setup_emulated_hdm(struct cxl_port *port,
-						   struct cxl_endpoint_dvsec_info *info)
+static bool should_emulate_decoders(struct cxl_endpoint_dvsec_info *info)
 {
-	struct device *dev = &port->dev;
 	struct cxl_hdm *cxlhdm;
+	void __iomem *hdm;
+	u32 ctrl;
+	int i;
 
-	if (!info->mem_enabled)
-		return ERR_PTR(-ENODEV);
+	if (!info)
+		return false;
 
-	cxlhdm = devm_kzalloc(dev, sizeof(*cxlhdm), GFP_KERNEL);
-	if (!cxlhdm)
-		return ERR_PTR(-ENOMEM);
+	cxlhdm = dev_get_drvdata(&info->port->dev);
+	hdm = cxlhdm->regs.hdm_decoder;
 
-	cxlhdm->port = port;
-	cxlhdm->decoder_count = info->ranges;
-	cxlhdm->target_count = info->ranges;
-	dev_set_drvdata(&port->dev, cxlhdm);
+	if (!hdm)
+		return true;
 
-	return cxlhdm;
+	/*
+	 * If HDM decoders are present and the driver is in control of
+	 * Mem_Enable skip DVSEC based emulation
+	 */
+	if (!info->mem_enabled)
+		return false;
+
+	/*
+	 * If any decoders are committed already, there should not be any
+	 * emulated DVSEC decoders.
+	 */
+	for (i = 0; i < cxlhdm->decoder_count; i++) {
+		ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(i));
+		if (FIELD_GET(CXL_HDM_DECODER0_CTRL_COMMITTED, ctrl))
+			return false;
+	}
+
+	return true;
 }
 
 /**
@@ -138,13 +153,14 @@ struct cxl_hdm *devm_cxl_setup_hdm(struct cxl_port *port,
 	cxlhdm = devm_kzalloc(dev, sizeof(*cxlhdm), GFP_KERNEL);
 	if (!cxlhdm)
 		return ERR_PTR(-ENOMEM);
-
 	cxlhdm->port = port;
-	crb = ioremap(port->component_reg_phys, CXL_COMPONENT_REG_BLOCK_SIZE);
-	if (!crb) {
-		if (info && info->mem_enabled)
-			return devm_cxl_setup_emulated_hdm(port, info);
+	dev_set_drvdata(dev, cxlhdm);
 
+	crb = ioremap(port->component_reg_phys, CXL_COMPONENT_REG_BLOCK_SIZE);
+	if (!crb && info && info->mem_enabled) {
+		cxlhdm->decoder_count = info->ranges;
+		return cxlhdm;
+	} else if (!crb) {
 		dev_err(dev, "No component registers mapped\n");
 		return ERR_PTR(-ENXIO);
 	}
@@ -160,7 +176,15 @@ struct cxl_hdm *devm_cxl_setup_hdm(struct cxl_port *port,
 		return ERR_PTR(-ENXIO);
 	}
 
-	dev_set_drvdata(dev, cxlhdm);
+	/*
+	 * Now that the hdm capability is parsed, decide if range
+	 * register emulation is needed and fixup cxlhdm accordingly.
+	 */
+	if (should_emulate_decoders(info)) {
+		dev_dbg(dev, "Fallback map %d range register%s\n", info->ranges,
+			info->ranges > 1 ? "s" : "");
+		cxlhdm->decoder_count = info->ranges;
+	}
 
 	return cxlhdm;
 }
@@ -714,14 +738,20 @@ static int cxl_decoder_reset(struct cxl_decoder *cxld)
 	return 0;
 }
 
-static int cxl_setup_hdm_decoder_from_dvsec(struct cxl_port *port,
-					    struct cxl_decoder *cxld, int which,
-					    struct cxl_endpoint_dvsec_info *info)
+static int cxl_setup_hdm_decoder_from_dvsec(
+	struct cxl_port *port, struct cxl_decoder *cxld, u64 *dpa_base,
+	int which, struct cxl_endpoint_dvsec_info *info)
 {
+	struct cxl_endpoint_decoder *cxled;
+	u64 len;
+	int rc;
+
 	if (!is_cxl_endpoint(port))
 		return -EOPNOTSUPP;
 
-	if (!range_len(&info->dvsec_range[which]))
+	cxled = to_cxl_endpoint_decoder(&cxld->dev);
+	len = range_len(&info->dvsec_range[which]);
+	if (!len)
 		return -ENOENT;
 
 	cxld->target_type = CXL_DECODER_EXPANDER;
@@ -736,40 +766,24 @@ static int cxl_setup_hdm_decoder_from_dvsec(struct cxl_port *port,
 	cxld->flags |= CXL_DECODER_F_ENABLE | CXL_DECODER_F_LOCK;
 	port->commit_end = cxld->id;
 
-	return 0;
-}
-
-static bool should_emulate_decoders(struct cxl_port *port)
-{
-	struct cxl_hdm *cxlhdm = dev_get_drvdata(&port->dev);
-	void __iomem *hdm = cxlhdm->regs.hdm_decoder;
-	u32 ctrl;
-	int i;
-
-	if (!is_cxl_endpoint(cxlhdm->port))
-		return false;
-
-	if (!hdm)
-		return true;
-
-	/*
-	 * If any decoders are committed already, there should not be any
-	 * emulated DVSEC decoders.
-	 */
-	for (i = 0; i < cxlhdm->decoder_count; i++) {
-		ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(i));
-		if (FIELD_GET(CXL_HDM_DECODER0_CTRL_COMMITTED, ctrl))
-			return false;
+	rc = devm_cxl_dpa_reserve(cxled, *dpa_base, len, 0);
+	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 + len - 1, rc);
+		return rc;
 	}
+	*dpa_base += len;
+	cxled->state = CXL_DECODER_STATE_AUTO;
 
-	return true;
+	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, struct cxl_endpoint_dvsec_info *info)
 {
-	struct cxl_endpoint_decoder *cxled = NULL;
+	struct cxl_endpoint_decoder *cxled;
 	u64 size, base, skip, dpa_size;
 	bool committed;
 	u32 remainder;
@@ -780,11 +794,9 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 		unsigned char target_id[8];
 	} target_list;
 
-	if (should_emulate_decoders(port))
-		return cxl_setup_hdm_decoder_from_dvsec(port, cxld, which, info);
-
-	if (is_endpoint_decoder(&cxld->dev))
-		cxled = to_cxl_endpoint_decoder(&cxld->dev);
+	if (should_emulate_decoders(info))
+		return cxl_setup_hdm_decoder_from_dvsec(port, cxld, dpa_base,
+							which, info);
 
 	ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(which));
 	base = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_BASE_LOW_OFFSET(which));
@@ -806,9 +818,6 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 		.end = base + size - 1,
 	};
 
-	if (cxled && !committed && range_len(&info->dvsec_range[which]))
-		return cxl_setup_hdm_decoder_from_dvsec(port, cxld, which, info);
-
 	/* decoders are enabled if committed */
 	if (committed) {
 		cxld->flags |= CXL_DECODER_F_ENABLE;
@@ -846,7 +855,7 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 	if (rc)
 		return rc;
 
-	if (!cxled) {
+	if (!info) {
 		target_list.value =
 			ioread64_hi_lo(hdm + CXL_HDM_DECODER0_TL_LOW(which));
 		for (i = 0; i < cxld->interleave_ways; i++)
@@ -866,6 +875,7 @@ static int init_hdm_decoder(struct cxl_port *port, struct cxl_decoder *cxld,
 		return -ENXIO;
 	}
 	skip = ioread64_hi_lo(hdm + CXL_HDM_DECODER0_SKIP_LOW(which));
+	cxled = to_cxl_endpoint_decoder(&cxld->dev);
 	rc = devm_cxl_dpa_reserve(cxled, *dpa_base + skip, dpa_size, skip);
 	if (rc) {
 		dev_err(&port->dev,
diff --git a/drivers/cxl/core/pci.c b/drivers/cxl/core/pci.c
index 7328a2552411..523d5b9fd7fc 100644
--- a/drivers/cxl/core/pci.c
+++ b/drivers/cxl/core/pci.c
@@ -462,7 +462,7 @@ static struct pci_doe_mb *find_cdat_doe(struct device *uport)
 	return NULL;
 }
 
-#define CDAT_DOE_REQ(entry_handle)					\
+#define CDAT_DOE_REQ(entry_handle) cpu_to_le32				\
 	(FIELD_PREP(CXL_DOE_TABLE_ACCESS_REQ_CODE,			\
 		    CXL_DOE_TABLE_ACCESS_REQ_CODE_READ) |		\
 	 FIELD_PREP(CXL_DOE_TABLE_ACCESS_TABLE_TYPE,			\
@@ -475,8 +475,8 @@ static void cxl_doe_task_complete(struct pci_doe_task *task)
 }
 
 struct cdat_doe_task {
-	u32 request_pl;
-	u32 response_pl[32];
+	__le32 request_pl;
+	__le32 response_pl[32];
 	struct completion c;
 	struct pci_doe_task task;
 };
@@ -510,10 +510,10 @@ static int cxl_cdat_get_length(struct device *dev,
 		return rc;
 	}
 	wait_for_completion(&t.c);
-	if (t.task.rv < sizeof(u32))
+	if (t.task.rv < 2 * sizeof(__le32))
 		return -EIO;
 
-	*length = t.response_pl[1];
+	*length = le32_to_cpu(t.response_pl[1]);
 	dev_dbg(dev, "CDAT length %zu\n", *length);
 
 	return 0;
@@ -524,13 +524,13 @@ static int cxl_cdat_read_table(struct device *dev,
 			       struct cxl_cdat *cdat)
 {
 	size_t length = cdat->length;
-	u32 *data = cdat->table;
+	__le32 *data = cdat->table;
 	int entry_handle = 0;
 
 	do {
 		DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(entry_handle), t);
+		struct cdat_entry_header *entry;
 		size_t entry_dw;
-		u32 *entry;
 		int rc;
 
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
@@ -539,26 +539,34 @@ static int cxl_cdat_read_table(struct device *dev,
 			return rc;
 		}
 		wait_for_completion(&t.c);
-		/* 1 DW header + 1 DW data min */
-		if (t.task.rv < (2 * sizeof(u32)))
+
+		/* 1 DW Table Access Response Header + CDAT entry */
+		entry = (struct cdat_entry_header *)(t.response_pl + 1);
+		if ((entry_handle == 0 &&
+		     t.task.rv != sizeof(__le32) + sizeof(struct cdat_header)) ||
+		    (entry_handle > 0 &&
+		     (t.task.rv < sizeof(__le32) + sizeof(*entry) ||
+		      t.task.rv != sizeof(__le32) + le16_to_cpu(entry->length))))
 			return -EIO;
 
 		/* Get the CXL table access header entry handle */
 		entry_handle = FIELD_GET(CXL_DOE_TABLE_ACCESS_ENTRY_HANDLE,
-					 t.response_pl[0]);
-		entry = t.response_pl + 1;
-		entry_dw = t.task.rv / sizeof(u32);
+					 le32_to_cpu(t.response_pl[0]));
+		entry_dw = t.task.rv / sizeof(__le32);
 		/* Skip Header */
 		entry_dw -= 1;
-		entry_dw = min(length / sizeof(u32), entry_dw);
+		entry_dw = min(length / sizeof(__le32), entry_dw);
 		/* Prevent length < 1 DW from causing a buffer overflow */
 		if (entry_dw) {
-			memcpy(data, entry, entry_dw * sizeof(u32));
-			length -= entry_dw * sizeof(u32);
+			memcpy(data, entry, entry_dw * sizeof(__le32));
+			length -= entry_dw * sizeof(__le32);
 			data += entry_dw;
 		}
 	} while (entry_handle != CXL_DOE_TABLE_ACCESS_LAST_ENTRY);
 
+	/* Length in CDAT header may exceed concatenation of CDAT entries */
+	cdat->length -= length;
+
 	return 0;
 }
 
diff --git a/drivers/cxl/core/pmem.c b/drivers/cxl/core/pmem.c
index c2e4b1093788..f8c38d997252 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 device *start)
+struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct cxl_memdev *cxlmd)
 {
-	struct cxl_port *port = find_cxl_root(start);
+	struct cxl_port *port = find_cxl_root(dev_get_drvdata(&cxlmd->dev));
 	struct device *dev;
 
 	if (!port)
@@ -253,7 +253,7 @@ int devm_cxl_add_nvdimm(struct cxl_memdev *cxlmd)
 	struct device *dev;
 	int rc;
 
-	cxl_nvb = cxl_find_nvdimm_bridge(&cxlmd->dev);
+	cxl_nvb = cxl_find_nvdimm_bridge(cxlmd);
 	if (!cxl_nvb)
 		return -ENODEV;
 
diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
index 8ee6b6e2e2a4..4d1f9c5b5029 100644
--- a/drivers/cxl/core/port.c
+++ b/drivers/cxl/core/port.c
@@ -823,41 +823,17 @@ static bool dev_is_cxl_root_child(struct device *dev)
 	return false;
 }
 
-/* Find a 2nd level CXL port that has a dport that is an ancestor of @match */
-static int match_root_child(struct device *dev, const void *match)
+struct cxl_port *find_cxl_root(struct cxl_port *port)
 {
-	const struct device *iter = NULL;
-	struct cxl_dport *dport;
-	struct cxl_port *port;
-
-	if (!dev_is_cxl_root_child(dev))
-		return 0;
-
-	port = to_cxl_port(dev);
-	iter = match;
-	while (iter) {
-		dport = cxl_find_dport_by_dev(port, iter);
-		if (dport)
-			break;
-		iter = iter->parent;
-	}
-
-	return !!iter;
-}
+	struct cxl_port *iter = port;
 
-struct cxl_port *find_cxl_root(struct device *dev)
-{
-	struct device *port_dev;
-	struct cxl_port *root;
+	while (iter && !is_cxl_root(iter))
+		iter = to_cxl_port(iter->dev.parent);
 
-	port_dev = bus_find_device(&cxl_bus_type, NULL, dev, match_root_child);
-	if (!port_dev)
+	if (!iter)
 		return NULL;
-
-	root = to_cxl_port(port_dev->parent);
-	get_device(&root->dev);
-	put_device(port_dev);
-	return root;
+	get_device(&iter->dev);
+	return iter;
 }
 EXPORT_SYMBOL_NS_GPL(find_cxl_root, CXL);
 
diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
index f29028148806..b2fd67fcebfb 100644
--- a/drivers/cxl/core/region.c
+++ b/drivers/cxl/core/region.c
@@ -134,9 +134,13 @@ static int cxl_region_decode_reset(struct cxl_region *cxlr, int count)
 		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_dev_state *cxlds = cxlmd->cxlds;
 		struct cxl_ep *ep;
 		int rc = 0;
 
+		if (cxlds->rcd)
+			goto endpoint_reset;
+
 		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
 			iter = to_cxl_port(iter->dev.parent);
 
@@ -153,6 +157,7 @@ static int cxl_region_decode_reset(struct cxl_region *cxlr, int count)
 				return rc;
 		}
 
+endpoint_reset:
 		rc = cxled->cxld.reset(&cxled->cxld);
 		if (rc)
 			return rc;
@@ -1199,6 +1204,7 @@ static void cxl_region_teardown_targets(struct cxl_region *cxlr)
 {
 	struct cxl_region_params *p = &cxlr->params;
 	struct cxl_endpoint_decoder *cxled;
+	struct cxl_dev_state *cxlds;
 	struct cxl_memdev *cxlmd;
 	struct cxl_port *iter;
 	struct cxl_ep *ep;
@@ -1214,6 +1220,10 @@ static void cxl_region_teardown_targets(struct cxl_region *cxlr)
 	for (i = 0; i < p->nr_targets; i++) {
 		cxled = p->targets[i];
 		cxlmd = cxled_to_memdev(cxled);
+		cxlds = cxlmd->cxlds;
+
+		if (cxlds->rcd)
+			continue;
 
 		iter = cxled_to_port(cxled);
 		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
@@ -1229,14 +1239,24 @@ static int cxl_region_setup_targets(struct cxl_region *cxlr)
 {
 	struct cxl_region_params *p = &cxlr->params;
 	struct cxl_endpoint_decoder *cxled;
+	struct cxl_dev_state *cxlds;
+	int i, rc, rch = 0, vh = 0;
 	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);
+		cxlds = cxlmd->cxlds;
+
+		/* validate that all targets agree on topology */
+		if (!cxlds->rcd) {
+			vh++;
+		} else {
+			rch++;
+			continue;
+		}
 
 		iter = cxled_to_port(cxled);
 		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
@@ -1256,6 +1276,12 @@ static int cxl_region_setup_targets(struct cxl_region *cxlr)
 		}
 	}
 
+	if (rch && vh) {
+		dev_err(&cxlr->dev, "mismatched CXL topologies detected\n");
+		cxl_region_teardown_targets(cxlr);
+		return -ENXIO;
+	}
+
 	return 0;
 }
 
@@ -1648,6 +1674,7 @@ static int cxl_region_attach(struct cxl_region *cxlr,
 		if (rc)
 			goto err_decrement;
 		p->state = CXL_CONFIG_ACTIVE;
+		set_bit(CXL_REGION_F_INCOHERENT, &cxlr->flags);
 	}
 
 	cxled->cxld.interleave_ways = p->interleave_ways;
@@ -1749,8 +1776,6 @@ static int attach_target(struct cxl_region *cxlr,
 
 	down_read(&cxl_dpa_rwsem);
 	rc = cxl_region_attach(cxlr, cxled, pos);
-	if (rc == 0)
-		set_bit(CXL_REGION_F_INCOHERENT, &cxlr->flags);
 	up_read(&cxl_dpa_rwsem);
 	up_write(&cxl_region_rwsem);
 	return rc;
@@ -2251,7 +2276,7 @@ static struct cxl_pmem_region *cxl_pmem_region_alloc(struct cxl_region *cxlr)
 		 * bridge for one device is the same for all.
 		 */
 		if (i == 0) {
-			cxl_nvb = cxl_find_nvdimm_bridge(&cxlmd->dev);
+			cxl_nvb = cxl_find_nvdimm_bridge(cxlmd);
 			if (!cxl_nvb) {
 				cxlr_pmem = ERR_PTR(-ENODEV);
 				goto out;
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index f2b0962a552d..044a92d9813e 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -658,7 +658,7 @@ 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);
-struct cxl_port *find_cxl_root(struct device *dev);
+struct cxl_port *find_cxl_root(struct cxl_port *port);
 int devm_cxl_enumerate_ports(struct cxl_memdev *cxlmd);
 void cxl_bus_rescan(void);
 void cxl_bus_drain(void);
@@ -695,13 +695,15 @@ int cxl_endpoint_autoremove(struct cxl_memdev *cxlmd, struct cxl_port *endpoint)
 
 /**
  * struct cxl_endpoint_dvsec_info - Cached DVSEC info
- * @mem_enabled: cached value of mem_enabled in the DVSEC, PCIE_DEVICE
+ * @mem_enabled: cached value of mem_enabled in the DVSEC at init time
  * @ranges: Number of active HDM ranges this device uses.
+ * @port: endpoint port associated with this info instance
  * @dvsec_range: cached attributes of the ranges in the DVSEC, PCIE_DEVICE
  */
 struct cxl_endpoint_dvsec_info {
 	bool mem_enabled;
 	int ranges;
+	struct cxl_port *port;
 	struct range dvsec_range[2];
 };
 
@@ -758,7 +760,7 @@ 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 cxl_memdev *cxlmd);
-struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct device *dev);
+struct cxl_nvdimm_bridge *cxl_find_nvdimm_bridge(struct cxl_memdev *cxlmd);
 
 #ifdef CONFIG_CXL_REGION
 bool is_cxl_pmem_region(struct device *dev);
diff --git a/drivers/cxl/cxlpci.h b/drivers/cxl/cxlpci.h
index be6a2ef3cce3..0465ef963cd6 100644
--- a/drivers/cxl/cxlpci.h
+++ b/drivers/cxl/cxlpci.h
@@ -68,6 +68,20 @@ enum cxl_regloc_type {
 	CXL_REGLOC_RBI_TYPES
 };
 
+struct cdat_header {
+	__le32 length;
+	u8 revision;
+	u8 checksum;
+	u8 reserved[6];
+	__le32 sequence;
+} __packed;
+
+struct cdat_entry_header {
+	u8 type;
+	u8 reserved;
+	__le16 length;
+} __packed;
+
 int devm_cxl_port_enumerate_dports(struct cxl_port *port);
 struct cxl_dev_state;
 int cxl_hdm_decode_init(struct cxl_dev_state *cxlds, struct cxl_hdm *cxlhdm,
diff --git a/drivers/cxl/port.c b/drivers/cxl/port.c
index 1049bb5ea496..22a7ab2bae7c 100644
--- a/drivers/cxl/port.c
+++ b/drivers/cxl/port.c
@@ -78,8 +78,8 @@ static int cxl_switch_port_probe(struct cxl_port *port)
 
 static int cxl_endpoint_port_probe(struct cxl_port *port)
 {
+	struct cxl_endpoint_dvsec_info info = { .port = port };
 	struct cxl_memdev *cxlmd = to_cxl_memdev(port->uport);
-	struct cxl_endpoint_dvsec_info info = { 0 };
 	struct cxl_dev_state *cxlds = cxlmd->cxlds;
 	struct cxl_hdm *cxlhdm;
 	struct cxl_port *root;
@@ -119,7 +119,7 @@ static int cxl_endpoint_port_probe(struct cxl_port *port)
 	 * This can't fail in practice as CXL root exit unregisters all
 	 * descendant ports and that in turn synchronizes with cxl_port_probe()
 	 */
-	root = find_cxl_root(&cxlmd->dev);
+	root = find_cxl_root(port);
 
 	/*
 	 * Now that all endpoint decoders are successfully enumerated, try to
diff --git a/drivers/pci/doe.c b/drivers/pci/doe.c
index 66d9ab288646..e5e9b287b976 100644
--- a/drivers/pci/doe.c
+++ b/drivers/pci/doe.c
@@ -128,7 +128,7 @@ static int pci_doe_send_req(struct pci_doe_mb *doe_mb,
 		return -EIO;
 
 	/* Length is 2 DW of header + length of payload in DW */
-	length = 2 + task->request_pl_sz / sizeof(u32);
+	length = 2 + task->request_pl_sz / sizeof(__le32);
 	if (length > PCI_DOE_MAX_LENGTH)
 		return -EIO;
 	if (length == PCI_DOE_MAX_LENGTH)
@@ -141,9 +141,9 @@ static int pci_doe_send_req(struct pci_doe_mb *doe_mb,
 	pci_write_config_dword(pdev, offset + PCI_DOE_WRITE,
 			       FIELD_PREP(PCI_DOE_DATA_OBJECT_HEADER_2_LENGTH,
 					  length));
-	for (i = 0; i < task->request_pl_sz / sizeof(u32); i++)
+	for (i = 0; i < task->request_pl_sz / sizeof(__le32); i++)
 		pci_write_config_dword(pdev, offset + PCI_DOE_WRITE,
-				       task->request_pl[i]);
+				       le32_to_cpu(task->request_pl[i]));
 
 	pci_doe_write_ctrl(doe_mb, PCI_DOE_CTRL_GO);
 
@@ -195,11 +195,11 @@ static int pci_doe_recv_resp(struct pci_doe_mb *doe_mb, struct pci_doe_task *tas
 
 	/* First 2 dwords have already been read */
 	length -= 2;
-	payload_length = min(length, task->response_pl_sz / sizeof(u32));
+	payload_length = min(length, task->response_pl_sz / sizeof(__le32));
 	/* Read the rest of the response payload */
 	for (i = 0; i < payload_length; i++) {
-		pci_read_config_dword(pdev, offset + PCI_DOE_READ,
-				      &task->response_pl[i]);
+		pci_read_config_dword(pdev, offset + PCI_DOE_READ, &val);
+		task->response_pl[i] = cpu_to_le32(val);
 		/* Prior to the last ack, ensure Data Object Ready */
 		if (i == (payload_length - 1) && !pci_doe_data_obj_ready(doe_mb))
 			return -EIO;
@@ -217,13 +217,14 @@ static int pci_doe_recv_resp(struct pci_doe_mb *doe_mb, struct pci_doe_task *tas
 	if (FIELD_GET(PCI_DOE_STATUS_ERROR, val))
 		return -EIO;
 
-	return min(length, task->response_pl_sz / sizeof(u32)) * sizeof(u32);
+	return min(length, task->response_pl_sz / sizeof(__le32)) * sizeof(__le32);
 }
 
 static void signal_task_complete(struct pci_doe_task *task, int rv)
 {
 	task->rv = rv;
 	task->complete(task);
+	destroy_work_on_stack(&task->work);
 }
 
 static void signal_task_abort(struct pci_doe_task *task, int rv)
@@ -317,14 +318,16 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 {
 	u32 request_pl = FIELD_PREP(PCI_DOE_DATA_OBJECT_DISC_REQ_3_INDEX,
 				    *index);
+	__le32 request_pl_le = cpu_to_le32(request_pl);
+	__le32 response_pl_le;
 	u32 response_pl;
 	DECLARE_COMPLETION_ONSTACK(c);
 	struct pci_doe_task task = {
 		.prot.vid = PCI_VENDOR_ID_PCI_SIG,
 		.prot.type = PCI_DOE_PROTOCOL_DISCOVERY,
-		.request_pl = &request_pl,
+		.request_pl = &request_pl_le,
 		.request_pl_sz = sizeof(request_pl),
-		.response_pl = &response_pl,
+		.response_pl = &response_pl_le,
 		.response_pl_sz = sizeof(response_pl),
 		.complete = pci_doe_task_complete,
 		.private = &c,
@@ -340,6 +343,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	if (task.rv != sizeof(response_pl))
 		return -EIO;
 
+	response_pl = le32_to_cpu(response_pl_le);
 	*vid = FIELD_GET(PCI_DOE_DATA_OBJECT_DISC_RSP_3_VID, response_pl);
 	*protocol = FIELD_GET(PCI_DOE_DATA_OBJECT_DISC_RSP_3_PROTOCOL,
 			      response_pl);
@@ -520,6 +524,8 @@ EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
  * task->complete will be called when the state machine is done processing this
  * task.
  *
+ * @task must be allocated on the stack.
+ *
  * Excess data will be discarded.
  *
  * RETURNS: 0 when task has been successfully queued, -ERRNO on error
@@ -533,15 +539,15 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
 	 */
-	if (task->request_pl_sz % sizeof(u32) ||
-	    task->response_pl_sz < sizeof(u32))
+	if (task->request_pl_sz % sizeof(__le32) ||
+	    task->response_pl_sz < sizeof(__le32))
 		return -EINVAL;
 
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
 	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
+	INIT_WORK_ONSTACK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
diff --git a/include/linux/pci-doe.h b/include/linux/pci-doe.h
index ed9b4df792b8..43765eaf2342 100644
--- a/include/linux/pci-doe.h
+++ b/include/linux/pci-doe.h
@@ -34,6 +34,10 @@ struct pci_doe_mb;
  * @work: Used internally by the mailbox
  * @doe_mb: Used internally by the mailbox
  *
+ * Payloads are treated as opaque byte streams which are transmitted verbatim,
+ * without byte-swapping.  If payloads contain little-endian register values,
+ * the caller is responsible for conversion with cpu_to_le32() / le32_to_cpu().
+ *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
  *
@@ -45,9 +49,9 @@ struct pci_doe_mb;
  */
 struct pci_doe_task {
 	struct pci_doe_protocol prot;
-	u32 *request_pl;
+	__le32 *request_pl;
 	size_t request_pl_sz;
-	u32 *response_pl;
+	__le32 *response_pl;
 	size_t response_pl_sz;
 	int rv;
 	void (*complete)(struct pci_doe_task *task);