Skip to contentSkip to conversation

cxl: Support Poison Inject & Clear by Region Offset

PATCH Superseded [email protected] 1 participant 6 messages Jul 22, 2025 CXL

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

v4 cover cxl: Support Poison Inject & Clear by Region Offset

From: Alison Schofield <[email protected]>

Built upon cxl repo branch cxl/next
Please excuse the checkpatch errors in new ACQUIRE syntax.

A CXL Unit Test update to cxl-poison.sh is posted separately. A stand-alone test of the SPA<=>HPA<=>DPA of calculations is here:
https://github.com/pmem/ndctl/commit/0b5ae6882d882145f671c6d6bd9a4fc1edb4b99d

Changes in v4:
Changelog appears per patch also.
Add new precursor patch creating a root decoder ops structure (DaveJ) Add /* Input validation ensures valid ways and gran */ (Jonathan) Use a temp instead of modifying hpa_offset in place (Jonathan) Reorder mask addends to match spec definition (Jonathan) Add a blank line, remove a blank line (Jonathan) s/Deivce/Device in commit log (Jonathan)
Use div64_u64_rem() for MOD operation (lkp)
Simplify return in cxl_[inject|clear]_poison() (Jonathan)

Changes in v3:
Check return of ways_to_eiw() & granularity_to_eig() (Jonathan) Collapse a header comment into pos and offset blocks (Jonathan) Calc pos for 3,6,12 using actual ways, not always 3 (Jonathan) Mask off bottom bits in < 8 offset case (Jonathan)
Wrap code comments closer to column 80 (Jonathan)
Use a continue to un-nest a for loop. (Jonathan)
Undo v2 return of rc from locked variants. Return 0. (Jonathan) Return rc from region_offset_to_dpa_result() if present (Jonathan) Remove the errant Documentation/ABI fixup (Jonathan) Add the above rc to the dev_dbg message

Changes in v2:
Rebased onto cxl repo branch for-6.17/cxl-acquire
As the ACQUIRE() set noted, ACQUIRE() syntax leads to a checkpatch
ERROR: do not use assignment in if condition
We'll have to endure that noise here until that set is merged with a checkpatch.pl update or a change to the syntax.

Changelog appears per patch also:
Simplify return using test_bit() in cxl_memdev_has_poison() (Jonathan) Use ACQUIRE() in the debugfs inject and clear funcs (DaveJ, Jonathan) Shift by (eig + eiw) not (eig + 8) in MOD 3 interleaves** (Jonathan) Simplify bottom bit handling by saving and restoring (Jonathan) Return the rc from locked variants in cxl_clear/inject_poison Fail if offset is in the extended linear cache (Jonathan) Calculate the pos and dpa_offset inline, not in a helper Remove the not customary __ prefix from locked variants Added 'cxl' to an existing ABI for memdev clear_poison
Add and use a root decoder callback for spa_to_hpa() Redefine ABI to take a region offset, not SPA (Dan)
Use div64_u64 instead of / to fix 32-bit ARM (lkp)
Use div64_u64_rem instead of % for arch safety
Remove KernelVersion field in ABI doc (Dan)
Pass pointer to results structures (DaveJ)
Add spec references and comments (DaveJ)
Warn against misuse in ABI doc (Dan)
Add validate_region_offset() helper

Begin Cover Letter

This series allows expert users to inject and clear poison by writing a Host Physical Address (HPA) to a region debugfs files. At the core of this new functionality is a helper that translates an HPA into a Device Physical Address (DPA) and a memdev based on the region's decoder configuration.

The set is not merely a convenience wrapper for these region poison operations as it enables these operations for XOR interleaved regions where they were previously impossible.

Patch 1 is a new precursor patch creating a root decoder ops structure.

Patch 2 defines a SPA->CXL HPA root decoder callback for XOR Math. It's a restructuring and renaming exercise that enables the reuse of an existing xormap function in either direction SPA<-->CXL HPA. It gets used in Patch 2.

Patch 3 introduces the translation logic capable of retrieving the memdev and a DPA for a region offset.

Patch 4 adds a locked variant of the inject and clear poison ops to support callers that must hold locks during the entire translation and operation sequence. It gets used in Patch 5.

Patch 5 exposes the capability through region debugfs attributes that only appear when all participating memdevs support the poison commands.

By the end of Patch 5 a region offset has been translated to a memdev and a DPA and can simply be passed through to the pre-existing per memdev inject and clear poison routines.

Alison Schofield (4):
cxl: Define a SPA->CXL HPA root decoder callback for XOR Math cxl/region: Introduce SPA to DPA address translation cxl/core: Add locked variants of the poison inject and clear funcs cxl/region: Add inject and clear poison by region offset

Dave Jiang (1):
cxl: Move hpa_to_spa callback to a new root decoder ops structure

 Documentation/ABI/testing/debugfs-cxl |  87 ++++++++++
 drivers/cxl/acpi.c                    |  35 ++--
 drivers/cxl/core/core.h               |   4 +
 drivers/cxl/core/memdev.c             |  60 +++++--
 drivers/cxl/core/port.c               |   2 +
 drivers/cxl/core/region.c             | 237 +++++++++++++++++++++++++-
 drivers/cxl/cxl.h                     |  14 +-
 drivers/cxl/cxlmem.h                  |   2 +
 8 files changed, 406 insertions(+), 35 deletions(-)

base-commit: 3a32c5b3bb7d2dfad5fab94817f59e8963e2b1a6

Signature · 1 line
2.37.3
author maintainer
#1 lore raw

PATCH v4 1/5 cxl: Move hpa_to_spa callback to a new root decoder ops structure

From: Dave Jiang <[email protected]>

The root decoder's HPA to SPA translation logic was implemented using a single function pointer. In preparation for additional per-decoder callbacks, convert this into a struct cxl_rd_ops and move the hpa_to_spa pointer into it.

To avoid maintaining a static ops instance populated with mostly NULL pointers, allocate the ops structure dynamically only when a platform requires overrides (e.g. XOR interleave decoding).

The setup can be extended as additional callbacks are added.

---

Changes in v4: this patch is new to v4

 drivers/cxl/acpi.c        | 10 +++++++---
 drivers/cxl/core/port.c   |  2 ++
 drivers/cxl/core/region.c | 11 ++++++++---
 drivers/cxl/cxl.h         | 12 +++++++++---
 4 files changed, 26 insertions(+), 9 deletions(-)
drivers/cxl/acpi.c +7 -3
index 712624cba2b6..b8d69460a368 100644
@@ -20,7 +20,6 @@ static const guid_t acpi_cxl_qtg_id_guid =
 	GUID_INIT(0xF365F9A6, 0xA7DE, 0x4071,
 		  0xA6, 0x6A, 0xB4, 0x0C, 0x0B, 0x4F, 0x8E, 0x52);
 
-
 static u64 cxl_xor_hpa_to_spa(struct cxl_root_decoder *cxlrd, u64 hpa)
 {
 	struct cxl_cxims_data *cximsd = cxlrd->platform_data;
@@ -472,8 +471,13 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,
 
 	cxlrd->qos_class = cfmws->qtg_id;
 
-	if (cfmws->interleave_arithmetic == ACPI_CEDT_CFMWS_ARITHMETIC_XOR)
-		cxlrd->hpa_to_spa = cxl_xor_hpa_to_spa;
+	if (cfmws->interleave_arithmetic == ACPI_CEDT_CFMWS_ARITHMETIC_XOR) {
+		cxlrd->ops = kzalloc(sizeof(*cxlrd->ops), GFP_KERNEL);
+		if (!cxlrd->ops)
+			return -ENOMEM;
+
+		cxlrd->ops->hpa_to_spa = cxl_xor_hpa_to_spa;
+	}
 
 	rc = cxl_decoder_add(cxld, target_map);
 	if (rc)
drivers/cxl/core/port.c +2 -0
index 29197376b18e..439473947423 100644
@@ -450,6 +450,7 @@ static void cxl_root_decoder_release(struct device *dev)
 	if (atomic_read(&cxlrd->region_id) >= 0)
 		memregion_free(atomic_read(&cxlrd->region_id));
 	__cxl_decoder_release(&cxlrd->cxlsd.cxld);
+	kfree(cxlrd->ops);
 	kfree(cxlrd);
 }
 
@@ -1833,6 +1834,7 @@ struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,
 
 	atomic_set(&cxlrd->region_id, rc);
 	cxlrd->qos_class = CXL_QOS_CLASS_INVALID;
+
 	return cxlrd;
 }
 EXPORT_SYMBOL_NS_GPL(cxl_root_decoder_alloc, "CXL");
drivers/cxl/core/region.c +8 -3
index e9bf42d91689..3a097ecda1ea 100644
@@ -2918,6 +2918,11 @@ static bool cxl_is_hpa_in_chunk(u64 hpa, struct cxl_region *cxlr, int pos)
 	return false;
 }
 
+static bool has_hpa_to_spa(struct cxl_root_decoder *cxlrd)
+{
+	return cxlrd->ops && cxlrd->ops->hpa_to_spa;
+}
+
 u64 cxl_dpa_to_hpa(struct cxl_region *cxlr, const struct cxl_memdev *cxlmd,
 		   u64 dpa)
 {
@@ -2972,8 +2977,8 @@ u64 cxl_dpa_to_hpa(struct cxl_region *cxlr, const struct cxl_memdev *cxlmd,
 	hpa = hpa_offset + p->res->start + p->cache_size;
 
 	/* Root decoder translation overrides typical modulo decode */
-	if (cxlrd->hpa_to_spa)
-		hpa = cxlrd->hpa_to_spa(cxlrd, hpa);
+	if (has_hpa_to_spa(cxlrd))
+		hpa = cxlrd->ops->hpa_to_spa(cxlrd, hpa);
 
 	if (!cxl_resource_contains_addr(p->res, hpa)) {
 		dev_dbg(&cxlr->dev,
@@ -2982,7 +2987,7 @@ u64 cxl_dpa_to_hpa(struct cxl_region *cxlr, const struct cxl_memdev *cxlmd,
 	}
 
 	/* Simple chunk check, by pos & gran, only applies to modulo decodes */
-	if (!cxlrd->hpa_to_spa && (!cxl_is_hpa_in_chunk(hpa, cxlr, pos)))
+	if (!has_hpa_to_spa(cxlrd) && (!cxl_is_hpa_in_chunk(hpa, cxlr, pos)))
 		return ULLONG_MAX;
 
 	return hpa;
drivers/cxl/cxl.h +9 -3
index b7111e3568d0..bb82777128a2 100644
@@ -418,27 +418,33 @@ struct cxl_switch_decoder {
 };
 
 struct cxl_root_decoder;
-typedef u64 (*cxl_hpa_to_spa_fn)(struct cxl_root_decoder *cxlrd, u64 hpa);
+/**
+ * struct cxl_rd_ops - CXL root decoder callback operations
+ * @hpa_to_spa: Convert host physical address to system physical address
+ */
+struct cxl_rd_ops {
+	u64 (*hpa_to_spa)(struct cxl_root_decoder *cxlrd, u64 hpa);
+};
 
 /**
  * struct cxl_root_decoder - Static platform CXL address decoder
  * @res: host / parent resource for region allocations
  * @cache_size: extended linear cache size if exists, otherwise zero.
  * @region_id: region id for next region provisioning event
- * @hpa_to_spa: translate CXL host-physical-address to Platform system-physical-address
  * @platform_data: platform specific configuration data
  * @range_lock: sync region autodiscovery by address range
  * @qos_class: QoS performance class cookie
+ * @ops: CXL root decoder operations
  * @cxlsd: base cxl switch decoder
  */
 struct cxl_root_decoder {
 	struct resource *res;
 	resource_size_t cache_size;
 	atomic_t region_id;
-	cxl_hpa_to_spa_fn hpa_to_spa;
 	void *platform_data;
 	struct mutex range_lock;
 	int qos_class;
+	struct cxl_rd_ops *ops;
 	struct cxl_switch_decoder cxlsd;
 };
 
Signature · 1 line
2.37.3
author maintainer
#2 lore raw

PATCH v4 2/5 cxl: Define a SPA->CXL HPA root decoder callback for XOR Math

In reply to [email protected]

From: Alison Schofield <[email protected]>

When DPA->SPA translation was introduced, it included a helper that applied the XOR maps to do the CXL HPA -> SPA translation for XOR region interleaves. In preparation for adding SPA->DPA address translation, introduce the reverse callback.

The root decoder callback is defined generically and not all usages may be self inverting like this XOR function. Add another root decoder callback that is the spa_to_hpa function.

Update the existing cxl_xor_hpa_to_spa() with a name that reflects what it does without directionality: cxl_apply_xor_maps(), a generic
parameter: addr replaces hpa, and code comments stating that the function supports the translation in either direction.

---

Changes in v4: none

 drivers/cxl/acpi.c | 27 ++++++++++++++++-----------
 drivers/cxl/cxl.h  |  2 ++
 2 files changed, 18 insertions(+), 11 deletions(-)
drivers/cxl/acpi.c +16 -11
index b8d69460a368..f1625212b08b 100644
@@ -20,7 +20,7 @@ static const guid_t acpi_cxl_qtg_id_guid =
 	GUID_INIT(0xF365F9A6, 0xA7DE, 0x4071,
 		  0xA6, 0x6A, 0xB4, 0x0C, 0x0B, 0x4F, 0x8E, 0x52);
 
-static u64 cxl_xor_hpa_to_spa(struct cxl_root_decoder *cxlrd, u64 hpa)
+static u64 cxl_apply_xor_maps(struct cxl_root_decoder *cxlrd, u64 addr)
 {
 	struct cxl_cxims_data *cximsd = cxlrd->platform_data;
 	int hbiw = cxlrd->cxlsd.nr_targets;
@@ -29,19 +29,23 @@ static u64 cxl_xor_hpa_to_spa(struct cxl_root_decoder *cxlrd, u64 hpa)
 
 	/* No xormaps for host bridge interleave ways of 1 or 3 */
 	if (hbiw == 1 || hbiw == 3)
-		return hpa;
+		return addr;
 
 	/*
-	 * For root decoders using xormaps (hbiw: 2,4,6,8,12,16) restore
-	 * the position bit to its value before the xormap was applied at
-	 * HPA->DPA translation.
+	 * In regions using XOR interleave arithmetic the CXL HPA may not
+	 * be the same as the SPA. This helper performs the SPA->CXL HPA
+	 * or the CXL HPA->SPA translation. Since XOR is self-inverting,
+	 * so is this function.
+	 *
+	 * For root decoders using xormaps (hbiw: 2,4,6,8,12,16) applying the
+	 * xormaps will toggle a position bit.
 	 *
 	 * pos is the lowest set bit in an XORMAP
-	 * val is the XORALLBITS(HPA & XORMAP)
+	 * val is the XORALLBITS(addr & XORMAP)
 	 *
 	 * XORALLBITS: The CXL spec (3.1 Table 9-22) defines XORALLBITS
 	 * as an operation that outputs a single bit by XORing all the
-	 * bits in the input (hpa & xormap). Implement XORALLBITS using
+	 * bits in the input (addr & xormap). Implement XORALLBITS using
 	 * hweight64(). If the hamming weight is even the XOR of those
 	 * bits results in val==0, if odd the XOR result is val==1.
 	 */
@@ -50,11 +54,11 @@ static u64 cxl_xor_hpa_to_spa(struct cxl_root_decoder *cxlrd, u64 hpa)
 		if (!cximsd->xormaps[i])
 			continue;
 		pos = __ffs(cximsd->xormaps[i]);
-		val = (hweight64(hpa & cximsd->xormaps[i]) & 1);
-		hpa = (hpa & ~(1ULL << pos)) | (val << pos);
+		val = (hweight64(addr & cximsd->xormaps[i]) & 1);
+		addr = (addr & ~(1ULL << pos)) | (val << pos);
 	}
 
-	return hpa;
+	return addr;
 }
 
 struct cxl_cxims_context {
@@ -476,7 +480,8 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,
 		if (!cxlrd->ops)
 			return -ENOMEM;
 
-		cxlrd->ops->hpa_to_spa = cxl_xor_hpa_to_spa;
+		cxlrd->ops->hpa_to_spa = cxl_apply_xor_maps;
+		cxlrd->ops->spa_to_hpa = cxl_apply_xor_maps;
 	}
 
 	rc = cxl_decoder_add(cxld, target_map);
drivers/cxl/cxl.h +2 -0
index bb82777128a2..f87628a22852 100644
@@ -421,9 +421,11 @@ struct cxl_root_decoder;
 /**
  * struct cxl_rd_ops - CXL root decoder callback operations
  * @hpa_to_spa: Convert host physical address to system physical address
+ * @spa_to_hpa: Convert system physical address to host physical address
  */
 struct cxl_rd_ops {
 	u64 (*hpa_to_spa)(struct cxl_root_decoder *cxlrd, u64 hpa);
+	u64 (*spa_to_hpa)(struct cxl_root_decoder *cxlrd, u64 spa);
 };
 
 /**
Signature · 1 line
2.37.3
author maintainer
#3 lore raw

PATCH v4 3/5 cxl/region: Introduce SPA to DPA address translation

In reply to [email protected]

From: Alison Schofield <[email protected]>

Add infrastructure to translate System Physical Addresses (SPA) to Device Physical Addresses (DPA) within CXL regions. This capability will be used by follow-on patches that add poison inject and clear operations at the region level.

The SPA-to-DPA translation process follows these steps:
1. Apply root decoder transformations (SPA to HPA) if configured.
2. Extract the position in region interleave from the HPA offset.
3. Extract the DPA offset from the HPA offset.
4. Use position to find endpoint decoder.
5. Use endpoint decoder to find memdev and calculate DPA from offset.
6. Return the result - a memdev and a DPA.

It is Step 1 above that makes this a driver level operation and not work we can push to user space. Rather than exporting the XOR maps for root decoders configured with XOR interleave, the driver performs this complex calculation for the user.

Steps 2 and 3 follow the CXL Spec 3.2 Section 8.2.4.20.13 Implementation Note: Device Decode Logic.

These calculations mirror much of the logic introduced earlier in DPA to SPA translation, see cxl_dpa_to_hpa(), where the driver needed to reverse the spec defined 'Device Decode Logic'.

---

Changes in v4:
Add /* Input validation ensures valid ways and gran */ (Jonathan) Use a temp instead of modifying hpa_offset in place (Jonathan) Reorder mask addends to match spec definition (Jonathan) Add a blank line, remove a blank line (Jonathan) s/Deivce/Device in commit log (Jonathan)
Use div64_u64_rem() for MOD operation (lkp)

 drivers/cxl/core/region.c | 101 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 101 insertions(+)
drivers/cxl/core/region.c +101 -0
index 3a097ecda1ea..da0759acb87e 100644
@@ -2923,6 +2923,11 @@ static bool has_hpa_to_spa(struct cxl_root_decoder *cxlrd)
 	return cxlrd->ops && cxlrd->ops->hpa_to_spa;
 }
 
+static bool has_spa_to_hpa(struct cxl_root_decoder *cxlrd)
+{
+	return cxlrd->ops && cxlrd->ops->spa_to_hpa;
+}
+
 u64 cxl_dpa_to_hpa(struct cxl_region *cxlr, const struct cxl_memdev *cxlmd,
 		   u64 dpa)
 {
@@ -2993,6 +2998,102 @@ u64 cxl_dpa_to_hpa(struct cxl_region *cxlr, const struct cxl_memdev *cxlmd,
 	return hpa;
 }
 
+struct dpa_result {
+	struct cxl_memdev *cxlmd;
+	u64 dpa;
+};
+
+static int __maybe_unused region_offset_to_dpa_result(struct cxl_region *cxlr,
+						      u64 offset,
+						      struct dpa_result *result)
+{
+	struct cxl_region_params *p = &cxlr->params;
+	struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(cxlr->dev.parent);
+	struct cxl_endpoint_decoder *cxled;
+	u64 hpa, hpa_offset, dpa_offset;
+	u64 bits_upper, bits_lower;
+	u64 shifted, rem, temp;
+	u16 eig = 0;
+	u8 eiw = 0;
+	int pos;
+
+	lockdep_assert_held(&cxl_rwsem.region);
+	lockdep_assert_held(&cxl_rwsem.dpa);
+
+	/* Input validation ensures valid ways and gran */
+	granularity_to_eig(p->interleave_granularity, &eig);
+	ways_to_eiw(p->interleave_ways, &eiw);
+
+	/*
+	 * If the root decoder has SPA to CXL HPA callback, use it. Otherwise
+	 * CXL HPA is assumed to equal SPA.
+	 */
+	if (has_spa_to_hpa(cxlrd)) {
+		hpa = cxlrd->ops->spa_to_hpa(cxlrd, p->res->start + offset);
+		hpa_offset = hpa - p->res->start;
+	} else {
+		hpa_offset = offset;
+	}
+	/*
+	 * Interleave position: CXL Spec 3.2 Section 8.2.4.20.13
+	 * eiw < 8
+	 *	Position is in the IW bits at HPA_OFFSET[IG+8+IW-1:IG+8].
+	 *	Per spec "remove IW bits starting with bit position IG+8"
+	 * eiw >= 8
+	 *	Position is not explicitly stored in HPA_OFFSET bits. It is
+	 *	derived from the modulo operation of the upper bits using
+	 *	the total number of interleave ways.
+	 */
+	if (eiw < 8) {
+		pos = (hpa_offset >> (eig + 8)) & GENMASK(eiw - 1, 0);
+	} else {
+		shifted = hpa_offset >> (eig + 8);
+		div64_u64_rem(shifted, p->interleave_ways, &rem);
+		pos = rem;
+	}
+	if (pos < 0 || pos >= p->nr_targets) {
+		dev_dbg(&cxlr->dev, "Invalid position %d for %d targets\n",
+			pos, p->nr_targets);
+		return -ENXIO;
+	}
+
+	/*
+	 * DPA offset: CXL Spec 3.2 Section 8.2.4.20.13
+	 * Lower bits [IG+7:0] pass through unchanged
+	 * (eiw < 8)
+	 *	Per spec: DPAOffset[51:IG+8] = (HPAOffset[51:IG+IW+8] >> IW)
+	 *	Clear the position bits to isolate upper section, then
+	 *	reverse the left shift by eiw that occurred during DPA->HPA
+	 * (eiw >= 8)
+	 *	Per spec: DPAOffset[51:IG+8] = HPAOffset[51:IG+IW] / 3
+	 *	Extract upper bits from the correct bit range and divide by 3
+	 *	to recover the original DPA upper bits
+	 */
+	bits_lower = hpa_offset & GENMASK_ULL(eig + 7, 0);
+	if (eiw < 8) {
+		temp = hpa_offset &= ~((u64)GENMASK(eig + eiw + 8 - 1, 0));
+		dpa_offset = temp >> eiw;
+	} else {
+		bits_upper = div64_u64(hpa_offset >> (eig + eiw), 3);
+		dpa_offset = bits_upper << (eig + 8);
+	}
+	dpa_offset |= bits_lower;
+
+	/* Look-up and return the result: a memdev and a DPA */
+	for (int i = 0; i < p->nr_targets; i++) {
+		cxled = p->targets[i];
+		if (cxled->pos != pos)
+			continue;
+		result->cxlmd = cxled_to_memdev(cxled);
+		result->dpa = cxl_dpa_resource_start(cxled) + dpa_offset;
+
+		return 0;
+	}
+	dev_err(&cxlr->dev, "No device found for position %d\n", pos);
+
+	return -ENXIO;
+}
+
 static struct lock_class_key cxl_pmem_region_key;
 
 static int cxl_pmem_region_alloc(struct cxl_region *cxlr)
Signature · 1 line
2.37.3
author maintainer
#4 lore raw

PATCH v4 4/5 cxl/core: Add locked variants of the poison inject and clear funcs

In reply to [email protected]

From: Alison Schofield <[email protected]>

The core functions that validate and send inject and clear commands to the memdev devices require holding both the dpa_rwsem and the region_rwsem.

In preparation for another caller of these functions that must hold the locks upon entry, split the work into a locked and unlocked pair.

Consideration was given to moving the locking to both callers, however, the existing caller is not in the core (mem.c) and cannot access the locks.

---

Changes in v4:
Simplify return in cxl_[inject|clear]_poison() (Jonathan)

 drivers/cxl/core/memdev.c | 52 +++++++++++++++++++++++++++------------
 drivers/cxl/cxlmem.h      |  2 ++
 2 files changed, 38 insertions(+), 16 deletions(-)
drivers/cxl/core/memdev.c +36 -16
index c569e00a511f..90d3390d9c7c 100644
@@ -276,7 +276,7 @@ static int cxl_validate_poison_dpa(struct cxl_memdev *cxlmd, u64 dpa)
 	return 0;
 }
 
-int cxl_inject_poison(struct cxl_memdev *cxlmd, u64 dpa)
+int cxl_inject_poison_locked(struct cxl_memdev *cxlmd, u64 dpa)
 {
 	struct cxl_mailbox *cxl_mbox = &cxlmd->cxlds->cxl_mbox;
 	struct cxl_mbox_inject_poison inject;
@@ -288,13 +288,8 @@ int cxl_inject_poison(struct cxl_memdev *cxlmd, u64 dpa)
 	if (!IS_ENABLED(CONFIG_DEBUG_FS))
 		return 0;
 
-	ACQUIRE(rwsem_read_intr, region_rwsem)(&cxl_rwsem.region);
-	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &region_rwsem)))
-		return rc;
-
-	ACQUIRE(rwsem_read_intr, dpa_rwsem)(&cxl_rwsem.dpa);
-	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &dpa_rwsem)))
-		return rc;
+	lockdep_assert_held(&cxl_rwsem.dpa);
+	lockdep_assert_held(&cxl_rwsem.region);
 
 	rc = cxl_validate_poison_dpa(cxlmd, dpa);
 	if (rc)
@@ -324,9 +319,24 @@ int cxl_inject_poison(struct cxl_memdev *cxlmd, u64 dpa)
 
 	return 0;
 }
+
+int cxl_inject_poison(struct cxl_memdev *cxlmd, u64 dpa)
+{
+	int rc;
+
+	ACQUIRE(rwsem_read_intr, region_rwsem)(&cxl_rwsem.region);
+	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &region_rwsem)))
+		return rc;
+
+	ACQUIRE(rwsem_read_intr, dpa_rwsem)(&cxl_rwsem.dpa);
+	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &dpa_rwsem)))
+		return rc;
+
+	return cxl_inject_poison_locked(cxlmd, dpa);
+}
 EXPORT_SYMBOL_NS_GPL(cxl_inject_poison, "CXL");
 
-int cxl_clear_poison(struct cxl_memdev *cxlmd, u64 dpa)
+int cxl_clear_poison_locked(struct cxl_memdev *cxlmd, u64 dpa)
 {
 	struct cxl_mailbox *cxl_mbox = &cxlmd->cxlds->cxl_mbox;
 	struct cxl_mbox_clear_poison clear;
@@ -338,13 +348,8 @@ int cxl_clear_poison(struct cxl_memdev *cxlmd, u64 dpa)
 	if (!IS_ENABLED(CONFIG_DEBUG_FS))
 		return 0;
 
-	ACQUIRE(rwsem_read_intr, region_rwsem)(&cxl_rwsem.region);
-	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &region_rwsem)))
-		return rc;
-
-	ACQUIRE(rwsem_read_intr, dpa_rwsem)(&cxl_rwsem.dpa);
-	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &dpa_rwsem)))
-		return rc;
+	lockdep_assert_held(&cxl_rwsem.dpa);
+	lockdep_assert_held(&cxl_rwsem.region);
 
 	rc = cxl_validate_poison_dpa(cxlmd, dpa);
 	if (rc)
@@ -383,6 +388,21 @@ int cxl_clear_poison(struct cxl_memdev *cxlmd, u64 dpa)
 
 	return 0;
 }
+
+int cxl_clear_poison(struct cxl_memdev *cxlmd, u64 dpa)
+{
+	int rc;
+
+	ACQUIRE(rwsem_read_intr, region_rwsem)(&cxl_rwsem.region);
+	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &region_rwsem)))
+		return rc;
+
+	ACQUIRE(rwsem_read_intr, dpa_rwsem)(&cxl_rwsem.dpa);
+	if ((rc = ACQUIRE_ERR(rwsem_read_intr, &dpa_rwsem)))
+		return rc;
+
+	return cxl_clear_poison_locked(cxlmd, dpa);
+}
 EXPORT_SYMBOL_NS_GPL(cxl_clear_poison, "CXL");
 
 static struct attribute *cxl_memdev_attributes[] = {
drivers/cxl/cxlmem.h +2 -0
index 751478dfc410..434031a0c1f7 100644
@@ -869,6 +869,8 @@ int cxl_mem_get_poison(struct cxl_memdev *cxlmd, u64 offset, u64 len,
 int cxl_trigger_poison_list(struct cxl_memdev *cxlmd);
 int cxl_inject_poison(struct cxl_memdev *cxlmd, u64 dpa);
 int cxl_clear_poison(struct cxl_memdev *cxlmd, u64 dpa);
+int cxl_inject_poison_locked(struct cxl_memdev *cxlmd, u64 dpa);
+int cxl_clear_poison_locked(struct cxl_memdev *cxlmd, u64 dpa);
 
 #ifdef CONFIG_CXL_EDAC_MEM_FEATURES
 int devm_cxl_memdev_edac_register(struct cxl_memdev *cxlmd);
Signature · 1 line
2.37.3
author maintainer
#5 lore raw

PATCH v4 5/5 cxl/region: Add inject and clear poison by region offset

In reply to [email protected]

From: Alison Schofield <[email protected]>

Add CXL region debugfs attributes to inject and clear poison based on an offset into the region. These new interfaces allow users to operate on poison at the region level without needing to resolve Device Physical Addresses (DPA) or target individual memdevs.

The implementation uses a new helper, region_offset_to_dpa_result() that applies decoder interleave logic, including XOR-based address decoding when applicable. Note that XOR decodes rely on driver internal xormaps which are not exposed to userspace. So, this support is not only a simplification of poison operations that could be done using existing per memdev operations, but also it enables this functionality for XOR interleaved regions for the first time.

New debugfs attributes are added in /sys/kernel/debug/cxl/regionX/:
inject_poison and clear_poison. These are only exposed if all memdevs participating in the region support both inject and clear commands, ensuring consistent and reliable behavior across multi-device regions.

If tracing is enabled, these operations are logged as cxl_poison events in /sys/kernel/tracing/trace.

The ABI documentation warns users of the significant risks that come with using these capabilities.

---

Changes in v4: none

 Documentation/ABI/testing/debugfs-cxl |  87 +++++++++++++++++
 drivers/cxl/core/core.h               |   4 +
 drivers/cxl/core/memdev.c             |   8 ++
 drivers/cxl/core/region.c             | 131 +++++++++++++++++++++++++-
 4 files changed, 227 insertions(+), 3 deletions(-)
Documentation/ABI/testing/debugfs-cxl +87 -0
index 12488c14be64..e1f69e727190 100644
@@ -19,6 +19,20 @@ Description:
 		is returned to the user. The inject_poison attribute is only
 		visible for devices supporting the capability.
 
+		TEST-ONLY INTERFACE: This interface is intended for testing
+		and validation purposes only. It is not a data repair mechanism
+		and should never be used on production systems or live data.
+
+		DATA LOSS RISK: For CXL persistent memory (PMEM) devices,
+		poison injection can result in permanent data loss. Injected
+		poison may render data permanently inaccessible even after
+		clearing, as the clear operation writes zeros and does not
+		recover original data.
+
+		SYSTEM STABILITY RISK: For volatile memory, poison injection
+		can cause kernel crashes, system instability, or unpredictable
+		behavior if the poisoned addresses are accessed by running code
+		or critical kernel structures.
 
 What:		/sys/kernel/debug/memX/clear_poison
 Date:		April, 2023
@@ -35,6 +49,79 @@ Description:
 		The clear_poison attribute is only visible for devices
 		supporting the capability.
 
+		TEST-ONLY INTERFACE: This interface is intended for testing
+		and validation purposes only. It is not a data repair mechanism
+		and should never be used on production systems or live data.
+
+		CLEAR IS NOT DATA RECOVERY: This operation writes zeros to the
+		specified address range and removes the address from the poison
+		list. It does NOT recover or restore original data that may have
+		been present before poison injection. Any original data at the
+		cleared address is permanently lost and replaced with zeros.
+
+		CLEAR IS NOT A REPAIR MECHANISM: This interface is for testing
+		purposes only and should not be used as a data repair tool.
+		Clearing poison is fundamentally different from data recovery
+		or error correction.
+
+What:		/sys/kernel/debug/cxl/regionX/inject_poison
+Date:		August, 2025
+Contact:	[email protected]
+Description:
+		(WO) When a Host Physical Address (HPA) is written to this
+		attribute, the region driver translates it to a Device
+		Physical Address (DPA) and identifies the corresponding
+		memdev. It then sends an inject poison command to that memdev
+		at the translated DPA. Refer to the memdev ABI entry at:
+		/sys/kernel/debug/cxl/memX/inject_poison for the detailed
+		behavior. This attribute is only visible if all memdevs
+		participating in the region support both inject and clear
+		poison commands.
+
+		TEST-ONLY INTERFACE: This interface is intended for testing
+		and validation purposes only. It is not a data repair mechanism
+		and should never be used on production systems or live data.
+
+		DATA LOSS RISK: For CXL persistent memory (PMEM) devices,
+		poison injection can result in permanent data loss. Injected
+		poison may render data permanently inaccessible even after
+		clearing, as the clear operation writes zeros and does not
+		recover original data.
+
+		SYSTEM STABILITY RISK: For volatile memory, poison injection
+		can cause kernel crashes, system instability, or unpredictable
+		behavior if the poisoned addresses are accessed by running code
+		or critical kernel structures.
+
+What:		/sys/kernel/debug/cxl/regionX/clear_poison
+Date:		August, 2025
+Contact:	[email protected]
+Description:
+		(WO) When a Host Physical Address (HPA) is written to this
+		attribute, the region driver translates it to a Device
+		Physical Address (DPA) and identifies the corresponding
+		memdev. It then sends a clear poison command to that memdev
+		at the translated DPA. Refer to the memdev ABI entry at:
+		/sys/kernel/debug/cxl/memX/clear_poison for the detailed
+		behavior. This attribute is only visible if all memdevs
+		participating in the region support both inject and clear
+		poison commands.
+
+		TEST-ONLY INTERFACE: This interface is intended for testing
+		and validation purposes only. It is not a data repair mechanism
+		and should never be used on production systems or live data.
+
+		CLEAR IS NOT DATA RECOVERY: This operation writes zeros to the
+		specified address range and removes the address from the poison
+		list. It does NOT recover or restore original data that may have
+		been present before poison injection. Any original data at the
+		cleared address is permanently lost and replaced with zeros.
+
+		CLEAR IS NOT A REPAIR MECHANISM: This interface is for testing
+		purposes only and should not be used as a data repair tool.
+		Clearing poison is fundamentally different from data recovery
+		or error correction.
+
 What:		/sys/kernel/debug/cxl/einj_types
 Date:		January, 2024
 KernelVersion:	v6.9
drivers/cxl/core/core.h +4 -0
index 2669f251d677..eac8cc1bdaa0 100644
@@ -135,6 +135,10 @@ enum cxl_poison_trace_type {
 	CXL_POISON_TRACE_CLEAR,
 };
 
+enum poison_cmd_enabled_bits;
+bool cxl_memdev_has_poison_cmd(struct cxl_memdev *cxlmd,
+			       enum poison_cmd_enabled_bits cmd);
+
 long cxl_pci_get_latency(struct pci_dev *pdev);
 int cxl_pci_get_bandwidth(struct pci_dev *pdev, struct access_coordinate *c);
 int cxl_update_hmat_access_coordinates(int nid, struct cxl_region *cxlr,
drivers/cxl/core/memdev.c +8 -0
index 90d3390d9c7c..e370d733e440 100644
@@ -200,6 +200,14 @@ static ssize_t security_erase_store(struct device *dev,
 static struct device_attribute dev_attr_security_erase =
 	__ATTR(erase, 0200, NULL, security_erase_store);
 
+bool cxl_memdev_has_poison_cmd(struct cxl_memdev *cxlmd,
+			       enum poison_cmd_enabled_bits cmd)
+{
+	struct cxl_memdev_state *mds = to_cxl_memdev_state(cxlmd->cxlds);
+
+	return test_bit(cmd, mds->poison.enabled_cmds);
+}
+
 static int cxl_get_poison_by_memdev(struct cxl_memdev *cxlmd)
 {
 	struct cxl_dev_state *cxlds = cxlmd->cxlds;
drivers/cxl/core/region.c +128 -3

161 lines not on this page · raw message

Signature · 1 line
2.37.3