Skip to contentSkip to conversation

ACPI: APEI: GHES: Collection of fixes for issues reported by sashiko

PATCH Superseded Dave Jiang 3 participants 11 messages Jul 9 – Jul 10

Subsystems (3)

v1 of 6 Superseded Latest is v6 →

lore
Share & tools
lore

Dave Jiang

author maintainer
#0 lore raw

cover ACPI: APEI: GHES: Collection of fixes for issues reported by sashiko

A collection of fixes for pre-existing issues reported by sashiko-bot while reviewing patches.

1/4: Use spin_lock_irqsave() instead of spin_lock() for register/unregister
of CXL CPER work.
2/4: Bound CXL check the firmware section length before copy. 3/4: Fix CONFIG_ACPI_APEI_PCIEAER guard typo in extlog.c.
4/4: Validate CXL protocol error section length before RAS capability copy

Dave Jiang (4):
ACPI: APEI: GHES: Use spin_lock_irqsave() for CXL CPER work locks
ACPI: APEI: GHES: Bound CXL event record copy to the firmware section
length
ACPI: extlog: Fix CONFIG_ACPI_APEI_PCIEAER guard typo
ACPI: APEI: GHES: Validate CXL protocol error section length before
RAS cap copy

 drivers/acpi/acpi_extlog.c       | 11 +++++-----
 drivers/acpi/apei/ghes.c         | 37 ++++++++++++++++++++++----------
 drivers/acpi/apei/ghes_helpers.c | 15 ++++++++++++-
 include/cxl/event.h              |  4 ++--
 4 files changed, 48 insertions(+), 19 deletions(-)

base-commit: 8cdeaa50eae8dad34885515f62559ee83e7e8dda

Signature · 1 line
2.54.0
author maintainer
#1 lore raw

PATCH 1/4 ACPI: APEI: GHES: Use spin_lock_irqsave() for CXL CPER work locks

cxl_cper_post_event() and cxl_cper_post_prot_err() acquire cxl_cper_work_lock and cxl_cper_prot_err_work_lock with spin_lock_irqsave(). These producers run from ghes_do_proc() in hard IRQ context (the GHES IRQ handler) and, for SEA/NMI notifications, from NMI context.

The register/unregister helpers take the same two locks with a plain spin_lock() from process context (cxl_ras_init()/cxl_ras_exit() at module load and unload). If a GHES error interrupt is delivered to a CPU that is holding one of these locks via the plain spin_lock(), the interrupt handler's spin_lock_irqsave() on the same lock spins forever against a lock the interrupted thread can no longer release.

Take these locks with spin_lock_irqsave() on the register/unregister paths as well.

Fixes 5e4a264bf8b5 ("acpi/ghes: Process CXL Component Events")
Fixes 36f257e3b0ba ("acpi/ghes, cxl/pci: Process CXL CPER Protocol Errors")
Assisted-by Claude:claude-opus-4-8
---
 drivers/acpi/apei/ghes.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/acpi/apei/ghes.c +4 -4
index 3236a3ce79d6..9b6917698ed2 100644
@@ -783,7 +783,7 @@ int cxl_cper_register_prot_err_work(struct work_struct *work)
 	if (cxl_cper_prot_err_work)
 		return -EINVAL;
 
-	guard(spinlock)(&cxl_cper_prot_err_work_lock);
+	guard(spinlock_irqsave)(&cxl_cper_prot_err_work_lock);
 	cxl_cper_prot_err_work = work;
 	return 0;
 }
@@ -794,7 +794,7 @@ int cxl_cper_unregister_prot_err_work(struct work_struct *work)
 	if (cxl_cper_prot_err_work != work)
 		return -EINVAL;
 
-	guard(spinlock)(&cxl_cper_prot_err_work_lock);
+	guard(spinlock_irqsave)(&cxl_cper_prot_err_work_lock);
 	cxl_cper_prot_err_work = NULL;
 	return 0;
 }
@@ -852,7 +852,7 @@ int cxl_cper_register_work(struct work_struct *work)
 	if (cxl_cper_work)
 		return -EINVAL;
 
-	guard(spinlock)(&cxl_cper_work_lock);
+	guard(spinlock_irqsave)(&cxl_cper_work_lock);
 	cxl_cper_work = work;
 	return 0;
 }
@@ -863,7 +863,7 @@ int cxl_cper_unregister_work(struct work_struct *work)
 	if (cxl_cper_work != work)
 		return -EINVAL;
 
-	guard(spinlock)(&cxl_cper_work_lock);
+	guard(spinlock_irqsave)(&cxl_cper_work_lock);
 	cxl_cper_work = NULL;
 	return 0;
 }
Signature · 1 line
2.54.0
#2 lore raw

Re: [PATCH 1/4] ACPI: APEI: GHES: Use spin_lock_irqsave() for CXL CPER work locks

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Missing bounds check on CPER record length and dvsec_len allows an out-of-bounds memory read that can be leaked to userspace via trace events. --

--- Patch [1]: [PATCH 1/4] ACPI: APEI: GHES: Use spin_lock_irqsave() for CXL CPER work locks --- commit 92925c77a9085c947d2733f2a851c54ec53288f8
Author: Dave Jiang <[email protected]>

ACPI: APEI: GHES: Use spin_lock_irqsave() for CXL CPER work locks

This patch switches the cxl_cper_work_lock and cxl_cper_prot_err_work_lock from plain spinlocks to spin_lock_irqsave(). This prevents a potential deadlock when a GHES error interrupt interrupts a thread holding the lock in process context.

Quoted 4 lines
drivers/acpi/apei/ghes.c +0 -0
index 3236a3ce79d6b..9b6917698ed24 100644

[ ... ]

drivers/acpi/apei/ghes.c int cxl_cper_unregister_work(struct work_struct *work) · line 869

Quoted 9 lines
drivers/acpi/apei/ghes.c +1 -1
@@ -863,7 +863,7 @@ int cxl_cper_unregister_work(struct work_struct *work)
 	if (cxl_cper_work != work)
 		return -EINVAL;
 
-	guard(spinlock)(&cxl_cper_work_lock);
+	guard(spinlock_irqsave)(&cxl_cper_work_lock);
 	cxl_cper_work = NULL;
 	return 0;
 }

[Severity: High]
This is a pre-existing issue, but looking at the CXL CPER work queue handling, is there a potential out-of-bounds read when processing malformed CPER records?

In drivers/acpi/apei/ghes_helpers.c:cxl_cper_setup_prot_err_work_data(), the cap_start pointer is computed using an untrusted dvsec_len from the firmware:

    dvsec_start = (u8 *)(prot_err + 1);
    cap_start = dvsec_start + prot_err->dvsec_len;

    memcpy(&wd->ras_cap, cap_start, sizeof(wd->ras_cap));

Since dvsec_len is provided by firmware, could a maliciously large value cause cap_start to point outside the mapped CPER record bounds?

Similarly, in drivers/acpi/apei/ghes.c:cxl_cper_post_event(), the code copies the event record:

    memcpy(&wd.rec, rec, sizeof(wd.rec));

If a hardware device provides a malformed record with a small error_data_length or an unusually large dvsec_len, could this unconditional memcpy() read unmapped memory or leak kernel memory contents to userspace via the kfifo trace events?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Reviewed-by on v1 1/4
#3 lore raw

Re: [PATCH 1/4] ACPI: APEI: GHES: Use spin_lock_irqsave() for CXL CPER work locks

In reply to Dave Jiang
Quoted 16 lines · Dave Jiang

cxl_cper_post_event() and cxl_cper_post_prot_err() acquire cxl_cper_work_lock and cxl_cper_prot_err_work_lock with spin_lock_irqsave(). These producers run from ghes_do_proc() in hard IRQ context (the GHES IRQ handler) and, for SEA/NMI notifications, from NMI context.

The register/unregister helpers take the same two locks with a plain spin_lock() from process context (cxl_ras_init()/cxl_ras_exit() at module load and unload). If a GHES error interrupt is delivered to a CPU that is holding one of these locks via the plain spin_lock(), the interrupt handler's spin_lock_irqsave() on the same lock spins forever against a lock the interrupted thread can no longer release.

Take these locks with spin_lock_irqsave() on the register/unregister paths as well.

Quoted 6 lines
Fixes 5e4a264bf8b5 ("acpi/ghes: Process CXL Component Events")
Fixes 36f257e3b0ba ("acpi/ghes, cxl/pci: Process CXL CPER Protocol Errors")
Assisted-by Claude:claude-opus-4-8

---

Since the Sashiko review is for a different issue and this makes sense to me:

author maintainer
#4 lore raw

PATCH 2/4 ACPI: APEI: GHES: Bound CXL event record copy to the firmware section length

In reply to Dave Jiang

cxl_cper_post_event() blindly copies a fixed
sizeof(struct cxl_cper_event_rec) out of the firmware-provided CPER section without checking the length. Pass gdata->err_data_length into the function and reject a section too small to hold the record before the copy.

Fixes 5e4a264bf8b5 ("acpi/ghes: Process CXL Component Events")
Assisted-by Claude:claude-opus-4-8
---
 drivers/acpi/apei/ghes.c | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)
drivers/acpi/apei/ghes.c +18 -4
index 9b6917698ed2..7b465e2afcb0 100644
@@ -815,10 +815,21 @@ static DEFINE_SPINLOCK(cxl_cper_work_lock);
 struct work_struct *cxl_cper_work;
 
 static void cxl_cper_post_event(enum cxl_event_type event_type,
-				struct cxl_cper_event_rec *rec)
+				struct cxl_cper_event_rec *rec, u32 len)
 {
 	struct cxl_cper_work_data wd;
 
+	/*
+	 * The record is copied whole below, so the firmware-provided section
+	 * must be at least as large as the record. cper_estatus_check() only
+	 * validates the section fits the error block, not that it satisfies a
+	 * given section type's size, so guard the fixed-size copy here.
+	 */
+	if (len < sizeof(*rec)) {
+		pr_err(FW_WARN "CXL CPER section too small (%u)\n", len);
+		return;
+	}
+
 	if (rec->hdr.length <= sizeof(rec->hdr) ||
 	    rec->hdr.length > sizeof(*rec)) {
 		pr_err(FW_WARN "CXL CPER Invalid section length (%u)\n",
@@ -949,15 +960,18 @@ static void ghes_do_proc(struct ghes *ghes,
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_GEN_MEDIA_GUID)) {
 			struct cxl_cper_event_rec *rec = acpi_hest_get_payload(gdata);
 
-			cxl_cper_post_event(CXL_CPER_EVENT_GEN_MEDIA, rec);
+			cxl_cper_post_event(CXL_CPER_EVENT_GEN_MEDIA, rec,
+					    gdata->error_data_length);
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_DRAM_GUID)) {
 			struct cxl_cper_event_rec *rec = acpi_hest_get_payload(gdata);
 
-			cxl_cper_post_event(CXL_CPER_EVENT_DRAM, rec);
+			cxl_cper_post_event(CXL_CPER_EVENT_DRAM, rec,
+					    gdata->error_data_length);
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_MEM_MODULE_GUID)) {
 			struct cxl_cper_event_rec *rec = acpi_hest_get_payload(gdata);
 
-			cxl_cper_post_event(CXL_CPER_EVENT_MEM_MODULE, rec);
+			cxl_cper_post_event(CXL_CPER_EVENT_MEM_MODULE, rec,
+					    gdata->error_data_length);
 		} else {
 			void *err = acpi_hest_get_payload(gdata);
 
Signature · 1 line
2.54.0
#5 lore raw

Re: [PATCH 2/4] ACPI: APEI: GHES: Bound CXL event record copy to the firmware section length

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Incomplete fix for out-of-bounds read vulnerability when parsing CXL Protocol Error and other CPER sections. --

--- Patch [2]: [PATCH 2/4] ACPI: APEI: GHES: Bound CXL event record copy to the firmware section length --- commit 4eb9f7aafa36f7ab79ba6d995ac9206e430ee1b0
Author: Dave Jiang <[email protected]>

ACPI: APEI: GHES: Bound CXL event record copy to the firmware section length

This commit addresses an issue where cxl_cper_post_event() copies a fixed-size record from the firmware-provided CPER section without validating the section length. It passes error_data_length to the function and rejects sections that are too small.

Quoted 4 lines
drivers/acpi/apei/ghes.c +0 -0
index 9b6917698ed24..7b465e2afcb04 100644

[ ... ]

drivers/acpi/apei/ghes.c static void ghes_do_proc(struct ghes *ghes,

Quoted 1 line
drivers/acpi/apei/ghes.c +0 -0
@@ -949,15 +960,18 @@ static void ghes_do_proc(struct ghes *ghes,

[Severity: High]
This isn't a bug introduced by this patch, but while bounds checking is being added for CXL media and DRAM events here, the same vulnerability appears to remain for CXL Protocol Error and other CPER sections nearby.

Does the code pass the payload to cxl_cper_post_prot_err() without validating if gdata->error_data_length is sufficient for the struct?

This could lead to an out-of-bounds memory read in cxl_cper_setup_prot_err_work_data() when copying the structs:

drivers/acpi/apei/ghes_helpers.c:cxl_cper_setup_prot_err_work_data() {
    ...
    cap_start = dvsec_start + prot_err->dvsec_len;
    memcpy(&wd->ras_cap, cap_start, sizeof(wd->ras_cap));
    ...
}

If the firmware provides a CPER_SEC_CXL_PROT_ERR section smaller than expected, could dvsec_len be read from uninitialized memory and used as an offset, potentially triggering a page fault?

Similar missing length validation also appears to be present when calling extlog_cxl_cper_handle_prot_err() in drivers/acpi/acpi_extlog.c:extlog_print(), and when accessing mem_err payloads in ghes_handle_memory_failure():

drivers/acpi/apei/ghes.c:ghes_handle_memory_failure() {
    ...
    if (!(mem_err->validation_bits & CPER_MEM_VALID_PA))
    ...
}
Quoted 12 lines
drivers/acpi/apei/ghes.c +4 -2
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_GEN_MEDIA_GUID)) {
 			struct cxl_cper_event_rec *rec = acpi_hest_get_payload(gdata);
 
-			cxl_cper_post_event(CXL_CPER_EVENT_GEN_MEDIA, rec);
+			cxl_cper_post_event(CXL_CPER_EVENT_GEN_MEDIA, rec,
+					    gdata->error_data_length);
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_DRAM_GUID)) {
 			struct cxl_cper_event_rec *rec = acpi_hest_get_payload(gdata);
 
-			cxl_cper_post_event(CXL_CPER_EVENT_DRAM, rec);
+			cxl_cper_post_event(CXL_CPER_EVENT_DRAM, rec,
+					    gdata->error_data_length);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
author maintainer
#6 lore raw

PATCH 3/4 ACPI: extlog: Fix CONFIG_ACPI_APEI_PCIEAER guard typo

In reply to Dave Jiang

"#ifdef ACPI_APEI_PCIEAER" is incorrect. It should be "#ifdef CONFIG_ACPI_APEI_PCIEAE". Currently the section wrapped will always evalutes to the stub because of this. Use the correct CONFIG_ prefixed symbol so the code is actually built.

Fixes 95350effc3ad ("ACPI: extlog: Trace CPER CXL Protocol Error Section")
Assisted-by Claude:claude-opus-4-8
---
 drivers/acpi/acpi_extlog.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/acpi/acpi_extlog.c +2 -2
index 7ad3b36013cc..60c2b90adf0a 100644
@@ -136,7 +136,7 @@ static int print_extlog_rcd(const char *pfx,
 static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 			      int severity)
 {
-#ifdef ACPI_APEI_PCIEAER
+#ifdef CONFIG_ACPI_APEI_PCIEAER
 	struct aer_capability_regs *aer;
 	struct pci_dev *pdev;
 	unsigned int devfn;
@@ -167,7 +167,7 @@ static void
 extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
 				int severity)
 {
-#ifdef ACPI_APEI_PCIEAER
+#ifdef CONFIG_ACPI_APEI_PCIEAER
 	struct cxl_cper_prot_err_work_data wd;
 
 	if (cxl_cper_sec_prot_err_valid(prot_err))
Signature · 1 line
2.54.0
#7 lore raw

Re: [PATCH 3/4] ACPI: extlog: Fix CONFIG_ACPI_APEI_PCIEAER guard typo

Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:

Pre-existing issues:
- [High] Missing bounds checking on firmware-provided CPER data lengths before accessing or copying PCIe and CXL protocol error structures.
- [High] Calling cxl_cper_handle_prot_err synchronously creates an AB-BA lock ordering deadlock with PCI device_lock and x86_mce_decoder_chain.rwsem.
- [High] Casting the raw hardware aer_info buffer directly to struct aer_capability_regs populates software-only metadata with unvalidated hardware data, triggering out-of-bounds array reads.
- [Low] Directly calling cxl_cper_handle_prot_err() causes a linkage failure when CONFIG_ACPI_EXTLOG=y and CONFIG_CXL_BUS is configured as a module.
--

--- Patch [3]: [PATCH 3/4] ACPI: extlog: Fix CONFIG_ACPI_APEI_PCIEAER guard typo --- commit d29f05a05af2b3add29bfc216f34a5de8a53eb2a
Author: Dave Jiang <[email protected]>

ACPI: extlog: Fix CONFIG_ACPI_APEI_PCIEAER guard typo

This patch fixes a typo in the preprocessor guard for ACPI_APEI_PCIEAER. It changes the guard to use the correct CONFIG_ prefix, ensuring the enclosed PCIe AER and CXL CPER handling code is actually built when enabled.

Quoted 4 lines
drivers/acpi/acpi_extlog.c +0 -0
index 7ad3b36013cc6..60c2b90adf0ae 100644

[ ... ]

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 139

Quoted 6 lines
drivers/acpi/acpi_extlog.c +1 -1
@@ -136,7 +136,7 @@ static int print_extlog_rcd(const char *pfx,
 static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 			      int severity)
 {
-#ifdef ACPI_APEI_PCIEAER
+#ifdef CONFIG_ACPI_APEI_PCIEAER

[Severity: High]
This isn't a bug introduced by this patch, but activating this code exposes a missing bounds check on firmware-provided CPER data lengths.

If the firmware provides a CPER record where gdata->error_data_length is smaller than the expected payload structs, or provides a maliciously large dvsec_len for a CXL Protocol Error, is there a risk of an out-of-bounds memory read?

When extlog_print() extracts payloads and dispatches to handlers without verifying gdata->error_data_length >= sizeof(...), the code casts the payload and executes cxl_cper_setup_prot_err_work_data(). This function performs pointer arithmetic using the unvalidated dvsec_len and executes a fixed-size memcpy(), which could lead to a kernel panic or data leak.

[Severity: High]
This is a pre-existing issue, but fixing the typo activates a code path in extlog_print_pcie() that can trigger out-of-bounds array reads.

Does casting the raw 96-byte aer_info hardware buffer directly to struct aer_capability_regs * populate software-only metadata with unvalidated hardware data?

In Linux, struct aer_capability_regs embeds struct pcie_tlp_log, placing software metadata (header_len and flit) at offset 0x54. Casting raw registers populates these software fields with unvalidated hardware state.

If flit evaluates to true and header_len is read as a large value (up to 255), pcie_print_tlp_log() will iterate up to 255 times, reading out-of-bounds past the 14-element log->dw array. This read traverses past the aer_info struct and can cross into unmapped memory.

Quoted 2 lines
drivers/acpi/acpi_extlog.c +0 -0
 	struct aer_capability_regs *aer;
 	struct pci_dev *pdev;

[ ... ]

drivers/acpi/acpi_extlog.c static void · line 170

Quoted 6 lines
drivers/acpi/acpi_extlog.c +1 -1
@@ -167,7 +167,7 @@ static void
 extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
 				int severity)
 {
-#ifdef ACPI_APEI_PCIEAER
+#ifdef CONFIG_ACPI_APEI_PCIEAER

[Severity: Low]
This is a pre-existing issue, but activating this previously dead code causes a linkage failure when CONFIG_ACPI_EXTLOG=y and CONFIG_CXL_BUS is configured as a module.

Because ACPI_EXTLOG does not depend on CXL_BUS in its Kconfig, is it possible that the direct symbol call to cxl_cper_handle_prot_err() from the built-in acpi_extlog.o to the module cxl_core.ko cannot be resolved? This would result in an undefined reference to cxl_cper_handle_prot_err.

[Severity: High]
This isn't a bug introduced by this patch, but activating this code exposes a potential AB-BA lock ordering deadlock with the PCI device_lock and x86_mce_decoder_chain.rwsem.

When extlog_cxl_cper_handle_prot_err() is called synchronously while the notifier rwsem is held, does this create a circular dependency?

extlog_print() is registered as an MCE decoder and is called with down_read(&x86_mce_decoder_chain.rwsem) held. It now synchronously calls cxl_cper_handle_prot_err(), which acquires the PCI device lock via guard(device)(&pdev->dev), establishing the lock order:
x86_mce_decoder_chain.rwsem -> device_lock.

Conversely, during system initialization, drivers (like AMD GPU or CXL via devm_cxl_register_mce_notifier()) execute .probe with the PCI device_lock held and call mce_register_decode_chain(), taking down_write(&x86_mce_decoder_chain.rwsem). Could this cause a hard deadlock if an MCE occurs during driver probing or initialization?

Quoted 3 lines
drivers/acpi/acpi_extlog.c +0 -0
 	struct cxl_cper_prot_err_work_data wd;
 
 	if (cxl_cper_sec_prot_err_valid(prot_err))
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
Reviewed-by on v1 3/4
#8 lore raw

Re: [PATCH 3/4] ACPI: extlog: Fix CONFIG_ACPI_APEI_PCIEAER guard typo

In reply to Dave Jiang
Quoted 26 lines · Dave Jiang

"#ifdef ACPI_APEI_PCIEAER" is incorrect. It should be "#ifdef CONFIG_ACPI_APEI_PCIEAE". Currently the section wrapped will always evalutes to the stub because of this. Use the correct CONFIG_ prefixed symbol so the code is actually built.

Fixes 95350effc3ad ("ACPI: extlog: Trace CPER CXL Protocol Error Section")
Assisted-by Claude:claude-opus-4-8
---
 drivers/acpi/acpi_extlog.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/acpi/acpi_extlog.c +1 -1
index 7ad3b36013cc..60c2b90adf0a 100644
@@ -136,7 +136,7 @@ static int print_extlog_rcd(const char *pfx,
 static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 			      int severity)
 {
-#ifdef ACPI_APEI_PCIEAER
+#ifdef CONFIG_ACPI_APEI_PCIEAER
 	struct aer_capability_regs *aer;
 	struct pci_dev *pdev;
 	unsigned int devfn;

drivers/acpi/acpi_extlog.c static void · line 170

Quoted 5 lines
drivers/acpi/acpi_extlog.c +1 -1
@@ -167,7 +167,7 @@ static void
 extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
 				int severity)
 {
-#ifdef ACPI_APEI_PCIEAER
+#ifdef CONFIG_ACPI_APEI_PCIEAER

Seems like Sashiko is complaining about bugs that it already highlighted in patch 2/4 (I think?). But it did point out a linker error here that needs to be fixed. Should be able to fix it by adding CONFIG_CXL_BUS to the #ifdef or a new nested #ifdef CONFIG_CXL_BUS. Either works for me, so with that applied:

Quoted 3 lines
drivers/acpi/acpi_extlog.c +0 -0
 	struct cxl_cper_prot_err_work_data wd;
 
 	if (cxl_cper_sec_prot_err_valid(prot_err))
author maintainer
#9 lore raw

PATCH 4/4 ACPI: APEI: GHES: Validate CXL protocol error section length before RAS cap copy

In reply to Dave Jiang

cxl_cper_setup_prot_err_work_data() locates the RAS Capability block by skipping a firmware-controlled DVSEC:

	dvsec_start = (u8 *)(prot_err + 1);
	cap_start = dvsec_start + prot_err->dvsec_len;
	memcpy(&wd->ras_cap, cap_start, sizeof(wd->ras_cap));

prot_err->dvsec_len is taken from the CPER section and is never validated.

Add the check to cxl_cper_sec_prot_err_valid() to enforce the check and avoid copying out of bounds.

Fixes 315c2f0b53ba ("acpi/ghes, cper: Recognize and cache CXL Protocol errors")
Assisted-by Claude:claude-opus-4-8
---
 drivers/acpi/acpi_extlog.c       |  7 ++++---
 drivers/acpi/apei/ghes.c         |  7 ++++---
 drivers/acpi/apei/ghes_helpers.c | 15 ++++++++++++++-
 include/cxl/event.h              |  4 ++--
 4 files changed, 24 insertions(+), 9 deletions(-)
drivers/acpi/acpi_extlog.c +4 -3
index 60c2b90adf0a..c65b54740bf5 100644
@@ -165,12 +165,12 @@ static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 
 static void
 extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
-				int severity)
+				int severity, u32 len)
 {
 #ifdef CONFIG_ACPI_APEI_PCIEAER
 	struct cxl_cper_prot_err_work_data wd;
 
-	if (cxl_cper_sec_prot_err_valid(prot_err))
+	if (cxl_cper_sec_prot_err_valid(prot_err, len))
 		return;
 
 	if (cxl_cper_setup_prot_err_work_data(&wd, prot_err, severity))
@@ -236,7 +236,8 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 				acpi_hest_get_payload(gdata);
 
 			extlog_cxl_cper_handle_prot_err(prot_err,
-							gdata->error_severity);
+							gdata->error_severity,
+							gdata->error_data_length);
 		} else if (guid_equal(sec_type, &CPER_SEC_PCIE)) {
 			struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata);
 
drivers/acpi/apei/ghes.c +4 -3
index 7b465e2afcb0..5af794fed242 100644
@@ -753,12 +753,12 @@ static DEFINE_SPINLOCK(cxl_cper_prot_err_work_lock);
 struct work_struct *cxl_cper_prot_err_work;
 
 static void cxl_cper_post_prot_err(struct cxl_cper_sec_prot_err *prot_err,
-				   int severity)
+				   int severity, u32 len)
 {
 #ifdef CONFIG_ACPI_APEI_PCIEAER
 	struct cxl_cper_prot_err_work_data wd;
 
-	if (cxl_cper_sec_prot_err_valid(prot_err))
+	if (cxl_cper_sec_prot_err_valid(prot_err, len))
 		return;
 
 	guard(spinlock_irqsave)(&cxl_cper_prot_err_work_lock);
@@ -956,7 +956,8 @@ static void ghes_do_proc(struct ghes *ghes,
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_PROT_ERR)) {
 			struct cxl_cper_sec_prot_err *prot_err = acpi_hest_get_payload(gdata);
 
-			cxl_cper_post_prot_err(prot_err, gdata->error_severity);
+			cxl_cper_post_prot_err(prot_err, gdata->error_severity,
+					       gdata->error_data_length);
 		} else if (guid_equal(sec_type, &CPER_SEC_CXL_GEN_MEDIA_GUID)) {
 			struct cxl_cper_event_rec *rec = acpi_hest_get_payload(gdata);
 
drivers/acpi/apei/ghes_helpers.c +14 -1
index bc7111b740af..8422f6daff86 100644
@@ -5,7 +5,7 @@
 #include <linux/aer.h>
 #include <cxl/event.h>
 
-int cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err)
+int cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err, u32 len)
 {
 	if (!(prot_err->valid_bits & PROT_ERR_VALID_AGENT_ADDRESS)) {
 		pr_err_ratelimited("CXL CPER invalid agent type\n");
@@ -23,6 +23,19 @@ int cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err)
 		return -EINVAL;
 	}
 
+	/*
+	 * The RAS Capability block follows a firmware-controlled DVSEC of
+	 * prot_err->dvsec_len bytes. Verify the header, the DVSEC and the RAS
+	 * Capability block all fit within the CPER section.
+	 */
+	if (sizeof(*prot_err) + prot_err->dvsec_len +
+	    sizeof(struct cxl_ras_capability_regs) > len) {
+		pr_err_ratelimited(FW_WARN
+				   "CXL CPER prot err section too small (%u)\n",
+				   len);
+		return -EINVAL;
+	}
+
 	if ((prot_err->agent_type == RCD || prot_err->agent_type == DEVICE ||
 	     prot_err->agent_type == LD || prot_err->agent_type == FMLD) &&
 	    !(prot_err->valid_bits & PROT_ERR_VALID_SERIAL_NUMBER))
include/cxl/event.h +2 -2
index ff97fea718d2..912305bee3bc 100644
@@ -321,13 +321,13 @@ static inline int cxl_cper_prot_err_kfifo_get(struct cxl_cper_prot_err_work_data
 #endif
 
 #ifdef CONFIG_ACPI_APEI_PCIEAER
-int cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err);
+int cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err, u32 len);
 int cxl_cper_setup_prot_err_work_data(struct cxl_cper_prot_err_work_data *wd,
 				      struct cxl_cper_sec_prot_err *prot_err,
 				      int severity);
 #else
 static inline int
-cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err)
+cxl_cper_sec_prot_err_valid(struct cxl_cper_sec_prot_err *prot_err, u32 len)
 {
 	return -EOPNOTSUPP;
 }
Signature · 1 line
2.54.0
Reviewed-by on v1 4/4
#10 lore raw

Re: [PATCH 4/4] ACPI: APEI: GHES: Validate CXL protocol error section length before RAS cap copy

Quoted 9 lines · Dave Jiang

cxl_cper_setup_prot_err_work_data() locates the RAS Capability block by skipping a firmware-controlled DVSEC:

	dvsec_start = (u8 *)(prot_err + 1);
	cap_start = dvsec_start + prot_err->dvsec_len;
	memcpy(&wd->ras_cap, cap_start, sizeof(wd->ras_cap));

prot_err->dvsec_len is taken from the CPER section and is never validated.

Quoted 8 lines

Add the check to cxl_cper_sec_prot_err_valid() to enforce the check and avoid copying out of bounds.

Fixes 315c2f0b53ba ("acpi/ghes, cper: Recognize and cache CXL Protocol errors")
Assisted-by Claude:claude-opus-4-8

---

You may want to check the compilation issue with the last patch isn't also present here. I think I've convinced myself it isn't, but it can't hurt. If it's fine: