Skip to contentSkip to conversation

ACPI / APEI: Boot Error Record Table processing was needlessly complicated

PATCH In review Luck, Tony 2 participants 6 messages Jun 15, 2017 – Jun 28, 2017

Subsystems (2)
ACPI ACPI APEI
lore
Share & tools
lore

Luck, Tony

author reviewer
#0 lore raw

PATCH v4 1/1 ACPI / APEI: Boot Error Record Table processing was needlessly complicated

From: Tony Luck <[email protected]>

Quoting version 6.1 of the ACPI specification. Section 18.3.1 "Boot Error Source" says:

The Boot Error Region is a range of addressable memory OSPM can access during initialization to determine if an unhandled error condition occurred. System firmware must report this memory range as firmware reserved. The format of the Boot Error Region follow that of an Error Status Block, this is defined in Section 18.3.2.7. The format of the error status block is described by Table 18-342.

This clarifies some points that were obfuscated in earlier versions. E.g. there is no longer a separate table to describe the format of the "Boot Error Region" (which was identical to the "Error Status Block"). Also saying "follow that of *an* Error Status Block" makes it clear that there is just one block (which can still contain multiple "Generic Error Data Entry structures").

The loop inside bert_print_all() is unnecessary (but probably harmless as the "while (remain > sizeof(struct acpi_bert_region))" loop should terminate after we skipped over the first entry.

We can drop the "bert_print_all()" function and just move the four relevant lines inline in "bert_init()".

Also this driver clears the "block_status" status field of the error status block in ACPI memory. I can see no justification for doing this in the ACPI spec. It is also highly kernel-ego-centric. There are user mode BERT harvesting tools that dig into /sys/firmware/acpi/tables/BERT which might like to see this summary value.

Drop the:

	estatus->block_status = 0;
Cc Len Brown <[email protected]>
Cc Huang Ying <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Tomasz Nowicki <[email protected]>
Cc Jonathan (Zhixiong) Zhang <[email protected]>
Cc Tyler Baicar <[email protected]>
Cc [email protected]
Signed-off-by Tony Luck <[email protected]>
---
 drivers/acpi/apei/bert.c | 59 ++++++++++--------------------------------------
 1 file changed, 12 insertions(+), 47 deletions(-)

v4: don't clear block_status

previous version was

Reviewed-by Borislav Petkov <[email protected]>
Tested-by Tyler Baicar <[email protected]>
drivers/acpi/apei/bert.c +12 -47
index 12771fcf0417..3b9b296dc84e 100644
@@ -34,50 +34,6 @@
 
 static int bert_disable;
 
-static void __init bert_print_all(struct acpi_bert_region *region,
-				  unsigned int region_len)
-{
-	struct acpi_hest_generic_status *estatus =
-		(struct acpi_hest_generic_status *)region;
-	int remain = region_len;
-	u32 estatus_len;
-
-	if (!estatus->block_status)
-		return;
-
-	while (remain > sizeof(struct acpi_bert_region)) {
-		if (cper_estatus_check(estatus)) {
-			pr_err(FW_BUG "Invalid error record.\n");
-			return;
-		}
-
-		estatus_len = cper_estatus_len(estatus);
-		if (remain < estatus_len) {
-			pr_err(FW_BUG "Truncated status block (length: %u).\n",
-			       estatus_len);
-			return;
-		}
-
-		pr_info_once("Error records from previous boot:\n");
-
-		cper_estatus_print(KERN_INFO HW_ERR, estatus);
-
-		/*
-		 * Because the boot error source is "one-time polled" type,
-		 * clear Block Status of current Generic Error Status Block,
-		 * once it's printed.
-		 */
-		estatus->block_status = 0;
-
-		estatus = (void *)estatus + estatus_len;
-		/* No more error records. */
-		if (!estatus->block_status)
-			return;
-
-		remain -= estatus_len;
-	}
-}
-
 static int __init setup_bert_disable(char *str)
 {
 	bert_disable = 1;

@@ -89,7 +45,7 @@ __setup("bert_disable", setup_bert_disable);
 static int __init bert_check_table(struct acpi_table_bert *bert_tab)
 {
 	if (bert_tab->header.length < sizeof(struct acpi_table_bert) ||
-	    bert_tab->region_length < sizeof(struct acpi_bert_region))
+	    bert_tab->region_length < sizeof(struct acpi_hest_generic_status))
 		return -EINVAL;
 
 	return 0;

@@ -98,7 +54,7 @@ static int __init bert_check_table(struct acpi_table_bert *bert_tab)
 static int __init bert_init(void)
 {
 	struct apei_resources bert_resources;
-	struct acpi_bert_region *boot_error_region;
+	struct acpi_hest_generic_status *boot_error_region;
 	struct acpi_table_bert *bert_tab;
 	unsigned int region_len;
 	acpi_status status;

@@ -138,7 +94,16 @@ static int __init bert_init(void)
 		goto out_fini;
 	boot_error_region = ioremap_cache(bert_tab->address, region_len);
 	if (boot_error_region) {
-		bert_print_all(boot_error_region, region_len);
+		if (boot_error_region->block_status) {
+			rc = cper_estatus_check(boot_error_region);
+			if (rc) {
+				pr_err(FW_BUG "Invalid error record.\n");
+				iounmap(boot_error_region);
+				return rc;
+			}
+			pr_info("Error records from previous boot:\n");
+			cper_estatus_print(KERN_INFO HW_ERR, boot_error_region);
+		}
 		iounmap(boot_error_region);
 	} else {
 		rc = -ENOMEM;
Signature · 1 line
2.11.0
#1 lore raw
Quoted 12 lines · Tony

From: Tony Luck <[email protected]>

Quoting version 6.1 of the ACPI specification. Section 18.3.1 "Boot Error Source" says:

The Boot Error Region is a range of addressable memory OSPM can access during initialization to determine if an unhandled error condition occurred. System firmware must report this memory range as firmware reserved. The format of the Boot Error Region follow that of an Error Status Block, this is defined in Section 18.3.2.7. The format of the error status block is described by Table 18-342.

Quoted 3 lines

This clarifies some points that were obfuscated in earlier versions. E.g. there is no longer a separate table to describe the format of the "Boot Error Region" (which was identical to the "Error Status Block").

Hmm, ok, so do we know whether some fw writers "misunderstood" the previous version of the spec and actually added that separate table and now we need to handle the different layout?

Probably not but I'm not going to be surprised someone did. At all.

Quoted 11 lines

Also saying "follow that of *an* Error Status Block" makes it clear that there is just one block (which can still contain multiple "Generic Error Data Entry structures").

The loop inside bert_print_all() is unnecessary (but probably harmless as the "while (remain > sizeof(struct acpi_bert_region))" loop should terminate after we skipped over the first entry.

We can drop the "bert_print_all()" function and just move the four relevant lines inline in "bert_init()".

Quoted 8 lines

Also this driver clears the "block_status" status field of the error status block in ACPI memory. I can see no justification for doing this in the ACPI spec. It is also highly kernel-ego-centric. There are user mode BERT harvesting tools that dig into /sys/firmware/acpi/tables/BERT which might like to see this summary value.

Drop the:

	estatus->block_status = 0;

This whole treatise of the clearing of block_status should most likely be a separate patch, though. As it is clearly a fix. Probably make it the first patch and the second one removing bert_print_all().

Something like that.

Thanks.

Signature · 5 lines
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
--
author reviewer
#2 lore raw

v5 cover Fixes for ACPI BERT table

From: Tony Luck <[email protected]>

At Boris' suggestion I've split this into two parts for version 5. First one stops the driver from clearing the block_status field as:
1) it is unnecessary
2) not mentioned in the spec
3) may upset user mode tools that extract information from the BERT table.

Second is the code simplification that you've seen before in ealier versions of this patch.

Cc Len Brown <[email protected]>
Cc Huang Ying <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Tomasz Nowicki <[email protected]>
Cc Jonathan (Zhixiong) Zhang <[email protected]>
Cc Tyler Baicar <[email protected]>
Cc [email protected]

Tony Luck (2):
ACPI / APEI / BERT: Don't clear "block_status" in header ACPI / APEI: Boot Error Record Table processing was needlessly
complicated

 drivers/acpi/apei/bert.c | 59 ++++++++++--------------------------------------
 1 file changed, 12 insertions(+), 47 deletions(-)
Signature · 1 line
2.11.0
author reviewer
#3 lore raw

PATCH v5 1/2 ACPI / APEI / BERT: Don't clear "block_status" in header

From: Tony Luck <[email protected]>

Although the ACPI specification says this is a "one time polled" error source, it makes no mention that the OS should stomp on fields in the error status block to enforce this.

There are user mode BERT harvesting tools that dig into /sys/firmware/acpi/tables/BERT which might like to see this summary value.

Drop the:

	estatus->block_status = 0;
Cc Len Brown <[email protected]>
Cc Huang Ying <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Tomasz Nowicki <[email protected]>
Cc Jonathan (Zhixiong) Zhang <[email protected]>
Cc Tyler Baicar <[email protected]>
Cc [email protected]
Signed-off-by Tony Luck <[email protected]>
---
 drivers/acpi/apei/bert.c | 7 -------
 1 file changed, 7 deletions(-)
drivers/acpi/apei/bert.c +0 -7
index 12771fcf0417..9197555dc3c7 100644
@@ -62,13 +62,6 @@ static void __init bert_print_all(struct acpi_bert_region *region,
 
 		cper_estatus_print(KERN_INFO HW_ERR, estatus);
 
-		/*
-		 * Because the boot error source is "one-time polled" type,
-		 * clear Block Status of current Generic Error Status Block,
-		 * once it's printed.
-		 */
-		estatus->block_status = 0;
-
 		estatus = (void *)estatus + estatus_len;
 		/* No more error records. */
 		if (!estatus->block_status)
Signature · 1 line
2.11.0
author reviewer
#4 lore raw

PATCH v5 2/2 ACPI / APEI: Boot Error Record Table processing was needlessly complicated

In reply to Luck, Tony

From: Tony Luck <[email protected]>

Quoting version 6.1 of the ACPI specification. Section 18.3.1 "Boot Error Source" says:

The Boot Error Region is a range of addressable memory OSPM can access during initialization to determine if an unhandled error condition occurred. System firmware must report this memory range as firmware reserved. The format of the Boot Error Region follow that of an Error Status Block, this is defined in Section 18.3.2.7. The format of the error status block is described by Table 18-342.

This clarifies some points that were obfuscated in earlier versions. E.g. there is no longer a separate table to describe the format of the "Boot Error Region" (which was identical to the "Error Status Block"). Also saying "follow that of *an* Error Status Block" makes it clear that there is just one block (which can still contain multiple "Generic Error Data Entry structures").

The loop inside bert_print_all() is unnecessary (but probably harmless as the "while (remain > sizeof(struct acpi_bert_region))" loop should terminate after we skipped over the first entry.

We can drop the "bert_print_all()" function and just move the four relevant lines inline in "bert_init()".

Cc Len Brown <[email protected]>
Cc Huang Ying <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Tomasz Nowicki <[email protected]>
Cc Jonathan (Zhixiong) Zhang <[email protected]>
Cc Tyler Baicar <[email protected]>
Cc [email protected]
Signed-off-by Tony Luck <[email protected]>
---
 drivers/acpi/apei/bert.c | 52 +++++++++++-------------------------------------
 1 file changed, 12 insertions(+), 40 deletions(-)
drivers/acpi/apei/bert.c +12 -40
index 9197555dc3c7..3b9b296dc84e 100644
@@ -34,43 +34,6 @@
 
 static int bert_disable;
 
-static void __init bert_print_all(struct acpi_bert_region *region,
-				  unsigned int region_len)
-{
-	struct acpi_hest_generic_status *estatus =
-		(struct acpi_hest_generic_status *)region;
-	int remain = region_len;
-	u32 estatus_len;
-
-	if (!estatus->block_status)
-		return;
-
-	while (remain > sizeof(struct acpi_bert_region)) {
-		if (cper_estatus_check(estatus)) {
-			pr_err(FW_BUG "Invalid error record.\n");
-			return;
-		}
-
-		estatus_len = cper_estatus_len(estatus);
-		if (remain < estatus_len) {
-			pr_err(FW_BUG "Truncated status block (length: %u).\n",
-			       estatus_len);
-			return;
-		}
-
-		pr_info_once("Error records from previous boot:\n");
-
-		cper_estatus_print(KERN_INFO HW_ERR, estatus);
-
-		estatus = (void *)estatus + estatus_len;
-		/* No more error records. */
-		if (!estatus->block_status)
-			return;
-
-		remain -= estatus_len;
-	}
-}
-
 static int __init setup_bert_disable(char *str)
 {
 	bert_disable = 1;

@@ -82,7 +45,7 @@ __setup("bert_disable", setup_bert_disable);
 static int __init bert_check_table(struct acpi_table_bert *bert_tab)
 {
 	if (bert_tab->header.length < sizeof(struct acpi_table_bert) ||
-	    bert_tab->region_length < sizeof(struct acpi_bert_region))
+	    bert_tab->region_length < sizeof(struct acpi_hest_generic_status))
 		return -EINVAL;
 
 	return 0;

@@ -91,7 +54,7 @@ static int __init bert_check_table(struct acpi_table_bert *bert_tab)
 static int __init bert_init(void)
 {
 	struct apei_resources bert_resources;
-	struct acpi_bert_region *boot_error_region;
+	struct acpi_hest_generic_status *boot_error_region;
 	struct acpi_table_bert *bert_tab;
 	unsigned int region_len;
 	acpi_status status;

@@ -131,7 +94,16 @@ static int __init bert_init(void)
 		goto out_fini;
 	boot_error_region = ioremap_cache(bert_tab->address, region_len);
 	if (boot_error_region) {
-		bert_print_all(boot_error_region, region_len);
+		if (boot_error_region->block_status) {
+			rc = cper_estatus_check(boot_error_region);
+			if (rc) {
+				pr_err(FW_BUG "Invalid error record.\n");
+				iounmap(boot_error_region);
+				return rc;
+			}
+			pr_info("Error records from previous boot:\n");
+			cper_estatus_print(KERN_INFO HW_ERR, boot_error_region);
+		}
 		iounmap(boot_error_region);
 	} else {
 		rc = -ENOMEM;
Signature · 1 line
2.11.0
Reviewed-by on the v5 series
#5 lore raw

Re: [PATCH v5 0/2] Fixes for ACPI BERT table

In reply to Luck, Tony
Quoted 19 lines · Tony

From: Tony Luck <[email protected]>

At Boris' suggestion I've split this into two parts for version 5. First one stops the driver from clearing the block_status field as:
1) it is unnecessary
2) not mentioned in the spec
3) may upset user mode tools that extract information from the BERT table.

Second is the code simplification that you've seen before in ealier versions of this patch.

Cc Len Brown <[email protected]>
Cc Huang Ying <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Tomasz Nowicki <[email protected]>
Cc Jonathan (Zhixiong) Zhang <[email protected]>
Cc Tyler Baicar <[email protected]>
Cc [email protected]
Quoted 4 lines

Tony Luck (2):
ACPI / APEI / BERT: Don't clear "block_status" in header ACPI / APEI: Boot Error Record Table processing was needlessly
complicated

Both:

Reviewed-by Borislav Petkov <[email protected]>
Signature · 5 lines
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
--