Linux-ACPI Archive on lore.kernel.org
 help / color / Atom feed
* [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup
@ 2020-01-23  1:41 Cao jin
  2020-01-23  1:41 ` [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing Cao jin
                   ` (3 more replies)
  0 siblings, 4 replies; 8+ messages in thread
From: Cao jin @ 2020-01-23  1:41 UTC (permalink / raw)
  To: x86, linux-pm, linux-kernel, linux-acpi
  Cc: rjw, len.brown, pavel, tglx, mingo, hpa

Logic in early_acpi_process_madt() & acpi_process_madt() is really hard to
follow now. Clean them up.

Done basic boot test on my x86-64 PC.

CCed linux-acpi@vger.kernel.org

Cao jin (2):
  x86/acpi: Improve code readablity of early madt processing
  x86/acpi: Cleanup acpi_process_madt()

 arch/x86/kernel/acpi/boot.c | 72 +++++++++++++++----------------------
 1 file changed, 28 insertions(+), 44 deletions(-)

-- 
2.21.0




^ permalink raw reply	[flat|nested] 8+ messages in thread

* [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing
  2020-01-23  1:41 [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
@ 2020-01-23  1:41 ` Cao jin
  2020-02-24 13:21   ` Borislav Petkov
  2020-01-23  1:41 ` [RFC PATCH 2/2] x86/acpi: Cleanup acpi_process_madt() Cao jin
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 8+ messages in thread
From: Cao jin @ 2020-01-23  1:41 UTC (permalink / raw)
  To: x86, linux-pm, linux-kernel, linux-acpi
  Cc: rjw, len.brown, pavel, tglx, mingo, hpa

Current processing logic is confusing.

Return value of early_acpi_parse_madt_lapic_addr_ovr() indicates error(< 0),
parsed entry number(>= 0). So, it makes no sense to initialize acpi_lapic &
smp_found_config seeing no override entry, instead, initialize them seeing
MADT.

Put register_lapic_address() into override entry processing doesn't make
sense either.

Improved all the related comments too.

Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
---
 arch/x86/kernel/acpi/boot.c | 28 ++++++++++------------------
 1 file changed, 10 insertions(+), 18 deletions(-)

diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 04205ce127a1..2131035bba98 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -1003,11 +1003,7 @@ static int __init acpi_parse_fadt(struct acpi_table_header *table)
 }
 
 #ifdef	CONFIG_X86_LOCAL_APIC
-/*
- * Parse LAPIC entries in MADT
- * returns 0 on success, < 0 on error
- */
-
+/* Returns >= 0 on success, indicates parsed entry number; < 0 on error. */
 static int __init early_acpi_parse_madt_lapic_addr_ovr(void)
 {
 	int count;
@@ -1025,11 +1021,8 @@ static int __init early_acpi_parse_madt_lapic_addr_ovr(void)
 	if (count < 0) {
 		printk(KERN_ERR PREFIX
 		       "Error parsing LAPIC address override entry\n");
-		return count;
 	}
 
-	register_lapic_address(acpi_lapic_addr);
-
 	return count;
 }
 
@@ -1234,19 +1227,16 @@ static inline int acpi_parse_madt_ioapic_entries(void)
 static void __init early_acpi_process_madt(void)
 {
 #ifdef CONFIG_X86_LOCAL_APIC
-	int error;
+	int ret;
 
 	if (!acpi_table_parse(ACPI_SIG_MADT, acpi_parse_madt)) {
+		/* Found MADT */
+		acpi_lapic = 1;
+		smp_found_config = 1;
 
-		/*
-		 * Parse MADT LAPIC entries
-		 */
-		error = early_acpi_parse_madt_lapic_addr_ovr();
-		if (!error) {
-			acpi_lapic = 1;
-			smp_found_config = 1;
-		}
-		if (error == -EINVAL) {
+		/* See if override entry exists. */
+		ret = early_acpi_parse_madt_lapic_addr_ovr();
+		if (ret == -EINVAL) {
 			/*
 			 * Dell Precision Workstation 410, 610 come here.
 			 */
@@ -1254,6 +1244,8 @@ static void __init early_acpi_process_madt(void)
 			       "Invalid BIOS MADT, disabling ACPI\n");
 			disable_acpi();
 		}
+
+		register_lapic_address(acpi_lapic_addr);
 	}
 #endif
 }
-- 
2.21.0




^ permalink raw reply	[flat|nested] 8+ messages in thread

* [RFC PATCH 2/2] x86/acpi: Cleanup acpi_process_madt()
  2020-01-23  1:41 [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
  2020-01-23  1:41 ` [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing Cao jin
@ 2020-01-23  1:41 ` Cao jin
  2020-02-19  7:32 ` [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
  2020-03-27 12:30 ` Cao jin
  3 siblings, 0 replies; 8+ messages in thread
From: Cao jin @ 2020-01-23  1:41 UTC (permalink / raw)
  To: x86, linux-pm, linux-kernel, linux-acpi
  Cc: rjw, len.brown, pavel, tglx, mingo, hpa

After many duct-taped patches, the readablity of these code chunks are
really hard. This patch does following improvements:

  1. Drop unnecessary comment that is self-documented by function name,
     while supply with necessary comments.
  2. Drop duplicated code: acpi_process_madt() has already been called
     in early madt processing, acpi_lapic & smp_found_config is also
     initialized there.
  3. Fix code logic: variable count's usage is quite confusing now, and
     some code logic is wrong, like count & x2count will never be negative
     after assigned from madt_proc[n].count

Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
---
 arch/x86/kernel/acpi/boot.c | 44 +++++++++++++++----------------------
 1 file changed, 18 insertions(+), 26 deletions(-)

diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 2131035bba98..844fc9f26064 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -1026,9 +1026,10 @@ static int __init early_acpi_parse_madt_lapic_addr_ovr(void)
 	return count;
 }
 
+/* Returns 0 on success, < 0 on error. */
 static int __init acpi_parse_madt_lapic_entries(void)
 {
-	int count;
+	int count = 0;
 	int x2count = 0;
 	int ret;
 	struct acpi_subtable_proc madt_proc[2];
@@ -1036,10 +1037,13 @@ static int __init acpi_parse_madt_lapic_entries(void)
 	if (!boot_cpu_has(X86_FEATURE_APIC))
 		return -ENODEV;
 
-	count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_SAPIC,
-				      acpi_parse_sapic, MAX_LOCAL_APIC);
+	/* SAPIC is commonly found on Intel Itanium processor. */
+	ret = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_SAPIC,
+				    acpi_parse_sapic, MAX_LOCAL_APIC);
 
-	if (!count) {
+
+	/* No SAPIC entries implies it is not Itanium processor. */
+	if (!ret) {
 		memset(madt_proc, 0, sizeof(madt_proc));
 		madt_proc[0].id = ACPI_MADT_TYPE_LOCAL_APIC;
 		madt_proc[0].handler = acpi_parse_lapic;
@@ -1056,15 +1060,14 @@ static int __init acpi_parse_madt_lapic_entries(void)
 
 		count = madt_proc[0].count;
 		x2count = madt_proc[1].count;
-	}
-	if (!count && !x2count) {
-		printk(KERN_ERR PREFIX "No LAPIC entries present\n");
-		/* TBD: Cleanup to allow fallback to MPS */
-		return -ENODEV;
-	} else if (count < 0 || x2count < 0) {
-		printk(KERN_ERR PREFIX "Error parsing LAPIC entry\n");
-		/* TBD: Cleanup to allow fallback to MPS */
-		return count;
+		if (!count && !x2count) {
+			printk(KERN_ERR PREFIX "No LAPIC entries present\n");
+			/* TBD: Cleanup to allow fallback to MPS */
+			return -ENODEV;
+		}
+	} else if (ret < 0) {
+		printk(KERN_ERR PREFIX "Error parsing SAPIC entries\n");
+		return ret;
 	}
 
 	x2count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_X2APIC_NMI,
@@ -1074,7 +1077,7 @@ static int __init acpi_parse_madt_lapic_entries(void)
 	if (count < 0 || x2count < 0) {
 		printk(KERN_ERR PREFIX "Error parsing LAPIC NMI entry\n");
 		/* TBD: Cleanup to allow fallback to MPS */
-		return count;
+		return (count < 0) ? count : x2count;
 	}
 	return 0;
 }
@@ -1255,25 +1258,14 @@ static void __init acpi_process_madt(void)
 #ifdef CONFIG_X86_LOCAL_APIC
 	int error;
 
-	if (!acpi_table_parse(ACPI_SIG_MADT, acpi_parse_madt)) {
-
-		/*
-		 * Parse MADT LAPIC entries
-		 */
+	if (acpi_lapic) {
 		error = acpi_parse_madt_lapic_entries();
 		if (!error) {
-			acpi_lapic = 1;
-
-			/*
-			 * Parse MADT IO-APIC entries
-			 */
 			mutex_lock(&acpi_ioapic_lock);
 			error = acpi_parse_madt_ioapic_entries();
 			mutex_unlock(&acpi_ioapic_lock);
 			if (!error) {
 				acpi_set_irq_model_ioapic();
-
-				smp_found_config = 1;
 			}
 		}
 		if (error == -EINVAL) {
-- 
2.21.0




^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup
  2020-01-23  1:41 [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
  2020-01-23  1:41 ` [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing Cao jin
  2020-01-23  1:41 ` [RFC PATCH 2/2] x86/acpi: Cleanup acpi_process_madt() Cao jin
@ 2020-02-19  7:32 ` Cao jin
  2020-03-27 12:30 ` Cao jin
  3 siblings, 0 replies; 8+ messages in thread
From: Cao jin @ 2020-02-19  7:32 UTC (permalink / raw)
  To: x86, linux-pm, linux-kernel, linux-acpi
  Cc: rjw, len.brown, pavel, tglx, mingo, hpa

Ping.

On 1/23/20 9:41 AM, Cao jin wrote:
> Logic in early_acpi_process_madt() & acpi_process_madt() is really hard to
> follow now. Clean them up.
> 
> Done basic boot test on my x86-64 PC.
> 
> CCed linux-acpi@vger.kernel.org
> 
> Cao jin (2):
>   x86/acpi: Improve code readablity of early madt processing
>   x86/acpi: Cleanup acpi_process_madt()
> 
>  arch/x86/kernel/acpi/boot.c | 72 +++++++++++++++----------------------
>  1 file changed, 28 insertions(+), 44 deletions(-)
> 


-- 
Sincerely,
Cao jin



^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing
  2020-01-23  1:41 ` [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing Cao jin
@ 2020-02-24 13:21   ` Borislav Petkov
  2020-02-25  7:02     ` Cao jin
  0 siblings, 1 reply; 8+ messages in thread
From: Borislav Petkov @ 2020-02-24 13:21 UTC (permalink / raw)
  To: Cao jin
  Cc: x86, linux-pm, linux-kernel, linux-acpi, rjw, len.brown, pavel,
	tglx, mingo, hpa

On Thu, Jan 23, 2020 at 09:41:43AM +0800, Cao jin wrote:
> Current processing logic is confusing.
> 
> Return value of early_acpi_parse_madt_lapic_addr_ovr() indicates error(< 0),
> parsed entry number(>= 0).

You mean, the count of table entries parsed successfully?

> So, it makes no sense to initialize acpi_lapic & smp_found_config
> seeing no override entry, instead, initialize them seeing MADT.

Err, that logical conclusion is not really clear to me - pls try
again with more detail. I kinda see what you mean by looking at
acpi_process_madt() but before I commit a change like that, I better
have the warm and fuzzy feeling that it is correct and properly
explained in its commit message.

So why did

  cbf9bd603ab1 ("acpi: get boot_cpu_id as early for k8_scan_nodes")

do it this way? Was it wrong or why?

I'm very wary about touching ACPI parsing code for no good reason
because, well, it is ACPI...

Thx.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing
  2020-02-24 13:21   ` Borislav Petkov
@ 2020-02-25  7:02     ` Cao jin
  2020-03-16  9:20       ` Cao jin
  0 siblings, 1 reply; 8+ messages in thread
From: Cao jin @ 2020-02-25  7:02 UTC (permalink / raw)
  To: Borislav Petkov
  Cc: x86, linux-pm, linux-kernel, linux-acpi, rjw, len.brown, pavel,
	tglx, mingo, hpa

On 2/24/20 9:21 PM, Borislav Petkov wrote:
> On Thu, Jan 23, 2020 at 09:41:43AM +0800, Cao jin wrote:
>> Current processing logic is confusing.
>>
>> Return value of early_acpi_parse_madt_lapic_addr_ovr() indicates error(< 0),
>> parsed entry number(>= 0).
> 
> You mean, the count of table entries parsed successfully?

Yes, 0 for no override sub-table.

> 
>> So, it makes no sense to initialize acpi_lapic & smp_found_config
>> seeing no override entry, instead, initialize them seeing MADT.
> 
> Err, that logical conclusion is not really clear to me - pls try
> again with more detail. I kinda see what you mean by looking at
> acpi_process_madt() but before I commit a change like that, I better
> have the warm and fuzzy feeling that it is correct and properly
> explained in its commit message.
> 

My understanding of early_acpi_process_madt(): mainly for getting APIC
register base address(acpi_lapic_addr) from MADT, then process it via
register_lapic_address().  acpi_lapic_addr could be got from one of
following 2 places:

  1. MADT header (32-bit address, always exist)
  2. MADT sub-table: Local APIC Address Override (64-bit address,
     optional, high priority and use it if present)

So the making-sense logic to me goes like:

  1. get (32-bit) acpi_lapic_addr from MADT header.
  2. check if there is MADT override structure & get 64-bit
     acpi_lapic_addr if present.
  3. register_lapic_address(acpi_lapic_addr);

Then, it looks weird to me putting register_lapic_address() into
early_acpi_parse_madt_lapic_addr_ovr(), the result is not wrong, but the
code logic is hard for newbie. (these 2 functions both does more than
its name tells, register_lapic_address() also get boot cpu APIC ID &
version.)

Variable acpi_lapic and its counterpart smp_found_config from MPS
indicate whether it is SMP system, right? The following code:


	error = early_acpi_parse_madt_lapic_addr_ovr();
	if (!error) {
		acpi_lapic = 1;
		smp_found_config = 1;
	}

means setting them when there is no override sub-table, so why can't
moving the setting operation out? Another issue: if there *is* override
sub-table, don't set those two?

> So why did
> 
>   cbf9bd603ab1 ("acpi: get boot_cpu_id as early for k8_scan_nodes")
> 
> do it this way? Was it wrong or why?

Not a clue... The title says it wants boot_cpu_physical_apicid, but did
many other things. Maybe Thomas could provide some insights?

> 
> I'm very wary about touching ACPI parsing code for no good reason
> because, well, it is ACPI...

I was expecting ACPI guys could help to confirm;) I also understand this
should be tested widely, but I just have a normal PC, so it is a RFC:)
-- 
Sincerely,
Cao jin



^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing
  2020-02-25  7:02     ` Cao jin
@ 2020-03-16  9:20       ` Cao jin
  0 siblings, 0 replies; 8+ messages in thread
From: Cao jin @ 2020-03-16  9:20 UTC (permalink / raw)
  To: Borislav Petkov
  Cc: x86, linux-pm, linux-kernel, linux-acpi, rjw, len.brown, pavel,
	tglx, mingo, hpa

Hello BP,

  Does the explanation make sense to you?
  BTW, also test it on i386, boots fine.

-- 
Sincerely,
Cao jin

On 2/25/20 3:02 PM, Cao jin wrote:
> On 2/24/20 9:21 PM, Borislav Petkov wrote:
>> On Thu, Jan 23, 2020 at 09:41:43AM +0800, Cao jin wrote:
>>> Current processing logic is confusing.
>>>
>>> Return value of early_acpi_parse_madt_lapic_addr_ovr() indicates error(< 0),
>>> parsed entry number(>= 0).
>>
>> You mean, the count of table entries parsed successfully?
> 
> Yes, 0 for no override sub-table.
> 
>>
>>> So, it makes no sense to initialize acpi_lapic & smp_found_config
>>> seeing no override entry, instead, initialize them seeing MADT.
>>
>> Err, that logical conclusion is not really clear to me - pls try
>> again with more detail. I kinda see what you mean by looking at
>> acpi_process_madt() but before I commit a change like that, I better
>> have the warm and fuzzy feeling that it is correct and properly
>> explained in its commit message.
>>
> 
> My understanding of early_acpi_process_madt(): mainly for getting APIC
> register base address(acpi_lapic_addr) from MADT, then process it via
> register_lapic_address().  acpi_lapic_addr could be got from one of
> following 2 places:
> 
>   1. MADT header (32-bit address, always exist)
>   2. MADT sub-table: Local APIC Address Override (64-bit address,
>      optional, high priority and use it if present)
> 
> So the making-sense logic to me goes like:
> 
>   1. get (32-bit) acpi_lapic_addr from MADT header.
>   2. check if there is MADT override structure & get 64-bit
>      acpi_lapic_addr if present.
>   3. register_lapic_address(acpi_lapic_addr);
> 
> Then, it looks weird to me putting register_lapic_address() into
> early_acpi_parse_madt_lapic_addr_ovr(), the result is not wrong, but the
> code logic is hard for newbie. (these 2 functions both does more than
> its name tells, register_lapic_address() also get boot cpu APIC ID &
> version.)
> 
> Variable acpi_lapic and its counterpart smp_found_config from MPS
> indicate whether it is SMP system, right? The following code:
> 
> 
> 	error = early_acpi_parse_madt_lapic_addr_ovr();
> 	if (!error) {
> 		acpi_lapic = 1;
> 		smp_found_config = 1;
> 	}
> 
> means setting them when there is no override sub-table, so why can't
> moving the setting operation out? Another issue: if there *is* override
> sub-table, don't set those two?
> 





^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup
  2020-01-23  1:41 [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
                   ` (2 preceding siblings ...)
  2020-02-19  7:32 ` [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
@ 2020-03-27 12:30 ` Cao jin
  3 siblings, 0 replies; 8+ messages in thread
From: Cao jin @ 2020-03-27 12:30 UTC (permalink / raw)
  To: x86, linux-pm, linux-kernel, linux-acpi
  Cc: rjw, len.brown, pavel, tglx, mingo, hpa, Borislav Petkov

Ping.

On 1/23/20 9:41 AM, Cao jin wrote:
> Logic in early_acpi_process_madt() & acpi_process_madt() is really hard to
> follow now. Clean them up.
> 
> Done basic boot test on my x86-64 PC.
> 
> CCed linux-acpi@vger.kernel.org
> 
> Cao jin (2):
>   x86/acpi: Improve code readablity of early madt processing
>   x86/acpi: Cleanup acpi_process_madt()
> 
>  arch/x86/kernel/acpi/boot.c | 72 +++++++++++++++----------------------
>  1 file changed, 28 insertions(+), 44 deletions(-)
> 


-- 
Sincerely,
Cao jin



^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, back to index

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-01-23  1:41 [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
2020-01-23  1:41 ` [RFC PATCH 1/2] x86/acpi: Improve code readablity of early madt processing Cao jin
2020-02-24 13:21   ` Borislav Petkov
2020-02-25  7:02     ` Cao jin
2020-03-16  9:20       ` Cao jin
2020-01-23  1:41 ` [RFC PATCH 2/2] x86/acpi: Cleanup acpi_process_madt() Cao jin
2020-02-19  7:32 ` [RFC PATCH 0/2] x86/boot: early ACPI MADT processing cleanup Cao jin
2020-03-27 12:30 ` Cao jin

Linux-ACPI Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-acpi/0 linux-acpi/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-acpi linux-acpi/ https://lore.kernel.org/linux-acpi \
		linux-acpi@vger.kernel.org
	public-inbox-index linux-acpi

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-acpi


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git