[PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

WARN_ONCE is more clear and simpler. Just replace it.

Signed-off-by: zhong jiang <zhongjiang@huawei.com>
---
 arch/x86/mm/ioremap.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
index a39dcdb..3b74599 100644
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
 		return NULL;
 
 	if (!phys_addr_valid(phys_addr)) {
-		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
-		       (unsigned long long)phys_addr);
-		WARN_ON_ONCE(1);
+		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
+			  (unsigned long long)phys_addr);
 		return NULL;
 	}
 
-- 
1.7.12.4

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[Borislav Petkov]

On Wed, Oct 30, 2019 at 04:57:18PM +0800, zhong jiang wrote:
> WARN_ONCE is more clear and simpler. Just replace it.
> 
> Signed-off-by: zhong jiang <zhongjiang@huawei.com>
> ---
>  arch/x86/mm/ioremap.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> index a39dcdb..3b74599 100644
> --- a/arch/x86/mm/ioremap.c
> +++ b/arch/x86/mm/ioremap.c
> @@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
>  		return NULL;
>  
>  	if (!phys_addr_valid(phys_addr)) {
> -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
> -		       (unsigned long long)phys_addr);
> -		WARN_ON_ONCE(1);
> +		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
> +			  (unsigned long long)phys_addr);

Does
	WARN_ONCE(!phys_addr_valid(phys_addr),
		  "ioremap: invalid physical address %llx\n",
		  (unsigned long long)phys_addr);

work too?

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

On 2019/10/31 19:03, Borislav Petkov wrote:
> On Wed, Oct 30, 2019 at 04:57:18PM +0800, zhong jiang wrote:
>> WARN_ONCE is more clear and simpler. Just replace it.
>>
>> Signed-off-by: zhong jiang <zhongjiang@huawei.com>
>> ---
>>  arch/x86/mm/ioremap.c | 5 ++---
>>  1 file changed, 2 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
>> index a39dcdb..3b74599 100644
>> --- a/arch/x86/mm/ioremap.c
>> +++ b/arch/x86/mm/ioremap.c
>> @@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
>>  		return NULL;
>>  
>>  	if (!phys_addr_valid(phys_addr)) {
>> -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
>> -		       (unsigned long long)phys_addr);
>> -		WARN_ON_ONCE(1);
>> +		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
>> +			  (unsigned long long)phys_addr);
> Does
> 	WARN_ONCE(!phys_addr_valid(phys_addr),
> 		  "ioremap: invalid physical address %llx\n",
> 		  (unsigned long long)phys_addr);
>
> work too?
>
Thanks, That is better. Will repost.

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

On 2019/10/31 19:03, Borislav Petkov wrote:
> On Wed, Oct 30, 2019 at 04:57:18PM +0800, zhong jiang wrote:
>> WARN_ONCE is more clear and simpler. Just replace it.
>>
>> Signed-off-by: zhong jiang <zhongjiang@huawei.com>
>> ---
>>  arch/x86/mm/ioremap.c | 5 ++---
>>  1 file changed, 2 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
>> index a39dcdb..3b74599 100644
>> --- a/arch/x86/mm/ioremap.c
>> +++ b/arch/x86/mm/ioremap.c
>> @@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
>>  		return NULL;
>>  
>>  	if (!phys_addr_valid(phys_addr)) {
>> -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
>> -		       (unsigned long long)phys_addr);
>> -		WARN_ON_ONCE(1);
>> +		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
>> +			  (unsigned long long)phys_addr);
> Does
> 	WARN_ONCE(!phys_addr_valid(phys_addr),
> 		  "ioremap: invalid physical address %llx\n",
> 		  (unsigned long long)phys_addr);
>
> work too?
>
Look at this again, It should not works. Because that will change the logical.
if phys_addr_valid is false, we should drop out in time.

WARN_ONCE should be just visible for user not should to handle with address.

Thanks,
zhong jiang

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[Joe Perches]

On Thu, 2019-10-31 at 19:36 +0800, zhong jiang wrote:
> On 2019/10/31 19:03, Borislav Petkov wrote:
> > On Wed, Oct 30, 2019 at 04:57:18PM +0800, zhong jiang wrote:
> > > WARN_ONCE is more clear and simpler. Just replace it.
[]
> > > diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
[]
> > > @@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
> > >  		return NULL;
> > >  
> > >  	if (!phys_addr_valid(phys_addr)) {
> > > -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
> > > -		       (unsigned long long)phys_addr);
> > > -		WARN_ON_ONCE(1);
> > > +		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
> > > +			  (unsigned long long)phys_addr);
> > Does
> > 	WARN_ONCE(!phys_addr_valid(phys_addr),
> > 		  "ioremap: invalid physical address %llx\n",
> > 		  (unsigned long long)phys_addr);
> > 
> > work too?
> > 
> Thanks, That is better. Will repost.

Perhaps this is not good patch concept as now each
invalid physical address will not be emitted.

Before:
	each invalid physical address printed
	one stack dump

After:
	one stck dump with first invalid physical address.

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

On 2019/10/31 20:00, Joe Perches wrote:
> On Thu, 2019-10-31 at 19:36 +0800, zhong jiang wrote:
>> On 2019/10/31 19:03, Borislav Petkov wrote:
>>> On Wed, Oct 30, 2019 at 04:57:18PM +0800, zhong jiang wrote:
>>>> WARN_ONCE is more clear and simpler. Just replace it.
> []
>>>> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> []
>>>> @@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
>>>>  		return NULL;
>>>>  
>>>>  	if (!phys_addr_valid(phys_addr)) {
>>>> -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
>>>> -		       (unsigned long long)phys_addr);
>>>> -		WARN_ON_ONCE(1);
>>>> +		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
>>>> +			  (unsigned long long)phys_addr);
>>> Does
>>> 	WARN_ONCE(!phys_addr_valid(phys_addr),
>>> 		  "ioremap: invalid physical address %llx\n",
>>> 		  (unsigned long long)phys_addr);
>>>
>>> work too?
>>>
>> Thanks, That is better. Will repost.
> Perhaps this is not good patch concept as now each
> invalid physical address will not be emitted.
>
> Before:
> 	each invalid physical address printed
> 	one stack dump
>
> After:
> 	one stck dump with first invalid physical address.
>
Yes,  I  has told that. 

How you think my above patch in the mail.  Thanks
>
> .
>

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[Borislav Petkov]

On Thu, Oct 31, 2019 at 07:54:09PM +0800, zhong jiang wrote:
> Look at this again, It should not works. Because that will change the logical.
> if phys_addr_valid is false, we should drop out in time.

That you can do too:

diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
index a39dcdb5ae34..13f44cc064af 100644
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -171,12 +171,10 @@ __ioremap_caller(resource_size_t phys_addr, unsigned long size,
 	if (!size || last_addr < phys_addr)
 		return NULL;
 
-	if (!phys_addr_valid(phys_addr)) {
-		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
-		       (unsigned long long)phys_addr);
-		WARN_ON_ONCE(1);
+	if (WARN_ONCE(!phys_addr_valid(phys_addr),
+	    "ioremap: invalid physical address %llx\n",
+	    (unsigned long long)phys_addr))
 		return NULL;
-	}
 
 	__ioremap_check_mem(phys_addr, size, &io_desc);
 
---

I'm not sure whether we care about printing every invalid address, as
Joe points out. Maybe we do... *shrug*

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

On 2019/10/31 23:49, Borislav Petkov wrote:
> On Thu, Oct 31, 2019 at 07:54:09PM +0800, zhong jiang wrote:
>> Look at this again, It should not works. Because that will change the logical.
>> if phys_addr_valid is false, we should drop out in time.
> That you can do too:
>
> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> index a39dcdb5ae34..13f44cc064af 100644
> --- a/arch/x86/mm/ioremap.c
> +++ b/arch/x86/mm/ioremap.c
> @@ -171,12 +171,10 @@ __ioremap_caller(resource_size_t phys_addr, unsigned long size,
>  	if (!size || last_addr < phys_addr)
>  		return NULL;
>  
> -	if (!phys_addr_valid(phys_addr)) {
> -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
> -		       (unsigned long long)phys_addr);
> -		WARN_ON_ONCE(1);
> +	if (WARN_ONCE(!phys_addr_valid(phys_addr),
> +	    "ioremap: invalid physical address %llx\n",
> +	    (unsigned long long)phys_addr))
>  		return NULL;
> -	}
>  
Yep,  WARN_ONCE alway return true in that case.

Thanks,
zhong jiang
>  	__ioremap_check_mem(phys_addr, size, &io_desc);
>  
> ---
>
> I'm not sure whether we care about printing every invalid address, as
> Joe points out. Maybe we do... *shrug*
>

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[Borislav Petkov]

On Thu, Oct 31, 2019 at 11:54:24PM +0800, zhong jiang wrote:
> Yep,  WARN_ONCE alway return true in that case.

Are you sure it does that always?

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

On 2019/10/31 20:00, Joe Perches wrote:
> On Thu, 2019-10-31 at 19:36 +0800, zhong jiang wrote:
>> On 2019/10/31 19:03, Borislav Petkov wrote:
>>> On Wed, Oct 30, 2019 at 04:57:18PM +0800, zhong jiang wrote:
>>>> WARN_ONCE is more clear and simpler. Just replace it.
> []
>>>> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> []
>>>> @@ -172,9 +172,8 @@ static void __ioremap_check_mem(resource_size_t addr, unsigned long size,
>>>>  		return NULL;
>>>>  
>>>>  	if (!phys_addr_valid(phys_addr)) {
>>>> -		printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
>>>> -		       (unsigned long long)phys_addr);
>>>> -		WARN_ON_ONCE(1);
>>>> +		WARN_ONCE(1, "ioremap: invalid physical address %llx\n",
>>>> +			  (unsigned long long)phys_addr);
>>> Does
>>> 	WARN_ONCE(!phys_addr_valid(phys_addr),
>>> 		  "ioremap: invalid physical address %llx\n",
>>> 		  (unsigned long long)phys_addr);
>>>
>>> work too?
>>>
>> Thanks, That is better. Will repost.
> Perhaps this is not good patch concept as now each
> invalid physical address will not be emitted.
>
> Before:
> 	each invalid physical address printed
> 	one stack dump
>
> After:
> 	one stck dump with first invalid physical address.
>
I misunderstand your meaning.  You're right.  My patch change its logical.
Thanks for your reminder.

Sincerely,
zhong jiang
>
> .
>

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[zhong jiang]

On 2019/11/1 16:45, Borislav Petkov wrote:
> On Thu, Oct 31, 2019 at 11:54:24PM +0800, zhong jiang wrote:
>> Yep,  WARN_ONCE alway return true in that case.
> Are you sure it does that always?
>
WARN_ONCE will alway return true if its condition is true.:-)

Re: [PATCH] mm/ioremap: Use WARN_ONCE instead of printk() + WARN_ON_ONCE()

[Borislav Petkov]

On Fri, Nov 01, 2019 at 11:32:11PM +0800, zhong jiang wrote:
> WARN_ONCE will alway return true if its condition is true.:-)

And if its condition is false?

Lemme save you some time: WARN_ONCE() returns the @condition value so
constructs like

	if (WARN_ONCE(condition,...))

are possible. Grep the code for examples.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--