mmap hw behavior

mmap hw behavior

[Michal Simek]

Hi All,

I wrote one email about my problem with mmap on Microblaze. (Especially
mmap01 ltp test)
I need to know how mmap should work with MMU or what the correct steps are.

When I open temp file in application is the kernel create file descriptor table and kernel take
care about setup tlb which are needed. Application get back just file descriptor.

When I call calloc it is called brk kernel syscall and is created one tlb for this place. 
Brk allocate more memory than is requested by calloc. Application got pointer to that memory place.


When I call mmap for that open file with pointer to calloc place (first parameter, + length zero)
it should be one tlb invalidation for calloc and new tlb which connect open file. 
We check it and we don't have any tlb invalidation that's why I think that kernel do different thigs.
Or is it there any copying? Or anything different?

Thanks for clarification,
Michal

-- 
Michal Simek, Ing. (M.Eng)
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663,+42-0-721842854 f: +61-7-30090663

mmap hw behavior

[Michal Simek]

Hi All,

I wrote one email about my problem with mmap on Microblaze. (Especially
mmap01 ltp test)
I need to know how mmap should work with MMU or what the correct steps are.

When I open temp file in application is the kernel create file descriptor table and kernel take
care about setup tlb which are needed. Application get back just file descriptor.

When I call calloc it is called brk kernel syscall and is created one tlb for this place. 
Brk allocate more memory than is requested by calloc. Application got pointer to that memory place.


When I call mmap for that open file with pointer to calloc place (first parameter, + length zero)
it should be one tlb invalidation for calloc and new tlb which connect open file. 
We check it and we don't have any tlb invalidation that's why I think that kernel do different thigs.
Or is it there any copying? Or anything different?

Thanks for clarification,
Michal

-- 
Michal Simek, Ing. (M.Eng)
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663,+42-0-721842854 f: +61-7-30090663

Re: mmap hw behavior

[David Miller]

From: Michal Simek <michal.simek@petalogix.com>
Date: Wed, 08 Jul 2009 19:03:11 +0200

> When I call mmap for that open file with pointer to calloc place
> (first parameter, + length zero) it should be one tlb invalidation
> for calloc and new tlb which connect open file.  We check it and we
> don't have any tlb invalidation that's why I think that kernel do
> different thigs.  Or is it there any copying? Or anything different?

There is no need to tlb flush the calloc area unless that memory area
is actually touched by the user application and thus the page is
faulted in.

Re: mmap hw behavior

[Michal Simek]

David Miller wrote:
> From: Michal Simek <michal.simek@petalogix.com>
> Date: Wed, 08 Jul 2009 19:03:11 +0200
>
>   
>> When I call mmap for that open file with pointer to calloc place
>> (first parameter, + length zero) it should be one tlb invalidation
>> for calloc and new tlb which connect open file.  We check it and we
>> don't have any tlb invalidation that's why I think that kernel do
>> different thigs.  Or is it there any copying? Or anything different?
>>     
>
> There is no need to tlb flush the calloc area unless that memory area
> is actually touched by the user application and thus the page is
> faulted in.
>   
That calloc area is filled by any value (in that test). Is it mean that
for this case when calloc area is touched
there must be tlb invalidation + remapping?

Michal


-- 
Michal Simek, Ing. (M.Eng)
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663,+42-0-721842854 f: +61-7-30090663

Re: mmap hw behavior

[David Miller]

From: Michal Simek <michal.simek-g5w7nrANp4BDPfheJLI6IQ@public.gmane.org>
Date: Wed, 08 Jul 2009 20:03:27 +0200

> David Miller wrote:
>> From: Michal Simek <michal.simek-g5w7nrANp4BDPfheJLI6IQ@public.gmane.org>
>> Date: Wed, 08 Jul 2009 19:03:11 +0200
>>
>>   
>>> When I call mmap for that open file with pointer to calloc place
>>> (first parameter, + length zero) it should be one tlb invalidation
>>> for calloc and new tlb which connect open file.  We check it and we
>>> don't have any tlb invalidation that's why I think that kernel do
>>> different thigs.  Or is it there any copying? Or anything different?
>>>     
>>
>> There is no need to tlb flush the calloc area unless that memory area
>> is actually touched by the user application and thus the page is
>> faulted in.
>>   
> That calloc area is filled by any value (in that test). Is it mean that
> for this case when calloc area is touched
> there must be tlb invalidation + remapping?

Yes, if the calloc area is written to by the application, there
should be a tlb flush when the mmap() overrides that virtual region
with a different mapping.

------------------------------------------------------------------------------
Enter the BlackBerry Developer Challenge  
This is your chance to win up to $100,000 in prizes! For a limited time, 
vendors submitting new applications to BlackBerry App World(TM) will have
the opportunity to enter the BlackBerry Developer Challenge. See full prize  
details at: http://p.sf.net/sfu/Challenge

Re: mmap hw behavior

[David Miller]

From: Michal Simek <michal.simek@petalogix.com>
Date: Wed, 08 Jul 2009 20:03:27 +0200

> David Miller wrote:
>> From: Michal Simek <michal.simek@petalogix.com>
>> Date: Wed, 08 Jul 2009 19:03:11 +0200
>>
>>   
>>> When I call mmap for that open file with pointer to calloc place
>>> (first parameter, + length zero) it should be one tlb invalidation
>>> for calloc and new tlb which connect open file.  We check it and we
>>> don't have any tlb invalidation that's why I think that kernel do
>>> different thigs.  Or is it there any copying? Or anything different?
>>>     
>>
>> There is no need to tlb flush the calloc area unless that memory area
>> is actually touched by the user application and thus the page is
>> faulted in.
>>   
> That calloc area is filled by any value (in that test). Is it mean that
> for this case when calloc area is touched
> there must be tlb invalidation + remapping?

Yes, if the calloc area is written to by the application, there
should be a tlb flush when the mmap() overrides that virtual region
with a different mapping.

Re: mmap hw behavior

[Michal Simek]

David Miller wrote:
> From: Michal Simek <michal.simek@petalogix.com>
> Date: Wed, 08 Jul 2009 20:03:27 +0200
>
>   
>> David Miller wrote:
>>     
>>> From: Michal Simek <michal.simek@petalogix.com>
>>> Date: Wed, 08 Jul 2009 19:03:11 +0200
>>>
>>>   
>>>       
>>>> When I call mmap for that open file with pointer to calloc place
>>>> (first parameter, + length zero) it should be one tlb invalidation
>>>> for calloc and new tlb which connect open file.  We check it and we
>>>> don't have any tlb invalidation that's why I think that kernel do
>>>> different thigs.  Or is it there any copying? Or anything different?
>>>>     
>>>>         
>>> There is no need to tlb flush the calloc area unless that memory area
>>> is actually touched by the user application and thus the page is
>>> faulted in.
>>>   
>>>       
>> That calloc area is filled by any value (in that test). Is it mean that
>> for this case when calloc area is touched
>> there must be tlb invalidation + remapping?
>>     
>
> Yes, if the calloc area is written to by the application, there
> should be a tlb flush when the mmap() overrides that virtual region
> with a different mapping.
>   
Can you please point me to any code which exactly do this? (for example
file mm/mmap.c line from ... to ... )

Thanks,
Michal

-- 
Michal Simek, Ing. (M.Eng)
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663,+42-0-721842854 f: +61-7-30090663

Re: mmap hw behavior

[David Miller]

From: Michal Simek <michal.simek@petalogix.com>
Date: Wed, 08 Jul 2009 22:22:22 +0200

> Can you please point me to any code which exactly do this? (for example
> file mm/mmap.c line from ... to ... )

Sorry I don't have time to read the source code for you.
I'd have to look it up too, just like you.

Re: mmap hw behavior

[Michal Simek]

David Miller wrote:
> From: Michal Simek <michal.simek@petalogix.com>
> Date: Wed, 08 Jul 2009 22:22:22 +0200
>
>   
>> Can you please point me to any code which exactly do this? (for example
>> file mm/mmap.c line from ... to ... )
>>     
>
> Sorry I don't have time to read the source code for you.
> I'd have to look it up too, just like you.
>   
It is ok - I didn't want to read the code from - just if you know any
functions which
could help me.

Thank you very much,
Michal

-- 
Michal Simek, Ing. (M.Eng)
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663,+42-0-721842854 f: +61-7-30090663

Re: mmap hw behavior

[Michal Simek]

David Miller wrote:
> From: Michal Simek <michal.simek@petalogix.com>
> Date: Wed, 08 Jul 2009 22:22:22 +0200
>
>   
>> Can you please point me to any code which exactly do this? (for example
>> file mm/mmap.c line from ... to ... )
>>     
>
> Sorry I don't have time to read the source code for you.
> I'd have to look it up too, just like you.
>   
I found it - Microblaze hasn't defined tlb_flush macro.

Thanks for your help,
Michal

-- 
Michal Simek, Ing. (M.Eng)
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663,+42-0-721842854 f: +61-7-30090663