[Help] : RSS/PSS showing 0 during smaps for Xorg

[Help] : RSS/PSS showing 0 during smaps for Xorg

[PINTU KUMAR]

Dear All,

I am facing one problem for one of my kernel module for our linux mobile with kernel2.6.36.

When I do cat /proc/<Xorg pid>/smaps | grep -A 11 /dev/ump , to track information for my ump module,
we always get Rss/Pss as 0 kB as shown below:
cat /proc/1731/smaps | grep -A 11 /dev/ump
414db000-415ff000 rw-s 00015000 00:12 6803       /dev/ump
Size:               1168 kB
Rss:                   0 kB
Pss:                   0 kB
track_rss_value = 0, iswalkcalled = 1, smap_pte_range_called = 1, swap_pte = 0, not_pte_present = 0, not_normal_page = 1
isspecial = 0, not_special = 1, isMixedMap = 0, pfnpages_null = 0, pfnoff_flag = 0, not_cow_mapping = 1, normal_page_end = 0
 
After tracing down the problem, I found out that during "show_smaps" in fs/proc/task_mmu.c and during call to smaps_pte_range the vm_normal_page() is always returning NULL for our /dev/ump driver.
(smaps_pte_range() is the place where Rss/Pss information is populated)
Thus mss->resident (Rss value) is never getting incremented.  
 
To trace the problem I added few flags during show_smaps & vm_normal_page() as shown above. The value of 1 indicates that the condition is executed.
Thus "normal_page_end" indicates that the "vm_normal_page" has never ended successfully and always returns from 
"!is_cow_mapping()".
 
So, I wanted to know the main cause for vm_normal_page() always returning NULL page for our ump driver. 
What is that I am missing in my driver ?
 
Can anyone please let me know what could be the problem in our driver.
 
Thanks.
 
With Regards,
Pintu

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[Hugh Dickins]

On Mon, 23 Jan 2012, PINTU KUMAR wrote:
> Dear All,
> 
> I am facing one problem for one of my kernel module for our linux mobile with kernel2.6.36.
> 
> When I do cat /proc/<Xorg pid>/smaps | grep -A 11 /dev/ump , to track information for my ump module,
> we always get Rss/Pss as 0 kB as shown below:
> cat /proc/1731/smaps | grep -A 11 /dev/ump
> 414db000-415ff000 rw-s 00015000 00:12 6803       /dev/ump
> Size:               1168 kB
> Rss:                   0 kB
> Pss:                   0 kB
> track_rss_value = 0, iswalkcalled = 1, smap_pte_range_called = 1, swap_pte = 0, not_pte_present = 0, not_normal_page = 1
> isspecial = 0, not_special = 1, isMixedMap = 0, pfnpages_null = 0, pfnoff_flag = 0, not_cow_mapping = 1, normal_page_end = 0
>  
> After tracing down the problem, I found out that during "show_smaps" in fs/proc/task_mmu.c and during call to smaps_pte_range the vm_normal_page() is always returning NULL for our /dev/ump driver.
> (smaps_pte_range() is the place where Rss/Pss information is populated)
> Thus mss->resident (Rss value) is never getting incremented.  
>  
> To trace the problem I added few flags during show_smaps & vm_normal_page() as shown above. The value of 1 indicates that the condition is executed.
> Thus "normal_page_end" indicates that the "vm_normal_page" has never ended successfully and always returns from 
> "!is_cow_mapping()".
>  
> So, I wanted to know the main cause for vm_normal_page() always returning NULL page for our ump driver. 
> What is that I am missing in my driver ?
>  
> Can anyone please let me know what could be the problem in our driver.

This not evidence of any problem in your driver.

vm_normal_page() returns NULL because the pages mapped by your driver
are not normal faultable and reclaimable pages, but an area of physical
memory mapped in by remap_pfn_range(), which sets the VM_PFNMAP flag.

The mm subsystem does not count such pages towards rss (or pss),
hence your 0s.

Hugh

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[PINTU KUMAR]

________________________________
>From: Hugh Dickins <hughd@google.com>
>To: PINTU KUMAR <pintu_agarwal@yahoo.com> 
>Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>; "linux-mm@kvack.org" <linux-mm@kvack.org> 
>Sent: Tuesday, 24 January 2012 1:01 AM
>Subject: Re: [Help] : RSS/PSS showing 0 during smaps for Xorg
>
>On Mon, 23 Jan 2012, PINTU KUMAR wrote:
>> Dear All,
>> 
>> I am facing one problem for one of my kernel module for our linux mobile with kernel2.6.36.
>> 
>> When I do cat /proc/<Xorg pid>/smaps | grep -A 11 /dev/ump , to track information for my ump module,
>> we always get Rss/Pss as 0 kB as shown below:
>> cat /proc/1731/smaps | grep -A 11 /dev/ump
>> 414db000-415ff000 rw-s 00015000 00:12 6803       /dev/ump
>> Size:               1168 kB
>> Rss:                   0 kB
>> Pss:                   0 kB
>> track_rss_value = 0, iswalkcalled = 1, smap_pte_range_called = 1, swap_pte = 0, not_pte_present = 0, not_normal_page = 1
>> isspecial = 0, not_special = 1, isMixedMap = 0, pfnpages_null = 0, pfnoff_flag = 0, not_cow_mapping = 1, normal_page_end = 0
>>  
>> After tracing down the problem, I found out that during "show_smaps" in fs/proc/task_mmu.c and during call to smaps_pte_range the vm_normal_page() is always returning NULL for our /dev/ump driver.
>> (smaps_pte_range() is the place where Rss/Pss information is populated)
>> Thus mss->resident (Rss value) is never getting incremented.  
>>  
>> To trace the problem I added few flags during show_smaps & vm_normal_page() as shown above. The value of 1 indicates that the condition is executed.
>> Thus "normal_page_end" indicates that the "vm_normal_page" has never ended successfully and always returns from 
>> "!is_cow_mapping()".
>>  
>> So, I wanted to know the main cause for vm_normal_page() always returning NULL page for our ump driver. 
>> What is that I am missing in my driver ?
>>  
>> Can anyone please let me know what could be the problem in our driver.
>
>This not evidence of any problem in your driver.
>
>vm_normal_page() returns NULL because the pages mapped by your driver
>are not normal faultable and reclaimable pages, but an area of physical
>memory mapped in by remap_pfn_range(), which sets the VM_PFNMAP flag.
>
>The mm subsystem does not count such pages towards rss (or pss),
>hence your 0s.
>
>Hugh
>

Dear Mr. Hugh,
 
Thank you very much for your reply.
Is there a way to convert our mapped pages to a normal pages. I tried pfn_to_page() but no effect.
I mean the page is considered normal only if it is associated with "struct page" right???
Is is possible to convert these pages to a normal struct pages so that we can get the Rss/Pss value??
 
Also, the VM_PFNMAP is being set for all dirvers during remap_pfn_range and stills shows Rss/Pss for other drivers.
Then why it is not shown for our driver?
How to avoid remap_pfn_range to not to set VM_PFNMAP for our driver?
 
Please let me know.
 
 
 
Thanks, Regards,
Pintu

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[Hugh Dickins]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1719 bytes --]

On Tue, 24 Jan 2012, PINTU KUMAR wrote:
>  
> Is there a way to convert our mapped pages to a normal pages. I tried pfn_to_page() but no effect.
> I mean the page is considered normal only if it is associated with "struct page" right???
> Is is possible to convert these pages to a normal struct pages so that we can get the Rss/Pss value??

I don't understand why you are so anxious to see non-0 numbers there.

I don't know if the pages you are mapping with remap_pfn_range() are
ordinary pages in normal memory, and so already have struct pages, or not.

>  
> Also, the VM_PFNMAP is being set for all dirvers during remap_pfn_range and stills shows Rss/Pss for other drivers.

I'm surprised.  It is possible to set up a private-writable VM_PFNMAP area,
which can then contain ordinary private copies of the underlying pages,
and these copies will count to Rss.  But I thought that was very unusual.

You don't mention which drivers these are that use remap_pfn_range yet
show Rss (and I don't particularly want to spend time researching them).

I can see three or four places in drivers/ where VM_PFNMAP is set,
perhaps without going through remap_pfn_range(): that seems prone
to error, I wouldn't recommend going that route.

> Then why it is not shown for our driver?
> How to avoid remap_pfn_range to not to set VM_PFNMAP for our driver?

If these are ordinary pages with struct pages, then you could probably
use a loop of vm_insert_page()s to insert them at mmap time, or a fault
routine to insert them on fault.  But as I said, I don't know if this
memory is part of the ordinary page pool or not.

Really, the question has to be, why do you need to see non-0s there?

Hugh

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[PINTU KUMAR]

Dear Hugh Dickins,
 
Thank you very much for your reply.
 
>If these are ordinary pages with struct pages, then you could probably
>use a loop of vm_insert_page()s to insert them at mmap time, or a fault
>routine to insert them on fault.  But as I said, I don't know if this
>memory is part of the ordinary page pool or not.
 
You suggestion about using vm_insert_page() instead of remap_pfn_range worked for me and I got the Rss/Pss information for my driver.
But still there is one problem related to page fault. 
If I remove remap_pfn_range then I get a page fault in the beginning. 
I tried to use the same vm_insert_page() during page_fault_handler for each vmf->virtual_address but it did not work.
So for time being I remove the page fault handler from my vm_operations.
But with these my menu screen(LCD screen) is not behaving properly (I get colorful lines on my LCD).
So I need to handle the page fault properly.
 
But I am not sure what is that I need to do inside page fault handler. Do you have any example or references or suggestions?
 
 
>Really, the question has to be, why do you need to see non-0s there?
I want Rss/Pss value to account for how much video memory is used by the driver for the menu-screen,Xorg processes.
 
 
 
Thanks,
Pintu
 

>________________________________
>From: Hugh Dickins <hughd@google.com>
>To: PINTU KUMAR <pintu_agarwal@yahoo.com> 
>Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>; "linux-mm@kvack.org" <linux-mm@kvack.org> 
>Sent: Thursday, 26 January 2012 6:29 AM
>Subject: Re: [Help] : RSS/PSS showing 0 during smaps for Xorg
>
>On Tue, 24 Jan 2012, PINTU KUMAR wrote:
>>  
>> Is there a way to convert our mapped pages to a normal pages. I tried pfn_to_page() but no effect.
>> I mean the page is considered normal only if it is associated with "struct page" right???
>> Is is possible to convert these pages to a normal struct pages so that we can get the Rss/Pss value??
>
>I don't understand why you are so anxious to see non-0 numbers there.
>
>I don't know if the pages you are mapping with remap_pfn_range() are
>ordinary pages in normal memory, and so already have struct pages, or not.
>
>>  
>> Also, the VM_PFNMAP is being set for all dirvers during remap_pfn_range and stills shows Rss/Pss for other drivers.
>
>I'm surprised.  It is possible to set up a private-writable VM_PFNMAP area,
>which can then contain ordinary private copies of the underlying pages,
>and these copies will count to Rss.  But I thought that was very unusual.
>
>You don't mention which drivers these are that use remap_pfn_range yet
>show Rss (and I don't particularly want to spend time researching them).
>
>I can see three or four places in drivers/ where VM_PFNMAP is set,
>perhaps without going through remap_pfn_range(): that seems prone
>to error, I wouldn't recommend going that route.
>
>> Then why it is not shown for our driver?
>> How to avoid remap_pfn_range to not to set VM_PFNMAP for our driver?
>
>If these are ordinary pages with struct pages, then you could probably
>use a loop of vm_insert_page()s to insert them at mmap time, or a fault
>routine to insert them on fault.  But as I said, I don't know if this
>memory is part of the ordinary page pool or not.
>
>Really, the question has to be, why do you need to see non-0s there?
>
>Hugh
>
>

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[Hugh Dickins]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 2331 bytes --]

On Mon, 30 Jan 2012, PINTU KUMAR wrote:
>  
> >If these are ordinary pages with struct pages, then you could probably
> >use a loop of vm_insert_page()s to insert them at mmap time, or a fault
> >routine to insert them on fault.  But as I said, I don't know if this
> >memory is part of the ordinary page pool or not.
>  
> You suggestion about using vm_insert_page() instead of remap_pfn_range worked for me and I got the Rss/Pss information for my driver.

Oh, I'm glad that happened to work for you.

> But still there is one problem related to page fault. 
> If I remove remap_pfn_range then I get a page fault in the beginning. 
> I tried to use the same vm_insert_page() during page_fault_handler for each vmf->virtual_address but it did not work.
> So for time being I remove the page fault handler from my vm_operations.
> But with these my menu screen(LCD screen) is not behaving properly (I get colorful lines on my LCD).
> So I need to handle the page fault properly.
>  
> But I am not sure what is that I need to do inside page fault handler. Do you have any example or references or suggestions?

Sounds like you're not using vm_insert_page() properly: I would not expect
you to get a page fault there once you've set up the area with a loop of
vm_insert_page()s.

Check the comments above it in mm/memory.c ("Your vma protection will
have to be set up correctly" might be relevant).

Compare how you're using it with other users of vm_insert_page() in
the kernel tree.  Sorry, I don't have time to do your debugging.

>  
> >Really, the question has to be, why do you need to see non-0s there?
> I want Rss/Pss value to account for how much video memory is used by the driver for the menu-screen,Xorg processes.

So, userspace does an mmap for a large-enough window, but only some part of
that is filled by the driver (whether by remap_pfn_range or vm_insert_pages),
and you'd like to communicate back how much via the Rss, instead of adding
some ioctl or sysfs interface to the driver?  Fair enough.

I expect userspace could also work it out by touching pages of the area
until it gets a SIGBUS, but that might be too dirty a way of finding out.

Hmm, SIGBUS: maybe that's related to the faults that are puzzling you:
perhaps you're mapping less than you need to.

Hugh

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[PINTU KUMAR]

Dear Hugh,
 
Thank you for your help/suggestion so far.
 
Please find further updates on this below.
 
>Sounds like you're not using vm_insert_page() properly: I would not expect
>you to get a page fault there once you've set up the area with a loop of
>vm_insert_page()s.
>perhaps you're mapping less than you need to.

1)  The page fault is not occuring now after calling vm_insert_page() in a loop for every page. The hint about "mapping less than you need to" stricked me. Please check the snapshot below.
     loop
        vm_insert_page(vma,start,page);
        start = start + PAGE_SIZE;
        size = size - PAGE_SIZE;
    until size > 0
 
    I verified other drivers in kernel code and found this is how it is done.
 
2)  But after doing all this also my menu-screen is not proper. Still I am getting colorful lines on the menu-screen.
     Can you point me out what could be the problem??
 
 
 
Thanks, Regards,
Pintu
 
 
>________________________________
>From: Hugh Dickins <hughd@google.com>
>To: PINTU KUMAR <pintu_agarwal@yahoo.com> 
>Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>; "linux-mm@kvack.org" <linux-mm@kvack.org> 
>Sent: Tuesday, 31 January 2012 12:19 AM
>Subject: Re: [Help] : RSS/PSS showing 0 during smaps for Xorg
>
>On Mon, 30 Jan 2012, PINTU KUMAR wrote:
>>  
>> >If these are ordinary pages with struct pages, then you could probably
>> >use a loop of vm_insert_page()s to insert them at mmap time, or a fault
>> >routine to insert them on fault.  But as I said, I don't know if this
>> >memory is part of the ordinary page pool or not.
>>  
>> You suggestion about using vm_insert_page() instead of remap_pfn_range worked for me and I got the Rss/Pss information for my driver.
>
>Oh, I'm glad that happened to work for you.
>
>> But still there is one problem related to page fault. 
>> If I remove remap_pfn_range then I get a page fault in the beginning. 
>> I tried to use the same vm_insert_page() during page_fault_handler for each vmf->virtual_address but it did not work.
>> So for time being I remove the page fault handler from my vm_operations.
>> But with these my menu screen(LCD screen) is not behaving properly (I get colorful lines on my LCD).
>> So I need to handle the page fault properly.
>>  
>> But I am not sure what is that I need to do inside page fault handler. Do you have any example or references or suggestions?
>
>Sounds like you're not using vm_insert_page() properly: I would not expect
>you to get a page fault there once you've set up the area with a loop of
>vm_insert_page()s.
>
>Check the comments above it in mm/memory.c ("Your vma protection will
>have to be set up correctly" might be relevant).
>
>Compare how you're using it with other users of vm_insert_page() in
>the kernel tree.  Sorry, I don't have time to do your debugging.
>
>>  
>> >Really, the question has to be, why do you need to see non-0s there?
>> I want Rss/Pss value to account for how much video memory is used by the driver for the menu-screen,Xorg processes.
>
>So, userspace does an mmap for a large-enough window, but only some part of
>that is filled by the driver (whether by remap_pfn_range or vm_insert_pages),
>and you'd like to communicate back how much via the Rss, instead of adding
>some ioctl or sysfs interface to the driver?  Fair enough.
>
>I expect userspace could also work it out by touching pages of the area
>until it gets a SIGBUS, but that might be too dirty a way of finding out.
>
>Hmm, SIGBUS: maybe that's related to the faults that are puzzling you:
>perhaps you're mapping less than you need to.
>
>Hugh
>
>

Re: [Help] : RSS/PSS showing 0 during smaps for Xorg

[Hugh Dickins]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 260 bytes --]

On Wed, 1 Feb 2012, PINTU KUMAR wrote:
>  
> 2)  But after doing all this also my menu-screen is not proper.
>     Still I am getting colorful lines on the menu-screen.
>     Can you point me out what could be the problem??

Sorry, I do not know.

Hugh