Re: [PATCH] MIPS: Always page align TASK_SIZE

[David Daney <ddaney.cavm@gmail.com>]

On 02/08/2016 10:15 AM, Harvey Hunt wrote:
> Hi David,
>
> On 02/08/2016 10:11 AM, David Daney wrote:
>> On 02/08/2016 10:05 AM, Harvey Hunt wrote:
>>> STACK_TOP_MAX is aligned on a 32k boundary. When __bprm_mm_init()
>>> creates an
>>> initial stack for a process, it does so using STACK_TOP_MAX as the end
>>> of the
>>> vma. A process's arguments and environment information are placed on
>>> the stack
>>> and then the stack is relocated and aligned on a page boundary. When
>>> using a 32
>>> bit kernel with 64k pages, the relocated stack has the process's args
>>> erroneously stored in the middle of the stack. This means that processes
>>> receive no arguments or environment variables, preventing them from
>>> running
>>> correctly.
>>>
>>> Fix this by aligning TASK_SIZE on a page boundary.
>>>
>>> Signed-off-by: Harvey Hunt <harvey.hunt@imgtec.com>
>>> Cc: David Daney <david.daney@cavium.com>
>>> Cc: Paul Burton <paul.burton@imgtec.com>
>>> Cc: James Hogan <james.hogan@imgtec.com>
>>> Cc: linux-kernel@vger.kernel.org
>>> ---
>>>   arch/mips/include/asm/processor.h | 6 +++---
>>>   1 file changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/arch/mips/include/asm/processor.h
>>> b/arch/mips/include/asm/processor.h
>>> index 3f832c3..b618b40 100644
>>> --- a/arch/mips/include/asm/processor.h
>>> +++ b/arch/mips/include/asm/processor.h
>>> @@ -39,13 +39,13 @@ extern unsigned int vced_count, vcei_count;
>>>   #ifdef CONFIG_32BIT
>>>   #ifdef CONFIG_KVM_GUEST
>>>   /* User space process size is limited to 1GB in KVM Guest Mode */
>>> -#define TASK_SIZE    0x3fff8000UL
>>> +#define TASK_SIZE    (0x40000000UL - PAGE_SIZE)
>>>   #else
>>>   /*
>>>    * User space process size: 2GB. This is hardcoded into a few places,
>>>    * so don't change it unless you know what you are doing.
>>>    */
>>> -#define TASK_SIZE    0x7fff8000UL
>>> +#define TASK_SIZE    (0x7fff8000UL & PAGE_SIZE)
>>
>> Can you check your math here.  This doesn't seem correct.
>
> Thanks for spotting that - it should have been:
>
> (0x7fff8000UL & PAGE_MASK)

This brings up an interesting point.  How was this tested?  Please note 
that in the change log.

Also look at the definition of PAGE_MASK in page.h

Is that correct?  Most of the other related symbols have an "_AC(1,UL)" 
in them.  Why is this not also appropriate for PAGE_MASK?

It may also be a good idea to prepare and test a patch that defines 
PAGE_MASK much in the same way HPAGE_MASK is defined.

David Daney


>
> I'll do a v2 now.
>
>>
>>>   #endif
>>>
>>>   #define STACK_TOP_MAX    TASK_SIZE
>>> @@ -62,7 +62,7 @@ extern unsigned int vced_count, vcei_count;
>>>    * support 16TB; the architectural reserve for future expansion is
>>>    * 8192EB ...
>>>    */
>>> -#define TASK_SIZE32    0x7fff8000UL
>>> +#define TASK_SIZE32    (0x7fff8000UL & PAGE_SIZE)
>>
>> Same here.
>
> As above.
>
>>
>>>   #define TASK_SIZE64    0x10000000000UL
>>>   #define TASK_SIZE (test_thread_flag(TIF_32BIT_ADDR) ? TASK_SIZE32 :
>>> TASK_SIZE64)
>>>   #define STACK_TOP_MAX    TASK_SIZE64
>>>
>>
>
> Thanks,
>
> Harvey