MIPS : die at free_initmem() function 3.9+

MIPS : die at free_initmem() function 3.9+

[EUNBONG SONG]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=euc-kr, Size: 3445 bytes --]


Hello. I try to boot my cavium board with david's patch. 
It's is not applied yet in linux tree, i got the patch from mailing list.
And the patch is as follow.


This is only very lightly tested, we need more testing before
declaring it the definitive fix.

 arch/mips/kernel/genex.S | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/arch/mips/kernel/genex.S b/arch/mips/kernel/genex.S
index ecb347c..57cda9a 100644
--- a/arch/mips/kernel/genex.S
+++ b/arch/mips/kernel/genex.S
@@ -132,12 +132,13 @@ LEAF(r4k_wait)
        .set    noreorder
        /* start of rollback region */
        LONG_L  t0, TI_FLAGS($28)
-       nop
        andi    t0, _TIF_NEED_RESCHED
        bnez    t0, 1f
         nop
-       nop
-       nop
+       /* Enable interrupts so WAIT will complete */
+       mfc0    t0, CP0_STATUS
+       ori     t0, ST0_IE
+       mtc0    t0, CP0_STATUS
        .set    mips3
        wait
        /* end of rollback region (the region size must be power of two) */

I think, it works well. But i encounter another problem at free_initmem(). 
The log messages are as follow.

[  132.134719] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W    3.9.0+ #29
[  132.141678] Stack : 0000000000000004 000000000000003f ffffffff80fa0000 ffffffff802924a8
          0000000000000000 ffffffff80fa0000 00000000000000ff ffffffff80293760
          0000000000000000 0000000000000000 ffffffff81080000 ffffffff81080000
          ffffffff80e2baf0 ffffffff80f93977 a80000004146cbb8 0000000000000020
          0000000000000003 0000000000000020 a800000041473da8 ffffffff810f0000
          a800000041473a10 ffffffff806ef910 a800000041473828 ffffffff80290920
          0000000000000000 ffffffff80293b90 000000000000000a ffffffff80e2baf0
          0000000000000000 a800000041473750 000000004146cef8 ffffffff805e7794
          0000000000000000 0000000000000000 0000000000000000 0000000000000000
          0000000000000000 ffffffff80272498 0000000000000000 0000000000000000
          ...
[  132.207201] Call Trace:
[  132.209655] [<ffffffff80272498>] show_stack+0x68/0x80
[  132.225943] [<ffffffff802bd4ac>] notifier_call_chain+0x5c/0xa8
[  132.231776] [<ffffffff802bdb84>] __atomic_notifier_call_chain+0x3c/0x58
[  132.238391] [<ffffffff802bdbe8>] notify_die+0x38/0x48
[  132.243442] [<ffffffff802716cc>] die+0x4c/0x148
[  132.247974] [<ffffffff8027f998>] do_page_fault+0x4b8/0x500
[  132.253461] [<ffffffff8026c764>] resume_userspace_check+0x0/0x10
[  132.259469] [<ffffffff80324a54>] free_reserved_area+0x8c/0x178
[  132.265304] [<ffffffff806e0dc8>] kernel_init+0x20/0x100
[  132.270529] [<ffffffff8026c7e0>] ret_from_kernel_thread+0x10/0x18

And i just changed free_initmem() functions as follow

diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
index 9b973e0..e246e9b 100644
--- a/arch/mips/mm/init.c
+++ b/arch/mips/mm/init.c
@@ -447,7 +447,10 @@ void free_initrd_mem(unsigned long start, unsigned long end)
 void __init_refok free_initmem(void)
 {
        prom_free_prom_memory();
-       free_initmem_default(POISON_FREE_INITMEM);
+
+       free_init_pages("unused kernel memory",
+                       __pa_symbol(&__init_begin),
+                       __pa_symbol(&__init_end));
 }

After that it works well. but i don't know why it works well.

Thanks. ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±þG«éÿŠ{ayº\x1dʇڙë,j\a­¢f£¢·hšïêÿ‘êçz_è®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿÿ¾\a«þG«éÿ¢¸?™¨è­Ú&£ø§~á¶iO•æ¬z·švØ^\x14\x04\x1a¶^[m§ÿÿÃ\fÿ¶ìÿ¢¸?–I¥

Re: MIPS : die at free_initmem() function 3.9+

[Jiang Liu]

Hi Eunbong,
	Thanks for reporting!
	I think this issue may be caused by __pa_symbol() on 64bits MIPS machines
due to following definition:
#ifdef CONFIG_64BIT
#define __pa(x)                                                         \
({                                                                      \
    unsigned long __x = (unsigned long)(x);                             \
    __x < CKSEG0 ? XPHYSADDR(__x) : CPHYSADDR(__x);                     \
})
#else
#define __pa(x)                                                         \
    ((unsigned long)(x) - PAGE_OFFSET + PHYS_OFFSET)
#endif
#define __va(x)         ((void *)((unsigned long)(x) + PAGE_OFFSET - PHYS_OFFSET))

So on 64bits MIPS platforms, __va(__pa(x)) may not equal to x, that may cause
trouble to free_initmem_default(). Could you please help to do another test
by changing
	free_initmem_default(POISON_FREE_INITMEM);
to
	free_initmem_default(0);

This test could help to identify whether this panic is caused by
	memset((void *)pos, poison, PAGE_SIZE);
in function free_reserved_area().

Thanks!
Gerry

On 05/03/2013 03:47 PM, EUNBONG SONG wrote:
> 
> Hello. I try to boot my cavium board with david's patch. 
> It's is not applied yet in linux tree, i got the patch from mailing list.
> And the patch is as follow.
> 
> 
> This is only very lightly tested, we need more testing before
> declaring it the definitive fix.
> 
>  arch/mips/kernel/genex.S | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/mips/kernel/genex.S b/arch/mips/kernel/genex.S
> index ecb347c..57cda9a 100644
> --- a/arch/mips/kernel/genex.S
> +++ b/arch/mips/kernel/genex.S
> @@ -132,12 +132,13 @@ LEAF(r4k_wait)
>         .set    noreorder
>         /* start of rollback region */
>         LONG_L  t0, TI_FLAGS($28)
> -       nop
>         andi    t0, _TIF_NEED_RESCHED
>         bnez    t0, 1f
>          nop
> -       nop
> -       nop
> +       /* Enable interrupts so WAIT will complete */
> +       mfc0    t0, CP0_STATUS
> +       ori     t0, ST0_IE
> +       mtc0    t0, CP0_STATUS
>         .set    mips3
>         wait
>         /* end of rollback region (the region size must be power of two) */
> 
> I think, it works well. But i encounter another problem at free_initmem(). 
> The log messages are as follow.
> 
> [  132.134719] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W    3.9.0+ #29
> [  132.141678] Stack : 0000000000000004 000000000000003f ffffffff80fa0000 ffffffff802924a8
>           0000000000000000 ffffffff80fa0000 00000000000000ff ffffffff80293760
>           0000000000000000 0000000000000000 ffffffff81080000 ffffffff81080000
>           ffffffff80e2baf0 ffffffff80f93977 a80000004146cbb8 0000000000000020
>           0000000000000003 0000000000000020 a800000041473da8 ffffffff810f0000
>           a800000041473a10 ffffffff806ef910 a800000041473828 ffffffff80290920
>           0000000000000000 ffffffff80293b90 000000000000000a ffffffff80e2baf0
>           0000000000000000 a800000041473750 000000004146cef8 ffffffff805e7794
>           0000000000000000 0000000000000000 0000000000000000 0000000000000000
>           0000000000000000 ffffffff80272498 0000000000000000 0000000000000000
>           ...
> [  132.207201] Call Trace:
> [  132.209655] [<ffffffff80272498>] show_stack+0x68/0x80
> [  132.225943] [<ffffffff802bd4ac>] notifier_call_chain+0x5c/0xa8
> [  132.231776] [<ffffffff802bdb84>] __atomic_notifier_call_chain+0x3c/0x58
> [  132.238391] [<ffffffff802bdbe8>] notify_die+0x38/0x48
> [  132.243442] [<ffffffff802716cc>] die+0x4c/0x148
> [  132.247974] [<ffffffff8027f998>] do_page_fault+0x4b8/0x500
> [  132.253461] [<ffffffff8026c764>] resume_userspace_check+0x0/0x10
> [  132.259469] [<ffffffff80324a54>] free_reserved_area+0x8c/0x178
> [  132.265304] [<ffffffff806e0dc8>] kernel_init+0x20/0x100
> [  132.270529] [<ffffffff8026c7e0>] ret_from_kernel_thread+0x10/0x18
> 
> And i just changed free_initmem() functions as follow
> 
> diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
> index 9b973e0..e246e9b 100644
> --- a/arch/mips/mm/init.c
> +++ b/arch/mips/mm/init.c
> @@ -447,7 +447,10 @@ void free_initrd_mem(unsigned long start, unsigned long end)
>  void __init_refok free_initmem(void)
>  {
>         prom_free_prom_memory();
> -       free_initmem_default(POISON_FREE_INITMEM);
> +
> +       free_init_pages("unused kernel memory",
> +                       __pa_symbol(&__init_begin),
> +                       __pa_symbol(&__init_end));
>  }
> 
> After that it works well. but i don't know why it works well.
> 
> Thanks. 
> 

Re: Re: MIPS : die at free_initmem() function 3.9+

[EUNBONG SONG]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=euc-kr, Size: 792 bytes --]


>So on 64bits MIPS platforms, __va(__pa(x)) may not equal to x, that may cause
>trouble to free_initmem_default(). Could you please help to do another test
>by changing
>free_initmem_default(POISON_FREE_INITMEM);
>to
>free_initmem_default(0);

>This test could help to identify whether this panic is caused by
>memset((void *)pos, poison, PAGE_SIZE);
>in function free_reserved_area().

Hello, as you said i changed  "free_initmem_default(POISON_FREE_INITMEM);" to
"free_initmem_default(0);". Panic still occurred. 
Actually, i put the some debug messages. and i confirmed panic occurs in __free_reserved_page() function.
Thanks!

ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±þG«éÿŠ{ayº\x1dʇڙë,j\a­¢f£¢·hšïêÿ‘êçz_è®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿÿ¾\a«þG«éÿ¢¸?™¨è­Ú&£ø§~á¶iO•æ¬z·švØ^\x14\x04\x1a¶^[m§ÿÿÃ\fÿ¶ìÿ¢¸?–I¥

Re: MIPS : die at free_initmem() function 3.9+

[Jiang Liu]

Hi Eunsong,
	Thanks for your help. I have done some code inspection and have 
found a possible reason for this issue. It seems that virt_to_page()
can't be used to handle address in compatible segments due to following
definitions.

#define virt_to_page(kaddr)     pfn_to_page(PFN_DOWN(virt_to_phys(kaddr)))
static inline unsigned long virt_to_phys(volatile const void *address)
{
        return (unsigned long)address - PAGE_OFFSET + PHYS_OFFSET;
}
#define __va(x)         ((void *)((unsigned long)(x) + PAGE_OFFSET - PHYS_OFFSET))
#define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
#define __pa(x)                                                         \
({                                                                      \
    unsigned long __x = (unsigned long)(x);                             \
    __x < CKSEG0 ? XPHYSADDR(__x) : CPHYSADDR(__x);                     \
})

#define CPHYSADDR(a)            ((_ACAST32_(a)) & 0x1fffffff)
#define XPHYSADDR(a)            ((_ACAST64_(a)) &                       \
                                 _CONST64_(0x000000ffffffffff))

#define XKUSEG                  _CONST64_(0x0000000000000000)
#define XKSSEG                  _CONST64_(0x4000000000000000)
#define XKPHYS                  _CONST64_(0x8000000000000000)
#define XKSEG                   _CONST64_(0xc000000000000000)
#define CKSEG0                  _CONST64_(0xffffffff80000000)
#define CKSEG1                  _CONST64_(0xffffffffa0000000)
#define CKSSEG                  _CONST64_(0xffffffffc0000000)
#define CKSEG3                  _CONST64_(0xffffffffe0000000)

So could you please help to prepare a formal patch for this bug and send
it to Andrew Morton for v3.10-rc1? Otherwise I could help to it too.

Regards!
Gerry


=On 05/06/2013 06:39 AM, EUNBONG SONG wrote:
> 
>> So on 64bits MIPS platforms, __va(__pa(x)) may not equal to x, that may cause
>> trouble to free_initmem_default(). Could you please help to do another test
>> by changing
>> free_initmem_default(POISON_FREE_INITMEM);
>> to
>> free_initmem_default(0);
> 
>> This test could help to identify whether this panic is caused by
>> memset((void *)pos, poison, PAGE_SIZE);
>> in function free_reserved_area().
> 
> Hello, as you said i changed  "free_initmem_default(POISON_FREE_INITMEM);" to
> "free_initmem_default(0);". Panic still occurred. 
> Actually, i put the some debug messages. and i confirmed panic occurs in __free_reserved_page() function.
> Thanks!
> 
> N떑꿩�r툤y鉉싕b쾊Ф푤v�^�)頻{.n�+돴쪐{콗喩zX㎍\x17썳變}찠꼿쟺�&j:+v돣�\a쳭喩zZ+�+zf＂톒쉱�~넮녬i�鎬z�\x1e췿ⅱ�?솳鈺�&�)刪^[f뷌^j푹y쬶끷@A첺뛴��\f0띠h�\x0f�i�
> 

Re: MIPS : die at free_initmem() function 3.9+

[David Daney]

On 05/07/2013 09:46 AM, Jiang Liu wrote:
> Hi Eunsong,
> 	Thanks for your help. I have done some code inspection and have
> found a possible reason for this issue. It seems that virt_to_page()
> can't be used to handle address in compatible segments due to following
> definitions.

This has caused multiple problems in the past.  I wrote a patch to 
correct this problem, but it looks like it was never applied.

I just sent the patch to you guys as a separate e-mail.

David Daney



>
> #define virt_to_page(kaddr)     pfn_to_page(PFN_DOWN(virt_to_phys(kaddr)))
> static inline unsigned long virt_to_phys(volatile const void *address)
> {
>          return (unsigned long)address - PAGE_OFFSET + PHYS_OFFSET;
> }
> #define __va(x)         ((void *)((unsigned long)(x) + PAGE_OFFSET - PHYS_OFFSET))
> #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
> #define __pa(x)                                                         \
> ({                                                                      \
>      unsigned long __x = (unsigned long)(x);                             \
>      __x < CKSEG0 ? XPHYSADDR(__x) : CPHYSADDR(__x);                     \
> })
>
> #define CPHYSADDR(a)            ((_ACAST32_(a)) & 0x1fffffff)
> #define XPHYSADDR(a)            ((_ACAST64_(a)) &                       \
>                                   _CONST64_(0x000000ffffffffff))
>
> #define XKUSEG                  _CONST64_(0x0000000000000000)
> #define XKSSEG                  _CONST64_(0x4000000000000000)
> #define XKPHYS                  _CONST64_(0x8000000000000000)
> #define XKSEG                   _CONST64_(0xc000000000000000)
> #define CKSEG0                  _CONST64_(0xffffffff80000000)
> #define CKSEG1                  _CONST64_(0xffffffffa0000000)
> #define CKSSEG                  _CONST64_(0xffffffffc0000000)
> #define CKSEG3                  _CONST64_(0xffffffffe0000000)
>
> So could you please help to prepare a formal patch for this bug and send
> it to Andrew Morton for v3.10-rc1? Otherwise I could help to it too.
>
> Regards!
> Gerry
>
>
> =On 05/06/2013 06:39 AM, EUNBONG SONG wrote:
>>
>>> So on 64bits MIPS platforms, __va(__pa(x)) may not equal to x, that may cause
>>> trouble to free_initmem_default(). Could you please help to do another test
>>> by changing
>>> free_initmem_default(POISON_FREE_INITMEM);
>>> to
>>> free_initmem_default(0);
>>
>>> This test could help to identify whether this panic is caused by
>>> memset((void *)pos, poison, PAGE_SIZE);
>>> in function free_reserved_area().
>>
>> Hello, as you said i changed  "free_initmem_default(POISON_FREE_INITMEM);" to
>> "free_initmem_default(0);". Panic still occurred.
>> Actually, i put the some debug messages. and i confirmed panic occurs in __free_reserved_page() function.
>> Thanks!
>>
>> N떑꿩�r툤y鉉싕b쾊Ф푤v�^�)頻{.n�+돴쪐{콗喩zX㎍\x17썳變}찠꼿쟺�&j:+v돣�\a쳭喩zZ+�+zf＂톒쉱�~넮녬i�鎬z�\x1e췿ⅱ�?솳鈺�&�)刪^[f뷌^j푹y쬶끷@A첺뛴��\f0띠h�\x0f�i�
>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
>