[U-Boot] SPL variant of sunxi nand module

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

Hi all,

While fighting with this A20-olinuxino nand boot process I've initially 
found that some essential nand-related parameters are apparently missing 
by default, preventing reasonable nand operation, so I started debugging 
this gradually.

 > U-Boot SPL 2019.01-rc2 (Dec 20 2018 - 16:30:46 +0300)
 > CPU: 912000000Hz, AXI/AHB/APB: 3/2/2
 > DRAM: 1024 MiB
 > Trying to boot from NAND

Now I've discovered that SPL-build variant of sunxi_nand module is 
missing the sunxi_nand_init() whatsoever. Apparently it is on purpose, 
because:

#ifndef CONFIG_SPL_BUILD
         sunxi_nand_init();
#endif

And I cannot simply un-ifdef sunxi_nand_init() here, because the SPL 
build would fail with undefined symbol then, but I'm wondering, where 
are all values (like ECC strength and such) are expected to come from 
instead? Is SPL nand loader is supposed to be usable at all currently?


Thank you,

Regards,
Nikolai

[U-Boot] SPL variant of sunxi nand module

[Michael Nazzareno Trimarchi]

Hi

On Sun, Dec 23, 2018 at 4:46 PM Nikolai Zhubr <n-a-zhubr@yandex.ru> wrote:
>
> Hi again,
>
> 23.12.2018 16:29, I wrote:
> >  > U-Boot SPL 2019.01-rc2 (Dec 20 2018 - 16:30:46 +0300)
> >  > CPU: 912000000Hz, AXI/AHB/APB: 3/2/2
> >  > DRAM: 1024 MiB
> >  > Trying to boot from NAND
> >
>
> Ok, discovered a special SPL-only sunxi_nand_spl variant, added some
> debugging, so the detection is visible:
>
> In nand_detect_config(), start detection...
> Considering addr_cycles=5, page_size=2048
>   Considering ecc_size=1024, ecc_strength=0 failed(a).
> Considering addr_cycles=5, page_size=2048 rejected.
> Considering addr_cycles=5, page_size=4096
>   Considering ecc_size=1024, ecc_strength=3 failed(a).
> Considering addr_cycles=5, page_size=4096 rejected.
> Considering addr_cycles=5, page_size=8192
>   Considering ecc_size=1024, ecc_strength=4 good(b).
> Considering addr_cycles=5, page_size=8192 accepted.
>
> I'm almost 100% sure that correct config would be page_size=8192,
> ecc_size=1024, ecc_strength=40 (from nand chip identification structure
> for regular linux kernel)

That is an index on an array. Am I wrong? so the max is 74

Michael

>
> Now the detection routine in sunxi_nand_spl apparently comes up with a
> value of ecc_strength=4 instead... Why is that?

n - 1 using an index, if the code that I have is aligned so

Michael

>
>
> Thank you,
>
> Regards,
> Nikolai
>
>
> >
> >
> > Thank you,
> >
> > Regards,
> > Nikolai
> > _______________________________________________
> > U-Boot mailing list
> > U-Boot at lists.denx.de
> > https://lists.denx.de/listinfo/u-boot
>
> _______________________________________________
> U-Boot mailing list
> U-Boot at lists.denx.de
> https://lists.denx.de/listinfo/u-boot



-- 
| Michael Nazzareno Trimarchi                     Amarula Solutions BV |
| COO  -  Founder                                      Cruquiuskade 47 |
| +31(0)851119172                                 Amsterdam 1018 AM NL |
|                  [`as] http://www.amarulasolutions.com               |

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

Hi again,

23.12.2018 16:29, I wrote:
>  > U-Boot SPL 2019.01-rc2 (Dec 20 2018 - 16:30:46 +0300)
>  > CPU: 912000000Hz, AXI/AHB/APB: 3/2/2
>  > DRAM: 1024 MiB
>  > Trying to boot from NAND
>

Ok, discovered a special SPL-only sunxi_nand_spl variant, added some 
debugging, so the detection is visible:

In nand_detect_config(), start detection...
Considering addr_cycles=5, page_size=2048
  Considering ecc_size=1024, ecc_strength=0 failed(a).
Considering addr_cycles=5, page_size=2048 rejected.
Considering addr_cycles=5, page_size=4096
  Considering ecc_size=1024, ecc_strength=3 failed(a).
Considering addr_cycles=5, page_size=4096 rejected.
Considering addr_cycles=5, page_size=8192
  Considering ecc_size=1024, ecc_strength=4 good(b).
Considering addr_cycles=5, page_size=8192 accepted.

I'm almost 100% sure that correct config would be page_size=8192, 
ecc_size=1024, ecc_strength=40 (from nand chip identification structure 
for regular linux kernel)

Now the detection routine in sunxi_nand_spl apparently comes up with a 
value of ecc_strength=4 instead... Why is that?


Thank you,

Regards,
Nikolai


>
>
> Thank you,
>
> Regards,
> Nikolai
> _______________________________________________
> U-Boot mailing list
> U-Boot at lists.denx.de
> https://lists.denx.de/listinfo/u-boot

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

Hi Michael,

23.12.2018 18:54, Michael Nazzareno Trimarchi:
[...]
>>    Considering ecc_size=1024, ecc_strength=4 good(b).
>> Considering addr_cycles=5, page_size=8192 accepted.
>>
>> I'm almost 100% sure that correct config would be page_size=8192,
>> ecc_size=1024, ecc_strength=40 (from nand chip identification structure
>> for regular linux kernel)
>
> That is an index on an array. Am I wrong? so the max is 74

Indeed, its an index, and therefore it appears SPL's detection actually 
gives correct values! Nevertheless, SPL reads all zeroes from nand.

On the other hand, normal U-boot variant reads correct data from nand, 
but I'm not yet sure if it uses correct parameters and specifically, if 
they match those of SPL version. Need to add yet more debugging output.


Thank you,

Regards,
Nikolai

> Michael
>
>>
>> Now the detection routine in sunxi_nand_spl apparently comes up with a
>> value of ecc_strength=4 instead... Why is that?
>
> n - 1 using an index, if the code that I have is aligned so
>
> Michael
>
>>
>>
>> Thank you,
>>
>> Regards,
>> Nikolai
>>
>>
>>>
>>>
>>> Thank you,
>>>
>>> Regards,
>>> Nikolai
>>> _______________________________________________
>>> U-Boot mailing list
>>> U-Boot at lists.denx.de
>>> https://lists.denx.de/listinfo/u-boot
>>
>> _______________________________________________
>> U-Boot mailing list
>> U-Boot at lists.denx.de
>> https://lists.denx.de/listinfo/u-boot
>
>
>

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

Hi again,

23.12.2018 21:11, I wrote:
[...]
> Indeed, its an index, and therefore it appears SPL's detection actually
> gives correct values! Nevertheless, SPL reads all zeroes from nand.

Regarding the sunxi_nand_spl.c module, I can not find any mention if it 
implies NAND_ECC_HW, NAND_ECC_HW_SYNDROME, or rather some other mode, or 
if these modes are irrelevant in this case?

Can not guess browsing through the code, it looks too different from 
sunxi_nand.c to compare.


Thank you,

Regards,
Nikolai


> On the other hand, normal U-boot variant reads correct data from nand,
> but I'm not yet sure if it uses correct parameters and specifically, if
> they match those of SPL version. Need to add yet more debugging output.
>
>
> Thank you,
>
> Regards,
> Nikolai
>
>> Michael
>>
>>>
>>> Now the detection routine in sunxi_nand_spl apparently comes up with a
>>> value of ecc_strength=4 instead... Why is that?
>>
>> n - 1 using an index, if the code that I have is aligned so
>>
>> Michael
>>
>>>
>>>
>>> Thank you,
>>>
>>> Regards,
>>> Nikolai
>>>
>>>
>>>>
>>>>
>>>> Thank you,
>>>>
>>>> Regards,
>>>> Nikolai
>>>> _______________________________________________
>>>> U-Boot mailing list
>>>> U-Boot at lists.denx.de
>>>> https://lists.denx.de/listinfo/u-boot
>>>
>>> _______________________________________________
>>> U-Boot mailing list
>>> U-Boot at lists.denx.de
>>> https://lists.denx.de/listinfo/u-boot
>>
>>
>>
>
> _______________________________________________
> U-Boot mailing list
> U-Boot at lists.denx.de
> https://lists.denx.de/listinfo/u-boot

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

23.12.2018 21:58, I wrote:
[...]
> Regarding the sunxi_nand_spl.c module, I can not find any mention if it
> implies NAND_ECC_HW, NAND_ECC_HW_SYNDROME, or rather some other mode, or
> if these modes are irrelevant in this case?

So far I'm observing that sunxi_nand_spl module refuses to load erased 
nand pages (500000 in the example below), which is correct, but it 
"successfully" loads filled nand pages as totally zeroed out, which is 
supposedly incorrect. This is in contrast with sunxi_nand module, that 
loads filled pages correctly, at least in my simple tests.

So while detection in sunxi_nand_spl seems fine, reading seems not.

I think digging deeper without some additional advice is a bit beyond my 
capabilities, although most probably some very tiny bits are missing...


Thank you,

Regards,
Nikolai

U-Boot SPL 2019.01-rc2 (Dec 23 2018 - 23:23:57 +0300)
CPU: 912000000Hz, AXI/AHB/APB: 3/2/2
DRAM: 1024 MiB
 >>SPL: board_init_r()
Trying to boot from NAND
spl: nand - using hw ecc
[zh]nand_spl_load_image(500000, 40)=-22
[zh]Read dump of page 768 (nand offs 600000, page_size 8192):
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
[zh]Read dump of page 768 (nand offs 600000, page_size 8192):
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
=====================================================
U-Boot 2019.01-rc2 (Dec 23 2018 - 22:30:35 +0300) Allwinner Technology

CPU:   Allwinner A20 (SUN7I)
Model: Olimex A20-Olinuxino Micro
I2C:   ready
DRAM:  1 GiB
NAND:  nand_base: device found, Manufacturer ID: 0xad, Chip ID: 0xd7
nand_base: Hynix H27UBG8T2BTR-BC 32G 3.3V 8-bit
nand_base: Hynix H27UBG8T2BTR-BC 32G 3.3V 8-bit
nand_base: 4096 MiB, MLC, erase size: 2048 KiB, page size: 8192, OOB 
size: 640
Using ECC step 1024, strength 40, mode 2
4096 MiB
In:    serial
Out:   serial
Err:   serial
SCSI:  SATA link 0 timeout.
AHCI 0001.0100 32 slots 1 ports 3 Gbps 0x1 impl SATA mode
flags: ncq stag pm led clo only pmp pio slum part ccc apst

Net:   Can't get reset: -524
eth0: ethernet at 1c50000
Hit any key to stop autoboot:  0
=> nand read  0x45000000 0x400000 0x1000

NAND read: device 0 offset 0x400000, size 0x1000
Scanning device for bad blocks
Bad eraseblock 0 at 0x0000001fe000
Bad eraseblock 1 at 0x0000003fe000
  4096 bytes read: OK
=> md.b 0x45000000 100
45000000: b8 00 00 ea 14 f0 9f e5 14 f0 9f e5 14 f0 9f e5 
................
=====================================================


>
> Can not guess browsing through the code, it looks too different from
> sunxi_nand.c to compare.
>
>
> Thank you,
>
> Regards,
> Nikolai
>
>
>> On the other hand, normal U-boot variant reads correct data from nand,
>> but I'm not yet sure if it uses correct parameters and specifically, if
>> they match those of SPL version. Need to add yet more debugging output.
>>
>>
>> Thank you,
>>
>> Regards,
>> Nikolai
>>
>>> Michael
>>>
>>>>
>>>> Now the detection routine in sunxi_nand_spl apparently comes up with a
>>>> value of ecc_strength=4 instead... Why is that?
>>>
>>> n - 1 using an index, if the code that I have is aligned so
>>>
>>> Michael
>>>
>>>>
>>>>
>>>> Thank you,
>>>>
>>>> Regards,
>>>> Nikolai
>>>>
>>>>
>>>>>
>>>>>
>>>>> Thank you,
>>>>>
>>>>> Regards,
>>>>> Nikolai
>>>>> _______________________________________________
>>>>> U-Boot mailing list
>>>>> U-Boot at lists.denx.de
>>>>> https://lists.denx.de/listinfo/u-boot
>>>>
>>>> _______________________________________________
>>>> U-Boot mailing list
>>>> U-Boot at lists.denx.de
>>>> https://lists.denx.de/listinfo/u-boot
>>>
>>>
>>>
>>
>> _______________________________________________
>> U-Boot mailing list
>> U-Boot at lists.denx.de
>> https://lists.denx.de/listinfo/u-boot
>

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

Hi again,

The following commit (titled "use PIO instead of DMA") apparently broke 
actual reading of nand pages in SPL at least for A20:

http://git.denx.de/?p=u-boot.git;a=commit;h=6ddbb1e936c78cdef1e7395039fa7020c5c75326

Instead of reading page contents, non-dma (current) version just feeds 
zeroes here.
Tested on A20-olinuxino-micro with a Hynix 4GB nand chip.

It'd be nice to somehow fix it back eventually :)

Other than that, nand boot seems usable (with added necessary chip id).


Thank you,

Regards,
Nikolai

24.12.2018 0:00, I wrote:
> 23.12.2018 21:58, I wrote:
> [...]
>> Regarding the sunxi_nand_spl.c module, I can not find any mention if it
>> implies NAND_ECC_HW, NAND_ECC_HW_SYNDROME, or rather some other mode, or
>> if these modes are irrelevant in this case?
>
> So far I'm observing that sunxi_nand_spl module refuses to load erased
> nand pages (500000 in the example below), which is correct, but it
> "successfully" loads filled nand pages as totally zeroed out, which is
> supposedly incorrect. This is in contrast with sunxi_nand module, that
> loads filled pages correctly, at least in my simple tests.
>
> So while detection in sunxi_nand_spl seems fine, reading seems not.
>
> I think digging deeper without some additional advice is a bit beyond my
> capabilities, although most probably some very tiny bits are missing...
>
>
> Thank you,
>
> Regards,
> Nikolai
>
> U-Boot SPL 2019.01-rc2 (Dec 23 2018 - 23:23:57 +0300)
> CPU: 912000000Hz, AXI/AHB/APB: 3/2/2
> DRAM: 1024 MiB
>  >>SPL: board_init_r()
> Trying to boot from NAND
> spl: nand - using hw ecc
> [zh]nand_spl_load_image(500000, 40)=-22
> [zh]Read dump of page 768 (nand offs 600000, page_size 8192):
> 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
> [zh]Read dump of page 768 (nand offs 600000, page_size 8192):
> 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
> =====================================================
> U-Boot 2019.01-rc2 (Dec 23 2018 - 22:30:35 +0300) Allwinner Technology
>
> CPU: Allwinner A20 (SUN7I)
> Model: Olimex A20-Olinuxino Micro
> I2C: ready
> DRAM: 1 GiB
> NAND: nand_base: device found, Manufacturer ID: 0xad, Chip ID: 0xd7
> nand_base: Hynix H27UBG8T2BTR-BC 32G 3.3V 8-bit
> nand_base: Hynix H27UBG8T2BTR-BC 32G 3.3V 8-bit
> nand_base: 4096 MiB, MLC, erase size: 2048 KiB, page size: 8192, OOB
> size: 640
> Using ECC step 1024, strength 40, mode 2
> 4096 MiB
> In: serial
> Out: serial
> Err: serial
> SCSI: SATA link 0 timeout.
> AHCI 0001.0100 32 slots 1 ports 3 Gbps 0x1 impl SATA mode
> flags: ncq stag pm led clo only pmp pio slum part ccc apst
>
> Net: Can't get reset: -524
> eth0: ethernet at 1c50000
> Hit any key to stop autoboot: 0
> => nand read 0x45000000 0x400000 0x1000
>
> NAND read: device 0 offset 0x400000, size 0x1000
> Scanning device for bad blocks
> Bad eraseblock 0 at 0x0000001fe000
> Bad eraseblock 1 at 0x0000003fe000
> 4096 bytes read: OK
> => md.b 0x45000000 100
> 45000000: b8 00 00 ea 14 f0 9f e5 14 f0 9f e5 14 f0 9f e5 ................
> =====================================================
>
>
>>
>> Can not guess browsing through the code, it looks too different from
>> sunxi_nand.c to compare.
>>
>>
>> Thank you,
>>
>> Regards,
>> Nikolai
>>
>>
>>> On the other hand, normal U-boot variant reads correct data from nand,
>>> but I'm not yet sure if it uses correct parameters and specifically, if
>>> they match those of SPL version. Need to add yet more debugging output.
>>>
>>>
>>> Thank you,
>>>
>>> Regards,
>>> Nikolai
>>>
>>>> Michael
>>>>
>>>>>
>>>>> Now the detection routine in sunxi_nand_spl apparently comes up with a
>>>>> value of ecc_strength=4 instead... Why is that?
>>>>
>>>> n - 1 using an index, if the code that I have is aligned so
>>>>
>>>> Michael
>>>>
>>>>>
>>>>>
>>>>> Thank you,
>>>>>
>>>>> Regards,
>>>>> Nikolai
>>>>>
>>>>>
>>>>>>
>>>>>>
>>>>>> Thank you,
>>>>>>
>>>>>> Regards,
>>>>>> Nikolai
>>>>>> _______________________________________________
>>>>>> U-Boot mailing list
>>>>>> U-Boot at lists.denx.de
>>>>>> https://lists.denx.de/listinfo/u-boot
>>>>>
>>>>> _______________________________________________
>>>>> U-Boot mailing list
>>>>> U-Boot at lists.denx.de
>>>>> https://lists.denx.de/listinfo/u-boot
>>>>
>>>>
>>>>
>>>
>>> _______________________________________________
>>> U-Boot mailing list
>>> U-Boot at lists.denx.de
>>> https://lists.denx.de/listinfo/u-boot
>>
>

[U-Boot] SPL variant of sunxi nand module

[Nikolai Zhubr]

Hi,

Its just to remind sunxi_nand_spl.c is still broken. (And probably has 
been for a couple releases or so).
I'm unable to fix it myself (other than reverting the mentioned commit 
locally.)


Thank you,

Regards,
Nikolai

24.12.2018 0:47, Nikolai Zhubr:
> Hi again,
>
> The following commit (titled "use PIO instead of DMA") apparently broke
> actual reading of nand pages in SPL at least for A20:
>
> http://git.denx.de/?p=u-boot.git;a=commit;h=6ddbb1e936c78cdef1e7395039fa7020c5c75326
>
>
> Instead of reading page contents, non-dma (current) version just feeds
> zeroes here.
> Tested on A20-olinuxino-micro with a Hynix 4GB nand chip.
>
> It'd be nice to somehow fix it back eventually :)
>
> Other than that, nand boot seems usable (with added necessary chip id).
>
>
> Thank you,
>
> Regards,
> Nikolai
>