Skip to contentSkip to conversation

A23 u-boot with SPL / dram init available in my personal git repo

U-Boot Hans de Goede 6 participants 15 messages Dec 7, 2014 – Dec 18, 2014

lore
Share & tools
lore
#0 lore raw

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Regards,

Hans

#1 lore raw
Quoted 4 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

Hip, Hip Hooray. Thank you.

How did you pull it off? Did you find documentation somewhere? Or by piecing things together?

Quoted 11 lines

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Regards,

Hans

#2 lore raw
In reply to Hans de Goede
Quoted 7 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

Hopefully I'll get around to testing this. BTW, what tablet do you have?

Quoted 3 lines

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Yes. With that fix my Hummingbird A31 boots properly without raising DCDC1 to 3.3V.

Thanks! I'll send the defconfig out later.

ChenYu

#3 lore raw

Hi,

Quoted 11 lines · Chen-Yu Tsai
Quoted 7 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

Hopefully I'll get around to testing this. BTW, what tablet do you have?

I've an Ippo q8h v1.2, there are at least 2 differences from the v5 you've. The dram clk speed and zq value are different, and the wifi is different. The wifi does not matter for u-boot, but does mean we need separate dtb files for the 2.

My u-boot sunxi-wip branch has a defconfig for the v1.2, you should be able to copy that over to the v5 defconfig, adjust dram clk and zq values with the ones from the v5 fix, change v1.2 to v5 in the CONFIG_FDTFILE setting and thats it.

Let me know if this works, then I'll also include an update for the v5 defconfig to enable the SPL when I send this upstream.

Quoted 8 lines
Quoted 3 lines

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Yes. With that fix my Hummingbird A31 boots properly without raising DCDC1 to 3.3V.

Thanks! I'll send the defconfig out later.

Good, so I guess this means that DCDC1 should be 3V for your board, since that is what the factory firmware uses, right ?

I guess it is time to make DCDC1 voltage configurable.

Regards,

Hans

#4 lore raw

Hi,

Quoted 24 lines · Hans de Goede

Hi,

Quoted 8 lines · Chen-Yu Tsai
Quoted 6 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

Quoted 6 lines
Quoted 2 lines

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

Hopefully I'll get around to testing this. BTW, what tablet do you have?

I've an Ippo q8h v1.2, there are at least 2 differences from the v5 you've. The dram clk speed and zq value are different, and the wifi is different. The wifi does not matter for u-boot, but does mean we need separate dtb files for the 2.

Quoted 7 lines

My u-boot sunxi-wip branch has a defconfig for the v1.2, you should be able to copy that over to the v5 defconfig, adjust dram clk and zq values with the ones from the v5 fix, change v1.2 to v5 in the CONFIG_FDTFILE setting and thats it.

Let me know if this works, then I'll also include an update for the v5 defconfig to enable the SPL when I send this upstream.

Just gave it a spin, and it works. The dram clock and zq can be found in sunxi-boards. Note that the actual settings fetched from stock u-boot has the highest bit of tpr13 set, disabling auto detection AFAICU. As you've not implemented that part, I used auto detection and it seems to work.

The only problem I have now is it has no connectivity to the outside world lol. I would've liked a v2, with a Realtek USB WiFi chip. What chip does the v1.2 have?

Quoted 8 lines
Quoted 8 lines
Quoted 3 lines

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Yes. With that fix my Hummingbird A31 boots properly without raising DCDC1 to 3.3V.

Quoted 7 lines
Quoted 1 line

Thanks! I'll send the defconfig out later.

Good, so I guess this means that DCDC1 should be 3V for your board, since that is what the factory firmware uses, right ?

I guess it is time to make DCDC1 voltage configurable.

3V seems to be the lowest tolerable voltage for some bits. Making it 3.3V might be a bit better. But with my light usage, I don't see any stability issues.

ChenYu

#5 lore raw

Hi,

Quoted 32 lines · Chen-Yu Tsai

Hi,

Quoted 29 lines · Hans de Goede

Hi,

Quoted 8 lines · Chen-Yu Tsai
Quoted 6 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

Quoted 6 lines
Quoted 2 lines

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

Hopefully I'll get around to testing this. BTW, what tablet do you have?

I've an Ippo q8h v1.2, there are at least 2 differences from the v5 you've. The dram clk speed and zq value are different, and the wifi is different. The wifi does not matter for u-boot, but does mean we need separate dtb files for the 2.

My u-boot sunxi-wip branch has a defconfig for the v1.2, you should be able to copy that over to the v5 defconfig, adjust dram clk and zq values with the ones from the v5 fix, change v1.2 to v5 in the CONFIG_FDTFILE setting and thats it.

Quoted 5 lines
Quoted 2 lines

Let me know if this works, then I'll also include an update for the v5 defconfig to enable the SPL when I send this upstream.

Just gave it a spin, and it works. The dram clock and zq can be found in sunxi-boards.

Good, I'll update my patch-set to reflect this.

Quoted 4 lines

Note that the actual settings fetched from
stock u-boot has the highest bit of tpr13 set, disabling auto detection AFAICU. As you've not implemented that part, I used auto detection and it seems to work.

It is implemented in the actual dram_init code, but for now I've all the dram_paras hardcoded except the clk and the zq value. If autodetect works I see no reason to change this, but for future models, if necessary the code for dealing with manual config is there, to test just put the values directly in the dram_para struct at the top of dram_sun8i.c .

Quoted 3 lines

The only problem I have now is it has no connectivity to the outside world lol. I would've liked a v2, with a Realtek USB WiFi chip. What chip does the v1.2 have?

An sdio based chip, the RDA RDA5990P, as mentioned here:

http://linux-sunxi.org/Ippo_q8h#Q8H-V1.2

It does have an empty spot on the pcb for a usb wifi module, I plan to hookup a usb connector to that one of these days, and see if I can get a USB host working this way.

Quoted 11 lines
Quoted 11 lines
Quoted 9 lines
Quoted 3 lines

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Yes. With that fix my Hummingbird A31 boots properly without raising DCDC1 to 3.3V.

Thanks! I'll send the defconfig out later.

Quoted 8 lines
Quoted 4 lines

Good, so I guess this means that DCDC1 should be 3V for your board, since that is what the factory firmware uses, right ?

I guess it is time to make DCDC1 voltage configurable.

3V seems to be the lowest tolerable voltage for some bits. Making it 3.3V might be a bit better. But with my light usage, I don't see any stability issues.

Right, I guess 3V is used to safe battery, which makes sense (in tablets / phones)

As said I'll make this configurable making 3V the default and set 3V3 in the defconfig where the original firmware does that.

Regards,

Hans

#6 lore raw
Quoted 55 lines · Hans de Goede
Quoted 31 lines · Chen-Yu Tsai
Quoted 30 lines · Hans de Goede
Quoted 9 lines · Chen-Yu Tsai
Quoted 7 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

Quoted 7 lines
Quoted 2 lines

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

Hopefully I'll get around to testing this. BTW, what tablet do you have?

I've an Ippo q8h v1.2, there are at least 2 differences from the v5 you've. The dram clk speed and zq value are different, and the wifi is different. The wifi does not matter for u-boot, but does mean we need separate dtb files for the 2.

My u-boot sunxi-wip branch has a defconfig for the v1.2, you should be able to copy that over to the v5 defconfig, adjust dram clk and zq values with the ones from the v5 fix, change v1.2 to v5 in the CONFIG_FDTFILE setting and thats it.

Quoted 6 lines
Quoted 2 lines

Let me know if this works, then I'll also include an update for the v5 defconfig to enable the SPL when I send this upstream.

Just gave it a spin, and it works. The dram clock and zq can be found in sunxi-boards.

Good, I'll update my patch-set to reflect this.

Quoted 4 lines

Note that the actual settings fetched from
stock u-boot has the highest bit of tpr13 set, disabling auto detection AFAICU. As you've not implemented that part, I used auto detection and it seems to work.

It is implemented in the actual dram_init code, but for now I've all the dram_paras hardcoded except the clk and the zq value. If autodetect works I see no reason to change this, but for future models, if necessary the code for dealing with manual config is there, to test just put the values directly in the dram_para struct at the top of dram_sun8i.c .

Quoted 8 lines
Quoted 3 lines

The only problem I have now is it has no connectivity to the outside world lol. I would've liked a v2, with a Realtek USB WiFi chip. What chip does the v1.2 have?

An sdio based chip, the RDA RDA5990P, as mentioned here:

http://linux-sunxi.org/Ippo_q8h#Q8H-V1.2

Not sure this would require a different DT. Probably the GPIOs are different.

Quoted 3 lines

It does have an empty spot on the pcb for a usb wifi module, I plan to hookup a usb connector to that one of these days, and see if I can get a USB host working this way.

I should do that too. Nice way to hook up ethernet. Small problem would be getting 5V from the board for VBUS.

Also I have a schematic file for A23 that someone provided me. I don't remember who...

It lists:
- ALDO2 connected to VCC-DLL @ 2.5V
- ALDO3 to VCC-PLL and AVCC (analog power) @ 3.0V
- DCDC1 to all the normal stuff @ 3V
- DCDC2 to VDD-SYS @ 1.1V
- DCDC3 to VDD-CPU (main cpu) @ 1.1V (1.26V in my fex file)
- DCDC5 to VCC-DRAM @ 1.1V (1.5V in my fex file)
- DC5LDO to VDD-CPUS (special cpu)

Quoted 24 lines
Quoted 16 lines
Quoted 16 lines
Quoted 10 lines
Quoted 3 lines

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Yes. With that fix my Hummingbird A31 boots properly without raising DCDC1 to 3.3V.

Thanks! I'll send the defconfig out later.

Good, so I guess this means that DCDC1 should be 3V for your board, since that is what the factory firmware uses, right ?

Quoted 6 lines
Quoted 1 line

I guess it is time to make DCDC1 voltage configurable.

3V seems to be the lowest tolerable voltage for some bits. Making it 3.3V might be a bit better. But with my light usage, I don't see any stability issues.

Quoted 5 lines

Right, I guess 3V is used to safe battery, which makes sense (in tablets / phones)

As said I'll make this configurable making 3V the default and set 3V3 in the defconfig where the original firmware does that.

Thanks!

ChenYu

#7 lore raw

Hi,

On 09-12-14 13:31, Chen-Yu Tsai wrote:

<snip>

Quoted 11 lines

Also I have a schematic file for A23 that someone provided me. I don't remember who...

It lists:
- ALDO2 connected to VCC-DLL @ 2.5V
- ALDO3 to VCC-PLL and AVCC (analog power) @ 3.0V
- DCDC1 to all the normal stuff @ 3V
- DCDC2 to VDD-SYS @ 1.1V
- DCDC3 to VDD-CPU (main cpu) @ 1.1V (1.26V in my fex file)
- DCDC5 to VCC-DRAM @ 1.1V (1.5V in my fex file)
- DC5LDO to VDD-CPUS (special cpu)

Thanks for this info!

I've prepared a sun8i / A23 set for u-boot which I'll send upstream as soon as I've verified that it does not break A31 support.

One thing which I've done is make u-boot explicitly disable power outputs which are unused, the specific target for this is DCDC4 which is enabled by default at power on and not used on sun8i.

As a side effect this also disabled ALDO1 which is enabled at power on too, and guess what, promptly my sdcard stopped working.

So it seems that ALDO1 is hookedup to the sdcard on the Ippo tablets, for now I've also enabled ALDO1 in the defconfig for V5, it would be great if you could give my current sunxi-wip branch a try on your V5 tablet, and if it works remove the setting of ALDO1 from the V5 defconfig and then try again. If it is like my V1.2 then it will stop booting when the SPL tries to load u-boot from the sdcard.

Regards,

Hans

#8 lore raw

Hi,

Quoted 31 lines · Hans de Goede

Hi,

On 09-12-14 13:31, Chen-Yu Tsai wrote:

<snip>

Quoted 11 lines

Also I have a schematic file for A23 that someone provided me. I don't remember who...

It lists:
- ALDO2 connected to VCC-DLL @ 2.5V
- ALDO3 to VCC-PLL and AVCC (analog power) @ 3.0V
- DCDC1 to all the normal stuff @ 3V
- DCDC2 to VDD-SYS @ 1.1V
- DCDC3 to VDD-CPU (main cpu) @ 1.1V (1.26V in my fex file)
- DCDC5 to VCC-DRAM @ 1.1V (1.5V in my fex file)
- DC5LDO to VDD-CPUS (special cpu)

Thanks for this info!

I've prepared a sun8i / A23 set for u-boot which I'll send upstream as soon as I've verified that it does not break A31 support.

One thing which I've done is make u-boot explicitly disable power outputs which are unused, the specific target for this is DCDC4 which is enabled by default at power on and not used on sun8i.

As a side effect this also disabled ALDO1 which is enabled at power on too, and guess what, promptly my sdcard stopped working.

Quoted 6 lines

So it seems that ALDO1 is hookedup to the sdcard on the Ippo tablets, for now I've also enabled ALDO1 in the defconfig for V5, it would be great if you could give my current sunxi-wip branch a try on your V5 tablet, and if it works remove the setting of ALDO1 from the V5 defconfig and then try again. If it is like my V1.2 then it will stop booting when the SPL tries to load u-boot from the sdcard.

I missed that one. ALDO1 is connected to VCC-IO (pin controller I think), VCC-PD (whatever that is, it isn't on the datasheet, but the schematic implies it's part of the LCD block), VCC-USB (USB block in the SoC), VCC-IO-WIFI (pull-ups and stuff for mmc wifi), and HPVCC (builtin headphone amp).

Others I failed to mention:

ELDO2 to DVDD1V8-CSI @ 1.8V
GPIO0 to AVDD-CSI @ 2.8V
DC1SW to VCC-LCD @ 3V (this is just a switchable output of DCDC1)

Hope this helps.

ChenYu

#9 lore raw

Hi,

Quoted 41 lines · Chen-Yu Tsai

Hi,

Quoted 31 lines · Hans de Goede

Hi,

On 09-12-14 13:31, Chen-Yu Tsai wrote:

<snip>

Quoted 11 lines

Also I have a schematic file for A23 that someone provided me. I don't remember who...

It lists:
- ALDO2 connected to VCC-DLL @ 2.5V
- ALDO3 to VCC-PLL and AVCC (analog power) @ 3.0V
- DCDC1 to all the normal stuff @ 3V
- DCDC2 to VDD-SYS @ 1.1V
- DCDC3 to VDD-CPU (main cpu) @ 1.1V (1.26V in my fex file)
- DCDC5 to VCC-DRAM @ 1.1V (1.5V in my fex file)
- DC5LDO to VDD-CPUS (special cpu)

Thanks for this info!

I've prepared a sun8i / A23 set for u-boot which I'll send upstream as soon as I've verified that it does not break A31 support.

One thing which I've done is make u-boot explicitly disable power outputs which are unused, the specific target for this is DCDC4 which is enabled by default at power on and not used on sun8i.

As a side effect this also disabled ALDO1 which is enabled at power on too, and guess what, promptly my sdcard stopped working.

Quoted 6 lines

So it seems that ALDO1 is hookedup to the sdcard on the Ippo tablets, for now I've also enabled ALDO1 in the defconfig for V5, it would be great if you could give my current sunxi-wip branch a try on your V5 tablet, and if it works remove the setting of ALDO1 from the V5 defconfig and then try again. If it is like my V1.2 then it will stop booting when the SPL tries to load u-boot from the sdcard.

Quoted 5 lines

I missed that one. ALDO1 is connected to VCC-IO (pin controller I think), VCC-PD (whatever that is, it isn't on the datasheet, but the schematic implies it's part of the LCD block), VCC-USB (USB block in the SoC), VCC-IO-WIFI (pull-ups and stuff for mmc wifi), and HPVCC (builtin headphone amp).

Thanks for the info. Weird thing is it is not listed in the fex, and android on my Ippo Q8H v1.2 seems to set it to 2.5 volt, which is somewhat lowish, it is below the minimum recommended voltage for Vusb (which is 2.7) and way below the minimum recommended voltage for HPVCC which is 3.0, according to the A31 datasheet (the A23 datasheet does not list it).

I've updated the defconfigs to set ALDO1 to 3.0 V and gave them a comment to explain to which voltages it is connected.

As for VCC-PD it is in the A23 datasheet, in the "pin description" chapter under GPIO-D, it powers the port D gpio pins.

Regards,

Hans

#10 lore raw

Hi,

I'm trying to boot from usb fel mode on A23. I have rebuilt your sunxi-wip branch, and try to run on my A23 (q88?) board, but it's failed in fel mode in executing.

I think the u-boot-spl.bin can be used as A20, this's the website I refer to:
http://docs.cubieboard.org/tutorials/cb2/development/booting_from_usb_in_fel_mode

my testing instructions are:

fel write address u-boot-spl.bin
fel exec address

the address for A20 is 0x2000, so that the DRAM will be initialized. but it's incorrect for A23 if the size reach 0x4000.

can you please let me know how to use fel and your u-boot-spl.bin?

Thanks!
Ricky

Quoted 71 lines · Hans de Goede

Hi,

Quoted 3 lines · Chen-Yu Tsai

Hi,

On Sat, Dec 13, 2014 at 9:58 PM, Hans de Goede <[email protected]

Quoted 28 lines · <javascript:>> wrote:
Quoted 28 lines

Hi,

On 09-12-14 13:31, Chen-Yu Tsai wrote:

<snip>

Quoted 11 lines

Also I have a schematic file for A23 that someone provided me. I don't remember who...

It lists:
- ALDO2 connected to VCC-DLL @ 2.5V
- ALDO3 to VCC-PLL and AVCC (analog power) @ 3.0V
- DCDC1 to all the normal stuff @ 3V
- DCDC2 to VDD-SYS @ 1.1V
- DCDC3 to VDD-CPU (main cpu) @ 1.1V (1.26V in my fex file)
- DCDC5 to VCC-DRAM @ 1.1V (1.5V in my fex file)
- DC5LDO to VDD-CPUS (special cpu)

Thanks for this info!

I've prepared a sun8i / A23 set for u-boot which I'll send upstream as soon as I've verified that it does not break A31 support.

One thing which I've done is make u-boot explicitly disable power outputs which are unused, the specific target for this is DCDC4 which is enabled by default at power on and not used on sun8i.

Quoted 5 lines
Quoted 5 lines

As a side effect this also disabled ALDO1 which is enabled at power on too, and guess what, promptly my sdcard stopped working.

So it seems that ALDO1 is hookedup to the sdcard on the Ippo tablets, for now I've also enabled ALDO1 in the defconfig for V5, it

would

Quoted 6 lines
Quoted 4 lines

be great if you could give my current sunxi-wip branch a try on your V5 tablet, and if it works remove the setting of ALDO1 from the V5 defconfig and then try again. If it is like my V1.2 then it will stop booting when the SPL tries to load u-boot from the sdcard.

I missed that one. ALDO1 is connected to VCC-IO (pin controller I

think),

Quoted 4 lines

VCC-PD (whatever that is, it isn't on the datasheet, but the schematic implies it's part of the LCD block), VCC-USB (USB block in the SoC), VCC-IO-WIFI (pull-ups and stuff for mmc wifi), and HPVCC (builtin headphone amp).

Thanks for the info. Weird thing is it is not listed in the fex, and android on my Ippo Q8H v1.2 seems to set it to 2.5 volt, which is somewhat lowish, it is below the minimum recommended voltage for Vusb (which is 2.7)
and way below the minimum recommended voltage for HPVCC which is 3.0, according to the A31 datasheet (the A23 datasheet does not list it).

I've updated the defconfigs to set ALDO1 to 3.0 V and gave them a comment to explain to which voltages it is connected.

As for VCC-PD it is in the A23 datasheet, in the "pin description" chapter under GPIO-D, it powers the port D gpio pins.

Regards,

Hans

#11 lore raw
Quoted 12 lines · On Sun, 14 Dec 2014 22:01:29 -0800 (PST) Ricky Xian <[email protected]> wrote:

Hi,

I'm trying to boot from usb fel mode on A23. I have rebuilt your sunxi-wip branch, and try to run on my A23 (q88?) board, but it's failed in fel mode in executing.

I think the u-boot-spl.bin can be used as A20, this's the website I refer to:
http://docs.cubieboard.org/tutorials/cb2/development/booting_from_usb_in_fel_mode

my testing instructions are:

Quoted 7 lines

fel write address u-boot-spl.bin
fel exec address

the address for A20 is 0x2000, so that the DRAM will be initialized. but it's incorrect for A23 if the size reach 0x4000.

can you please let me know how to use fel and your u-boot-spl.bin?

The size 0x4000 is just too large for the SPL binary. You can have a look at the SRAM memory map that is used in the FEL mode here:
https://github.com/hno/Allwinner-Info/blob/master/FEL-usb/USB-protocol.txt

Basically, there are two areas. One starts at 0x2000 and ends somewhere around 0x5D00 (may be slightly different for different SoC variants). The stack pointer is set right at the end of this area by the BROM code before passing control to SPL. So in practice it means that SPL code, data and stack must all fit there in just ~15K.

Another free area is around 0x8000-0xbfff, at least on A10/A13/A20.

The rest of SRAM is used by the BROM code for implementing FEL USB protocol and we should not touch it.

In order to provide more room for SPL in the FEL mode, the following things can be done:
1) Compile the FEL SPL in Thumb2 mode instead of ARM mode, that's going to save around 30% of code size (depending on the compiler version).
2) Take the extra 0x8000-0xbfff area into use.

I had a patch for this earlier (it is currently self-NAKED):

    http://lists.denx.de/pipermail/u-boot/2014-July/183985.html

Switching to Thumb2 should be perfectly safe. But before using the 0x8000-0xbfff area, we need to confirm that it is really free to use on every Allwinner SoC variant and does not clash with the data structures used by the FEL BROM code. The FEL SRAM memory map details can be probably clarified by the Allwinner representatives.

I was kind of reluctant to settle with the Thumb2 part alone until the sunxi custodians acknowledge that the SPL code size is actually a real practical problem. For example, this is important for making other decisions and having some progress with

    http://lists.denx.de/pipermail/u-boot/2014-August/185722.html    

To sum it up. Just try to reduce the SPL size. Maybe try a different gcc version to compile it. Maybe try to apply the Thumb2 tweak.

Signature · 2 lines
Best regards,
Siarhei Siamashka
#12 lore raw

Thanks Siarhei ~!

I think it's what Hans's working on, I want to know the debugging instructions of this branch so that I can try with my board.

the uart is combined with sd card, so I can't boot from sd card.

Quoted 72 lines · Siarhei Siamashka
Quoted 3 lines · On Sun, 14 Dec 2014 22:01:29 -0800 (PST) Ricky Xian <[email protected]…

Hi,

I'm trying to boot from usb fel mode on A23. I have rebuilt your

sunxi-wip

Quoted 1 line

branch, and try to run on my A23 (q88?) board, but it's failed in fel

mode

Quoted 3 lines

in executing.

I think the u-boot-spl.bin can be used as A20, this's the website I

refer

Quoted 2 lines

to:

http://docs.cubieboard.org/tutorials/cb2/development/booting_from_usb_in_fel_mode

Quoted 10 lines

my testing instructions are:

fel write address u-boot-spl.bin
fel exec address

the address for A20 is 0x2000, so that the DRAM will be initialized. but it's incorrect for A23 if the size reach 0x4000.

can you please let me know how to use fel and your u-boot-spl.bin?

The size 0x4000 is just too large for the SPL binary. You can have a look at the SRAM memory map that is used in the FEL mode here:

https://github.com/hno/Allwinner-Info/blob/master/FEL-usb/USB-protocol.txt

Basically, there are two areas. One starts at 0x2000 and ends somewhere around 0x5D00 (may be slightly different for different SoC variants). The stack pointer is set right at the end of this area by the BROM code before passing control to SPL. So in practice it means that SPL code, data and stack must all fit there in just ~15K.

Another free area is around 0x8000-0xbfff, at least on A10/A13/A20.

The rest of SRAM is used by the BROM code for implementing FEL USB protocol and we should not touch it.

In order to provide more room for SPL in the FEL mode, the following things can be done:
1) Compile the FEL SPL in Thumb2 mode instead of ARM mode, that's going to save around 30% of code size (depending on the compiler version).
2) Take the extra 0x8000-0xbfff area into use.

I had a patch for this earlier (it is currently self-NAKED):

    http://lists.denx.de/pipermail/u-boot/2014-July/183985.html 

Switching to Thumb2 should be perfectly safe. But before using the 0x8000-0xbfff area, we need to confirm that it is really free to use on every Allwinner SoC variant and does not clash with the data structures used by the FEL BROM code. The FEL SRAM memory map details can be probably clarified by the Allwinner representatives.

I was kind of reluctant to settle with the Thumb2 part alone until the sunxi custodians acknowledge that the SPL code size is actually a real practical problem. For example, this is important for making other decisions and having some progress with

    http://lists.denx.de/pipermail/u-boot/2014-August/185722.html     

To sum it up. Just try to reduce the SPL size. Maybe try a different gcc version to compile it. Maybe try to apply the Thumb2 tweak.

Signature · 2 lines
Best regards, 
Siarhei Siamashka 
#13 lore raw
In reply to Ricky Xian

Hi,

Quoted 12 lines · Ricky Xian

Hi,

I'm trying to boot from usb fel mode on A23. I have rebuilt your sunxi-wip branch, and try to run on my A23 (q88?) board, but it's failed in fel mode in executing.

I think the u-boot-spl.bin can be used as A20, this's the website I refer to:
http://docs.cubieboard.org/tutorials/cb2/development/booting_from_usb_in_fel_mode

my testing instructions are:

Quoted 7 lines

fel write address u-boot-spl.bin
fel exec address

the address for A20 is 0x2000, so that the DRAM will be initialized. but it's incorrect for A23 if the size reach 0x4000.

can you please let me know how to use fel and your u-boot-spl.bin?

I've just tested this, and it works fine for me using the following commands:

make -j4 CROSS_COMPILE=arm-linux-gnu- Ippo_q8h_v1_2_felconfig

Note the _felconfig rather then _defconfig at the end !

And then:

make -j4 CROSS_COMPILE=arm-linux-gnu-

sudo ~hans/bin/fel write 0x2000 spl/u-boot-spl.bin sudo ~hans/bin/fel exe 0x2000
sleep 0.5
sudo ~hans/bin/fel write 0x4a000000 u-boot.bin sudo ~hans/bin/fel exe 0x4a000000

Regards,

Hans

#14 lore raw
In reply to Hans de Goede
Quoted 11 lines · Hans de Goede

Hi,

This is still a bit rough around the edges, I'll clean it up as time permits and then post it upstream.

In the mean time people interested can find $subject here:
https://github.com/jwrdegoede/u-boot-sunxi/commits/sunxi-wip

ChenYu, this also has a mmc fix which you may find interesting, it may explain some of the problems with mmc you've been having on both the A80 board, as well as the A31 dev board you've.

Wow, very cool, including PMIC support et al.

Thanks a lot :)

Maxime

Signature · 10 lines
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20141208/1c713dbc/attachment.pgp>