Skip to contentSkip to conversation

nios2 and openrisc in barebox

DISCUSSION Sascha Hauer 7 participants 13 messages Feb 23, 2021 – Mar 1, 2021

lore
Share & tools
lore
#0 lore raw

Hi Franck, Antony

Are you still interested in barebox support for NIOS2 and openrisc? If not I would consider removing it. Background is that we are planning to introduce green threads and implement pollers with them which makes green threads rather mandatory. Green threads need a setjmp/longjmp implementation that would have to be added for these architectures.

Keeping the architectures is no big deal normally, but I don't want to add untested new functionality, also I don't want to implement things that nobody uses, so if you are still interested it would be great if you volunteer to implement setjmp/longjmp.

Thanks
Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#1 lore raw
Quoted 9 lines · Sascha Hauer

Hi Franck, Antony

Are you still interested in barebox support for NIOS2 and openrisc? If not I would consider removing it. Background is that we are planning to introduce green threads and implement pollers with them which makes green threads rather mandatory. Green threads need a setjmp/longjmp implementation that would have to be added for these architectures.

Quoted 8 lines

Keeping the architectures is no big deal normally, but I don't want to add untested new functionality, also I don't want to implement things that nobody uses, so if you are still interested it would be great if you volunteer to implement setjmp/longjmp.

Thanks
Sascha

Hi,

Thanks for asking.

I don't think anyone is using (has ever used ?) Barebox on these architectures.

I do have a product which uses Barebox and a NIOS2 in the field. It's working for 10+ years already but will never be updated.

I think you can drop support for them. I won't work on this myself. Let's wait for Antony's view on this.

Franck.

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#2 lore raw
Quoted 23 lines · On Tue, 23 Feb 2021 14:13:44 +0100 Franck Jullien <[email protected]>…
Quoted 9 lines · Sascha Hauer

Hi Franck, Antony

Are you still interested in barebox support for NIOS2 and openrisc? If not I would consider removing it. Background is that we are planning to introduce green threads and implement pollers with them which makes green threads rather mandatory. Green threads need a setjmp/longjmp implementation that would have to be added for these architectures.

Quoted 8 lines

Keeping the architectures is no big deal normally, but I don't want to add untested new functionality, also I don't want to implement things that nobody uses, so if you are still interested it would be great if you volunteer to implement setjmp/longjmp.

Thanks
Sascha

Hi,

Thanks for asking.

Quoted 7 lines

I don't think anyone is using (has ever used ?) Barebox on these architectures.

I do have a product which uses Barebox and a NIOS2 in the field. It's working for 10+ years already but will never be updated.

I think you can drop support for them. I won't work on this myself. Let's wait for Antony's view on this.

Hi!

I experimented with openrisc before widespread use of RISC-V. I have never used nios2.

IMHO we can drop openrisc and nios2.

Signature · 7 lines
Best regards,
  Antony Pavlov

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#3 lore raw
Quoted 34 lines · Antony Pavlov
Quoted 23 lines · On Tue, 23 Feb 2021 14:13:44 +0100 Franck Jullien <[email protected]>…
Quoted 9 lines · Sascha Hauer

Hi Franck, Antony

Are you still interested in barebox support for NIOS2 and openrisc? If not I would consider removing it. Background is that we are planning to introduce green threads and implement pollers with them which makes green threads rather mandatory. Green threads need a setjmp/longjmp implementation that would have to be added for these architectures.

Quoted 8 lines

Keeping the architectures is no big deal normally, but I don't want to add untested new functionality, also I don't want to implement things that nobody uses, so if you are still interested it would be great if you volunteer to implement setjmp/longjmp.

Thanks
Sascha

Hi,

Thanks for asking.

Quoted 7 lines

I don't think anyone is using (has ever used ?) Barebox on these architectures.

I do have a product which uses Barebox and a NIOS2 in the field. It's working for 10+ years already but will never be updated.

I think you can drop support for them. I won't work on this myself. Let's wait for Antony's view on this.

Quoted 6 lines

Hi!

I experimented with openrisc before widespread use of RISC-V. I have never used nios2.

IMHO we can drop openrisc and nios2.

Ok, thanks for your input. Let's go for it.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#4 lore raw
In reply to Sascha Hauer

Hello,

I am adding Stafford Horne in CC, as I think he might want to help about openrisc port :)

Cheers

Yann

Quoted 8 lines · Sascha Hauer

Hi Franck, Antony

Are you still interested in barebox support for NIOS2 and openrisc? If not I would consider removing it. Background is that we are planning to introduce green threads and implement pollers with them which makes green threads rather mandatory. Green threads need a setjmp/longjmp implementation that would have to be added for these architectures.

Quoted 8 lines

Keeping the architectures is no big deal normally, but I don't want to add untested new functionality, also I don't want to implement things that nobody uses, so if you are still interested it would be great if you volunteer to implement setjmp/longjmp.

Thanks
Sascha

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#5 lore raw

Hello,

Quoted 22 lines · Yann Sionneau

Hello,

I am adding Stafford Horne in CC, as I think he might want to help about openrisc port :)

Cheers

Yann

Quoted 12 lines · Sascha Hauer

Hi Franck, Antony

Are you still interested in barebox support for NIOS2 and openrisc? If not I would consider removing it. Background is that we are planning to introduce green threads and implement pollers with them which makes green threads rather mandatory. Green threads need a setjmp/longjmp implementation that would have to be added for these architectures.

Keeping the architectures is no big deal normally, but I don't want to add untested new functionality, also I don't want to implement things that nobody uses, so if you are still interested it would be great if you volunteer to implement setjmp/longjmp.

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

That said, we have a setjmp/longjmp implementation in our glibc port which I have been testing for upstreaming lately.

https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/setjmp.S
https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/__longjmp.S

I had a look at the barebox docs and openrisc port, it seems like there is support for our standard platform, or1ksim, which is the same as my de0-nano fpga board/soc and qemu. I should be able to make use of this.

Is there a branch where you are doing the green thread work? I could add and test the OpenRISC implementation.

-Stafford

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#6 lore raw
Quoted 12 lines · Stafford Horne

Hello,

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

That said, we have a setjmp/longjmp implementation in our glibc port which I have been testing for upstreaming lately.

https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/setjmp.S
https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/__longjmp.S

Quoted 6 lines

I had a look at the barebox docs and openrisc port, it seems like there is support for our standard platform, or1ksim, which is the same as my de0-nano fpga board/soc and qemu. I should be able to make use of this.

Is there a branch where you are doing the green thread work? I could add and test the OpenRISC implementation.

Hi,

I've done this yesterday, I've push a branch[1] that has the green thread paches applied (except patches on usb mass storage gadget).
https://github.com/jmaselbas/barebox/tree/test_setjmp

To test greenthread your can run the poller -c command in barebox's prompt. To do you you'll need to enable CONFIG_POLLER_YIELD and define CONFIG_HAS_ARCH_SJLJ

Best,
Jules

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#7 lore raw

Hello,

Quoted 17 lines · Jules Maselbas
Quoted 16 lines · Stafford Horne

Hello,

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

That said, we have a setjmp/longjmp implementation in our glibc port which I have been testing for upstreaming lately.

https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/setjmp.S
https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/__longjmp.S

I had a look at the barebox docs and openrisc port, it seems like there is support for our standard platform, or1ksim, which is the same as my de0-nano fpga board/soc and qemu. I should be able to make use of this.

Quoted 7 lines
Quoted 2 lines

Is there a branch where you are doing the green thread work? I could add and test the OpenRISC implementation.

Hi,

I've done this yesterday, I've push a branch[1] that has the green thread paches applied (except patches on usb mass storage gadget).
https://github.com/jmaselbas/barebox/tree/test_setjmp

Great. Thanks. I'm preparing an updated version that will clean it up a bit, but the heart part of architectures providing setjmp/longjmp/initjmp will stay the same.

Small request:
If you implement initjmp, could you have it save into jmp_buf a frame pointer value of NULL if applicable? That way, should backtrace be implemented (or done via JTAG), they have a defined end.

Cheers,
Ahmad

Quoted 14 lines

To test greenthread your can run the poller -c command in barebox's prompt. To do you you'll need to enable CONFIG_POLLER_YIELD and define CONFIG_HAS_ARCH_SJLJ

Best,
Jules

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#8 lore raw
Quoted 10 lines · On Thu, 25 Feb 2021 16:33:06 +0100 Ahmad Fatoum <[email protected]> wrote:

Hello,

Quoted 7 lines · Jules Maselbas
Quoted 6 lines · Stafford Horne

Hello,

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

There is barebox for litex SoC with RISC-V CPU core, please see:

https://github.com/enjoy-digital/litex/issues/413#issuecomment-609047340

Quoted 55 lines
Quoted 10 lines
Quoted 10 lines

That said, we have a setjmp/longjmp implementation in our glibc port which I have been testing for upstreaming lately.

https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/setjmp.S
https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/__longjmp.S

I had a look at the barebox docs and openrisc port, it seems like there is support for our standard platform, or1ksim, which is the same as my de0-nano fpga board/soc and qemu. I should be able to make use of this.

Quoted 7 lines
Quoted 2 lines

Is there a branch where you are doing the green thread work? I could add and test the OpenRISC implementation.

Hi,

I've done this yesterday, I've push a branch[1] that has the green thread paches applied (except patches on usb mass storage gadget).
https://github.com/jmaselbas/barebox/tree/test_setjmp

Great. Thanks. I'm preparing an updated version that will clean it up a bit, but the heart part of architectures providing setjmp/longjmp/initjmp will stay the same.

Small request:
If you implement initjmp, could you have it save into jmp_buf a frame pointer value of NULL if applicable? That way, should backtrace be implemented (or done via JTAG), they have a defined end.

Cheers,
Ahmad

Quoted 14 lines

To test greenthread your can run the poller -c command in barebox's prompt. To do you you'll need to enable CONFIG_POLLER_YIELD and define CONFIG_HAS_ARCH_SJLJ

Best,
Jules

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
Signature · 7 lines
Best regards,
  Antony Pavlov

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#9 lore raw
Quoted 9 lines · Antony Pavlov
Quoted 6 lines · On Thu, 25 Feb 2021 16:33:06 +0100 Ahmad Fatoum <[email protected]> wrote:

Hello,

Quoted 3 lines · Jules Maselbas
Quoted 2 lines · Stafford Horne

Hello,

Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

There is barebox for litex SoC with RISC-V CPU core, please see:

https://github.com/enjoy-digital/litex/issues/413#issuecomment-609047340

Hi Antony,

Great thanks, I see that is a two stage bootloader Litex BIOS -> Barebox. That's great and gives me an easier target.

-Stafford

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#10 lore raw
In reply to Ahmad Fatoum

Hello,

Quoted 26 lines · Ahmad Fatoum
Quoted 17 lines · Jules Maselbas
Quoted 16 lines · Stafford Horne

Hello,

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

That said, we have a setjmp/longjmp implementation in our glibc port which I have been testing for upstreaming lately.

https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/setjmp.S
https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/__longjmp.S

I had a look at the barebox docs and openrisc port, it seems like there is support for our standard platform, or1ksim, which is the same as my de0-nano fpga board/soc and qemu. I should be able to make use of this.

Quoted 7 lines
Quoted 2 lines

Is there a branch where you are doing the green thread work? I could add and test the OpenRISC implementation.

Hi,

I've done this yesterday, I've push a branch[1] that has the green thread paches applied (except patches on usb mass storage gadget).
https://github.com/jmaselbas/barebox/tree/test_setjmp

Quoted 8 lines

Great. Thanks. I'm preparing an updated version that will clean it up a bit, but the heart part of architectures providing setjmp/longjmp/initjmp will stay the same.

Small request:
If you implement initjmp, could you have it save into jmp_buf a frame pointer value of NULL if applicable? That way, should backtrace be implemented (or done via JTAG), they have a defined end.

Dismiss this last request. C code will just zero-initialize jmp_buf, so initjmp needs only fill out return address and stack pointer. I just sent out a series, which can be used to develop this one.

@Jules, any chance you could take over the kvx port? :-)

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#11 lore raw

Hi Ahmad,

Quoted 7 lines · Ahmad Fatoum

Hello,

Quoted 4 lines

Small request:
If you implement initjmp, could you have it save into jmp_buf a frame pointer value of NULL if applicable? That way, should backtrace be implemented (or done via JTAG), they have a defined end.

Quoted 5 lines

Dismiss this last request. C code will just zero-initialize jmp_buf, so initjmp needs only fill out return address and stack pointer. I just sent out a series, which can be used to develop this one.

@Jules, any chance you could take over the kvx port? :-)

Yes, I've got a patch almost ready to be sent.

Jules

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#12 lore raw
In reply to Ahmad Fatoum
Quoted 34 lines · Ahmad Fatoum

Hello,

Quoted 30 lines · Ahmad Fatoum
Quoted 17 lines · Jules Maselbas
Quoted 16 lines · Stafford Horne

Hello,

I am the maintainer of most OpenRISC things right now. I would be happy to help, but I don't really use barebox right now. Most of my deployments I use with litex FPGA SoC's that has its own bootloader.

That said, we have a setjmp/longjmp implementation in our glibc port which I have been testing for upstreaming lately.

https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/setjmp.S
https://github.com/stffrdhrn/or1k-glibc/blob/or1k-port-2/sysdeps/or1k/__longjmp.S

I had a look at the barebox docs and openrisc port, it seems like there is support for our standard platform, or1ksim, which is the same as my de0-nano fpga board/soc and qemu. I should be able to make use of this.

Quoted 7 lines
Quoted 2 lines

Is there a branch where you are doing the green thread work? I could add and test the OpenRISC implementation.

Hi,

I've done this yesterday, I've push a branch[1] that has the green thread paches applied (except patches on usb mass storage gadget).
https://github.com/jmaselbas/barebox/tree/test_setjmp

Great. Thanks. I'm preparing an updated version that will clean it up a bit, but the heart part of architectures providing setjmp/longjmp/initjmp will stay the same.

Quoted 8 lines
Quoted 4 lines

Small request:
If you implement initjmp, could you have it save into jmp_buf a frame pointer value of NULL if applicable? That way, should backtrace be implemented (or done via JTAG), they have a defined end.

Dismiss this last request. C code will just zero-initialize jmp_buf, so initjmp needs only fill out return address and stack pointer. I just sent out a series, which can be used to develop this one.

Thanks,

I will start looking at this now, I'll let you know when I get progress on openrisc.

Quoted 7 lines

@Jules, any chance you could take over the kvx port? :-)

Signature · 4 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox