All of lore.kernel.org
 help / color / mirror / Atom feed
* RiscV SBI incompatibilities?
@ 2026-11-11 14:35 Martin Husemann
  2026-11-11 15:49 ` Yao Zi
  0 siblings, 1 reply; 5+ messages in thread
From: Martin Husemann @ 2026-11-11 14:35 UTC (permalink / raw)
  To: u-boot

Hi folks,

maybe I tried something stupid, but it felt like a good idea at first.

I am updating my VisionFive 2 board from an oldish installation with
U-Boot 2024.10.

To have as few moving parts as possible, I build 2026.10 with the same SBI
that I had lying around from the original 2024.10 build, which happend
to be OpenSBI 1.4.

This failed spectacularily: most sbc_* calls for HSM fail with 
-3 (SBI_ERR_INVALID_PARAM).

A few calls worked, like sbi_hart_start() and sbi_hart_get_status(),
which told me the hartids I was using were valid, but then doing
e.g. sbi_send_ipi() or sbi_remote_sfence_vma() failed.

Does this ring a bell? Is it expected? Did I overlook a "minimal SBI version
1.X" somewhere?

Martin

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: RiscV SBI incompatibilities?
  2026-11-11 14:35 RiscV SBI incompatibilities? Martin Husemann
@ 2026-11-11 15:49 ` Yao Zi
  2026-11-12 10:08   ` Martin Husemann
  2026-11-12 10:59   ` Yao Zi
  0 siblings, 2 replies; 5+ messages in thread
From: Yao Zi @ 2026-11-11 15:49 UTC (permalink / raw)
  To: Martin Husemann, u-boot; +Cc: Yao Zi

On Wed, Nov 11, 2026 at 03:35:48PM +0100, Martin Husemann wrote:
> Hi folks,
> 
> maybe I tried something stupid, but it felt like a good idea at first.
> 
> I am updating my VisionFive 2 board from an oldish installation with
> U-Boot 2024.10.
> 
> To have as few moving parts as possible, I build 2026.10 with the same SBI
> that I had lying around from the original 2024.10 build, which happend
> to be OpenSBI 1.4.
> 
> This failed spectacularily: most sbc_* calls for HSM fail with 
> -3 (SBI_ERR_INVALID_PARAM).
> 
> A few calls worked, like sbi_hart_start() and sbi_hart_get_status(),
> which told me the hartids I was using were valid, but then doing
> e.g. sbi_send_ipi() or sbi_remote_sfence_vma() failed.
> 
> Does this ring a bell? Is it expected? Did I overlook a "minimal SBI version
> 1.X" somewhere?

doc/board/starfive/jh7110_common.rst says,

> 3.  U-Boot for JH-7110 requires OpenSBI v1.5+ generic platform object
>     fw_dynamic.bin to be included in the Flattened Image Tree blob. OpenSBI may
>     first be built as below:

I had a brief look through the changes from v1.4 to v1.5, and suspect it
has something to do with OpenSBI commit c5be0e1ed1fa ("lib: utils/ipi:
Skip initialize ipi when dt is not enabled")[1], but don't confirm it :)

> Martin

-- 
Thanks,
Yao Zi

Meow-mrrp-mrow, I'm a cat.

[1]: https://github.com/riscv-software-src/opensbi/commit/c5be0e1ed1fa17b9f9abb49e80db014e067523e4

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: RiscV SBI incompatibilities?
  2026-11-11 15:49 ` Yao Zi
@ 2026-11-12 10:08   ` Martin Husemann
  2026-11-12 10:59   ` Yao Zi
  1 sibling, 0 replies; 5+ messages in thread
From: Martin Husemann @ 2026-11-12 10:08 UTC (permalink / raw)
  To: Yao Zi; +Cc: u-boot

On Wed, Nov 11, 2026 at 03:49:52PM +0000, Yao Zi wrote:
> I had a brief look through the changes from v1.4 to v1.5, and suspect it
> has something to do with OpenSBI commit c5be0e1ed1fa ("lib: utils/ipi:
> Skip initialize ipi when dt is not enabled")[1], but don't confirm it :)

Thanks, indeed with OpenSBI 1.9 it works a lot better.

Martin

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: RiscV SBI incompatibilities?
  2026-11-11 15:49 ` Yao Zi
  2026-11-12 10:08   ` Martin Husemann
@ 2026-11-12 10:59   ` Yao Zi
  2026-11-12 11:09     ` Martin Husemann
  1 sibling, 1 reply; 5+ messages in thread
From: Yao Zi @ 2026-11-12 10:59 UTC (permalink / raw)
  To: Martin Husemann, u-boot; +Cc: Yao Zi

On Wed, Nov 11, 2026 at 03:49:52PM +0000, Yao Zi wrote:
> On Wed, Nov 11, 2026 at 03:35:48PM +0100, Martin Husemann wrote:
> > Hi folks,
> > 
> > maybe I tried something stupid, but it felt like a good idea at first.
> > 
> > I am updating my VisionFive 2 board from an oldish installation with
> > U-Boot 2024.10.
> > 
> > To have as few moving parts as possible, I build 2026.10 with the same SBI
> > that I had lying around from the original 2024.10 build, which happend
> > to be OpenSBI 1.4.
> > 
> > This failed spectacularily: most sbc_* calls for HSM fail with 
> > -3 (SBI_ERR_INVALID_PARAM).
> > 
> > A few calls worked, like sbi_hart_start() and sbi_hart_get_status(),
> > which told me the hartids I was using were valid, but then doing
> > e.g. sbi_send_ipi() or sbi_remote_sfence_vma() failed.
> > 
> > Does this ring a bell? Is it expected? Did I overlook a "minimal SBI version
> > 1.X" somewhere?
> 
> doc/board/starfive/jh7110_common.rst says,
> 
> > 3.  U-Boot for JH-7110 requires OpenSBI v1.5+ generic platform object
> >     fw_dynamic.bin to be included in the Flattened Image Tree blob. OpenSBI may
> >     first be built as below:
> 
> I had a brief look through the changes from v1.4 to v1.5, and suspect it
> has something to do with OpenSBI commit c5be0e1ed1fa ("lib: utils/ipi:
> Skip initialize ipi when dt is not enabled")[1], but don't confirm it :)

Totally unnecessary, just out of curiosity, could you please send a list
of failed sbi calls and their corresponding error codes? From this I
could probably find out whether it's related to the OpenSBI commit :)

Feel free to ignore this message if you don't want to :)

> > Martin
> 
> -- 
> Thanks,
> Yao Zi
> 
> Meow-mrrp-mrow, I'm a cat.
> 
> [1]: https://github.com/riscv-software-src/opensbi/commit/c5be0e1ed1fa17b9f9abb49e80db014e067523e4

-- 
Thanks,
Yao Zi

Meow-mrrp-mrow, I'm a cat.

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: RiscV SBI incompatibilities?
  2026-11-12 10:59   ` Yao Zi
@ 2026-11-12 11:09     ` Martin Husemann
  0 siblings, 0 replies; 5+ messages in thread
From: Martin Husemann @ 2026-11-12 11:09 UTC (permalink / raw)
  To: Yao Zi; +Cc: u-boot

On Thu, Nov 12, 2026 at 10:59:50AM +0000, Yao Zi wrote:
> Totally unnecessary, just out of curiosity, could you please send a list
> of failed sbi calls and their corresponding error codes? From this I
> could probably find out whether it's related to the OpenSBI commit :)

I didn't create a full list, but verified two that did fail:

sbi_send_ipi()
sbi_remote_sfence_vma()


Unrelated:
I still have a strange problem (identical in 2024.10 and 2026.10) that
sometimes the NVMe is not readable after reboot; I initially thought it
would be related to a USB storage device being present, but it seems to
be mostly random. Maybe the USB device enhances the chance for failure,
wich might point at a power supply problem - I should replace the cable.

Martin

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2026-11-12 11:09 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-11 14:35 RiscV SBI incompatibilities? Martin Husemann
2026-11-11 15:49 ` Yao Zi
2026-11-12 10:08   ` Martin Husemann
2026-11-12 10:59   ` Yao Zi
2026-11-12 11:09     ` Martin Husemann

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.