From: Yao Zi <me@ziyao.cc>
To: E Shattow <e@freeshell.de>, Martin Husemann <martin@NetBSD.org>,
u-boot@lists.u-boot-project.org
Cc: Yao Zi <me@ziyao.cc>
Subject: Re: StarFive VisionFive 2: NVME not responding randomly
Date: Wed, 18 Nov 2026 15:45:20 +0000 [thread overview]
Message-ID: <av3IEIY_T08Dv1xY@pie> (raw)
In-Reply-To: <5e9eeca5-6772-4aae-b2a7-d49b1676f6d5@freeshell.de>
On Tue, Nov 17, 2026 at 11:55:36PM -0800, E Shattow wrote:
> Hi Martin,
>
> On 11/17/26 05:29, Martin Husemann wrote:
> > Hi folks,
> >
> > I boot my StarFive VisionFive 2 from NVME usually, but "sometimes"
> > (which is actually qite often) u-boot does not recognize the NVME
> > and can't read the partition.
> >
> > Full boot lo of a working boot below. When it doesn't work it
> > says:
> >
> > [..]
> > USB XHCI 1.00
> > Bus xhci_pci: 2 USB Device(s) found
> > scanning usb for storage devices... 0 Storage Device(s) found
> > Working FDT set to ff6f4430
> > Hit any key to stop autoboot: 0
> > Card did not respond to voltage select! : -110
> > Card did not respond to voltage select! : -110
> > <------- all normal untill here, now the difference starts:
> > ** Invalid partition 1 **
> > Couldn't find partition nvme 0:1
> > ** Invalid partition 1 **
> > Couldn't find partition nvme 0:1
> > Cannot persist EFI variables without system partition
> > ** Invalid partition 1 **
> > Couldn't find partition nvme 0:1
> > ** Invalid partition 1 **
> > Couldn't find partition nvme 0:1
> > ** Booting bootflow '<NULL>' with efi_mgr
> > Loading Boot0000 'nvme 0' failed
> > EFI boot manager: Cannot load any image
> > Boot failed (err=-14)
> >
> >
> > I tried everything obvious: software reset, reset button, removing power for
> > a minute - nothing made a difference. Sometime later it just works again.
> >
> > I can build a debug version, but would like to hear some hints what
> > debug stuff I should enable for that, or any hints where to look.
> >
> > Thanks,
> >
> > Martin
...
> My understanding of PCIe on JH-7110 SoC is there's a hardware design
> flaw and the "fix" is a delay-and-wait that was rejected by Linux
> upstream. This presents more readily with *some* NVMe drives. It was a
> hot topic for discussion and based on that I selected Samsung 970
> EVOPlus 500GB and 250GB models and I've never had an issue appear with
> the 250GB on which I do all my RISC-V development as with a Home
> Assistant Yellow baseboard hosting the Milk-V Mars CM Lite (JH-7110 SoC)
> system-on-module.
>
> Is this the issue you're experiencing? What exact make/model is that
> NVMe module?
I don't think these are likely the same. The issue E referring to is
probably mentioned in this patch[1], which says,
> As the Starfive JH7110 hardware can't keep two inbound post write in
> order all the time, such as MSI messages and NVMe completions. If the
> NVMe completion update later than the MSI, an NVMe IRQ handle will miss.
and its follow-up[2] adds,
> Not any two of inbound post requests can`t be kept in order in JH7110 SoC,
> the only one case we found is NVMe completions with MSI interrupts.
> To be more precise, they are the pending status in nvme_completion struct and
> nvme_irq handler in nvme/host/pci.c.
So it's the order between writes for MSI interrupts, and writes for
MSI completions, that could be broken.
However, IIRC, U-Boot's NVMe driver always polls for updates to the NVMe
command ring, which is natural since we don't have any interrupting
facilities at all, so it's impossible for us to suffer from any kinds of
ordering issues related to interrupts.
> -E
--
Thanks,
Yao Zi
Meow-mrrp-mrow, I'm a cat.
[1]: https://lore.kernel.org/linux-pci/20240227103522.80915-23-minda.chen@starfivetech.com/
[2]: https://lore.kernel.org/linux-pci/ZQ0PR01MB0981BC562E837B232B419AC28229A@ZQ0PR01MB0981.CHNPR01.prod.partner.outlook.cn/
next prev parent reply other threads:[~2026-11-18 15:46 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-17 13:29 StarFive VisionFive 2: NVME not responding randomly Martin Husemann
2026-11-18 7:55 ` E Shattow
2026-11-18 8:38 ` Martin Husemann
2026-11-18 15:45 ` Yao Zi [this message]
2026-11-20 5:19 ` Yao Zi
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=av3IEIY_T08Dv1xY@pie \
--to=me@ziyao.cc \
--cc=e@freeshell.de \
--cc=martin@NetBSD.org \
--cc=u-boot@lists.u-boot-project.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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.