[BUG] NVMe: Discard timeout/hang on a801 during mkfs

[Sune Brian]

On Sun, Mar 1, 2026 at 10:25 PM Sune Brian <briansune at gmail.com> wrote:
>
> On Sun, Mar 1, 2026 at 9:36 PM Maurizio Lombardi
> <mlombard at bsdbackstore.eu> wrote:
> >
> > On Sat Feb 28, 2026 at 1:22 PM CET, Sune Brian wrote:
> > > Issue: A fresh NVMe is format below instructions
> > > always result in series of "opcode # (Write) QID 1 timeout"
> > > While normal mount and read write shows no stall
> > > or hang.
> > >
> > > dmesg log is placed at the end.
> > >
> > > During testing it shows that DRAM-less NVMe will result in
> > > timeout but DRAM NVMe does not.
> > >
> > > Please maintainers help to see if there are any possible
> > > causes that could remove these possible PCIe reset.
> > >
> > > Test Kernel: 6.12.33-g3234b1ed8956-dirty
> > > CPU: ARM7l
> >
> > I am not sure from where this kernel comes from but a search on google
> > hints to the Altera FPGA github tree
> > Is this the Altera's kernel?
>
> Hi Maurizio,
>
> Absolutely correct kernels are clones from the Altera trunk.
> I am hoping there are patches missing and let me give it a try.
>
> >
> > I looked at their 6.12 LTS branch

Hi Maurizio,

I had done a quick diff:
https://github.com/altera-fpga/linux-socfpga/blob/socfpga-6.19/drivers/nvme/host/pci.c
They had that missing return in the more latest kernel.

Had done a quick wipe and format test works no more timeout very repeatable.

The interesting part is DRAM NVMe does not have this issue, only DRAM-less.

Issue solved~

Thanks a lot,
Brian

> > https://github.com/altera-fpga/linux-socfpga/blob/socfpga-6.12.43-lts/drivers/nvme/host/pci.c
> >
> > And they miss a patch that fixes some problems with I/O timeouts in
> > the nvme-pci driver.
>
> >
> > https://lore.kernel.org/linux-nvme/CABXGCsMd_xv8jPDF_sFYhwd8GtANZ23nbaSJuCxQRO7cjPtgWg@mail.gmail.com/
> >
> > https://lore.kernel.org/linux-nvme/20250331162859.1407725-1-mlombard@redhat.com/
>
> Thanks,
> Brian
>
> >
> > Maurizio
> >
> > >
> > > 1) sudo wipefs -a /dev/nvme0n1
> > >
> > > /dev/nvme0n1: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49
> > > 20 50 41 52 54
> > > /dev/nvme0n1: 8 bytes were erased at offset 0x1dcf855e00 (gpt): 45 46
> > > 49 20 50 41 52 54
> > > /dev/nvme0n1: 2 bytes were erased at offset 0x000001fe (PMBR): 55 a
> > >
> > > 2) sudo fdisk /dev/nvme0n1
> > > g
> > > n
> > >
> > > w
> > >
> > > 3) sudo mkfs.ext4 /dev/nvme0n1p1
> > >
> > > mke2fs 1.47.2 (1-Jan-2025)
> > > Discarding device blocks: done
> > > Creating filesystem with 31258368 4k blocks and 7815168 inodes
> > > Filesystem UUID: 8f469cc5-d6e1-4606-bfda-869806245d9f
> > > Superblock backups stored on blocks:
> > >         32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
> > >         4096000, 7962624, 11239424, 20480000, 23887872
> > >
> > > Allocating group tables: done
> > > Writing inode tables: done
> > > Creating journal (131072 blocks):
> > > done
> > > Writing superblocks and filesystem accounting information: done
> > >
> > > Thanks,
> > > Brian