| Message ID | 20260522045949.184811-1-aleksandar.nikolic22@pm.me |
|---|---|
| State | New |
| Headers | show |
| Series | [scarthgap,v2] wic: Fix updating fstab for nvme0n1 | expand |
On Fri, 2026-05-22 at 06:59 +0200, Aleksandar Nikolic via lists.openembedded.org wrote: > From: Aleksandar Nikolic <aleksandar.nikolic@zeiss.com> > > In case wks file references nvme0n1, update_fstab() function creates wrong entries > in /etc/fstab, as it is not adding prefix 'p' before the partition number (at the > moment this happens only for mmcblk). This leads that entries in /etc/fstab start > with: > > /dev/nvme0n11 > /dev/nvme0n13 > > instead with: > > /dev/nvme0n1p1 > /dev/nvme0n1p3 > > The patch fixes this as it extends the if condition and adds prefix 'p' for > both mmcblk and nvme0n1. This is an interesting case for our backport policy, as wic is no longer embedded within OE-core on wrynose or master. However, I think this should be submitted against the standalone wic repository first (see the readme [1] for contributing instructions) and we should get feedback from the maintainer. [1]: https://git.yoctoproject.org/wic/tree/README.md?h=master This should also handle other NVMe devices beyond the first, what about /dev/nvme1n1? Or an NVMe device with a second namespace /dev/nvme1n2? Best regards,
On Fri, May 22, 2026 at 12:53 PM, Paul Barker wrote: > > On Fri, 2026-05-22 at 06:59 +0200, Aleksandar Nikolic via > lists.openembedded.org wrote: > >> From: Aleksandar Nikolic <aleksandar.nikolic@zeiss.com> >> >> In case wks file references nvme0n1, update_fstab() function creates wrong >> entries >> in /etc/fstab, as it is not adding prefix 'p' before the partition number >> (at the >> moment this happens only for mmcblk). This leads that entries in >> /etc/fstab start >> with: >> >> /dev/nvme0n11 >> /dev/nvme0n13 >> >> instead with: >> >> /dev/nvme0n1p1 >> /dev/nvme0n1p3 >> >> The patch fixes this as it extends the if condition and adds prefix 'p' >> for >> both mmcblk and nvme0n1. > > This is an interesting case for our backport policy, as wic is no longer > embedded within OE-core on wrynose or master. However, I think this > should be submitted against the standalone wic repository first (see the > readme [1] for contributing instructions) and we should get feedback > from the maintainer. Yes, makes sense. I'll post a patch there as soon as I have added your suggestions. When/if it gets accepted we can backport it to scarthgap and whinlatter. > > [1]: https://git.yoctoproject.org/wic/tree/README.md?h=master > > This should also handle other NVMe devices beyond the first, what about > /dev/nvme1n1? Or an NVMe device with a second namespace /dev/nvme1n2? Thanks, will add your suggestions in the patch for the wic repository. Cheers, Aleksandar > > Best regards, > > -- > Paul Barker
diff --git a/scripts/lib/wic/plugins/imager/direct.py b/scripts/lib/wic/plugins/imager/direct.py index a1d152659b..5600d72e61 100644 --- a/scripts/lib/wic/plugins/imager/direct.py +++ b/scripts/lib/wic/plugins/imager/direct.py @@ -133,8 +133,8 @@ class DirectPlugin(ImagerPlugin): elif part.use_label: device_name = "LABEL=%s" % part.label else: - # mmc device partitions are named mmcblk0p1, mmcblk0p2.. - prefix = 'p' if part.disk.startswith('mmcblk') else '' + # mmc and nvme0n1 device partitions are named <disk>p1, <disk>p2.. + prefix = 'p' if part.disk.startswith(('mmcblk', 'nvme0n1')) else '' device_name = "/dev/%s%s%d" % (part.disk, prefix, part.realnum) opts = part.fsopts if part.fsopts else "defaults"