Unlike what your commit message says, you _can_ run at least some of the functionality with no HW and no specially configured qemu: append memmap=4G!20G to the kernel's cmdline, where "4G" is the size of an emulated nvdimm, and 20G is its start in physical memory. This stanza can be used multiple times, obviously for non-overlapping regions.
Such an emulated nvdimm has no label. Without a label, you can't partition the nvdimm -- there can be only exactly one namespace per memmap stanza (so you can't delete it, but create-namespace -e works same as delete+create). ndctl: * changing namespace's type: ✓ * partitioning: ✗ * error injection: requires "nfit_test" kernel module * on-chip encryption: ✗ * on-chip scrub: ✗ daxctl: * switching devdax↔system-ram: ✓ Thus, compared to qemu, you lose labels, gain very clunky error simulation. This way of emulating nvdimms is even more useful for higher layers of the stack, as you don't need to muck with qemu but can use your regular bare-metal dev environment. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1853506 Title: [MIR] ndctl To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/ndctl/+bug/1853506/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs