Re: failure to setup devdax over fake nvdimm

Le 16/03/2019 à 02:19, Dan Williams a écrit : > > Hmm, ok. Can you boot with libnvdimm.dyndbg and see if it squawks > about the problem? If this is 1ffc664f9b8c then you need to have the > dax_pmem_compat driver or the dax_pmem driver to attach, and this also > won't work if you've hotplugged it with the kmem driver. Things work when dax_pmem_compat is loaded. [ 93.387078] nd_bus ndbus0: START: dax_pmem_compat.probe(dax0.0) [ 93.387082] nd_bus ndbus0: END: dax_pmem_compat.probe(dax0.0) = -19 [ 93.387101] nd_bus ndbus0: START: dax_pmem_compat.probe(dax1.0) [ 93.387103] nd_bus ndbus0: END: dax_pmem_compat.probe(dax1.0) = -19 [ 96.741806] nd_bus ndbus0: nd_pmem.remove(namespace1.0) = 0 [ 96.940567] nd namespace1.0: fail (-6) [ 96.941343] nd namespace1.0: fail (-6) [ 96.941488] nd dax1.0: result: 0 wrote: 1cdae3a1-5ede-45d6-a3fc-31bb5c4c891a [ 96.941583] nd dax1.0: result: 0 wrote: pmem [ 96.941650] nd dax1.0: result: 0 wrote: 2097152 [ 96.941745] nd dax1.0: result: 13 wrote: namespace1.0 [ 96.942445] nd_bus ndbus0: START: dax_pmem_compat.probe(dax1.0) [ 97.087764] dax1.0 initialised, 6192640 pages in 124ms [ 97.093554] nd_bus ndbus0: END: dax_pmem_compat.probe(dax1.0) = 0 [ 97.093673] nd_bus ndbus0: START: dax_pmem_compat.probe(dax1.1) [ 97.093677] nd_bus ndbus0: END: dax_pmem_compat.probe(dax1.1) = -19 When only dax_pmem is loaded (compat is blacklisted), it fails with: [ 36.141032] nd_bus ndbus0: nd_pmem.remove(namespace1.0) = 0 [ 36.343444] nd namespace1.0: fail (-6) [ 36.344268] nd namespace1.0: fail (-6) [ 36.344440] nd dax1.0: result: 0 wrote: bf1ed9eb-5fc1-4637-85e5-c67a59c7c59d [ 36.344528] nd dax1.0: result: 0 wrote: pmem [ 36.344595] nd dax1.0: result: 0 wrote: 2097152 [ 36.344704] nd dax1.0: result: 13 wrote: namespace1.0 [ 36.345802] nd dax1.0: result: 2 wrote: Boot dmesg below (before running ndctl).