On Wed, 2019-05-08 at 16:02 -0400, Pavel Tatashin wrote:
> Hi Vishal,
>
> I am trying to use your changes, and getting the following error:
> libdaxctl: __sysfs_device_parse: dax0.0: add_dev() failed
>
> Here is full output:
> # ndctl create-namespace --mode devdax --map mem -e namespace0.0 -f
> [ 26.891054] dax0.0 initialised, 524288 pages in 9ms
> [ 26.894239] random: ndctl: uninitialized urandom read (4 bytes read)
> libdaxctl: __sysfs_device_parse: dax0.0: add_dev() failed
> {
> "dev":"namespace0.0",
> "mode":"devdax",
> "map":"mem",
> "size":"2046.00 MiB (2145.39 MB)",
> "uuid":"6684b3b0-4ab1-45ba-9ce6-48aa046b5fc1",
> "daxregion":{
> "id":0,
> "size":"2046.00 MiB (2145.39 MB)",
> "align":2097152
> },
> "align":2097152
> }
Thanks for the report!
I suspect you may be using the "legacy" device-model. You can
confirm by seeing if that device even shows up on the "dax" bus.
Compare /sys/bus/dax/devices/ vs /sys/class/dax/
If you see stuff in /sys/class/dax and nothing in
/sys/bus/dax/devices then you need to configure the system to switch
to the device-dax bus-model by running either of the following
commands:
# daxctl migrate-device-model
...or:
# cat <<EOF > /etc/modprobe.d/daxctl.conf
blacklist dax_pmem_compat
alias nd:t7* dax_pmem
EOF
...and then reload the subsystem:
ndctl disable-region all
ndctl enable-region all
That said, if this was the cause, we should improve the failure mode in
this case - I'll look into that, and test by forcing dax-class too.
Thanks,
-Vishal
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
