Re: [v5 0/3] "Hotremove" persistent memory

[Pavel Tatashin]

On Thu, May 2, 2019 at 6:29 PM Verma, Vishal L <vishal.l.verma@xxxxxxxxx> wrote:
>
> On Thu, 2019-05-02 at 17:44 -0400, Pavel Tatashin wrote:
>
> > > In running with these patches, and testing the offlining part, I ran
> > > into the following lockdep below.
> > >
> > > This is with just these three patches on top of -rc7.
> >
> > Hi Verma,
> >
> > Thank you for testing. I wonder if there is a command sequence that I
> > could run to reproduce it?
> > Also, could you please send your config and qemu arguments.
> >
> Yes, here is the qemu config:
>
> qemu-system-x86_64
>         -machine accel=kvm
>         -machine pc-i440fx-2.6,accel=kvm,usb=off,vmport=off,dump-guest-core=off,nvdimm
>         -cpu Haswell-noTSX
>         -m 12G,slots=3,maxmem=44G
>         -realtime mlock=off
>         -smp 8,sockets=2,cores=4,threads=1
>         -numa node,nodeid=0,cpus=0-3,mem=6G
>         -numa node,nodeid=1,cpus=4-7,mem=6G
>         -numa node,nodeid=2
>         -numa node,nodeid=3
>         -drive file=/virt/fedora-test.qcow2,format=qcow2,if=none,id=drive-virtio-disk1
>         -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x9,drive=drive-virtio-disk1,id=virtio-disk1,bootindex=1
>         -object memory-backend-file,id=mem1,share,mem-path=/virt/nvdimm1,size=16G,align=128M
>         -device nvdimm,memdev=mem1,id=nv1,label-size=2M,node=2
>         -object memory-backend-file,id=mem2,share,mem-path=/virt/nvdimm2,size=16G,align=128M
>         -device nvdimm,memdev=mem2,id=nv2,label-size=2M,node=3
>         -serial stdio
>         -display none
>
> For the command list - I'm using WIP patches to ndctl/daxctl to add the
> command I mentioned earlier. Using this command, I can reproduce the
> lockdep issue. I thought I should be able to reproduce the issue by
> onlining/offlining through sysfs directly too - something like:
>
>    node="$(cat /sys/bus/dax/devices/dax0.0/target_node)"
>    for mem in /sys/devices/system/node/node"$node"/memory*; do
>      echo "offline" > $mem/state
>    done
>
> But with that I can't reproduce the problem.
>
> I'll try to dig a bit deeper into what might be happening, the daxctl
> modifications simply amount to doing the same thing as above in C, so
> I'm not immediately sure what might be happening.
>
> If you're interested, I can post the ndctl patches - maybe as an RFC -
> to test with.

I could apply the patches and test with them. Also, could you please
send your kernel config.

Thank you,
Pasha

>
> Thanks,
> -Vishal
>
>
>