On 2/4/22 09:15, Shivaprasad G Bhat wrote:
If the device backend is not persistent memory for the nvdimm, there is need for explicit IO flushes to ensure persistence. On SPAPR, the issue is addressed by adding a new hcall to request for an explicit flush from the guest when the backend is not pmem. So, the approach here is to convey when the hcall flush is required in a device tree property. The guest once it knows the device needs explicit flushes, makes the hcall as and when required. It was suggested to create a new device type to address the explicit flush for such backends on PPC instead of extending the generic nvdimm device with new property. So, the patch introduces the spapr-nvdimm device. The new device inherits the nvdimm device with the new bahviour such that if the backend has pmem=no, the device tree property is set by default. The below demonstration shows the map_sync behavior for non-pmem backends. (https://github.com/avocado-framework-tests/avocado-misc-tests/blob/master/memory/ndctl.py.data/map_sync.c) The pmem0 is from spapr-nvdimm with with backend pmem=on, and pmem1 is from spapr-nvdimm with pmem=off, mounted as /dev/pmem0 on /mnt1 type xfs (rw,relatime,attr2,dax=always,inode64,logbufs=8,logbsize=32k,noquota) /dev/pmem1 on /mnt2 type xfs (rw,relatime,attr2,dax=always,inode64,logbufs=8,logbsize=32k,noquota) [root@atest-guest ~]# ./mapsync /mnt1/newfile ----> When pmem=on [root@atest-guest ~]# ./mapsync /mnt2/newfile ----> when pmem=off Failed to mmap with Operation not supported First patch adds the realize/unrealize call backs to the generic device for the new device's vmstate registration. The second patch implements the hcall, adds the necessary vmstate properties to spapr machine structure for carrying the hcall status during save-restore. The nature of the hcall being asynchronus, the patch uses aio utilities to offload the flush. The third patch introduces the spapr-nvdimm device, adds the device tree property for the guest when spapr-nvdimm is used with pmem=no on the backend. Also adds new property pmem-override(?, suggest if you have better name) to the spapr-nvdimm which hints at forcing the hcall based flushes even on pmem backed devices. The kernel changes to exploit this hcall is at https://github.com/linuxppc/linux/commit/75b7c05ebf9026.patch
Applied for ppc-7.0 Thanks, C.
