Re: [RFC/Patch 4/5] libndctl Make interfaces to use Translate SPA
> On Fri, Aug 4, 2017 at 2:12 AM, Yasunori Goto wrote: > > ndctl:libndctl Make interfaces to use Translate SPA. > > A process note, please separate kernel patches and ndctl patches into > a different set. The expectation is that all the [PATCH 1/n] to [PATCH > n/n] should apply to the same project. > > I have typically been marking my ndctl patches with a prefix like > [ndctl PATCH 1/n], where the kernel patches just say [PATCH]. > > > > > This patch makes 3 new interfaces : > > - to ask bus has translate SPA feature. > > - to call translate SPA. > > - to find DIMM by SPA address. > > > > > > Note) I'm not sure how many buffer should be prepared, because > > it depends on max # of mirroring way. > > This patch assume maxmum # is 4 way. > > I think it is ok for now to just reserve space for 1-DIMM. When / if > mirroring support materializes in the kernel you should then be able > to detect how many DIMMs are in the mirror set. Since mirror support > would require changes to the Namespace Label specification I think we > have time to circle back and fix up ndctl for the multiple DIMM per > Translate SPA request case. I see. I'll reserve space for 1-DIMM. > > Otherwise, the kernel will handle the case where the provided buffer > is too small and return an error. > ___ Linux-nvdimm mailing list Linux-nvdimm@lists.01.org https://lists.01.org/mailman/listinfo/linux-nvdimm
Re: [RFC/Patch 4/5] libndctl Make interfaces to use Translate SPA
On Fri, Aug 4, 2017 at 2:12 AM, Yasunori Goto wrote: > ndctl:libndctl Make interfaces to use Translate SPA. A process note, please separate kernel patches and ndctl patches into a different set. The expectation is that all the [PATCH 1/n] to [PATCH n/n] should apply to the same project. I have typically been marking my ndctl patches with a prefix like [ndctl PATCH 1/n], where the kernel patches just say [PATCH]. > > This patch makes 3 new interfaces : > - to ask bus has translate SPA feature. > - to call translate SPA. > - to find DIMM by SPA address. > > > Note) I'm not sure how many buffer should be prepared, because > it depends on max # of mirroring way. > This patch assume maxmum # is 4 way. I think it is ok for now to just reserve space for 1-DIMM. When / if mirroring support materializes in the kernel you should then be able to detect how many DIMMs are in the mirror set. Since mirror support would require changes to the Namespace Label specification I think we have time to circle back and fix up ndctl for the multiple DIMM per Translate SPA request case. Otherwise, the kernel will handle the case where the provided buffer is too small and return an error. ___ Linux-nvdimm mailing list Linux-nvdimm@lists.01.org https://lists.01.org/mailman/listinfo/linux-nvdimm
