Message ID | 13e82227-1de3-fd48-bd0f-99f07fca21db@suse.com (mailing list archive) |
---|---|
Headers | show |
Series | x86: Protected Processor Inventory Number (PPIN) support | expand |
On 08/11/2019 15:22, Jan Beulich wrote: > 1: include the PPIN in MCE records when available > 2: explicitly disallow guest access to PPIN > 3: provide Dom0 access to PPIN via XENPF_resource_op > > I have yet to get around to post the Linux side consumer > patch of the interface addition in patch 1. What about printing the PPIN information when Xen writes MCE details to the console? ~Andrew
On 13.12.2019 20:50, Andrew Cooper wrote: > On 08/11/2019 15:22, Jan Beulich wrote: >> 1: include the PPIN in MCE records when available >> 2: explicitly disallow guest access to PPIN >> 3: provide Dom0 access to PPIN via XENPF_resource_op >> >> I have yet to get around to post the Linux side consumer >> patch of the interface addition in patch 1. > > What about printing the PPIN information when Xen writes MCE details to > the console? Do you mean in x86_mcinfo_dump(), or mc_panic(), or yet somewhere else? In any event if we did so I'd want to arrange for each PPIN to get logged at most once, to reduce unnecessary redundancy. Of course there's a comment ahead of x86_mcinfo_dump() mentioning that the output is to be parseable by mcelog, so I'm not sure if we can reasonably change what we produce. Otoh I'm also not sure the comment hasn't become stale with its (presumed) porting from Linux anyway. Jan
On 16/12/2019 11:33, Jan Beulich wrote: > On 13.12.2019 20:50, Andrew Cooper wrote: >> On 08/11/2019 15:22, Jan Beulich wrote: >>> 1: include the PPIN in MCE records when available >>> 2: explicitly disallow guest access to PPIN >>> 3: provide Dom0 access to PPIN via XENPF_resource_op >>> >>> I have yet to get around to post the Linux side consumer >>> patch of the interface addition in patch 1. >> What about printing the PPIN information when Xen writes MCE details to >> the console? > Do you mean in x86_mcinfo_dump(), or mc_panic(), or yet somewhere > else? Something which ends up on the console. > In any event if we did so I'd want to arrange for each PPIN > to get logged at most once, to reduce unnecessary redundancy. Getting it printed, even redundantly is more helpful than not having it printed. > Of course there's a comment ahead of x86_mcinfo_dump() mentioning > that the output is to be parseable by mcelog, so I'm not sure if > we can reasonably change what we produce. Otoh I'm also not sure > the comment hasn't become stale with its (presumed) porting from > Linux anyway. That comment is long obsolete. What Xen prints needs substantial rearranging for mcelog to parse it these days. ~Andrew