On Fri, 20 Jan 2023 14:24:50 +0000
Jonathan Cameron via <qemu-devel@nongnu.org> wrote:
> CXL uses PCI AER Internal errors to signal to the host that an error has
> occurred. The host can then read more detailed status from the CXL RAS
> capability.
>
> For uncorrectable errors: support multiple injection in one operation
> as this is needed to reliably test multiple header logging support in an
> OS. The equivalent feature doesn't exist for correctable errors, so only
> one error need be injected at a time.
>
> Note:
> - Header content needs to be manually specified in a fashion that
> matches the specification for what can be in the header for each
> error type.
>
> Injection via QMP:
> { "execute": "qmp_capabilities" }
> ...
> { "execute": "cxl-inject-uncorrectable-errors",
> "arguments": {
> "path": "/machine/peripheral/cxl-pmem0",
> "errors": [
> {
> "type": "cache-address-parity",
> "header": [ 3, 4]
> },
> {
> "type": "cache-data-parity",
> "header":
> [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
> },
> {
> "type": "internal",
> "header": [ 1, 2, 4]
> }
> ]
> }}
> ...
> { "execute": "cxl-inject-correctable-error",
> "arguments": {
> "path": "/machine/peripheral/cxl-pmem0",
> "type": "physical",
> "header": [ 3, 4]
Correctable errors don't have header logging (I misread the spec a long time
back and somehow didn't notice this was still here) I'll fix this up for v3
along with cleaning up the messy endian handling that was pointed out off list.
I don't intent (at this point) to deal with big endian support in general,
but it would be good not to make the situation worse than it already is
even if no one is shouting for big endian yet.
Thanks,
Jonathan
