On Tue, Jul 19, 2016 at 11:52:53AM -0700, Dan Williams wrote: > On Tue, Jul 19, 2016 at 11:50 AM, Linda Knippers <linda.knippers@xxxxxxx> wrote: > > > > > > On 7/19/2016 1:11 PM, Jerry Hoemann wrote: > >> On Fri, Jun 24, 2016 at 10:44:25AM -0700, Dan Williams wrote: > >>> QEMU 2.6 implements nascent support for nvdimm DSMs. Depending on > >>> configuration it may only implement the function0 dsm to indicate that > >>> no other DSMs are available. Commit 31eca76ba2fc "nfit, libnvdimm: > >>> limited/whitelisted dimm command marshaling mechanism" breaks QEMU, but > >>> QEMU is spec compliant. Per the spec the way to indicate that no > >>> functions are supported is: > >>> > >>> If Function Index is zero, the return is a buffer containing one bit > >>> for each function index, starting with zero. Bit 0 indicates whether > >>> there is support for any functions other than function 0 for the > >>> specified UUID and Revision ID. If set to zero, no functions are > >>> supported (other than function zero) for the specified UUID and > >>> Revision ID. > >> > >> > >> Dan, > >> > >> This breaks calling DSM on HPE platforms and is a regression. > >> > >> E-mail context can be found at: > >> > >> https://lists.01.org/pipermail/linux-nvdimm/2016-June/006099.html > >> > >> The problem with this change is that it assumes that ACPI returning an > >> object means that the UUID is supported on that platform. > >> > >> However, looking at ACPI v 6.1 section 9.1.1, the example for > >> evaluating a _DSM shows that if the UUID is not supported at all, > >> a zeroed out buffer of length 1 is returned: > >> > >> // > >> // If not one of the UUIDs we recognize, then return a buffer > >> // with bit 0 set to 0 indicating no functions supported. > >> // > >> return(Buffer(){0}) > >> > >> HPE firmware has been following this practice for a long long time. > >> I suspect other manufacturer's firmware do so as well. > >> > >> The problem is that this creates an ambiguity and the linux code > >> is no longer differentiating the case where the DSM/UUID is > >> supported but only implements function 0 (the QEMU case you're > >> trying to accommodate) and the case that the DSM/UUID is not > >> supported at all. > >> > >> The result is that the code in acpi_nfit_add_dimm: > >> > >> for (i = NVDIMM_FAMILY_INTEL; i <= NVDIMM_FAMILY_HPE2; i++) > >> - if (acpi_check_dsm(adev_dimm->handle, to_nfit_uuid(i), 1, 1)) > >> + if (acpi_check_dsm(adev_dimm->handle, to_nfit_uuid(i), 1, 0)) > >> break; > >> > >> > >> always matches NVDIMM_FAMILY_INTEL. This is either because its > >> is an Intel nvdimm, or because the unsupported UUID returns back > >> a zeroed out buffer of length 1. > >> > >> > >> As nfit_mem->family always equals NVDIMM_FAMILY_INTEL, no subsequent > >> DSM call will succeed for NVDIMM_FAMILY_HPE1 or any other > >> family. > >> > >> I don't have a fix as of yet, but wanted to make you aware of > >> the problem. > > > > Could we try the all known UUIDs looking for one that returns a non-zero > > value? > > > > Yes, that seems like the way forward, and also make not finding a DSM > family non-fatal. Reverting this change would address for HPE, but I do not fully understand the nature of the problem this change was attempting to address. Can you expand a bit? -- ----------------------------------------------------------------------------- Jerry Hoemann Software Engineer Hewlett Packard Enterprise ----------------------------------------------------------------------------- -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
