RE: [Question] Regarding presence of duplicate ACPI CPU entries at two nodes \\_SB.CXXX and \\_SB.CPUS.CXXX of namespace

[Salil Mehta <salil.mehta@huawei.com>]

Please ignore this!


> From: Linuxarm [mailto:linuxarm-bounces@huawei.com] On Behalf Of Salil Mehta
> Sent: Wednesday, January 29, 2020 11:51 AM
> To: qemu-devel@nongnu.org; qemu-arm@nongnu.org
> Cc: peter.maydell@linaro.org; drjones@redhat.com; sameo@linux.intel.com;
> mst@redhat.com; Linuxarm <linuxarm@huawei.com>; eric.auger@redhat.com;
> pbonzini <pbonzini@redhat.com>; Igor Mammedov <imammedo@redhat.com>;
> lersek@redhat.com
> Subject: [Question] Regarding presence of duplicate ACPI CPU entries at two nodes
> \\_SB.CXXX and \\_SB.CPUS.CXXX of namespace
> 
> Hello,
> 
> Observation:
> If we launch Linux Guest VM using QEMU(running on any type host. I am using x86)
> with
> CPU based on any ARM64 architecture then I could see QEMU populating ACPI nodes
> related to same CPU at 2 places of the ACPI namespace:
> 1. \\_SB.CXXX
> 2. \\_SB.CPUS.CXXX
> 
> Above results in Guest VM showing duplicate CPU entries in the sysfs for the
> same CPUS.
> I could make out the entries under \\_SB.CPUS.XXX are part of the container.
> 
> estuary:/$ ls -al /sys/bus/acpi/devices/
> 
> Observation 1: (belongs to \\_SB.C00X)
> ACPI0007:00 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0007:00
> ACPI0007:01 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0007:01
> ACPI0007:02 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0007:02
> ACPI0007:03 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0007:03
> ACPI0007:04 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0007:04
> ACPI0007:05 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0007:05
> 
> Observation 2: (belongs to \\_SB.CPUS.C00X and under container ACPI0010:00 part
> of \\_SB.CPUS)
> ACPI0007:06
> -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00/ACPI0007:06
> ACPI0007:07
> -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00/ACPI0007:07
> ACPI0007:08
> -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00/ACPI0007:08
> ACPI0007:09
> -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00/ACPI0007:09
> ACPI0007:0a
> -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00/ACPI0007:0a
> ACPI0007:0b
> -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00/ACPI0007:0b
> ACPI0010:00 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0010:00
> 
> 
> estuary:/$ cat /sys/bus/acpi/devices/ACPI0010\:00/path
> \_SB_.CPUS
> 
> estuary:/$ cat /sys/bus/acpi/devices/ACPI0007\:00/uid
> 0
> estuary:/$ cat /sys/bus/acpi/devices/ACPI0007\:00/path
> \_SB_.C000
> 
> estuary:/$ cat /sys/bus/acpi/devices/ACPI0007\:06/uid
> 0
> estuary:/$ cat /sys/bus/acpi/devices/ACPI0007\:06/path
> \_SB_.CPUS.C000
> 
> 
> 
> QEMU Code Excerpt:
> I could trace with in QEMU AML code, the CPUS are being appended at 2 places:
> 
> 
> Code 1. File: hw/arm/virt-acpi-build.c (Cause of Observation 1 i.e. \\_SB.CXXX )
> 
> static void acpi_dsdt_add_cpus(Aml *scope, int smp_cpus)
> {
>     uint16_t i;
> 
>     for (i = 0; i < smp_cpus; i++) { --->{should be possible cpus anyways}
>         Aml *dev = aml_device("C%.03X", i);
>         aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0007")));
>         aml_append(dev, aml_name_decl("_UID", aml_int(i)));
>         aml_append(scope, dev);
>     }
> }
> 
> 
> Code 2. File: hw/acpi/cpu.c (Cause of Observation 2 i.e. \\_SB.CPUS.CXXX)
> 
> void build_cpus_aml(..)
> {
> [...]
>   cpus_dev = aml_device("\\_SB.CPUS");
>   {
>     [...]
>             /* build Processor object for each processor */
>         for (i = 0; i < arch_ids->len; i++) {
>             Aml *dev;
>             Aml *uid = aml_int(i);
>             GArray *madt_buf = g_array_new(0, 1, 1);
>             int arch_id = arch_ids->cpus[i].arch_id;
> 
>             if (opts.acpi_1_compatible && arch_id < 255) {
>                 dev = aml_processor(i, 0, 0, CPU_NAME_FMT, i);
>             } else {
>                 dev = aml_device(CPU_NAME_FMT, i);
>                 aml_append(dev, aml_name_decl("_HID",
> aml_string("ACPI0007")));
>                 aml_append(dev, aml_name_decl("_UID", uid));
>             }
>     [...]
>   }
> [...]
> }
> 
> 
> 
> Questions:
> Q1. I could not understand the purpose of keeping acpi_dsdt_add_cpus() after
> the code 2.
>      was introduced as part of the below change and which is already adding CPUS
> related
>      AML to \\_SB.CPUS namespace?
> 
> acpi: cpuhp: add CPU devices AML with _STA method
> commit 5e1b5d93887b52eede156f846b6c4c5c8bbcfcdb
> 
> Q2. Do we really require CPUs being added by acpi_dsdt_add_cpus() in \\_SB.CXXX
>      Namespace OR is it a stray code left?
> 
> 
> 
> Please help to correct if there is a gap in my understanding here and please
> forgive me if I have terribly missed out something very basic here.
> 
> Many thanks!
> 
> Best Regards
> Salil
>