Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Thomas Gleixner <tglx@linutronix.de>
To:     Sohil Mehta <sohil.mehta@intel.com>,
        LKML <linux-kernel@vger.kernel.org>
Cc:     x86@kernel.org, Tom Lendacky <thomas.lendacky@amd.com>,
        Andrew Cooper <andrew.cooper3@citrix.com>,
        Arjan van de Ven <arjan@linux.intel.com>,
        Huang Rui <ray.huang@amd.com>, Juergen Gross <jgross@suse.com>,
        Dimitri Sivanich <dimitri.sivanich@hpe.com>,
        Michael Kelley <mikelley@microsoft.com>,
        K Prateek Nayak <kprateek.nayak@amd.com>,
        Kan Liang <kan.liang@linux.intel.com>,
        Zhang Rui <rui.zhang@intel.com>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Feng Tang <feng.tang@intel.com>,
        Andy Shevchenko <andy@infradead.org>
Subject: Re: [patch 00/53] x86/topology: The final installment
In-Reply-To: <c8ca9e16-ce03-1fe0-cf9f-f6cc2302dbac@intel.com>
References: <20230807130108.853357011@linutronix.de>
 <c8ca9e16-ce03-1fe0-cf9f-f6cc2302dbac@intel.com>
Date:   Tue, 08 Aug 2023 21:10:41 +0200
Message-ID: <87h6p9l58u.ffs@tglx>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On Tue, Aug 08 2023 at 11:29, Sohil Mehta wrote:
> On 8/7/2023 6:52 AM, Thomas Gleixner wrote:
> However, I am a bit confused with the dmesg results.
>
> Dmesg output
> ------------
> CPU topo: Max. logical packages:   4
> CPU topo: Max. logical dies:       4
> CPU topo: Max. dies per package:   1
> CPU topo: Max. threads per core:   2
> CPU topo: Num. cores per package:    16
> CPU topo: Num. threads per package:  32

That's indeed weird. Can you please provide:

     - the output of 'cpuid -r'
     - the output of /sys/kernel/debug/x86/topo/domains
     - the APIC IDs of all CPUs (see below patch)

> Questions
> ---------
> 1) Before this series, the Max logical packages used to be logged as 8
> in dmesg. But now it shows up as 4. Is that expected?
> To me, it seems that to get to 160 potential CPUs with 10cores/20threads
> per package, the Max logical packages should be 8.

As the number of cores per package is not consistent, this is not a
surprise. 160/32 = 5, but yes something is fishy there.

Oh. Is this perhaps one of those machines where the APICs are enumerated
twice. Can you apply the patch below?

The ACPI part is a modified variant of:

    https://lore.kernel.org/r/23ccb5aad770fb52014dca9fb1e4353b5c3cede9.camel@intel.com

Please apply the topology hunks first without the ACPI changes and then
try the ACPI change on top.

Thanks,

        tglx
---
 arch/x86/kernel/acpi/boot.c    |   31 ++++++++++++++++---------------
 arch/x86/kernel/cpu/topology.c |    2 ++
 2 files changed, 18 insertions(+), 15 deletions(-)

--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -171,6 +171,8 @@ static bool __init acpi_is_processor_usa
 	return false;
 }
 
+static bool has_lapic_cpus;
+
 static int __init
 acpi_parse_x2apic(union acpi_subtable_headers *header, const unsigned long end)
 {
@@ -241,6 +243,14 @@ acpi_parse_lapic(union acpi_subtable_hea
 		return 0;
 
 	/*
+	 * According to https://uefi.org/specs/ACPI/6.5/05_ACPI_Software_Programming_Model.html#processor-local-x2apic-structure
+	 * when MADT provides both valid LAPIC and x2APIC entries, the APIC ID
+	 * in x2APIC must be equal or greater than 0xff.
+	 */
+	if (has_lapic_cpus && apic_id < 0xff)
+		return 0;
+
+	/*
 	 * We need to register disabled CPU as well to permit
 	 * counting disabled CPUs. This allows us to size
 	 * cpus_possible_map more accurately, to permit
@@ -1084,21 +1094,12 @@ static int __init acpi_parse_madt_lapic_
 				      acpi_parse_sapic, MAX_LOCAL_APIC);
 
 	if (!count) {
-		memset(madt_proc, 0, sizeof(madt_proc));
-		madt_proc[0].id = ACPI_MADT_TYPE_LOCAL_APIC;
-		madt_proc[0].handler = acpi_parse_lapic;
-		madt_proc[1].id = ACPI_MADT_TYPE_LOCAL_X2APIC;
-		madt_proc[1].handler = acpi_parse_x2apic;
-		ret = acpi_table_parse_entries_array(ACPI_SIG_MADT,
-				sizeof(struct acpi_table_madt),
-				madt_proc, ARRAY_SIZE(madt_proc), MAX_LOCAL_APIC);
-		if (ret < 0) {
-			pr_err("Error parsing LAPIC/X2APIC entries\n");
-			return ret;
-		}
-
-		count = madt_proc[0].count;
-		x2count = madt_proc[1].count;
+		count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_LAPIC,
+					acpi_parse_lapic, MAX_LOCAL_APIC);
+		if (count)
+			has_lapic_cpus = true;
+		x2count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_X2APIC,
+					acpi_parse_x2apic, MAX_LOCAL_APIC);
 	}
 	if (!count && !x2count) {
 		pr_err("No LAPIC entries present\n");
--- a/arch/x86/kernel/cpu/topology.c
+++ b/arch/x86/kernel/cpu/topology.c
@@ -159,6 +159,8 @@ void __init topology_register_apic(u32 a
 		return;
 	}
 
+	pr_info("Register %03x %d\n", apic_id, present);
+
 	if (present) {
 		/*
 		 * Prevent double registration, which is valid in case of