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:   Conor Dooley <conor.dooley@microchip.com>
To:     <stable@vger.kernel.org>
CC:     <conor.dooley@microchip.com>, <Brice.Goglin@inria.fr>,
        <atishp@atishpatra.org>, <catalin.marinas@arm.com>,
        <gregkh@linuxfoundation.org>,
        <linux-arm-kernel@lists.infradead.org>,
        <linux-kernel@vger.kernel.org>, <linux-riscv@lists.infradead.org>,
        <palmer@dabbelt.com>, <sudeep.holla@arm.com>, <will@kernel.org>,
        Atish Patra <atishp@rivosinc.com>
Subject: [PATCH 5.4 2/2] riscv: topology: fix default topology reporting
Date:   Wed, 19 Oct 2022 13:52:10 +0100
Message-ID: <20221019125209.2844943-2-conor.dooley@microchip.com>
In-Reply-To: <20221019125209.2844943-1-conor.dooley@microchip.com>
References: <20221019125209.2844943-1-conor.dooley@microchip.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain
Precedence: bulk

commit fbd92809997a391f28075f1c8b5ee314c225557c upstream.

RISC-V has no sane defaults to fall back on where there is no cpu-map
in the devicetree.
Without sane defaults, the package, core and thread IDs are all set to
-1. This causes user-visible inaccuracies for tools like hwloc/lstopo
which rely on the sysfs cpu topology files to detect a system's
topology.

On a PolarFire SoC, which should have 4 harts with a thread each,
lstopo currently reports:

Machine (793MB total)
  Package L#0
    NUMANode L#0 (P#0 793MB)
    Core L#0
      L1d L#0 (32KB) + L1i L#0 (32KB) + PU L#0 (P#0)
      L1d L#1 (32KB) + L1i L#1 (32KB) + PU L#1 (P#1)
      L1d L#2 (32KB) + L1i L#2 (32KB) + PU L#2 (P#2)
      L1d L#3 (32KB) + L1i L#3 (32KB) + PU L#3 (P#3)

Adding calls to store_cpu_topology() in {boot,smp} hart bringup code
results in the correct topolgy being reported:

Machine (793MB total)
  Package L#0
    NUMANode L#0 (P#0 793MB)
    L1d L#0 (32KB) + L1i L#0 (32KB) + Core L#0 + PU L#0 (P#0)
    L1d L#1 (32KB) + L1i L#1 (32KB) + Core L#1 + PU L#1 (P#1)
    L1d L#2 (32KB) + L1i L#2 (32KB) + Core L#2 + PU L#2 (P#2)
    L1d L#3 (32KB) + L1i L#3 (32KB) + Core L#3 + PU L#3 (P#3)

CC: stable@vger.kernel.org # 456797da792f: arm64: topology: move store_cpu_topology() to shared code
Fixes: 03f11f03dbfe ("RISC-V: Parse cpu topology during boot.")
Reported-by: Brice Goglin <Brice.Goglin@inria.fr>
Link: https://github.com/open-mpi/hwloc/issues/536
Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
---
I just resolved the conflicts. Tested in QEMU only. SMP doesn't seem to work in
5.4, using the same command that works for 5.10.
---
 arch/riscv/Kconfig          | 2 +-
 arch/riscv/kernel/smpboot.c | 4 +++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index b21549a34447..e0a77af5c130 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -51,7 +51,7 @@ config RISCV
 	select PCI_MSI if PCI
 	select RISCV_TIMER
 	select GENERIC_IRQ_MULTI_HANDLER
-	select GENERIC_ARCH_TOPOLOGY if SMP
+	select GENERIC_ARCH_TOPOLOGY
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_MMIOWB
 	select HAVE_EBPF_JIT if 64BIT
diff --git a/arch/riscv/kernel/smpboot.c b/arch/riscv/kernel/smpboot.c
index 261f4087cc39..0576a6b2bcc5 100644
--- a/arch/riscv/kernel/smpboot.c
+++ b/arch/riscv/kernel/smpboot.c
@@ -46,6 +46,8 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 {
 	int cpuid;
 
+	store_cpu_topology(smp_processor_id());
+
 	/* This covers non-smp usecase mandated by "nosmp" option */
 	if (max_cpus == 0)
 		return;
@@ -142,8 +144,8 @@ asmlinkage __visible void __init smp_callin(void)
 	current->active_mm = mm;
 
 	trap_init();
+	store_cpu_topology(smp_processor_id());
 	notify_cpu_starting(smp_processor_id());
-	update_siblings_masks(smp_processor_id());
 	set_cpu_online(smp_processor_id(), 1);
 	/*
 	 * Remote TLB flushes are ignored while the CPU is offline, so emit
-- 
2.38.0