2018-08-19 16:13:07

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 00/16] Add support for Hygon Dhyana Family 18h processor

As a new x86 CPU Vendor, Chengdu Haiguang IC Design Co., Ltd (Hygon)
is a Joint Venture between AMD and Haiguang Information Technology Co.,
Ltd., and aims at providing high performance x86 processor for China
server market.

The first generation Hygon's processor(Dhyana) originates from AMD
technology and shares most of the architecture with AMD's family 17h,
but with different CPU Vendor ID("HygonGenuine")/PCIE Device Vendor ID
(0x1D94)/Family series number (Family 18h).

To enable the support of Linux kernel to Hygon's CPU, we added a new
vendor type (X86_VENDOR_HYGON, with value of 9) in arch/x86/include/
asm/processor.h, and shared most of kernel support codes with AMD
family 17h.

This patch series have been applied and tested successfully in Hygon's
Dhyana SoC silicon. Also tested on AMD's EPYC (Family 17h) processor
works fine and makes no harm to existing codes.


v3->v4:
- Rebased on 4.18.3 and tested against it.
- Merge patchs 05/17 perfctr and 10/17 events in v3 to patch 05/16
PMU for better patch function group.
- Add hygon_get_topology_early() in patch 01/16.
- Rework vendor checking and refine coding style.
- Add Acked-by from Bjorn Helgaas for pci.
- Add Acked-by from Rafael J. Wysocki for cpufreq and acpi.

v2->v3:
- Rebased on 4.18-rc8 and tested against it.
- Rework vendor checking codes to improve consistency.

v1->v2:
- Rebased on 4.18-rc6 and tested against it.
- Split the patchset to small series of patches.
- Rework patch descriptions.
- Create a separated arch/x86/kernel/cpu/hygon.c for Dhyana CPU
initialization to reduce long-term maintenance effort.


Pu Wen (16):
x86/cpu: create Dhyana init file and register new cpu_dev to system
x86/cache: get cache size/leaves and setup cache cpumap for Dhyana
x86/mtrr: get MTRR number and support TOP_MEM2
x86/smpboot: smp init nodelay and no flush caches before sleep
x86/pmu: enable Hygon support to PMU infrastructure
x86/nops: init ideal_nops for Hygon
x86/pci: add Hygon PCI vendor and northbridge support
x86/apic: add modern APIC support for Hygon
x86/bugs: add lfence mitigation to spectre v2 and no meltdown for
Hygon
x86/mce: enable Hygon support to MCE infrastructure
x86/kvm: enable Hygon support to KVM infrastructure
x86/xen: enable Hygon support to Xen
driver/acpi: enable Hygon support to ACPI driver
driver/cpufreq: enable Hygon support to cpufreq driver
driver/edac: enable Hygon support to AMD64 EDAC driver
tools/cpupower: enable Hygon support to cpupower tool

MAINTAINERS | 6 +
arch/x86/Kconfig | 2 +-
arch/x86/Kconfig.cpu | 13 +
arch/x86/events/amd/core.c | 6 +
arch/x86/events/amd/uncore.c | 15 +-
arch/x86/events/core.c | 4 +
arch/x86/include/asm/cacheinfo.h | 1 +
arch/x86/include/asm/kvm_emulate.h | 4 +
arch/x86/include/asm/mce.h | 5 +
arch/x86/include/asm/nospec-branch.h | 4 +-
arch/x86/include/asm/processor.h | 3 +-
arch/x86/include/asm/virtext.h | 5 +-
arch/x86/kernel/alternative.c | 4 +
arch/x86/kernel/amd_nb.c | 51 ++-
arch/x86/kernel/apic/apic.c | 19 +-
arch/x86/kernel/apic/probe_32.c | 1 +
arch/x86/kernel/cpu/Makefile | 1 +
arch/x86/kernel/cpu/bugs.c | 28 +-
arch/x86/kernel/cpu/cacheinfo.c | 31 +-
arch/x86/kernel/cpu/common.c | 1 +
arch/x86/kernel/cpu/cpu.h | 1 +
arch/x86/kernel/cpu/hygon.c | 409 +++++++++++++++++++++
arch/x86/kernel/cpu/mcheck/mce-severity.c | 3 +-
arch/x86/kernel/cpu/mcheck/mce.c | 21 +-
arch/x86/kernel/cpu/mtrr/cleanup.c | 3 +-
arch/x86/kernel/cpu/mtrr/generic.c | 5 +-
arch/x86/kernel/cpu/mtrr/mtrr.c | 2 +-
arch/x86/kernel/cpu/perfctr-watchdog.c | 2 +
arch/x86/kernel/smpboot.c | 4 +-
arch/x86/kvm/emulate.c | 11 +-
arch/x86/pci/amd_bus.c | 6 +-
arch/x86/xen/pmu.c | 12 +-
drivers/acpi/acpi_pad.c | 1 +
drivers/acpi/processor_idle.c | 1 +
drivers/cpufreq/acpi-cpufreq.c | 5 +
drivers/cpufreq/amd_freq_sensitivity.c | 9 +-
drivers/edac/amd64_edac.c | 52 ++-
drivers/edac/amd64_edac.h | 5 +
drivers/edac/mce_amd.c | 15 +-
include/linux/pci_ids.h | 2 +
tools/power/cpupower/utils/cpufreq-info.c | 6 +-
tools/power/cpupower/utils/helpers/amd.c | 6 +-
tools/power/cpupower/utils/helpers/cpuid.c | 8 +-
tools/power/cpupower/utils/helpers/helpers.h | 2 +-
tools/power/cpupower/utils/helpers/misc.c | 3 +-
.../cpupower/utils/idle_monitor/mperf_monitor.c | 3 +-
46 files changed, 732 insertions(+), 69 deletions(-)
create mode 100644 arch/x86/kernel/cpu/hygon.c

--
2.7.4



2018-08-19 16:13:07

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 03/16] x86/mtrr: get MTRR number and support TOP_MEM2

Hygon CPU have a special magic MSR way to force WB for memory >4GB,
and also support TOP_MEM2. Therefore, it is necessary to add Hygon
support in amd_special_default_mtrr().

The MtrrFixDramModEn bit on Hygon platform should also be set to 1
during BIOS initialization of the fixed MTRRs, then cleared to 0 for
operation.

The number of variable MTRRs for Hygon is 2 as AMD's.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/kernel/cpu/mtrr/cleanup.c | 3 ++-
arch/x86/kernel/cpu/mtrr/generic.c | 5 +++--
arch/x86/kernel/cpu/mtrr/mtrr.c | 2 +-
3 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kernel/cpu/mtrr/cleanup.c b/arch/x86/kernel/cpu/mtrr/cleanup.c
index 765afd5..3668c5d 100644
--- a/arch/x86/kernel/cpu/mtrr/cleanup.c
+++ b/arch/x86/kernel/cpu/mtrr/cleanup.c
@@ -831,7 +831,8 @@ int __init amd_special_default_mtrr(void)
{
u32 l, h;

- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
+ boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
return 0;
if (boot_cpu_data.x86 < 0xf)
return 0;
diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index e12ee86..77c3eaa 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -49,8 +49,9 @@ static inline void k8_check_syscfg_dram_mod_en(void)
{
u32 lo, hi;

- if (!((boot_cpu_data.x86_vendor == X86_VENDOR_AMD) &&
- (boot_cpu_data.x86 >= 0x0f)))
+ if (!((boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
+ boot_cpu_data.x86 >= 0x0f) ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON))
return;

rdmsr(MSR_K8_SYSCFG, lo, hi);
diff --git a/arch/x86/kernel/cpu/mtrr/mtrr.c b/arch/x86/kernel/cpu/mtrr/mtrr.c
index 9a19c80..507039c 100644
--- a/arch/x86/kernel/cpu/mtrr/mtrr.c
+++ b/arch/x86/kernel/cpu/mtrr/mtrr.c
@@ -127,7 +127,7 @@ static void __init set_num_var_ranges(void)

if (use_intel())
rdmsr(MSR_MTRRcap, config, dummy);
- else if (is_cpu(AMD))
+ else if (is_cpu(AMD) || is_cpu(HYGON))
config = 2;
else if (is_cpu(CYRIX) || is_cpu(CENTAUR))
config = 8;
--
2.7.4


2018-08-19 16:13:07

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 02/16] x86/cache: get cache size/leaves and setup cache cpumap for Dhyana

Hygon Dhyana processor has the topology extensions bit in CPUID.
With this bit kernel can get the cache info. So add support
in cpuid4_cache_lookup_regs() to get the correct cache size.

Dhyana also find num_cache_leaves via CPUID leaf 0x8000001d, so
add Hygon support in find_num_cache_leaves().

Also add cacheinfo_hygon_init_llc_id() and init_hygon_cacheinfo()
functions to initialize Dhyana cache info. Setup cache cpumap in
the same way as AMD does.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/include/asm/cacheinfo.h | 1 +
arch/x86/kernel/cpu/cacheinfo.c | 31 +++++++++++++++++++++++++++++--
arch/x86/kernel/cpu/cpu.h | 1 +
arch/x86/kernel/cpu/hygon.c | 3 +++
4 files changed, 34 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/cacheinfo.h b/arch/x86/include/asm/cacheinfo.h
index e958e28..86b63c7 100644
--- a/arch/x86/include/asm/cacheinfo.h
+++ b/arch/x86/include/asm/cacheinfo.h
@@ -3,5 +3,6 @@
#define _ASM_X86_CACHEINFO_H

void cacheinfo_amd_init_llc_id(struct cpuinfo_x86 *c, int cpu, u8 node_id);
+void cacheinfo_hygon_init_llc_id(struct cpuinfo_x86 *c, int cpu, u8 node_id);

#endif /* _ASM_X86_CACHEINFO_H */
diff --git a/arch/x86/kernel/cpu/cacheinfo.c b/arch/x86/kernel/cpu/cacheinfo.c
index 0c5fcbd..dc1b934 100644
--- a/arch/x86/kernel/cpu/cacheinfo.c
+++ b/arch/x86/kernel/cpu/cacheinfo.c
@@ -602,6 +602,10 @@ cpuid4_cache_lookup_regs(int index, struct _cpuid4_info_regs *this_leaf)
else
amd_cpuid4(index, &eax, &ebx, &ecx);
amd_init_l3_cache(this_leaf, index);
+ } else if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) {
+ cpuid_count(0x8000001d, index, &eax.full,
+ &ebx.full, &ecx.full, &edx);
+ amd_init_l3_cache(this_leaf, index);
} else {
cpuid_count(4, index, &eax.full, &ebx.full, &ecx.full, &edx);
}
@@ -625,7 +629,8 @@ static int find_num_cache_leaves(struct cpuinfo_x86 *c)
union _cpuid4_leaf_eax cache_eax;
int i = -1;

- if (c->x86_vendor == X86_VENDOR_AMD)
+ if (c->x86_vendor == X86_VENDOR_AMD ||
+ c->x86_vendor == X86_VENDOR_HYGON)
op = 0x8000001d;
else
op = 4;
@@ -678,6 +683,22 @@ void cacheinfo_amd_init_llc_id(struct cpuinfo_x86 *c, int cpu, u8 node_id)
}
}

+void cacheinfo_hygon_init_llc_id(struct cpuinfo_x86 *c, int cpu, u8 node_id)
+{
+ /*
+ * We may have multiple LLCs if L3 caches exist, so check if we
+ * have an L3 cache by looking at the L3 cache CPUID leaf.
+ */
+ if (!cpuid_edx(0x80000006))
+ return;
+
+ /*
+ * LLC is at the core complex level.
+ * Core complex ID is ApicId[3] for these processors.
+ */
+ per_cpu(cpu_llc_id, cpu) = c->apicid >> 3;
+}
+
void init_amd_cacheinfo(struct cpuinfo_x86 *c)
{

@@ -691,6 +712,11 @@ void init_amd_cacheinfo(struct cpuinfo_x86 *c)
}
}

+void init_hygon_cacheinfo(struct cpuinfo_x86 *c)
+{
+ num_cache_leaves = find_num_cache_leaves(c);
+}
+
void init_intel_cacheinfo(struct cpuinfo_x86 *c)
{
/* Cache sizes */
@@ -913,7 +939,8 @@ static void __cache_cpumap_setup(unsigned int cpu, int index,
int index_msb, i;
struct cpuinfo_x86 *c = &cpu_data(cpu);

- if (c->x86_vendor == X86_VENDOR_AMD) {
+ if (c->x86_vendor == X86_VENDOR_AMD ||
+ c->x86_vendor == X86_VENDOR_HYGON) {
if (__cache_amd_cpumap_setup(cpu, index, base))
return;
}
diff --git a/arch/x86/kernel/cpu/cpu.h b/arch/x86/kernel/cpu/cpu.h
index 7b229af..da5446a 100644
--- a/arch/x86/kernel/cpu/cpu.h
+++ b/arch/x86/kernel/cpu/cpu.h
@@ -54,6 +54,7 @@ extern u32 get_scattered_cpuid_leaf(unsigned int level,
enum cpuid_regs_idx reg);
extern void init_intel_cacheinfo(struct cpuinfo_x86 *c);
extern void init_amd_cacheinfo(struct cpuinfo_x86 *c);
+extern void init_hygon_cacheinfo(struct cpuinfo_x86 *c);

extern void detect_num_cpu_cores(struct cpuinfo_x86 *c);
extern int detect_extended_topology_early(struct cpuinfo_x86 *c);
diff --git a/arch/x86/kernel/cpu/hygon.c b/arch/x86/kernel/cpu/hygon.c
index 50bdce5..4e6d64e 100644
--- a/arch/x86/kernel/cpu/hygon.c
+++ b/arch/x86/kernel/cpu/hygon.c
@@ -90,6 +90,7 @@ static void hygon_get_topology(struct cpuinfo_x86 *c)
if (!err)
c->x86_coreid_bits = get_count_order(c->x86_max_cores);

+ cacheinfo_amd_init_llc_id(c, cpu, node_id);
} else if (cpu_has(c, X86_FEATURE_NODEID_MSR)) {
u64 value;

@@ -320,6 +321,8 @@ static void init_hygon(struct cpuinfo_x86 *c)
hygon_get_topology(c);
srat_detect_node(c);

+ init_hygon_cacheinfo(c);
+
set_cpu_cap(c, X86_FEATURE_K8);

if (cpu_has(c, X86_FEATURE_XMM2)) {
--
2.7.4


2018-08-19 16:13:07

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 04/16] x86/smpboot: smp init nodelay and no flush caches before sleep

Dhyana use no delay in smp_quirk_init_udelay(), and return in
mwait_play_dead() as AMD does.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/kernel/smpboot.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index f02ecaf..5369d7f 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -676,6 +676,7 @@ static void __init smp_quirk_init_udelay(void)

/* if modern processor, use no delay */
if (((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) && (boot_cpu_data.x86 == 6)) ||
+ ((boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) && (boot_cpu_data.x86 >= 0x18)) ||
((boot_cpu_data.x86_vendor == X86_VENDOR_AMD) && (boot_cpu_data.x86 >= 0xF))) {
init_udelay = 0;
return;
@@ -1592,7 +1593,8 @@ static inline void mwait_play_dead(void)
void *mwait_ptr;
int i;

- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
return;
if (!this_cpu_has(X86_FEATURE_MWAIT))
return;
--
2.7.4


2018-08-19 16:13:07

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 01/16] x86/cpu: create Dhyana init file and register new cpu_dev to system

Add x86 architecture support for new processor Hygon Dhyana Family 18h.
Rework to create a separated file(arch/x86/kernel/cpu/hygon.c) from the
AMD init one(arch/x86/kernel/cpu/amd.c) to initialize Dhyana CPU. In
this way we can remove old AMD architecture support codes from Hygon
code path and generate a clear initialization flow for Hygon processors.
It also reduce long-term maintenance effort.
Also add Maintainer information for hygon.c in accordance.

To identify Hygon processors, add a new vendor type X86_VENDOR_HYGON(9)
for system recognition.

To enable Hygon processor config, add a separated Kconfig entry
(CPU_SUP_HYGON) for Dhyana CPU in kernel config setup.

Signed-off-by: Pu Wen <[email protected]>
---
MAINTAINERS | 6 +
arch/x86/Kconfig.cpu | 13 ++
arch/x86/include/asm/processor.h | 3 +-
arch/x86/kernel/cpu/Makefile | 1 +
arch/x86/kernel/cpu/hygon.c | 406 +++++++++++++++++++++++++++++++++++++++
5 files changed, 428 insertions(+), 1 deletion(-)
create mode 100644 arch/x86/kernel/cpu/hygon.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 544cac8..26955c6 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -6605,6 +6605,12 @@ S: Maintained
F: mm/memory-failure.c
F: mm/hwpoison-inject.c

+HYGON PROCESSOR SUPPORT
+M: Pu Wen <[email protected]>
+L: [email protected]
+S: Supported
+F: arch/x86/kernel/cpu/hygon.c
+
Hyper-V CORE AND DRIVERS
M: "K. Y. Srinivasan" <[email protected]>
M: Haiyang Zhang <[email protected]>
diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu
index 638411f..2e92267 100644
--- a/arch/x86/Kconfig.cpu
+++ b/arch/x86/Kconfig.cpu
@@ -426,6 +426,19 @@ config CPU_SUP_AMD

If unsure, say N.

+config CPU_SUP_HYGON
+ default y
+ bool "Support Hygon processors" if PROCESSOR_SELECT
+ help
+ This enables detection, tunings and quirks for Hygon processors
+
+ You need this enabled if you want your kernel to run on an
+ Hygon CPU. Disabling this option on other types of CPUs
+ makes the kernel a tiny bit smaller. Disabling it on an Hygon
+ CPU might render the kernel unbootable.
+
+ If unsure, say N.
+
config CPU_SUP_CENTAUR
default y
bool "Support Centaur processors" if PROCESSOR_SELECT
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 79e4099..0fbdcc0 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -153,7 +153,8 @@ enum cpuid_regs_idx {
#define X86_VENDOR_CENTAUR 5
#define X86_VENDOR_TRANSMETA 7
#define X86_VENDOR_NSC 8
-#define X86_VENDOR_NUM 9
+#define X86_VENDOR_HYGON 9
+#define X86_VENDOR_NUM 10

#define X86_VENDOR_UNKNOWN 0xff

diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile
index 7a40196..889ce58 100644
--- a/arch/x86/kernel/cpu/Makefile
+++ b/arch/x86/kernel/cpu/Makefile
@@ -30,6 +30,7 @@ obj-$(CONFIG_X86_FEATURE_NAMES) += capflags.o powerflags.o

obj-$(CONFIG_CPU_SUP_INTEL) += intel.o intel_pconfig.o
obj-$(CONFIG_CPU_SUP_AMD) += amd.o
+obj-$(CONFIG_CPU_SUP_HYGON) += hygon.o
obj-$(CONFIG_CPU_SUP_CYRIX_32) += cyrix.o
obj-$(CONFIG_CPU_SUP_CENTAUR) += centaur.o
obj-$(CONFIG_CPU_SUP_TRANSMETA_32) += transmeta.o
diff --git a/arch/x86/kernel/cpu/hygon.c b/arch/x86/kernel/cpu/hygon.c
new file mode 100644
index 0000000..50bdce5
--- /dev/null
+++ b/arch/x86/kernel/cpu/hygon.c
@@ -0,0 +1,406 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Hygon Processor Support for Linux
+ *
+ * Copyright (c) Chengdu Haiguang IC Design Co., Ltd.
+ *
+ * Author: 2018 Pu Wen <[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General
+ * License v2.0 or later. See file COPYING for details.
+ */
+#include <linux/io.h>
+
+#include <asm/cpu.h>
+#include <asm/smp.h>
+#include <asm/cacheinfo.h>
+#include <asm/spec-ctrl.h>
+#include <asm/delay.h>
+#ifdef CONFIG_X86_64
+# include <asm/set_memory.h>
+#endif
+
+#include "cpu.h"
+
+/*
+ * nodes_per_socket: Stores the number of nodes per socket.
+ * Refer to CPUID Fn8000_001E_ECX Node Identifiers[10:8]
+ */
+static u32 nodes_per_socket = 1;
+
+#ifdef CONFIG_NUMA
+/*
+ * To workaround broken NUMA config. Read the comment in
+ * srat_detect_node().
+ */
+static int nearby_node(int apicid)
+{
+ int i, node;
+
+ for (i = apicid - 1; i >= 0; i--) {
+ node = __apicid_to_node[i];
+ if (node != NUMA_NO_NODE && node_online(node))
+ return node;
+ }
+ for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
+ node = __apicid_to_node[i];
+ if (node != NUMA_NO_NODE && node_online(node))
+ return node;
+ }
+ return first_node(node_online_map); /* Shouldn't happen */
+}
+#endif
+
+static void hygon_get_topology_early(struct cpuinfo_x86 *c)
+{
+ if (cpu_has(c, X86_FEATURE_TOPOEXT))
+ smp_num_siblings = ((cpuid_ebx(0x8000001e) >> 8) & 0xff) + 1;
+}
+
+/*
+ * Fixup core topology information for
+ * (1) Hygon multi-node processors
+ * Assumption: Number of cores in each internal node is the same.
+ * (2) Hygon processors supporting compute units
+ */
+static void hygon_get_topology(struct cpuinfo_x86 *c)
+{
+ u8 node_id;
+ int cpu = smp_processor_id();
+
+ /* get information required for multi-node processors */
+ if (boot_cpu_has(X86_FEATURE_TOPOEXT)) {
+ int err;
+ u32 eax, ebx, ecx, edx;
+
+ cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
+
+ node_id = ecx & 0xff;
+
+ c->cpu_core_id = ebx & 0xff;
+
+ if (smp_num_siblings > 1)
+ c->x86_max_cores /= smp_num_siblings;
+
+ /*
+ * In case leaf B is available, use it to derive
+ * topology information.
+ */
+ err = detect_extended_topology(c);
+ if (!err)
+ c->x86_coreid_bits = get_count_order(c->x86_max_cores);
+
+ } else if (cpu_has(c, X86_FEATURE_NODEID_MSR)) {
+ u64 value;
+
+ rdmsrl(MSR_FAM10H_NODE_ID, value);
+ node_id = value & 7;
+
+ per_cpu(cpu_llc_id, cpu) = node_id;
+ } else
+ return;
+
+ if (nodes_per_socket > 1)
+ set_cpu_cap(c, X86_FEATURE_AMD_DCM);
+}
+
+/*
+ * On Hygon setup the lower bits of the APIC id distinguish the cores.
+ * Assumes number of cores is a power of two.
+ */
+static void hygon_detect_cmp(struct cpuinfo_x86 *c)
+{
+ unsigned int bits;
+ int cpu = smp_processor_id();
+
+ bits = c->x86_coreid_bits;
+ /* Low order bits define the core id (index of core in socket) */
+ c->cpu_core_id = c->initial_apicid & ((1 << bits)-1);
+ /* Convert the initial APIC ID into the socket ID */
+ c->phys_proc_id = c->initial_apicid >> bits;
+ /* use socket ID also for last level cache */
+ per_cpu(cpu_llc_id, cpu) = c->phys_proc_id;
+}
+
+static void srat_detect_node(struct cpuinfo_x86 *c)
+{
+#ifdef CONFIG_NUMA
+ int cpu = smp_processor_id();
+ int node;
+ unsigned int apicid = c->apicid;
+
+ node = numa_cpu_node(cpu);
+ if (node == NUMA_NO_NODE)
+ node = per_cpu(cpu_llc_id, cpu);
+
+ /*
+ * On multi-fabric platform (e.g. Numascale NumaChip) a
+ * platform-specific handler needs to be called to fixup some
+ * IDs of the CPU.
+ */
+ if (x86_cpuinit.fixup_cpu_id)
+ x86_cpuinit.fixup_cpu_id(c, node);
+
+ if (!node_online(node)) {
+ /*
+ * Two possibilities here:
+ *
+ * - The CPU is missing memory and no node was created. In
+ * that case try picking one from a nearby CPU.
+ *
+ * - The APIC IDs differ from the HyperTransport node IDs.
+ * Assume they are all increased by a constant offset, but
+ * in the same order as the HT nodeids. If that doesn't
+ * result in a usable node fall back to the path for the
+ * previous case.
+ *
+ * This workaround operates directly on the mapping between
+ * APIC ID and NUMA node, assuming certain relationship
+ * between APIC ID, HT node ID and NUMA topology. As going
+ * through CPU mapping may alter the outcome, directly
+ * access __apicid_to_node[].
+ */
+ int ht_nodeid = c->initial_apicid;
+
+ if (__apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
+ node = __apicid_to_node[ht_nodeid];
+ /* Pick a nearby node */
+ if (!node_online(node))
+ node = nearby_node(apicid);
+ }
+ numa_set_node(cpu, node);
+#endif
+}
+
+static void early_init_hygon_mc(struct cpuinfo_x86 *c)
+{
+#ifdef CONFIG_SMP
+ unsigned int bits, ecx;
+
+ /* Multi core CPU? */
+ if (c->extended_cpuid_level < 0x80000008)
+ return;
+
+ ecx = cpuid_ecx(0x80000008);
+
+ c->x86_max_cores = (ecx & 0xff) + 1;
+
+ /* CPU telling us the core id bits shift? */
+ bits = (ecx >> 12) & 0xF;
+
+ /* Otherwise recompute */
+ if (bits == 0) {
+ while ((1 << bits) < c->x86_max_cores)
+ bits++;
+ }
+
+ c->x86_coreid_bits = bits;
+#endif
+}
+
+static void bsp_init_hygon(struct cpuinfo_x86 *c)
+{
+#ifdef CONFIG_X86_64
+ unsigned long long tseg;
+
+ /*
+ * Split up direct mapping around the TSEG SMM area.
+ * Don't do it for gbpages because there seems very little
+ * benefit in doing so.
+ */
+ if (!rdmsrl_safe(MSR_K8_TSEG_ADDR, &tseg)) {
+ unsigned long pfn = tseg >> PAGE_SHIFT;
+
+ pr_debug("tseg: %010llx\n", tseg);
+ if (pfn_range_is_mapped(pfn, pfn + 1))
+ set_memory_4k((unsigned long)__va(tseg), 1);
+ }
+#endif
+
+ if (cpu_has(c, X86_FEATURE_CONSTANT_TSC)) {
+ u64 val;
+
+ rdmsrl(MSR_K7_HWCR, val);
+ if (!(val & BIT(24)))
+ pr_warn(FW_BUG "TSC doesn't count with P0 frequency!\n");
+ }
+
+ if (cpu_has(c, X86_FEATURE_MWAITX))
+ use_mwaitx_delay();
+
+ if (boot_cpu_has(X86_FEATURE_TOPOEXT)) {
+ u32 ecx;
+
+ ecx = cpuid_ecx(0x8000001e);
+ nodes_per_socket = ((ecx >> 8) & 7) + 1;
+ } else if (boot_cpu_has(X86_FEATURE_NODEID_MSR)) {
+ u64 value;
+
+ rdmsrl(MSR_FAM10H_NODE_ID, value);
+ nodes_per_socket = ((value >> 3) & 7) + 1;
+ }
+
+ if (!boot_cpu_has(X86_FEATURE_AMD_SSBD) &&
+ !boot_cpu_has(X86_FEATURE_VIRT_SSBD)) {
+ /*
+ * Try to cache the base value so further operations can
+ * avoid RMW. If that faults, do not enable SSBD.
+ */
+ if (!rdmsrl_safe(MSR_AMD64_LS_CFG, &x86_amd_ls_cfg_base)) {
+ setup_force_cpu_cap(X86_FEATURE_LS_CFG_SSBD);
+ setup_force_cpu_cap(X86_FEATURE_SSBD);
+ x86_amd_ls_cfg_ssbd_mask = 1ULL << 10;
+ }
+ }
+}
+
+static void early_init_hygon(struct cpuinfo_x86 *c)
+{
+ u32 dummy;
+
+ early_init_hygon_mc(c);
+
+ rdmsr_safe(MSR_AMD64_PATCH_LEVEL, &c->microcode, &dummy);
+
+ /*
+ * c->x86_power is 8000_0007 edx. Bit 8 is TSC runs at constant rate
+ * with P/T states and does not stop in deep C-states
+ */
+ if (c->x86_power & (1 << 8)) {
+ set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
+ set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
+ }
+
+ /* Bit 12 of 8000_0007 edx is accumulated power mechanism. */
+ if (c->x86_power & BIT(12))
+ set_cpu_cap(c, X86_FEATURE_ACC_POWER);
+
+ set_cpu_cap(c, X86_FEATURE_SYSCALL32);
+
+#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PCI)
+ /*
+ * ApicID can always be treated as an 8-bit value for Hygon APIC So, we
+ * can safely set X86_FEATURE_EXTD_APICID unconditionally.
+ */
+ if (boot_cpu_has(X86_FEATURE_APIC))
+ set_cpu_cap(c, X86_FEATURE_EXTD_APICID);
+#endif
+
+ /*
+ * This is only needed to tell the kernel whether to use VMCALL
+ * and VMMCALL. VMMCALL is never executed except under virt, so
+ * we can set it unconditionally.
+ */
+ set_cpu_cap(c, X86_FEATURE_VMMCALL);
+
+ hygon_get_topology_early(c);
+}
+
+static void init_hygon(struct cpuinfo_x86 *c)
+{
+ early_init_hygon(c);
+
+ /*
+ * Bit 31 in normal CPUID used for nonstandard 3DNow ID;
+ * 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway
+ */
+ clear_cpu_cap(c, 0*32+31);
+
+ set_cpu_cap(c, X86_FEATURE_REP_GOOD);
+
+ /* get apicid instead of initial apic id from cpuid */
+ c->apicid = hard_smp_processor_id();
+
+ set_cpu_cap(c, X86_FEATURE_ZEN);
+ set_cpu_cap(c, X86_FEATURE_CPB);
+
+ cpu_detect_cache_sizes(c);
+
+ hygon_detect_cmp(c);
+ hygon_get_topology(c);
+ srat_detect_node(c);
+
+ set_cpu_cap(c, X86_FEATURE_K8);
+
+ if (cpu_has(c, X86_FEATURE_XMM2)) {
+ unsigned long long val;
+ int ret;
+
+ /*
+ * A serializing LFENCE has less overhead than MFENCE, so
+ * use it for execution serialization. On families which
+ * don't have that MSR, LFENCE is already serializing.
+ * msr_set_bit() uses the safe accessors, too, even if the MSR
+ * is not present.
+ */
+ msr_set_bit(MSR_F10H_DECFG,
+ MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT);
+
+ /*
+ * Verify that the MSR write was successful (could be running
+ * under a hypervisor) and only then assume that LFENCE is
+ * serializing.
+ */
+ ret = rdmsrl_safe(MSR_F10H_DECFG, &val);
+ if (!ret && (val & MSR_F10H_DECFG_LFENCE_SERIALIZE)) {
+ /* A serializing LFENCE stops RDTSC speculation */
+ set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC);
+ } else {
+ /* MFENCE stops RDTSC speculation */
+ set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC);
+ }
+ }
+
+ /*
+ * Hygon processors have APIC timer running in deep C states.
+ */
+ set_cpu_cap(c, X86_FEATURE_ARAT);
+
+ /* Hygon CPUs don't reset SS attributes on SYSRET, Xen does. */
+ if (!cpu_has(c, X86_FEATURE_XENPV))
+ set_cpu_bug(c, X86_BUG_SYSRET_SS_ATTRS);
+}
+
+static void cpu_detect_tlb_hygon(struct cpuinfo_x86 *c)
+{
+ u32 ebx, eax, ecx, edx;
+ u16 mask = 0xfff;
+
+ if (c->extended_cpuid_level < 0x80000006)
+ return;
+
+ cpuid(0x80000006, &eax, &ebx, &ecx, &edx);
+
+ tlb_lld_4k[ENTRIES] = (ebx >> 16) & mask;
+ tlb_lli_4k[ENTRIES] = ebx & mask;
+
+ /* Handle DTLB 2M and 4M sizes, fall back to L1 if L2 is disabled */
+ if (!((eax >> 16) & mask))
+ tlb_lld_2m[ENTRIES] = (cpuid_eax(0x80000005) >> 16) & 0xff;
+ else
+ tlb_lld_2m[ENTRIES] = (eax >> 16) & mask;
+
+ /* a 4M entry uses two 2M entries */
+ tlb_lld_4m[ENTRIES] = tlb_lld_2m[ENTRIES] >> 1;
+
+ /* Handle ITLB 2M and 4M sizes, fall back to L1 if L2 is disabled */
+ if (!(eax & mask)) {
+ cpuid(0x80000005, &eax, &ebx, &ecx, &edx);
+ tlb_lli_2m[ENTRIES] = eax & 0xff;
+ } else
+ tlb_lli_2m[ENTRIES] = eax & mask;
+
+ tlb_lli_4m[ENTRIES] = tlb_lli_2m[ENTRIES] >> 1;
+}
+
+static const struct cpu_dev hygon_cpu_dev = {
+ .c_vendor = "Hygon",
+ .c_ident = { "HygonGenuine" },
+ .c_early_init = early_init_hygon,
+ .c_detect_tlb = cpu_detect_tlb_hygon,
+ .c_bsp_init = bsp_init_hygon,
+ .c_init = init_hygon,
+ .c_x86_vendor = X86_VENDOR_HYGON,
+};
+
+cpu_dev_register(hygon_cpu_dev);
--
2.7.4


2018-08-19 16:13:39

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 06/16] x86/nops: init ideal_nops for Hygon

The ideal_nops for Dhyana processors should be p6_nops.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/kernel/alternative.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index a481763..8f4925b 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -222,6 +222,10 @@ void __init arch_init_ideal_nops(void)
}
break;

+ case X86_VENDOR_HYGON:
+ ideal_nops = p6_nops;
+ return;
+
case X86_VENDOR_AMD:
if (boot_cpu_data.x86 > 0xf) {
ideal_nops = p6_nops;
--
2.7.4


2018-08-19 16:13:39

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 08/16] x86/apic: add modern APIC support for Hygon

Hygon processors use modern APIC, so just return in modern_apic() and
sync_Arb_IDs(). And should break in switch case in detect_init_API().

When running on 32 bit mode, should set bigsmp if there are more than
8 cores.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/kernel/apic/apic.c | 19 ++++++++++++++-----
arch/x86/kernel/apic/probe_32.c | 1 +
2 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 3b3a2d0..9480ffa 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -220,10 +220,15 @@ static inline int lapic_is_integrated(void)
*/
static int modern_apic(void)
{
- /* AMD systems use old APIC versions, so check the CPU */
- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
- boot_cpu_data.x86 >= 0xf)
+ /*
+ * Old AMD systems use old APIC versions, newer AMD systems
+ * and Hygon systems use modern APIC, so check the CPU
+ */
+ if ((boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
+ boot_cpu_data.x86 >= 0xf) ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
return 1;
+
return lapic_get_version() >= 0x14;
}

@@ -1211,9 +1216,11 @@ void __init sync_Arb_IDs(void)
{
/*
* Unsupported on P4 - see Intel Dev. Manual Vol. 3, Ch. 8.6.1 And not
- * needed on AMD.
+ * needed on AMD or Hygon.
*/
- if (modern_apic() || boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ if (modern_apic() ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_AMD ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
return;

/*
@@ -1912,6 +1919,8 @@ static int __init detect_init_APIC(void)
(boot_cpu_data.x86 >= 15))
break;
goto no_apic;
+ case X86_VENDOR_HYGON:
+ break;
case X86_VENDOR_INTEL:
if (boot_cpu_data.x86 == 6 || boot_cpu_data.x86 == 15 ||
(boot_cpu_data.x86 == 5 && boot_cpu_has(X86_FEATURE_APIC)))
diff --git a/arch/x86/kernel/apic/probe_32.c b/arch/x86/kernel/apic/probe_32.c
index 02e8acb..47ff297 100644
--- a/arch/x86/kernel/apic/probe_32.c
+++ b/arch/x86/kernel/apic/probe_32.c
@@ -185,6 +185,7 @@ void __init default_setup_apic_routing(void)
break;
}
/* If P4 and above fall through */
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
def_to_bigsmp = 1;
}
--
2.7.4


2018-08-19 16:13:51

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 07/16] x86/pci: add Hygon PCI vendor and northbridge support

As Hygon register its PCI Vendor ID as a new one "0x1d94", so add a new
definition PCI_VENDOR_ID_HYGON in include/linux/pci_ids.h.

Also Hygon PCI Device ID(0x1450/0x1463/0x1464) for Host bridge is added
to amd_nb.c. And it need to define new arrays for Hygon:
hygon_root_ids[], hygon_nb_misc_ids[], hygon_nb_link_ids[].

To enable Hygon north bridge support, add new variable root_ids, and
assign its value based on whether CPU vendor is AMD or Hygon. Modify
the CONFIG_AMD_NB to depends on either AMD or Hygon.

Add Hygon support in amd_postcore_init(), early_root_info_init().

Acked-by: Bjorn Helgaas <[email protected]> # pci_ids.h
Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/Kconfig | 2 +-
arch/x86/kernel/amd_nb.c | 51 ++++++++++++++++++++++++++++++++++++++++++------
arch/x86/pci/amd_bus.c | 6 ++++--
include/linux/pci_ids.h | 2 ++
4 files changed, 52 insertions(+), 9 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 6b8065d..34d05cd 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -2823,7 +2823,7 @@ endif # X86_32

config AMD_NB
def_bool y
- depends on CPU_SUP_AMD && PCI
+ depends on (CPU_SUP_AMD || CPU_SUP_HYGON) && PCI

source "drivers/pcmcia/Kconfig"

diff --git a/arch/x86/kernel/amd_nb.c b/arch/x86/kernel/amd_nb.c
index b481b95..d9867b2 100644
--- a/arch/x86/kernel/amd_nb.c
+++ b/arch/x86/kernel/amd_nb.c
@@ -20,6 +20,10 @@
#define PCI_DEVICE_ID_AMD_17H_M10H_DF_F3 0x15eb
#define PCI_DEVICE_ID_AMD_17H_M10H_DF_F4 0x15ec

+#define PCI_DEVICE_ID_HYGON_18H_ROOT 0x1450
+#define PCI_DEVICE_ID_HYGON_18H_DF_F3 0x1463
+#define PCI_DEVICE_ID_HYGON_18H_DF_F4 0x1464
+
/* Protect the PCI config register pairs used for SMN and DF indirect access. */
static DEFINE_MUTEX(smn_mutex);

@@ -61,6 +65,21 @@ static const struct pci_device_id amd_nb_link_ids[] = {
{}
};

+static const struct pci_device_id hygon_root_ids[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_HYGON, PCI_DEVICE_ID_HYGON_18H_ROOT) },
+ {}
+};
+
+const struct pci_device_id hygon_nb_misc_ids[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_HYGON, PCI_DEVICE_ID_HYGON_18H_DF_F3) },
+ {}
+};
+
+static const struct pci_device_id hygon_nb_link_ids[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_HYGON, PCI_DEVICE_ID_HYGON_18H_DF_F4) },
+ {}
+};
+
const struct amd_nb_bus_dev_range amd_nb_bus_dev_ranges[] __initconst = {
{ 0x00, 0x18, 0x20 },
{ 0xff, 0x00, 0x20 },
@@ -197,12 +216,25 @@ int amd_cache_northbridges(void)
u16 i = 0;
struct amd_northbridge *nb;
struct pci_dev *root, *misc, *link;
+ const struct pci_device_id *root_ids = NULL;
+ const struct pci_device_id *misc_ids = NULL;
+ const struct pci_device_id *link_ids = NULL;
+
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
+ root_ids = amd_root_ids;
+ misc_ids = amd_nb_misc_ids;
+ link_ids = amd_nb_link_ids;
+ } else if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) {
+ root_ids = hygon_root_ids;
+ misc_ids = hygon_nb_misc_ids;
+ link_ids = hygon_nb_link_ids;
+ }

if (amd_northbridges.num)
return 0;

misc = NULL;
- while ((misc = next_northbridge(misc, amd_nb_misc_ids)) != NULL)
+ while ((misc = next_northbridge(misc, misc_ids)) != NULL)
i++;

if (!i)
@@ -218,11 +250,11 @@ int amd_cache_northbridges(void)
link = misc = root = NULL;
for (i = 0; i != amd_northbridges.num; i++) {
node_to_amd_nb(i)->root = root =
- next_northbridge(root, amd_root_ids);
+ next_northbridge(root, root_ids);
node_to_amd_nb(i)->misc = misc =
- next_northbridge(misc, amd_nb_misc_ids);
+ next_northbridge(misc, misc_ids);
node_to_amd_nb(i)->link = link =
- next_northbridge(link, amd_nb_link_ids);
+ next_northbridge(link, link_ids);
}

if (amd_gart_present())
@@ -263,9 +295,15 @@ bool __init early_is_amd_nb(u32 device)
{
const struct pci_device_id *id;
u32 vendor = device & 0xffff;
+ const struct pci_device_id *misc_ids = NULL;
+
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ misc_ids = amd_nb_misc_ids;
+ else if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
+ misc_ids = hygon_nb_misc_ids;

device >>= 16;
- for (id = amd_nb_misc_ids; id->vendor; id++)
+ for (id = misc_ids; id->vendor; id++)
if (vendor == id->vendor && device == id->device)
return true;
return false;
@@ -277,7 +315,8 @@ struct resource *amd_get_mmconfig_range(struct resource *res)
u64 base, msr;
unsigned int segn_busn_bits;

- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
+ boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
return NULL;

/* assume all cpus from fam10h have mmconfig */
diff --git a/arch/x86/pci/amd_bus.c b/arch/x86/pci/amd_bus.c
index 649bdde..bfa50e6 100644
--- a/arch/x86/pci/amd_bus.c
+++ b/arch/x86/pci/amd_bus.c
@@ -93,7 +93,8 @@ static int __init early_root_info_init(void)
vendor = id & 0xffff;
device = (id>>16) & 0xffff;

- if (vendor != PCI_VENDOR_ID_AMD)
+ if (vendor != PCI_VENDOR_ID_AMD &&
+ vendor != PCI_VENDOR_ID_HYGON)
continue;

if (hb_probes[i].device == device) {
@@ -390,7 +391,8 @@ static int __init pci_io_ecs_init(void)

static int __init amd_postcore_init(void)
{
- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
+ boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
return 0;

early_root_info_init();
diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h
index 2950223..87883cf 100644
--- a/include/linux/pci_ids.h
+++ b/include/linux/pci_ids.h
@@ -2559,6 +2559,8 @@

#define PCI_VENDOR_ID_AMAZON 0x1d0f

+#define PCI_VENDOR_ID_HYGON 0x1d94
+
#define PCI_VENDOR_ID_TEKRAM 0x1de1
#define PCI_DEVICE_ID_TEKRAM_DC290 0xdc29

--
2.7.4


2018-08-19 16:15:00

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 10/16] x86/mce: enable Hygon support to MCE infrastructure

Hygon machine check arch is similar to AMD family 17h. To enable the MCE
infrastructure support, add CPU vendor check for Hygon to share the code
path of AMD.

Add hygon mce init function mce_hygon_feature_init() to minimize further
maintenance effort.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/include/asm/mce.h | 5 +++++
arch/x86/kernel/cpu/mcheck/mce-severity.c | 3 ++-
arch/x86/kernel/cpu/mcheck/mce.c | 21 +++++++++++++++------
3 files changed, 22 insertions(+), 7 deletions(-)

diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index 8c7b3e5..0af3b0e 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -213,6 +213,11 @@ static inline void mce_amd_feature_init(struct cpuinfo_x86 *c) { }
static inline int umc_normaddr_to_sysaddr(u64 norm_addr, u16 nid, u8 umc, u64 *sys_addr) { return -EINVAL; };
#endif

+static inline void mce_hygon_feature_init(struct cpuinfo_x86 *c)
+{
+ return mce_amd_feature_init(c);
+}
+
int mce_available(struct cpuinfo_x86 *c);
bool mce_is_memory_error(struct mce *m);

diff --git a/arch/x86/kernel/cpu/mcheck/mce-severity.c b/arch/x86/kernel/cpu/mcheck/mce-severity.c
index f34d89c..44396d5 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-severity.c
+++ b/arch/x86/kernel/cpu/mcheck/mce-severity.c
@@ -336,7 +336,8 @@ int (*mce_severity)(struct mce *m, int tolerant, char **msg, bool is_excp) =

void __init mcheck_vendor_init_severity(void)
{
- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
mce_severity = mce_severity_amd;
}

diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 8c50754..65d13f8 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -274,7 +274,8 @@ static void print_mce(struct mce *m)
{
__print_mce(m);

- if (m->cpuvendor != X86_VENDOR_AMD)
+ if (m->cpuvendor != X86_VENDOR_AMD &&
+ m->cpuvendor != X86_VENDOR_HYGON)
pr_emerg_ratelimited(HW_ERR "Run the above through 'mcelog --ascii'\n");
}

@@ -512,9 +513,9 @@ static int mce_usable_address(struct mce *m)

bool mce_is_memory_error(struct mce *m)
{
- if (m->cpuvendor == X86_VENDOR_AMD) {
+ if (m->cpuvendor == X86_VENDOR_AMD ||
+ m->cpuvendor == X86_VENDOR_HYGON) {
return amd_mce_is_memory_error(m);
-
} else if (m->cpuvendor == X86_VENDOR_INTEL) {
/*
* Intel SDM Volume 3B - 15.9.2 Compound Error Codes
@@ -540,7 +541,9 @@ EXPORT_SYMBOL_GPL(mce_is_memory_error);

static bool mce_is_correctable(struct mce *m)
{
- if (m->cpuvendor == X86_VENDOR_AMD && m->status & MCI_STATUS_DEFERRED)
+ if ((m->cpuvendor == X86_VENDOR_AMD ||
+ m->cpuvendor == X86_VENDOR_HYGON) &&
+ (m->status & MCI_STATUS_DEFERRED))
return false;

if (m->status & MCI_STATUS_UC)
@@ -1725,7 +1728,8 @@ static int __mcheck_cpu_ancient_init(struct cpuinfo_x86 *c)
*/
static void __mcheck_cpu_init_early(struct cpuinfo_x86 *c)
{
- if (c->x86_vendor == X86_VENDOR_AMD) {
+ if (c->x86_vendor == X86_VENDOR_AMD ||
+ c->x86_vendor == X86_VENDOR_HYGON) {
mce_flags.overflow_recov = !!cpu_has(c, X86_FEATURE_OVERFLOW_RECOV);
mce_flags.succor = !!cpu_has(c, X86_FEATURE_SUCCOR);
mce_flags.smca = !!cpu_has(c, X86_FEATURE_SMCA);
@@ -1766,6 +1770,9 @@ static void __mcheck_cpu_init_vendor(struct cpuinfo_x86 *c)
mce_amd_feature_init(c);
break;
}
+ case X86_VENDOR_HYGON:
+ mce_hygon_feature_init(c);
+ break;
case X86_VENDOR_CENTAUR:
mce_centaur_feature_init(c);
break;
@@ -1991,12 +1998,14 @@ static void mce_disable_error_reporting(void)
static void vendor_disable_error_reporting(void)
{
/*
- * Don't clear on Intel or AMD CPUs. Some of these MSRs are socket-wide.
+ * Don't clear on Intel or AMD or Hygon CPUs. Some of these MSRs
+ * are socket-wide.
* Disabling them for just a single offlined CPU is bad, since it will
* inhibit reporting for all shared resources on the socket like the
* last level cache (LLC), the integrated memory controller (iMC), etc.
*/
if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON ||
boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
return;

--
2.7.4


2018-08-19 16:15:00

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 09/16] x86/bugs: add lfence mitigation to spectre v2 and no meltdown for Hygon

To share codes between AMD and Hygon to mitigate Spectre V2 Retpoline
vulnerability, rename macros SPECTRE_V2_RETPOLINE_MINIMAL_AMD to
SPECTRE_V2_RETPOLINE_MINIMAL_LFENCE, and SPECTRE_V2_CMD_RETPOLINE_AMD
to SPECTRE_V2_CMD_RETPOLINE_LFENCE.

As Hygon processors are not affected by meltdown vulnerability as AMD's,
so add exception in array cpu_no_meltdown[] for Hygon.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/include/asm/nospec-branch.h | 4 ++--
arch/x86/kernel/cpu/bugs.c | 28 +++++++++++++++-------------
arch/x86/kernel/cpu/common.c | 1 +
3 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index f6f6c63..aad6b9c 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -211,9 +211,9 @@
enum spectre_v2_mitigation {
SPECTRE_V2_NONE,
SPECTRE_V2_RETPOLINE_MINIMAL,
- SPECTRE_V2_RETPOLINE_MINIMAL_AMD,
+ SPECTRE_V2_RETPOLINE_MINIMAL_LFENCE,
SPECTRE_V2_RETPOLINE_GENERIC,
- SPECTRE_V2_RETPOLINE_AMD,
+ SPECTRE_V2_RETPOLINE_LFENCE,
SPECTRE_V2_IBRS,
};

diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
index 664f161..a663a1b 100644
--- a/arch/x86/kernel/cpu/bugs.c
+++ b/arch/x86/kernel/cpu/bugs.c
@@ -132,15 +132,15 @@ enum spectre_v2_mitigation_cmd {
SPECTRE_V2_CMD_FORCE,
SPECTRE_V2_CMD_RETPOLINE,
SPECTRE_V2_CMD_RETPOLINE_GENERIC,
- SPECTRE_V2_CMD_RETPOLINE_AMD,
+ SPECTRE_V2_CMD_RETPOLINE_LFENCE,
};

static const char *spectre_v2_strings[] = {
[SPECTRE_V2_NONE] = "Vulnerable",
[SPECTRE_V2_RETPOLINE_MINIMAL] = "Vulnerable: Minimal generic ASM retpoline",
- [SPECTRE_V2_RETPOLINE_MINIMAL_AMD] = "Vulnerable: Minimal AMD ASM retpoline",
+ [SPECTRE_V2_RETPOLINE_MINIMAL_LFENCE] = "Vulnerable: Minimal LFENCE ASM retpoline",
[SPECTRE_V2_RETPOLINE_GENERIC] = "Mitigation: Full generic retpoline",
- [SPECTRE_V2_RETPOLINE_AMD] = "Mitigation: Full AMD retpoline",
+ [SPECTRE_V2_RETPOLINE_LFENCE] = "Mitigation: Full LFENCE retpoline",
};

#undef pr_fmt
@@ -271,7 +271,7 @@ static const struct {
{ "off", SPECTRE_V2_CMD_NONE, false },
{ "on", SPECTRE_V2_CMD_FORCE, true },
{ "retpoline", SPECTRE_V2_CMD_RETPOLINE, false },
- { "retpoline,amd", SPECTRE_V2_CMD_RETPOLINE_AMD, false },
+ { "retpoline,lfence", SPECTRE_V2_CMD_RETPOLINE_LFENCE, false },
{ "retpoline,generic", SPECTRE_V2_CMD_RETPOLINE_GENERIC, false },
{ "auto", SPECTRE_V2_CMD_AUTO, false },
};
@@ -303,16 +303,17 @@ static enum spectre_v2_mitigation_cmd __init spectre_v2_parse_cmdline(void)
}

if ((cmd == SPECTRE_V2_CMD_RETPOLINE ||
- cmd == SPECTRE_V2_CMD_RETPOLINE_AMD ||
+ cmd == SPECTRE_V2_CMD_RETPOLINE_LFENCE ||
cmd == SPECTRE_V2_CMD_RETPOLINE_GENERIC) &&
!IS_ENABLED(CONFIG_RETPOLINE)) {
pr_err("%s selected but not compiled in. Switching to AUTO select\n", mitigation_options[i].option);
return SPECTRE_V2_CMD_AUTO;
}

- if (cmd == SPECTRE_V2_CMD_RETPOLINE_AMD &&
+ if (cmd == SPECTRE_V2_CMD_RETPOLINE_LFENCE &&
+ boot_cpu_data.x86_vendor != X86_VENDOR_HYGON &&
boot_cpu_data.x86_vendor != X86_VENDOR_AMD) {
- pr_err("retpoline,amd selected but CPU is not AMD. Switching to AUTO select\n");
+ pr_err("retpoline,lfence selected but CPU is not AMD or Hygon. Switching to AUTO select\n");
return SPECTRE_V2_CMD_AUTO;
}

@@ -346,9 +347,9 @@ static void __init spectre_v2_select_mitigation(void)
if (IS_ENABLED(CONFIG_RETPOLINE))
goto retpoline_auto;
break;
- case SPECTRE_V2_CMD_RETPOLINE_AMD:
+ case SPECTRE_V2_CMD_RETPOLINE_LFENCE:
if (IS_ENABLED(CONFIG_RETPOLINE))
- goto retpoline_amd;
+ goto retpoline_lfence;
break;
case SPECTRE_V2_CMD_RETPOLINE_GENERIC:
if (IS_ENABLED(CONFIG_RETPOLINE))
@@ -363,14 +364,15 @@ static void __init spectre_v2_select_mitigation(void)
return;

retpoline_auto:
- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
- retpoline_amd:
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD ||
+ boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) {
+ retpoline_lfence:
if (!boot_cpu_has(X86_FEATURE_LFENCE_RDTSC)) {
pr_err("Spectre mitigation: LFENCE not serializing, switching to generic retpoline\n");
goto retpoline_generic;
}
- mode = retp_compiler() ? SPECTRE_V2_RETPOLINE_AMD :
- SPECTRE_V2_RETPOLINE_MINIMAL_AMD;
+ mode = retp_compiler() ? SPECTRE_V2_RETPOLINE_LFENCE :
+ SPECTRE_V2_RETPOLINE_MINIMAL_LFENCE;
setup_force_cpu_cap(X86_FEATURE_RETPOLINE_AMD);
setup_force_cpu_cap(X86_FEATURE_RETPOLINE);
} else {
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index b41b72b..6c7a2cd 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -962,6 +962,7 @@ static const __initconst struct x86_cpu_id cpu_no_speculation[] = {

static const __initconst struct x86_cpu_id cpu_no_meltdown[] = {
{ X86_VENDOR_AMD },
+ { X86_VENDOR_HYGON },
{}
};

--
2.7.4


2018-08-19 16:15:29

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 11/16] x86/kvm: enable Hygon support to KVM infrastructure

Hygon Dhyana CPU has the SVM feature as AMD family 17h does.
Add Hygon support in the KVM infrastructure.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/include/asm/kvm_emulate.h | 4 ++++
arch/x86/include/asm/virtext.h | 5 +++--
arch/x86/kvm/emulate.c | 11 ++++++++++-
3 files changed, 17 insertions(+), 3 deletions(-)

diff --git a/arch/x86/include/asm/kvm_emulate.h b/arch/x86/include/asm/kvm_emulate.h
index 0f82cd9..93c4bf5 100644
--- a/arch/x86/include/asm/kvm_emulate.h
+++ b/arch/x86/include/asm/kvm_emulate.h
@@ -364,6 +364,10 @@ struct x86_emulate_ctxt {
#define X86EMUL_CPUID_VENDOR_AMDisbetterI_ecx 0x21726574
#define X86EMUL_CPUID_VENDOR_AMDisbetterI_edx 0x74656273

+#define X86EMUL_CPUID_VENDOR_HygonGenuine_ebx 0x6f677948
+#define X86EMUL_CPUID_VENDOR_HygonGenuine_ecx 0x656e6975
+#define X86EMUL_CPUID_VENDOR_HygonGenuine_edx 0x6e65476e
+
#define X86EMUL_CPUID_VENDOR_GenuineIntel_ebx 0x756e6547
#define X86EMUL_CPUID_VENDOR_GenuineIntel_ecx 0x6c65746e
#define X86EMUL_CPUID_VENDOR_GenuineIntel_edx 0x49656e69
diff --git a/arch/x86/include/asm/virtext.h b/arch/x86/include/asm/virtext.h
index 0116b2e..e05e0d3 100644
--- a/arch/x86/include/asm/virtext.h
+++ b/arch/x86/include/asm/virtext.h
@@ -83,9 +83,10 @@ static inline void cpu_emergency_vmxoff(void)
*/
static inline int cpu_has_svm(const char **msg)
{
- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD) {
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
+ boot_cpu_data.x86_vendor != X86_VENDOR_HYGON) {
if (msg)
- *msg = "not amd";
+ *msg = "not amd or hygon";
return 0;
}

diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index 4c4f426..1ab7758 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -2711,7 +2711,16 @@ static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
edx == X86EMUL_CPUID_VENDOR_AMDisbetterI_edx)
return true;

- /* default: (not Intel, not AMD), apply Intel's stricter rules... */
+ /* Hygon ("HygonGenuine") */
+ if (ebx == X86EMUL_CPUID_VENDOR_HygonGenuine_ebx &&
+ ecx == X86EMUL_CPUID_VENDOR_HygonGenuine_ecx &&
+ edx == X86EMUL_CPUID_VENDOR_HygonGenuine_edx)
+ return true;
+
+ /*
+ * default: (not Intel, not AMD, not Hygon), apply Intel's
+ * stricter rules...
+ */
return false;
}

--
2.7.4


2018-08-19 16:16:50

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 13/16] driver/acpi: enable Hygon support to ACPI driver

For Dhyana processors have NONSTOP TSC feature, so enable the support
to ACPI driver.

Acked-by: Rafael J. Wysocki <[email protected]>
Signed-off-by: Pu Wen <[email protected]>
---
drivers/acpi/acpi_pad.c | 1 +
drivers/acpi/processor_idle.c | 1 +
2 files changed, 2 insertions(+)

diff --git a/drivers/acpi/acpi_pad.c b/drivers/acpi/acpi_pad.c
index 552c1f7..a47676a 100644
--- a/drivers/acpi/acpi_pad.c
+++ b/drivers/acpi/acpi_pad.c
@@ -70,6 +70,7 @@ static void power_saving_mwait_init(void)

#if defined(CONFIG_X86)
switch (boot_cpu_data.x86_vendor) {
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
case X86_VENDOR_INTEL:
/*
diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index abb559c..b2131c4 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -205,6 +205,7 @@ static void lapic_timer_state_broadcast(struct acpi_processor *pr,
static void tsc_check_state(int state)
{
switch (boot_cpu_data.x86_vendor) {
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
case X86_VENDOR_INTEL:
case X86_VENDOR_CENTAUR:
--
2.7.4


2018-08-19 16:16:50

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 12/16] x86/xen: enable Hygon support to Xen

To make Xen works functionly on Hygon platforms, reuse AMD's Xen support
code path for Hygon.

There are six core performance events counters per thread, so there are
six MSRs for these counters(0-5). Also there are four legacy PMC MSRs,
they are alias of the counters(0-3).

In this version of kernel Hygon use the lagacy and safe versions of MSR
access. It works fine when VPMU enabled in Xen on Hygon platforms by
testing with perf.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/xen/pmu.c | 12 +++++++++---
1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/arch/x86/xen/pmu.c b/arch/x86/xen/pmu.c
index 7d00d4a..9403854 100644
--- a/arch/x86/xen/pmu.c
+++ b/arch/x86/xen/pmu.c
@@ -90,6 +90,12 @@ static void xen_pmu_arch_init(void)
k7_counters_mirrored = 0;
break;
}
+ } else if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) {
+ amd_num_counters = F10H_NUM_COUNTERS;
+ amd_counters_base = MSR_K7_PERFCTR0;
+ amd_ctrls_base = MSR_K7_EVNTSEL0;
+ amd_msr_step = 1;
+ k7_counters_mirrored = 0;
} else {
uint32_t eax, ebx, ecx, edx;

@@ -285,7 +291,7 @@ static bool xen_amd_pmu_emulate(unsigned int msr, u64 *val, bool is_read)

bool pmu_msr_read(unsigned int msr, uint64_t *val, int *err)
{
- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) {
if (is_amd_pmu_msr(msr)) {
if (!xen_amd_pmu_emulate(msr, val, 1))
*val = native_read_msr_safe(msr, err);
@@ -308,7 +314,7 @@ bool pmu_msr_write(unsigned int msr, uint32_t low, uint32_t high, int *err)
{
uint64_t val = ((uint64_t)high << 32) | low;

- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) {
if (is_amd_pmu_msr(msr)) {
if (!xen_amd_pmu_emulate(msr, &val, 0))
*err = native_write_msr_safe(msr, low, high);
@@ -379,7 +385,7 @@ static unsigned long long xen_intel_read_pmc(int counter)

unsigned long long xen_read_pmc(int counter)
{
- if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
return xen_amd_read_pmc(counter);
else
return xen_intel_read_pmc(counter);
--
2.7.4


2018-08-19 16:17:54

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 16/16] tools/cpupower: enable Hygon support to cpupower tool

Tool cpupower is useful to get CPU frequency information and monitor
power stats on Hygon platforms. So enable platform support to cpupower
for Hygon Dhyana Family 18h processors by checking vendor ID & family
and vendor string along with AMD.

Signed-off-by: Pu Wen <[email protected]>
---
tools/power/cpupower/utils/cpufreq-info.c | 6 ++++--
tools/power/cpupower/utils/helpers/amd.c | 6 ++++--
tools/power/cpupower/utils/helpers/cpuid.c | 8 +++++---
tools/power/cpupower/utils/helpers/helpers.h | 2 +-
tools/power/cpupower/utils/helpers/misc.c | 3 ++-
tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 3 ++-
6 files changed, 18 insertions(+), 10 deletions(-)

diff --git a/tools/power/cpupower/utils/cpufreq-info.c b/tools/power/cpupower/utils/cpufreq-info.c
index df43cd4..56e54ea 100644
--- a/tools/power/cpupower/utils/cpufreq-info.c
+++ b/tools/power/cpupower/utils/cpufreq-info.c
@@ -170,6 +170,7 @@ static int get_boost_mode(unsigned int cpu)
unsigned long pstates[MAX_HW_PSTATES] = {0,};

if (cpupower_cpu_info.vendor != X86_VENDOR_AMD &&
+ cpupower_cpu_info.vendor != X86_VENDOR_HYGON &&
cpupower_cpu_info.vendor != X86_VENDOR_INTEL)
return 0;

@@ -190,8 +191,9 @@ static int get_boost_mode(unsigned int cpu)
printf(_(" Supported: %s\n"), support ? _("yes") : _("no"));
printf(_(" Active: %s\n"), active ? _("yes") : _("no"));

- if (cpupower_cpu_info.vendor == X86_VENDOR_AMD &&
- cpupower_cpu_info.family >= 0x10) {
+ if ((cpupower_cpu_info.vendor == X86_VENDOR_AMD &&
+ cpupower_cpu_info.family >= 0x10) ||
+ cpupower_cpu_info.vendor == X86_VENDOR_HYGON) {
ret = decode_pstates(cpu, cpupower_cpu_info.family, b_states,
pstates, &pstate_no);
if (ret)
diff --git a/tools/power/cpupower/utils/helpers/amd.c b/tools/power/cpupower/utils/helpers/amd.c
index bb41cdd..d9327ee 100644
--- a/tools/power/cpupower/utils/helpers/amd.c
+++ b/tools/power/cpupower/utils/helpers/amd.c
@@ -45,7 +45,8 @@ static int get_did(int family, union msr_pstate pstate)

if (family == 0x12)
t = pstate.val & 0xf;
- else if (family == 0x17)
+ else if (family == 0x17 ||
+ (cpupower_cpu_info.vendor == X86_VENDOR_HYGON && family == 0x18))
t = pstate.fam17h_bits.did;
else
t = pstate.bits.did;
@@ -59,7 +60,8 @@ static int get_cof(int family, union msr_pstate pstate)
int fid, did, cof;

did = get_did(family, pstate);
- if (family == 0x17) {
+ if (family == 0x17 ||
+ (cpupower_cpu_info.vendor == X86_VENDOR_HYGON && family == 0x18)) {
fid = pstate.fam17h_bits.fid;
cof = 200 * fid / did;
} else {
diff --git a/tools/power/cpupower/utils/helpers/cpuid.c b/tools/power/cpupower/utils/helpers/cpuid.c
index 732b0b4..5cc39d4 100644
--- a/tools/power/cpupower/utils/helpers/cpuid.c
+++ b/tools/power/cpupower/utils/helpers/cpuid.c
@@ -8,7 +8,7 @@
#include "helpers/helpers.h"

static const char *cpu_vendor_table[X86_VENDOR_MAX] = {
- "Unknown", "GenuineIntel", "AuthenticAMD",
+ "Unknown", "GenuineIntel", "AuthenticAMD", "HygonGenuine",
};

#if defined(__i386__) || defined(__x86_64__)
@@ -109,6 +109,7 @@ int get_cpu_info(struct cpupower_cpu_info *cpu_info)
fclose(fp);
/* Get some useful CPU capabilities from cpuid */
if (cpu_info->vendor != X86_VENDOR_AMD &&
+ cpu_info->vendor != X86_VENDOR_HYGON &&
cpu_info->vendor != X86_VENDOR_INTEL)
return ret;

@@ -124,8 +125,9 @@ int get_cpu_info(struct cpupower_cpu_info *cpu_info)
if (cpuid_level >= 6 && (cpuid_ecx(6) & 0x1))
cpu_info->caps |= CPUPOWER_CAP_APERF;

- /* AMD Boost state enable/disable register */
- if (cpu_info->vendor == X86_VENDOR_AMD) {
+ /* AMD or Hygon Boost state enable/disable register */
+ if (cpu_info->vendor == X86_VENDOR_AMD ||
+ cpu_info->vendor == X86_VENDOR_HYGON) {
if (ext_cpuid_level >= 0x80000007 &&
(cpuid_edx(0x80000007) & (1 << 9)))
cpu_info->caps |= CPUPOWER_CAP_AMD_CBP;
diff --git a/tools/power/cpupower/utils/helpers/helpers.h b/tools/power/cpupower/utils/helpers/helpers.h
index 41da392..9021396 100644
--- a/tools/power/cpupower/utils/helpers/helpers.h
+++ b/tools/power/cpupower/utils/helpers/helpers.h
@@ -61,7 +61,7 @@ extern int be_verbose;

/* cpuid and cpuinfo helpers **************************/
enum cpupower_cpu_vendor {X86_VENDOR_UNKNOWN = 0, X86_VENDOR_INTEL,
- X86_VENDOR_AMD, X86_VENDOR_MAX};
+ X86_VENDOR_AMD, X86_VENDOR_HYGON, X86_VENDOR_MAX};

#define CPUPOWER_CAP_INV_TSC 0x00000001
#define CPUPOWER_CAP_APERF 0x00000002
diff --git a/tools/power/cpupower/utils/helpers/misc.c b/tools/power/cpupower/utils/helpers/misc.c
index 80fdf55..596bbb4 100644
--- a/tools/power/cpupower/utils/helpers/misc.c
+++ b/tools/power/cpupower/utils/helpers/misc.c
@@ -26,7 +26,8 @@ int cpufreq_has_boost_support(unsigned int cpu, int *support, int *active,
* has Hardware determined variable increments instead.
*/

- if (cpu_info.family == 0x17) {
+ if (cpu_info.family == 0x17 ||
+ (cpu_info.vendor == X86_VENDOR_HYGON && cpu_info.family == 0x18)) {
if (!read_msr(cpu, MSR_AMD_HWCR, &val)) {
if (!(val & CPUPOWER_AMD_CPBDIS))
*active = 1;
diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
index d7c2a6d..f2a7e9c 100644
--- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
@@ -241,7 +241,8 @@ static int init_maxfreq_mode(void)
if (!(cpupower_cpu_info.caps & CPUPOWER_CAP_INV_TSC))
goto use_sysfs;

- if (cpupower_cpu_info.vendor == X86_VENDOR_AMD) {
+ if (cpupower_cpu_info.vendor == X86_VENDOR_AMD ||
+ cpupower_cpu_info.vendor == X86_VENDOR_HYGON) {
/* MSR_AMD_HWCR tells us whether TSC runs at P0/mperf
* freq.
* A test whether hwcr is accessable/available would be:
--
2.7.4


2018-08-19 16:21:44

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 05/16] x86/pmu: enable Hygon support to PMU infrastructure

Hygon PMU arch is similar to AMD Family 17h. To support Hygon PMU, the
initialization flow for it just call amd_pmu_init() and change PMU name
to "HYGON". To share AMD's flow, add code check for Hygon family ID 18h
to run the code path of AMD family 17h in core/uncore functions.

Also it returns the bit offset of the performance counter register and
event selection register for Hygon CPU in the similar way as AMD does.

Signed-off-by: Pu Wen <[email protected]>
---
arch/x86/events/amd/core.c | 6 ++++++
arch/x86/events/amd/uncore.c | 15 ++++++++++-----
arch/x86/events/core.c | 4 ++++
arch/x86/kernel/cpu/perfctr-watchdog.c | 2 ++
4 files changed, 22 insertions(+), 5 deletions(-)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index c84584b..6c13c9d 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -669,6 +669,12 @@ static int __init amd_core_pmu_init(void)
* We fallback to using default amd_get_event_constraints.
*/
break;
+ case 0x18:
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) {
+ pr_cont("Fam18h ");
+ /* Using default amd_get_event_constraints. */
+ break;
+ }
default:
pr_err("core perfctr but no constraints; unknown hardware!\n");
return -ENODEV;
diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
index 981ba5e..9f2eb43 100644
--- a/arch/x86/events/amd/uncore.c
+++ b/arch/x86/events/amd/uncore.c
@@ -507,17 +507,22 @@ static int __init amd_uncore_init(void)
{
int ret = -ENODEV;

- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
+ boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
return -ENODEV;

if (!boot_cpu_has(X86_FEATURE_TOPOEXT))
return -ENODEV;

- if (boot_cpu_data.x86 == 0x17) {
+ if ((boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
+ boot_cpu_data.x86 == 0x17) ||
+ (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON &&
+ boot_cpu_data.x86 == 0x18)) {
/*
- * For F17h, the Northbridge counters are repurposed as Data
- * Fabric counters. Also, L3 counters are supported too. The PMUs
- * are exported based on family as either L2 or L3 and NB or DF.
+ * For AMD F17h or Hygon F18h, the Northbridge counters are
+ * repurposed as DataFabric counters. Also, L3 counters
+ * are supported too. The PMUs are exported based on
+ * family as either L2 or L3 and NB or DF.
*/
num_counters_nb = NUM_COUNTERS_NB;
num_counters_llc = NUM_COUNTERS_L3;
diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index 5f4829f..93e026b 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -1776,6 +1776,10 @@ static int __init init_hw_perf_events(void)
case X86_VENDOR_AMD:
err = amd_pmu_init();
break;
+ case X86_VENDOR_HYGON:
+ err = amd_pmu_init();
+ x86_pmu.name = "HYGON";
+ break;
default:
err = -ENOTSUPP;
}
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c
index d389083..9556930 100644
--- a/arch/x86/kernel/cpu/perfctr-watchdog.c
+++ b/arch/x86/kernel/cpu/perfctr-watchdog.c
@@ -46,6 +46,7 @@ static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr)
{
/* returns the bit offset of the performance counter register */
switch (boot_cpu_data.x86_vendor) {
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
if (msr >= MSR_F15H_PERF_CTR)
return (msr - MSR_F15H_PERF_CTR) >> 1;
@@ -74,6 +75,7 @@ static inline unsigned int nmi_evntsel_msr_to_bit(unsigned int msr)
{
/* returns the bit offset of the event selection register */
switch (boot_cpu_data.x86_vendor) {
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
if (msr >= MSR_F15H_PERF_CTL)
return (msr - MSR_F15H_PERF_CTL) >> 1;
--
2.7.4


2018-08-19 16:24:10

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

To make AMD64 EDAC and MCE drivers working on Hygon platforms, add
vendor checking for Hygon by using the code path of AMD 0x17. Add a
vendor field to struct amd64_pvt and initialize it in per_family_init
for vendor checking.

Also Hygon PCI Device ID DF_F0/DF_F6(0x1460/0x1466) of Host bridges
is needed for edac driver.

Signed-off-by: Pu Wen <[email protected]>
---
drivers/edac/amd64_edac.c | 52 ++++++++++++++++++++++++++++++++++++++++++++++-
drivers/edac/amd64_edac.h | 5 +++++
drivers/edac/mce_amd.c | 15 +++++++++++++-
3 files changed, 70 insertions(+), 2 deletions(-)

diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
index 18aeabb..3cefb25 100644
--- a/drivers/edac/amd64_edac.c
+++ b/drivers/edac/amd64_edac.c
@@ -211,7 +211,8 @@ static int __set_scrub_rate(struct amd64_pvt *pvt, u32 new_bw, u32 min_rate)

scrubval = scrubrates[i].scrubval;

- if (pvt->fam == 0x17) {
+ if (pvt->fam == 0x17 ||
+ (pvt->vendor == X86_VENDOR_HYGON && pvt->fam == 0x18)) {
__f17h_set_scrubval(pvt, scrubval);
} else if (pvt->fam == 0x15 && pvt->model == 0x60) {
f15h_select_dct(pvt, 0);
@@ -274,6 +275,21 @@ static int get_scrub_rate(struct mem_ctl_info *mci)
}
break;

+ case 0x18:
+ if (pvt->vendor == X86_VENDOR_HYGON) {
+ amd64_read_pci_cfg(pvt->F6, F17H_SCR_BASE_ADDR, &scrubval);
+ if (scrubval & BIT(0)) {
+ amd64_read_pci_cfg(pvt->F6, F17H_SCR_LIMIT_ADDR, &scrubval);
+ scrubval &= 0xF;
+ scrubval += 0x5;
+ } else {
+ scrubval = 0;
+ }
+ break;
+ }
+
+ /* fall through */
+
default:
amd64_read_pci_cfg(pvt->F3, SCRCTRL, &scrubval);
break;
@@ -1052,6 +1068,19 @@ static void determine_memory_type(struct amd64_pvt *pvt)
pvt->dram_type = MEM_DDR4;
return;

+ case 0x18:
+ if (pvt->vendor == X86_VENDOR_HYGON) {
+ if ((pvt->umc[0].dimm_cfg | pvt->umc[1].dimm_cfg) & BIT(5))
+ pvt->dram_type = MEM_LRDDR4;
+ else if ((pvt->umc[0].dimm_cfg | pvt->umc[1].dimm_cfg) & BIT(4))
+ pvt->dram_type = MEM_RDDR4;
+ else
+ pvt->dram_type = MEM_DDR4;
+ return;
+ }
+
+ /* fall through */
+
default:
WARN(1, KERN_ERR "%s: Family??? 0x%x\n", __func__, pvt->fam);
pvt->dram_type = MEM_EMPTY;
@@ -2200,6 +2229,16 @@ static struct amd64_family_type family_types[] = {
.dbam_to_cs = f17_base_addr_to_cs_size,
}
},
+ [HYGON_F18_CPUS] = {
+ /* Hygon F18h uses the same AMD F17h support */
+ .ctl_name = "Hygon_F18h",
+ .f0_id = PCI_DEVICE_ID_HYGON_18H_DF_F0,
+ .f6_id = PCI_DEVICE_ID_HYGON_18H_DF_F6,
+ .ops = {
+ .early_channel_count = f17_early_channel_count,
+ .dbam_to_cs = f17_base_addr_to_cs_size,
+ }
+ },
};

/*
@@ -3149,6 +3188,7 @@ static struct amd64_family_type *per_family_init(struct amd64_pvt *pvt)
pvt->ext_model = boot_cpu_data.x86_model >> 4;
pvt->stepping = boot_cpu_data.x86_stepping;
pvt->model = boot_cpu_data.x86_model;
+ pvt->vendor = boot_cpu_data.x86_vendor;
pvt->fam = boot_cpu_data.x86;

switch (pvt->fam) {
@@ -3192,6 +3232,15 @@ static struct amd64_family_type *per_family_init(struct amd64_pvt *pvt)
pvt->ops = &family_types[F17_CPUS].ops;
break;

+ case 0x18:
+ if (pvt->vendor == X86_VENDOR_HYGON) {
+ fam_type = &family_types[HYGON_F18_CPUS];
+ pvt->ops = &family_types[HYGON_F18_CPUS].ops;
+ break;
+ }
+
+ /* fall through */
+
default:
amd64_err("Unsupported family!\n");
return NULL;
@@ -3428,6 +3477,7 @@ static const struct x86_cpu_id amd64_cpuids[] = {
{ X86_VENDOR_AMD, 0x15, X86_MODEL_ANY, X86_FEATURE_ANY, 0 },
{ X86_VENDOR_AMD, 0x16, X86_MODEL_ANY, X86_FEATURE_ANY, 0 },
{ X86_VENDOR_AMD, 0x17, X86_MODEL_ANY, X86_FEATURE_ANY, 0 },
+ { X86_VENDOR_HYGON, 0x18, X86_MODEL_ANY, X86_FEATURE_ANY, 0 },
{ }
};
MODULE_DEVICE_TABLE(x86cpu, amd64_cpuids);
diff --git a/drivers/edac/amd64_edac.h b/drivers/edac/amd64_edac.h
index 1d4b74e..5176b51 100644
--- a/drivers/edac/amd64_edac.h
+++ b/drivers/edac/amd64_edac.h
@@ -116,6 +116,9 @@
#define PCI_DEVICE_ID_AMD_17H_DF_F0 0x1460
#define PCI_DEVICE_ID_AMD_17H_DF_F6 0x1466

+#define PCI_DEVICE_ID_HYGON_18H_DF_F0 0x1460
+#define PCI_DEVICE_ID_HYGON_18H_DF_F6 0x1466
+
/*
* Function 1 - Address Map
*/
@@ -281,6 +284,7 @@ enum amd_families {
F16_CPUS,
F16_M30H_CPUS,
F17_CPUS,
+ HYGON_F18_CPUS,
NUM_FAMILIES,
};

@@ -328,6 +332,7 @@ struct amd64_pvt {
struct pci_dev *F0, *F1, *F2, *F3, *F6;

u16 mc_node_id; /* MC index of this MC node */
+ u8 vendor; /* CPU vendor */
u8 fam; /* CPU family */
u8 model; /* ... model */
u8 stepping; /* ... stepping */
diff --git a/drivers/edac/mce_amd.c b/drivers/edac/mce_amd.c
index 2ab4d61..886e2bb 100644
--- a/drivers/edac/mce_amd.c
+++ b/drivers/edac/mce_amd.c
@@ -1059,7 +1059,8 @@ static int __init mce_amd_init(void)
{
struct cpuinfo_x86 *c = &boot_cpu_data;

- if (c->x86_vendor != X86_VENDOR_AMD)
+ if (c->x86_vendor != X86_VENDOR_AMD &&
+ c->x86_vendor != X86_VENDOR_HYGON)
return -ENODEV;

fam_ops = kzalloc(sizeof(struct amd_decoder_ops), GFP_KERNEL);
@@ -1120,6 +1121,18 @@ static int __init mce_amd_init(void)
}
break;

+ case 0x18:
+ if (c->x86_vendor == X86_VENDOR_HYGON) {
+ xec_mask = 0x3f;
+ if (!boot_cpu_has(X86_FEATURE_SMCA)) {
+ pr_warn("Decoding supported only on Scalable MCA processors.\n");
+ goto err_out;
+ }
+ break;
+ }
+
+ /* fall through */
+
default:
printk(KERN_WARNING "Huh? What family is it: 0x%x?!\n", c->x86);
goto err_out;
--
2.7.4


2018-08-19 16:58:31

by Wen Pu

[permalink] [raw]
Subject: [PATCH v4 14/16] driver/cpufreq: enable Hygon support to cpufreq driver

Enable ACPI cpufreq driver support for Hygon by adding family ID check
along with AMD.

As Hygon platforms have SMBus device(PCI device ID 0x790b), enable Hygon
support to function amd_freq_sensitivity_init().

Acked-by: Rafael J. Wysocki <[email protected]>
Signed-off-by: Pu Wen <[email protected]>
---
drivers/cpufreq/acpi-cpufreq.c | 5 +++++
drivers/cpufreq/amd_freq_sensitivity.c | 9 +++++++--
2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
index b61f4ec..d62fd37 100644
--- a/drivers/cpufreq/acpi-cpufreq.c
+++ b/drivers/cpufreq/acpi-cpufreq.c
@@ -61,6 +61,7 @@ enum {

#define INTEL_MSR_RANGE (0xffff)
#define AMD_MSR_RANGE (0x7)
+#define HYGON_MSR_RANGE (0x7)

#define MSR_K7_HWCR_CPB_DIS (1ULL << 25)

@@ -95,6 +96,7 @@ static bool boost_state(unsigned int cpu)
rdmsr_on_cpu(cpu, MSR_IA32_MISC_ENABLE, &lo, &hi);
msr = lo | ((u64)hi << 32);
return !(msr & MSR_IA32_MISC_ENABLE_TURBO_DISABLE);
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
rdmsr_on_cpu(cpu, MSR_K7_HWCR, &lo, &hi);
msr = lo | ((u64)hi << 32);
@@ -113,6 +115,7 @@ static int boost_set_msr(bool enable)
msr_addr = MSR_IA32_MISC_ENABLE;
msr_mask = MSR_IA32_MISC_ENABLE_TURBO_DISABLE;
break;
+ case X86_VENDOR_HYGON:
case X86_VENDOR_AMD:
msr_addr = MSR_K7_HWCR;
msr_mask = MSR_K7_HWCR_CPB_DIS;
@@ -225,6 +228,8 @@ static unsigned extract_msr(struct cpufreq_policy *policy, u32 msr)

if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
msr &= AMD_MSR_RANGE;
+ else if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
+ msr &= HYGON_MSR_RANGE;
else
msr &= INTEL_MSR_RANGE;

diff --git a/drivers/cpufreq/amd_freq_sensitivity.c b/drivers/cpufreq/amd_freq_sensitivity.c
index be926d9..4ac7c3c 100644
--- a/drivers/cpufreq/amd_freq_sensitivity.c
+++ b/drivers/cpufreq/amd_freq_sensitivity.c
@@ -111,11 +111,16 @@ static int __init amd_freq_sensitivity_init(void)
{
u64 val;
struct pci_dev *pcidev;
+ unsigned int pci_vendor;

- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
+ if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ pci_vendor = PCI_VENDOR_ID_AMD;
+ else if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
+ pci_vendor = PCI_VENDOR_ID_HYGON;
+ else
return -ENODEV;

- pcidev = pci_get_device(PCI_VENDOR_ID_AMD,
+ pcidev = pci_get_device(pci_vendor,
PCI_DEVICE_ID_AMD_KERNCZ_SMBUS, NULL);

if (!pcidev) {
--
2.7.4


2018-08-21 01:44:41

by Boris Ostrovsky

[permalink] [raw]
Subject: Re: [Xen-devel] [PATCH v4 12/16] x86/xen: enable Hygon support to Xen

On 08/19/2018 12:13 PM, Pu Wen wrote:
> To make Xen works functionly on Hygon platforms, reuse AMD's Xen support
> code path for Hygon.
>
> There are six core performance events counters per thread, so there are
> six MSRs for these counters(0-5). Also there are four legacy PMC MSRs,
> they are alias of the counters(0-3).
>
> In this version of kernel Hygon use the lagacy and safe versions of MSR
> access. It works fine when VPMU enabled in Xen on Hygon platforms by
> testing with perf.
>
> Signed-off-by: Pu Wen <[email protected]>

Reviewed-by: Boris Ostrovsky <[email protected]>



2018-08-21 08:15:24

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On Mon, Aug 20, 2018 at 12:14:58AM +0800, Pu Wen wrote:
> To make AMD64 EDAC and MCE drivers working on Hygon platforms, add
> vendor checking for Hygon by using the code path of AMD 0x17. Add a
> vendor field to struct amd64_pvt and initialize it in per_family_init
> for vendor checking.
>
> Also Hygon PCI Device ID DF_F0/DF_F6(0x1460/0x1466) of Host bridges
> is needed for edac driver.
>
> Signed-off-by: Pu Wen <[email protected]>
> ---
> drivers/edac/amd64_edac.c | 52 ++++++++++++++++++++++++++++++++++++++++++++++-
> drivers/edac/amd64_edac.h | 5 +++++
> drivers/edac/mce_amd.c | 15 +++++++++++++-
> 3 files changed, 70 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
> index 18aeabb..3cefb25 100644
> --- a/drivers/edac/amd64_edac.c
> +++ b/drivers/edac/amd64_edac.c
> @@ -211,7 +211,8 @@ static int __set_scrub_rate(struct amd64_pvt *pvt, u32 new_bw, u32 min_rate)
>
> scrubval = scrubrates[i].scrubval;
>
> - if (pvt->fam == 0x17) {
> + if (pvt->fam == 0x17 ||
> + (pvt->vendor == X86_VENDOR_HYGON && pvt->fam == 0x18)) {

Didn't we say that AMD won't do family 0x18 so you don't need the vendor check?

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.
--

2018-08-21 11:31:20

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On Tue, Aug 21, 2018 at 07:04:23PM +0800, Pu Wen wrote:
> Sure, JV will negotiate with AMD and make sure only JV use family 18h and
> AMD won't use family 0x18h, which will make the patch tight and clear.
>
> What's the best way to adapt for EDAC driver?
> * To simplify the code based on AMD won't do family 0x18 and remove
> vendor checking.

That, provided AMD won't use family 0x18.

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.
--

2018-08-21 11:32:35

by Paolo Bonzini

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On 21/08/2018 13:20, Borislav Petkov wrote:
> On Tue, Aug 21, 2018 at 07:04:23PM +0800, Pu Wen wrote:
>> Sure, JV will negotiate with AMD and make sure only JV use family 18h and
>> AMD won't use family 0x18h, which will make the patch tight and clear.
>>
>> What's the best way to adapt for EDAC driver?
>> * To simplify the code based on AMD won't do family 0x18 and remove
>> vendor checking.
>
> That, provided AMD won't use family 0x18.
>

But then I don't see the point of adding the Hygon vendor, since any
check can be simplified:

(AMD || HYGON) -->
AMD

(AMD && fam = 17h) || (HYGON && fam = 18h) -->
AMD && (fam = 17h || fam = 18h)

(AMD && fam = 17h) || HYGON -->
AMD && (fam = 17h || fam = 18h)

etc. Anyway, I don't care much either way.

Paolo

2018-08-21 11:39:01

by Wen Pu

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On 2018-08-21 16:13, Borislav Petkov wrote:
>On Mon, Aug 20, 2018 at 12:14:58AM +0800, Pu Wen wrote:
...
>> diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
>> index 18aeabb..3cefb25 100644
>> --- a/drivers/edac/amd64_edac.c
>> +++ b/drivers/edac/amd64_edac.c
>> @@ -211,7 +211,8 @@ static int __set_scrub_rate(struct amd64_pvt
*pvt, u32 new_bw, u32 min_rate)
>>
>> scrubval = scrubrates[i].scrubval;
>>
>> - if (pvt->fam == 0x17) {
>> + if (pvt->fam == 0x17 ||
>> + (pvt->vendor == X86_VENDOR_HYGON && pvt->fam == 0x18)) {
>
>Didn't we say that AMD won't do family 0x18 so you don't need the
vendor check?

Hi Boris:

Based on the feedback from Paolo Bonzini:
"if the x86 maintainers prefer to have a new X86_VENDOR_*
constant, I'd just ignore the fact that AMD will skip family 18h, and
introduce vendor checks along the lines below. This has the advantage
that it's not an issue if AMD ends up _not_ skipping family 18h."
http://lkml.iu.edu/hypermail/linux/kernel/1807.3/06042.html

and the suggestions from Michael Jin:
"The condition case for family 18h and vendor Hygon looks better
because it is more clear."
http://lkml.iu.edu/hypermail/linux/kernel/1808.1/03232.html

we reworked the patch.

Sure, JV will negotiate with AMD and make sure only JV use family 18h and
AMD won't use family 0x18h, which will make the patch tight and clear.

What's the best way to adapt for EDAC driver?
* To simplify the code based on AMD won't do family 0x18 and remove
vendor checking.
* Or add vendor checking in codes?

Thanks,
Pu Wen


2018-08-21 14:20:53

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On Tue, Aug 21, 2018 at 01:26:13PM +0200, Paolo Bonzini wrote:
> But then I don't see the point of adding the Hygon vendor, since any
> check can be simplified:

I think Hygon wanted to superficially show it is not really an AMD. For
example, the Hygon thing doesn't do SME/SEV. AFAIK.

So we can just as well check only family but I'd say the vendor thing is
laying the grounds for the future where reportedly it will differ more
from an AMD. And then we can start splitting code more based on vendor
and not look at family at all.

But for right now I think we should strive to keep the changes as small
as possible and only do real splitting when they start adding new
functionality. Which would mean having a hygon_edac.c too, for example.

All, IMHO, of course. Sharing code between vendors is always yucky.

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.
--

2018-08-21 19:06:20

by Pavel Machek

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

> On Tue, Aug 21, 2018 at 01:26:13PM +0200, Paolo Bonzini wrote:
> > But then I don't see the point of adding the Hygon vendor, since any
> > check can be simplified:
>
> I think Hygon wanted to superficially show it is not really an AMD. For
> example, the Hygon thing doesn't do SME/SEV. AFAIK.
>
> So we can just as well check only family but I'd say the vendor thing is
> laying the grounds for the future where reportedly it will differ more
> from an AMD. And then we can start splitting code more based on vendor
> and not look at family at all.
>
> But for right now I think we should strive to keep the changes as small
> as possible and only do real splitting when they start adding new
> functionality. Which would mean having a hygon_edac.c too, for example.
>
> All, IMHO, of course. Sharing code between vendors is always yucky.

Dunno, people do not know that 0x18 is reserved on AMD, so resulting
code is quite confusing.

Explicit vendor check is way to go, long term.
Pavel

--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

2018-08-22 13:21:21

by Wen Pu

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On 2018/8/22 2:07, Pavel Machek wrote>> But for right now I think we
should strive to keep the changes as small
>> as possible and only do real splitting when they start adding new
>> functionality. Which would mean having a hygon_edac.c too, for example.
>>
>> All, IMHO, of course. Sharing code between vendors is always yucky.
>
> Dunno, people do not know that 0x18 is reserved on AMD, so resulting
> code is quite confusing.
>
> Explicit vendor check is way to go, long term.
> Pavel

Hi Pavel,

For amd64_edac_init() entry, there will be a x86_match_cpu(amd64_cpuids)
which checking whether amd64_edac module fit current CPU. Only Hygon has
registered X86_VENDOR_HYGON with family 0x18h in amd64_cpuids[] to use
AMD edac codes.

Also with the assumption that Hygon will negotiate with AMD that only
Hygon will use Family 18h. So we assume in later codes vendor checking
can be omitted.

You are right, for short term, we are trying to keep the tight change and
enable Dhyana with minimal effort. For long term, agreed with that.

So based on the assumption that only Hygon will use family 0x18, we will
rework the patch, make the assumption clear and try to keep it small.
If that's ok, we will rework this patch.

Thanks for all the suggestions.

Regards,
Pu Wen


2018-08-22 13:59:12

by Wen Pu

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On 2018/8/21 21:04, Borislav Petkov wrote:
> On Tue, Aug 21, 2018 at 01:26:13PM +0200, Paolo Bonzini wrote:
>> But then I don't see the point of adding the Hygon vendor, since any
>> check can be simplified:
>
> I think Hygon wanted to superficially show it is not really an AMD. For
> example, the Hygon thing doesn't do SME/SEV. AFAIK.
>
> So we can just as well check only family but I'd say the vendor thing is
> laying the grounds for the future where reportedly it will differ more
> from an AMD. And then we can start splitting code more based on vendor
> and not look at family at all.
>
> But for right now I think we should strive to keep the changes as small
> as possible and only do real splitting when they start adding new
> functionality. Which would mean having a hygon_edac.c too, for example.
>
> All, IMHO, of course. Sharing code between vendors is always yucky.

Thanks for Boris's clarification.

It's hard to find the balance between short term(sharing codes) and long
term(splitting codes).
Current I think we'd better try to follow the following way:
For current version, we will try to minimize the modification and share
codes,
For later big modification, will try to split codes to make code path
more clear.

Regards
Pu Wen


2018-08-23 08:29:22

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v4 15/16] driver/edac: enable Hygon support to AMD64 EDAC driver

On Wed, Aug 22, 2018 at 09:07:20PM +0800, Pu Wen wrote:
> Current I think we'd better try to follow the following way:
> For current version, we will try to minimize the modification and share
> codes,
> For later big modification, will try to split codes to make code path
> more clear.

Yes, makes sense.

Thx.

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.
--