AuxXxilium/linux_dsm_epyc7002
1
0
Fork 0
mirror of https://github.com/AuxXxilium/linux_dsm_epyc7002.git synced 2024-12-28 11:18:45 +07:00
Code Issues Actions Packages Projects Releases Wiki Activity

AMD thresholding fixes for 3.6

Browse Source 
Those are a bunch of patches which give the MCE thresholding code a
 hard look and a scrubbing to remove a couple of annoyances like sysfs
 warnings when running CPU off-/online tests and the threshold_bank4 node
 under /sys/devices/system/machinecheck/ is a symlink.
 
 It also gives proper names to the thresholding banks instead of simply
 enumerating them, like this:
 
      /sys/devices/system/machinecheck/machinecheck0/
      |-- bank0
      |-- bank1
      |-- bank2
      |-- bank3
      |-- bank4
      |-- bank5
      |-- bank6
      |-- check_interval
      |-- cmci_disabled
      |-- combined_unit
      |   |-- combined_unit
      |       |-- error_count
      |       |-- threshold_limit
      |-- dont_log_ce
      |-- execution_unit
      |   |-- execution_unit
      |       |-- error_count
      |       |-- threshold_limit
      |-- ignore_ce
      |-- insn_fetch
      |   |-- insn_fetch
      |       |-- error_count
      |       |-- threshold_limit
      |-- load_store
      |   |-- load_store
      |       |-- error_count
      |       |-- threshold_limit
      |-- monarch_timeout
      |-- northbridge
      |   |-- dram
      |   |   |-- error_count
      |   |   |-- interrupt_enable
      |   |   |-- threshold_limit
      |   |-- ht_links
      |   |   |-- error_count
      |   |   |-- interrupt_enable
      |   |   |-- threshold_limit
      |   |-- l3_cache
      |       |-- error_count
      |       |-- interrupt_enable
      |       |-- threshold_limit
     ...
 
 It is tested on all our families >= K8.
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.4.12 (GNU/Linux)
 
 iQIcBAABAgAGBQJP0Jw9AAoJEBLB8Bhh3lVKMa8P/1ZPWkFZVFIdilyViQdSR/1/
 6MPy6BcZAACBl4rgrvjtFhmNv8C2dCGoPYRksHiO9sjgsilhQe/L92rmORifrNB4
 kvqR1QfKH2Hw2X1B/0fWXthh7UV37h1TdrVNJNlzhmi3wO+MHlX54iZcwpsaceFx
 QdzSqdHbaKfkfttojxIdgSfl7M2aCRnkmMOUG4X9HCsIK0C3ChdHLhJDnLT0xYb8
 fdA8dkXMktli0GC+KfevOXILZGLhUQuigu4iYKRm689N98N1Ejfa7BvMCVqLr0kF
 4fNmC+BtZmdw8MYd7EiuYXhA0Unu+CAg23ADQpn0AEyGQcM5h7/9/4GKvgjjsV1h
 /2r1WU+UVGZSUQ3FRDbzD37QVAa9FoOv967Gks6Fa31K7kEPC8yIRhWl72wXQXpa
 hFk+Hf3RlKtaO06iH/2RD2JA+W6xntiFo8CZ+AUMoLWfIQaYSAFP039lpjJp/Hzd
 CDdNWKCchAaMYI1MBmbRZ65mSgsVLLioNrf55+kdWT/CbuXJua95YxRRmllNFv5k
 MHjPoTajL0WKZhYxUSjCH87rqZHyNBH5s8iZlIt7wR//kqBGYfmRvGSDe31yMrL8
 PH/MgEIBVmrLQSWcojF+pU6ep+sQELVNsbu1+doZd/ruD/hzsZeu+MANWtJgrrVs
 +rsPRDWTcC3ca/V5Y1UO
 =XN3W
 -----END PGP SIGNATURE-----

Merge tag 'amd-thresholding-fixes-for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras into x86/mce

Pull in AMD MCE thresholding fixes for v3.6, from Borislav Petkov:

" Those are a bunch of patches which give the MCE thresholding code a
  hard look and a scrubbing to remove a couple of annoyances like sysfs
  warnings when running CPU off-/online tests and the threshold_bank4 node
  under /sys/devices/system/machinecheck/ is a symlink.

  It also gives proper names to the thresholding banks instead of simply
  enumerating them, like this:

     /sys/devices/system/machinecheck/machinecheck0/
     |-- bank0
     |-- bank1
     |-- bank2
     |-- bank3
     |-- bank4
     |-- bank5
     |-- bank6
     |-- check_interval
     |-- cmci_disabled
     |-- combined_unit
     |   |-- combined_unit
     |       |-- error_count
     |       |-- threshold_limit
     |-- dont_log_ce
     |-- execution_unit
     |   |-- execution_unit
     |       |-- error_count
     |       |-- threshold_limit
     |-- ignore_ce
     |-- insn_fetch
     |   |-- insn_fetch
     |       |-- error_count
     |       |-- threshold_limit
     |-- load_store
     |   |-- load_store
     |       |-- error_count
     |       |-- threshold_limit
     |-- monarch_timeout
     |-- northbridge
     |   |-- dram
     |   |   |-- error_count
     |   |   |-- interrupt_enable
     |   |   |-- threshold_limit
     |   |-- ht_links
     |   |   |-- error_count
     |   |   |-- interrupt_enable
     |   |   |-- threshold_limit
     |   |-- l3_cache
     |       |-- error_count
     |       |-- interrupt_enable
     |       |-- threshold_limit
    ...

  It is tested on all our families >= K8."

Signed-off-by: Ingo Molnar <mingo@kernel.org>
This commit is contained in:
Ingo Molnar 2012-06-08 12:29:47 +02:00
commit 707ecec1dc
5 changed files with 150 additions and 148 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
arch/x86/include/asm/amd_nb.h
View File

@ -26,10 +26,31 @@ struct amd_l3_cache {
u8 subcaches[4];
};
struct threshold_block {
unsigned int block;
unsigned int bank;
unsigned int cpu;
u32 address;
u16 interrupt_enable;
bool interrupt_capable;
u16 threshold_limit;
struct kobject kobj;
struct list_head miscj;
};
struct threshold_bank {
struct kobject *kobj;
struct threshold_block *blocks;
/* initialized to the number of CPUs on the node sharing this bank */
atomic_t cpus;
};
struct amd_northbridge {
struct pci_dev *misc;
struct pci_dev *link;
struct amd_l3_cache l3_cache;
struct threshold_bank *bank4;
};
struct amd_northbridge_info {

1
arch/x86/kernel/amd_nb.c
View File

@ -16,6 +16,7 @@ const struct pci_device_id amd_nb_misc_ids[] = {
{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MISC) },
{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC) },
{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_NB_F3) },
{ PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_M10H_F3) },
{}
};
EXPORT_SYMBOL(amd_nb_misc_ids);

270
arch/x86/kernel/cpu/mcheck/mce_amd.c
View File

@ -1,15 +1,17 @@
/*
* (c) 2005, 2006 Advanced Micro Devices, Inc.
* (c) 2005-2012 Advanced Micro Devices, Inc.
* Your use of this code is subject to the terms and conditions of the
* GNU general public license version 2. See "COPYING" or
* http://www.gnu.org/licenses/gpl.html
*
* Written by Jacob Shin - AMD, Inc.
*
* Support : jacob.shin@amd.com
* Support: borislav.petkov@amd.com
*
* April 2006
* - added support for AMD Family 0x10 processors
* May 2012
* - major scrubbing
*
* All MC4_MISCi registers are shared between multi-cores
*/
@ -25,6 +27,7 @@
#include <linux/cpu.h>
#include <linux/smp.h>
#include <asm/amd_nb.h>
#include <asm/apic.h>
#include <asm/idle.h>
#include <asm/mce.h>
@ -45,23 +48,15 @@
#define MASK_BLKPTR_LO 0xFF000000
#define MCG_XBLK_ADDR 0xC0000400
struct threshold_block {
unsigned int block;
unsigned int bank;
unsigned int cpu;
u32 address;
u16 interrupt_enable;
bool interrupt_capable;
u16 threshold_limit;
struct kobject kobj;
struct list_head miscj;
static const char * const th_names[] = {
"load_store",
"insn_fetch",
"combined_unit",
"",
"northbridge",
"execution_unit",
};
struct threshold_bank {
struct kobject *kobj;
struct threshold_block *blocks;
cpumask_var_t cpus;
};
static DEFINE_PER_CPU(struct threshold_bank * [NR_BANKS], threshold_banks);
static unsigned char shared_bank[NR_BANKS] = {
@ -84,6 +79,26 @@ struct thresh_restart {
u16 old_limit;
};
static const char * const bank4_names(struct threshold_block *b)
{
switch (b->address) {
/* MSR4_MISC0 */
case 0x00000413:
return "dram";
case 0xc0000408:
return "ht_links";
case 0xc0000409:
return "l3_cache";
default:
WARN(1, "Funny MSR: 0x%08x\n", b->address);
return "";
}
};
static bool lvt_interrupt_supported(unsigned int bank, u32 msr_high_bits)
{
/*
@ -224,8 +239,6 @@ void mce_amd_feature_init(struct cpuinfo_x86 *c)
if (!block)
per_cpu(bank_map, cpu) |= (1 << bank);
if (shared_bank[bank] && c->cpu_core_id)
break;
memset(&b, 0, sizeof(b));
b.cpu = cpu;
@ -326,7 +339,7 @@ struct threshold_attr {
#define SHOW_FIELDS(name) \
static ssize_t show_ ## name(struct threshold_block *b, char *buf) \
{ \
return sprintf(buf, "%lx\n", (unsigned long) b->name); \
return sprintf(buf, "%lu\n", (unsigned long) b->name); \
}
SHOW_FIELDS(interrupt_enable)
SHOW_FIELDS(threshold_limit)
@ -377,37 +390,20 @@ store_threshold_limit(struct threshold_block *b, const char *buf, size_t size)
return size;
}
struct threshold_block_cross_cpu {
struct threshold_block *tb;
long retval;
};
static void local_error_count_handler(void *_tbcc)
{
struct threshold_block_cross_cpu *tbcc = _tbcc;
struct threshold_block *b = tbcc->tb;
u32 low, high;
rdmsr(b->address, low, high);
tbcc->retval = (high & 0xFFF) - (THRESHOLD_MAX - b->threshold_limit);
}
static ssize_t show_error_count(struct threshold_block *b, char *buf)
{
struct threshold_block_cross_cpu tbcc = { .tb = b, };
u32 lo, hi;
smp_call_function_single(b->cpu, local_error_count_handler, &tbcc, 1);
return sprintf(buf, "%lx\n", tbcc.retval);
rdmsr_on_cpu(b->cpu, b->address, &lo, &hi);
return sprintf(buf, "%u\n", ((hi & THRESHOLD_MAX) -
(THRESHOLD_MAX - b->threshold_limit)));
}
static ssize_t store_error_count(struct threshold_block *b,
const char *buf, size_t count)
{
struct thresh_restart tr = { .b = b, .reset = 1, .old_limit = 0 };
smp_call_function_single(b->cpu, threshold_restart_bank, &tr, 1);
return 1;
}
static struct threshold_attr error_count = {
.attr = {.name = __stringify(error_count), .mode = 0444 },
.show = show_error_count,
};
#define RW_ATTR(val) \
static struct threshold_attr val = { \
@ -418,7 +414,6 @@ static struct threshold_attr val = { \
RW_ATTR(interrupt_enable);
RW_ATTR(threshold_limit);
RW_ATTR(error_count);
static struct attribute *default_attrs[] = {
&threshold_limit.attr,
@ -517,7 +512,7 @@ static __cpuinit int allocate_threshold_blocks(unsigned int cpu,
err = kobject_init_and_add(&b->kobj, &threshold_ktype,
per_cpu(threshold_banks, cpu)[bank]->kobj,
"misc%i", block);
(bank == 4 ? bank4_names(b) : th_names[bank]));
if (err)
goto out_free;
recurse:
@ -548,98 +543,91 @@ static __cpuinit int allocate_threshold_blocks(unsigned int cpu,
return err;
}
static __cpuinit long
local_allocate_threshold_blocks(int cpu, unsigned int bank)
static __cpuinit int __threshold_add_blocks(struct threshold_bank *b)
{
return allocate_threshold_blocks(cpu, bank, 0,
MSR_IA32_MC0_MISC + bank * 4);
struct list_head *head = &b->blocks->miscj;
struct threshold_block *pos = NULL;
struct threshold_block *tmp = NULL;
int err = 0;
err = kobject_add(&b->blocks->kobj, b->kobj, b->blocks->kobj.name);
if (err)
return err;
list_for_each_entry_safe(pos, tmp, head, miscj) {
err = kobject_add(&pos->kobj, b->kobj, pos->kobj.name);
if (err) {
list_for_each_entry_safe_reverse(pos, tmp, head, miscj)
kobject_del(&pos->kobj);
return err;
}
}
return err;
}
/* symlinks sibling shared banks to first core. first core owns dir/files. */
static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
{
int i, err = 0;
struct threshold_bank *b = NULL;
struct device *dev = per_cpu(mce_device, cpu);
char name[32];
struct amd_northbridge *nb = NULL;
struct threshold_bank *b = NULL;
const char *name = th_names[bank];
int err = 0;
sprintf(name, "threshold_bank%i", bank);
if (shared_bank[bank]) {
#ifdef CONFIG_SMP
if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */
i = cpumask_first(cpu_llc_shared_mask(cpu));
nb = node_to_amd_nb(amd_get_nb_id(cpu));
WARN_ON(!nb);
/* threshold descriptor already initialized on this node? */
if (nb->bank4) {
/* yes, use it */
b = nb->bank4;
err = kobject_add(b->kobj, &dev->kobj, name);
if (err)
goto out;
per_cpu(threshold_banks, cpu)[bank] = b;
atomic_inc(&b->cpus);
err = __threshold_add_blocks(b);
/* first core not up yet */
if (cpu_data(i).cpu_core_id)
goto out;
/* already linked */
if (per_cpu(threshold_banks, cpu)[bank])
goto out;
b = per_cpu(threshold_banks, i)[bank];
if (!b)
goto out;
err = sysfs_create_link(&dev->kobj, b->kobj, name);
if (err)
goto out;
cpumask_copy(b->cpus, cpu_llc_shared_mask(cpu));
per_cpu(threshold_banks, cpu)[bank] = b;
goto out;
}
}
#endif
b = kzalloc(sizeof(struct threshold_bank), GFP_KERNEL);
if (!b) {
err = -ENOMEM;
goto out;
}
if (!zalloc_cpumask_var(&b->cpus, GFP_KERNEL)) {
kfree(b);
err = -ENOMEM;
goto out;
}
b->kobj = kobject_create_and_add(name, &dev->kobj);
if (!b->kobj)
if (!b->kobj) {
err = -EINVAL;
goto out_free;
#ifndef CONFIG_SMP
cpumask_setall(b->cpus);
#else
cpumask_set_cpu(cpu, b->cpus);
#endif
}
per_cpu(threshold_banks, cpu)[bank] = b;
err = local_allocate_threshold_blocks(cpu, bank);
if (err)
goto out_free;
if (shared_bank[bank]) {
atomic_set(&b->cpus, 1);
for_each_cpu(i, b->cpus) {
if (i == cpu)
continue;
dev = per_cpu(mce_device, i);
if (dev)
err = sysfs_create_link(&dev->kobj,b->kobj, name);
if (err)
goto out;
per_cpu(threshold_banks, i)[bank] = b;
/* nb is already initialized, see above */
WARN_ON(nb->bank4);
nb->bank4 = b;
}
goto out;
err = allocate_threshold_blocks(cpu, bank, 0,
MSR_IA32_MC0_MISC + bank * 4);
if (!err)
goto out;
out_free:
per_cpu(threshold_banks, cpu)[bank] = NULL;
free_cpumask_var(b->cpus);
out_free:
kfree(b);
out:
out:
return err;
}
@ -660,12 +648,6 @@ static __cpuinit int threshold_create_device(unsigned int cpu)
return err;
}
/*
* let's be hotplug friendly.
* in case of multiple core processors, the first core always takes ownership
* of shared sysfs dir/files, and rest of the cores will be symlinked to it.
*/
static void deallocate_threshold_block(unsigned int cpu,
unsigned int bank)
{
@ -686,41 +668,42 @@ static void deallocate_threshold_block(unsigned int cpu,
per_cpu(threshold_banks, cpu)[bank]->blocks = NULL;
}
static void __threshold_remove_blocks(struct threshold_bank *b)
{
struct threshold_block *pos = NULL;
struct threshold_block *tmp = NULL;
kobject_del(b->kobj);
list_for_each_entry_safe(pos, tmp, &b->blocks->miscj, miscj)
kobject_del(&pos->kobj);
}
static void threshold_remove_bank(unsigned int cpu, int bank)
{
struct amd_northbridge *nb;
struct threshold_bank *b;
struct device *dev;
char name[32];
int i = 0;
b = per_cpu(threshold_banks, cpu)[bank];
if (!b)
return;
if (!b->blocks)
goto free_out;
sprintf(name, "threshold_bank%i", bank);
#ifdef CONFIG_SMP
/* sibling symlink */
if (shared_bank[bank] && b->blocks->cpu != cpu) {
dev = per_cpu(mce_device, cpu);
sysfs_remove_link(&dev->kobj, name);
per_cpu(threshold_banks, cpu)[bank] = NULL;
return;
}
#endif
/* remove all sibling symlinks before unregistering */
for_each_cpu(i, b->cpus) {
if (i == cpu)
continue;
dev = per_cpu(mce_device, i);
if (dev)
sysfs_remove_link(&dev->kobj, name);
per_cpu(threshold_banks, i)[bank] = NULL;
if (shared_bank[bank]) {
if (!atomic_dec_and_test(&b->cpus)) {
__threshold_remove_blocks(b);
per_cpu(threshold_banks, cpu)[bank] = NULL;
return;
} else {
/*
* the last CPU on this node using the shared bank is
* going away, remove that bank now.
*/
nb = node_to_amd_nb(amd_get_nb_id(cpu));
nb->bank4 = NULL;
}
}
deallocate_threshold_block(cpu, bank);
@ -728,7 +711,6 @@ static void threshold_remove_bank(unsigned int cpu, int bank)
free_out:
kobject_del(b->kobj);
kobject_put(b->kobj);
free_cpumask_var(b->cpus);
kfree(b);
per_cpu(threshold_banks, cpu)[bank] = NULL;
}

5
drivers/hwmon/k10temp.c
View File

@ -33,9 +33,6 @@ static bool force;
module_param(force, bool, 0444);
MODULE_PARM_DESC(force, "force loading on processors with erratum 319");
/* PCI-IDs for Northbridge devices not used anywhere else */
#define PCI_DEVICE_ID_AMD_15H_M10H_NB_F3 0x1403
/* CPUID function 0x80000001, ebx */
#define CPUID_PKGTYPE_MASK 0xf0000000
#define CPUID_PKGTYPE_F 0x00000000
@ -213,7 +210,7 @@ static DEFINE_PCI_DEVICE_TABLE(k10temp_id_table) = {
{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_11H_NB_MISC) },
{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_CNB17H_F3) },
{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_NB_F3) },
{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M10H_NB_F3) },
{ PCI_VDEVICE(AMD, PCI_DEVICE_ID_AMD_15H_M10H_F3) },
{}
};
MODULE_DEVICE_TABLE(pci, k10temp_id_table);

1
include/linux/pci_ids.h
View File

@ -517,6 +517,7 @@
#define PCI_DEVICE_ID_AMD_11H_NB_DRAM 0x1302
#define PCI_DEVICE_ID_AMD_11H_NB_MISC 0x1303
#define PCI_DEVICE_ID_AMD_11H_NB_LINK 0x1304
#define PCI_DEVICE_ID_AMD_15H_M10H_F3 0x1403
#define PCI_DEVICE_ID_AMD_15H_NB_F0 0x1600
#define PCI_DEVICE_ID_AMD_15H_NB_F1 0x1601
#define PCI_DEVICE_ID_AMD_15H_NB_F2 0x1602