linux/arch/x86/kernel/cpu/amd.c

1356 lines
35 KiB
C
Raw Permalink Normal View History

// SPDX-License-Identifier: GPL-2.0-only
#include <linux/export.h>
#include <linux/bitops.h>
#include <linux/elf.h>
#include <linux/mm.h>
#include <linux/io.h>
#include <linux/sched.h>
#include <linux/sched/clock.h>
x86/mm: Improve AMD Bulldozer ASLR workaround The ASLR implementation needs to special-case AMD F15h processors by clearing out bits [14:12] of the virtual address in order to avoid I$ cross invalidations and thus performance penalty for certain workloads. For details, see: dfb09f9b7ab0 ("x86, amd: Avoid cache aliasing penalties on AMD family 15h") This special case reduces the mmapped file's entropy by 3 bits. The following output is the run on an AMD Opteron 62xx class CPU processor under x86_64 Linux 4.0.0: $ for i in `seq 1 10`; do cat /proc/self/maps | grep "r-xp.*libc" ; done b7588000-b7736000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b7570000-b771e000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b75d0000-b777e000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b75b0000-b775e000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b7578000-b7726000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 ... Bits [12:14] are always 0, i.e. the address always ends in 0x8000 or 0x0000. 32-bit systems, as in the example above, are especially sensitive to this issue because 32-bit randomness for VA space is 8 bits (see mmap_rnd()). With the Bulldozer special case, this diminishes to only 32 different slots of mmap virtual addresses. This patch randomizes per boot the three affected bits rather than setting them to zero. Since all the shared pages have the same value at bits [12..14], there is no cache aliasing problems. This value gets generated during system boot and it is thus not known to a potential remote attacker. Therefore, the impact from the Bulldozer workaround gets diminished and ASLR randomness increased. More details at: http://hmarco.org/bugs/AMD-Bulldozer-linux-ASLR-weakness-reducing-mmaped-files-by-eight.html Original white paper by AMD dealing with the issue: http://developer.amd.com/wordpress/media/2012/10/SharedL1InstructionCacheonAMD15hCPU.pdf Mentored-by: Ismael Ripoll <iripoll@disca.upv.es> Signed-off-by: Hector Marco-Gisbert <hecmargi@upv.es> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Kees Cook <keescook@chromium.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jan-Simon <dl9pf@gmx.de> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-fsdevel@vger.kernel.org Link: http://lkml.kernel.org/r/1427456301-3764-1-git-send-email-hecmargi@upv.es Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-03-27 12:38:21 +01:00
#include <linux/random.h>
sched/topology: Improve load balancing on AMD EPYC systems SD_BALANCE_{FORK,EXEC} and SD_WAKE_AFFINE are stripped in sd_init() for any sched domains with a NUMA distance greater than 2 hops (RECLAIM_DISTANCE). The idea being that it's expensive to balance across domains that far apart. However, as is rather unfortunately explained in: commit 32e45ff43eaf ("mm: increase RECLAIM_DISTANCE to 30") the value for RECLAIM_DISTANCE is based on node distance tables from 2011-era hardware. Current AMD EPYC machines have the following NUMA node distances: node distances: node 0 1 2 3 4 5 6 7 0: 10 16 16 16 32 32 32 32 1: 16 10 16 16 32 32 32 32 2: 16 16 10 16 32 32 32 32 3: 16 16 16 10 32 32 32 32 4: 32 32 32 32 10 16 16 16 5: 32 32 32 32 16 10 16 16 6: 32 32 32 32 16 16 10 16 7: 32 32 32 32 16 16 16 10 where 2 hops is 32. The result is that the scheduler fails to load balance properly across NUMA nodes on different sockets -- 2 hops apart. For example, pinning 16 busy threads to NUMA nodes 0 (CPUs 0-7) and 4 (CPUs 32-39) like so, $ numactl -C 0-7,32-39 ./spinner 16 causes all threads to fork and remain on node 0 until the active balancer kicks in after a few seconds and forcibly moves some threads to node 4. Override node_reclaim_distance for AMD Zen. Signed-off-by: Matt Fleming <matt@codeblueprint.co.uk> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Mel Gorman <mgorman@techsingularity.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@surriel.com> Cc: Suravee.Suthikulpanit@amd.com Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Thomas.Lendacky@amd.com Cc: Tony Luck <tony.luck@intel.com> Link: https://lkml.kernel.org/r/20190808195301.13222-3-matt@codeblueprint.co.uk Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-08-08 20:53:01 +01:00
#include <linux/topology.h>
#include <linux/platform_data/x86/amd-fch.h>
#include <asm/processor.h>
#include <asm/apic.h>
#include <asm/cacheinfo.h>
#include <asm/cpu.h>
#include <asm/cpu_device_id.h>
#include <asm/spec-ctrl.h>
#include <asm/smp.h>
2020-08-06 14:35:11 +02:00
#include <asm/numa.h>
#include <asm/pci-direct.h>
x86/asm/delay: Introduce an MWAITX-based delay with a configurable timer MWAITX can enable a timer and a corresponding timer value specified in SW P0 clocks. The SW P0 frequency is the same as TSC. The timer provides an upper bound on how long the instruction waits before exiting. This way, a delay function in the kernel can leverage that MWAITX timer of MWAITX. When a CPU core executes MWAITX, it will be quiesced in a waiting phase, diminishing its power consumption. This way, we can save power in comparison to our default TSC-based delays. A simple test shows that: $ cat /sys/bus/pci/devices/0000\:00\:18.4/hwmon/hwmon0/power1_acc $ sleep 10000s $ cat /sys/bus/pci/devices/0000\:00\:18.4/hwmon/hwmon0/power1_acc Results: * TSC-based default delay: 485115 uWatts average power * MWAITX-based delay: 252738 uWatts average power Thus, that's about 240 milliWatts less power consumption. The test method relies on the support of AMD CPU accumulated power algorithm in fam15h_power for which patches are forthcoming. Suggested-by: Andy Lutomirski <luto@amacapital.net> Suggested-by: Borislav Petkov <bp@suse.de> Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Huang Rui <ray.huang@amd.com> [ Fix delay truncation. ] Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Aaron Lu <aaron.lu@intel.com> Cc: Andreas Herrmann <herrmann.der.user@gmail.com> Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: Fengguang Wu <fengguang.wu@intel.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hector Marco-Gisbert <hecmargi@upv.es> Cc: Jacob Shin <jacob.w.shin@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tony Li <tony.li@amd.com> Link: http://lkml.kernel.org/r/1438744732-1459-3-git-send-email-ray.huang@amd.com Link: http://lkml.kernel.org/r/1439201994-28067-4-git-send-email-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-08-10 12:19:54 +02:00
#include <asm/delay.h>
#include <asm/debugreg.h>
#include <asm/resctrl.h>
#include <asm/msr.h>
x86/sev: Add SEV-SNP host initialization support The memory integrity guarantees of SEV-SNP are enforced through a new structure called the Reverse Map Table (RMP). The RMP is a single data structure shared across the system that contains one entry for every 4K page of DRAM that may be used by SEV-SNP VMs. The APM Volume 2 section on Secure Nested Paging (SEV-SNP) details a number of steps needed to detect/enable SEV-SNP and RMP table support on the host: - Detect SEV-SNP support based on CPUID bit - Initialize the RMP table memory reported by the RMP base/end MSR registers and configure IOMMU to be compatible with RMP access restrictions - Set the MtrrFixDramModEn bit in SYSCFG MSR - Set the SecureNestedPagingEn and VMPLEn bits in the SYSCFG MSR - Configure IOMMU RMP table entry format is non-architectural and it can vary by processor. It is defined by the PPR document for each respective CPU family. Restrict SNP support to CPU models/families which are compatible with the current RMP table entry format to guard against any undefined behavior when running on other system types. Future models/support will handle this through an architectural mechanism to allow for broader compatibility. SNP host code depends on CONFIG_KVM_AMD_SEV config flag which may be enabled even when CONFIG_AMD_MEM_ENCRYPT isn't set, so update the SNP-specific IOMMU helpers used here to rely on CONFIG_KVM_AMD_SEV instead of CONFIG_AMD_MEM_ENCRYPT. Signed-off-by: Brijesh Singh <brijesh.singh@amd.com> Co-developed-by: Ashish Kalra <ashish.kalra@amd.com> Signed-off-by: Ashish Kalra <ashish.kalra@amd.com> Co-developed-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Co-developed-by: Borislav Petkov (AMD) <bp@alien8.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Co-developed-by: Michael Roth <michael.roth@amd.com> Signed-off-by: Michael Roth <michael.roth@amd.com> Link: https://lore.kernel.org/r/20240126041126.1927228-5-michael.roth@amd.com
2024-01-25 22:11:04 -06:00
#include <asm/sev.h>
#ifdef CONFIG_X86_64
# include <asm/mmconfig.h>
#endif
#include "cpu.h"
u16 invlpgb_count_max __ro_after_init = 1;
static inline int rdmsrq_amd_safe(unsigned msr, u64 *p)
{
u32 gprs[8] = { 0 };
int err;
WARN_ONCE((boot_cpu_data.x86 != 0xf),
"%s should only be used on K8!\n", __func__);
gprs[1] = msr;
gprs[7] = 0x9c5a203a;
err = rdmsr_safe_regs(gprs);
*p = gprs[0] | ((u64)gprs[2] << 32);
return err;
}
static inline int wrmsrq_amd_safe(unsigned msr, u64 val)
{
u32 gprs[8] = { 0 };
WARN_ONCE((boot_cpu_data.x86 != 0xf),
"%s should only be used on K8!\n", __func__);
gprs[0] = (u32)val;
gprs[1] = msr;
gprs[2] = val >> 32;
gprs[7] = 0x9c5a203a;
return wrmsr_safe_regs(gprs);
}
/*
* B step AMD K6 before B 9730xxxx have hardware bugs that can cause
* misexecution of code under Linux. Owners of such processors should
* contact AMD for precise details and a CPU swap.
*
* See http://www.multimania.com/poulot/k6bug.html
* and section 2.6.2 of "AMD-K6 Processor Revision Guide - Model 6"
* (Publication # 21266 Issue Date: August 1998)
*
* The following test is erm.. interesting. AMD neglected to up
* the chip setting when fixing the bug but they also tweaked some
* performance at the same time..
*/
#ifdef CONFIG_X86_32
extern __visible void vide(void);
__asm__(".text\n"
".globl vide\n"
".type vide, @function\n"
".align 4\n"
"vide: ret\n");
#endif
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void init_amd_k5(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_32
/*
* General Systems BIOSen alias the cpu frequency registers
* of the Elan at 0x000df000. Unfortunately, one of the Linux
* drivers subsequently pokes it, and changes the CPU speed.
* Workaround : Remove the unneeded alias.
*/
#define CBAR (0xfffc) /* Configuration Base Address (32-bit) */
#define CBAR_ENB (0x80000000)
#define CBAR_KEY (0X000000CB)
if (c->x86_model == 9 || c->x86_model == 10) {
if (inl(CBAR) & CBAR_ENB)
outl(0 | CBAR_KEY, CBAR);
}
#endif
}
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void init_amd_k6(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_32
u32 l, h;
int mbytes = get_num_physpages() >> (20-PAGE_SHIFT);
if (c->x86_model < 6) {
/* Based on AMD doc 20734R - June 2000 */
if (c->x86_model == 0) {
clear_cpu_cap(c, X86_FEATURE_APIC);
set_cpu_cap(c, X86_FEATURE_PGE);
}
return;
}
if (c->x86_model == 6 && c->x86_stepping == 1) {
const int K6_BUG_LOOP = 1000000;
int n;
void (*f_vide)(void);
u64 d, d2;
pr_info("AMD K6 stepping B detected - ");
/*
* It looks like AMD fixed the 2.6.2 bug and improved indirect
* calls at the same time.
*/
n = K6_BUG_LOOP;
f_vide = vide;
OPTIMIZER_HIDE_VAR(f_vide);
d = rdtsc();
while (n--)
f_vide();
d2 = rdtsc();
d = d2-d;
if (d > 20*K6_BUG_LOOP)
pr_cont("system stability may be impaired when more than 32 MB are used.\n");
else
pr_cont("probably OK (after B9730xxxx).\n");
}
/* K6 with old style WHCR */
if (c->x86_model < 8 ||
(c->x86_model == 8 && c->x86_stepping < 8)) {
/* We can only write allocate on the low 508Mb */
if (mbytes > 508)
mbytes = 508;
rdmsr(MSR_K6_WHCR, l, h);
if ((l&0x0000FFFF) == 0) {
unsigned long flags;
l = (1<<0)|((mbytes/4)<<1);
local_irq_save(flags);
wbinvd();
wrmsr(MSR_K6_WHCR, l, h);
local_irq_restore(flags);
pr_info("Enabling old style K6 write allocation for %d Mb\n",
mbytes);
}
return;
}
if ((c->x86_model == 8 && c->x86_stepping > 7) ||
c->x86_model == 9 || c->x86_model == 13) {
/* The more serious chips .. */
if (mbytes > 4092)
mbytes = 4092;
rdmsr(MSR_K6_WHCR, l, h);
if ((l&0xFFFF0000) == 0) {
unsigned long flags;
l = ((mbytes>>2)<<22)|(1<<16);
local_irq_save(flags);
wbinvd();
wrmsr(MSR_K6_WHCR, l, h);
local_irq_restore(flags);
pr_info("Enabling new style K6 write allocation for %d Mb\n",
mbytes);
}
return;
}
if (c->x86_model == 10) {
/* AMD Geode LX is model 10 */
/* placeholder for any needed mods */
return;
}
#endif
}
static void init_amd_k7(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_32
u32 l, h;
/*
* Bit 15 of Athlon specific MSR 15, needs to be 0
* to enable SSE on Palomino/Morgan/Barton CPU's.
* If the BIOS didn't enable it already, enable it here.
*/
if (c->x86_model >= 6 && c->x86_model <= 10) {
if (!cpu_has(c, X86_FEATURE_XMM)) {
pr_info("Enabling disabled K7/SSE Support.\n");
msr_clear_bit(MSR_K7_HWCR, 15);
set_cpu_cap(c, X86_FEATURE_XMM);
}
}
/*
* It's been determined by AMD that Athlons since model 8 stepping 1
* are more robust with CLK_CTL set to 200xxxxx instead of 600xxxxx
* As per AMD technical note 27212 0.2
*/
if ((c->x86_model == 8 && c->x86_stepping >= 1) || (c->x86_model > 8)) {
rdmsr(MSR_K7_CLK_CTL, l, h);
if ((l & 0xfff00000) != 0x20000000) {
pr_info("CPU: CLK_CTL MSR was %x. Reprogramming to %x\n",
l, ((l & 0x000fffff)|0x20000000));
wrmsr(MSR_K7_CLK_CTL, (l & 0x000fffff)|0x20000000, h);
}
}
/* calling is from identify_secondary_cpu() ? */
if (!c->cpu_index)
return;
/*
* Certain Athlons might work (for various values of 'work') in SMP
* but they are not certified as MP capable.
*/
/* Athlon 660/661 is valid. */
if ((c->x86_model == 6) && ((c->x86_stepping == 0) ||
(c->x86_stepping == 1)))
return;
/* Duron 670 is valid */
if ((c->x86_model == 7) && (c->x86_stepping == 0))
return;
/*
* Athlon 662, Duron 671, and Athlon >model 7 have capability
* bit. It's worth noting that the A5 stepping (662) of some
* Athlon XP's have the MP bit set.
* See http://www.heise.de/newsticker/data/jow-18.10.01-000 for
* more.
*/
if (((c->x86_model == 6) && (c->x86_stepping >= 2)) ||
((c->x86_model == 7) && (c->x86_stepping >= 1)) ||
(c->x86_model > 7))
if (cpu_has(c, X86_FEATURE_MP))
return;
/* If we get here, not a certified SMP capable AMD system. */
/*
* Don't taint if we are running SMP kernel on a single non-MP
* approved Athlon
*/
WARN_ONCE(1, "WARNING: This combination of AMD"
" processors is not suitable for SMP.\n");
add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_NOW_UNRELIABLE);
#endif
}
#ifdef CONFIG_NUMA
/*
* To workaround broken NUMA config. Read the comment in
* srat_detect_node().
*/
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
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
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void srat_detect_node(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_NUMA
int cpu = smp_processor_id();
int node;
unsigned apicid = c->topo.apicid;
node = numa_cpu_node(cpu);
if (node == NUMA_NO_NODE)
node = per_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
* which the K8 northbridge parsing fills in. 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->topo.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 bsp_determine_snp(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_ARCH_HAS_CC_PLATFORM
cc_vendor = CC_VENDOR_AMD;
if (cpu_has(c, X86_FEATURE_SEV_SNP)) {
/*
* RMP table entry format is not architectural and is defined by the
* per-processor PPR. Restrict SNP support on the known CPU models
* for which the RMP table entry format is currently defined or for
* processors which support the architecturally defined RMPREAD
* instruction.
*/
if (!cpu_has(c, X86_FEATURE_HYPERVISOR) &&
(cpu_feature_enabled(X86_FEATURE_ZEN3) ||
cpu_feature_enabled(X86_FEATURE_ZEN4) ||
cpu_feature_enabled(X86_FEATURE_RMPREAD)) &&
snp_probe_rmptable_info()) {
cc_platform_set(CC_ATTR_HOST_SEV_SNP);
} else {
setup_clear_cpu_cap(X86_FEATURE_SEV_SNP);
cc_platform_clear(CC_ATTR_HOST_SEV_SNP);
}
}
#endif
}
#define ZEN_MODEL_STEP_UCODE(fam, model, step, ucode) \
X86_MATCH_VFM_STEPS(VFM_MAKE(X86_VENDOR_AMD, fam, model), \
step, step, ucode)
static const struct x86_cpu_id amd_tsa_microcode[] = {
ZEN_MODEL_STEP_UCODE(0x19, 0x01, 0x1, 0x0a0011d7),
ZEN_MODEL_STEP_UCODE(0x19, 0x01, 0x2, 0x0a00123b),
ZEN_MODEL_STEP_UCODE(0x19, 0x08, 0x2, 0x0a00820d),
ZEN_MODEL_STEP_UCODE(0x19, 0x11, 0x1, 0x0a10114c),
ZEN_MODEL_STEP_UCODE(0x19, 0x11, 0x2, 0x0a10124c),
ZEN_MODEL_STEP_UCODE(0x19, 0x18, 0x1, 0x0a108109),
ZEN_MODEL_STEP_UCODE(0x19, 0x21, 0x0, 0x0a20102e),
ZEN_MODEL_STEP_UCODE(0x19, 0x21, 0x2, 0x0a201211),
ZEN_MODEL_STEP_UCODE(0x19, 0x44, 0x1, 0x0a404108),
ZEN_MODEL_STEP_UCODE(0x19, 0x50, 0x0, 0x0a500012),
ZEN_MODEL_STEP_UCODE(0x19, 0x61, 0x2, 0x0a60120a),
ZEN_MODEL_STEP_UCODE(0x19, 0x74, 0x1, 0x0a704108),
ZEN_MODEL_STEP_UCODE(0x19, 0x75, 0x2, 0x0a705208),
ZEN_MODEL_STEP_UCODE(0x19, 0x78, 0x0, 0x0a708008),
ZEN_MODEL_STEP_UCODE(0x19, 0x7c, 0x0, 0x0a70c008),
ZEN_MODEL_STEP_UCODE(0x19, 0xa0, 0x2, 0x0aa00216),
{},
};
static void tsa_init(struct cpuinfo_x86 *c)
{
if (cpu_has(c, X86_FEATURE_HYPERVISOR))
return;
if (cpu_has(c, X86_FEATURE_ZEN3) ||
cpu_has(c, X86_FEATURE_ZEN4)) {
if (x86_match_min_microcode_rev(amd_tsa_microcode))
setup_force_cpu_cap(X86_FEATURE_VERW_CLEAR);
else
pr_debug("%s: current revision: 0x%x\n", __func__, c->microcode);
} else {
setup_force_cpu_cap(X86_FEATURE_TSA_SQ_NO);
setup_force_cpu_cap(X86_FEATURE_TSA_L1_NO);
}
}
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void bsp_init_amd(struct cpuinfo_x86 *c)
{
if (cpu_has(c, X86_FEATURE_CONSTANT_TSC)) {
if (c->x86 > 0x10 ||
(c->x86 == 0x10 && c->x86_model >= 0x2)) {
u64 val;
rdmsrq(MSR_K7_HWCR, val);
if (!(val & BIT(24)))
pr_warn(FW_BUG "TSC doesn't count with P0 frequency!\n");
}
}
if (c->x86 == 0x15) {
unsigned long upperbit;
u32 cpuid, assoc;
cpuid = cpuid_edx(0x80000005);
assoc = cpuid >> 16 & 0xff;
upperbit = ((cpuid >> 24) << 10) / assoc;
va_align.mask = (upperbit - 1) & PAGE_MASK;
va_align.flags = ALIGN_VA_32 | ALIGN_VA_64;
x86/mm: Improve AMD Bulldozer ASLR workaround The ASLR implementation needs to special-case AMD F15h processors by clearing out bits [14:12] of the virtual address in order to avoid I$ cross invalidations and thus performance penalty for certain workloads. For details, see: dfb09f9b7ab0 ("x86, amd: Avoid cache aliasing penalties on AMD family 15h") This special case reduces the mmapped file's entropy by 3 bits. The following output is the run on an AMD Opteron 62xx class CPU processor under x86_64 Linux 4.0.0: $ for i in `seq 1 10`; do cat /proc/self/maps | grep "r-xp.*libc" ; done b7588000-b7736000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b7570000-b771e000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b75d0000-b777e000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b75b0000-b775e000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 b7578000-b7726000 r-xp 00000000 00:01 4924 /lib/i386-linux-gnu/libc.so.6 ... Bits [12:14] are always 0, i.e. the address always ends in 0x8000 or 0x0000. 32-bit systems, as in the example above, are especially sensitive to this issue because 32-bit randomness for VA space is 8 bits (see mmap_rnd()). With the Bulldozer special case, this diminishes to only 32 different slots of mmap virtual addresses. This patch randomizes per boot the three affected bits rather than setting them to zero. Since all the shared pages have the same value at bits [12..14], there is no cache aliasing problems. This value gets generated during system boot and it is thus not known to a potential remote attacker. Therefore, the impact from the Bulldozer workaround gets diminished and ASLR randomness increased. More details at: http://hmarco.org/bugs/AMD-Bulldozer-linux-ASLR-weakness-reducing-mmaped-files-by-eight.html Original white paper by AMD dealing with the issue: http://developer.amd.com/wordpress/media/2012/10/SharedL1InstructionCacheonAMD15hCPU.pdf Mentored-by: Ismael Ripoll <iripoll@disca.upv.es> Signed-off-by: Hector Marco-Gisbert <hecmargi@upv.es> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Kees Cook <keescook@chromium.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jan-Simon <dl9pf@gmx.de> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-fsdevel@vger.kernel.org Link: http://lkml.kernel.org/r/1427456301-3764-1-git-send-email-hecmargi@upv.es Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-03-27 12:38:21 +01:00
/* A random value per boot for bit slice [12:upper_bit) */
va_align.bits = get_random_u32() & va_align.mask;
}
x86/asm/delay: Introduce an MWAITX-based delay with a configurable timer MWAITX can enable a timer and a corresponding timer value specified in SW P0 clocks. The SW P0 frequency is the same as TSC. The timer provides an upper bound on how long the instruction waits before exiting. This way, a delay function in the kernel can leverage that MWAITX timer of MWAITX. When a CPU core executes MWAITX, it will be quiesced in a waiting phase, diminishing its power consumption. This way, we can save power in comparison to our default TSC-based delays. A simple test shows that: $ cat /sys/bus/pci/devices/0000\:00\:18.4/hwmon/hwmon0/power1_acc $ sleep 10000s $ cat /sys/bus/pci/devices/0000\:00\:18.4/hwmon/hwmon0/power1_acc Results: * TSC-based default delay: 485115 uWatts average power * MWAITX-based delay: 252738 uWatts average power Thus, that's about 240 milliWatts less power consumption. The test method relies on the support of AMD CPU accumulated power algorithm in fam15h_power for which patches are forthcoming. Suggested-by: Andy Lutomirski <luto@amacapital.net> Suggested-by: Borislav Petkov <bp@suse.de> Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Huang Rui <ray.huang@amd.com> [ Fix delay truncation. ] Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Aaron Lu <aaron.lu@intel.com> Cc: Andreas Herrmann <herrmann.der.user@gmail.com> Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: Fengguang Wu <fengguang.wu@intel.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hector Marco-Gisbert <hecmargi@upv.es> Cc: Jacob Shin <jacob.w.shin@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tony Li <tony.li@amd.com> Link: http://lkml.kernel.org/r/1438744732-1459-3-git-send-email-ray.huang@amd.com Link: http://lkml.kernel.org/r/1439201994-28067-4-git-send-email-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-08-10 12:19:54 +02:00
if (cpu_has(c, X86_FEATURE_MWAITX))
use_mwaitx_delay();
perf/x86/amd: Move nodes_per_socket into bsp_init_amd() nodes_per_socket is static and it needn't be initialized many times during every CPU core init. So move its initialization into bsp_init_amd(). Signed-off-by: Huang Rui <ray.huang@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Aaron Lu <aaron.lu@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andreas Herrmann <herrmann.der.user@googlemail.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: David Ahern <dsahern@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: Fengguang Wu <fengguang.wu@intel.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hector Marco-Gisbert <hecmargi@upv.es> Cc: Jacob Shin <jacob.w.shin@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Robert Richter <rric@kernel.org> Cc: Stephane Eranian <eranian@google.com> Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vincent.weaver@maine.edu> Cc: spg_linux_kernel@amd.com Link: http://lkml.kernel.org/r/1452739808-11871-2-git-send-email-ray.huang@amd.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-01-14 10:50:04 +08:00
if (!boot_cpu_has(X86_FEATURE_AMD_SSBD) &&
!boot_cpu_has(X86_FEATURE_VIRT_SSBD) &&
c->x86 >= 0x15 && c->x86 <= 0x17) {
unsigned int bit;
switch (c->x86) {
case 0x15: bit = 54; break;
case 0x16: bit = 33; break;
case 0x17: bit = 10; break;
default: return;
}
/*
* Try to cache the base value so further operations can
* avoid RMW. If that faults, do not enable SSBD.
*/
if (!rdmsrq_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 << bit;
}
}
resctrl_cpu_detect(c);
/* Figure out Zen generations: */
switch (c->x86) {
case 0x17:
switch (c->x86_model) {
case 0x00 ... 0x2f:
case 0x50 ... 0x5f:
setup_force_cpu_cap(X86_FEATURE_ZEN1);
break;
case 0x30 ... 0x4f:
case 0x60 ... 0x7f:
case 0x90 ... 0x91:
case 0xa0 ... 0xaf:
setup_force_cpu_cap(X86_FEATURE_ZEN2);
break;
default:
goto warn;
}
break;
case 0x19:
switch (c->x86_model) {
case 0x00 ... 0x0f:
case 0x20 ... 0x5f:
setup_force_cpu_cap(X86_FEATURE_ZEN3);
break;
case 0x10 ... 0x1f:
case 0x60 ... 0xaf:
setup_force_cpu_cap(X86_FEATURE_ZEN4);
break;
default:
goto warn;
}
break;
case 0x1a:
switch (c->x86_model) {
case 0x00 ... 0x2f:
case 0x40 ... 0x4f:
case 0x60 ... 0x7f:
setup_force_cpu_cap(X86_FEATURE_ZEN5);
break;
case 0x50 ... 0x5f:
case 0x90 ... 0xaf:
case 0xc0 ... 0xcf:
setup_force_cpu_cap(X86_FEATURE_ZEN6);
break;
default:
goto warn;
}
break;
default:
break;
}
bsp_determine_snp(c);
tsa_init(c);
if (cpu_has(c, X86_FEATURE_GP_ON_USER_CPUID))
setup_force_cpu_cap(X86_FEATURE_CPUID_FAULT);
return;
warn:
WARN_ONCE(1, "Family 0x%x, model: 0x%x??\n", c->x86, c->x86_model);
}
static void early_detect_mem_encrypt(struct cpuinfo_x86 *c)
{
u64 msr;
/*
* BIOS support is required for SME and SEV.
* For SME: If BIOS has enabled SME then adjust x86_phys_bits by
* the SME physical address space reduction value.
* If BIOS has not enabled SME then don't advertise the
* SME feature (set in scattered.c).
* If the kernel has not enabled SME via any means then
* don't advertise the SME feature.
* For SEV: If BIOS has not enabled SEV then don't advertise SEV and
* any additional functionality based on it.
*
* In all cases, since support for SME and SEV requires long mode,
* don't advertise the feature under CONFIG_X86_32.
*/
if (cpu_has(c, X86_FEATURE_SME) || cpu_has(c, X86_FEATURE_SEV)) {
/* Check if memory encryption is enabled */
rdmsrq(MSR_AMD64_SYSCFG, msr);
if (!(msr & MSR_AMD64_SYSCFG_MEM_ENCRYPT))
goto clear_all;
/*
* Always adjust physical address bits. Even though this
* will be a value above 32-bits this is still done for
* CONFIG_X86_32 so that accurate values are reported.
*/
c->x86_phys_bits -= (cpuid_ebx(0x8000001f) >> 6) & 0x3f;
if (IS_ENABLED(CONFIG_X86_32))
goto clear_all;
if (!sme_me_mask)
setup_clear_cpu_cap(X86_FEATURE_SME);
rdmsrq(MSR_K7_HWCR, msr);
if (!(msr & MSR_K7_HWCR_SMMLOCK))
goto clear_sev;
return;
clear_all:
setup_clear_cpu_cap(X86_FEATURE_SME);
clear_sev:
setup_clear_cpu_cap(X86_FEATURE_SEV);
setup_clear_cpu_cap(X86_FEATURE_SEV_ES);
setup_clear_cpu_cap(X86_FEATURE_SEV_SNP);
}
}
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void early_init_amd(struct cpuinfo_x86 *c)
{
x86/cpu/AMD: Make the microcode level available earlier in the boot Move the setting of the cpuinfo_x86.microcode field from amd_init() to early_amd_init() so that it is available earlier in the boot process. This avoids having to read MSR_AMD64_PATCH_LEVEL directly during early boot. Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Brijesh Singh <brijesh.singh@amd.com> Cc: Dave Young <dyoung@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Radim Krčmář <rkrcmar@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Toshimitsu Kani <toshi.kani@hpe.com> Cc: kasan-dev@googlegroups.com Cc: kvm@vger.kernel.org Cc: linux-arch@vger.kernel.org Cc: linux-doc@vger.kernel.org Cc: linux-efi@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/7b7525fa12593dac5f4b01fcc25c95f97e93862f.1500319216.git.thomas.lendacky@amd.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-07-17 16:10:23 -05:00
u32 dummy;
x86/jump_label: Initialize static branching early Static branching is useful to runtime patch branches that are used in hot path, but are infrequently changed. The x86 clock framework is one example that uses static branches to setup the best clock during boot and never changes it again. It is desired to enable the TSC based sched clock early to allow fine grained boot time analysis early on. That requires the static branching functionality to be functional early as well. Static branching requires patching nop instructions, thus, arch_init_ideal_nops() must be called prior to jump_label_init(). Do all the necessary steps to call arch_init_ideal_nops() right after early_cpu_init(), which also allows to insert a call to jump_label_init() right after that. jump_label_init() will be called again from the generic init code, but the code is protected against reinitialization already. [ tglx: Massaged changelog ] Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: steven.sistare@oracle.com Cc: daniel.m.jordan@oracle.com Cc: linux@armlinux.org.uk Cc: schwidefsky@de.ibm.com Cc: heiko.carstens@de.ibm.com Cc: john.stultz@linaro.org Cc: sboyd@codeaurora.org Cc: hpa@zytor.com Cc: douly.fnst@cn.fujitsu.com Cc: prarit@redhat.com Cc: feng.tang@intel.com Cc: pmladek@suse.com Cc: gnomes@lxorguk.ukuu.org.uk Cc: linux-s390@vger.kernel.org Cc: boris.ostrovsky@oracle.com Cc: jgross@suse.com Cc: pbonzini@redhat.com Link: https://lkml.kernel.org/r/20180719205545.16512-10-pasha.tatashin@oracle.com
2018-07-19 16:55:28 -04:00
if (c->x86 >= 0xf)
set_cpu_cap(c, X86_FEATURE_K8);
x86/cpu/AMD: Make the microcode level available earlier in the boot Move the setting of the cpuinfo_x86.microcode field from amd_init() to early_amd_init() so that it is available earlier in the boot process. This avoids having to read MSR_AMD64_PATCH_LEVEL directly during early boot. Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Brijesh Singh <brijesh.singh@amd.com> Cc: Dave Young <dyoung@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Radim Krčmář <rkrcmar@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Toshimitsu Kani <toshi.kani@hpe.com> Cc: kasan-dev@googlegroups.com Cc: kvm@vger.kernel.org Cc: linux-arch@vger.kernel.org Cc: linux-doc@vger.kernel.org Cc: linux-efi@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/7b7525fa12593dac5f4b01fcc25c95f97e93862f.1500319216.git.thomas.lendacky@amd.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-07-17 16:10:23 -05:00
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);
}
x86/cpufeature, perf/x86: Add AMD Accumulated Power Mechanism feature flag AMD CPU family 15h model 0x60 introduces a mechanism for measuring accumulated power. It is used to report the processor power consumption and support for it is indicated by CPUID Fn8000_0007_EDX[12]. Signed-off-by: Huang Rui <ray.huang@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Aaron Lu <aaron.lu@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andreas Herrmann <herrmann.der.user@googlemail.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: David Ahern <dsahern@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: Fengguang Wu <fengguang.wu@intel.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Hector Marco-Gisbert <hecmargi@upv.es> Cc: Jacob Shin <jacob.w.shin@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Kristen Carlson Accardi <kristen@linux.intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Robert Richter <rric@kernel.org> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Stephane Eranian <eranian@google.com> Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vincent.weaver@maine.edu> Cc: Wan Zongshun <Vincent.Wan@amd.com> Cc: spg_linux_kernel@amd.com Link: http://lkml.kernel.org/r/1452739808-11871-4-git-send-email-ray.huang@amd.com [ Resolved conflict and moved the synthetic CPUID slot to 19. ] Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-01-14 10:50:06 +08:00
/* Bit 12 of 8000_0007 edx is accumulated power mechanism. */
if (c->x86_power & BIT(12))
set_cpu_cap(c, X86_FEATURE_ACC_POWER);
/* Bit 14 indicates the Runtime Average Power Limit interface. */
if (c->x86_power & BIT(14))
set_cpu_cap(c, X86_FEATURE_RAPL);
#ifdef CONFIG_X86_64
set_cpu_cap(c, X86_FEATURE_SYSCALL32);
#else
/* Set MTRR capability flag if appropriate */
if (c->x86 == 5)
if (c->x86_model == 13 || c->x86_model == 9 ||
(c->x86_model == 8 && c->x86_stepping >= 8))
set_cpu_cap(c, X86_FEATURE_K6_MTRR);
#endif
#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PCI)
/*
* ApicID can always be treated as an 8-bit value for AMD APIC versions
* >= 0x10, but even old K8s came out of reset with version 0x10. So, we
* can safely set X86_FEATURE_EXTD_APICID unconditionally for families
* after 16h.
*/
if (boot_cpu_has(X86_FEATURE_APIC)) {
if (c->x86 > 0x16)
set_cpu_cap(c, X86_FEATURE_EXTD_APICID);
else if (c->x86 >= 0xf) {
/* check CPU config space for extended APIC ID */
unsigned int val;
val = read_pci_config(0, 24, 0, 0x68);
if ((val >> 17 & 0x3) == 0x3)
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);
/* F16h erratum 793, CVE-2013-6885 */
if (c->x86 == 0x16 && c->x86_model <= 0xf)
msr_set_bit(MSR_AMD64_LS_CFG, 15);
early_detect_mem_encrypt(c);
if (!cpu_has(c, X86_FEATURE_HYPERVISOR) && !cpu_has(c, X86_FEATURE_IBPB_BRTYPE)) {
if (c->x86 == 0x17 && boot_cpu_has(X86_FEATURE_AMD_IBPB))
setup_force_cpu_cap(X86_FEATURE_IBPB_BRTYPE);
else if (c->x86 >= 0x19 && !wrmsrq_safe(MSR_IA32_PRED_CMD, PRED_CMD_SBPB)) {
setup_force_cpu_cap(X86_FEATURE_IBPB_BRTYPE);
setup_force_cpu_cap(X86_FEATURE_SBPB);
}
}
x86/bugs: Separate AMD E400 erratum and C1E bug The workaround for the AMD Erratum E400 (Local APIC timer stops in C1E state) is a two step process: - Selection of the E400 aware idle routine - Detection whether the platform is affected The idle routine selection happens for possibly affected CPUs depending on family/model/stepping information. These range of CPUs is not necessarily affected as the decision whether to enable the C1E feature is made by the firmware. Unfortunately there is no way to query this at early boot. The current implementation polls a MSR in the E400 aware idle routine to detect whether the CPU is affected. This is inefficient on non affected CPUs because every idle entry has to do the MSR read. There is a better way to detect this before going idle for the first time which requires to seperate the bug flags: X86_BUG_AMD_E400 - Selects the E400 aware idle routine and enables the detection X86_BUG_AMD_APIC_C1E - Set when the platform is affected by E400 Replace the current X86_BUG_AMD_APIC_C1E usage by the new X86_BUG_AMD_E400 bug bit to select the idle routine which currently does an unconditional detection poll. X86_BUG_AMD_APIC_C1E is going to be used in later patches to remove the MSR polling and simplify the handling of this misfeature. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Jiri Olsa <jolsa@redhat.com> Link: http://lkml.kernel.org/r/20161209182912.2726-3-bp@alien8.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-12-09 19:29:09 +01:00
}
static void init_amd_k8(struct cpuinfo_x86 *c)
{
u32 level;
u64 value;
/* On C+ stepping K8 rep microcode works well for copy/memset */
level = cpuid_eax(1);
if ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58)
set_cpu_cap(c, X86_FEATURE_REP_GOOD);
/*
* Some BIOSes incorrectly force this feature, but only K8 revision D
* (model = 0x14) and later actually support it.
* (AMD Erratum #110, docId: 25759).
*/
x86/cpu: Don't clear X86_FEATURE_LAHF_LM flag in init_amd_k8() on AMD when running in a virtual machine When running in a virtual machine, we might see the original hardware CPU vendor string (i.e. "AuthenticAMD"), but a model and family ID set by the hypervisor. In case we run on AMD hardware and the hypervisor sets a model ID < 0x14, the LAHF cpu feature is eliminated from the the list of CPU capabilities present to circumvent a bug with some BIOSes in conjunction with AMD K8 processors. Parsing the flags list from /proc/cpuinfo seems to be happening mostly in bash scripts and prebuilt Docker containers, as it does not need to have additionals tools present – even though more reliable ways like using "kcpuid", which calls the CPUID instruction instead of parsing a list, should be preferred. Scripts, that use /proc/cpuinfo to determine if the current CPU is "compliant" with defined microarchitecture levels like x86-64-v2 will falsely claim the CPU is incapable of modern CPU instructions when "lahf_lm" is missing in that flags list. This can prevent some docker containers from starting or build scripts to create unoptimized binaries. Admittably, this is more a small inconvenience than a severe bug in the kernel and the shoddy scripts that rely on parsing /proc/cpuinfo should be fixed instead. This patch adds an additional check to see if we're running inside a virtual machine (X86_FEATURE_HYPERVISOR is present), which, to my understanding, can't be present on a real K8 processor as it was introduced only with the later/other Athlon64 models. Example output with the "lahf_lm" flag missing in the flags list (should be shown between "hypervisor" and "abm"): $ cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 15 model : 6 model name : Common KVM processor stepping : 1 microcode : 0x1000065 cpu MHz : 2599.998 cache size : 512 KB physical id : 0 siblings : 1 core id : 0 cpu cores : 1 apicid : 0 initial apicid : 0 fpu : yes fpu_exception : yes cpuid level : 13 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx rdtscp lm rep_good nopl cpuid extd_apicid tsc_known_freq pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c hypervisor abm 3dnowprefetch vmmcall bmi1 avx2 bmi2 xsaveopt ... while kcpuid shows the feature to be present in the CPU: # kcpuid -d | grep lahf lahf_lm - LAHF/SAHF available in 64-bit mode [ mingo: Updated the comment a bit, incorporated Boris's review feedback. ] Signed-off-by: Max Grobecker <max@grobecker.info> Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: linux-kernel@vger.kernel.org Cc: Borislav Petkov <bp@alien8.de>
2025-02-27 21:45:05 +01:00
if (c->x86_model < 0x14 && cpu_has(c, X86_FEATURE_LAHF_LM) && !cpu_has(c, X86_FEATURE_HYPERVISOR)) {
clear_cpu_cap(c, X86_FEATURE_LAHF_LM);
if (!rdmsrq_amd_safe(0xc001100d, &value)) {
value &= ~BIT_64(32);
wrmsrq_amd_safe(0xc001100d, value);
}
}
if (!c->x86_model_id[0])
strscpy(c->x86_model_id, "Hammer");
#ifdef CONFIG_SMP
/*
* Disable TLB flush filter by setting HWCR.FFDIS on K8
* bit 6 of msr C001_0015
*
* Errata 63 for SH-B3 steppings
* Errata 122 for all steppings (F+ have it disabled by default)
*/
msr_set_bit(MSR_K7_HWCR, 6);
#endif
set_cpu_bug(c, X86_BUG_SWAPGS_FENCE);
/*
* Check models and steppings affected by erratum 400. This is
* used to select the proper idle routine and to enable the
* check whether the machine is affected in arch_post_acpi_subsys_init()
* which sets the X86_BUG_AMD_APIC_C1E bug depending on the MSR check.
*/
if (c->x86_model > 0x41 ||
(c->x86_model == 0x41 && c->x86_stepping >= 0x2))
setup_force_cpu_bug(X86_BUG_AMD_E400);
}
static void init_amd_gh(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_MMCONF_FAM10H
/* do this for boot cpu */
if (c == &boot_cpu_data)
check_enable_amd_mmconf_dmi();
fam10h_check_enable_mmcfg();
#endif
/*
* Disable GART TLB Walk Errors on Fam10h. We do this here because this
* is always needed when GART is enabled, even in a kernel which has no
* MCE support built in. BIOS should disable GartTlbWlk Errors already.
* If it doesn't, we do it here as suggested by the BKDG.
*
* Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=33012
*/
msr_set_bit(MSR_AMD64_MCx_MASK(4), 10);
/*
* On family 10h BIOS may not have properly enabled WC+ support, causing
* it to be converted to CD memtype. This may result in performance
* degradation for certain nested-paging guests. Prevent this conversion
* by clearing bit 24 in MSR_AMD64_BU_CFG2.
*
* NOTE: we want to use the _safe accessors so as not to #GP kvm
* guests on older kvm hosts.
*/
msr_clear_bit(MSR_AMD64_BU_CFG2, 24);
set_cpu_bug(c, X86_BUG_AMD_TLB_MMATCH);
/*
* Check models and steppings affected by erratum 400. This is
* used to select the proper idle routine and to enable the
* check whether the machine is affected in arch_post_acpi_subsys_init()
* which sets the X86_BUG_AMD_APIC_C1E bug depending on the MSR check.
*/
if (c->x86_model > 0x2 ||
(c->x86_model == 0x2 && c->x86_stepping >= 0x1))
setup_force_cpu_bug(X86_BUG_AMD_E400);
}
static void init_amd_ln(struct cpuinfo_x86 *c)
{
/*
* Apply erratum 665 fix unconditionally so machines without a BIOS
* fix work.
*/
msr_set_bit(MSR_AMD64_DE_CFG, 31);
}
x86/CPU/AMD: Clear RDRAND CPUID bit on AMD family 15h/16h There have been reports of RDRAND issues after resuming from suspend on some AMD family 15h and family 16h systems. This issue stems from a BIOS not performing the proper steps during resume to ensure RDRAND continues to function properly. RDRAND support is indicated by CPUID Fn00000001_ECX[30]. This bit can be reset by clearing MSR C001_1004[62]. Any software that checks for RDRAND support using CPUID, including the kernel, will believe that RDRAND is not supported. Update the CPU initialization to clear the RDRAND CPUID bit for any family 15h and 16h processor that supports RDRAND. If it is known that the family 15h or family 16h system does not have an RDRAND resume issue or that the system will not be placed in suspend, the "rdrand=force" kernel parameter can be used to stop the clearing of the RDRAND CPUID bit. Additionally, update the suspend and resume path to save and restore the MSR C001_1004 value to ensure that the RDRAND CPUID setting remains in place after resuming from suspend. Note, that clearing the RDRAND CPUID bit does not prevent a processor that normally supports the RDRAND instruction from executing it. So any code that determined the support based on family and model won't #UD. Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Chen Yu <yu.c.chen@intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@chromium.org> Cc: "linux-doc@vger.kernel.org" <linux-doc@vger.kernel.org> Cc: "linux-pm@vger.kernel.org" <linux-pm@vger.kernel.org> Cc: Nathan Chancellor <natechancellor@gmail.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: <stable@vger.kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "x86@kernel.org" <x86@kernel.org> Link: https://lkml.kernel.org/r/7543af91666f491547bd86cebb1e17c66824ab9f.1566229943.git.thomas.lendacky@amd.com
2019-08-19 15:52:35 +00:00
static bool rdrand_force;
static int __init rdrand_cmdline(char *str)
{
if (!str)
return -EINVAL;
if (!strcmp(str, "force"))
rdrand_force = true;
else
return -EINVAL;
return 0;
}
early_param("rdrand", rdrand_cmdline);
static void clear_rdrand_cpuid_bit(struct cpuinfo_x86 *c)
{
/*
* Saving of the MSR used to hide the RDRAND support during
* suspend/resume is done by arch/x86/power/cpu.c, which is
* dependent on CONFIG_PM_SLEEP.
*/
if (!IS_ENABLED(CONFIG_PM_SLEEP))
return;
/*
* The self-test can clear X86_FEATURE_RDRAND, so check for
x86/CPU/AMD: Clear RDRAND CPUID bit on AMD family 15h/16h There have been reports of RDRAND issues after resuming from suspend on some AMD family 15h and family 16h systems. This issue stems from a BIOS not performing the proper steps during resume to ensure RDRAND continues to function properly. RDRAND support is indicated by CPUID Fn00000001_ECX[30]. This bit can be reset by clearing MSR C001_1004[62]. Any software that checks for RDRAND support using CPUID, including the kernel, will believe that RDRAND is not supported. Update the CPU initialization to clear the RDRAND CPUID bit for any family 15h and 16h processor that supports RDRAND. If it is known that the family 15h or family 16h system does not have an RDRAND resume issue or that the system will not be placed in suspend, the "rdrand=force" kernel parameter can be used to stop the clearing of the RDRAND CPUID bit. Additionally, update the suspend and resume path to save and restore the MSR C001_1004 value to ensure that the RDRAND CPUID setting remains in place after resuming from suspend. Note, that clearing the RDRAND CPUID bit does not prevent a processor that normally supports the RDRAND instruction from executing it. So any code that determined the support based on family and model won't #UD. Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Chen Yu <yu.c.chen@intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@chromium.org> Cc: "linux-doc@vger.kernel.org" <linux-doc@vger.kernel.org> Cc: "linux-pm@vger.kernel.org" <linux-pm@vger.kernel.org> Cc: Nathan Chancellor <natechancellor@gmail.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: <stable@vger.kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "x86@kernel.org" <x86@kernel.org> Link: https://lkml.kernel.org/r/7543af91666f491547bd86cebb1e17c66824ab9f.1566229943.git.thomas.lendacky@amd.com
2019-08-19 15:52:35 +00:00
* RDRAND support using the CPUID function directly.
*/
if (!(cpuid_ecx(1) & BIT(30)) || rdrand_force)
return;
msr_clear_bit(MSR_AMD64_CPUID_FN_1, 62);
/*
* Verify that the CPUID change has occurred in case the kernel is
* running virtualized and the hypervisor doesn't support the MSR.
*/
if (cpuid_ecx(1) & BIT(30)) {
pr_info_once("BIOS may not properly restore RDRAND after suspend, but hypervisor does not support hiding RDRAND via CPUID.\n");
return;
}
clear_cpu_cap(c, X86_FEATURE_RDRAND);
pr_info_once("BIOS may not properly restore RDRAND after suspend, hiding RDRAND via CPUID. Use rdrand=force to reenable.\n");
}
static void init_amd_jg(struct cpuinfo_x86 *c)
{
/*
* Some BIOS implementations do not restore proper RDRAND support
* across suspend and resume. Check on whether to hide the RDRAND
* instruction support via CPUID.
*/
clear_rdrand_cpuid_bit(c);
}
static void init_amd_bd(struct cpuinfo_x86 *c)
{
u64 value;
/*
* The way access filter has a performance penalty on some workloads.
* Disable it on the affected CPUs.
*/
if ((c->x86_model >= 0x02) && (c->x86_model < 0x20)) {
if (!rdmsrq_safe(MSR_F15H_IC_CFG, &value) && !(value & 0x1E)) {
value |= 0x1E;
wrmsrq_safe(MSR_F15H_IC_CFG, value);
}
}
x86/CPU/AMD: Clear RDRAND CPUID bit on AMD family 15h/16h There have been reports of RDRAND issues after resuming from suspend on some AMD family 15h and family 16h systems. This issue stems from a BIOS not performing the proper steps during resume to ensure RDRAND continues to function properly. RDRAND support is indicated by CPUID Fn00000001_ECX[30]. This bit can be reset by clearing MSR C001_1004[62]. Any software that checks for RDRAND support using CPUID, including the kernel, will believe that RDRAND is not supported. Update the CPU initialization to clear the RDRAND CPUID bit for any family 15h and 16h processor that supports RDRAND. If it is known that the family 15h or family 16h system does not have an RDRAND resume issue or that the system will not be placed in suspend, the "rdrand=force" kernel parameter can be used to stop the clearing of the RDRAND CPUID bit. Additionally, update the suspend and resume path to save and restore the MSR C001_1004 value to ensure that the RDRAND CPUID setting remains in place after resuming from suspend. Note, that clearing the RDRAND CPUID bit does not prevent a processor that normally supports the RDRAND instruction from executing it. So any code that determined the support based on family and model won't #UD. Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Chen Yu <yu.c.chen@intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@chromium.org> Cc: "linux-doc@vger.kernel.org" <linux-doc@vger.kernel.org> Cc: "linux-pm@vger.kernel.org" <linux-pm@vger.kernel.org> Cc: Nathan Chancellor <natechancellor@gmail.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: <stable@vger.kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "x86@kernel.org" <x86@kernel.org> Link: https://lkml.kernel.org/r/7543af91666f491547bd86cebb1e17c66824ab9f.1566229943.git.thomas.lendacky@amd.com
2019-08-19 15:52:35 +00:00
/*
* Some BIOS implementations do not restore proper RDRAND support
* across suspend and resume. Check on whether to hide the RDRAND
* instruction support via CPUID.
*/
clear_rdrand_cpuid_bit(c);
}
static const struct x86_cpu_id erratum_1386_microcode[] = {
X86_MATCH_VFM_STEPS(VFM_MAKE(X86_VENDOR_AMD, 0x17, 0x01), 0x2, 0x2, 0x0800126e),
X86_MATCH_VFM_STEPS(VFM_MAKE(X86_VENDOR_AMD, 0x17, 0x31), 0x0, 0x0, 0x08301052),
{}
};
static void fix_erratum_1386(struct cpuinfo_x86 *c)
{
/*
* Work around Erratum 1386. The XSAVES instruction malfunctions in
* certain circumstances on Zen1/2 uarch, and not all parts have had
* updated microcode at the time of writing (March 2023).
*
* Affected parts all have no supervisor XSAVE states, meaning that
* the XSAVEC instruction (which works fine) is equivalent.
*
* Clear the feature flag only on microcode revisions which
* don't have the fix.
*/
if (x86_match_min_microcode_rev(erratum_1386_microcode))
return;
clear_cpu_cap(c, X86_FEATURE_XSAVES);
}
void init_spectral_chicken(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_MITIGATION_UNRET_ENTRY
u64 value;
/*
* On Zen2 we offer this chicken (bit) on the altar of Speculation.
*
* This suppresses speculation from the middle of a basic block, i.e. it
* suppresses non-branch predictions.
*/
if (!cpu_has(c, X86_FEATURE_HYPERVISOR)) {
if (!rdmsrq_safe(MSR_ZEN2_SPECTRAL_CHICKEN, &value)) {
value |= MSR_ZEN2_SPECTRAL_CHICKEN_BIT;
wrmsrq_safe(MSR_ZEN2_SPECTRAL_CHICKEN, value);
}
}
#endif
}
static void init_amd_zen_common(void)
{
setup_force_cpu_cap(X86_FEATURE_ZEN);
sched/topology: Improve load balancing on AMD EPYC systems SD_BALANCE_{FORK,EXEC} and SD_WAKE_AFFINE are stripped in sd_init() for any sched domains with a NUMA distance greater than 2 hops (RECLAIM_DISTANCE). The idea being that it's expensive to balance across domains that far apart. However, as is rather unfortunately explained in: commit 32e45ff43eaf ("mm: increase RECLAIM_DISTANCE to 30") the value for RECLAIM_DISTANCE is based on node distance tables from 2011-era hardware. Current AMD EPYC machines have the following NUMA node distances: node distances: node 0 1 2 3 4 5 6 7 0: 10 16 16 16 32 32 32 32 1: 16 10 16 16 32 32 32 32 2: 16 16 10 16 32 32 32 32 3: 16 16 16 10 32 32 32 32 4: 32 32 32 32 10 16 16 16 5: 32 32 32 32 16 10 16 16 6: 32 32 32 32 16 16 10 16 7: 32 32 32 32 16 16 16 10 where 2 hops is 32. The result is that the scheduler fails to load balance properly across NUMA nodes on different sockets -- 2 hops apart. For example, pinning 16 busy threads to NUMA nodes 0 (CPUs 0-7) and 4 (CPUs 32-39) like so, $ numactl -C 0-7,32-39 ./spinner 16 causes all threads to fork and remain on node 0 until the active balancer kicks in after a few seconds and forcibly moves some threads to node 4. Override node_reclaim_distance for AMD Zen. Signed-off-by: Matt Fleming <matt@codeblueprint.co.uk> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Mel Gorman <mgorman@techsingularity.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@surriel.com> Cc: Suravee.Suthikulpanit@amd.com Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Thomas.Lendacky@amd.com Cc: Tony Luck <tony.luck@intel.com> Link: https://lkml.kernel.org/r/20190808195301.13222-3-matt@codeblueprint.co.uk Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-08-08 20:53:01 +01:00
#ifdef CONFIG_NUMA
node_reclaim_distance = 32;
#endif
}
static void init_amd_zen1(struct cpuinfo_x86 *c)
{
fix_erratum_1386(c);
sched/topology: Improve load balancing on AMD EPYC systems SD_BALANCE_{FORK,EXEC} and SD_WAKE_AFFINE are stripped in sd_init() for any sched domains with a NUMA distance greater than 2 hops (RECLAIM_DISTANCE). The idea being that it's expensive to balance across domains that far apart. However, as is rather unfortunately explained in: commit 32e45ff43eaf ("mm: increase RECLAIM_DISTANCE to 30") the value for RECLAIM_DISTANCE is based on node distance tables from 2011-era hardware. Current AMD EPYC machines have the following NUMA node distances: node distances: node 0 1 2 3 4 5 6 7 0: 10 16 16 16 32 32 32 32 1: 16 10 16 16 32 32 32 32 2: 16 16 10 16 32 32 32 32 3: 16 16 16 10 32 32 32 32 4: 32 32 32 32 10 16 16 16 5: 32 32 32 32 16 10 16 16 6: 32 32 32 32 16 16 10 16 7: 32 32 32 32 16 16 16 10 where 2 hops is 32. The result is that the scheduler fails to load balance properly across NUMA nodes on different sockets -- 2 hops apart. For example, pinning 16 busy threads to NUMA nodes 0 (CPUs 0-7) and 4 (CPUs 32-39) like so, $ numactl -C 0-7,32-39 ./spinner 16 causes all threads to fork and remain on node 0 until the active balancer kicks in after a few seconds and forcibly moves some threads to node 4. Override node_reclaim_distance for AMD Zen. Signed-off-by: Matt Fleming <matt@codeblueprint.co.uk> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Mel Gorman <mgorman@techsingularity.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@surriel.com> Cc: Suravee.Suthikulpanit@amd.com Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Thomas.Lendacky@amd.com Cc: Tony Luck <tony.luck@intel.com> Link: https://lkml.kernel.org/r/20190808195301.13222-3-matt@codeblueprint.co.uk Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-08-08 20:53:01 +01:00
/* Fix up CPUID bits, but only if not virtualised. */
if (!cpu_has(c, X86_FEATURE_HYPERVISOR)) {
/* Erratum 1076: CPB feature bit not being set in CPUID. */
if (!cpu_has(c, X86_FEATURE_CPB))
set_cpu_cap(c, X86_FEATURE_CPB);
}
pr_notice_once("AMD Zen1 DIV0 bug detected. Disable SMT for full protection.\n");
setup_force_cpu_bug(X86_BUG_DIV0);
/*
* Turn off the Instructions Retired free counter on machines that are
* susceptible to erratum #1054 "Instructions Retired Performance
* Counter May Be Inaccurate".
*/
if (c->x86_model < 0x30) {
msr_clear_bit(MSR_K7_HWCR, MSR_K7_HWCR_IRPERF_EN_BIT);
clear_cpu_cap(c, X86_FEATURE_IRPERF);
}
}
static bool cpu_has_zenbleed_microcode(void)
{
u32 good_rev = 0;
switch (boot_cpu_data.x86_model) {
case 0x30 ... 0x3f: good_rev = 0x0830107b; break;
case 0x60 ... 0x67: good_rev = 0x0860010c; break;
case 0x68 ... 0x6f: good_rev = 0x08608107; break;
case 0x70 ... 0x7f: good_rev = 0x08701033; break;
case 0xa0 ... 0xaf: good_rev = 0x08a00009; break;
default:
return false;
}
if (boot_cpu_data.microcode < good_rev)
return false;
return true;
}
static void zen2_zenbleed_check(struct cpuinfo_x86 *c)
{
if (cpu_has(c, X86_FEATURE_HYPERVISOR))
return;
if (!cpu_has(c, X86_FEATURE_AVX))
return;
if (!cpu_has_zenbleed_microcode()) {
pr_notice_once("Zenbleed: please update your microcode for the most optimal fix\n");
msr_set_bit(MSR_AMD64_DE_CFG, MSR_AMD64_DE_CFG_ZEN2_FP_BACKUP_FIX_BIT);
} else {
msr_clear_bit(MSR_AMD64_DE_CFG, MSR_AMD64_DE_CFG_ZEN2_FP_BACKUP_FIX_BIT);
}
}
static void init_amd_zen2(struct cpuinfo_x86 *c)
{
init_spectral_chicken(c);
fix_erratum_1386(c);
zen2_zenbleed_check(c);
/* Disable RDSEED on AMD Cyan Skillfish because of an error. */
if (c->x86_model == 0x47 && c->x86_stepping == 0x0) {
clear_cpu_cap(c, X86_FEATURE_RDSEED);
msr_clear_bit(MSR_AMD64_CPUID_FN_7, 18);
pr_emerg("RDSEED is not reliable on this platform; disabling.\n");
}
/* Correct misconfigured CPUID on some clients. */
clear_cpu_cap(c, X86_FEATURE_INVLPGB);
}
static void init_amd_zen3(struct cpuinfo_x86 *c)
{
if (!cpu_has(c, X86_FEATURE_HYPERVISOR)) {
/*
* Zen3 (Fam19 model < 0x10) parts are not susceptible to
* Branch Type Confusion, but predate the allocation of the
* BTC_NO bit.
*/
if (!cpu_has(c, X86_FEATURE_BTC_NO))
set_cpu_cap(c, X86_FEATURE_BTC_NO);
}
}
static void init_amd_zen4(struct cpuinfo_x86 *c)
{
if (!cpu_has(c, X86_FEATURE_HYPERVISOR))
msr_set_bit(MSR_ZEN4_BP_CFG, MSR_ZEN4_BP_CFG_SHARED_BTB_FIX_BIT);
/*
* These Zen4 SoCs advertise support for virtualized VMLOAD/VMSAVE
* in some BIOS versions but they can lead to random host reboots.
*/
switch (c->x86_model) {
case 0x18 ... 0x1f:
case 0x60 ... 0x7f:
clear_cpu_cap(c, X86_FEATURE_V_VMSAVE_VMLOAD);
break;
}
}
static void init_amd_zen5(struct cpuinfo_x86 *c)
{
}
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void init_amd(struct cpuinfo_x86 *c)
{
u64 vm_cr;
early_init_amd(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);
if (c->x86 >= 0x10)
set_cpu_cap(c, X86_FEATURE_REP_GOOD);
/* AMD FSRM also implies FSRS */
if (cpu_has(c, X86_FEATURE_FSRM))
set_cpu_cap(c, X86_FEATURE_FSRS);
/* K6s reports MCEs but don't actually have all the MSRs */
if (c->x86 < 6)
clear_cpu_cap(c, X86_FEATURE_MCE);
switch (c->x86) {
case 4: init_amd_k5(c); break;
case 5: init_amd_k6(c); break;
case 6: init_amd_k7(c); break;
case 0xf: init_amd_k8(c); break;
case 0x10: init_amd_gh(c); break;
case 0x12: init_amd_ln(c); break;
case 0x15: init_amd_bd(c); break;
x86/CPU/AMD: Clear RDRAND CPUID bit on AMD family 15h/16h There have been reports of RDRAND issues after resuming from suspend on some AMD family 15h and family 16h systems. This issue stems from a BIOS not performing the proper steps during resume to ensure RDRAND continues to function properly. RDRAND support is indicated by CPUID Fn00000001_ECX[30]. This bit can be reset by clearing MSR C001_1004[62]. Any software that checks for RDRAND support using CPUID, including the kernel, will believe that RDRAND is not supported. Update the CPU initialization to clear the RDRAND CPUID bit for any family 15h and 16h processor that supports RDRAND. If it is known that the family 15h or family 16h system does not have an RDRAND resume issue or that the system will not be placed in suspend, the "rdrand=force" kernel parameter can be used to stop the clearing of the RDRAND CPUID bit. Additionally, update the suspend and resume path to save and restore the MSR C001_1004 value to ensure that the RDRAND CPUID setting remains in place after resuming from suspend. Note, that clearing the RDRAND CPUID bit does not prevent a processor that normally supports the RDRAND instruction from executing it. So any code that determined the support based on family and model won't #UD. Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Chen Yu <yu.c.chen@intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kees Cook <keescook@chromium.org> Cc: "linux-doc@vger.kernel.org" <linux-doc@vger.kernel.org> Cc: "linux-pm@vger.kernel.org" <linux-pm@vger.kernel.org> Cc: Nathan Chancellor <natechancellor@gmail.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: <stable@vger.kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "x86@kernel.org" <x86@kernel.org> Link: https://lkml.kernel.org/r/7543af91666f491547bd86cebb1e17c66824ab9f.1566229943.git.thomas.lendacky@amd.com
2019-08-19 15:52:35 +00:00
case 0x16: init_amd_jg(c); break;
}
/*
* Save up on some future enablement work and do common Zen
* settings.
*/
if (c->x86 >= 0x17)
init_amd_zen_common();
if (boot_cpu_has(X86_FEATURE_ZEN1))
init_amd_zen1(c);
else if (boot_cpu_has(X86_FEATURE_ZEN2))
init_amd_zen2(c);
else if (boot_cpu_has(X86_FEATURE_ZEN3))
init_amd_zen3(c);
else if (boot_cpu_has(X86_FEATURE_ZEN4))
init_amd_zen4(c);
else if (boot_cpu_has(X86_FEATURE_ZEN5))
init_amd_zen5(c);
/*
* Enable workaround for FXSAVE leak on CPUs
* without a XSaveErPtr feature
*/
if ((c->x86 >= 6) && (!cpu_has(c, X86_FEATURE_XSAVEERPTR)))
set_cpu_bug(c, X86_BUG_FXSAVE_LEAK);
cpu_detect_cache_sizes(c);
srat_detect_node(c);
init_amd_cacheinfo(c);
if (cpu_has(c, X86_FEATURE_SVM)) {
rdmsrq(MSR_VM_CR, vm_cr);
if (vm_cr & SVM_VM_CR_SVM_DIS_MASK) {
pr_notice_once("SVM disabled (by BIOS) in MSR_VM_CR\n");
clear_cpu_cap(c, X86_FEATURE_SVM);
}
}
if (!cpu_has(c, X86_FEATURE_LFENCE_RDTSC) && cpu_has(c, X86_FEATURE_XMM2)) {
/*
x86: Remove X86_FEATURE_MFENCE_RDTSC AMD and Intel both have serializing lfence (X86_FEATURE_LFENCE_RDTSC). They've both had it for a long time, and AMD has had it enabled in Linux since Spectre v1 was announced. Back then, there was a proposal to remove the serializing mfence feature bit (X86_FEATURE_MFENCE_RDTSC), since both AMD and Intel have serializing lfence. At the time, it was (ahem) speculated that some hypervisors might not yet support its removal, so it remained for the time being. Now a year-and-a-half later, it should be safe to remove. I asked Andrew Cooper about whether it's still needed: So if you're virtualised, you've got no choice in the matter.  lfence is either dispatch-serialising or not on AMD, and you won't be able to change it. Furthermore, you can't accurately tell what state the bit is in, because the MSR might not be virtualised at all, or may not reflect the true state in hardware.  Worse still, attempting to set the bit may not be successful even if there isn't a fault for doing so. Xen sets the DE_CFG bit unconditionally, as does Linux by the looks of things (see MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT).  ISTR other hypervisor vendors saying the same, but I don't have any information to hand. If you are running under a hypervisor which has been updated, then lfence will almost certainly be dispatch-serialising in practice, and you'll almost certainly see the bit already set in DE_CFG.  If you're running under a hypervisor which hasn't been patched since Spectre, you've already lost in many more ways. I'd argue that X86_FEATURE_MFENCE_RDTSC is not worth keeping. So remove it. This will reduce some code rot, and also make it easier to hook barrier_nospec() up to a cmdline disable for performance raisins, without having to need an alternative_3() macro. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/d990aa51e40063acb9888e8c1b688e41355a9588.1562255067.git.jpoimboe@redhat.com
2019-07-04 10:46:37 -05:00
* Use LFENCE 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_AMD64_DE_CFG,
MSR_AMD64_DE_CFG_LFENCE_SERIALIZE_BIT);
x86: Remove X86_FEATURE_MFENCE_RDTSC AMD and Intel both have serializing lfence (X86_FEATURE_LFENCE_RDTSC). They've both had it for a long time, and AMD has had it enabled in Linux since Spectre v1 was announced. Back then, there was a proposal to remove the serializing mfence feature bit (X86_FEATURE_MFENCE_RDTSC), since both AMD and Intel have serializing lfence. At the time, it was (ahem) speculated that some hypervisors might not yet support its removal, so it remained for the time being. Now a year-and-a-half later, it should be safe to remove. I asked Andrew Cooper about whether it's still needed: So if you're virtualised, you've got no choice in the matter.  lfence is either dispatch-serialising or not on AMD, and you won't be able to change it. Furthermore, you can't accurately tell what state the bit is in, because the MSR might not be virtualised at all, or may not reflect the true state in hardware.  Worse still, attempting to set the bit may not be successful even if there isn't a fault for doing so. Xen sets the DE_CFG bit unconditionally, as does Linux by the looks of things (see MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT).  ISTR other hypervisor vendors saying the same, but I don't have any information to hand. If you are running under a hypervisor which has been updated, then lfence will almost certainly be dispatch-serialising in practice, and you'll almost certainly see the bit already set in DE_CFG.  If you're running under a hypervisor which hasn't been patched since Spectre, you've already lost in many more ways. I'd argue that X86_FEATURE_MFENCE_RDTSC is not worth keeping. So remove it. This will reduce some code rot, and also make it easier to hook barrier_nospec() up to a cmdline disable for performance raisins, without having to need an alternative_3() macro. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/d990aa51e40063acb9888e8c1b688e41355a9588.1562255067.git.jpoimboe@redhat.com
2019-07-04 10:46:37 -05:00
/* A serializing LFENCE stops RDTSC speculation */
set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC);
}
/*
* Family 0x12 and above processors have APIC timer
* running in deep C states.
*/
if (c->x86 > 0x11)
set_cpu_cap(c, X86_FEATURE_ARAT);
/* 3DNow or LM implies PREFETCHW */
if (!cpu_has(c, X86_FEATURE_3DNOWPREFETCH))
if (cpu_has(c, X86_FEATURE_3DNOW) || cpu_has(c, X86_FEATURE_LM))
set_cpu_cap(c, X86_FEATURE_3DNOWPREFETCH);
/* AMD CPUs don't reset SS attributes on SYSRET, Xen does. */
if (!cpu_feature_enabled(X86_FEATURE_XENPV))
set_cpu_bug(c, X86_BUG_SYSRET_SS_ATTRS);
x86/cpu/amd: Enable the fixed Instructions Retired counter IRPERF Commit aaf248848db50 ("perf/x86/msr: Add AMD IRPERF (Instructions Retired) performance counter") added support for access to the free-running counter via 'perf -e msr/irperf/', but when exercised, it always returns a 0 count: BEFORE: $ perf stat -e instructions,msr/irperf/ true Performance counter stats for 'true': 624,833 instructions 0 msr/irperf/ Simply set its enable bit - HWCR bit 30 - to make it start counting. Enablement is restricted to all machines advertising IRPERF capability, except those susceptible to an erratum that makes the IRPERF return bad values. That erratum occurs in Family 17h models 00-1fh [1], but not in F17h models 20h and above [2]. AFTER (on a family 17h model 31h machine): $ perf stat -e instructions,msr/irperf/ true Performance counter stats for 'true': 621,690 instructions 622,490 msr/irperf/ [1] Revision Guide for AMD Family 17h Models 00h-0Fh Processors [2] Revision Guide for AMD Family 17h Models 30h-3Fh Processors The revision guides are available from the bugzilla Link below. [ bp: Massage commit message. ] Fixes: aaf248848db50 ("perf/x86/msr: Add AMD IRPERF (Instructions Retired) performance counter") Signed-off-by: Kim Phillips <kim.phillips@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537 Link: http://lkml.kernel.org/r/20200214201805.13830-1-kim.phillips@amd.com
2020-02-19 18:52:43 +01:00
/* Enable the Instructions Retired free counter */
if (cpu_has(c, X86_FEATURE_IRPERF))
x86/cpu/amd: Enable the fixed Instructions Retired counter IRPERF Commit aaf248848db50 ("perf/x86/msr: Add AMD IRPERF (Instructions Retired) performance counter") added support for access to the free-running counter via 'perf -e msr/irperf/', but when exercised, it always returns a 0 count: BEFORE: $ perf stat -e instructions,msr/irperf/ true Performance counter stats for 'true': 624,833 instructions 0 msr/irperf/ Simply set its enable bit - HWCR bit 30 - to make it start counting. Enablement is restricted to all machines advertising IRPERF capability, except those susceptible to an erratum that makes the IRPERF return bad values. That erratum occurs in Family 17h models 00-1fh [1], but not in F17h models 20h and above [2]. AFTER (on a family 17h model 31h machine): $ perf stat -e instructions,msr/irperf/ true Performance counter stats for 'true': 621,690 instructions 622,490 msr/irperf/ [1] Revision Guide for AMD Family 17h Models 00h-0Fh Processors [2] Revision Guide for AMD Family 17h Models 30h-3Fh Processors The revision guides are available from the bugzilla Link below. [ bp: Massage commit message. ] Fixes: aaf248848db50 ("perf/x86/msr: Add AMD IRPERF (Instructions Retired) performance counter") Signed-off-by: Kim Phillips <kim.phillips@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537 Link: http://lkml.kernel.org/r/20200214201805.13830-1-kim.phillips@amd.com
2020-02-19 18:52:43 +01:00
msr_set_bit(MSR_K7_HWCR, MSR_K7_HWCR_IRPERF_EN_BIT);
check_null_seg_clears_base(c);
/*
* Make sure EFER[AIBRSE - Automatic IBRS Enable] is set. The APs are brought up
* using the trampoline code and as part of it, MSR_EFER gets prepared there in
* order to be replicated onto them. Regardless, set it here again, if not set,
* to protect against any future refactoring/code reorganization which might
* miss setting this important bit.
*/
if (spectre_v2_in_eibrs_mode(spectre_v2_enabled) &&
cpu_has(c, X86_FEATURE_AUTOIBRS))
WARN_ON_ONCE(msr_set_bit(MSR_EFER, _EFER_AUTOIBRS) < 0);
x86/barrier: Do not serialize MSR accesses on AMD AMD does not have the requirement for a synchronization barrier when acccessing a certain group of MSRs. Do not incur that unnecessary penalty there. There will be a CPUID bit which explicitly states that a MFENCE is not needed. Once that bit is added to the APM, this will be extended with it. While at it, move to processor.h to avoid include hell. Untangling that file properly is a matter for another day. Some notes on the performance aspect of why this is relevant, courtesy of Kishon VijayAbraham <Kishon.VijayAbraham@amd.com>: On a AMD Zen4 system with 96 cores, a modified ipi-bench[1] on a VM shows x2AVIC IPI rate is 3% to 4% lower than AVIC IPI rate. The ipi-bench is modified so that the IPIs are sent between two vCPUs in the same CCX. This also requires to pin the vCPU to a physical core to prevent any latencies. This simulates the use case of pinning vCPUs to the thread of a single CCX to avoid interrupt IPI latency. In order to avoid run-to-run variance (for both x2AVIC and AVIC), the below configurations are done: 1) Disable Power States in BIOS (to prevent the system from going to lower power state) 2) Run the system at fixed frequency 2500MHz (to prevent the system from increasing the frequency when the load is more) With the above configuration: *) Performance measured using ipi-bench for AVIC: Average Latency: 1124.98ns [Time to send IPI from one vCPU to another vCPU] Cumulative throughput: 42.6759M/s [Total number of IPIs sent in a second from 48 vCPUs simultaneously] *) Performance measured using ipi-bench for x2AVIC: Average Latency: 1172.42ns [Time to send IPI from one vCPU to another vCPU] Cumulative throughput: 40.9432M/s [Total number of IPIs sent in a second from 48 vCPUs simultaneously] From above, x2AVIC latency is ~4% more than AVIC. However, the expectation is x2AVIC performance to be better or equivalent to AVIC. Upon analyzing the perf captures, it is observed significant time is spent in weak_wrmsr_fence() invoked by x2apic_send_IPI(). With the fix to skip weak_wrmsr_fence() *) Performance measured using ipi-bench for x2AVIC: Average Latency: 1117.44ns [Time to send IPI from one vCPU to another vCPU] Cumulative throughput: 42.9608M/s [Total number of IPIs sent in a second from 48 vCPUs simultaneously] Comparing the performance of x2AVIC with and without the fix, it can be seen the performance improves by ~4%. Performance captured using an unmodified ipi-bench using the 'mesh-ipi' option with and without weak_wrmsr_fence() on a Zen4 system also showed significant performance improvement without weak_wrmsr_fence(). The 'mesh-ipi' option ignores CCX or CCD and just picks random vCPU. Average throughput (10 iterations) with weak_wrmsr_fence(), Cumulative throughput: 4933374 IPI/s Average throughput (10 iterations) without weak_wrmsr_fence(), Cumulative throughput: 6355156 IPI/s [1] https://github.com/bytedance/kvm-utils/tree/master/microbenchmark/ipi-bench Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20230622095212.20940-1-bp@alien8.de
2023-10-27 14:24:16 +02:00
/* AMD CPUs don't need fencing after x2APIC/TSC_DEADLINE MSR writes. */
clear_cpu_cap(c, X86_FEATURE_APIC_MSRS_FENCE);
/* Enable Translation Cache Extension */
if (cpu_has(c, X86_FEATURE_TCE))
msr_set_bit(MSR_EFER, _EFER_TCE);
}
#ifdef CONFIG_X86_32
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static unsigned int amd_size_cache(struct cpuinfo_x86 *c, unsigned int size)
{
/* AMD errata T13 (order #21922) */
if (c->x86 == 6) {
/* Duron Rev A0 */
if (c->x86_model == 3 && c->x86_stepping == 0)
size = 64;
/* Tbird rev A1/A2 */
if (c->x86_model == 4 &&
(c->x86_stepping == 0 || c->x86_stepping == 1))
size = 256;
}
return size;
}
#endif
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static void cpu_detect_tlb_amd(struct cpuinfo_x86 *c)
{
u32 ebx, eax, ecx, edx;
u16 mask = 0xfff;
if (c->x86 < 0xf)
return;
if (c->extended_cpuid_level < 0x80000006)
return;
cpuid(0x80000006, &eax, &ebx, &ecx, &edx);
tlb_lld_4k = (ebx >> 16) & mask;
tlb_lli_4k = ebx & mask;
/*
* K8 doesn't have 2M/4M entries in the L2 TLB so read out the L1 TLB
* characteristics from the CPUID function 0x80000005 instead.
*/
if (c->x86 == 0xf) {
cpuid(0x80000005, &eax, &ebx, &ecx, &edx);
mask = 0xff;
}
/* Handle DTLB 2M and 4M sizes, fall back to L1 if L2 is disabled */
if (!((eax >> 16) & mask))
tlb_lld_2m = (cpuid_eax(0x80000005) >> 16) & 0xff;
else
tlb_lld_2m = (eax >> 16) & mask;
/* a 4M entry uses two 2M entries */
tlb_lld_4m = tlb_lld_2m >> 1;
/* Handle ITLB 2M and 4M sizes, fall back to L1 if L2 is disabled */
if (!(eax & mask)) {
/* Erratum 658 */
if (c->x86 == 0x15 && c->x86_model <= 0x1f) {
tlb_lli_2m = 1024;
} else {
cpuid(0x80000005, &eax, &ebx, &ecx, &edx);
tlb_lli_2m = eax & 0xff;
}
} else
tlb_lli_2m = eax & mask;
tlb_lli_4m = tlb_lli_2m >> 1;
/* Max number of pages INVLPGB can invalidate in one shot */
if (cpu_has(c, X86_FEATURE_INVLPGB))
invlpgb_count_max = (cpuid_edx(0x80000008) & 0xffff) + 1;
}
x86: delete __cpuinit usage from all x86 files The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/x86 uses of the __cpuinit macros from all C files. x86 only had the one __CPUINIT used in assembly files, and it wasn't paired off with a .previous or a __FINIT, so we can delete it directly w/o any corresponding additional change there. [1] https://lkml.org/lkml/2013/5/20/589 Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: x86@kernel.org Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-18 18:23:59 -04:00
static const struct cpu_dev amd_cpu_dev = {
.c_vendor = "AMD",
.c_ident = { "AuthenticAMD" },
#ifdef CONFIG_X86_32
.legacy_models = {
{ .family = 4, .model_names =
{
[3] = "486 DX/2",
[7] = "486 DX/2-WB",
[8] = "486 DX/4",
[9] = "486 DX/4-WB",
[14] = "Am5x86-WT",
[15] = "Am5x86-WB"
}
},
},
.legacy_cache_size = amd_size_cache,
#endif
.c_early_init = early_init_amd,
.c_detect_tlb = cpu_detect_tlb_amd,
.c_bsp_init = bsp_init_amd,
.c_init = init_amd,
.c_x86_vendor = X86_VENDOR_AMD,
};
cpu_dev_register(amd_cpu_dev);
static DEFINE_PER_CPU_READ_MOSTLY(unsigned long[4], amd_dr_addr_mask);
static unsigned int amd_msr_dr_addr_masks[] = {
MSR_F16H_DR0_ADDR_MASK,
MSR_F16H_DR1_ADDR_MASK,
MSR_F16H_DR1_ADDR_MASK + 1,
MSR_F16H_DR1_ADDR_MASK + 2
};
void amd_set_dr_addr_mask(unsigned long mask, unsigned int dr)
{
int cpu = smp_processor_id();
if (!cpu_feature_enabled(X86_FEATURE_BPEXT))
return;
if (WARN_ON_ONCE(dr >= ARRAY_SIZE(amd_msr_dr_addr_masks)))
return;
if (per_cpu(amd_dr_addr_mask, cpu)[dr] == mask)
return;
wrmsrq(amd_msr_dr_addr_masks[dr], mask);
per_cpu(amd_dr_addr_mask, cpu)[dr] = mask;
}
unsigned long amd_get_dr_addr_mask(unsigned int dr)
{
if (!cpu_feature_enabled(X86_FEATURE_BPEXT))
return 0;
if (WARN_ON_ONCE(dr >= ARRAY_SIZE(amd_msr_dr_addr_masks)))
return 0;
return per_cpu(amd_dr_addr_mask[dr], smp_processor_id());
}
EXPORT_SYMBOL_GPL(amd_get_dr_addr_mask);
static void zenbleed_check_cpu(void *unused)
{
struct cpuinfo_x86 *c = &cpu_data(smp_processor_id());
zen2_zenbleed_check(c);
}
void amd_check_microcode(void)
{
if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
return;
if (cpu_feature_enabled(X86_FEATURE_ZEN2))
on_each_cpu(zenbleed_check_cpu, NULL, 1);
}
x86/CPU/AMD: Print the reason for the last reset The following register contains bits that indicate the cause for the previous reset. PMx000000C0 (FCH::PM::S5_RESET_STATUS) This is useful for debug. The reasons for reset are broken into 6 high level categories. Decode it by category and print during boot. Specifics within a category are split off into debugging documentation. The register is accessed indirectly through a "PM" port in the FCH. Use MMIO access in order to avoid restrictions with legacy port access. Use a late_initcall() to ensure that MMIO has been set up before trying to access the register. This register was introduced with AMD Family 17h, so avoid access on older families. There is no CPUID feature bit for this register. [ bp: Simplify the reason dumping loop. - merge a fix to not access an array element after the last one: https://lore.kernel.org/r/20250505133609.83933-1-superm1@kernel.org Reported-by: James Dutton <james.dutton@gmail.com> ] [ mingo: - Use consistent .rst formatting - Fix 'Sleep' class field to 'ACPI-State' - Standardize pin messages around the 'tripped' verbiage - Remove reference to ring-buffer printing & simplify the wording - Use curly braces for multi-line conditional statements ] Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Co-developed-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/20250422234830.2840784-6-superm1@kernel.org
2025-04-22 18:48:30 -05:00
static const char * const s5_reset_reason_txt[] = {
[0] = "thermal pin BP_THERMTRIP_L was tripped",
[1] = "power button was pressed for 4 seconds",
[2] = "shutdown pin was tripped",
[4] = "remote ASF power off command was received",
[9] = "internal CPU thermal limit was tripped",
[16] = "system reset pin BP_SYS_RST_L was tripped",
[17] = "software issued PCI reset",
[18] = "software wrote 0x4 to reset control register 0xCF9",
[19] = "software wrote 0x6 to reset control register 0xCF9",
[20] = "software wrote 0xE to reset control register 0xCF9",
[21] = "ACPI power state transition occurred",
[22] = "keyboard reset pin KB_RST_L was tripped",
[23] = "internal CPU shutdown event occurred",
[24] = "system failed to boot before failed boot timer expired",
[25] = "hardware watchdog timer expired",
[26] = "remote ASF reset command was received",
[27] = "an uncorrected error caused a data fabric sync flood event",
[29] = "FCH and MP1 failed warm reset handshake",
[30] = "a parity error occurred",
[31] = "a software sync flood event occurred",
};
static __init int print_s5_reset_status_mmio(void)
{
unsigned long value;
void __iomem *addr;
int i;
if (!cpu_feature_enabled(X86_FEATURE_ZEN))
return 0;
addr = ioremap(FCH_PM_BASE + FCH_PM_S5_RESET_STATUS, sizeof(value));
if (!addr)
return 0;
value = ioread32(addr);
iounmap(addr);
for (i = 0; i < ARRAY_SIZE(s5_reset_reason_txt); i++) {
if (!(value & BIT(i)))
continue;
if (s5_reset_reason_txt[i]) {
pr_info("x86/amd: Previous system reset reason [0x%08lx]: %s\n",
value, s5_reset_reason_txt[i]);
}
}
return 0;
}
late_initcall(print_s5_reset_status_mmio);