Commit 9c6a73c7 authored by Tom Lendacky's avatar Tom Lendacky Committed by Thomas Gleixner

x86/cpu/AMD: Use LFENCE_RDTSC in preference to MFENCE_RDTSC

With LFENCE now a serializing instruction, use LFENCE_RDTSC in preference
to MFENCE_RDTSC.  However, since the kernel could be running under a
hypervisor that does not support writing that MSR, read the MSR back and
verify that the bit has been set successfully.  If the MSR can be read
and the bit is set, then set the LFENCE_RDTSC feature, otherwise set the
MFENCE_RDTSC feature.
Signed-off-by: default avatarTom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: default avatarThomas Gleixner <tglx@linutronix.de>
Reviewed-by: default avatarReviewed-by: Borislav Petkov <bp@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@linux-foundation.org>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Paul Turner <pjt@google.com>
Link: https://lkml.kernel.org/r/20180108220932.12580.52458.stgit@tlendack-t1.amdoffice.net
parent e4d0e84e
...@@ -354,6 +354,7 @@ ...@@ -354,6 +354,7 @@
#define MSR_FAM10H_NODE_ID 0xc001100c #define MSR_FAM10H_NODE_ID 0xc001100c
#define MSR_F10H_DECFG 0xc0011029 #define MSR_F10H_DECFG 0xc0011029
#define MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT 1 #define MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT 1
#define MSR_F10H_DECFG_LFENCE_SERIALIZE BIT_ULL(MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT)
/* K8 MSRs */ /* K8 MSRs */
#define MSR_K8_TOP_MEM1 0xc001001a #define MSR_K8_TOP_MEM1 0xc001001a
......
...@@ -829,6 +829,9 @@ static void init_amd(struct cpuinfo_x86 *c) ...@@ -829,6 +829,9 @@ static void init_amd(struct cpuinfo_x86 *c)
set_cpu_cap(c, X86_FEATURE_K8); set_cpu_cap(c, X86_FEATURE_K8);
if (cpu_has(c, X86_FEATURE_XMM2)) { if (cpu_has(c, X86_FEATURE_XMM2)) {
unsigned long long val;
int ret;
/* /*
* A serializing LFENCE has less overhead than MFENCE, so * A serializing LFENCE has less overhead than MFENCE, so
* use it for execution serialization. On families which * use it for execution serialization. On families which
...@@ -839,8 +842,19 @@ static void init_amd(struct cpuinfo_x86 *c) ...@@ -839,8 +842,19 @@ static void init_amd(struct cpuinfo_x86 *c)
msr_set_bit(MSR_F10H_DECFG, msr_set_bit(MSR_F10H_DECFG,
MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT); MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT);
/* MFENCE stops RDTSC speculation */ /*
set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC); * Verify that the MSR write was successful (could be running
* under a hypervisor) and only then assume that LFENCE is
* serializing.
*/
ret = rdmsrl_safe(MSR_F10H_DECFG, &val);
if (!ret && (val & MSR_F10H_DECFG_LFENCE_SERIALIZE)) {
/* A serializing LFENCE stops RDTSC speculation */
set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC);
} else {
/* MFENCE stops RDTSC speculation */
set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC);
}
} }
/* /*
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment