On Wed, 12 Sep 2018, Jiri Kosina wrote:
> case X86_BUG_SPECTRE_V2:
> - return sprintf(buf, "%s%s%s%s\n",
> spectre_v2_strings[spectre_v2_enabled],
> + mutex_lock(&spec_ctrl_mutex);
> + ret = sprintf(buf, "%s%s%s%s%s\n",
> spectre_v2_strings[spectre_v2_enabled],
> boot_cpu_has(X86_FEATURE_USE_IBPB) ? ", IBPB" :
> "",
> boot_cpu_has(X86_FEATURE_USE_IBRS_FW) ? ",
> IBRS_FW" : "",
> + (x86_spec_ctrl_base & SPEC_CTRL_STIBP) ? ",
> STIBP" : "",
> spectre_v2_module_string());
> + mutex_unlock(&spec_ctrl_mutex);
The mutex for this printing is overkill. It's a read after all and if there
is a concurrent SMT control fiddling going on then you have a chance of
getting the wrong information as well. I'll zap it.
Thanks,
tglx
