Re: [PATCH 11/16] x86/cpu: Print VMX features as separate line item in /proc/cpuinfo

[Sean Christopherson]

On Tue, Oct 08, 2019 at 08:57:30AM +0200, Paolo Bonzini wrote:
> On 07/10/19 21:56, Sean Christopherson wrote:
> > On Mon, Oct 07, 2019 at 07:12:37PM +0200, Paolo Bonzini wrote:
> >> On 04/10/19 23:56, Sean Christopherson wrote:
> >>> diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
> >>> index cb2e49810d68..4eec8889b0ff 100644
> >>> --- a/arch/x86/kernel/cpu/proc.c
> >>> +++ b/arch/x86/kernel/cpu/proc.c
> >>> @@ -7,6 +7,10 @@
> >>>  
> >>>  #include "cpu.h"
> >>>  
> >>> +#ifdef CONFIG_X86_VMX_FEATURE_NAMES
> >>> +extern const char * const x86_vmx_flags[NVMXINTS*32];
> >>> +#endif
> >>> +
> >>>  /*
> >>>   *	Get CPU information for use by the procfs.
> >>>   */
> >>> @@ -102,6 +106,17 @@ static int show_cpuinfo(struct seq_file *m, void *v)
> >>>  		if (cpu_has(c, i) && x86_cap_flags[i] != NULL)
> >>>  			seq_printf(m, " %s", x86_cap_flags[i]);
> >>
> >> I'm afraid this is going to break some scripts in the wild.  I would
> >> simply remove the seq_puts below.
> > 
> > Can you elaborate?  I'm having trouble connecting the dots...
> 
> Somebody is bound to have scripts doing "grep ^flags.*ept /proc/cpuinfo"
> or checking for VMX flags under some kind of "if (/^flags/)", so it's
> safer not to separate VMX and non-VMX flags.

Are the names of the flags considered ABI?  If so, then the rename of
"vnmi" to "virtual_nmis" also needs to be dropped.  :-(