linux-kernel - Re: [PATCH] tools/power turbostat: Fix MSRs with CONFIG

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <a813fd9a41dcfc0afc8b68f9ac8d3e07d0b96530.camel@intel.com>
Date: Wed, 18 Jun 2025 03:28:18 +0000
From: "Zhang, Rui" <rui.zhang@...el.com>
To: "calvin@...nvd.org" <calvin@...nvd.org>
CC: "linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"lenb@...nel.org" <lenb@...nel.org>
Subject: Re: [PATCH] tools/power turbostat: Fix MSRs with CONFIG_MULTIUSER=n

On Sun, 2025-06-15 at 21:38 -0700, Calvin Owens wrote:
> On Monday 06/16 at 01:30 +0000, Zhang, Rui wrote:
> > On Fri, 2025-06-13 at 19:20 -0700, Calvin Owens wrote:
> > > Handle ENOSYS from cap_get_proc() in check_for_cap_sys_rawio(), so
> > > turbostat can display temperatures when running on kernels compiled
> > > without multiuser support.
> > > 
> > > Signed-off-by: Calvin Owens <calvin@...nvd.org>
> > > ---
> > >  tools/power/x86/turbostat/turbostat.c | 7 ++++++-
> > >  1 file changed, 6 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/tools/power/x86/turbostat/turbostat.c
> > > b/tools/power/x86/turbostat/turbostat.c
> > > index 925556b90770..f7d665913a52 100644
> > > --- a/tools/power/x86/turbostat/turbostat.c
> > > +++ b/tools/power/x86/turbostat/turbostat.c
> > > @@ -6496,8 +6496,13 @@ int check_for_cap_sys_rawio(void)
> > >  	int ret = 0;
> > >  
> > >  	caps = cap_get_proc();
> > > -	if (caps == NULL)
> > > +	if (caps == NULL) {
> > > +		/* Support CONFIG_MULTIUSER=n */
> > > +		if (errno == ENOSYS)
> > 
> > Can you point me where this knowledge comes from?
> > 
> > I downloaded the libcap source and didn't see how ENOSYS is set.
> 
> Hi Rui,
> 
> When the kernel is built without multiuser support, the capget() et al.
> syscalls are #ifdef'd out:
> 
>    
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/kernel/capability.c#n32
> 
> ...so userspace will get -ENOSYS when it tries to call them, and that
> ends up being propagated to errno in userspace.
> 
> Admittedly it is sort of implicit. Maybe a better way to "fix" this
> would be to warn the user if the capability check fails, but still
> attempt to access the MSR devices? I can do that if you prefer.

if cap_get_proc returns -ENOSYS only when CONFIG_MULTIUSER=n, then I
think the patch is good, maybe we just need some more detailed comment.

and I don't think we should continue when the capability check fails in
general.

> 
> That is my only problem here: when check_for_cap_sys_rawio() fails, the
> current code doesn't attempt to access the MSR devices at all, even
> though in my case it would actually work.

Or maybe we can cover this with the "--force" parameter?
say, does it make sense to address the problem by running "turbostat --
force"?
I'll leave this question to Len.

thanks,
rui
> 
> I realize this is very weird: it came up when I was recently including
> turbostat as part of an extremely tiny bootable utility image.
> 
> Thanks,
> Calvin
> 
> > thanks,
> > rui
> > > +			return 0;
> > > +
> > >  		return 1;
> > > +	}
> > >  
> > >  	if (cap_get_flag(caps, CAP_SYS_RAWIO, CAP_EFFECTIVE,
> > > &cap_flag_value)) {
> > >  		ret = 1;
> >