[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAGtprH8Ts9ScKUTraw-V20mLvdGjT_ggZCwWQD393KgfVa1nNA@mail.gmail.com>
Date: Thu, 13 Feb 2025 09:40:31 -0800
From: Vishal Annapurve <vannapurve@...gle.com>
To: "Kirill A. Shutemov" <kirill@...temov.name>
Cc: x86@...nel.org, linux-kernel@...r.kernel.org, pbonzini@...hat.com,
seanjc@...gle.com, erdemaktas@...gle.com, ackerleytng@...gle.com,
jxgao@...gle.com, sagis@...gle.com, oupton@...gle.com, pgonda@...gle.com,
dave.hansen@...ux.intel.com, linux-coco@...ts.linux.dev,
chao.p.peng@...ux.intel.com, isaku.yamahata@...il.com,
sathyanarayanan.kuppuswamy@...ux.intel.com
Subject: Re: [PATCH V4 4/4] x86/tdx: Remove TDX specific idle routine
On Wed, Feb 12, 2025 at 5:03 AM Kirill A. Shutemov <kirill@...temov.name> wrote:
>
> On Wed, Feb 12, 2025 at 12:07:47AM +0000, Vishal Annapurve wrote:
> > With explicit dependency on CONFIG_PARAVIRT and TDX specific
> > halt()/safe_halt() routines in place, default_idle() is safe to execute for
> > TDX VMs. Remove TDX specific idle routine override which is now
> > redundant.
>
> I am not convinced that it is good idea.
>
> It adds two needless flipping of IF in the hot path: first enabling
> interrupts in tdx_safe_halt() and disabling it back in default_idle().
>
IIUC, the aspect of flipping IF in the hot path twice is also
applicable to default_idle() that gets executed for non-confidential
VMs. I agree that it's redundant for TDX VMs and if there is strong
consensus here, I can drop this cleanup in the next version.
> --
> Kiryl Shutsemau / Kirill A. Shutemov
Powered by blists - more mailing lists