linux-kernel - Re: [PATCH 2/2] hyperv: Do not overlap the input and output hypercall areas in get

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: <Z2SRiEW6yU2Nf4hs@liuwe-devbox-debian-v2>
Date: Thu, 19 Dec 2024 21:35:04 +0000
From: Wei Liu <wei.liu@...nel.org>
To: Roman Kisel <romank@...ux.microsoft.com>
Cc: Wei Liu <wei.liu@...nel.org>, hpa@...or.com, kys@...rosoft.com,
	bp@...en8.de, dave.hansen@...ux.intel.com, decui@...rosoft.com,
	eahariha@...ux.microsoft.com, haiyangz@...rosoft.com,
	mingo@...hat.com, mhklinux@...look.com,
	nunodasneves@...ux.microsoft.com, tglx@...utronix.de,
	tiala@...rosoft.com, linux-hyperv@...r.kernel.org,
	linux-kernel@...r.kernel.org, x86@...nel.org, apais@...rosoft.com,
	benhill@...rosoft.com, ssengar@...rosoft.com,
	sunilmut@...rosoft.com, vdso@...bites.dev
Subject: Re: [PATCH 2/2] hyperv: Do not overlap the input and output
 hypercall areas in get_vtl(void)

On Thu, Dec 19, 2024 at 10:19:07AM -0800, Roman Kisel wrote:
> 
> 
> On 12/18/2024 6:42 PM, Wei Liu wrote:
> > On Wed, Dec 18, 2024 at 12:54:21PM -0800, Roman Kisel wrote:
> > > The Top-Level Functional Specification for Hyper-V, Section 3.6 [1, 2], disallows
> > > overlapping of the input and output hypercall areas, and get_vtl(void) does
> > > overlap them.
> > > 
> > > To fix this, enable allocation of the output hypercall pages when running in
> > > the VTL mode and use the output hypercall page of the current vCPU for the
> > > hypercall.
> > > 
> > > [1] https://learn.microsoft.com/en-us/virtualization/hyper-v-on-windows/tlfs/hypercall-interface
> > > [2] https://github.com/MicrosoftDocs/Virtualization-Documentation/tree/main/tlfs
> > > 
> > > Fixes: 8387ce06d70b ("x86/hyperv: Set Virtual Trust Level in VMBus init message")
> > > Signed-off-by: Roman Kisel <romank@...ux.microsoft.com>
> > > ---
> > >   arch/x86/hyperv/hv_init.c | 2 +-
> > >   drivers/hv/hv_common.c    | 6 +++---
> > >   2 files changed, 4 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/arch/x86/hyperv/hv_init.c b/arch/x86/hyperv/hv_init.c
> > > index c7185c6a290b..90c9ea00273e 100644
> > > --- a/arch/x86/hyperv/hv_init.c
> > > +++ b/arch/x86/hyperv/hv_init.c
> > > @@ -422,7 +422,7 @@ static u8 __init get_vtl(void)
> > >   	local_irq_save(flags);
> > >   	input = *this_cpu_ptr(hyperv_pcpu_input_arg);
> > > -	output = (struct hv_get_vp_registers_output *)input;
> > > +	output = *this_cpu_ptr(hyperv_pcpu_output_arg);
> > 
> > You can do
> > 
> > 	output = (char *)input + HV_HYP_PAGE_SIZE / 2;
> > 
> > to avoid the extra allocation.
> > 
> > The input and output structures surely won't take up half of the page.
> Agreed on the both counts! I do think that the attempt to save here
> won't help much: the hypercall output per-CPU pages in the VTL mode are
> needed just as in the dom0/root partition mode because this hypercall
> isn't going to be the only one required.
> 
> In other words, we will have to allocate these pages anyway as we evolve
> the code; we are trying to save here what is going to be spent anyway. Sort
> of, kicking the can down the road as the saying goes :)
> 

If you want this patch to be backported, then the smaller the change the
better.

In this particular case, I don't have a strong opinion. Your original
patch is small enough to be backported easily.

You can keep the patch as-is.

Thanks,
Wei.