Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Thu, 29 Jul 2021 15:28:35 +0200
From:   Claudio Imbrenda <imbrenda@linux.ibm.com>
To:     Janosch Frank <frankja@linux.ibm.com>
Cc:     kvm@vger.kernel.org, cohuck@redhat.com, borntraeger@de.ibm.com,
        thuth@redhat.com, pasic@linux.ibm.com, david@redhat.com,
        linux-s390@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 05/13] KVM: s390: pv: handle secure storage
 exceptions for normal guests
Message-ID: <20210729152835.1f470ba8@p-imbrenda>
In-Reply-To: <103c158c-dba6-7421-af8d-4d771c1cf087@linux.ibm.com>
References: <20210728142631.41860-1-imbrenda@linux.ibm.com>
        <20210728142631.41860-6-imbrenda@linux.ibm.com>
        <103c158c-dba6-7421-af8d-4d771c1cf087@linux.ibm.com>
Organization: IBM
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Thu, 29 Jul 2021 14:17:11 +0200
Janosch Frank <frankja@linux.ibm.com> wrote:

> On 7/28/21 4:26 PM, Claudio Imbrenda wrote:
> > With upcoming patches, normal guests might touch secure pages.
> > 
> > This patch extends the existing exception handler to convert the
> > pages to non secure also when the exception is triggered by a
> > normal guest.
> > 
> > This can happen for example when a secure guest reboots; the first
> > stage of a secure guest is non secure, and in general a secure guest
> > can reboot into non-secure mode.
> > 
> > If the secure memory of the previous boot has not been cleared up
> > completely yet, a non-secure guest might touch secure memory, which
> > will need to be handled properly.
> > 
> > Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> > ---
> >  arch/s390/mm/fault.c | 12 +++++++++++-
> >  1 file changed, 11 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c
> > index eb68b4f36927..b89d625ea2ec 100644
> > --- a/arch/s390/mm/fault.c
> > +++ b/arch/s390/mm/fault.c
> > @@ -767,6 +767,7 @@ void do_secure_storage_access(struct pt_regs
> > *regs) struct vm_area_struct *vma;
> >  	struct mm_struct *mm;
> >  	struct page *page;
> > +	struct gmap *gmap;
> >  	int rc;
> >  
> >  	/*
> > @@ -796,6 +797,16 @@ void do_secure_storage_access(struct pt_regs
> > *regs) }
> >  
> >  	switch (get_fault_type(regs)) {
> > +	case GMAP_FAULT:
> > +		gmap = (struct gmap *)S390_lowcore.gmap;
> > +		/*
> > +		 * Very unlikely, but if it happens, simply try
> > again.
> > +		 * The next attempt will trigger a different
> > exception.
> > +		 */  
> 
> If we keep this the way it currently is then the comment needs to go
> to the EFAULT check since it makes no sense above the
> gmap_translate().
> 
> > +		addr = __gmap_translate(gmap, addr);  
> 
> So we had a valid gmap PTE to end up here where the guest touched a
> secure page and triggered the exception. But we suddenly can't
> translate the gaddr to a vmaddr because the gmap tracking doesn't
> have an entry for the address.
> 
> My first instinct is to SIGSEGV the process since I can't come up
> with a way out of this situation except for the process to map this
> back in. The only reason I can think of that it was removed from the
> mapping is malicious intent or a bug.
> 
> I think this is needs a VM_FAULT_BADMAP and a do_fault_error() call.

fair enough, the next version will have that

> > +		if (addr == -EFAULT)
> > +			break;
> > +		fallthrough;
> >  	case USER_FAULT:
> >  		mm = current->mm;
> >  		mmap_read_lock(mm);
> > @@ -824,7 +835,6 @@ void do_secure_storage_access(struct pt_regs
> > *regs) if (rc)
> >  			BUG();
> >  		break;
> > -	case GMAP_FAULT:
> >  	default:
> >  		do_fault_error(regs, VM_READ | VM_WRITE,
> > VM_FAULT_BADMAP); WARN_ON_ONCE(1);
> >   
>