linux-kernel - Re: [PATCH v3 13/25] x86/sgx: Add helpers to expose ECREATE and EINIT to KVM

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: <20210406205958.084147e365d04d066e4357c1@intel.com>
Date:   Tue, 6 Apr 2021 20:59:58 +1200
From:   Kai Huang <kai.huang@...el.com>
To:     Borislav Petkov <bp@...en8.de>
Cc:     kvm@...r.kernel.org, linux-sgx@...r.kernel.org, x86@...nel.org,
        linux-kernel@...r.kernel.org, seanjc@...gle.com, jarkko@...nel.org,
        luto@...nel.org, dave.hansen@...el.com, rick.p.edgecombe@...el.com,
        haitao.huang@...el.com, pbonzini@...hat.com, tglx@...utronix.de,
        mingo@...hat.com, hpa@...or.com
Subject: Re: [PATCH v3 13/25] x86/sgx: Add helpers to expose ECREATE and
 EINIT to KVM

On Tue, 6 Apr 2021 09:40:38 +0200 Borislav Petkov wrote:
> On Tue, Apr 06, 2021 at 09:44:21AM +1200, Kai Huang wrote:
> > The intention was to catch KVM bug, since KVM is the only caller, and in current
> > implementation KVM won't call this function if @secs is not a valid userspace
> > pointer. But yes we can also return here, but in this case an exception number
> > must also be specified to *trapnr so that KVM can inject to guest. It's not that
> > straightforward to decide which exception should we inject, but I think #GP
> > should be OK. Please see below.
> 
> Why should you inject anything in that case?
> 
> AFAICT, you can handle the return value in __handle_encls_ecreate() and
> inject only when the return value is EFAULT. If it is another negative
> error value, you pass it back up to its caller, handle_encls_ecreate()
> which returns other error values like -ENOMEM too. Which means, its
> callchain can stomach negative values just fine.
> 

OK. My thinking was that, returning negative error value basically means guest
will be killed.  For the case access_ok() fails for @secs or other user
pointers, it seems killing guest is a little it overkill, but since this code's
purpose is to catch KVM bug, I think killing guest is also OK from this
perspective (like -ENOMEM case, it is kernel/kvm internal error). So yes I
guess we can make handle_encls_xx() to stomach negative values, and only inject
upon -EFAULT.