linux-kernel - Re: [PATCH 5/5] KVM: x86/xen: Explicitly pad struct compat_vcpu

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]

Date:   Wed, 10 Feb 2021 20:51:42 +0000
From:   "Woodhouse, David" <dwmw@...zon.co.uk>
To:     "seanjc@...gle.com" <seanjc@...gle.com>,
        "pbonzini@...hat.com" <pbonzini@...hat.com>
CC:     "jmattson@...gle.com" <jmattson@...gle.com>,
        "joro@...tes.org" <joro@...tes.org>,
        "kvm@...r.kernel.org" <kvm@...r.kernel.org>,
        "vkuznets@...hat.com" <vkuznets@...hat.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "wanpengli@...cent.com" <wanpengli@...cent.com>
Subject: Re: [PATCH 5/5] KVM: x86/xen: Explicitly pad struct compat_vcpu_info to 64
 bytes

On Wed, 2021-02-10 at 10:26 -0800, Sean Christopherson wrote:
> Add a 2 byte pad to struct compat_vcpu_info so that the sum size of its
> fields is actually 64 bytes.  The effective size without the padding is
> also 64 bytes due to the compiler aligning evtchn_pending_sel to a 4-byte
> boundary, but depending on compiler alignment is subtle and unnecessary.

I think there's at least one BUILD_BUG_ON() which would have triggered
if the compiler ever did stop honouring the ELF ABI. And in fact in a
parallel universe where the ABI permits such packing, the padding would
be *wrong*, since the original Xen struct doesn't have the padding. 

It *does* have an explicit uint8_t to replace evtchn_upcall_mask but it
doesn't have the following two bytes; canonically we *are* supposed to
take our chances with the ABI there. Although of course the relevant
ABI is the *32-bit* ABI in the compat case, not the 64-bit ABI. They
both align 32-bit values to 32 bits though.

    uint8_t evtchn_upcall_pending;
#ifdef XEN_HAVE_PV_UPCALL_MASK
    uint8_t evtchn_upcall_mask;
#else /* XEN_HAVE_PV_UPCALL_MASK */
    uint8_t pad0;
#endif /* XEN_HAVE_PV_UPCALL_MASK */
    xen_ulong_t evtchn_pending_sel;
    struct arch_vcpu_info arch;
    struct vcpu_time_info time;
}; /* 64 bytes (x86) */

So it isn't clear the additionally padding really buys us anything; if
we play this game without knowing the ABI we'd be screwed anyway. But
it doesn't hurt.

> Opportunistically replace spaces with tables in the other fields.

That part I certainly approve of. 

Reviewed-by: David Woodhouse <dwmw@...zon.co.uk>

Amazon Development Centre (London) Ltd. Registered in England and Wales with registration number 04543232 with its registered office at 1 Principal Place, Worship Street, London EC2A 2FA, United Kingdom.