DMARC-Filter: OpenDMARC Filter v1.3.2 mx1.redhat.com E33D07E427
Date: Wed, 13 Sep 2017 16:39:56 +0200
From: Radim =?utf-8?B?S3LEjW3DocWZ?= <rkrcmar@redhat.com>
To: changbin.du@intel.com
Cc: pbonzini@redhat.com, tglx@linutronix.de, mingo@redhat.com,
        hpa@zytor.com, x86@kernel.org, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [RESEND][PATCH] kvm: x86: Do not handle MMIO request in
 fast_page_fault
Message-ID: <20170913143955.GB2673@flask>
References: <1504607862-8497-1-git-send-email-changbin.du@intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1504607862-8497-1-git-send-email-changbin.du@intel.com>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1697
Lines: 50

2017-09-05 18:37+0800, changbin.du@intel.com:
> From: Changbin Du <changbin.du@intel.com>
> 
> If it is a MMIO request, it should be handled by slow path. This patch
> actually fixed below warning when mmu debug is enabled.
> 
> WARNING: CPU: 5 PID: 2282 at arch/x86/kvm/mmu.c:226 fast_page_fault+0x41b/0x520
> CPU: 5 PID: 2282 Comm: qemu-system-x86 Not tainted 4.13.0-rc6+ #34
> task: ffff9b47f5286000 task.stack: ffffb18d03b28000
> RIP: 0010:fast_page_fault+0x41b/0x520
> Call Trace:
>   tdp_page_fault+0xfb/0x290
>   kvm_mmu_page_fault+0x61/0x120
>   handle_ept_misconfig+0x1ba/0x1c0
>   vmx_handle_exit+0xb8/0xd70
>   ? kvm_arch_vcpu_ioctl_run+0x9b6/0x18e0
>   kvm_arch_vcpu_ioctl_run+0xa5a/0x18e0
>   ? kvm_arch_vcpu_load+0x62/0x230
>   kvm_vcpu_ioctl+0x340/0x6c0
>   ? kvm_vcpu_ioctl+0x340/0x6c0
>   ? lock_acquire+0xf5/0x1f0
>   do_vfs_ioctl+0xa2/0x670
>   ? __fget+0x107/0x200
>   SyS_ioctl+0x79/0x90
>   entry_SYSCALL_64_fastpath+0x23/0xc2
> 
> Signed-off-by: Changbin Du <changbin.du@intel.com>
> ---
>  arch/x86/kvm/mmu.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 9d3f275..63c3360 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -3180,6 +3180,9 @@ static bool fast_page_fault(struct kvm_vcpu *vcpu, gva_t gva, int level,
>  			    iterator.level < level)
>  				break;
>  
> +		if (is_mmio_spte(spte))
> +			break;

Hm, handle_ept_misconfig() calls kvm_mmu_page_fault with error_code =
PFERR_RSVD_MASK.  This error_code gets propagated and checked at the
beginning of page_fault_can_be_fast(), where it should break the
function execution.

How did the execution get all the way to the loop?

Thanks.