linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Avi Kivity <avi.kivity@gmail.com>
To: Paolo Bonzini <pbonzini@redhat.com>,
	linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Cc: rkrcmar@redhat.com, bsd@redhat.com,
	guangrong.xiao@linux.intel.com,
	Yang Zhang <yang.z.zhang@intel.com>,
	wanpeng.li@linux.intel.com
Subject: Re: [PATCH 12/13] KVM: x86: add KVM_MEM_X86_SMRAM memory slot flag
Date: Fri, 15 May 2015 23:32:16 +0300	[thread overview]
Message-ID: <555657D0.5080500@gmail.com> (raw)
In-Reply-To: <1430393772-27208-13-git-send-email-pbonzini@redhat.com>

On 04/30/2015 02:36 PM, Paolo Bonzini wrote:
> This adds an arch-specific memslot flag that hides slots unless the
> VCPU is in system management mode.
>
> Some care is needed in order to limit the overhead of x86_gfn_to_memslot
> when compared with gfn_to_memslot.  Thankfully, we have __gfn_to_memslot
> and search_memslots which are the same, so we can add some extra output
> to search_memslots.  The compiler will optimize it as dead code in
> __gfn_to_memslot, and will use it to thread jumps in x86_gfn_to_memslot.
>
>   
>   struct kvm_memory_slot *x86_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn)
>   {
> -	struct kvm_memory_slot *slot = gfn_to_memslot(vcpu->kvm, gfn);
> +	/* By using search_memslots directly the compiler can optimize away
> +	 * the "if (found)" check below.
> +         *
> +	 * It cannot do the same for gfn_to_memslot because it is not inlined,
> +	 * and it also cannot do the same for __gfn_to_memslot because the
> +	 * kernel is compiled with -fno-delete-null-pointer-checks.
> +	 */
> +	bool found;
> +	struct kvm_memslots *memslots = kvm_memslots(vcpu->kvm);
> +	struct kvm_memory_slot *slot = search_memslots(memslots, gfn, &found);
> +
> +	if (found && unlikely(slot->flags & KVM_MEM_X86_SMRAM) && !is_smm(vcpu))
> +		return NULL;
>   

Alternative approach: store the memslot pointer in the vcpu, instead of 
of vcpu->kvm.  The uapi interface code can generate two memslot tables 
to be stored in kvm->, one for smm and one for !smm.

This is a little more generic and can be used for other asymmetric 
memory maps causes (the only one I can think of, mapping the APIC to 
different physical addresses, can't be efficiently supported).

>   	return slot;
>   }
> @@ -112,7 +125,15 @@ EXPORT_SYMBOL_GPL(x86_read_guest);
>   int x86_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
>   			      gpa_t gpa, unsigned long len)
>   {
> -	return kvm_gfn_to_hva_cache_init(kvm, ghc, gpa, len);
> +	int r = kvm_gfn_to_hva_cache_init(kvm, ghc, gpa, len);
> +
> +	if (r < 0)
> +		return r;
> +
> +	/* Use slow path for reads and writes to SMRAM.  */
> +	if (ghc->memslot && (ghc->memslot->flags & KVM_MEM_X86_SMRAM))
> +		ghc->memslot = NULL;
> +	return r;
>   }
>   EXPORT_SYMBOL_GPL(x86_gfn_to_hva_cache_init);
>   
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index e7cfee09970a..ea3db5d89f67 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -800,16 +800,18 @@ static inline void kvm_guest_exit(void)
>    * gfn_to_memslot() itself isn't here as an inline because that would
>    * bloat other code too much.
>    */
> -static inline struct kvm_memory_slot *
> -search_memslots(struct kvm_memslots *slots, gfn_t gfn)
> +static __always_inline struct kvm_memory_slot *
> +search_memslots(struct kvm_memslots *slots, gfn_t gfn, bool *found)
>   {
>   	int start = 0, end = slots->used_slots;
>   	int slot = atomic_read(&slots->lru_slot);
>   	struct kvm_memory_slot *memslots = slots->memslots;
>   
>   	if (gfn >= memslots[slot].base_gfn &&
> -	    gfn < memslots[slot].base_gfn + memslots[slot].npages)
> +	    gfn < memslots[slot].base_gfn + memslots[slot].npages) {
> +		*found = true;
>   		return &memslots[slot];
> +	}
>   
>   	while (start < end) {
>   		slot = start + (end - start) / 2;
> @@ -823,16 +825,20 @@ search_memslots(struct kvm_memslots *slots, gfn_t gfn)
>   	if (gfn >= memslots[start].base_gfn &&
>   	    gfn < memslots[start].base_gfn + memslots[start].npages) {
>   		atomic_set(&slots->lru_slot, start);
> +		*found = true;
>   		return &memslots[start];
>   	}
>   
> +	*found = false;
>   	return NULL;
>   }
>   
>   static inline struct kvm_memory_slot *
>   __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
>   {
> -	return search_memslots(slots, gfn);
> +	bool found;
> +
> +	return search_memslots(slots, gfn, &found);
>   }
>   
>   static inline unsigned long
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index e82c46b9860f..b8b76106a003 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -716,6 +716,10 @@ static int check_memory_region_flags(struct kvm_userspace_memory_region *mem)
>   #ifdef __KVM_HAVE_READONLY_MEM
>   	valid_flags |= KVM_MEM_READONLY;
>   #endif
> +#ifdef __KVM_ARCH_VALID_FLAGS
> +	BUILD_BUG_ON(__KVM_ARCH_VALID_FLAGS & KVM_MEMSLOT_INVALID);
> +	valid_flags |= __KVM_ARCH_VALID_FLAGS;
> +#endif
>   
>   	if (mem->flags & ~valid_flags)
>   		return -EINVAL;


  parent reply	other threads:[~2015-05-15 20:32 UTC|newest]

Thread overview: 54+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-04-30 11:35 [RFC PATCH 00/13] KVM: x86: SMM support Paolo Bonzini
2015-04-30 11:36 ` [PATCH 01/13] KVM: MMU: fix for CR4.SMEP=1, CR0.WP=0? Paolo Bonzini
2015-05-08  2:52   ` Xiao Guangrong
2015-04-30 11:36 ` [PATCH 02/13] KVM: reuse memslot in kvm_write_guest_page Paolo Bonzini
2015-05-05 15:03   ` Bandan Das
2015-05-05 16:29     ` Radim Krčmář
2015-04-30 11:36 ` [PATCH 03/13] KVM: export __gfn_to_pfn_memslot, drop gfn_to_pfn_async Paolo Bonzini
2015-04-30 11:36 ` [PATCH 04/13] KVM: remove unnecessary arg from mark_page_dirty_in_slot, export it Paolo Bonzini
2015-04-30 11:36 ` [PATCH 05/13] KVM: x86: pass host_initiated to functions that read MSRs Paolo Bonzini
2015-05-04 14:01   ` Radim Krčmář
2015-05-04 16:04     ` Paolo Bonzini
2015-04-30 11:36 ` [PATCH 06/13] KVM: x86: pass the whole hflags field to emulator and back Paolo Bonzini
2015-05-05 15:47   ` Bandan Das
2015-05-05 16:16     ` Paolo Bonzini
2015-05-06 16:49       ` Bandan Das
2015-04-30 11:36 ` [PATCH 07/13] KVM: x86: API changes for SMM support Paolo Bonzini
2015-05-04 15:37   ` Radim Krčmář
2015-05-04 16:02     ` Paolo Bonzini
2015-05-05 16:36   ` Bandan Das
2015-05-05 16:45     ` Paolo Bonzini
2015-04-30 11:36 ` [PATCH 08/13] KVM: x86: stubs " Paolo Bonzini
2015-05-04 17:51   ` Radim Krčmář
2015-05-05  9:37     ` Paolo Bonzini
2015-05-05 18:38     ` Bandan Das
2015-05-05 18:48       ` Radim Krčmář
2015-04-30 11:36 ` [PATCH 09/13] KVM: x86: save/load state on SMM switch Paolo Bonzini
2015-05-04 19:59   ` Radim Krčmář
2015-05-05  9:37     ` Paolo Bonzini
2015-05-05 12:48       ` Radim Krčmář
2015-05-05 13:18         ` Paolo Bonzini
2015-05-05 20:44   ` Bandan Das
2015-05-06 10:39     ` Paolo Bonzini
2015-05-06 17:55       ` Bandan Das
2015-05-06 19:38         ` Paolo Bonzini
2015-05-12 23:56           ` Bandan Das
2015-05-13  6:58             ` Paolo Bonzini
2015-04-30 11:36 ` [PATCH 10/13] KVM: x86: add vcpu-specific functions to read/write/translate GFNs Paolo Bonzini
2015-04-30 11:36 ` [PATCH 11/13] KVM: x86: add SMM to the MMU role Paolo Bonzini
2015-04-30 11:36 ` [PATCH 12/13] KVM: x86: add KVM_MEM_X86_SMRAM memory slot flag Paolo Bonzini
2015-05-05 17:17   ` Radim Krčmář
2015-05-06  9:47     ` Paolo Bonzini
2015-05-06 16:24       ` Radim Krčmář
2015-05-06 18:15         ` Bandan Das
2015-05-06 19:43         ` Paolo Bonzini
2015-05-15 20:32   ` Avi Kivity [this message]
2015-05-18  8:31     ` Paolo Bonzini
2015-04-30 11:36 ` [PATCH 13/13] KVM: x86: advertise KVM_CAP_X86_SMM Paolo Bonzini
2015-05-05 18:40 ` [RFC PATCH 00/13] KVM: x86: SMM support Radim Krčmář
2015-05-06 11:18   ` Paolo Bonzini
2015-05-06 17:14     ` Radim Krčmář
2015-05-19 14:25 ` Zhang, Yang Z
2015-05-19 14:27   ` Paolo Bonzini
2015-05-20  1:03     ` Zhang, Yang Z
2015-05-20 15:22     ` Andi Kleen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=555657D0.5080500@gmail.com \
    --to=avi.kivity@gmail.com \
    --cc=bsd@redhat.com \
    --cc=guangrong.xiao@linux.intel.com \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=rkrcmar@redhat.com \
    --cc=wanpeng.li@linux.intel.com \
    --cc=yang.z.zhang@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).