Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756810Ab0GFKz5 (ORCPT ); Tue, 6 Jul 2010 06:55:57 -0400 Received: from cn.fujitsu.com ([222.73.24.84]:59263 "EHLO song.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1754731Ab0GFKzz (ORCPT ); Tue, 6 Jul 2010 06:55:55 -0400 Message-ID: <4C330AD3.7090402@cn.fujitsu.com> Date: Tue, 06 Jul 2010 18:52:03 +0800 From: Xiao Guangrong User-Agent: Thunderbird 2.0.0.24 (Windows/20100228) MIME-Version: 1.0 To: Avi Kivity CC: Marcelo Tosatti , LKML , KVM list Subject: [PATCH v5 9/9] KVM: MMU: trace pte prefetch References: <4C330918.6040709@cn.fujitsu.com> In-Reply-To: <4C330918.6040709@cn.fujitsu.com> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 7733 Lines: 256 Trace pte prefetch, it can help us to improve the prefetch's performance Signed-off-by: Xiao Guangrong --- arch/x86/kvm/mmu.c | 40 ++++++++++++++++++++++++++++++---------- arch/x86/kvm/mmutrace.h | 33 +++++++++++++++++++++++++++++++++ arch/x86/kvm/paging_tmpl.h | 23 +++++++++++++++++------ 3 files changed, 80 insertions(+), 16 deletions(-) diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c index 66e225d..9e3bc13 100644 --- a/arch/x86/kvm/mmu.c +++ b/arch/x86/kvm/mmu.c @@ -90,6 +90,11 @@ module_param(oos_shadow, bool, 0644); #endif #define PTE_PREFETCH_NUM 16 +#define PREFETCH_SUCCESS 0 +#define PREFETCH_ERR_GFN2PFN 1 +#define PREFETCH_ERR_ALLOC_MEM 2 +#define PREFETCH_ERR_RSVD_BITS_SET 3 +#define PREFETCH_ERR_MMIO 4 #define PT_FIRST_AVAIL_BITS_SHIFT 9 #define PT64_SECOND_AVAIL_BITS_SHIFT 52 @@ -2040,7 +2045,7 @@ static void nonpaging_new_cr3(struct kvm_vcpu *vcpu) static int direct_pte_prefetch_many(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, - u64 *start, u64 *end) + u64 *start, u64 *end, u64 address) { gfn_t gfn; struct page *pages[PTE_PREFETCH_NUM]; @@ -2052,31 +2057,44 @@ static int direct_pte_prefetch_many(struct kvm_vcpu *vcpu, ret = gfn_to_page_many_atomic(vcpu->kvm, gfn, pages, end - start, &enough); - if (ret <= 0) + if (ret <= 0) { + trace_pte_prefetch(true, address, 0, ret == -1 ? + PREFETCH_ERR_MMIO : PREFETCH_ERR_GFN2PFN); return -1; + } - for (j = 0; j < ret; j++, gfn++, start++) + for (j = 0; j < ret; j++, gfn++, start++) { + trace_pte_prefetch(true, address, 0, + PREFETCH_SUCCESS); mmu_set_spte(vcpu, start, ACC_ALL, sp->role.access, 0, 0, 1, NULL, sp->role.level, gfn, page_to_pfn(pages[j]), true, true); + } - if (!enough) + if (!enough) { + trace_pte_prefetch(true, address, 0, + PREFETCH_ERR_GFN2PFN); return -1; + } } return 0; } static void __direct_pte_prefetch(struct kvm_vcpu *vcpu, - struct kvm_mmu_page *sp, u64 *sptep) + struct kvm_mmu_page *sp, u64 *sptep, + u64 addr) { u64 *start = NULL; int index, i, max; WARN_ON(!sp->role.direct); - if (pte_prefetch_topup_memory_cache(vcpu)) + if (pte_prefetch_topup_memory_cache(vcpu)) { + trace_pte_prefetch(true, addr, 0, + PREFETCH_ERR_ALLOC_MEM); return; + } index = sptep - sp->spt; i = index & ~(PTE_PREFETCH_NUM - 1); @@ -2088,7 +2106,8 @@ static void __direct_pte_prefetch(struct kvm_vcpu *vcpu, if (*spte != shadow_trap_nonpresent_pte || spte == sptep) { if (!start) continue; - if (direct_pte_prefetch_many(vcpu, sp, start, spte) < 0) + if (direct_pte_prefetch_many(vcpu, sp, start, spte, + addr) < 0) break; start = NULL; } else if (!start) @@ -2096,7 +2115,7 @@ static void __direct_pte_prefetch(struct kvm_vcpu *vcpu, } } -static void direct_pte_prefetch(struct kvm_vcpu *vcpu, u64 *sptep) +static void direct_pte_prefetch(struct kvm_vcpu *vcpu, u64 *sptep, u64 addr) { struct kvm_mmu_page *sp; @@ -2113,7 +2132,7 @@ static void direct_pte_prefetch(struct kvm_vcpu *vcpu, u64 *sptep) if (sp->role.level > PT_PAGE_TABLE_LEVEL) return; - __direct_pte_prefetch(vcpu, sp, sptep); + __direct_pte_prefetch(vcpu, sp, sptep, addr); } static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write, @@ -2129,7 +2148,8 @@ static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write, mmu_set_spte(vcpu, iterator.sptep, ACC_ALL, ACC_ALL, 0, write, 1, &pt_write, level, gfn, pfn, false, true); - direct_pte_prefetch(vcpu, iterator.sptep); + direct_pte_prefetch(vcpu, iterator.sptep, + gfn << PAGE_SHIFT); ++vcpu->stat.pf_fixed; break; } diff --git a/arch/x86/kvm/mmutrace.h b/arch/x86/kvm/mmutrace.h index 3aab0f0..c07b6a6 100644 --- a/arch/x86/kvm/mmutrace.h +++ b/arch/x86/kvm/mmutrace.h @@ -195,6 +195,39 @@ DEFINE_EVENT(kvm_mmu_page_class, kvm_mmu_prepare_zap_page, TP_ARGS(sp) ); + +#define pte_prefetch_err \ + {PREFETCH_SUCCESS, "SUCCESS" }, \ + {PREFETCH_ERR_GFN2PFN, "ERR_GFN2PFN" }, \ + {PREFETCH_ERR_ALLOC_MEM, "ERR_ALLOC_MEM" }, \ + {PREFETCH_ERR_RSVD_BITS_SET, "ERR_RSVD_BITS_SET"}, \ + {PREFETCH_ERR_MMIO, "ERR_MMIO" } + +TRACE_EVENT( + pte_prefetch, + TP_PROTO(bool direct, u64 addr, u64 gpte, int err_code), + + TP_ARGS(direct, addr, gpte, err_code), + + TP_STRUCT__entry( + __field(bool, direct) + __field(u64, addr) + __field(u64, gpte) + __field(int, err_code) + ), + + TP_fast_assign( + __entry->direct = direct; + __entry->addr = addr; + __entry->gpte = gpte; + __entry->err_code = err_code; + ), + + TP_printk("%s address:%llx gpte:%llx %s", + __entry->direct ? "direct" : "indirect", + __entry->addr, __entry->gpte, + __print_symbolic(__entry->err_code, pte_prefetch_err)) + ); #endif /* _TRACE_KVMMMU_H */ #undef TRACE_INCLUDE_PATH diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h index a1e6d91..e7fe5fe 100644 --- a/arch/x86/kvm/paging_tmpl.h +++ b/arch/x86/kvm/paging_tmpl.h @@ -293,7 +293,7 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, } static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw, - u64 *sptep) + u64 *sptep, u64 addr) { struct kvm_mmu_page *sp; pt_element_t *gptep; @@ -306,7 +306,7 @@ static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw, return; if (sp->role.direct) - return __direct_pte_prefetch(vcpu, sp, sptep); + return __direct_pte_prefetch(vcpu, sp, sptep, addr); index = sptep - sp->spt; i = index & ~(PTE_PREFETCH_NUM - 1); @@ -314,8 +314,10 @@ static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw, gptep = gw->prefetch_ptes; - if (pte_prefetch_topup_memory_cache(vcpu)) + if (pte_prefetch_topup_memory_cache(vcpu)) { + trace_pte_prefetch(false, addr, 0, PREFETCH_ERR_ALLOC_MEM); return; + } for (j = 0; i < max; i++, j++) { pt_element_t gpte; @@ -332,15 +334,21 @@ static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw, gpte = gptep[j]; - if (is_rsvd_bits_set(vcpu, gpte, PT_PAGE_TABLE_LEVEL)) + if (is_rsvd_bits_set(vcpu, gpte, PT_PAGE_TABLE_LEVEL)) { + trace_pte_prefetch(false, addr, gpte, + PREFETCH_ERR_RSVD_BITS_SET); break; + } if (!(gpte & PT_ACCESSED_MASK)) continue; if (!is_present_gpte(gpte)) { - if (!sp->unsync) + if (!sp->unsync) { + trace_pte_prefetch(false, addr, gpte, + PREFETCH_SUCCESS); __set_spte(spte, shadow_notrap_nonpresent_pte); + } continue; } @@ -348,10 +356,13 @@ static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw, pfn = gfn_to_pfn_atomic(vcpu->kvm, gfn); if (is_error_pfn(pfn)) { + trace_pte_prefetch(false, addr, gpte, + PREFETCH_ERR_GFN2PFN); kvm_release_pfn_clean(pfn); break; } + trace_pte_prefetch(false, addr, gpte, PREFETCH_SUCCESS); pte_access = sp->role.access & FNAME(gpte_access)(vcpu, gpte); mmu_set_spte(vcpu, spte, sp->role.access, pte_access, 0, 0, is_dirty_gpte(gpte), NULL, sp->role.level, gfn, @@ -491,7 +502,7 @@ check_set_spte: user_fault, write_fault, dirty, ptwrite, level, gw->gfn, pfn, false, true); - FNAME(pte_prefetch)(vcpu, gw, sptep); + FNAME(pte_prefetch)(vcpu, gw, sptep, addr); break; } } -- 1.6.1.2 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/