From patchwork Fri Sep 20 04:52:45 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Paul Mackerras X-Patchwork-Id: 276223 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 3C7E82C011C for ; Fri, 20 Sep 2013 14:54:08 +1000 (EST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754147Ab3ITExk (ORCPT ); Fri, 20 Sep 2013 00:53:40 -0400 Received: from ozlabs.org ([203.10.76.45]:57346 "EHLO ozlabs.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754046Ab3ITExM (ORCPT ); Fri, 20 Sep 2013 00:53:12 -0400 Received: from iris.ozlabs.ibm.com (ibmaus65.lnk.telstra.net [165.228.126.9]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTPSA id E55402C0127; Fri, 20 Sep 2013 14:53:10 +1000 (EST) From: Paul Mackerras To: Alexander Graf , kvm-ppc@vger.kernel.org Cc: kvm@vger.kernel.org Subject: [PATCH 09/18] KVM: PPC: Book3S PR: Use 64k host pages where possible Date: Fri, 20 Sep 2013 14:52:45 +1000 Message-Id: <1379652774-10106-10-git-send-email-paulus@samba.org> X-Mailer: git-send-email 1.8.4.rc3 In-Reply-To: <1379652774-10106-1-git-send-email-paulus@samba.org> References: <1379652774-10106-1-git-send-email-paulus@samba.org> Sender: kvm-ppc-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm-ppc@vger.kernel.org Currently, PR KVM uses 4k pages for the host-side mappings of guest memory, regardless of the host page size. When the host page size is 64kB, we might as well use 64k host page mappings for guest mappings of 64kB and larger pages and for guest real-mode mappings. However, the magic page has to remain a 4k page. To implement this, we first add another flag bit to the guest VSID values we use, to indicate that this segment is one where host pages should be mapped using 64k pages. For segments with this bit set we set the bits in the shadow SLB entry to indicate a 64k base page size. When faulting in host HPTEs for this segment, we make them 64k HPTEs instead of 4k. We record the pagesize in struct hpte_cache for use when invalidating the HPTE. For now we restrict the segment containing the magic page (if any) to 4k pages. It should be possible to lift this restriction in future by ensuring that the magic 4k page is appropriately positioned within a host 64k page. Signed-off-by: Paul Mackerras --- arch/powerpc/include/asm/kvm_book3s.h | 6 ++++-- arch/powerpc/kvm/book3s_32_mmu.c | 1 + arch/powerpc/kvm/book3s_64_mmu.c | 35 ++++++++++++++++++++++++++++++----- arch/powerpc/kvm/book3s_64_mmu_host.c | 27 +++++++++++++++++++++------ arch/powerpc/kvm/book3s_pr.c | 1 + 5 files changed, 57 insertions(+), 13 deletions(-) diff --git a/arch/powerpc/include/asm/kvm_book3s.h b/arch/powerpc/include/asm/kvm_book3s.h index 1d4a120..6bf20b4 100644 --- a/arch/powerpc/include/asm/kvm_book3s.h +++ b/arch/powerpc/include/asm/kvm_book3s.h @@ -66,6 +66,7 @@ struct hpte_cache { u64 pfn; ulong slot; struct kvmppc_pte pte; + int pagesize; }; struct kvmppc_vcpu_book3s { @@ -113,8 +114,9 @@ struct kvmppc_vcpu_book3s { #define CONTEXT_GUEST 1 #define CONTEXT_GUEST_END 2 -#define VSID_REAL 0x0fffffffffc00000ULL -#define VSID_BAT 0x0fffffffffb00000ULL +#define VSID_REAL 0x07ffffffffc00000ULL +#define VSID_BAT 0x07ffffffffb00000ULL +#define VSID_64K 0x0800000000000000ULL #define VSID_1T 0x1000000000000000ULL #define VSID_REAL_DR 0x2000000000000000ULL #define VSID_REAL_IR 0x4000000000000000ULL diff --git a/arch/powerpc/kvm/book3s_32_mmu.c b/arch/powerpc/kvm/book3s_32_mmu.c index c8cefdd..af04553 100644 --- a/arch/powerpc/kvm/book3s_32_mmu.c +++ b/arch/powerpc/kvm/book3s_32_mmu.c @@ -308,6 +308,7 @@ static int kvmppc_mmu_book3s_32_xlate(struct kvm_vcpu *vcpu, gva_t eaddr, ulong mp_ea = vcpu->arch.magic_page_ea; pte->eaddr = eaddr; + pte->page_size = MMU_PAGE_4K; /* Magic page override */ if (unlikely(mp_ea) && diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c index ee1cfe2..50506ed 100644 --- a/arch/powerpc/kvm/book3s_64_mmu.c +++ b/arch/powerpc/kvm/book3s_64_mmu.c @@ -542,6 +542,16 @@ static void kvmppc_mmu_book3s_64_tlbie(struct kvm_vcpu *vcpu, ulong va, kvmppc_mmu_pte_vflush(vcpu, va >> 12, mask); } +#ifdef CONFIG_PPC_64K_PAGES +static int segment_contains_magic_page(struct kvm_vcpu *vcpu, ulong esid) +{ + ulong mp_ea = vcpu->arch.magic_page_ea; + + return mp_ea && !(vcpu->arch.shared->msr & MSR_PR) && + (mp_ea >> SID_SHIFT) == esid; +} +#endif + static int kvmppc_mmu_book3s_64_esid_to_vsid(struct kvm_vcpu *vcpu, ulong esid, u64 *vsid) { @@ -549,11 +559,13 @@ static int kvmppc_mmu_book3s_64_esid_to_vsid(struct kvm_vcpu *vcpu, ulong esid, struct kvmppc_slb *slb; u64 gvsid = esid; ulong mp_ea = vcpu->arch.magic_page_ea; + int pagesize = MMU_PAGE_64K; if (vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) { slb = kvmppc_mmu_book3s_64_find_slbe(vcpu, ea); if (slb) { gvsid = slb->vsid; + pagesize = slb->base_page_size; if (slb->tb) { gvsid <<= SID_SHIFT_1T - SID_SHIFT; gvsid |= esid & ((1ul << (SID_SHIFT_1T - SID_SHIFT)) - 1); @@ -564,28 +576,41 @@ static int kvmppc_mmu_book3s_64_esid_to_vsid(struct kvm_vcpu *vcpu, ulong esid, switch (vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) { case 0: - *vsid = VSID_REAL | esid; + gvsid = VSID_REAL | esid; break; case MSR_IR: - *vsid = VSID_REAL_IR | gvsid; + gvsid |= VSID_REAL_IR; break; case MSR_DR: - *vsid = VSID_REAL_DR | gvsid; + gvsid |= VSID_REAL_DR; break; case MSR_DR|MSR_IR: if (!slb) goto no_slb; - *vsid = gvsid; break; default: BUG(); break; } +#ifdef CONFIG_PPC_64K_PAGES + /* + * Mark this as a 64k segment if the host is using + * 64k pages, the host MMU supports 64k pages and + * the guest segment page size is >= 64k, + * but not if this segment contains the magic page. + */ + if (pagesize >= MMU_PAGE_64K && + mmu_psize_defs[MMU_PAGE_64K].shift && + !segment_contains_magic_page(vcpu, esid)) + gvsid |= VSID_64K; +#endif + if (vcpu->arch.shared->msr & MSR_PR) - *vsid |= VSID_PR; + gvsid |= VSID_PR; + *vsid = gvsid; return 0; no_slb: diff --git a/arch/powerpc/kvm/book3s_64_mmu_host.c b/arch/powerpc/kvm/book3s_64_mmu_host.c index e524052..6bda504 100644 --- a/arch/powerpc/kvm/book3s_64_mmu_host.c +++ b/arch/powerpc/kvm/book3s_64_mmu_host.c @@ -34,7 +34,7 @@ void kvmppc_mmu_invalidate_pte(struct kvm_vcpu *vcpu, struct hpte_cache *pte) { ppc_md.hpte_invalidate(pte->slot, pte->host_vpn, - MMU_PAGE_4K, MMU_PAGE_4K, MMU_SEGSIZE_256M, + pte->pagesize, pte->pagesize, MMU_SEGSIZE_256M, false); } @@ -90,6 +90,7 @@ int kvmppc_mmu_map_page(struct kvm_vcpu *vcpu, struct kvmppc_pte *orig_pte) int attempt = 0; struct kvmppc_sid_map *map; int r = 0; + int hpsize = MMU_PAGE_4K; /* Get host physical address for gpa */ hpaddr = kvmppc_gfn_to_pfn(vcpu, orig_pte->raddr >> PAGE_SHIFT); @@ -99,7 +100,6 @@ int kvmppc_mmu_map_page(struct kvm_vcpu *vcpu, struct kvmppc_pte *orig_pte) goto out; } hpaddr <<= PAGE_SHIFT; - hpaddr |= orig_pte->raddr & (~0xfffULL & ~PAGE_MASK); /* and write the mapping ea -> hpa into the pt */ vcpu->arch.mmu.esid_to_vsid(vcpu, orig_pte->eaddr >> SID_SHIFT, &vsid); @@ -117,8 +117,7 @@ int kvmppc_mmu_map_page(struct kvm_vcpu *vcpu, struct kvmppc_pte *orig_pte) goto out; } - vsid = map->host_vsid; - vpn = hpt_vpn(orig_pte->eaddr, vsid, MMU_SEGSIZE_256M); + vpn = hpt_vpn(orig_pte->eaddr, map->host_vsid, MMU_SEGSIZE_256M); if (!orig_pte->may_write) rflags |= HPTE_R_PP; @@ -130,7 +129,16 @@ int kvmppc_mmu_map_page(struct kvm_vcpu *vcpu, struct kvmppc_pte *orig_pte) else kvmppc_mmu_flush_icache(hpaddr >> PAGE_SHIFT); - hash = hpt_hash(vpn, PTE_SIZE, MMU_SEGSIZE_256M); + /* + * Use 64K pages if possible; otherwise, on 64K page kernels, + * we need to transfer 4 more bits from guest real to host real addr. + */ + if (vsid & VSID_64K) + hpsize = MMU_PAGE_64K; + else + hpaddr |= orig_pte->raddr & (~0xfffULL & ~PAGE_MASK); + + hash = hpt_hash(vpn, mmu_psize_defs[hpsize].shift, MMU_SEGSIZE_256M); map_again: hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP); @@ -143,7 +151,7 @@ map_again: } ret = ppc_md.hpte_insert(hpteg, vpn, hpaddr, rflags, vflags, - MMU_PAGE_4K, MMU_PAGE_4K, MMU_SEGSIZE_256M); + hpsize, hpsize, MMU_SEGSIZE_256M); if (ret < 0) { /* If we couldn't map a primary PTE, try a secondary */ @@ -168,6 +176,7 @@ map_again: pte->host_vpn = vpn; pte->pte = *orig_pte; pte->pfn = hpaddr >> PAGE_SHIFT; + pte->pagesize = hpsize; kvmppc_mmu_hpte_cache_map(vcpu, pte); } @@ -291,6 +300,12 @@ int kvmppc_mmu_map_segment(struct kvm_vcpu *vcpu, ulong eaddr) slb_vsid &= ~SLB_VSID_KP; slb_esid |= slb_index; +#ifdef CONFIG_PPC_64K_PAGES + /* Set host segment base page size to 64K if possible */ + if (gvsid & VSID_64K) + slb_vsid |= mmu_psize_defs[MMU_PAGE_64K].sllp; +#endif + svcpu->slb[slb_index].esid = slb_esid; svcpu->slb[slb_index].vsid = slb_vsid; diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c index 6cc9958..e9e8c74 100644 --- a/arch/powerpc/kvm/book3s_pr.c +++ b/arch/powerpc/kvm/book3s_pr.c @@ -422,6 +422,7 @@ int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu, pte.raddr = eaddr & KVM_PAM; pte.eaddr = eaddr; pte.vpage = eaddr >> 12; + pte.page_size = MMU_PAGE_64K; } switch (vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) {