From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mtagate4.de.ibm.com (mtagate4.de.ibm.com [195.212.29.153]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client CN "mtagate4.de.ibm.com", Issuer "Equifax" (verified OK)) by ozlabs.org (Postfix) with ESMTPS id 92A90DDE1F for ; Wed, 23 Jul 2008 18:37:07 +1000 (EST) Received: from d12nrmr1607.megacenter.de.ibm.com (d12nrmr1607.megacenter.de.ibm.com [9.149.167.49]) by mtagate4.de.ibm.com (8.13.8/8.13.8) with ESMTP id m6N8amnb264658 for ; Wed, 23 Jul 2008 08:36:48 GMT Received: from d12av03.megacenter.de.ibm.com (d12av03.megacenter.de.ibm.com [9.149.165.213]) by d12nrmr1607.megacenter.de.ibm.com (8.13.8/8.13.8/NCO v9.0) with ESMTP id m6N8alMQ2285604 for ; Wed, 23 Jul 2008 10:36:47 +0200 Received: from d12av03.megacenter.de.ibm.com (loopback [127.0.0.1]) by d12av03.megacenter.de.ibm.com (8.12.11.20060308/8.13.3) with ESMTP id m6N8alsu021877 for ; Wed, 23 Jul 2008 10:36:47 +0200 From: ehrhardt@linux.vnet.ibm.com To: kvm-ppc@vger.kernel.org, embedded-hypervisor@power.org, linuxppc-dev@ozlabs.org Subject: [PATCH 5/6] kvmppc: magic page paravirtualization - guest part Date: Wed, 23 Jul 2008 10:36:46 +0200 Message-Id: <1216802207-32675-6-git-send-email-ehrhardt@linux.vnet.ibm.com> In-Reply-To: <1216802207-32675-1-git-send-email-ehrhardt@linux.vnet.ibm.com> References: <1216802207-32675-1-git-send-email-ehrhardt@linux.vnet.ibm.com> Cc: hollisb@us.ibm.com List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , From: Christian Ehrhardt This patch adds the guest handling for the magic page mechanism. A Hypervisor can modify the device tree passed to the guest. Using that already existing interface a guest can simply detect available hypervisor features and agree on the supported ones using hypercalls. In this example it is checked for the feature switch "feature,pv-magicpage" in the hypervisor node and additional data which represents the size the hypervisor requests in "data,pv-magicpage-size". When the guest read that data and wants to support it the memory is allocated and passed to the hypervisor using the KVM_HCALL_RESERVE_MAGICPAGE hypercall. Signed-off-by: Christian Ehrhardt --- [diffstat] arch/powerpc/kernel/kvm.c | 48 +++++++++++++++++++++++++++++++++++++++++ include/asm-powerpc/kvm_para.h | 27 ++++++++++++++++++++++- 2 files changed, 74 insertions(+), 1 deletion(-) [diff] diff --git a/arch/powerpc/kernel/kvm.c b/arch/powerpc/kernel/kvm.c --- a/arch/powerpc/kernel/kvm.c +++ b/arch/powerpc/kernel/kvm.c @@ -22,9 +22,57 @@ #include #include #include +#include + +/* + * this is guest memory granted to the hypervisor; + * the hypervisor can place data in this area and rewrite + * privileged instructions to read from this area without + * trapping. + * Only the Hypervisor needs to be aware of the structure layout + * which makes the guest more felxible - the guest only guarantees + * the size which is requested by the hypervisor and read from a + * device tree entry. + */ +void *kvm_magicpage; + +static void __init kvmppc_register_magic_page(void) +{ + unsigned long paddr; + int size; + long err; + + size = kvmppc_pv_read_data(KVM_PVDATA_MAGICPAGE_SIZE); + if (size < 0) { + printk(KERN_ERR"%s: couldn't read size for kvmppc style " + "paravirtualization support (got %d)\n", + __func__, size); + return; + } + + /* FIXME Guest SMP needs that percpu + * On SMP we might also need a free implementation */ + kvm_magicpage = alloc_bootmem(size); + if (!kvm_magicpage) { + printk(KERN_ERR"%s - failed to allocate %d bytes\n", + __func__, size); + return; + } + + paddr = (unsigned long)__pa(kvm_magicpage); + err = kvm_hypercall1(KVM_HCALL_RESERVE_MAGICPAGE, paddr); + if (err) + printk(KERN_ERR"%s: couldn't register magic page\n", __func__); + else + printk(KERN_NOTICE"%s: registered %d bytes for " + "virtualization support\n", __func__, size); +} void __init kvm_guest_init(void) { if (!kvm_para_available()) return; + + if (kvm_para_has_feature(KVM_FEATURE_PPCPV_MAGICPAGE)) + kvmppc_register_magic_page(); } diff --git a/include/asm-powerpc/kvm_para.h b/include/asm-powerpc/kvm_para.h --- a/include/asm-powerpc/kvm_para.h +++ b/include/asm-powerpc/kvm_para.h @@ -28,10 +28,18 @@ #define KVM_HYPERCALL_BIN 0x03ffffff +#define KVM_HCALL_RESERVE_MAGICPAGE 0 + +#define KVM_PVDATA_MAGICPAGE_SIZE "data,pv-magicpage-size" + +/* List of PV features supported, returned as a bitfield */ +#define KVM_FEATURE_PPCPV_MAGICPAGE 0 + static struct kvmppc_para_features { char *dtcell; int feature; } para_features[] = { + { "feature,pv-magicpage", KVM_FEATURE_PPCPV_MAGICPAGE } }; static inline int kvm_para_available(void) @@ -54,13 +62,30 @@ if (!dn) return 0; - for (i = 0; i < ARRAY_SIZE(para_features)-1; i++) { + for (i = 0; i < ARRAY_SIZE(para_features); i++) { dtval = of_get_property(dn, para_features[i].dtcell, NULL); if (dtval && *dtval == 1) features |= (1 << para_features[i].feature); } return features; +} + +/* reads the specified data field out of the hypervisor node */ +static inline int kvmppc_pv_read_data(char *dtcell) +{ + struct device_node *dn; + const int *dtval; + + dn = of_find_node_by_path("/hypervisor"); + if (!dn) + return -EINVAL; + + dtval = of_get_property(dn, dtcell, NULL); + if (dtval) + return *dtval; + else + return -EINVAL; } void kvm_guest_init(void);