diff --git a/sys/x86/xen/hvm.c b/sys/x86/xen/hvm.c index 9b5b588234c4..569b113364b1 100644 --- a/sys/x86/xen/hvm.c +++ b/sys/x86/xen/hvm.c @@ -1,535 +1,491 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2008, 2013 Citrix Systems, Inc. * Copyright (c) 2012 Spectra Logic Corporation * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /*--------------------------- Forward Declarations ---------------------------*/ static void xen_hvm_cpu_init(void); /*-------------------------------- Global Data -------------------------------*/ enum xen_domain_type xen_domain_type = XEN_NATIVE; #ifdef SMP struct cpu_ops xen_hvm_cpu_ops = { .cpu_init = xen_hvm_cpu_init, .cpu_resume = xen_hvm_cpu_init }; #endif static MALLOC_DEFINE(M_XENHVM, "xen_hvm", "Xen HVM PV Support"); /** * If non-zero, the hypervisor has been configured to use a direct * IDT event callback for interrupt injection. */ int xen_vector_callback_enabled; /** * Start info flags. ATM this only used to store the initial domain flag for * PVHv2, and it's always empty for HVM guests. */ uint32_t hvm_start_flags; /** * Signal whether the vector injected for the event channel upcall requires to * be EOI'ed on the local APIC. */ bool xen_evtchn_needs_ack; /*------------------------------- Per-CPU Data -------------------------------*/ DPCPU_DEFINE(struct vcpu_info, vcpu_local_info); DPCPU_DEFINE(struct vcpu_info *, vcpu_info); /*------------------ Hypervisor Access Shared Memory Regions -----------------*/ shared_info_t *HYPERVISOR_shared_info; /*------------------------------ Sysctl tunables -----------------------------*/ int xen_disable_pv_disks = 0; int xen_disable_pv_nics = 0; TUNABLE_INT("hw.xen.disable_pv_disks", &xen_disable_pv_disks); TUNABLE_INT("hw.xen.disable_pv_nics", &xen_disable_pv_nics); /*---------------------- XEN Hypervisor Probe and Setup ----------------------*/ static uint32_t cpuid_base; static uint32_t xen_hvm_cpuid_base(void) { uint32_t base, regs[4]; for (base = 0x40000000; base < 0x40010000; base += 0x100) { do_cpuid(base, regs); if (!memcmp("XenVMMXenVMM", ®s[1], 12) && (regs[0] - base) >= 2) return (base); } return (0); } static void hypervisor_quirks(unsigned int major, unsigned int minor) { #ifdef SMP if (((major < 4) || (major == 4 && minor <= 5)) && msix_disable_migration == -1) { /* * Xen hypervisors prior to 4.6.0 do not properly * handle updates to enabled MSI-X table entries, * so disable MSI-X interrupt migration in that * case. */ if (bootverbose) printf( "Disabling MSI-X interrupt migration due to Xen hypervisor bug.\n" "Set machdep.msix_disable_migration=0 to forcefully enable it.\n"); msix_disable_migration = 1; } #endif } static void hypervisor_version(void) { uint32_t regs[4]; int major, minor; do_cpuid(cpuid_base + 1, regs); major = regs[0] >> 16; minor = regs[0] & 0xffff; printf("XEN: Hypervisor version %d.%d detected.\n", major, minor); hypervisor_quirks(major, minor); } /* * Allocate and fill in the hypcall page. */ int xen_hvm_init_hypercall_stubs(enum xen_hvm_init_type init_type) { uint32_t regs[4]; /* Legacy PVH will get here without the cpuid leaf being set. */ if (cpuid_base == 0) cpuid_base = xen_hvm_cpuid_base(); if (cpuid_base == 0) return (ENXIO); if (xen_domain() && init_type == XEN_HVM_INIT_LATE) { /* * If the domain type is already set we can assume that the * hypercall page has been populated too, so just print the * version (and apply any quirks) and exit. */ hypervisor_version(); return 0; } if (init_type == XEN_HVM_INIT_LATE) hypervisor_version(); /* * Find the hypercall pages. */ do_cpuid(cpuid_base + 2, regs); if (regs[0] != 1) return (EINVAL); wrmsr(regs[1], (init_type == XEN_HVM_INIT_EARLY) ? ((vm_paddr_t)&hypercall_page - KERNBASE) : vtophys(&hypercall_page)); return (0); } static void xen_hvm_init_shared_info_page(void) { struct xen_add_to_physmap xatp; if (xen_pv_domain()) { /* * Already setup in the PV case, shared_info is passed inside * of the start_info struct at start of day. */ return; } if (HYPERVISOR_shared_info == NULL) { HYPERVISOR_shared_info = malloc(PAGE_SIZE, M_XENHVM, M_NOWAIT); if (HYPERVISOR_shared_info == NULL) panic("Unable to allocate Xen shared info page"); } xatp.domid = DOMID_SELF; xatp.idx = 0; xatp.space = XENMAPSPACE_shared_info; xatp.gpfn = vtophys(HYPERVISOR_shared_info) >> PAGE_SHIFT; if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp)) panic("HYPERVISOR_memory_op failed"); } static int set_percpu_callback(unsigned int vcpu) { struct xen_hvm_evtchn_upcall_vector vec; int error; vec.vcpu = vcpu; vec.vector = IDT_EVTCHN; error = HYPERVISOR_hvm_op(HVMOP_set_evtchn_upcall_vector, &vec); return (error != 0 ? xen_translate_error(error) : 0); } /* * Tell the hypervisor how to contact us for event channel callbacks. */ void xen_hvm_set_callback(device_t dev) { struct xen_hvm_param xhp; int irq; if (xen_vector_callback_enabled) return; xhp.domid = DOMID_SELF; xhp.index = HVM_PARAM_CALLBACK_IRQ; if (xen_feature(XENFEAT_hvm_callback_vector) != 0) { int error; error = set_percpu_callback(0); if (error == 0) { xen_evtchn_needs_ack = true; /* Trick toolstack to think we are enlightened */ xhp.value = 1; } else xhp.value = HVM_CALLBACK_VECTOR(IDT_EVTCHN); error = HYPERVISOR_hvm_op(HVMOP_set_param, &xhp); if (error == 0) { xen_vector_callback_enabled = 1; return; } else if (xen_evtchn_needs_ack) panic("Unable to setup fake HVM param: %d", error); printf("Xen HVM callback vector registration failed (%d). " "Falling back to emulated device interrupt\n", error); } xen_vector_callback_enabled = 0; if (dev == NULL) { /* * Called from early boot or resume. * xenpci will invoke us again later. */ return; } irq = pci_get_irq(dev); if (irq < 16) { xhp.value = HVM_CALLBACK_GSI(irq); } else { u_int slot; u_int pin; slot = pci_get_slot(dev); pin = pci_get_intpin(dev) - 1; xhp.value = HVM_CALLBACK_PCI_INTX(slot, pin); } if (HYPERVISOR_hvm_op(HVMOP_set_param, &xhp) != 0) panic("Can't set evtchn callback"); } #define XEN_MAGIC_IOPORT 0x10 enum { XMI_MAGIC = 0x49d2, XMI_UNPLUG_IDE_DISKS = 0x01, XMI_UNPLUG_NICS = 0x02, XMI_UNPLUG_IDE_EXCEPT_PRI_MASTER = 0x04 }; static void xen_hvm_disable_emulated_devices(void) { u_short disable_devs = 0; if (xen_pv_domain()) { /* * No emulated devices in the PV case, so no need to unplug * anything. */ if (xen_disable_pv_disks != 0 || xen_disable_pv_nics != 0) printf("PV devices cannot be disabled in PV guests\n"); return; } if (inw(XEN_MAGIC_IOPORT) != XMI_MAGIC) return; if (xen_disable_pv_disks == 0) { if (bootverbose) printf("XEN: disabling emulated disks\n"); disable_devs |= XMI_UNPLUG_IDE_DISKS; } if (xen_disable_pv_nics == 0) { if (bootverbose) printf("XEN: disabling emulated nics\n"); disable_devs |= XMI_UNPLUG_NICS; } if (disable_devs != 0) outw(XEN_MAGIC_IOPORT, disable_devs); } static void xen_hvm_init(enum xen_hvm_init_type init_type) { int error; int i; if (init_type == XEN_HVM_INIT_CANCELLED_SUSPEND) return; error = xen_hvm_init_hypercall_stubs(init_type); switch (init_type) { case XEN_HVM_INIT_LATE: if (error != 0) return; /* * If xen_domain_type is not set at this point * it means we are inside a (PV)HVM guest, because * for PVH the guest type is set much earlier * (see hammer_time_xen). */ if (!xen_domain()) { xen_domain_type = XEN_HVM_DOMAIN; vm_guest = VM_GUEST_XEN; } setup_xen_features(); #ifdef SMP cpu_ops = xen_hvm_cpu_ops; #endif break; case XEN_HVM_INIT_RESUME: if (error != 0) panic("Unable to init Xen hypercall stubs on resume"); /* Clear stale vcpu_info. */ CPU_FOREACH(i) DPCPU_ID_SET(i, vcpu_info, NULL); break; default: panic("Unsupported HVM initialization type"); } xen_vector_callback_enabled = 0; xen_evtchn_needs_ack = false; xen_hvm_set_callback(NULL); /* * On (PV)HVM domains we need to request the hypervisor to * fill the shared info page, for PVH guest the shared_info page * is passed inside the start_info struct and is already set, so this * functions are no-ops. */ xen_hvm_init_shared_info_page(); xen_hvm_disable_emulated_devices(); } void xen_hvm_suspend(void) { } void xen_hvm_resume(bool suspend_cancelled) { xen_hvm_init(suspend_cancelled ? XEN_HVM_INIT_CANCELLED_SUSPEND : XEN_HVM_INIT_RESUME); /* Register vcpu_info area for CPU#0. */ xen_hvm_cpu_init(); } static void xen_hvm_sysinit(void *arg __unused) { xen_hvm_init(XEN_HVM_INIT_LATE); } SYSINIT(xen_hvm_init, SI_SUB_HYPERVISOR, SI_ORDER_FIRST, xen_hvm_sysinit, NULL); static void xen_hvm_cpu_init(void) { struct vcpu_register_vcpu_info info; struct vcpu_info *vcpu_info; uint32_t regs[4]; int cpu, rc; if (!xen_domain()) return; if (DPCPU_GET(vcpu_info) != NULL) { /* * vcpu_info is already set. We're resuming * from a failed migration and our pre-suspend * configuration is still valid. */ return; } /* * Set vCPU ID. If available fetch the ID from CPUID, if not just use * the ACPI ID. */ KASSERT(cpuid_base != 0, ("Invalid base Xen CPUID leaf")); cpuid_count(cpuid_base + 4, 0, regs); KASSERT((regs[0] & XEN_HVM_CPUID_VCPU_ID_PRESENT) || !xen_pv_domain(), ("Xen PV domain without vcpu_id in cpuid")); PCPU_SET(vcpu_id, (regs[0] & XEN_HVM_CPUID_VCPU_ID_PRESENT) ? regs[1] : PCPU_GET(acpi_id)); if (xen_evtchn_needs_ack && !IS_BSP()) { /* * Setup the per-vpcu event channel upcall vector. This is only * required when using the new HVMOP_set_evtchn_upcall_vector * hypercall, which allows using a different vector for each * vCPU. Note that FreeBSD uses the same vector for all vCPUs * because it's not dynamically allocated. */ rc = set_percpu_callback(PCPU_GET(vcpu_id)); if (rc != 0) panic("Event channel upcall vector setup failed: %d", rc); } /* * Set the vCPU info. * * NB: the vCPU info for vCPUs < 32 can be fetched from the shared info * page, but in order to make sure the mapping code is correct always * attempt to map the vCPU info at a custom place. */ vcpu_info = DPCPU_PTR(vcpu_local_info); cpu = PCPU_GET(vcpu_id); info.mfn = vtophys(vcpu_info) >> PAGE_SHIFT; info.offset = vtophys(vcpu_info) - trunc_page(vtophys(vcpu_info)); rc = HYPERVISOR_vcpu_op(VCPUOP_register_vcpu_info, cpu, &info); if (rc != 0) DPCPU_SET(vcpu_info, &HYPERVISOR_shared_info->vcpu_info[cpu]); else DPCPU_SET(vcpu_info, vcpu_info); } SYSINIT(xen_hvm_cpu_init, SI_SUB_INTR, SI_ORDER_FIRST, xen_hvm_cpu_init, NULL); - -/* HVM/PVH start_info accessors */ -static vm_paddr_t -hvm_get_xenstore_mfn(void) -{ - - return (hvm_get_parameter(HVM_PARAM_STORE_PFN)); -} - -static evtchn_port_t -hvm_get_xenstore_evtchn(void) -{ - - return (hvm_get_parameter(HVM_PARAM_STORE_EVTCHN)); -} - -static vm_paddr_t -hvm_get_console_mfn(void) -{ - - return (hvm_get_parameter(HVM_PARAM_CONSOLE_PFN)); -} - -static evtchn_port_t -hvm_get_console_evtchn(void) -{ - - return (hvm_get_parameter(HVM_PARAM_CONSOLE_EVTCHN)); -} - -static uint32_t -hvm_get_start_flags(void) -{ - - return (hvm_start_flags); -} - -struct hypervisor_info hypervisor_info = { - .get_xenstore_mfn = hvm_get_xenstore_mfn, - .get_xenstore_evtchn = hvm_get_xenstore_evtchn, - .get_console_mfn = hvm_get_console_mfn, - .get_console_evtchn = hvm_get_console_evtchn, - .get_start_flags = hvm_get_start_flags, -}; diff --git a/sys/xen/xen-os.h b/sys/xen/xen-os.h index 874c3b71b861..cd8a53aab54f 100644 --- a/sys/xen/xen-os.h +++ b/sys/xen/xen-os.h @@ -1,191 +1,176 @@ /****************************************************************************** * xen/xen-os.h * * Random collection of macros and definition * * Copyright (c) 2003, 2004 Keir Fraser (on behalf of the Xen team) * All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to * deal in the Software without restriction, including without limitation the * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or * sell copies of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * * $FreeBSD$ */ #ifndef _XEN_XEN_OS_H_ #define _XEN_XEN_OS_H_ #if !defined(__XEN_INTERFACE_VERSION__) #define __XEN_INTERFACE_VERSION__ 0x00030208 #endif #define GRANT_REF_INVALID 0xffffffff #ifdef LOCORE #define __ASSEMBLY__ #endif #include #ifndef __ASSEMBLY__ +#include #include -struct hypervisor_info { - vm_paddr_t (*get_xenstore_mfn)(void); - evtchn_port_t (*get_xenstore_evtchn)(void); - vm_paddr_t (*get_console_mfn)(void); - evtchn_port_t (*get_console_evtchn)(void); - uint32_t (*get_start_flags)(void); -}; -extern struct hypervisor_info hypervisor_info; - static inline vm_paddr_t xen_get_xenstore_mfn(void) { - return (hypervisor_info.get_xenstore_mfn()); + return (hvm_get_parameter(HVM_PARAM_STORE_PFN)); } static inline evtchn_port_t xen_get_xenstore_evtchn(void) { - return (hypervisor_info.get_xenstore_evtchn()); + return (hvm_get_parameter(HVM_PARAM_STORE_EVTCHN)); } static inline vm_paddr_t xen_get_console_mfn(void) { - return (hypervisor_info.get_console_mfn()); + return (hvm_get_parameter(HVM_PARAM_CONSOLE_PFN)); } static inline evtchn_port_t xen_get_console_evtchn(void) { - return (hypervisor_info.get_console_evtchn()); -} - -static inline uint32_t -xen_get_start_flags(void) -{ - - return (hypervisor_info.get_start_flags()); + return (hvm_get_parameter(HVM_PARAM_CONSOLE_EVTCHN)); } #endif #include /* Everything below this point is not included by assembler (.S) files. */ #ifndef __ASSEMBLY__ extern shared_info_t *HYPERVISOR_shared_info; extern int xen_disable_pv_disks; extern int xen_disable_pv_nics; extern bool xen_suspend_cancelled; enum xen_domain_type { XEN_NATIVE, /* running on bare hardware */ XEN_PV_DOMAIN, /* running in a PV domain */ XEN_HVM_DOMAIN, /* running in a Xen hvm domain */ }; extern enum xen_domain_type xen_domain_type; static inline int xen_domain(void) { return (xen_domain_type != XEN_NATIVE); } static inline int xen_pv_domain(void) { return (xen_domain_type == XEN_PV_DOMAIN); } static inline int xen_hvm_domain(void) { return (xen_domain_type == XEN_HVM_DOMAIN); } static inline bool xen_initial_domain(void) { - return (xen_domain() && (xen_get_start_flags() & SIF_INITDOMAIN) != 0); + return (xen_domain() && (hvm_start_flags & SIF_INITDOMAIN) != 0); } /* * Based on ofed/include/linux/bitops.h * * Those helpers are prefixed by xen_ because xen-os.h is widely included * and we don't want the other drivers using them. * */ #define NBPL (NBBY * sizeof(long)) static inline bool xen_test_bit(int bit, volatile long *addr) { unsigned long mask = 1UL << (bit % NBPL); return !!(atomic_load_acq_long(&addr[bit / NBPL]) & mask); } static inline void xen_set_bit(int bit, volatile long *addr) { atomic_set_long(&addr[bit / NBPL], 1UL << (bit % NBPL)); } static inline void xen_clear_bit(int bit, volatile long *addr) { atomic_clear_long(&addr[bit / NBPL], 1UL << (bit % NBPL)); } #undef NBPL /* * Functions to allocate/free unused memory in order * to map memory from other domains. */ struct resource *xenmem_alloc(device_t dev, int *res_id, size_t size); int xenmem_free(device_t dev, int res_id, struct resource *res); /* Debug/emergency function, prints directly to hypervisor console */ void xc_printf(const char *, ...) __printflike(1, 2); #ifndef xen_mb #define xen_mb() mb() #endif #ifndef xen_rmb #define xen_rmb() rmb() #endif #ifndef xen_wmb #define xen_wmb() wmb() #endif #endif /* !__ASSEMBLY__ */ #endif /* _XEN_XEN_OS_H_ */