ia64/xen-unstable
changeset 15758:778985f246a0
merge with xen-unstable.hg
| author | Alex Williamson <alex.williamson@hp.com> |
|---|---|
| date | Thu Aug 16 10:47:33 2007 -0600 (2007-08-16) |
| parents | b5dbf184df6c 256160ff19b7 |
| children | cb3c7f006077 |
| files |
line diff
1.1 --- a/tools/libfsimage/fat/fsys_fat.c Thu Aug 16 10:03:26 2007 -0600 1.2 +++ b/tools/libfsimage/fat/fsys_fat.c Thu Aug 16 10:47:33 2007 -0600 1.3 @@ -228,15 +228,15 @@ fat_read (fsi_file_t *ffi, char *buf, in 1.4 if (!devread (ffi, sector, 0, FAT_CACHE_SIZE, (char*) FAT_BUF)) 1.5 return 0; 1.6 } 1.7 - next_cluster = * (unsigned long *) (FAT_BUF + (cached_pos >> 1)); 1.8 + next_cluster = ((__u16 *) (FAT_BUF + (cached_pos >> 1)))[0]; 1.9 if (FAT_SUPER->fat_size == 3) 1.10 { 1.11 if (cached_pos & 1) 1.12 next_cluster >>= 4; 1.13 next_cluster &= 0xFFF; 1.14 } 1.15 - else if (FAT_SUPER->fat_size == 4) 1.16 - next_cluster &= 0xFFFF; 1.17 + else if (FAT_SUPER->fat_size > 4) 1.18 + next_cluster |= ((__u16 *) (FAT_BUF + (cached_pos >> 1)))[1] << 16; 1.19 1.20 if (next_cluster >= FAT_SUPER->clust_eof_marker) 1.21 return ret;
2.1 --- a/tools/libxc/xc_domain_restore.c Thu Aug 16 10:03:26 2007 -0600 2.2 +++ b/tools/libxc/xc_domain_restore.c Thu Aug 16 10:47:33 2007 -0600 2.3 @@ -272,7 +272,8 @@ int xc_domain_restore(int xc_handle, int 2.4 /* The new domain's shared-info frame number. */ 2.5 unsigned long shared_info_frame; 2.6 unsigned char shared_info_page[PAGE_SIZE]; /* saved contents from file */ 2.7 - shared_info_t *shared_info = (shared_info_t *)shared_info_page; 2.8 + shared_info_t *old_shared_info = (shared_info_t *)shared_info_page; 2.9 + shared_info_t *new_shared_info; 2.10 2.11 /* A copy of the CPU context of the guest. */ 2.12 vcpu_guest_context_t ctxt; 2.13 @@ -286,9 +287,6 @@ int xc_domain_restore(int xc_handle, int 2.14 /* Types of the pfns in the current region */ 2.15 unsigned long region_pfn_type[MAX_BATCH_SIZE]; 2.16 2.17 - /* A temporary mapping, and a copy, of one frame of guest memory. */ 2.18 - unsigned long *page = NULL; 2.19 - 2.20 /* A copy of the pfn-to-mfn table frame list. */ 2.21 xen_pfn_t *p2m_frame_list = NULL; 2.22 2.23 @@ -1084,17 +1082,30 @@ int xc_domain_restore(int xc_handle, int 2.24 goto out; 2.25 } 2.26 2.27 - /* clear any pending events and the selector */ 2.28 - memset(&(shared_info->evtchn_pending[0]), 0, 2.29 - sizeof (shared_info->evtchn_pending)); 2.30 - for ( i = 0; i < MAX_VIRT_CPUS; i++ ) 2.31 - shared_info->vcpu_info[i].evtchn_pending_sel = 0; 2.32 + /* Restore contents of shared-info page. No checking needed. */ 2.33 + new_shared_info = xc_map_foreign_range( 2.34 + xc_handle, dom, PAGE_SIZE, PROT_WRITE, shared_info_frame); 2.35 + 2.36 + /* restore saved vcpu_info and arch specific info */ 2.37 + memcpy(&new_shared_info->vcpu_info, 2.38 + &old_shared_info->vcpu_info, 2.39 + sizeof(new_shared_info->vcpu_info)); 2.40 + memcpy(&new_shared_info->arch, 2.41 + &old_shared_info->arch, 2.42 + sizeof(new_shared_info->arch)); 2.43 2.44 - /* Copy saved contents of shared-info page. No checking needed. */ 2.45 - page = xc_map_foreign_range( 2.46 - xc_handle, dom, PAGE_SIZE, PROT_WRITE, shared_info_frame); 2.47 - memcpy(page, shared_info, PAGE_SIZE); 2.48 - munmap(page, PAGE_SIZE); 2.49 + /* clear any pending events and the selector */ 2.50 + memset(&(new_shared_info->evtchn_pending[0]), 0, 2.51 + sizeof (new_shared_info->evtchn_pending)); 2.52 + for ( i = 0; i < MAX_VIRT_CPUS; i++ ) 2.53 + new_shared_info->vcpu_info[i].evtchn_pending_sel = 0; 2.54 + 2.55 + /* mask event channels */ 2.56 + memset(&(new_shared_info->evtchn_mask[0]), 0xff, 2.57 + sizeof (new_shared_info->evtchn_mask)); 2.58 + 2.59 + /* leave wallclock time. set by hypervisor */ 2.60 + munmap(new_shared_info, PAGE_SIZE); 2.61 2.62 /* Uncanonicalise the pfn-to-mfn table frame-number list. */ 2.63 for ( i = 0; i < P2M_FL_ENTRIES; i++ )
3.1 --- a/tools/libxc/xc_evtchn.c Thu Aug 16 10:03:26 2007 -0600 3.2 +++ b/tools/libxc/xc_evtchn.c Thu Aug 16 10:47:33 2007 -0600 3.3 @@ -33,9 +33,10 @@ static int do_evtchn_op(int xc_handle, i 3.4 } 3.5 3.6 3.7 -int xc_evtchn_alloc_unbound(int xc_handle, 3.8 - uint32_t dom, 3.9 - uint32_t remote_dom) 3.10 +evtchn_port_or_error_t 3.11 +xc_evtchn_alloc_unbound(int xc_handle, 3.12 + uint32_t dom, 3.13 + uint32_t remote_dom) 3.14 { 3.15 int rc; 3.16 struct evtchn_alloc_unbound arg = {
4.1 --- a/tools/libxc/xc_linux.c Thu Aug 16 10:03:26 2007 -0600 4.2 +++ b/tools/libxc/xc_linux.c Thu Aug 16 10:47:33 2007 -0600 4.3 @@ -254,7 +254,8 @@ int xc_evtchn_notify(int xce_handle, evt 4.4 return ioctl(xce_handle, IOCTL_EVTCHN_NOTIFY, ¬ify); 4.5 } 4.6 4.7 -evtchn_port_t xc_evtchn_bind_unbound_port(int xce_handle, int domid) 4.8 +evtchn_port_or_error_t 4.9 +xc_evtchn_bind_unbound_port(int xce_handle, int domid) 4.10 { 4.11 struct ioctl_evtchn_bind_unbound_port bind; 4.12 4.13 @@ -263,8 +264,9 @@ evtchn_port_t xc_evtchn_bind_unbound_por 4.14 return ioctl(xce_handle, IOCTL_EVTCHN_BIND_UNBOUND_PORT, &bind); 4.15 } 4.16 4.17 -evtchn_port_t xc_evtchn_bind_interdomain(int xce_handle, int domid, 4.18 - evtchn_port_t remote_port) 4.19 +evtchn_port_or_error_t 4.20 +xc_evtchn_bind_interdomain(int xce_handle, int domid, 4.21 + evtchn_port_t remote_port) 4.22 { 4.23 struct ioctl_evtchn_bind_interdomain bind; 4.24 4.25 @@ -274,6 +276,16 @@ evtchn_port_t xc_evtchn_bind_interdomain 4.26 return ioctl(xce_handle, IOCTL_EVTCHN_BIND_INTERDOMAIN, &bind); 4.27 } 4.28 4.29 +evtchn_port_or_error_t 4.30 +xc_evtchn_bind_virq(int xce_handle, unsigned int virq) 4.31 +{ 4.32 + struct ioctl_evtchn_bind_virq bind; 4.33 + 4.34 + bind.virq = virq; 4.35 + 4.36 + return ioctl(xce_handle, IOCTL_EVTCHN_BIND_VIRQ, &bind); 4.37 +} 4.38 + 4.39 int xc_evtchn_unbind(int xce_handle, evtchn_port_t port) 4.40 { 4.41 struct ioctl_evtchn_unbind unbind; 4.42 @@ -283,15 +295,6 @@ int xc_evtchn_unbind(int xce_handle, evt 4.43 return ioctl(xce_handle, IOCTL_EVTCHN_UNBIND, &unbind); 4.44 } 4.45 4.46 -evtchn_port_t xc_evtchn_bind_virq(int xce_handle, unsigned int virq) 4.47 -{ 4.48 - struct ioctl_evtchn_bind_virq bind; 4.49 - 4.50 - bind.virq = virq; 4.51 - 4.52 - return ioctl(xce_handle, IOCTL_EVTCHN_BIND_VIRQ, &bind); 4.53 -} 4.54 - 4.55 static int dorw(int fd, char *data, size_t size, int do_write) 4.56 { 4.57 size_t offset = 0; 4.58 @@ -317,7 +320,8 @@ static int dorw(int fd, char *data, size 4.59 return 0; 4.60 } 4.61 4.62 -evtchn_port_t xc_evtchn_pending(int xce_handle) 4.63 +evtchn_port_or_error_t 4.64 +xc_evtchn_pending(int xce_handle) 4.65 { 4.66 evtchn_port_t port; 4.67
5.1 --- a/tools/libxc/xc_solaris.c Thu Aug 16 10:03:26 2007 -0600 5.2 +++ b/tools/libxc/xc_solaris.c Thu Aug 16 10:47:33 2007 -0600 5.3 @@ -165,7 +165,8 @@ int xc_evtchn_notify(int xce_handle, evt 5.4 return ioctl(xce_handle, IOCTL_EVTCHN_NOTIFY, ¬ify); 5.5 } 5.6 5.7 -evtchn_port_t xc_evtchn_bind_unbound_port(int xce_handle, int domid) 5.8 +evtchn_port_or_error_t 5.9 +xc_evtchn_bind_unbound_port(int xce_handle, int domid) 5.10 { 5.11 struct ioctl_evtchn_bind_unbound_port bind; 5.12 5.13 @@ -174,8 +175,9 @@ evtchn_port_t xc_evtchn_bind_unbound_por 5.14 return ioctl(xce_handle, IOCTL_EVTCHN_BIND_UNBOUND_PORT, &bind); 5.15 } 5.16 5.17 -evtchn_port_t xc_evtchn_bind_interdomain(int xce_handle, int domid, 5.18 - evtchn_port_t remote_port) 5.19 +evtchn_port_or_error_t 5.20 +xc_evtchn_bind_interdomain(int xce_handle, int domid, 5.21 + evtchn_port_t remote_port) 5.22 { 5.23 struct ioctl_evtchn_bind_interdomain bind; 5.24 5.25 @@ -185,6 +187,16 @@ evtchn_port_t xc_evtchn_bind_interdomain 5.26 return ioctl(xce_handle, IOCTL_EVTCHN_BIND_INTERDOMAIN, &bind); 5.27 } 5.28 5.29 +evtchn_port_or_error_t 5.30 +xc_evtchn_bind_virq(int xce_handle, unsigned int virq) 5.31 +{ 5.32 + struct ioctl_evtchn_bind_virq bind; 5.33 + 5.34 + bind.virq = virq; 5.35 + 5.36 + return ioctl(xce_handle, IOCTL_EVTCHN_BIND_VIRQ, &bind); 5.37 +} 5.38 + 5.39 int xc_evtchn_unbind(int xce_handle, evtchn_port_t port) 5.40 { 5.41 struct ioctl_evtchn_unbind unbind; 5.42 @@ -194,15 +206,6 @@ int xc_evtchn_unbind(int xce_handle, evt 5.43 return ioctl(xce_handle, IOCTL_EVTCHN_UNBIND, &unbind); 5.44 } 5.45 5.46 -evtchn_port_t xc_evtchn_bind_virq(int xce_handle, unsigned int virq) 5.47 -{ 5.48 - struct ioctl_evtchn_bind_virq bind; 5.49 - 5.50 - bind.virq = virq; 5.51 - 5.52 - return ioctl(xce_handle, IOCTL_EVTCHN_BIND_VIRQ, &bind); 5.53 -} 5.54 - 5.55 static int dorw(int fd, char *data, size_t size, int do_write) 5.56 { 5.57 size_t offset = 0; 5.58 @@ -228,7 +231,8 @@ static int dorw(int fd, char *data, size 5.59 return 0; 5.60 } 5.61 5.62 -evtchn_port_t xc_evtchn_pending(int xce_handle) 5.63 +evtchn_port_or_error_t 5.64 +xc_evtchn_pending(int xce_handle) 5.65 { 5.66 evtchn_port_t port; 5.67
6.1 --- a/tools/libxc/xenctrl.h Thu Aug 16 10:03:26 2007 -0600 6.2 +++ b/tools/libxc/xenctrl.h Thu Aug 16 10:47:33 2007 -0600 6.3 @@ -449,6 +449,9 @@ int xc_domain_setdebugging(int xc_handle 6.4 * EVENT CHANNEL FUNCTIONS 6.5 */ 6.6 6.7 +/* A port identifier is guaranteed to fit in 31 bits. */ 6.8 +typedef int evtchn_port_or_error_t; 6.9 + 6.10 /** 6.11 * This function allocates an unbound port. Ports are named endpoints used for 6.12 * interdomain communication. This function is most useful in opening a 6.13 @@ -463,13 +466,78 @@ int xc_domain_setdebugging(int xc_handle 6.14 * @parm remote_dom the ID of the domain who will later bind 6.15 * @return allocated port (in @dom) on success, -1 on failure 6.16 */ 6.17 -int xc_evtchn_alloc_unbound(int xc_handle, 6.18 - uint32_t dom, 6.19 - uint32_t remote_dom); 6.20 +evtchn_port_or_error_t 6.21 +xc_evtchn_alloc_unbound(int xc_handle, 6.22 + uint32_t dom, 6.23 + uint32_t remote_dom); 6.24 6.25 int xc_evtchn_reset(int xc_handle, 6.26 uint32_t dom); 6.27 6.28 +/* 6.29 + * Return a handle to the event channel driver, or -1 on failure, in which case 6.30 + * errno will be set appropriately. 6.31 + */ 6.32 +int xc_evtchn_open(void); 6.33 + 6.34 +/* 6.35 + * Close a handle previously allocated with xc_evtchn_open(). 6.36 + */ 6.37 +int xc_evtchn_close(int xce_handle); 6.38 + 6.39 +/* 6.40 + * Return an fd that can be select()ed on for further calls to 6.41 + * xc_evtchn_pending(). 6.42 + */ 6.43 +int xc_evtchn_fd(int xce_handle); 6.44 + 6.45 +/* 6.46 + * Notify the given event channel. Returns -1 on failure, in which case 6.47 + * errno will be set appropriately. 6.48 + */ 6.49 +int xc_evtchn_notify(int xce_handle, evtchn_port_t port); 6.50 + 6.51 +/* 6.52 + * Returns a new event port awaiting interdomain connection from the given 6.53 + * domain ID, or -1 on failure, in which case errno will be set appropriately. 6.54 + */ 6.55 +evtchn_port_or_error_t 6.56 +xc_evtchn_bind_unbound_port(int xce_handle, int domid); 6.57 + 6.58 +/* 6.59 + * Returns a new event port bound to the remote port for the given domain ID, 6.60 + * or -1 on failure, in which case errno will be set appropriately. 6.61 + */ 6.62 +evtchn_port_or_error_t 6.63 +xc_evtchn_bind_interdomain(int xce_handle, int domid, 6.64 + evtchn_port_t remote_port); 6.65 + 6.66 +/* 6.67 + * Bind an event channel to the given VIRQ. Returns the event channel bound to 6.68 + * the VIRQ, or -1 on failure, in which case errno will be set appropriately. 6.69 + */ 6.70 +evtchn_port_or_error_t 6.71 +xc_evtchn_bind_virq(int xce_handle, unsigned int virq); 6.72 + 6.73 +/* 6.74 + * Unbind the given event channel. Returns -1 on failure, in which case errno 6.75 + * will be set appropriately. 6.76 + */ 6.77 +int xc_evtchn_unbind(int xce_handle, evtchn_port_t port); 6.78 + 6.79 +/* 6.80 + * Return the next event channel to become pending, or -1 on failure, in which 6.81 + * case errno will be set appropriately. 6.82 + */ 6.83 +evtchn_port_or_error_t 6.84 +xc_evtchn_pending(int xce_handle); 6.85 + 6.86 +/* 6.87 + * Unmask the given event channel. Returns -1 on failure, in which case errno 6.88 + * will be set appropriately. 6.89 + */ 6.90 +int xc_evtchn_unmask(int xce_handle, evtchn_port_t port); 6.91 + 6.92 int xc_physdev_pci_access_modify(int xc_handle, 6.93 uint32_t domid, 6.94 int bus, 6.95 @@ -699,66 +767,6 @@ int xc_version(int xc_handle, int cmd, v 6.96 6.97 int xc_acm_op(int xc_handle, int cmd, void *arg, unsigned long arg_size); 6.98 6.99 -/* 6.100 - * Return a handle to the event channel driver, or -1 on failure, in which case 6.101 - * errno will be set appropriately. 6.102 - */ 6.103 -int xc_evtchn_open(void); 6.104 - 6.105 -/* 6.106 - * Close a handle previously allocated with xc_evtchn_open(). 6.107 - */ 6.108 -int xc_evtchn_close(int xce_handle); 6.109 - 6.110 -/* 6.111 - * Return an fd that can be select()ed on for further calls to 6.112 - * xc_evtchn_pending(). 6.113 - */ 6.114 -int xc_evtchn_fd(int xce_handle); 6.115 - 6.116 -/* 6.117 - * Notify the given event channel. Returns -1 on failure, in which case 6.118 - * errno will be set appropriately. 6.119 - */ 6.120 -int xc_evtchn_notify(int xce_handle, evtchn_port_t port); 6.121 - 6.122 -/* 6.123 - * Returns a new event port awaiting interdomain connection from the given 6.124 - * domain ID, or -1 on failure, in which case errno will be set appropriately. 6.125 - */ 6.126 -evtchn_port_t xc_evtchn_bind_unbound_port(int xce_handle, int domid); 6.127 - 6.128 -/* 6.129 - * Returns a new event port bound to the remote port for the given domain ID, 6.130 - * or -1 on failure, in which case errno will be set appropriately. 6.131 - */ 6.132 -evtchn_port_t xc_evtchn_bind_interdomain(int xce_handle, int domid, 6.133 - evtchn_port_t remote_port); 6.134 - 6.135 -/* 6.136 - * Unbind the given event channel. Returns -1 on failure, in which case errno 6.137 - * will be set appropriately. 6.138 - */ 6.139 -int xc_evtchn_unbind(int xce_handle, evtchn_port_t port); 6.140 - 6.141 -/* 6.142 - * Bind an event channel to the given VIRQ. Returns the event channel bound to 6.143 - * the VIRQ, or -1 on failure, in which case errno will be set appropriately. 6.144 - */ 6.145 -evtchn_port_t xc_evtchn_bind_virq(int xce_handle, unsigned int virq); 6.146 - 6.147 -/* 6.148 - * Return the next event channel to become pending, or -1 on failure, in which 6.149 - * case errno will be set appropriately. 6.150 - */ 6.151 -evtchn_port_t xc_evtchn_pending(int xce_handle); 6.152 - 6.153 -/* 6.154 - * Unmask the given event channel. Returns -1 on failure, in which case errno 6.155 - * will be set appropriately. 6.156 - */ 6.157 -int xc_evtchn_unmask(int xce_handle, evtchn_port_t port); 6.158 - 6.159 /************************** 6.160 * GRANT TABLE OPERATIONS * 6.161 **************************/
7.1 --- a/tools/python/xen/util/acmpolicy.py Thu Aug 16 10:03:26 2007 -0600 7.2 +++ b/tools/python/xen/util/acmpolicy.py Thu Aug 16 10:47:33 2007 -0600 7.3 @@ -191,20 +191,21 @@ class ACMPolicy(XSPolicy): 7.4 acmpol_old.policy_get_virtualmachinelabel_names_sorted() 7.5 del_array = "" 7.6 chg_array = "" 7.7 + 7.8 for o in oldvmnames: 7.9 if o not in newvmnames: 7.10 - old_idx = oldvmnames.index(o) + 1 # for _NULL_LABEL_ 7.11 + old_idx = oldvmnames.index(o) 7.12 if vmlabel_map.has_key(o): 7.13 #not a deletion, but a renaming 7.14 new = vmlabel_map[o] 7.15 - new_idx = newvmnames.index(new) + 1 # for _NULL_LABEL_ 7.16 + new_idx = newvmnames.index(new) 7.17 chg_array += struct.pack("ii", old_idx, new_idx) 7.18 else: 7.19 del_array += struct.pack("i", old_idx) 7.20 for v in newvmnames: 7.21 if v in oldvmnames: 7.22 - old_idx = oldvmnames.index(v) + 1 # for _NULL_LABEL_ 7.23 - new_idx = newvmnames.index(v) + 1 # for _NULL_LABEL_ 7.24 + old_idx = oldvmnames.index(v) 7.25 + new_idx = newvmnames.index(v) 7.26 if old_idx != new_idx: 7.27 chg_array += struct.pack("ii", old_idx, new_idx) 7.28 7.29 @@ -348,7 +349,7 @@ class ACMPolicy(XSPolicy): 7.30 ssidref = xsconstants.INVALID_SSIDREF 7.31 names = self.policy_get_virtualmachinelabel_names_sorted() 7.32 try: 7.33 - vmidx = names.index(vm_label) + 1 # for _NULL_LABEL_ 7.34 + vmidx = names.index(vm_label) 7.35 ssidref = (vmidx << 16) | vmidx 7.36 except: 7.37 pass 7.38 @@ -618,6 +619,9 @@ class ACMPolicy(XSPolicy): 7.39 vmnames.remove(bootstrap) 7.40 vmnames.sort() 7.41 vmnames.insert(0, bootstrap) 7.42 + if ACM_LABEL_UNLABELED in vmnames: 7.43 + vmnames.remove(ACM_LABEL_UNLABELED) 7.44 + vmnames.insert(0, ACM_LABEL_UNLABELED) 7.45 return vmnames 7.46 7.47 def policy_get_virtualmachinelabel_names_sorted(self): 7.48 @@ -625,7 +629,10 @@ class ACMPolicy(XSPolicy): 7.49 label will be the first one in that list, followed 7.50 by an alphabetically sorted list of VM label names """ 7.51 vmnames = self.policy_get_virtualmachinelabel_names() 7.52 - return self.policy_sort_virtualmachinelabel_names(vmnames) 7.53 + res = self.policy_sort_virtualmachinelabel_names(vmnames) 7.54 + if res[0] != ACM_LABEL_UNLABELED: 7.55 + res.insert(0, ACM_LABEL_UNLABELED) 7.56 + return res 7.57 7.58 def policy_get_virtualmachinelabels(self): 7.59 """ Get a list of all virtual machine labels in this policy """ 7.60 @@ -906,7 +913,7 @@ class ACMPolicy(XSPolicy): 7.61 allvmtypes = self.policy_get_virtualmachinelabel_names_sorted() 7.62 except: 7.63 return None 7.64 - return allvmtypes[chwall_ref-1] # skip _NULL_LABEL_ 7.65 + return allvmtypes[chwall_ref] 7.66 7.67 def policy_get_domain_label_formatted(self, domid): 7.68 label = self.policy_get_domain_label(domid)
8.1 --- a/tools/python/xen/util/security.py Thu Aug 16 10:03:26 2007 -0600 8.2 +++ b/tools/python/xen/util/security.py Thu Aug 16 10:47:33 2007 -0600 8.3 @@ -838,13 +838,28 @@ def set_resource_label_xapi(resource, re 8.4 8.5 8.6 def is_resource_in_use(resource): 8.7 - """ Investigate all running domains whether they use this device """ 8.8 + """ 8.9 + Domain-0 'owns' resources of type 'VLAN', the rest are owned by 8.10 + the guests. 8.11 + """ 8.12 from xen.xend import XendDomain 8.13 - dominfos = XendDomain.instance().list('all') 8.14 lst = [] 8.15 - for dominfo in dominfos: 8.16 - if is_resource_in_use_by_dom(dominfo, resource): 8.17 - lst.append(dominfo) 8.18 + if resource.startswith('vlan'): 8.19 + from xen.xend.XendXSPolicyAdmin import XSPolicyAdminInstance 8.20 + curpol = XSPolicyAdminInstance().get_loaded_policy() 8.21 + policytype, label, policy = get_res_label(resource) 8.22 + if curpol and \ 8.23 + policytype == xsconstants.ACM_POLICY_ID and \ 8.24 + policy == curpol.get_name() and \ 8.25 + label in curpol.policy_get_resourcelabel_names(): 8.26 + # VLAN is in use. 8.27 + lst.append(XendDomain.instance(). 8.28 + get_vm_by_uuid(XendDomain.DOM0_UUID)) 8.29 + else: 8.30 + dominfos = XendDomain.instance().list('all') 8.31 + for dominfo in dominfos: 8.32 + if is_resource_in_use_by_dom(dominfo, resource): 8.33 + lst.append(dominfo) 8.34 return lst 8.35 8.36 def devices_equal(res1, res2, mustexist=True): 8.37 @@ -892,6 +907,10 @@ def get_domain_resources(dominfo): 8.38 if sec_lab: 8.39 resources[typ].append(sec_lab) 8.40 else: 8.41 + # !!! This should really get the label of the domain 8.42 + # or at least a resource label that has the same STE type 8.43 + # as the domain has 8.44 + from xen.util.acmpolicy import ACM_LABEL_UNLABELED 8.45 resources[typ].append("%s:%s:%s" % 8.46 (xsconstants.ACM_POLICY_ID, 8.47 active_policy, 8.48 @@ -924,7 +943,8 @@ def resources_compatible_with_vmlabel(xs 8.49 8.50 8.51 def __resources_compatible_with_vmlabel(xspol, dominfo, vmlabel, 8.52 - access_control): 8.53 + access_control, 8.54 + is_policy_update=False): 8.55 """ 8.56 Check whether the resources' labels are compatible with the 8.57 given VM label. The access_control parameter provides a 8.58 @@ -955,15 +975,23 @@ def __resources_compatible_with_vmlabel( 8.59 elif key in [ 'vif' ]: 8.60 for xapi_label in value: 8.61 label = xapi_label.split(":") 8.62 - if not collect_labels(reslabels, label, polname): 8.63 - return False 8.64 + from xen.util.acmpolicy import ACM_LABEL_UNLABELED 8.65 + if not (is_policy_update and \ 8.66 + label[2] == ACM_LABEL_UNLABELED): 8.67 + if not collect_labels(reslabels, label, polname): 8.68 + return False 8.69 else: 8.70 log.error("Unhandled device type: %s" % key) 8.71 return False 8.72 8.73 # Check that all resource labes have a common STE type with the 8.74 # vmlabel 8.75 - rc = xspol.policy_check_vmlabel_against_reslabels(vmlabel, reslabels) 8.76 + if len(reslabels) > 0: 8.77 + rc = xspol.policy_check_vmlabel_against_reslabels(vmlabel, reslabels) 8.78 + else: 8.79 + rc = True 8.80 + log.info("vmlabel=%s, reslabels=%s, rc=%s" % 8.81 + (vmlabel, reslabels, str(rc))) 8.82 return rc; 8.83 8.84 def set_resource_label(resource, policytype, policyref, reslabel, \ 8.85 @@ -1234,11 +1262,12 @@ def change_acm_policy(bin_pol, del_array 8.86 compatible = __resources_compatible_with_vmlabel(new_acmpol, 8.87 dominfo, 8.88 new_vmlabel, 8.89 - access_control) 8.90 + access_control, 8.91 + is_policy_update=True) 8.92 log.info("Domain %s with new label '%s' can access its " 8.93 "resources? : %s" % 8.94 (name, new_vmlabel, str(compatible))) 8.95 - log.info("VM labels in new domain: %s" % 8.96 + log.info("VM labels in new policy: %s" % 8.97 new_acmpol.policy_get_virtualmachinelabel_names()) 8.98 if not compatible: 8.99 return (-xsconstants.XSERR_RESOURCE_ACCESS, "") 8.100 @@ -1252,11 +1281,16 @@ def change_acm_policy(bin_pol, del_array 8.101 sec_lab, new_seclab = labels 8.102 if sec_lab != new_seclab: 8.103 log.info("Updating domain %s to new label '%s'." % \ 8.104 - (sec_lab, new_seclab)) 8.105 + (dominfo.getName(), new_seclab)) 8.106 # This better be working! 8.107 - dominfo.set_security_label(new_seclab, 8.108 - sec_lab, 8.109 - new_acmpol) 8.110 + res = dominfo.set_security_label(new_seclab, 8.111 + sec_lab, 8.112 + new_acmpol, 8.113 + cur_acmpol) 8.114 + if res[0] != xsconstants.XSERR_SUCCESS: 8.115 + log.info("ERROR: Could not chg label on domain %s: %s" % 8.116 + (dominfo.getName(), 8.117 + xsconstants.xserr2string(-int(res[0])))) 8.118 finally: 8.119 log.info("----------------------------------------------") 8.120 mapfile_unlock()
9.1 --- a/tools/python/xen/xend/XendCheckpoint.py Thu Aug 16 10:03:26 2007 -0600 9.2 +++ b/tools/python/xen/xend/XendCheckpoint.py Thu Aug 16 10:47:33 2007 -0600 9.3 @@ -98,6 +98,9 @@ def save(fd, dominfo, network, live, dst 9.4 log.info("Domain %d suspended.", dominfo.getDomid()) 9.5 dominfo.migrateDevices(network, dst, DEV_MIGRATE_STEP3, 9.6 domain_name) 9.7 + if hvm: 9.8 + dominfo.image.saveDeviceModel() 9.9 + 9.10 tochild.write("done\n") 9.11 tochild.flush() 9.12 log.debug('Written done')
10.1 --- a/tools/python/xen/xend/XendDomainInfo.py Thu Aug 16 10:03:26 2007 -0600 10.2 +++ b/tools/python/xen/xend/XendDomainInfo.py Thu Aug 16 10:47:33 2007 -0600 10.3 @@ -558,9 +558,64 @@ class XendDomainInfo: 10.4 for devclass in XendDevices.valid_devices(): 10.5 self.getDeviceController(devclass).waitForDevices() 10.6 10.7 - def destroyDevice(self, deviceClass, devid, force = False): 10.8 - log.debug("dev = %s", devid) 10.9 - return self.getDeviceController(deviceClass).destroyDevice(devid, force) 10.10 + def destroyDevice(self, deviceClass, devid, force = False, rm_cfg = False): 10.11 + log.debug("XendDomainInfo.destroyDevice: deviceClass = %s, device = %s", 10.12 + deviceClass, devid) 10.13 + 10.14 + if rm_cfg: 10.15 + # Convert devid to device number. A device number is 10.16 + # needed to remove its configuration. 10.17 + dev = self.getDeviceController(deviceClass).convertToDeviceNumber(devid) 10.18 + 10.19 + # Save current sxprs. A device number and a backend 10.20 + # path are needed to remove its configuration but sxprs 10.21 + # do not have those after calling destroyDevice. 10.22 + sxprs = self.getDeviceSxprs(deviceClass) 10.23 + 10.24 + rc = None 10.25 + if self.domid is not None: 10.26 + rc = self.getDeviceController(deviceClass).destroyDevice(devid, force) 10.27 + if not force and rm_cfg: 10.28 + # The backend path, other than the device itself, 10.29 + # has to be passed because its accompanied frontend 10.30 + # path may be void until its removal is actually 10.31 + # issued. It is probable because destroyDevice is 10.32 + # issued first. 10.33 + for dev_num, dev_info in sxprs: 10.34 + dev_num = int(dev_num) 10.35 + if dev_num == dev: 10.36 + for x in dev_info: 10.37 + if x[0] == 'backend': 10.38 + backend = x[1] 10.39 + break 10.40 + break 10.41 + self._waitForDevice_destroy(deviceClass, devid, backend) 10.42 + 10.43 + if rm_cfg: 10.44 + if deviceClass == 'vif': 10.45 + if self.domid is not None: 10.46 + for dev_num, dev_info in sxprs: 10.47 + dev_num = int(dev_num) 10.48 + if dev_num == dev: 10.49 + for x in dev_info: 10.50 + if x[0] == 'mac': 10.51 + mac = x[1] 10.52 + break 10.53 + break 10.54 + dev_info = self.getDeviceInfo_vif(mac) 10.55 + else: 10.56 + _, dev_info = sxprs[dev] 10.57 + else: # 'vbd' or 'tap' 10.58 + dev_info = self.getDeviceInfo_vbd(dev) 10.59 + if dev_info is None: 10.60 + return rc 10.61 + 10.62 + dev_uuid = sxp.child_value(dev_info, 'uuid') 10.63 + del self.info['devices'][dev_uuid] 10.64 + self.info['%s_refs' % deviceClass].remove(dev_uuid) 10.65 + xen.xend.XendDomain.instance().managed_config_save(self) 10.66 + 10.67 + return rc 10.68 10.69 def getDeviceSxprs(self, deviceClass): 10.70 if self._stateGet() in (DOM_STATE_RUNNING, DOM_STATE_PAUSED): 10.71 @@ -574,6 +629,23 @@ class XendDomainInfo: 10.72 dev_num += 1 10.73 return sxprs 10.74 10.75 + def getDeviceInfo_vif(self, mac): 10.76 + for dev_type, dev_info in self.info.all_devices_sxpr(): 10.77 + if dev_type != 'vif': 10.78 + continue 10.79 + if mac == sxp.child_value(dev_info, 'mac'): 10.80 + return dev_info 10.81 + 10.82 + def getDeviceInfo_vbd(self, devid): 10.83 + for dev_type, dev_info in self.info.all_devices_sxpr(): 10.84 + if dev_type != 'vbd' and dev_type != 'tap': 10.85 + continue 10.86 + dev = sxp.child_value(dev_info, 'dev') 10.87 + dev = dev.split(':')[0] 10.88 + dev = self.getDeviceController(dev_type).convertToDeviceNumber(dev) 10.89 + if devid == dev: 10.90 + return dev_info 10.91 + 10.92 10.93 def setMemoryTarget(self, target): 10.94 """Set the memory target of this domain. 10.95 @@ -1112,8 +1184,6 @@ class XendDomainInfo: 10.96 self._clearRestart() 10.97 10.98 if reason == 'suspend': 10.99 - if self._stateGet() != DOM_STATE_SUSPENDED: 10.100 - self.image.saveDeviceModel() 10.101 self._stateSet(DOM_STATE_SUSPENDED) 10.102 # Don't destroy the domain. XendCheckpoint will do 10.103 # this once it has finished. However, stop watching 10.104 @@ -1321,6 +1391,10 @@ class XendDomainInfo: 10.105 deviceClass, config = self.info['devices'].get(dev_uuid) 10.106 self._waitForDevice(deviceClass, config['devid']) 10.107 10.108 + def _waitForDevice_destroy(self, deviceClass, devid, backpath): 10.109 + return self.getDeviceController(deviceClass).waitForDevice_destroy( 10.110 + devid, backpath) 10.111 + 10.112 def _reconfigureDevice(self, deviceClass, devid, devconfig): 10.113 return self.getDeviceController(deviceClass).reconfigureDevice( 10.114 devid, devconfig) 10.115 @@ -2187,11 +2261,18 @@ class XendDomainInfo: 10.116 return self.metrics.get_uuid(); 10.117 10.118 10.119 - def get_security_label(self): 10.120 + def get_security_label(self, xspol=None): 10.121 + """ 10.122 + Get the security label of a domain 10.123 + @param xspol The policy to use when converting the ssid into 10.124 + a label; only to be passed during the updating 10.125 + of the policy 10.126 + """ 10.127 domid = self.getDomid() 10.128 10.129 - from xen.xend.XendXSPolicyAdmin import XSPolicyAdminInstance 10.130 - xspol = XSPolicyAdminInstance().get_loaded_policy() 10.131 + if not xspol: 10.132 + from xen.xend.XendXSPolicyAdmin import XSPolicyAdminInstance 10.133 + xspol = XSPolicyAdminInstance().get_loaded_policy() 10.134 10.135 if domid == 0: 10.136 if xspol: 10.137 @@ -2202,7 +2283,8 @@ class XendDomainInfo: 10.138 label = self.info.get('security_label', '') 10.139 return label 10.140 10.141 - def set_security_label(self, seclab, old_seclab, xspol=None): 10.142 + def set_security_label(self, seclab, old_seclab, xspol=None, 10.143 + xspol_old=None): 10.144 """ 10.145 Set the security label of a domain from its old to 10.146 a new value. 10.147 @@ -2213,6 +2295,8 @@ class XendDomainInfo: 10.148 @param xspol An optional policy under which this 10.149 update should be done. If not given, 10.150 then the current active policy is used. 10.151 + @param xspol_old The old policy; only to be passed during 10.152 + the updating of a policy 10.153 @return Returns return code, a string with errors from 10.154 the hypervisor's operation, old label of the 10.155 domain 10.156 @@ -2223,6 +2307,7 @@ class XendDomainInfo: 10.157 new_ssidref = 0 10.158 domid = self.getDomid() 10.159 res_labels = None 10.160 + is_policy_update = (xspol_old != None) 10.161 10.162 from xen.xend.XendXSPolicyAdmin import XSPolicyAdminInstance 10.163 from xen.util import xsconstants 10.164 @@ -2276,13 +2361,16 @@ class XendDomainInfo: 10.165 10.166 # Check that all used resources are accessible under the 10.167 # new label 10.168 - if not security.resources_compatible_with_vmlabel(xspol, 10.169 + if not is_policy_update and \ 10.170 + not security.resources_compatible_with_vmlabel(xspol, 10.171 self, label): 10.172 return (-xsconstants.XSERR_BAD_LABEL, "", "", 0) 10.173 10.174 #Check label against expected one. 10.175 - old_label = self.get_security_label() 10.176 + old_label = self.get_security_label(xspol_old) 10.177 if old_label != old_seclab: 10.178 + log.info("old_label != old_seclab: %s != %s" % 10.179 + (old_label, old_seclab)) 10.180 return (-xsconstants.XSERR_BAD_LABEL, "", "", 0) 10.181 10.182 # relabel domain in the hypervisor
11.1 --- a/tools/python/xen/xend/server/DevController.py Thu Aug 16 10:03:26 2007 -0600 11.2 +++ b/tools/python/xen/xend/server/DevController.py Thu Aug 16 10:47:33 2007 -0600 11.3 @@ -28,17 +28,19 @@ from xen.xend.xenstore.xswatch import xs 11.4 11.5 import os 11.6 11.7 -DEVICE_CREATE_TIMEOUT = 100 11.8 +DEVICE_CREATE_TIMEOUT = 100 11.9 +DEVICE_DESTROY_TIMEOUT = 100 11.10 HOTPLUG_STATUS_NODE = "hotplug-status" 11.11 HOTPLUG_ERROR_NODE = "hotplug-error" 11.12 HOTPLUG_STATUS_ERROR = "error" 11.13 HOTPLUG_STATUS_BUSY = "busy" 11.14 11.15 -Connected = 1 11.16 -Error = 2 11.17 -Missing = 3 11.18 -Timeout = 4 11.19 -Busy = 5 11.20 +Connected = 1 11.21 +Error = 2 11.22 +Missing = 3 11.23 +Timeout = 4 11.24 +Busy = 5 11.25 +Disconnected = 6 11.26 11.27 xenbusState = { 11.28 'Unknown' : 0, 11.29 @@ -185,6 +187,18 @@ class DevController: 11.30 (devid, self.deviceClass, err)) 11.31 11.32 11.33 + def waitForDevice_destroy(self, devid, backpath): 11.34 + log.debug("Waiting for %s - destroyDevice.", devid) 11.35 + 11.36 + if not self.hotplug: 11.37 + return 11.38 + 11.39 + status = self.waitForBackend_destroy(backpath) 11.40 + 11.41 + if status == Timeout: 11.42 + raise VmError("Device %s (%s) could not be disconnected. " % 11.43 + (devid, self.deviceClass)) 11.44 + 11.45 11.46 def reconfigureDevice(self, devid, config): 11.47 """Reconfigure the specified device. 11.48 @@ -209,12 +223,7 @@ class DevController: 11.49 here. 11.50 """ 11.51 11.52 - try: 11.53 - dev = int(devid) 11.54 - except ValueError: 11.55 - # Does devid contain devicetype/deviceid? 11.56 - # Propogate exception if unable to find an integer devid 11.57 - dev = int(type(devid) is str and devid.split('/')[-1] or None) 11.58 + dev = self.convertToDeviceNumber(devid) 11.59 11.60 # Modify online status /before/ updating state (latter is watched by 11.61 # drivers, so this ordering avoids a race). 11.62 @@ -283,6 +292,15 @@ class DevController: 11.63 all_configs[devid] = config_dict 11.64 return all_configs 11.65 11.66 + 11.67 + def convertToDeviceNumber(self, devid): 11.68 + try: 11.69 + return int(devid) 11.70 + except ValueError: 11.71 + # Does devid contain devicetype/deviceid? 11.72 + # Propogate exception if unable to find an integer devid 11.73 + return int(type(devid) is str and devid.split('/')[-1] or None) 11.74 + 11.75 ## protected: 11.76 11.77 def getDeviceDetails(self, config): 11.78 @@ -513,6 +531,19 @@ class DevController: 11.79 return (Missing, None) 11.80 11.81 11.82 + def waitForBackend_destroy(self, backpath): 11.83 + 11.84 + statusPath = backpath + '/' + HOTPLUG_STATUS_NODE 11.85 + ev = Event() 11.86 + result = { 'status': Timeout } 11.87 + 11.88 + xswatch(statusPath, deviceDestroyCallback, ev, result) 11.89 + 11.90 + ev.wait(DEVICE_DESTROY_TIMEOUT) 11.91 + 11.92 + return result['status'] 11.93 + 11.94 + 11.95 def backendPath(self, backdom, devid): 11.96 """Construct backend path given the backend domain and device id. 11.97 11.98 @@ -561,3 +592,19 @@ def hotplugStatusCallback(statusPath, ev 11.99 11.100 ev.set() 11.101 return 0 11.102 + 11.103 + 11.104 +def deviceDestroyCallback(statusPath, ev, result): 11.105 + log.debug("deviceDestroyCallback %s.", statusPath) 11.106 + 11.107 + status = xstransact.Read(statusPath) 11.108 + 11.109 + if status is None: 11.110 + result['status'] = Disconnected 11.111 + else: 11.112 + return 1 11.113 + 11.114 + log.debug("deviceDestroyCallback %d.", result['status']) 11.115 + 11.116 + ev.set() 11.117 + return 0
12.1 --- a/tools/python/xen/xend/server/blkif.py Thu Aug 16 10:03:26 2007 -0600 12.2 +++ b/tools/python/xen/xend/server/blkif.py Thu Aug 16 10:47:33 2007 -0600 12.3 @@ -165,11 +165,23 @@ class BlkifController(DevController): 12.4 try: 12.5 DevController.destroyDevice(self, devid, force) 12.6 except ValueError: 12.7 - devid_end = type(devid) is str and devid.split('/')[-1] or None 12.8 + dev = self.convertToDeviceNumber(devid) 12.9 12.10 for i in self.deviceIDs(): 12.11 - d = self.readBackend(i, 'dev') 12.12 - if d == devid or (devid_end and d == devid_end): 12.13 + if i == dev: 12.14 DevController.destroyDevice(self, i, force) 12.15 return 12.16 raise VmError("Device %s not connected" % devid) 12.17 + 12.18 + def convertToDeviceNumber(self, devid): 12.19 + try: 12.20 + dev = int(devid) 12.21 + except ValueError: 12.22 + if type(devid) is not str: 12.23 + raise VmError("devid %s is wrong type" % str(devid)) 12.24 + try: 12.25 + dev = devid.split('/')[-1] 12.26 + dev = int(dev) 12.27 + except ValueError: 12.28 + dev = blkif.blkdev_name_to_number(dev) 12.29 + return dev
13.1 --- a/tools/python/xen/xm/main.py Thu Aug 16 10:03:26 2007 -0600 13.2 +++ b/tools/python/xen/xm/main.py Thu Aug 16 10:47:33 2007 -0600 13.3 @@ -876,7 +876,7 @@ def parse_doms_info(info): 13.4 if len(tmp) != 3: 13.5 seclabel = "" 13.6 else: 13.7 - seclabel = tmp[2] 13.8 + seclabel = security_label 13.9 parsed_info['seclabel'] = seclabel 13.10 13.11 if serverType == SERVER_XEN_API: 13.12 @@ -2186,6 +2186,7 @@ def xm_network_attach(args): 13.13 13.14 13.15 def detach(args, deviceClass): 13.16 + rm_cfg = True 13.17 dom = args[0] 13.18 dev = args[1] 13.19 try: 13.20 @@ -2196,7 +2197,7 @@ def detach(args, deviceClass): 13.21 except IndexError: 13.22 force = None 13.23 13.24 - server.xend.domain.destroyDevice(dom, deviceClass, dev, force) 13.25 + server.xend.domain.destroyDevice(dom, deviceClass, dev, force, rm_cfg) 13.26 13.27 13.28 def xm_block_detach(args):
14.1 --- a/tools/xm-test/configure.ac Thu Aug 16 10:03:26 2007 -0600 14.2 +++ b/tools/xm-test/configure.ac Thu Aug 16 10:47:33 2007 -0600 14.3 @@ -85,6 +85,13 @@ AC_SUBST(NET_IP_RANGE) 14.4 AC_SUBST(NETWORK_ADDRESS) 14.5 AC_SUBST(NETMASK) 14.6 14.7 +DOM0_INTF="vif0.0" 14.8 +AC_ARG_WITH(dom0-intf, 14.9 + [ --with-dom0-intf=intf Set dom0 interface name [[default="vif0.0"]]], 14.10 + [ DOM0_INTF="$withval" ]) 14.11 + 14.12 +AC_SUBST(DOM0_INTF) 14.13 + 14.14 AC_ARG_WITH(hvm-kernel, 14.15 [[ --with-hvm-kernel=kernel Use this kernel for hvm disk.img testing]], 14.16 HVMKERNEL=$withval,
15.1 --- a/tools/xm-test/lib/XmTestLib/NetConfig.py Thu Aug 16 10:03:26 2007 -0600 15.2 +++ b/tools/xm-test/lib/XmTestLib/NetConfig.py Thu Aug 16 10:47:33 2007 -0600 15.3 @@ -104,8 +104,8 @@ class NetConfig: 15.4 if self.network == "169.254.0.0": 15.5 checkZeroconfAddresses() 15.6 15.7 - # Clean out any aliases in the network range for vif0.0. If 15.8 - # an alias exists, a test xendevice add command could fail. 15.9 + # Clean out any aliases in the network range for dom0's interface. 15.10 + # If an alias exists, a test xendevice add command could fail. 15.11 if NETWORK_IP_RANGE != "dhcp": 15.12 self.__cleanDom0Aliases() 15.13 15.14 @@ -139,20 +139,22 @@ class NetConfig: 15.15 15.16 def __cleanDom0Aliases(self): 15.17 # Remove any aliases within the supplied network IP range on dom0 15.18 - scmd = 'ip addr show dev vif0.0' 15.19 + scmd = 'ip addr show dev %s' % (DOM0_INTF) 15.20 15.21 status, output = traceCommand(scmd) 15.22 if status: 15.23 - raise NetworkError("Failed to show vif0.0 aliases: %d" % status) 15.24 + raise NetworkError("Failed to show %s aliases: %d" % 15.25 + (DOM0_INTF, status)) 15.26 15.27 lines = output.split("\n") 15.28 for line in lines: 15.29 ip = re.search('(\d+\.\d+\.\d+\.\d+)', line) 15.30 if ip and self.isIPInRange(ip.group(1)) == True: 15.31 - dcmd = 'ip addr del %s dev vif0.0' % ip.group(1) 15.32 + dcmd = 'ip addr del %s dev %s' % (ip.group(1), DOM0_INTF) 15.33 dstatus, doutput = traceCommand(dcmd) 15.34 if dstatus: 15.35 - raise NetworkError("Failed to remove vif0.0 aliases: %d" % status) 15.36 + raise NetworkError("Failed to remove %s aliases: %d" % 15.37 + (DOM0_INTF, status)) 15.38 15.39 def getNetEnv(self): 15.40 return self.netenv
16.1 --- a/tools/xm-test/lib/XmTestLib/XenDevice.py Thu Aug 16 10:03:26 2007 -0600 16.2 +++ b/tools/xm-test/lib/XmTestLib/XenDevice.py Thu Aug 16 10:47:33 2007 -0600 16.3 @@ -214,7 +214,7 @@ class XenNetDevice(XenDevice): 16.4 def removeDevice(self): 16.5 self.releaseNetDevIP() 16.6 16.7 - def addDom0AliasCmd(self, dev="vif0.0"): 16.8 + def addDom0AliasCmd(self, dev=DOM0_INTF): 16.9 # Method to add start and remove dom0 alias cmds 16.10 acmd = 'ip addr add %s dev %s' % (self.dom0_alias_ip, dev) 16.11 rcmd = 'ip addr del %s dev %s' % (self.dom0_alias_ip, dev)
17.1 --- a/tools/xm-test/lib/XmTestLib/config.py.in Thu Aug 16 10:03:26 2007 -0600 17.2 +++ b/tools/xm-test/lib/XmTestLib/config.py.in Thu Aug 16 10:47:33 2007 -0600 17.3 @@ -4,3 +4,4 @@ ENABLE_HVM_SUPPORT = @ENABLE_HVM@ 17.4 NETWORK_IP_RANGE = "@NET_IP_RANGE@" 17.5 NETWORK = "@NETWORK_ADDRESS@" 17.6 NETMASK = "@NETMASK@" 17.7 +DOM0_INTF = "@DOM0_INTF@"
18.1 --- a/unmodified_drivers/linux-2.6/platform-pci/machine_reboot.c Thu Aug 16 10:03:26 2007 -0600 18.2 +++ b/unmodified_drivers/linux-2.6/platform-pci/machine_reboot.c Thu Aug 16 10:47:33 2007 -0600 18.3 @@ -1,5 +1,6 @@ 18.4 #include <linux/config.h> 18.5 -#include <linux/stop_machine.h> 18.6 +#include <linux/cpumask.h> 18.7 +#include <linux/preempt.h> 18.8 #include <xen/evtchn.h> 18.9 #include <xen/gnttab.h> 18.10 #include <xen/xenbus.h>
19.1 --- a/xen/acm/acm_policy.c Thu Aug 16 10:03:26 2007 -0600 19.2 +++ b/xen/acm/acm_policy.c Thu Aug 16 10:47:33 2007 -0600 19.3 @@ -710,12 +710,12 @@ acm_change_policy(struct acm_change_poli 19.4 goto acm_chg_policy_exit; 19.5 } 19.6 19.7 - if ( copy_from_guest((u8 *)dels.array, 19.8 + if ( copy_from_guest(dels.array, 19.9 chgpolicy->del_array, 19.10 - chgpolicy->delarray_size) || 19.11 - copy_from_guest((u8 *)ssidmap.array, 19.12 + dels.num_items) || 19.13 + copy_from_guest(ssidmap.array, 19.14 chgpolicy->chg_array, 19.15 - chgpolicy->chgarray_size) || 19.16 + ssidmap.num_items) || 19.17 copy_from_guest(binpolicy, 19.18 chgpolicy->policy_pushcache, 19.19 chgpolicy->policy_pushcache_size )) 19.20 @@ -844,9 +844,9 @@ acm_relabel_domains(struct acm_relabel_d 19.21 memset(errors.array, 0x0, sizeof(uint32_t) * errors.num_items); 19.22 } 19.23 19.24 - if ( copy_from_guest((u8 *)relabels.array, 19.25 + if ( copy_from_guest(relabels.array, 19.26 relabel->relabel_map, 19.27 - relabel->relabel_map_size) ) 19.28 + relabels.num_items) ) 19.29 { 19.30 rc = -EFAULT; 19.31 goto acm_relabel_doms_exit;
20.1 --- a/xen/arch/x86/acpi/boot.c Thu Aug 16 10:03:26 2007 -0600 20.2 +++ b/xen/arch/x86/acpi/boot.c Thu Aug 16 10:47:33 2007 -0600 20.3 @@ -423,7 +423,7 @@ acpi_fadt_parse_sleep_info(struct fadt_d 20.4 goto bad; 20.5 20.6 if (strncmp(facs->signature, "FACS", 4)) { 20.7 - printk(KERN_ERR PREFIX "Invalid FACS signature %s\n", 20.8 + printk(KERN_ERR PREFIX "Invalid FACS signature %.4s\n", 20.9 facs->signature); 20.10 goto bad; 20.11 } 20.12 @@ -451,12 +451,13 @@ acpi_fadt_parse_sleep_info(struct fadt_d 20.13 acpi_sinfo.vector_width = 64; 20.14 } 20.15 20.16 - printk (KERN_INFO PREFIX 20.17 - "ACPI SLEEP INFO: pm1x_cnt[%x,%x], pm1x_evt[%x,%x]\n" 20.18 - " wakeup_vec[%"PRIx64"], vec_size[%x]\n", 20.19 - acpi_sinfo.pm1a_cnt, acpi_sinfo.pm1b_cnt, 20.20 - acpi_sinfo.pm1a_evt, acpi_sinfo.pm1b_cnt, 20.21 - acpi_sinfo.wakeup_vector, acpi_sinfo.vector_width); 20.22 + printk(KERN_INFO PREFIX 20.23 + "ACPI SLEEP INFO: pm1x_cnt[%x,%x], pm1x_evt[%x,%x]\n", 20.24 + acpi_sinfo.pm1a_cnt, acpi_sinfo.pm1b_cnt, 20.25 + acpi_sinfo.pm1a_evt, acpi_sinfo.pm1b_cnt); 20.26 + printk(KERN_INFO PREFIX 20.27 + " wakeup_vec[%"PRIx64"], vec_size[%x]\n", 20.28 + acpi_sinfo.wakeup_vector, acpi_sinfo.vector_width); 20.29 return; 20.30 bad: 20.31 memset(&acpi_sinfo, 0, sizeof(acpi_sinfo));
21.1 --- a/xen/arch/x86/hvm/hvm.c Thu Aug 16 10:03:26 2007 -0600 21.2 +++ b/xen/arch/x86/hvm/hvm.c Thu Aug 16 10:47:33 2007 -0600 21.3 @@ -76,13 +76,6 @@ void hvm_enable(struct hvm_function_tabl 21.4 hvm_enabled = 1; 21.5 } 21.6 21.7 -void hvm_stts(struct vcpu *v) 21.8 -{ 21.9 - /* FPU state already dirty? Then no need to setup_fpu() lazily. */ 21.10 - if ( !v->fpu_dirtied ) 21.11 - hvm_funcs.stts(v); 21.12 -} 21.13 - 21.14 void hvm_set_guest_time(struct vcpu *v, u64 gtime) 21.15 { 21.16 u64 host_tsc; 21.17 @@ -112,7 +105,8 @@ void hvm_do_resume(struct vcpu *v) 21.18 { 21.19 ioreq_t *p; 21.20 21.21 - hvm_stts(v); 21.22 + if ( !v->fpu_dirtied ) 21.23 + hvm_funcs.stts(v); 21.24 21.25 pt_thaw_time(v); 21.26 21.27 @@ -520,6 +514,174 @@ void hvm_triple_fault(void) 21.28 domain_shutdown(v->domain, SHUTDOWN_reboot); 21.29 } 21.30 21.31 +int hvm_set_cr0(unsigned long value) 21.32 +{ 21.33 + struct vcpu *v = current; 21.34 + unsigned long mfn, old_base_mfn, old_value = v->arch.hvm_vcpu.guest_cr[0]; 21.35 + 21.36 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR0 value = %lx", value); 21.37 + 21.38 + if ( (u32)value != value ) 21.39 + { 21.40 + HVM_DBG_LOG(DBG_LEVEL_1, 21.41 + "Guest attempts to set upper 32 bits in CR0: %lx", 21.42 + value); 21.43 + hvm_inject_exception(TRAP_gp_fault, 0, 0); 21.44 + return 0; 21.45 + } 21.46 + 21.47 + value &= ~HVM_CR0_GUEST_RESERVED_BITS; 21.48 + 21.49 + /* ET is reserved and should be always be 1. */ 21.50 + value |= X86_CR0_ET; 21.51 + 21.52 + if ( (value & (X86_CR0_PE|X86_CR0_PG)) == X86_CR0_PG ) 21.53 + { 21.54 + hvm_inject_exception(TRAP_gp_fault, 0, 0); 21.55 + return 0; 21.56 + } 21.57 + 21.58 + if ( (value & X86_CR0_PG) && !(old_value & X86_CR0_PG) ) 21.59 + { 21.60 + if ( v->arch.hvm_vcpu.guest_efer & EFER_LME ) 21.61 + { 21.62 + if ( !(v->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PAE) ) 21.63 + { 21.64 + HVM_DBG_LOG(DBG_LEVEL_1, "Enable paging before PAE enable"); 21.65 + hvm_inject_exception(TRAP_gp_fault, 0, 0); 21.66 + return 0; 21.67 + } 21.68 + HVM_DBG_LOG(DBG_LEVEL_1, "Enabling long mode"); 21.69 + v->arch.hvm_vcpu.guest_efer |= EFER_LMA; 21.70 + hvm_update_guest_efer(v); 21.71 + } 21.72 + 21.73 + if ( !paging_mode_hap(v->domain) ) 21.74 + { 21.75 + /* The guest CR3 must be pointing to the guest physical. */ 21.76 + mfn = get_mfn_from_gpfn(v->arch.hvm_vcpu.guest_cr[3]>>PAGE_SHIFT); 21.77 + if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain)) 21.78 + { 21.79 + gdprintk(XENLOG_ERR, "Invalid CR3 value = %lx (mfn=%lx)\n", 21.80 + v->arch.hvm_vcpu.guest_cr[3], mfn); 21.81 + domain_crash(v->domain); 21.82 + return 0; 21.83 + } 21.84 + 21.85 + /* Now arch.guest_table points to machine physical. */ 21.86 + old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 21.87 + v->arch.guest_table = pagetable_from_pfn(mfn); 21.88 + if ( old_base_mfn ) 21.89 + put_page(mfn_to_page(old_base_mfn)); 21.90 + 21.91 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx, mfn = %lx", 21.92 + v->arch.hvm_vcpu.guest_cr[3], mfn); 21.93 + } 21.94 + } 21.95 + else if ( !(value & X86_CR0_PG) && (old_value & X86_CR0_PG) ) 21.96 + { 21.97 + /* When CR0.PG is cleared, LMA is cleared immediately. */ 21.98 + if ( hvm_long_mode_enabled(v) ) 21.99 + { 21.100 + v->arch.hvm_vcpu.guest_efer &= ~EFER_LMA; 21.101 + hvm_update_guest_efer(v); 21.102 + } 21.103 + 21.104 + if ( !paging_mode_hap(v->domain) ) 21.105 + { 21.106 + put_page(mfn_to_page(get_mfn_from_gpfn( 21.107 + v->arch.hvm_vcpu.guest_cr[3] >> PAGE_SHIFT))); 21.108 + v->arch.guest_table = pagetable_null(); 21.109 + } 21.110 + } 21.111 + 21.112 + v->arch.hvm_vcpu.guest_cr[0] = value; 21.113 + hvm_update_guest_cr(v, 0); 21.114 + 21.115 + if ( (value ^ old_value) & X86_CR0_PG ) 21.116 + paging_update_paging_modes(v); 21.117 + 21.118 + return 1; 21.119 +} 21.120 + 21.121 +int hvm_set_cr3(unsigned long value) 21.122 +{ 21.123 + unsigned long old_base_mfn, mfn; 21.124 + struct vcpu *v = current; 21.125 + 21.126 + if ( paging_mode_hap(v->domain) || !hvm_paging_enabled(v) ) 21.127 + { 21.128 + /* Nothing to do. */ 21.129 + } 21.130 + else if ( value == v->arch.hvm_vcpu.guest_cr[3] ) 21.131 + { 21.132 + /* Shadow-mode TLB flush. Invalidate the shadow. */ 21.133 + mfn = get_mfn_from_gpfn(value >> PAGE_SHIFT); 21.134 + if ( mfn != pagetable_get_pfn(v->arch.guest_table) ) 21.135 + goto bad_cr3; 21.136 + } 21.137 + else 21.138 + { 21.139 + /* Shadow-mode CR3 change. Check PDBR and then make a new shadow. */ 21.140 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 value = %lx", value); 21.141 + mfn = get_mfn_from_gpfn(value >> PAGE_SHIFT); 21.142 + if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 21.143 + goto bad_cr3; 21.144 + 21.145 + old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 21.146 + v->arch.guest_table = pagetable_from_pfn(mfn); 21.147 + 21.148 + if ( old_base_mfn ) 21.149 + put_page(mfn_to_page(old_base_mfn)); 21.150 + 21.151 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx", value); 21.152 + } 21.153 + 21.154 + v->arch.hvm_vcpu.guest_cr[3] = value; 21.155 + paging_update_cr3(v); 21.156 + return 1; 21.157 + 21.158 + bad_cr3: 21.159 + gdprintk(XENLOG_ERR, "Invalid CR3\n"); 21.160 + domain_crash(v->domain); 21.161 + return 0; 21.162 +} 21.163 + 21.164 +int hvm_set_cr4(unsigned long value) 21.165 +{ 21.166 + struct vcpu *v = current; 21.167 + unsigned long old_cr; 21.168 + 21.169 + if ( value & HVM_CR4_GUEST_RESERVED_BITS ) 21.170 + { 21.171 + HVM_DBG_LOG(DBG_LEVEL_1, 21.172 + "Guest attempts to set reserved bit in CR4: %lx", 21.173 + value); 21.174 + goto gpf; 21.175 + } 21.176 + 21.177 + if ( !(value & X86_CR4_PAE) && hvm_long_mode_enabled(v) ) 21.178 + { 21.179 + HVM_DBG_LOG(DBG_LEVEL_1, "Guest cleared CR4.PAE while " 21.180 + "EFER.LMA is set"); 21.181 + goto gpf; 21.182 + } 21.183 + 21.184 + old_cr = v->arch.hvm_vcpu.guest_cr[4]; 21.185 + v->arch.hvm_vcpu.guest_cr[4] = value; 21.186 + hvm_update_guest_cr(v, 4); 21.187 + 21.188 + /* Modifying CR4.{PSE,PAE,PGE} invalidates all TLB entries, inc. Global. */ 21.189 + if ( (old_cr ^ value) & (X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE) ) 21.190 + paging_update_paging_modes(v); 21.191 + 21.192 + return 1; 21.193 + 21.194 + gpf: 21.195 + hvm_inject_exception(TRAP_gp_fault, 0, 0); 21.196 + return 0; 21.197 +} 21.198 + 21.199 /* 21.200 * __hvm_copy(): 21.201 * @buf = hypervisor buffer 21.202 @@ -668,7 +830,6 @@ typedef unsigned long hvm_hypercall_t( 21.203 static hvm_hypercall_t *hvm_hypercall32_table[NR_hypercalls] = { 21.204 HYPERCALL(memory_op), 21.205 [ __HYPERVISOR_grant_table_op ] = (hvm_hypercall_t *)hvm_grant_table_op, 21.206 - HYPERCALL(multicall), 21.207 HYPERCALL(xen_version), 21.208 HYPERCALL(grant_table_op), 21.209 HYPERCALL(event_channel_op), 21.210 @@ -813,12 +974,6 @@ int hvm_do_hypercall(struct cpu_user_reg 21.211 flush ? HVM_HCALL_invalidate : HVM_HCALL_completed); 21.212 } 21.213 21.214 -void hvm_update_guest_cr3(struct vcpu *v, unsigned long guest_cr3) 21.215 -{ 21.216 - v->arch.hvm_vcpu.hw_cr3 = guest_cr3; 21.217 - hvm_funcs.update_guest_cr3(v); 21.218 -} 21.219 - 21.220 static void hvm_latch_shinfo_size(struct domain *d) 21.221 { 21.222 /*
22.1 --- a/xen/arch/x86/hvm/svm/svm.c Thu Aug 16 10:03:26 2007 -0600 22.2 +++ b/xen/arch/x86/hvm/svm/svm.c Thu Aug 16 10:47:33 2007 -0600 22.3 @@ -59,8 +59,9 @@ int inst_copy_from_guest(unsigned char * 22.4 int inst_len); 22.5 asmlinkage void do_IRQ(struct cpu_user_regs *); 22.6 22.7 -static int svm_reset_to_realmode(struct vcpu *v, 22.8 - struct cpu_user_regs *regs); 22.9 +static int svm_reset_to_realmode( 22.10 + struct vcpu *v, struct cpu_user_regs *regs); 22.11 +static void svm_update_guest_cr(struct vcpu *v, unsigned int cr); 22.12 22.13 /* va of hardware host save area */ 22.14 static void *hsa[NR_CPUS] __read_mostly; 22.15 @@ -78,7 +79,7 @@ static void svm_inject_exception( 22.16 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.17 22.18 if ( trap == TRAP_page_fault ) 22.19 - HVMTRACE_2D(PF_INJECT, v, v->arch.hvm_svm.cpu_cr2, error_code); 22.20 + HVMTRACE_2D(PF_INJECT, v, v->arch.hvm_vcpu.guest_cr[2], error_code); 22.21 else 22.22 HVMTRACE_2D(INJ_EXC, v, trap, error_code); 22.23 22.24 @@ -97,55 +98,14 @@ static void svm_cpu_down(void) 22.25 write_efer(read_efer() & ~EFER_SVME); 22.26 } 22.27 22.28 -#ifdef __x86_64__ 22.29 - 22.30 static int svm_lme_is_set(struct vcpu *v) 22.31 { 22.32 - u64 guest_efer = v->arch.hvm_svm.cpu_shadow_efer; 22.33 +#ifdef __x86_64__ 22.34 + u64 guest_efer = v->arch.hvm_vcpu.guest_efer; 22.35 return guest_efer & EFER_LME; 22.36 -} 22.37 - 22.38 -static int svm_long_mode_enabled(struct vcpu *v) 22.39 -{ 22.40 - u64 guest_efer = v->arch.hvm_svm.cpu_shadow_efer; 22.41 - return guest_efer & EFER_LMA; 22.42 -} 22.43 - 22.44 -#else /* __i386__ */ 22.45 - 22.46 -static int svm_lme_is_set(struct vcpu *v) 22.47 -{ return 0; } 22.48 -static int svm_long_mode_enabled(struct vcpu *v) 22.49 -{ return 0; } 22.50 - 22.51 +#else 22.52 + return 0; 22.53 #endif 22.54 - 22.55 -static int svm_cr4_pae_is_set(struct vcpu *v) 22.56 -{ 22.57 - unsigned long guest_cr4 = v->arch.hvm_svm.cpu_shadow_cr4; 22.58 - return guest_cr4 & X86_CR4_PAE; 22.59 -} 22.60 - 22.61 -static int svm_paging_enabled(struct vcpu *v) 22.62 -{ 22.63 - unsigned long guest_cr0 = v->arch.hvm_svm.cpu_shadow_cr0; 22.64 - return (guest_cr0 & X86_CR0_PE) && (guest_cr0 & X86_CR0_PG); 22.65 -} 22.66 - 22.67 -static int svm_pae_enabled(struct vcpu *v) 22.68 -{ 22.69 - unsigned long guest_cr4 = v->arch.hvm_svm.cpu_shadow_cr4; 22.70 - return svm_paging_enabled(v) && (guest_cr4 & X86_CR4_PAE); 22.71 -} 22.72 - 22.73 -static int svm_nx_enabled(struct vcpu *v) 22.74 -{ 22.75 - return v->arch.hvm_svm.cpu_shadow_efer & EFER_NX; 22.76 -} 22.77 - 22.78 -static int svm_pgbit_test(struct vcpu *v) 22.79 -{ 22.80 - return v->arch.hvm_svm.cpu_shadow_cr0 & X86_CR0_PG; 22.81 } 22.82 22.83 static void svm_store_cpu_guest_regs( 22.84 @@ -165,10 +125,10 @@ static void svm_store_cpu_guest_regs( 22.85 if ( crs != NULL ) 22.86 { 22.87 /* Returning the guest's regs */ 22.88 - crs[0] = v->arch.hvm_svm.cpu_shadow_cr0; 22.89 - crs[2] = v->arch.hvm_svm.cpu_cr2; 22.90 - crs[3] = v->arch.hvm_svm.cpu_cr3; 22.91 - crs[4] = v->arch.hvm_svm.cpu_shadow_cr4; 22.92 + crs[0] = v->arch.hvm_vcpu.guest_cr[0]; 22.93 + crs[2] = v->arch.hvm_vcpu.guest_cr[2]; 22.94 + crs[3] = v->arch.hvm_vcpu.guest_cr[3]; 22.95 + crs[4] = v->arch.hvm_vcpu.guest_cr[4]; 22.96 } 22.97 } 22.98 22.99 @@ -202,7 +162,8 @@ static enum handler_return long_mode_do_ 22.100 if ( (msr_content & EFER_LME) && !svm_lme_is_set(v) ) 22.101 { 22.102 /* EFER.LME transition from 0 to 1. */ 22.103 - if ( svm_paging_enabled(v) || !svm_cr4_pae_is_set(v) ) 22.104 + if ( hvm_paging_enabled(v) || 22.105 + !(v->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PAE) ) 22.106 { 22.107 gdprintk(XENLOG_WARNING, "Trying to set LME bit when " 22.108 "in paging mode or PAE bit is not set\n"); 22.109 @@ -212,7 +173,7 @@ static enum handler_return long_mode_do_ 22.110 else if ( !(msr_content & EFER_LME) && svm_lme_is_set(v) ) 22.111 { 22.112 /* EFER.LME transistion from 1 to 0. */ 22.113 - if ( svm_paging_enabled(v) ) 22.114 + if ( hvm_paging_enabled(v) ) 22.115 { 22.116 gdprintk(XENLOG_WARNING, 22.117 "Trying to clear EFER.LME while paging enabled\n"); 22.118 @@ -220,9 +181,9 @@ static enum handler_return long_mode_do_ 22.119 } 22.120 } 22.121 22.122 - v->arch.hvm_svm.cpu_shadow_efer = msr_content; 22.123 + v->arch.hvm_vcpu.guest_efer = msr_content; 22.124 vmcb->efer = msr_content | EFER_SVME; 22.125 - if ( !svm_paging_enabled(v) ) 22.126 + if ( !hvm_paging_enabled(v) ) 22.127 vmcb->efer &= ~(EFER_LME | EFER_LMA); 22.128 22.129 break; 22.130 @@ -297,10 +258,10 @@ int svm_vmcb_save(struct vcpu *v, struct 22.131 c->rsp = vmcb->rsp; 22.132 c->rflags = vmcb->rflags; 22.133 22.134 - c->cr0 = v->arch.hvm_svm.cpu_shadow_cr0; 22.135 - c->cr2 = v->arch.hvm_svm.cpu_cr2; 22.136 - c->cr3 = v->arch.hvm_svm.cpu_cr3; 22.137 - c->cr4 = v->arch.hvm_svm.cpu_shadow_cr4; 22.138 + c->cr0 = v->arch.hvm_vcpu.guest_cr[0]; 22.139 + c->cr2 = v->arch.hvm_vcpu.guest_cr[2]; 22.140 + c->cr3 = v->arch.hvm_vcpu.guest_cr[3]; 22.141 + c->cr4 = v->arch.hvm_vcpu.guest_cr[4]; 22.142 22.143 #ifdef HVM_DEBUG_SUSPEND 22.144 printk("%s: cr3=0x%"PRIx64", cr0=0x%"PRIx64", cr4=0x%"PRIx64".\n", 22.145 @@ -383,58 +344,31 @@ int svm_vmcb_restore(struct vcpu *v, str 22.146 vmcb->rsp = c->rsp; 22.147 vmcb->rflags = c->rflags; 22.148 22.149 - v->arch.hvm_svm.cpu_shadow_cr0 = c->cr0; 22.150 - vmcb->cr0 = c->cr0 | X86_CR0_WP | X86_CR0_ET | X86_CR0_PG; 22.151 - 22.152 - v->arch.hvm_svm.cpu_cr2 = c->cr2; 22.153 + v->arch.hvm_vcpu.guest_cr[0] = c->cr0 | X86_CR0_ET; 22.154 + v->arch.hvm_vcpu.guest_cr[2] = c->cr2; 22.155 + v->arch.hvm_vcpu.guest_cr[3] = c->cr3; 22.156 + v->arch.hvm_vcpu.guest_cr[4] = c->cr4; 22.157 + svm_update_guest_cr(v, 0); 22.158 + svm_update_guest_cr(v, 2); 22.159 + svm_update_guest_cr(v, 4); 22.160 22.161 #ifdef HVM_DEBUG_SUSPEND 22.162 printk("%s: cr3=0x%"PRIx64", cr0=0x%"PRIx64", cr4=0x%"PRIx64".\n", 22.163 - __func__, 22.164 - c->cr3, 22.165 - c->cr0, 22.166 - c->cr4); 22.167 + __func__, c->cr3, c->cr0, c->cr4); 22.168 #endif 22.169 22.170 - if ( !svm_paging_enabled(v) ) 22.171 - { 22.172 - printk("%s: paging not enabled.\n", __func__); 22.173 - goto skip_cr3; 22.174 - } 22.175 - 22.176 - if ( c->cr3 == v->arch.hvm_svm.cpu_cr3 ) 22.177 + if ( hvm_paging_enabled(v) && !paging_mode_hap(v->domain) ) 22.178 { 22.179 - /* 22.180 - * This is simple TLB flush, implying the guest has 22.181 - * removed some translation or changed page attributes. 22.182 - * We simply invalidate the shadow. 22.183 - */ 22.184 - mfn = gmfn_to_mfn(v->domain, c->cr3 >> PAGE_SHIFT); 22.185 - if ( mfn != pagetable_get_pfn(v->arch.guest_table) ) 22.186 - goto bad_cr3; 22.187 - } 22.188 - else 22.189 - { 22.190 - /* 22.191 - * If different, make a shadow. Check if the PDBR is valid 22.192 - * first. 22.193 - */ 22.194 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 c->cr3 = %"PRIx64, c->cr3); 22.195 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 = %"PRIx64, c->cr3); 22.196 mfn = gmfn_to_mfn(v->domain, c->cr3 >> PAGE_SHIFT); 22.197 if( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 22.198 goto bad_cr3; 22.199 - 22.200 old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 22.201 v->arch.guest_table = pagetable_from_pfn(mfn); 22.202 - if (old_base_mfn) 22.203 + if ( old_base_mfn ) 22.204 put_page(mfn_to_page(old_base_mfn)); 22.205 - v->arch.hvm_svm.cpu_cr3 = c->cr3; 22.206 } 22.207 22.208 - skip_cr3: 22.209 - vmcb->cr4 = c->cr4 | HVM_CR4_HOST_MASK; 22.210 - v->arch.hvm_svm.cpu_shadow_cr4 = c->cr4; 22.211 - 22.212 vmcb->idtr.limit = c->idtr_limit; 22.213 vmcb->idtr.base = c->idtr_base; 22.214 22.215 @@ -488,10 +422,6 @@ int svm_vmcb_restore(struct vcpu *v, str 22.216 22.217 if ( paging_mode_hap(v->domain) ) 22.218 { 22.219 - vmcb->cr0 = v->arch.hvm_svm.cpu_shadow_cr0; 22.220 - vmcb->cr4 = (v->arch.hvm_svm.cpu_shadow_cr4 | 22.221 - (HVM_CR4_HOST_MASK & ~X86_CR4_PAE)); 22.222 - vmcb->cr3 = c->cr3; 22.223 vmcb->np_enable = 1; 22.224 vmcb->g_pat = 0x0007040600070406ULL; /* guest PAT */ 22.225 vmcb->h_cr3 = pagetable_get_paddr(v->domain->arch.phys_table); 22.226 @@ -521,7 +451,6 @@ int svm_vmcb_restore(struct vcpu *v, str 22.227 } 22.228 22.229 paging_update_paging_modes(v); 22.230 - svm_asid_g_update_paging(v); 22.231 22.232 return 0; 22.233 22.234 @@ -540,7 +469,7 @@ static void svm_save_cpu_state(struct vc 22.235 data->msr_star = vmcb->star; 22.236 data->msr_cstar = vmcb->cstar; 22.237 data->msr_syscall_mask = vmcb->sfmask; 22.238 - data->msr_efer = v->arch.hvm_svm.cpu_shadow_efer; 22.239 + data->msr_efer = v->arch.hvm_vcpu.guest_efer; 22.240 data->msr_flags = -1ULL; 22.241 22.242 data->tsc = hvm_get_guest_time(v); 22.243 @@ -556,7 +485,7 @@ static void svm_load_cpu_state(struct vc 22.244 vmcb->star = data->msr_star; 22.245 vmcb->cstar = data->msr_cstar; 22.246 vmcb->sfmask = data->msr_syscall_mask; 22.247 - v->arch.hvm_svm.cpu_shadow_efer = data->msr_efer; 22.248 + v->arch.hvm_vcpu.guest_efer = data->msr_efer; 22.249 vmcb->efer = data->msr_efer | EFER_SVME; 22.250 /* VMCB's EFER.LME isn't set unless we're actually in long mode 22.251 * (see long_mode_do_msr_write()) */ 22.252 @@ -605,11 +534,11 @@ static int svm_guest_x86_mode(struct vcp 22.253 { 22.254 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.255 22.256 - if ( unlikely(!(v->arch.hvm_svm.cpu_shadow_cr0 & X86_CR0_PE)) ) 22.257 + if ( unlikely(!(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_PE)) ) 22.258 return 0; 22.259 if ( unlikely(vmcb->rflags & X86_EFLAGS_VM) ) 22.260 return 1; 22.261 - if ( svm_long_mode_enabled(v) && likely(vmcb->cs.attr.fields.l) ) 22.262 + if ( hvm_long_mode_enabled(v) && likely(vmcb->cs.attr.fields.l) ) 22.263 return 8; 22.264 return (likely(vmcb->cs.attr.fields.db) ? 4 : 2); 22.265 } 22.266 @@ -619,9 +548,45 @@ static void svm_update_host_cr3(struct v 22.267 /* SVM doesn't have a HOST_CR3 equivalent to update. */ 22.268 } 22.269 22.270 -static void svm_update_guest_cr3(struct vcpu *v) 22.271 +static void svm_update_guest_cr(struct vcpu *v, unsigned int cr) 22.272 { 22.273 - v->arch.hvm_svm.vmcb->cr3 = v->arch.hvm_vcpu.hw_cr3; 22.274 + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.275 + 22.276 + switch ( cr ) 22.277 + { 22.278 + case 0: 22.279 + vmcb->cr0 = v->arch.hvm_vcpu.guest_cr[0]; 22.280 + if ( !paging_mode_hap(v->domain) ) 22.281 + vmcb->cr0 |= X86_CR0_PG | X86_CR0_WP; 22.282 + break; 22.283 + case 2: 22.284 + vmcb->cr2 = v->arch.hvm_vcpu.guest_cr[2]; 22.285 + break; 22.286 + case 3: 22.287 + vmcb->cr3 = v->arch.hvm_vcpu.hw_cr[3]; 22.288 + svm_asid_inv_asid(v); 22.289 + break; 22.290 + case 4: 22.291 + vmcb->cr4 = HVM_CR4_HOST_MASK; 22.292 + if ( paging_mode_hap(v->domain) ) 22.293 + vmcb->cr4 &= ~X86_CR4_PAE; 22.294 + vmcb->cr4 |= v->arch.hvm_vcpu.guest_cr[4]; 22.295 + break; 22.296 + default: 22.297 + BUG(); 22.298 + } 22.299 +} 22.300 + 22.301 +static void svm_update_guest_efer(struct vcpu *v) 22.302 +{ 22.303 +#ifdef __x86_64__ 22.304 + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.305 + 22.306 + if ( v->arch.hvm_vcpu.guest_efer & EFER_LMA ) 22.307 + vmcb->efer |= EFER_LME | EFER_LMA; 22.308 + else 22.309 + vmcb->efer &= ~(EFER_LME | EFER_LMA); 22.310 +#endif 22.311 } 22.312 22.313 static void svm_flush_guest_tlbs(void) 22.314 @@ -639,24 +604,6 @@ static void svm_update_vtpr(struct vcpu 22.315 vmcb->vintr.fields.tpr = value & 0x0f; 22.316 } 22.317 22.318 -static unsigned long svm_get_ctrl_reg(struct vcpu *v, unsigned int num) 22.319 -{ 22.320 - switch ( num ) 22.321 - { 22.322 - case 0: 22.323 - return v->arch.hvm_svm.cpu_shadow_cr0; 22.324 - case 2: 22.325 - return v->arch.hvm_svm.cpu_cr2; 22.326 - case 3: 22.327 - return v->arch.hvm_svm.cpu_cr3; 22.328 - case 4: 22.329 - return v->arch.hvm_svm.cpu_shadow_cr4; 22.330 - default: 22.331 - BUG(); 22.332 - } 22.333 - return 0; /* dummy */ 22.334 -} 22.335 - 22.336 static void svm_sync_vmcb(struct vcpu *v) 22.337 { 22.338 struct arch_svm_struct *arch_svm = &v->arch.hvm_svm; 22.339 @@ -674,7 +621,7 @@ static void svm_sync_vmcb(struct vcpu *v 22.340 static unsigned long svm_get_segment_base(struct vcpu *v, enum x86_segment seg) 22.341 { 22.342 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.343 - int long_mode = vmcb->cs.attr.fields.l && svm_long_mode_enabled(v); 22.344 + int long_mode = vmcb->cs.attr.fields.l && hvm_long_mode_enabled(v); 22.345 22.346 switch ( seg ) 22.347 { 22.348 @@ -748,7 +695,7 @@ static void svm_stts(struct vcpu *v) 22.349 * then this is not necessary: no FPU activity can occur until the guest 22.350 * clears CR0.TS, and we will initialise the FPU when that happens. 22.351 */ 22.352 - if ( !(v->arch.hvm_svm.cpu_shadow_cr0 & X86_CR0_TS) ) 22.353 + if ( !(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_TS) ) 22.354 { 22.355 v->arch.hvm_svm.vmcb->exception_intercepts |= 1U << TRAP_no_device; 22.356 vmcb->cr0 |= X86_CR0_TS; 22.357 @@ -949,7 +896,7 @@ static void svm_hvm_inject_exception( 22.358 { 22.359 struct vcpu *v = current; 22.360 if ( trapnr == TRAP_page_fault ) 22.361 - v->arch.hvm_svm.vmcb->cr2 = v->arch.hvm_svm.cpu_cr2 = cr2; 22.362 + v->arch.hvm_svm.vmcb->cr2 = v->arch.hvm_vcpu.guest_cr[2] = cr2; 22.363 svm_inject_exception(v, trapnr, (errcode != -1), errcode); 22.364 } 22.365 22.366 @@ -970,17 +917,13 @@ static struct hvm_function_table svm_fun 22.367 .load_cpu_guest_regs = svm_load_cpu_guest_regs, 22.368 .save_cpu_ctxt = svm_save_vmcb_ctxt, 22.369 .load_cpu_ctxt = svm_load_vmcb_ctxt, 22.370 - .paging_enabled = svm_paging_enabled, 22.371 - .long_mode_enabled = svm_long_mode_enabled, 22.372 - .pae_enabled = svm_pae_enabled, 22.373 - .nx_enabled = svm_nx_enabled, 22.374 .interrupts_enabled = svm_interrupts_enabled, 22.375 .guest_x86_mode = svm_guest_x86_mode, 22.376 - .get_guest_ctrl_reg = svm_get_ctrl_reg, 22.377 .get_segment_base = svm_get_segment_base, 22.378 .get_segment_register = svm_get_segment_register, 22.379 .update_host_cr3 = svm_update_host_cr3, 22.380 - .update_guest_cr3 = svm_update_guest_cr3, 22.381 + .update_guest_cr = svm_update_guest_cr, 22.382 + .update_guest_efer = svm_update_guest_efer, 22.383 .flush_guest_tlbs = svm_flush_guest_tlbs, 22.384 .update_vtpr = svm_update_vtpr, 22.385 .stts = svm_stts, 22.386 @@ -1075,7 +1018,7 @@ static void svm_do_no_device_fault(struc 22.387 setup_fpu(v); 22.388 vmcb->exception_intercepts &= ~(1U << TRAP_no_device); 22.389 22.390 - if ( !(v->arch.hvm_svm.cpu_shadow_cr0 & X86_CR0_TS) ) 22.391 + if ( !(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_TS) ) 22.392 vmcb->cr0 &= ~X86_CR0_TS; 22.393 } 22.394 22.395 @@ -1347,7 +1290,7 @@ static int svm_get_io_address( 22.396 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.397 22.398 /* If we're in long mode, don't check the segment presence & limit */ 22.399 - long_mode = vmcb->cs.attr.fields.l && svm_long_mode_enabled(v); 22.400 + long_mode = vmcb->cs.attr.fields.l && hvm_long_mode_enabled(v); 22.401 22.402 /* d field of cs.attr is 1 for 32-bit, 0 for 16 or 64 bit. 22.403 * l field combined with EFER_LMA says whether it's 16 or 64 bit. 22.404 @@ -1650,31 +1593,11 @@ static void svm_io_instruction(struct vc 22.405 static int svm_set_cr0(unsigned long value) 22.406 { 22.407 struct vcpu *v = current; 22.408 - unsigned long mfn, old_value = v->arch.hvm_svm.cpu_shadow_cr0; 22.409 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.410 - unsigned long old_base_mfn; 22.411 - 22.412 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR0 value = %lx", value); 22.413 + int rc = hvm_set_cr0(value); 22.414 22.415 - if ( (u32)value != value ) 22.416 - { 22.417 - HVM_DBG_LOG(DBG_LEVEL_1, 22.418 - "Guest attempts to set upper 32 bits in CR0: %lx", 22.419 - value); 22.420 - svm_inject_exception(v, TRAP_gp_fault, 1, 0); 22.421 + if ( rc == 0 ) 22.422 return 0; 22.423 - } 22.424 - 22.425 - value &= ~HVM_CR0_GUEST_RESERVED_BITS; 22.426 - 22.427 - /* ET is reserved and should be always be 1. */ 22.428 - value |= X86_CR0_ET; 22.429 - 22.430 - if ( (value & (X86_CR0_PE|X86_CR0_PG)) == X86_CR0_PG ) 22.431 - { 22.432 - svm_inject_exception(v, TRAP_gp_fault, 1, 0); 22.433 - return 0; 22.434 - } 22.435 22.436 /* TS cleared? Then initialise FPU now. */ 22.437 if ( !(value & X86_CR0_TS) ) 22.438 @@ -1683,76 +1606,9 @@ static int svm_set_cr0(unsigned long val 22.439 vmcb->exception_intercepts &= ~(1U << TRAP_no_device); 22.440 } 22.441 22.442 - if ( (value & X86_CR0_PG) && !(old_value & X86_CR0_PG) ) 22.443 - { 22.444 - if ( svm_lme_is_set(v) ) 22.445 - { 22.446 - if ( !svm_cr4_pae_is_set(v) ) 22.447 - { 22.448 - HVM_DBG_LOG(DBG_LEVEL_1, "Enable paging before PAE enable"); 22.449 - svm_inject_exception(v, TRAP_gp_fault, 1, 0); 22.450 - return 0; 22.451 - } 22.452 - HVM_DBG_LOG(DBG_LEVEL_1, "Enable the Long mode"); 22.453 - v->arch.hvm_svm.cpu_shadow_efer |= EFER_LMA; 22.454 - vmcb->efer |= EFER_LMA | EFER_LME; 22.455 - } 22.456 - 22.457 - if ( !paging_mode_hap(v->domain) ) 22.458 - { 22.459 - /* The guest CR3 must be pointing to the guest physical. */ 22.460 - mfn = get_mfn_from_gpfn(v->arch.hvm_svm.cpu_cr3 >> PAGE_SHIFT); 22.461 - if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain)) 22.462 - { 22.463 - gdprintk(XENLOG_ERR, "Invalid CR3 value = %lx (mfn=%lx)\n", 22.464 - v->arch.hvm_svm.cpu_cr3, mfn); 22.465 - domain_crash(v->domain); 22.466 - return 0; 22.467 - } 22.468 - 22.469 - /* Now arch.guest_table points to machine physical. */ 22.470 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 22.471 - v->arch.guest_table = pagetable_from_pfn(mfn); 22.472 - if ( old_base_mfn ) 22.473 - put_page(mfn_to_page(old_base_mfn)); 22.474 - 22.475 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx, mfn = %lx", 22.476 - v->arch.hvm_vmx.cpu_cr3, mfn); 22.477 - } 22.478 - } 22.479 - else if ( !(value & X86_CR0_PG) && (old_value & X86_CR0_PG) ) 22.480 - { 22.481 - /* When CR0.PG is cleared, LMA is cleared immediately. */ 22.482 - if ( svm_long_mode_enabled(v) ) 22.483 - { 22.484 - vmcb->efer &= ~(EFER_LME | EFER_LMA); 22.485 - v->arch.hvm_svm.cpu_shadow_efer &= ~EFER_LMA; 22.486 - } 22.487 - 22.488 - if ( !paging_mode_hap(v->domain) && v->arch.hvm_svm.cpu_cr3 ) 22.489 - { 22.490 - put_page(mfn_to_page(get_mfn_from_gpfn( 22.491 - v->arch.hvm_svm.cpu_cr3 >> PAGE_SHIFT))); 22.492 - v->arch.guest_table = pagetable_null(); 22.493 - } 22.494 - } 22.495 - 22.496 - vmcb->cr0 = v->arch.hvm_svm.cpu_shadow_cr0 = value; 22.497 - if ( !paging_mode_hap(v->domain) ) 22.498 - vmcb->cr0 |= X86_CR0_PG | X86_CR0_WP; 22.499 - 22.500 - if ( (value ^ old_value) & X86_CR0_PG ) 22.501 - { 22.502 - paging_update_paging_modes(v); 22.503 - svm_asid_g_update_paging(v); 22.504 - } 22.505 - 22.506 return 1; 22.507 } 22.508 22.509 -/* 22.510 - * Read from control registers. CR0 and CR4 are read from the shadow. 22.511 - */ 22.512 static void mov_from_cr(int cr, int gp, struct cpu_user_regs *regs) 22.513 { 22.514 unsigned long value = 0; 22.515 @@ -1763,16 +1619,16 @@ static void mov_from_cr(int cr, int gp, 22.516 switch ( cr ) 22.517 { 22.518 case 0: 22.519 - value = v->arch.hvm_svm.cpu_shadow_cr0; 22.520 + value = v->arch.hvm_vcpu.guest_cr[0]; 22.521 break; 22.522 case 2: 22.523 value = vmcb->cr2; 22.524 break; 22.525 case 3: 22.526 - value = (unsigned long)v->arch.hvm_svm.cpu_cr3; 22.527 + value = (unsigned long)v->arch.hvm_vcpu.guest_cr[3]; 22.528 break; 22.529 case 4: 22.530 - value = (unsigned long)v->arch.hvm_svm.cpu_shadow_cr4; 22.531 + value = (unsigned long)v->arch.hvm_vcpu.guest_cr[4]; 22.532 break; 22.533 case 8: 22.534 value = (unsigned long)vlapic_get_reg(vlapic, APIC_TASKPRI); 22.535 @@ -1791,13 +1647,9 @@ static void mov_from_cr(int cr, int gp, 22.536 HVM_DBG_LOG(DBG_LEVEL_VMMU, "mov_from_cr: CR%d, value = %lx", cr, value); 22.537 } 22.538 22.539 - 22.540 -/* 22.541 - * Write to control registers 22.542 - */ 22.543 static int mov_to_cr(int gpreg, int cr, struct cpu_user_regs *regs) 22.544 { 22.545 - unsigned long value, old_cr, old_base_mfn, mfn; 22.546 + unsigned long value; 22.547 struct vcpu *v = current; 22.548 struct vlapic *vlapic = vcpu_vlapic(v); 22.549 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.550 @@ -1815,131 +1667,10 @@ static int mov_to_cr(int gpreg, int cr, 22.551 return svm_set_cr0(value); 22.552 22.553 case 3: 22.554 - if ( paging_mode_hap(v->domain) ) 22.555 - { 22.556 - vmcb->cr3 = v->arch.hvm_svm.cpu_cr3 = value; 22.557 - break; 22.558 - } 22.559 - 22.560 - /* If paging is not enabled yet, simply copy the value to CR3. */ 22.561 - if ( !svm_paging_enabled(v) ) 22.562 - { 22.563 - v->arch.hvm_svm.cpu_cr3 = value; 22.564 - break; 22.565 - } 22.566 - 22.567 - /* We make a new one if the shadow does not exist. */ 22.568 - if ( value == v->arch.hvm_svm.cpu_cr3 ) 22.569 - { 22.570 - /* 22.571 - * This is simple TLB flush, implying the guest has 22.572 - * removed some translation or changed page attributes. 22.573 - * We simply invalidate the shadow. 22.574 - */ 22.575 - mfn = get_mfn_from_gpfn(value >> PAGE_SHIFT); 22.576 - if ( mfn != pagetable_get_pfn(v->arch.guest_table) ) 22.577 - goto bad_cr3; 22.578 - paging_update_cr3(v); 22.579 - /* signal paging update to ASID handler */ 22.580 - svm_asid_g_mov_to_cr3 (v); 22.581 - } 22.582 - else 22.583 - { 22.584 - /* 22.585 - * If different, make a shadow. Check if the PDBR is valid 22.586 - * first. 22.587 - */ 22.588 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 value = %lx", value); 22.589 - mfn = get_mfn_from_gpfn(value >> PAGE_SHIFT); 22.590 - if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 22.591 - goto bad_cr3; 22.592 - 22.593 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 22.594 - v->arch.guest_table = pagetable_from_pfn(mfn); 22.595 - 22.596 - if ( old_base_mfn ) 22.597 - put_page(mfn_to_page(old_base_mfn)); 22.598 - 22.599 - v->arch.hvm_svm.cpu_cr3 = value; 22.600 - update_cr3(v); 22.601 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx", value); 22.602 - /* signal paging update to ASID handler */ 22.603 - svm_asid_g_mov_to_cr3 (v); 22.604 - } 22.605 - break; 22.606 - 22.607 - case 4: /* CR4 */ 22.608 - if ( value & HVM_CR4_GUEST_RESERVED_BITS ) 22.609 - { 22.610 - HVM_DBG_LOG(DBG_LEVEL_1, 22.611 - "Guest attempts to set reserved bit in CR4: %lx", 22.612 - value); 22.613 - svm_inject_exception(v, TRAP_gp_fault, 1, 0); 22.614 - break; 22.615 - } 22.616 + return hvm_set_cr3(value); 22.617 22.618 - if ( paging_mode_hap(v->domain) ) 22.619 - { 22.620 - v->arch.hvm_svm.cpu_shadow_cr4 = value; 22.621 - vmcb->cr4 = value | (HVM_CR4_HOST_MASK & ~X86_CR4_PAE); 22.622 - paging_update_paging_modes(v); 22.623 - /* signal paging update to ASID handler */ 22.624 - svm_asid_g_update_paging (v); 22.625 - break; 22.626 - } 22.627 - 22.628 - old_cr = v->arch.hvm_svm.cpu_shadow_cr4; 22.629 - if ( value & X86_CR4_PAE && !(old_cr & X86_CR4_PAE) ) 22.630 - { 22.631 - if ( svm_pgbit_test(v) ) 22.632 - { 22.633 -#if CONFIG_PAGING_LEVELS >= 3 22.634 - /* The guest is a 32-bit PAE guest. */ 22.635 - unsigned long mfn, old_base_mfn; 22.636 - mfn = get_mfn_from_gpfn(v->arch.hvm_svm.cpu_cr3 >> PAGE_SHIFT); 22.637 - if ( !mfn_valid(mfn) || 22.638 - !get_page(mfn_to_page(mfn), v->domain) ) 22.639 - goto bad_cr3; 22.640 - 22.641 - /* 22.642 - * Now arch.guest_table points to machine physical. 22.643 - */ 22.644 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 22.645 - v->arch.guest_table = pagetable_from_pfn(mfn); 22.646 - if ( old_base_mfn ) 22.647 - put_page(mfn_to_page(old_base_mfn)); 22.648 - paging_update_paging_modes(v); 22.649 - /* signal paging update to ASID handler */ 22.650 - svm_asid_g_update_paging (v); 22.651 - 22.652 - HVM_DBG_LOG(DBG_LEVEL_VMMU, 22.653 - "Update CR3 value = %lx, mfn = %lx", 22.654 - v->arch.hvm_svm.cpu_cr3, mfn); 22.655 -#endif 22.656 - } 22.657 - } 22.658 - else if ( !(value & X86_CR4_PAE) ) 22.659 - { 22.660 - if ( svm_long_mode_enabled(v) ) 22.661 - { 22.662 - svm_inject_exception(v, TRAP_gp_fault, 1, 0); 22.663 - } 22.664 - } 22.665 - 22.666 - v->arch.hvm_svm.cpu_shadow_cr4 = value; 22.667 - vmcb->cr4 = value | HVM_CR4_HOST_MASK; 22.668 - 22.669 - /* 22.670 - * Writing to CR4 to modify the PSE, PGE, or PAE flag invalidates 22.671 - * all TLB entries except global entries. 22.672 - */ 22.673 - if ((old_cr ^ value) & (X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE)) 22.674 - { 22.675 - paging_update_paging_modes(v); 22.676 - /* signal paging update to ASID handler */ 22.677 - svm_asid_g_update_paging (v); 22.678 - } 22.679 - break; 22.680 + case 4: 22.681 + return hvm_set_cr4(value); 22.682 22.683 case 8: 22.684 vlapic_set_reg(vlapic, APIC_TASKPRI, ((value & 0x0F) << 4)); 22.685 @@ -1953,19 +1684,11 @@ static int mov_to_cr(int gpreg, int cr, 22.686 } 22.687 22.688 return 1; 22.689 - 22.690 - bad_cr3: 22.691 - gdprintk(XENLOG_ERR, "Invalid CR3\n"); 22.692 - domain_crash(v->domain); 22.693 - return 0; 22.694 } 22.695 22.696 - 22.697 -#define ARR_SIZE(x) (sizeof(x) / sizeof(x[0])) 22.698 - 22.699 - 22.700 -static int svm_cr_access(struct vcpu *v, unsigned int cr, unsigned int type, 22.701 - struct cpu_user_regs *regs) 22.702 +static void svm_cr_access( 22.703 + struct vcpu *v, unsigned int cr, unsigned int type, 22.704 + struct cpu_user_regs *regs) 22.705 { 22.706 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.707 int inst_len = 0; 22.708 @@ -1990,12 +1713,12 @@ static int svm_cr_access(struct vcpu *v, 22.709 if ( type == TYPE_MOV_TO_CR ) 22.710 { 22.711 inst_len = __get_instruction_length_from_list( 22.712 - v, list_a, ARR_SIZE(list_a), &buffer[index], &match); 22.713 + v, list_a, ARRAY_SIZE(list_a), &buffer[index], &match); 22.714 } 22.715 else /* type == TYPE_MOV_FROM_CR */ 22.716 { 22.717 inst_len = __get_instruction_length_from_list( 22.718 - v, list_b, ARR_SIZE(list_b), &buffer[index], &match); 22.719 + v, list_b, ARRAY_SIZE(list_b), &buffer[index], &match); 22.720 } 22.721 22.722 ASSERT(inst_len > 0); 22.723 @@ -2008,7 +1731,8 @@ static int svm_cr_access(struct vcpu *v, 22.724 22.725 HVM_DBG_LOG(DBG_LEVEL_1, "eip = %lx", (unsigned long) vmcb->rip); 22.726 22.727 - switch (match) 22.728 + switch ( match ) 22.729 + 22.730 { 22.731 case INSTR_MOV2CR: 22.732 gpreg = decode_src_reg(prefix, buffer[index+2]); 22.733 @@ -2025,18 +1749,18 @@ static int svm_cr_access(struct vcpu *v, 22.734 setup_fpu(current); 22.735 vmcb->exception_intercepts &= ~(1U << TRAP_no_device); 22.736 vmcb->cr0 &= ~X86_CR0_TS; /* clear TS */ 22.737 - v->arch.hvm_svm.cpu_shadow_cr0 &= ~X86_CR0_TS; /* clear TS */ 22.738 + v->arch.hvm_vcpu.guest_cr[0] &= ~X86_CR0_TS; /* clear TS */ 22.739 break; 22.740 22.741 case INSTR_LMSW: 22.742 gpreg = decode_src_reg(prefix, buffer[index+2]); 22.743 value = get_reg(gpreg, regs, vmcb) & 0xF; 22.744 - value = (v->arch.hvm_svm.cpu_shadow_cr0 & ~0xF) | value; 22.745 + value = (v->arch.hvm_vcpu.guest_cr[0] & ~0xF) | value; 22.746 result = svm_set_cr0(value); 22.747 break; 22.748 22.749 case INSTR_SMSW: 22.750 - value = v->arch.hvm_svm.cpu_shadow_cr0 & 0xFFFF; 22.751 + value = v->arch.hvm_vcpu.guest_cr[0] & 0xFFFF; 22.752 modrm = buffer[index+2]; 22.753 addr_size = svm_guest_x86_mode(v); 22.754 if ( addr_size < 2 ) 22.755 @@ -2099,9 +1823,8 @@ static int svm_cr_access(struct vcpu *v, 22.756 22.757 ASSERT(inst_len); 22.758 22.759 - __update_guest_eip(vmcb, inst_len); 22.760 - 22.761 - return result; 22.762 + if ( result ) 22.763 + __update_guest_eip(vmcb, inst_len); 22.764 } 22.765 22.766 static void svm_do_msr_access( 22.767 @@ -2129,7 +1852,7 @@ static void svm_do_msr_access( 22.768 break; 22.769 22.770 case MSR_EFER: 22.771 - msr_content = v->arch.hvm_svm.cpu_shadow_efer; 22.772 + msr_content = v->arch.hvm_vcpu.guest_efer; 22.773 break; 22.774 22.775 case MSR_K8_MC4_MISC: /* Threshold register */ 22.776 @@ -2319,8 +2042,7 @@ void svm_handle_invlpg(const short invlp 22.777 HVMTRACE_3D(INVLPG, v, (invlpga?1:0), g_vaddr, (invlpga?regs->ecx:0)); 22.778 22.779 paging_invlpg(v, g_vaddr); 22.780 - /* signal invplg to ASID handler */ 22.781 - svm_asid_g_invlpg (v, g_vaddr); 22.782 + svm_asid_g_invlpg(v, g_vaddr); 22.783 } 22.784 22.785 22.786 @@ -2335,29 +2057,23 @@ static int svm_reset_to_realmode(struct 22.787 { 22.788 struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb; 22.789 22.790 - /* clear the vmcb and user regs */ 22.791 memset(regs, 0, sizeof(struct cpu_user_regs)); 22.792 - 22.793 - /* VMCB State */ 22.794 - vmcb->cr0 = X86_CR0_ET | X86_CR0_PG | X86_CR0_WP; 22.795 - v->arch.hvm_svm.cpu_shadow_cr0 = X86_CR0_ET; 22.796 + 22.797 + v->arch.hvm_vcpu.guest_cr[0] = X86_CR0_ET; 22.798 + svm_update_guest_cr(v, 0); 22.799 22.800 - vmcb->cr2 = 0; 22.801 + v->arch.hvm_vcpu.guest_cr[2] = 0; 22.802 + svm_update_guest_cr(v, 2); 22.803 + 22.804 + v->arch.hvm_vcpu.guest_cr[4] = 0; 22.805 + svm_update_guest_cr(v, 4); 22.806 + 22.807 vmcb->efer = EFER_SVME; 22.808 22.809 - vmcb->cr4 = HVM_CR4_HOST_MASK; 22.810 - v->arch.hvm_svm.cpu_shadow_cr4 = 0; 22.811 - 22.812 - if ( paging_mode_hap(v->domain) ) { 22.813 - vmcb->cr0 = v->arch.hvm_svm.cpu_shadow_cr0; 22.814 - vmcb->cr4 = v->arch.hvm_svm.cpu_shadow_cr4 | 22.815 - (HVM_CR4_HOST_MASK & ~X86_CR4_PAE); 22.816 - } 22.817 - 22.818 /* This will jump to ROMBIOS */ 22.819 vmcb->rip = 0xFFF0; 22.820 22.821 - /* setup the segment registers and all their hidden states */ 22.822 + /* Set up the segment registers and all their hidden states. */ 22.823 vmcb->cs.sel = 0xF000; 22.824 vmcb->cs.attr.bytes = 0x089b; 22.825 vmcb->cs.limit = 0xffff; 22.826 @@ -2483,7 +2199,7 @@ asmlinkage void svm_vmexit_handler(struc 22.827 unsigned long va; 22.828 va = vmcb->exitinfo2; 22.829 regs->error_code = vmcb->exitinfo1; 22.830 - HVM_DBG_LOG(DBG_LEVEL_VMMU, 22.831 + HVM_DBG_LOG(DBG_LEVEL_VMMU, 22.832 "eax=%lx, ebx=%lx, ecx=%lx, edx=%lx, esi=%lx, edi=%lx", 22.833 (unsigned long)regs->eax, (unsigned long)regs->ebx, 22.834 (unsigned long)regs->ecx, (unsigned long)regs->edx, 22.835 @@ -2495,7 +2211,7 @@ asmlinkage void svm_vmexit_handler(struc 22.836 break; 22.837 } 22.838 22.839 - v->arch.hvm_svm.cpu_cr2 = vmcb->cr2 = va; 22.840 + v->arch.hvm_vcpu.guest_cr[2] = vmcb->cr2 = va; 22.841 svm_inject_exception(v, TRAP_page_fault, 1, regs->error_code); 22.842 break; 22.843 }
23.1 --- a/xen/arch/x86/hvm/svm/vmcb.c Thu Aug 16 10:03:26 2007 -0600 23.2 +++ b/xen/arch/x86/hvm/svm/vmcb.c Thu Aug 16 10:47:33 2007 -0600 23.3 @@ -111,7 +111,7 @@ static int construct_vmcb(struct vcpu *v 23.4 svm_segment_attributes_t attrib; 23.5 23.6 /* TLB control, and ASID assigment. */ 23.7 - svm_asid_init_vcpu (v); 23.8 + svm_asid_init_vcpu(v); 23.9 23.10 vmcb->general1_intercepts = 23.11 GENERAL1_INTERCEPT_INTR | GENERAL1_INTERCEPT_NMI | 23.12 @@ -216,27 +216,19 @@ static int construct_vmcb(struct vcpu *v 23.13 vmcb->tr.base = 0; 23.14 vmcb->tr.limit = 0xff; 23.15 23.16 - /* Guest CR0. */ 23.17 - vmcb->cr0 = read_cr0(); 23.18 - arch_svm->cpu_shadow_cr0 = vmcb->cr0 & ~(X86_CR0_PG | X86_CR0_TS); 23.19 - vmcb->cr0 |= X86_CR0_WP; 23.20 + v->arch.hvm_vcpu.guest_cr[0] = X86_CR0_PE | X86_CR0_TS; 23.21 + hvm_update_guest_cr(v, 0); 23.22 23.23 - /* Guest CR4. */ 23.24 - arch_svm->cpu_shadow_cr4 = 23.25 - read_cr4() & ~(X86_CR4_PGE | X86_CR4_PSE | X86_CR4_PAE); 23.26 - vmcb->cr4 = arch_svm->cpu_shadow_cr4 | HVM_CR4_HOST_MASK; 23.27 + v->arch.hvm_vcpu.guest_cr[4] = 0; 23.28 + hvm_update_guest_cr(v, 4); 23.29 23.30 paging_update_paging_modes(v); 23.31 - vmcb->cr3 = v->arch.hvm_vcpu.hw_cr3; 23.32 23.33 if ( paging_mode_hap(v->domain) ) 23.34 { 23.35 - vmcb->cr0 = arch_svm->cpu_shadow_cr0; 23.36 vmcb->np_enable = 1; /* enable nested paging */ 23.37 vmcb->g_pat = 0x0007040600070406ULL; /* guest PAT */ 23.38 vmcb->h_cr3 = pagetable_get_paddr(v->domain->arch.phys_table); 23.39 - vmcb->cr4 = arch_svm->cpu_shadow_cr4 = 23.40 - (HVM_CR4_HOST_MASK & ~X86_CR4_PAE); 23.41 vmcb->exception_intercepts = HVM_TRAP_MASK; 23.42 23.43 /* No point in intercepting CR3/4 reads, because the hardware
24.1 --- a/xen/arch/x86/hvm/vioapic.c Thu Aug 16 10:03:26 2007 -0600 24.2 +++ b/xen/arch/x86/hvm/vioapic.c Thu Aug 16 10:47:33 2007 -0600 24.3 @@ -43,10 +43,6 @@ 24.4 /* HACK: Route IRQ0 only to VCPU0 to prevent time jumps. */ 24.5 #define IRQ0_SPECIAL_ROUTING 1 24.6 24.7 -#if defined(__ia64__) 24.8 -#define opt_hvm_debug_level opt_vmx_debug_level 24.9 -#endif 24.10 - 24.11 static void vioapic_deliver(struct hvm_hw_vioapic *vioapic, int irq); 24.12 24.13 static unsigned long vioapic_read_indirect(struct hvm_hw_vioapic *vioapic,
25.1 --- a/xen/arch/x86/hvm/vmx/vmcs.c Thu Aug 16 10:03:26 2007 -0600 25.2 +++ b/xen/arch/x86/hvm/vmx/vmcs.c Thu Aug 16 10:47:33 2007 -0600 25.3 @@ -315,34 +315,69 @@ void vmx_cpu_down(void) 25.4 local_irq_restore(flags); 25.5 } 25.6 25.7 +struct foreign_vmcs { 25.8 + struct vcpu *v; 25.9 + unsigned int count; 25.10 +}; 25.11 +static DEFINE_PER_CPU(struct foreign_vmcs, foreign_vmcs); 25.12 + 25.13 void vmx_vmcs_enter(struct vcpu *v) 25.14 { 25.15 + struct foreign_vmcs *fv; 25.16 + 25.17 /* 25.18 * NB. We must *always* run an HVM VCPU on its own VMCS, except for 25.19 * vmx_vmcs_enter/exit critical regions. 25.20 */ 25.21 - if ( v == current ) 25.22 + if ( likely(v == current) ) 25.23 return; 25.24 25.25 - vcpu_pause(v); 25.26 - spin_lock(&v->arch.hvm_vmx.vmcs_lock); 25.27 + fv = &this_cpu(foreign_vmcs); 25.28 + 25.29 + if ( fv->v == v ) 25.30 + { 25.31 + BUG_ON(fv->count == 0); 25.32 + } 25.33 + else 25.34 + { 25.35 + BUG_ON(fv->v != NULL); 25.36 + BUG_ON(fv->count != 0); 25.37 25.38 - vmx_clear_vmcs(v); 25.39 - vmx_load_vmcs(v); 25.40 + vcpu_pause(v); 25.41 + spin_lock(&v->arch.hvm_vmx.vmcs_lock); 25.42 + 25.43 + vmx_clear_vmcs(v); 25.44 + vmx_load_vmcs(v); 25.45 + 25.46 + fv->v = v; 25.47 + } 25.48 + 25.49 + fv->count++; 25.50 } 25.51 25.52 void vmx_vmcs_exit(struct vcpu *v) 25.53 { 25.54 - if ( v == current ) 25.55 + struct foreign_vmcs *fv; 25.56 + 25.57 + if ( likely(v == current) ) 25.58 return; 25.59 25.60 - /* Don't confuse vmx_do_resume (for @v or @current!) */ 25.61 - vmx_clear_vmcs(v); 25.62 - if ( is_hvm_vcpu(current) ) 25.63 - vmx_load_vmcs(current); 25.64 + fv = &this_cpu(foreign_vmcs); 25.65 + BUG_ON(fv->v != v); 25.66 + BUG_ON(fv->count == 0); 25.67 25.68 - spin_unlock(&v->arch.hvm_vmx.vmcs_lock); 25.69 - vcpu_unpause(v); 25.70 + if ( --fv->count == 0 ) 25.71 + { 25.72 + /* Don't confuse vmx_do_resume (for @v or @current!) */ 25.73 + vmx_clear_vmcs(v); 25.74 + if ( is_hvm_vcpu(current) ) 25.75 + vmx_load_vmcs(current); 25.76 + 25.77 + spin_unlock(&v->arch.hvm_vmx.vmcs_lock); 25.78 + vcpu_unpause(v); 25.79 + 25.80 + fv->v = NULL; 25.81 + } 25.82 } 25.83 25.84 struct xgt_desc { 25.85 @@ -380,7 +415,6 @@ static void vmx_set_host_env(struct vcpu 25.86 25.87 static void construct_vmcs(struct vcpu *v) 25.88 { 25.89 - unsigned long cr0, cr4; 25.90 union vmcs_arbytes arbytes; 25.91 25.92 vmx_vmcs_enter(v); 25.93 @@ -504,19 +538,11 @@ static void construct_vmcs(struct vcpu * 25.94 25.95 __vmwrite(EXCEPTION_BITMAP, HVM_TRAP_MASK | (1U << TRAP_page_fault)); 25.96 25.97 - /* Guest CR0. */ 25.98 - cr0 = read_cr0(); 25.99 - v->arch.hvm_vmx.cpu_cr0 = cr0; 25.100 - __vmwrite(GUEST_CR0, v->arch.hvm_vmx.cpu_cr0); 25.101 - v->arch.hvm_vmx.cpu_shadow_cr0 = cr0 & ~(X86_CR0_PG | X86_CR0_TS); 25.102 - __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr0); 25.103 + v->arch.hvm_vcpu.guest_cr[0] = X86_CR0_PE | X86_CR0_ET; 25.104 + hvm_update_guest_cr(v, 0); 25.105 25.106 - /* Guest CR4. */ 25.107 - cr4 = read_cr4(); 25.108 - __vmwrite(GUEST_CR4, cr4 & ~X86_CR4_PSE); 25.109 - v->arch.hvm_vmx.cpu_shadow_cr4 = 25.110 - cr4 & ~(X86_CR4_PGE | X86_CR4_VMXE | X86_CR4_PAE); 25.111 - __vmwrite(CR4_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr4); 25.112 + v->arch.hvm_vcpu.guest_cr[4] = 0; 25.113 + hvm_update_guest_cr(v, 4); 25.114 25.115 if ( cpu_has_vmx_tpr_shadow ) 25.116 {
26.1 --- a/xen/arch/x86/hvm/vmx/vmx.c Thu Aug 16 10:03:26 2007 -0600 26.2 +++ b/xen/arch/x86/hvm/vmx/vmx.c Thu Aug 16 10:47:33 2007 -0600 26.3 @@ -61,6 +61,8 @@ static void vmx_ctxt_switch_to(struct vc 26.4 static int vmx_alloc_vlapic_mapping(struct domain *d); 26.5 static void vmx_free_vlapic_mapping(struct domain *d); 26.6 static void vmx_install_vlapic_mapping(struct vcpu *v); 26.7 +static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr); 26.8 +static void vmx_update_guest_efer(struct vcpu *v); 26.9 26.10 static int vmx_domain_initialise(struct domain *d) 26.11 { 26.12 @@ -100,63 +102,8 @@ static void vmx_vcpu_destroy(struct vcpu 26.13 vmx_destroy_vmcs(v); 26.14 } 26.15 26.16 -static int vmx_paging_enabled(struct vcpu *v) 26.17 -{ 26.18 - unsigned long cr0 = v->arch.hvm_vmx.cpu_shadow_cr0; 26.19 - return (cr0 & (X86_CR0_PE | X86_CR0_PG)) == (X86_CR0_PE | X86_CR0_PG); 26.20 -} 26.21 - 26.22 -static int vmx_pgbit_test(struct vcpu *v) 26.23 -{ 26.24 - unsigned long cr0 = v->arch.hvm_vmx.cpu_shadow_cr0; 26.25 - return cr0 & X86_CR0_PG; 26.26 -} 26.27 - 26.28 -static int vmx_pae_enabled(struct vcpu *v) 26.29 -{ 26.30 - unsigned long cr4 = v->arch.hvm_vmx.cpu_shadow_cr4; 26.31 - return vmx_paging_enabled(v) && (cr4 & X86_CR4_PAE); 26.32 -} 26.33 - 26.34 -static int vmx_nx_enabled(struct vcpu *v) 26.35 -{ 26.36 - return v->arch.hvm_vmx.efer & EFER_NX; 26.37 -} 26.38 - 26.39 #ifdef __x86_64__ 26.40 26.41 -static int vmx_lme_is_set(struct vcpu *v) 26.42 -{ 26.43 - return v->arch.hvm_vmx.efer & EFER_LME; 26.44 -} 26.45 - 26.46 -static int vmx_long_mode_enabled(struct vcpu *v) 26.47 -{ 26.48 - return v->arch.hvm_vmx.efer & EFER_LMA; 26.49 -} 26.50 - 26.51 -static void vmx_enable_long_mode(struct vcpu *v) 26.52 -{ 26.53 - unsigned long vm_entry_value; 26.54 - 26.55 - vm_entry_value = __vmread(VM_ENTRY_CONTROLS); 26.56 - vm_entry_value |= VM_ENTRY_IA32E_MODE; 26.57 - __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value); 26.58 - 26.59 - v->arch.hvm_vmx.efer |= EFER_LMA; 26.60 -} 26.61 - 26.62 -static void vmx_disable_long_mode(struct vcpu *v) 26.63 -{ 26.64 - unsigned long vm_entry_value; 26.65 - 26.66 - vm_entry_value = __vmread(VM_ENTRY_CONTROLS); 26.67 - vm_entry_value &= ~VM_ENTRY_IA32E_MODE; 26.68 - __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value); 26.69 - 26.70 - v->arch.hvm_vmx.efer &= ~EFER_LMA; 26.71 -} 26.72 - 26.73 static DEFINE_PER_CPU(struct vmx_msr_state, host_msr_state); 26.74 26.75 static u32 msr_index[VMX_MSR_COUNT] = 26.76 @@ -190,7 +137,7 @@ static enum handler_return long_mode_do_ 26.77 switch ( ecx ) 26.78 { 26.79 case MSR_EFER: 26.80 - msr_content = v->arch.hvm_vmx.efer; 26.81 + msr_content = v->arch.hvm_vcpu.guest_efer; 26.82 break; 26.83 26.84 case MSR_FS_BASE: 26.85 @@ -204,7 +151,7 @@ static enum handler_return long_mode_do_ 26.86 case MSR_SHADOW_GS_BASE: 26.87 msr_content = v->arch.hvm_vmx.shadow_gs; 26.88 check_long_mode: 26.89 - if ( !(vmx_long_mode_enabled(v)) ) 26.90 + if ( !(hvm_long_mode_enabled(v)) ) 26.91 { 26.92 vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.93 return HNDL_exception_raised; 26.94 @@ -263,9 +210,9 @@ static enum handler_return long_mode_do_ 26.95 } 26.96 26.97 if ( (msr_content & EFER_LME) 26.98 - && !(v->arch.hvm_vmx.efer & EFER_LME) ) 26.99 + && !(v->arch.hvm_vcpu.guest_efer & EFER_LME) ) 26.100 { 26.101 - if ( unlikely(vmx_paging_enabled(v)) ) 26.102 + if ( unlikely(hvm_paging_enabled(v)) ) 26.103 { 26.104 gdprintk(XENLOG_WARNING, 26.105 "Trying to set EFER.LME with paging enabled\n"); 26.106 @@ -273,9 +220,9 @@ static enum handler_return long_mode_do_ 26.107 } 26.108 } 26.109 else if ( !(msr_content & EFER_LME) 26.110 - && (v->arch.hvm_vmx.efer & EFER_LME) ) 26.111 + && (v->arch.hvm_vcpu.guest_efer & EFER_LME) ) 26.112 { 26.113 - if ( unlikely(vmx_paging_enabled(v)) ) 26.114 + if ( unlikely(hvm_paging_enabled(v)) ) 26.115 { 26.116 gdprintk(XENLOG_WARNING, 26.117 "Trying to clear EFER.LME with paging enabled\n"); 26.118 @@ -283,17 +230,17 @@ static enum handler_return long_mode_do_ 26.119 } 26.120 } 26.121 26.122 - if ( (msr_content ^ v->arch.hvm_vmx.efer) & (EFER_NX|EFER_SCE) ) 26.123 + if ( (msr_content ^ v->arch.hvm_vcpu.guest_efer) & (EFER_NX|EFER_SCE) ) 26.124 write_efer((read_efer() & ~(EFER_NX|EFER_SCE)) | 26.125 (msr_content & (EFER_NX|EFER_SCE))); 26.126 26.127 - v->arch.hvm_vmx.efer = msr_content; 26.128 + v->arch.hvm_vcpu.guest_efer = msr_content; 26.129 break; 26.130 26.131 case MSR_FS_BASE: 26.132 case MSR_GS_BASE: 26.133 case MSR_SHADOW_GS_BASE: 26.134 - if ( !vmx_long_mode_enabled(v) ) 26.135 + if ( !hvm_long_mode_enabled(v) ) 26.136 goto gp_fault; 26.137 26.138 if ( !is_canonical_address(msr_content) ) 26.139 @@ -394,27 +341,18 @@ static void vmx_restore_guest_msrs(struc 26.140 clear_bit(i, &guest_flags); 26.141 } 26.142 26.143 - if ( (v->arch.hvm_vmx.efer ^ read_efer()) & (EFER_NX | EFER_SCE) ) 26.144 + if ( (v->arch.hvm_vcpu.guest_efer ^ read_efer()) & (EFER_NX | EFER_SCE) ) 26.145 { 26.146 HVM_DBG_LOG(DBG_LEVEL_2, 26.147 "restore guest's EFER with value %lx", 26.148 - v->arch.hvm_vmx.efer); 26.149 + v->arch.hvm_vcpu.guest_efer); 26.150 write_efer((read_efer() & ~(EFER_NX | EFER_SCE)) | 26.151 - (v->arch.hvm_vmx.efer & (EFER_NX | EFER_SCE))); 26.152 + (v->arch.hvm_vcpu.guest_efer & (EFER_NX | EFER_SCE))); 26.153 } 26.154 } 26.155 26.156 #else /* __i386__ */ 26.157 26.158 -static int vmx_lme_is_set(struct vcpu *v) 26.159 -{ return 0; } 26.160 -static int vmx_long_mode_enabled(struct vcpu *v) 26.161 -{ return 0; } 26.162 -static void vmx_enable_long_mode(struct vcpu *v) 26.163 -{ BUG(); } 26.164 -static void vmx_disable_long_mode(struct vcpu *v) 26.165 -{ BUG(); } 26.166 - 26.167 #define vmx_save_host_msrs() ((void)0) 26.168 26.169 static void vmx_restore_host_msrs(void) 26.170 @@ -427,13 +365,13 @@ static void vmx_restore_host_msrs(void) 26.171 26.172 static void vmx_restore_guest_msrs(struct vcpu *v) 26.173 { 26.174 - if ( (v->arch.hvm_vmx.efer ^ read_efer()) & EFER_NX ) 26.175 + if ( (v->arch.hvm_vcpu.guest_efer ^ read_efer()) & EFER_NX ) 26.176 { 26.177 HVM_DBG_LOG(DBG_LEVEL_2, 26.178 "restore guest's EFER with value %lx", 26.179 - v->arch.hvm_vmx.efer); 26.180 + v->arch.hvm_vcpu.guest_efer); 26.181 write_efer((read_efer() & ~EFER_NX) | 26.182 - (v->arch.hvm_vmx.efer & EFER_NX)); 26.183 + (v->arch.hvm_vcpu.guest_efer & EFER_NX)); 26.184 } 26.185 } 26.186 26.187 @@ -444,7 +382,7 @@ static enum handler_return long_mode_do_ 26.188 26.189 switch ( regs->ecx ) { 26.190 case MSR_EFER: 26.191 - msr_content = v->arch.hvm_vmx.efer; 26.192 + msr_content = v->arch.hvm_vcpu.guest_efer; 26.193 break; 26.194 26.195 default: 26.196 @@ -475,10 +413,10 @@ static enum handler_return long_mode_do_ 26.197 return HNDL_exception_raised; 26.198 } 26.199 26.200 - if ( (msr_content ^ v->arch.hvm_vmx.efer) & EFER_NX ) 26.201 + if ( (msr_content ^ v->arch.hvm_vcpu.guest_efer) & EFER_NX ) 26.202 write_efer((read_efer() & ~EFER_NX) | (msr_content & EFER_NX)); 26.203 26.204 - v->arch.hvm_vmx.efer = msr_content; 26.205 + v->arch.hvm_vcpu.guest_efer = msr_content; 26.206 break; 26.207 26.208 default: 26.209 @@ -501,12 +439,12 @@ static int vmx_guest_x86_mode(struct vcp 26.210 26.211 ASSERT(v == current); 26.212 26.213 - if ( unlikely(!(v->arch.hvm_vmx.cpu_shadow_cr0 & X86_CR0_PE)) ) 26.214 + if ( unlikely(!(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_PE)) ) 26.215 return 0; 26.216 if ( unlikely(__vmread(GUEST_RFLAGS) & X86_EFLAGS_VM) ) 26.217 return 1; 26.218 cs_ar_bytes = __vmread(GUEST_CS_AR_BYTES); 26.219 - if ( vmx_long_mode_enabled(v) && 26.220 + if ( hvm_long_mode_enabled(v) && 26.221 likely(cs_ar_bytes & X86_SEG_AR_CS_LM_ACTIVE) ) 26.222 return 8; 26.223 return (likely(cs_ar_bytes & X86_SEG_AR_DEF_OP_SIZE) ? 4 : 2); 26.224 @@ -551,12 +489,12 @@ void vmx_vmcs_save(struct vcpu *v, struc 26.225 c->rsp = __vmread(GUEST_RSP); 26.226 c->rflags = __vmread(GUEST_RFLAGS); 26.227 26.228 - c->cr0 = v->arch.hvm_vmx.cpu_shadow_cr0; 26.229 - c->cr2 = v->arch.hvm_vmx.cpu_cr2; 26.230 - c->cr3 = v->arch.hvm_vmx.cpu_cr3; 26.231 - c->cr4 = v->arch.hvm_vmx.cpu_shadow_cr4; 26.232 - 26.233 - c->msr_efer = v->arch.hvm_vmx.efer; 26.234 + c->cr0 = v->arch.hvm_vcpu.guest_cr[0]; 26.235 + c->cr2 = v->arch.hvm_vcpu.guest_cr[2]; 26.236 + c->cr3 = v->arch.hvm_vcpu.guest_cr[3]; 26.237 + c->cr4 = v->arch.hvm_vcpu.guest_cr[4]; 26.238 + 26.239 + c->msr_efer = v->arch.hvm_vcpu.guest_efer; 26.240 26.241 #ifdef HVM_DEBUG_SUSPEND 26.242 printk("%s: cr3=0x%"PRIx64", cr0=0x%"PRIx64", cr4=0x%"PRIx64".\n", 26.243 @@ -635,51 +573,33 @@ int vmx_vmcs_restore(struct vcpu *v, str 26.244 __vmwrite(GUEST_RSP, c->rsp); 26.245 __vmwrite(GUEST_RFLAGS, c->rflags); 26.246 26.247 - v->arch.hvm_vmx.cpu_cr0 = (c->cr0 | X86_CR0_PE | X86_CR0_PG | 26.248 - X86_CR0_NE | X86_CR0_WP | X86_CR0_ET); 26.249 - __vmwrite(GUEST_CR0, v->arch.hvm_vmx.cpu_cr0); 26.250 - v->arch.hvm_vmx.cpu_shadow_cr0 = c->cr0; 26.251 - __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr0); 26.252 - 26.253 - v->arch.hvm_vmx.cpu_cr2 = c->cr2; 26.254 - 26.255 - v->arch.hvm_vmx.efer = c->msr_efer; 26.256 + v->arch.hvm_vcpu.guest_cr[0] = c->cr0 | X86_CR0_ET; 26.257 + v->arch.hvm_vcpu.guest_cr[2] = c->cr2; 26.258 + v->arch.hvm_vcpu.guest_cr[3] = c->cr3; 26.259 + v->arch.hvm_vcpu.guest_cr[4] = c->cr4; 26.260 + vmx_update_guest_cr(v, 0); 26.261 + vmx_update_guest_cr(v, 2); 26.262 + vmx_update_guest_cr(v, 4); 26.263 26.264 #ifdef HVM_DEBUG_SUSPEND 26.265 printk("%s: cr3=0x%"PRIx64", cr0=0x%"PRIx64", cr4=0x%"PRIx64".\n", 26.266 __func__, c->cr3, c->cr0, c->cr4); 26.267 #endif 26.268 26.269 - if ( !vmx_paging_enabled(v) ) 26.270 + if ( hvm_paging_enabled(v) ) 26.271 { 26.272 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "%s: paging not enabled.", __func__); 26.273 - goto skip_cr3; 26.274 - } 26.275 - 26.276 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 = %"PRIx64, c->cr3); 26.277 - /* current!=vcpu as not called by arch_vmx_do_launch */ 26.278 - mfn = gmfn_to_mfn(v->domain, c->cr3 >> PAGE_SHIFT); 26.279 - if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 26.280 - { 26.281 - gdprintk(XENLOG_ERR, "Invalid CR3 value=0x%"PRIx64".\n", c->cr3); 26.282 - vmx_vmcs_exit(v); 26.283 - return -EINVAL; 26.284 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 = %"PRIx64, c->cr3); 26.285 + mfn = gmfn_to_mfn(v->domain, c->cr3 >> PAGE_SHIFT); 26.286 + if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 26.287 + goto bad_cr3; 26.288 + old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 26.289 + v->arch.guest_table = pagetable_from_pfn(mfn); 26.290 + if ( old_base_mfn ) 26.291 + put_page(mfn_to_page(old_base_mfn)); 26.292 } 26.293 26.294 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 26.295 - v->arch.guest_table = pagetable_from_pfn(mfn); 26.296 - if ( old_base_mfn ) 26.297 - put_page(mfn_to_page(old_base_mfn)); 26.298 - 26.299 - skip_cr3: 26.300 - v->arch.hvm_vmx.cpu_cr3 = c->cr3; 26.301 - 26.302 - if ( vmx_long_mode_enabled(v) ) 26.303 - vmx_enable_long_mode(v); 26.304 - 26.305 - __vmwrite(GUEST_CR4, (c->cr4 | HVM_CR4_HOST_MASK)); 26.306 - v->arch.hvm_vmx.cpu_shadow_cr4 = c->cr4; 26.307 - __vmwrite(CR4_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr4); 26.308 + v->arch.hvm_vcpu.guest_efer = c->msr_efer; 26.309 + vmx_update_guest_efer(v); 26.310 26.311 __vmwrite(GUEST_IDTR_LIMIT, c->idtr_limit); 26.312 __vmwrite(GUEST_IDTR_BASE, c->idtr_base); 26.313 @@ -760,6 +680,11 @@ int vmx_vmcs_restore(struct vcpu *v, str 26.314 } 26.315 26.316 return 0; 26.317 + 26.318 + bad_cr3: 26.319 + gdprintk(XENLOG_ERR, "Invalid CR3 value=0x%"PRIx64"\n", c->cr3); 26.320 + vmx_vmcs_exit(v); 26.321 + return -EINVAL; 26.322 } 26.323 26.324 #if defined(__x86_64__) && defined(HVM_DEBUG_SUSPEND) 26.325 @@ -884,10 +809,10 @@ static void vmx_store_cpu_guest_regs( 26.326 26.327 if ( crs != NULL ) 26.328 { 26.329 - crs[0] = v->arch.hvm_vmx.cpu_shadow_cr0; 26.330 - crs[2] = v->arch.hvm_vmx.cpu_cr2; 26.331 - crs[3] = v->arch.hvm_vmx.cpu_cr3; 26.332 - crs[4] = v->arch.hvm_vmx.cpu_shadow_cr4; 26.333 + crs[0] = v->arch.hvm_vcpu.guest_cr[0]; 26.334 + crs[2] = v->arch.hvm_vcpu.guest_cr[2]; 26.335 + crs[3] = v->arch.hvm_vcpu.guest_cr[3]; 26.336 + crs[4] = v->arch.hvm_vcpu.guest_cr[4]; 26.337 } 26.338 26.339 vmx_vmcs_exit(v); 26.340 @@ -928,24 +853,6 @@ static void vmx_load_cpu_guest_regs(stru 26.341 vmx_vmcs_exit(v); 26.342 } 26.343 26.344 -static unsigned long vmx_get_ctrl_reg(struct vcpu *v, unsigned int num) 26.345 -{ 26.346 - switch ( num ) 26.347 - { 26.348 - case 0: 26.349 - return v->arch.hvm_vmx.cpu_cr0; 26.350 - case 2: 26.351 - return v->arch.hvm_vmx.cpu_cr2; 26.352 - case 3: 26.353 - return v->arch.hvm_vmx.cpu_cr3; 26.354 - case 4: 26.355 - return v->arch.hvm_vmx.cpu_shadow_cr4; 26.356 - default: 26.357 - BUG(); 26.358 - } 26.359 - return 0; /* dummy */ 26.360 -} 26.361 - 26.362 static unsigned long vmx_get_segment_base(struct vcpu *v, enum x86_segment seg) 26.363 { 26.364 unsigned long base = 0; 26.365 @@ -953,7 +860,7 @@ static unsigned long vmx_get_segment_bas 26.366 26.367 ASSERT(v == current); 26.368 26.369 - if ( vmx_long_mode_enabled(v) && 26.370 + if ( hvm_long_mode_enabled(v) && 26.371 (__vmread(GUEST_CS_AR_BYTES) & X86_SEG_AR_CS_LM_ACTIVE) ) 26.372 long_mode = 1; 26.373 26.374 @@ -1045,6 +952,9 @@ static void vmx_get_segment_register(str 26.375 } 26.376 26.377 reg->attr.bytes = (attr & 0xff) | ((attr >> 4) & 0xf00); 26.378 + /* Unusable flag is folded into Present flag. */ 26.379 + if ( attr & (1u<<16) ) 26.380 + reg->attr.fields.p = 0; 26.381 } 26.382 26.383 /* Make sure that xen intercepts any FP accesses from current */ 26.384 @@ -1059,10 +969,10 @@ static void vmx_stts(struct vcpu *v) 26.385 * then this is not necessary: no FPU activity can occur until the guest 26.386 * clears CR0.TS, and we will initialise the FPU when that happens. 26.387 */ 26.388 - if ( !(v->arch.hvm_vmx.cpu_shadow_cr0 & X86_CR0_TS) ) 26.389 + if ( !(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_TS) ) 26.390 { 26.391 - v->arch.hvm_vmx.cpu_cr0 |= X86_CR0_TS; 26.392 - __vmwrite(GUEST_CR0, v->arch.hvm_vmx.cpu_cr0); 26.393 + v->arch.hvm_vcpu.hw_cr[0] |= X86_CR0_TS; 26.394 + __vmwrite(GUEST_CR0, v->arch.hvm_vcpu.hw_cr[0]); 26.395 __vm_set_bit(EXCEPTION_BITMAP, TRAP_no_device); 26.396 } 26.397 } 26.398 @@ -1135,14 +1045,60 @@ static void vmx_update_host_cr3(struct v 26.399 vmx_vmcs_exit(v); 26.400 } 26.401 26.402 -static void vmx_update_guest_cr3(struct vcpu *v) 26.403 +static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr) 26.404 { 26.405 ASSERT((v == current) || !vcpu_runnable(v)); 26.406 + 26.407 vmx_vmcs_enter(v); 26.408 - __vmwrite(GUEST_CR3, v->arch.hvm_vcpu.hw_cr3); 26.409 + 26.410 + switch ( cr ) 26.411 + { 26.412 + case 0: 26.413 + v->arch.hvm_vcpu.hw_cr[0] = 26.414 + v->arch.hvm_vcpu.guest_cr[0] | 26.415 + X86_CR0_PE | X86_CR0_NE | X86_CR0_PG | X86_CR0_WP; 26.416 + __vmwrite(GUEST_CR0, v->arch.hvm_vcpu.hw_cr[0]); 26.417 + __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vcpu.guest_cr[0]); 26.418 + break; 26.419 + case 2: 26.420 + /* CR2 is updated in exit stub. */ 26.421 + break; 26.422 + case 3: 26.423 + __vmwrite(GUEST_CR3, v->arch.hvm_vcpu.hw_cr[3]); 26.424 + break; 26.425 + case 4: 26.426 + v->arch.hvm_vcpu.hw_cr[4] = 26.427 + v->arch.hvm_vcpu.guest_cr[4] | HVM_CR4_HOST_MASK; 26.428 + __vmwrite(GUEST_CR4, v->arch.hvm_vcpu.hw_cr[4]); 26.429 + __vmwrite(CR4_READ_SHADOW, v->arch.hvm_vcpu.guest_cr[4]); 26.430 + break; 26.431 + default: 26.432 + BUG(); 26.433 + } 26.434 + 26.435 vmx_vmcs_exit(v); 26.436 } 26.437 26.438 +static void vmx_update_guest_efer(struct vcpu *v) 26.439 +{ 26.440 +#ifdef __x86_64__ 26.441 + unsigned long vm_entry_value; 26.442 + 26.443 + ASSERT((v == current) || !vcpu_runnable(v)); 26.444 + 26.445 + vmx_vmcs_enter(v); 26.446 + 26.447 + vm_entry_value = __vmread(VM_ENTRY_CONTROLS); 26.448 + if ( v->arch.hvm_vcpu.guest_efer & EFER_LMA ) 26.449 + vm_entry_value |= VM_ENTRY_IA32E_MODE; 26.450 + else 26.451 + vm_entry_value &= ~VM_ENTRY_IA32E_MODE; 26.452 + __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value); 26.453 + 26.454 + vmx_vmcs_exit(v); 26.455 +#endif 26.456 +} 26.457 + 26.458 static void vmx_flush_guest_tlbs(void) 26.459 { 26.460 /* No tagged TLB support on VMX yet. The fact that we're in Xen 26.461 @@ -1156,7 +1112,7 @@ static void vmx_inject_exception( 26.462 struct vcpu *v = current; 26.463 vmx_inject_hw_exception(v, trapnr, errcode); 26.464 if ( trapnr == TRAP_page_fault ) 26.465 - v->arch.hvm_vmx.cpu_cr2 = cr2; 26.466 + v->arch.hvm_vcpu.guest_cr[2] = cr2; 26.467 } 26.468 26.469 static void vmx_update_vtpr(struct vcpu *v, unsigned long value) 26.470 @@ -1200,17 +1156,13 @@ static struct hvm_function_table vmx_fun 26.471 .load_cpu_guest_regs = vmx_load_cpu_guest_regs, 26.472 .save_cpu_ctxt = vmx_save_vmcs_ctxt, 26.473 .load_cpu_ctxt = vmx_load_vmcs_ctxt, 26.474 - .paging_enabled = vmx_paging_enabled, 26.475 - .long_mode_enabled = vmx_long_mode_enabled, 26.476 - .pae_enabled = vmx_pae_enabled, 26.477 - .nx_enabled = vmx_nx_enabled, 26.478 .interrupts_enabled = vmx_interrupts_enabled, 26.479 .guest_x86_mode = vmx_guest_x86_mode, 26.480 - .get_guest_ctrl_reg = vmx_get_ctrl_reg, 26.481 .get_segment_base = vmx_get_segment_base, 26.482 .get_segment_register = vmx_get_segment_register, 26.483 .update_host_cr3 = vmx_update_host_cr3, 26.484 - .update_guest_cr3 = vmx_update_guest_cr3, 26.485 + .update_guest_cr = vmx_update_guest_cr, 26.486 + .update_guest_efer = vmx_update_guest_efer, 26.487 .flush_guest_tlbs = vmx_flush_guest_tlbs, 26.488 .update_vtpr = vmx_update_vtpr, 26.489 .stts = vmx_stts, 26.490 @@ -1315,10 +1267,10 @@ static void vmx_do_no_device_fault(void) 26.491 __vm_clear_bit(EXCEPTION_BITMAP, TRAP_no_device); 26.492 26.493 /* Disable TS in guest CR0 unless the guest wants the exception too. */ 26.494 - if ( !(v->arch.hvm_vmx.cpu_shadow_cr0 & X86_CR0_TS) ) 26.495 + if ( !(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_TS) ) 26.496 { 26.497 - v->arch.hvm_vmx.cpu_cr0 &= ~X86_CR0_TS; 26.498 - __vmwrite(GUEST_CR0, v->arch.hvm_vmx.cpu_cr0); 26.499 + v->arch.hvm_vcpu.hw_cr[0] &= ~X86_CR0_TS; 26.500 + __vmwrite(GUEST_CR0, v->arch.hvm_vcpu.hw_cr[0]); 26.501 } 26.502 } 26.503 26.504 @@ -1773,7 +1725,7 @@ static void vmx_do_str_pio(unsigned long 26.505 26.506 sign = regs->eflags & X86_EFLAGS_DF ? -1 : 1; 26.507 ar_bytes = __vmread(GUEST_CS_AR_BYTES); 26.508 - if ( vmx_long_mode_enabled(current) && 26.509 + if ( hvm_long_mode_enabled(current) && 26.510 (ar_bytes & X86_SEG_AR_CS_LM_ACTIVE) ) 26.511 long_mode = 1; 26.512 addr = __vmread(GUEST_LINEAR_ADDRESS); 26.513 @@ -1900,9 +1852,9 @@ static void vmx_world_save(struct vcpu * 26.514 c->esp = __vmread(GUEST_RSP); 26.515 c->eflags = __vmread(GUEST_RFLAGS) & ~X86_EFLAGS_RF; 26.516 26.517 - c->cr0 = v->arch.hvm_vmx.cpu_shadow_cr0; 26.518 - c->cr3 = v->arch.hvm_vmx.cpu_cr3; 26.519 - c->cr4 = v->arch.hvm_vmx.cpu_shadow_cr4; 26.520 + c->cr0 = v->arch.hvm_vcpu.guest_cr[0]; 26.521 + c->cr3 = v->arch.hvm_vcpu.guest_cr[3]; 26.522 + c->cr4 = v->arch.hvm_vcpu.guest_cr[4]; 26.523 26.524 c->idtr_limit = __vmread(GUEST_IDTR_LIMIT); 26.525 c->idtr_base = __vmread(GUEST_IDTR_BASE); 26.526 @@ -1959,30 +1911,15 @@ static int vmx_world_restore(struct vcpu 26.527 __vmwrite(GUEST_RSP, c->esp); 26.528 __vmwrite(GUEST_RFLAGS, c->eflags); 26.529 26.530 - v->arch.hvm_vmx.cpu_shadow_cr0 = c->cr0; 26.531 - __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr0); 26.532 - 26.533 - if ( !vmx_paging_enabled(v) ) 26.534 - goto skip_cr3; 26.535 - 26.536 - if ( c->cr3 == v->arch.hvm_vmx.cpu_cr3 ) 26.537 + v->arch.hvm_vcpu.guest_cr[0] = c->cr0; 26.538 + v->arch.hvm_vcpu.guest_cr[3] = c->cr3; 26.539 + v->arch.hvm_vcpu.guest_cr[4] = c->cr4; 26.540 + vmx_update_guest_cr(v, 0); 26.541 + vmx_update_guest_cr(v, 4); 26.542 + 26.543 + if ( hvm_paging_enabled(v) ) 26.544 { 26.545 - /* 26.546 - * This is simple TLB flush, implying the guest has 26.547 - * removed some translation or changed page attributes. 26.548 - * We simply invalidate the shadow. 26.549 - */ 26.550 - mfn = get_mfn_from_gpfn(c->cr3 >> PAGE_SHIFT); 26.551 - if ( mfn != pagetable_get_pfn(v->arch.guest_table) ) 26.552 - goto bad_cr3; 26.553 - } 26.554 - else 26.555 - { 26.556 - /* 26.557 - * If different, make a shadow. Check if the PDBR is valid 26.558 - * first. 26.559 - */ 26.560 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 c->cr3 = %x", c->cr3); 26.561 + HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 = %x", c->cr3); 26.562 mfn = get_mfn_from_gpfn(c->cr3 >> PAGE_SHIFT); 26.563 if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 26.564 goto bad_cr3; 26.565 @@ -1990,19 +1927,8 @@ static int vmx_world_restore(struct vcpu 26.566 v->arch.guest_table = pagetable_from_pfn(mfn); 26.567 if ( old_base_mfn ) 26.568 put_page(mfn_to_page(old_base_mfn)); 26.569 - v->arch.hvm_vmx.cpu_cr3 = c->cr3; 26.570 } 26.571 26.572 - skip_cr3: 26.573 - if ( !vmx_paging_enabled(v) ) 26.574 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "switching to vmxassist. use phys table"); 26.575 - else 26.576 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %x", c->cr3); 26.577 - 26.578 - __vmwrite(GUEST_CR4, (c->cr4 | HVM_CR4_HOST_MASK)); 26.579 - v->arch.hvm_vmx.cpu_shadow_cr4 = c->cr4; 26.580 - __vmwrite(CR4_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr4); 26.581 - 26.582 __vmwrite(GUEST_IDTR_LIMIT, c->idtr_limit); 26.583 __vmwrite(GUEST_IDTR_BASE, c->idtr_base); 26.584 26.585 @@ -2149,33 +2075,11 @@ static int vmx_assist(struct vcpu *v, in 26.586 static int vmx_set_cr0(unsigned long value) 26.587 { 26.588 struct vcpu *v = current; 26.589 - unsigned long mfn; 26.590 unsigned long eip; 26.591 - int paging_enabled; 26.592 - unsigned long old_cr0; 26.593 - unsigned long old_base_mfn; 26.594 - 26.595 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR0 value = %lx", value); 26.596 - 26.597 - if ( (u32)value != value ) 26.598 - { 26.599 - HVM_DBG_LOG(DBG_LEVEL_1, 26.600 - "Guest attempts to set upper 32 bits in CR0: %lx", 26.601 - value); 26.602 - vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.603 + int rc = hvm_set_cr0(value); 26.604 + 26.605 + if ( rc == 0 ) 26.606 return 0; 26.607 - } 26.608 - 26.609 - value &= ~HVM_CR0_GUEST_RESERVED_BITS; 26.610 - 26.611 - /* ET is reserved and should be always be 1. */ 26.612 - value |= X86_CR0_ET; 26.613 - 26.614 - if ( (value & (X86_CR0_PE | X86_CR0_PG)) == X86_CR0_PG ) 26.615 - { 26.616 - vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.617 - return 0; 26.618 - } 26.619 26.620 /* TS cleared? Then initialise FPU now. */ 26.621 if ( !(value & X86_CR0_TS) ) 26.622 @@ -2184,88 +2088,13 @@ static int vmx_set_cr0(unsigned long val 26.623 __vm_clear_bit(EXCEPTION_BITMAP, TRAP_no_device); 26.624 } 26.625 26.626 - old_cr0 = v->arch.hvm_vmx.cpu_shadow_cr0; 26.627 - paging_enabled = old_cr0 & X86_CR0_PG; 26.628 - 26.629 - v->arch.hvm_vmx.cpu_cr0 = (value | X86_CR0_PE | X86_CR0_PG 26.630 - | X86_CR0_NE | X86_CR0_WP); 26.631 - __vmwrite(GUEST_CR0, v->arch.hvm_vmx.cpu_cr0); 26.632 - 26.633 - v->arch.hvm_vmx.cpu_shadow_cr0 = value; 26.634 - __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr0); 26.635 - 26.636 - /* Trying to enable paging. */ 26.637 - if ( (value & X86_CR0_PE) && (value & X86_CR0_PG) && !paging_enabled ) 26.638 - { 26.639 - if ( vmx_lme_is_set(v) && !vmx_long_mode_enabled(v) ) 26.640 - { 26.641 - if ( !(v->arch.hvm_vmx.cpu_shadow_cr4 & X86_CR4_PAE) ) 26.642 - { 26.643 - HVM_DBG_LOG(DBG_LEVEL_1, "Guest enabled paging " 26.644 - "with EFER.LME set but not CR4.PAE"); 26.645 - vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.646 - return 0; 26.647 - } 26.648 - 26.649 - HVM_DBG_LOG(DBG_LEVEL_1, "Enabling long mode"); 26.650 - vmx_enable_long_mode(v); 26.651 - } 26.652 - 26.653 - /* 26.654 - * The guest CR3 must be pointing to the guest physical. 26.655 - */ 26.656 - mfn = get_mfn_from_gpfn(v->arch.hvm_vmx.cpu_cr3 >> PAGE_SHIFT); 26.657 - if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 26.658 - { 26.659 - gdprintk(XENLOG_ERR, "Invalid CR3 value = %lx (mfn=%lx)\n", 26.660 - v->arch.hvm_vmx.cpu_cr3, mfn); 26.661 - domain_crash(v->domain); 26.662 - return 0; 26.663 - } 26.664 - 26.665 - /* 26.666 - * Now arch.guest_table points to machine physical. 26.667 - */ 26.668 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 26.669 - v->arch.guest_table = pagetable_from_pfn(mfn); 26.670 - if ( old_base_mfn ) 26.671 - put_page(mfn_to_page(old_base_mfn)); 26.672 - 26.673 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx, mfn = %lx", 26.674 - v->arch.hvm_vmx.cpu_cr3, mfn); 26.675 - 26.676 - paging_update_paging_modes(v); 26.677 - } 26.678 - 26.679 - /* Trying to disable paging. */ 26.680 - if ( ((value & (X86_CR0_PE | X86_CR0_PG)) != (X86_CR0_PE | X86_CR0_PG)) && 26.681 - paging_enabled ) 26.682 - { 26.683 - /* When CR0.PG is cleared, LMA is cleared immediately. */ 26.684 - if ( vmx_long_mode_enabled(v) ) 26.685 - vmx_disable_long_mode(v); 26.686 - 26.687 - if ( v->arch.hvm_vmx.cpu_cr3 ) 26.688 - { 26.689 - put_page(mfn_to_page(get_mfn_from_gpfn( 26.690 - v->arch.hvm_vmx.cpu_cr3 >> PAGE_SHIFT))); 26.691 - v->arch.guest_table = pagetable_null(); 26.692 - } 26.693 - } 26.694 - 26.695 /* 26.696 * VMX does not implement real-mode virtualization. We emulate 26.697 * real-mode by performing a world switch to VMXAssist whenever 26.698 * a partition disables the CR0.PE bit. 26.699 */ 26.700 - if ( (value & X86_CR0_PE) == 0 ) 26.701 + if ( !(value & X86_CR0_PE) ) 26.702 { 26.703 - if ( value & X86_CR0_PG ) 26.704 - { 26.705 - vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.706 - return 0; 26.707 - } 26.708 - 26.709 if ( vmx_assist(v, VMX_ASSIST_INVOKE) ) 26.710 { 26.711 eip = __vmread(GUEST_RIP); 26.712 @@ -2286,8 +2115,6 @@ static int vmx_set_cr0(unsigned long val 26.713 return 0; /* do not update eip! */ 26.714 } 26.715 } 26.716 - else if ( (value & (X86_CR0_PE | X86_CR0_PG)) == X86_CR0_PE ) 26.717 - paging_update_paging_modes(v); 26.718 26.719 return 1; 26.720 } 26.721 @@ -2316,12 +2143,9 @@ static int vmx_set_cr0(unsigned long val 26.722 CASE_ ## T ## ET_REG(R15, r15) 26.723 #endif 26.724 26.725 -/* 26.726 - * Write to control registers 26.727 - */ 26.728 static int mov_to_cr(int gp, int cr, struct cpu_user_regs *regs) 26.729 { 26.730 - unsigned long value, old_cr, old_base_mfn, mfn; 26.731 + unsigned long value; 26.732 struct vcpu *v = current; 26.733 struct vlapic *vlapic = vcpu_vlapic(v); 26.734 26.735 @@ -2353,108 +2177,10 @@ static int mov_to_cr(int gp, int cr, str 26.736 return vmx_set_cr0(value); 26.737 26.738 case 3: 26.739 - /* 26.740 - * If paging is not enabled yet, simply copy the value to CR3. 26.741 - */ 26.742 - if ( !vmx_paging_enabled(v) ) 26.743 - { 26.744 - v->arch.hvm_vmx.cpu_cr3 = value; 26.745 - break; 26.746 - } 26.747 - 26.748 - /* 26.749 - * We make a new one if the shadow does not exist. 26.750 - */ 26.751 - if ( value == v->arch.hvm_vmx.cpu_cr3 ) { 26.752 - /* 26.753 - * This is simple TLB flush, implying the guest has 26.754 - * removed some translation or changed page attributes. 26.755 - * We simply invalidate the shadow. 26.756 - */ 26.757 - mfn = get_mfn_from_gpfn(value >> PAGE_SHIFT); 26.758 - if ( mfn != pagetable_get_pfn(v->arch.guest_table) ) 26.759 - goto bad_cr3; 26.760 - paging_update_cr3(v); 26.761 - } else { 26.762 - /* 26.763 - * If different, make a shadow. Check if the PDBR is valid 26.764 - * first. 26.765 - */ 26.766 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 value = %lx", value); 26.767 - mfn = get_mfn_from_gpfn(value >> PAGE_SHIFT); 26.768 - if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) ) 26.769 - goto bad_cr3; 26.770 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 26.771 - v->arch.guest_table = pagetable_from_pfn(mfn); 26.772 - if ( old_base_mfn ) 26.773 - put_page(mfn_to_page(old_base_mfn)); 26.774 - v->arch.hvm_vmx.cpu_cr3 = value; 26.775 - update_cr3(v); 26.776 - HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx", value); 26.777 - } 26.778 - break; 26.779 - 26.780 - case 4: /* CR4 */ 26.781 - old_cr = v->arch.hvm_vmx.cpu_shadow_cr4; 26.782 - 26.783 - if ( value & HVM_CR4_GUEST_RESERVED_BITS ) 26.784 - { 26.785 - HVM_DBG_LOG(DBG_LEVEL_1, 26.786 - "Guest attempts to set reserved bit in CR4: %lx", 26.787 - value); 26.788 - vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.789 - return 0; 26.790 - } 26.791 - 26.792 - if ( (value & X86_CR4_PAE) && !(old_cr & X86_CR4_PAE) ) 26.793 - { 26.794 - if ( vmx_pgbit_test(v) ) 26.795 - { 26.796 -#if CONFIG_PAGING_LEVELS >= 3 26.797 - /* The guest is a 32-bit PAE guest. */ 26.798 - unsigned long mfn, old_base_mfn; 26.799 - mfn = get_mfn_from_gpfn(v->arch.hvm_vmx.cpu_cr3 >> PAGE_SHIFT); 26.800 - if ( !mfn_valid(mfn) || 26.801 - !get_page(mfn_to_page(mfn), v->domain) ) 26.802 - goto bad_cr3; 26.803 - 26.804 - /* 26.805 - * Now arch.guest_table points to machine physical. 26.806 - */ 26.807 - old_base_mfn = pagetable_get_pfn(v->arch.guest_table); 26.808 - v->arch.guest_table = pagetable_from_pfn(mfn); 26.809 - if ( old_base_mfn ) 26.810 - put_page(mfn_to_page(old_base_mfn)); 26.811 - 26.812 - HVM_DBG_LOG(DBG_LEVEL_VMMU, 26.813 - "Update CR3 value = %lx, mfn = %lx", 26.814 - v->arch.hvm_vmx.cpu_cr3, mfn); 26.815 -#endif 26.816 - } 26.817 - } 26.818 - else if ( !(value & X86_CR4_PAE) ) 26.819 - { 26.820 - if ( unlikely(vmx_long_mode_enabled(v)) ) 26.821 - { 26.822 - HVM_DBG_LOG(DBG_LEVEL_1, "Guest cleared CR4.PAE while " 26.823 - "EFER.LMA is set"); 26.824 - vmx_inject_hw_exception(v, TRAP_gp_fault, 0); 26.825 - return 0; 26.826 - } 26.827 - } 26.828 - 26.829 - __vmwrite(GUEST_CR4, value | HVM_CR4_HOST_MASK); 26.830 - v->arch.hvm_vmx.cpu_shadow_cr4 = value; 26.831 - __vmwrite(CR4_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr4); 26.832 - 26.833 - /* 26.834 - * Writing to CR4 to modify the PSE, PGE, or PAE flag invalidates 26.835 - * all TLB entries except global entries. 26.836 - */ 26.837 - if ( (old_cr ^ value) & (X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE) ) 26.838 - paging_update_paging_modes(v); 26.839 - 26.840 - break; 26.841 + return hvm_set_cr3(value); 26.842 + 26.843 + case 4: 26.844 + return hvm_set_cr4(value); 26.845 26.846 case 8: 26.847 vlapic_set_reg(vlapic, APIC_TASKPRI, ((value & 0x0F) << 4)); 26.848 @@ -2462,14 +2188,11 @@ static int mov_to_cr(int gp, int cr, str 26.849 26.850 default: 26.851 gdprintk(XENLOG_ERR, "invalid cr: %d\n", cr); 26.852 - domain_crash(v->domain); 26.853 - return 0; 26.854 + goto exit_and_crash; 26.855 } 26.856 26.857 return 1; 26.858 26.859 - bad_cr3: 26.860 - gdprintk(XENLOG_ERR, "Invalid CR3\n"); 26.861 exit_and_crash: 26.862 domain_crash(v->domain); 26.863 return 0; 26.864 @@ -2487,7 +2210,7 @@ static void mov_from_cr(int cr, int gp, 26.865 switch ( cr ) 26.866 { 26.867 case 3: 26.868 - value = (unsigned long)v->arch.hvm_vmx.cpu_cr3; 26.869 + value = (unsigned long)v->arch.hvm_vcpu.guest_cr[3]; 26.870 break; 26.871 case 8: 26.872 value = (unsigned long)vlapic_get_reg(vlapic, APIC_TASKPRI); 26.873 @@ -2530,7 +2253,8 @@ static int vmx_cr_access(unsigned long e 26.874 unsigned long value; 26.875 struct vcpu *v = current; 26.876 26.877 - switch ( exit_qualification & CONTROL_REG_ACCESS_TYPE ) { 26.878 + switch ( exit_qualification & CONTROL_REG_ACCESS_TYPE ) 26.879 + { 26.880 case TYPE_MOV_TO_CR: 26.881 gp = exit_qualification & CONTROL_REG_ACCESS_REG; 26.882 cr = exit_qualification & CONTROL_REG_ACCESS_NUM; 26.883 @@ -2545,14 +2269,14 @@ static int vmx_cr_access(unsigned long e 26.884 setup_fpu(v); 26.885 __vm_clear_bit(EXCEPTION_BITMAP, TRAP_no_device); 26.886 26.887 - v->arch.hvm_vmx.cpu_cr0 &= ~X86_CR0_TS; /* clear TS */ 26.888 - __vmwrite(GUEST_CR0, v->arch.hvm_vmx.cpu_cr0); 26.889 - 26.890 - v->arch.hvm_vmx.cpu_shadow_cr0 &= ~X86_CR0_TS; /* clear TS */ 26.891 - __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vmx.cpu_shadow_cr0); 26.892 + v->arch.hvm_vcpu.hw_cr[0] &= ~X86_CR0_TS; /* clear TS */ 26.893 + __vmwrite(GUEST_CR0, v->arch.hvm_vcpu.hw_cr[0]); 26.894 + 26.895 + v->arch.hvm_vcpu.guest_cr[0] &= ~X86_CR0_TS; /* clear TS */ 26.896 + __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vcpu.guest_cr[0]); 26.897 break; 26.898 case TYPE_LMSW: 26.899 - value = v->arch.hvm_vmx.cpu_shadow_cr0; 26.900 + value = v->arch.hvm_vcpu.guest_cr[0]; 26.901 value = (value & ~0xF) | 26.902 (((exit_qualification & LMSW_SOURCE_DATA) >> 16) & 0xF); 26.903 return vmx_set_cr0(value); 26.904 @@ -2943,7 +2667,7 @@ asmlinkage void vmx_vmexit_handler(struc 26.905 break; 26.906 } 26.907 26.908 - v->arch.hvm_vmx.cpu_cr2 = exit_qualification; 26.909 + v->arch.hvm_vcpu.guest_cr[2] = exit_qualification; 26.910 vmx_inject_hw_exception(v, TRAP_page_fault, regs->error_code); 26.911 break; 26.912 case TRAP_nmi:
27.1 --- a/xen/arch/x86/hvm/vmx/x86_32/exits.S Thu Aug 16 10:03:26 2007 -0600 27.2 +++ b/xen/arch/x86/hvm/vmx/x86_32/exits.S Thu Aug 16 10:47:33 2007 -0600 27.3 @@ -74,7 +74,7 @@ ENTRY(vmx_asm_do_vmentry) 27.4 jnz vmx_process_softirqs 27.5 27.6 call vmx_intr_assist 27.7 - movl VCPU_vmx_cr2(%ebx),%eax 27.8 + movl VCPU_hvm_guest_cr2(%ebx),%eax 27.9 movl %eax,%cr2 27.10 call vmx_trace_vmentry 27.11
28.1 --- a/xen/arch/x86/hvm/vmx/x86_64/exits.S Thu Aug 16 10:03:26 2007 -0600 28.2 +++ b/xen/arch/x86/hvm/vmx/x86_64/exits.S Thu Aug 16 10:47:33 2007 -0600 28.3 @@ -88,7 +88,7 @@ ENTRY(vmx_asm_do_vmentry) 28.4 jnz vmx_process_softirqs 28.5 28.6 call vmx_intr_assist 28.7 - movq VCPU_vmx_cr2(%rbx),%rax 28.8 + movq VCPU_hvm_guest_cr2(%rbx),%rax 28.9 movq %rax,%cr2 28.10 call vmx_trace_vmentry 28.11
29.1 --- a/xen/arch/x86/mm.c Thu Aug 16 10:03:26 2007 -0600 29.2 +++ b/xen/arch/x86/mm.c Thu Aug 16 10:47:33 2007 -0600 29.3 @@ -394,8 +394,8 @@ void write_ptbase(struct vcpu *v) 29.4 write_cr3(v->arch.cr3); 29.5 } 29.6 29.7 -/* Should be called after CR3 is updated. 29.8 - * Updates vcpu->arch.cr3 and, for HVM guests, vcpu->arch.hvm_vcpu.cpu_cr3. 29.9 +/* 29.10 + * Should be called after CR3 is updated. 29.11 * 29.12 * Uses values found in vcpu->arch.(guest_table and guest_table_user), and 29.13 * for HVM guests, arch.monitor_table and hvm's guest CR3.
30.1 --- a/xen/arch/x86/mm/hap/guest_walk.c Thu Aug 16 10:03:26 2007 -0600 30.2 +++ b/xen/arch/x86/mm/hap/guest_walk.c Thu Aug 16 10:47:33 2007 -0600 30.3 @@ -62,7 +62,7 @@ unsigned long hap_gva_to_gfn(GUEST_PAGIN 30.4 unsigned long hap_gva_to_gfn(GUEST_PAGING_LEVELS)( 30.5 struct vcpu *v, unsigned long gva) 30.6 { 30.7 - unsigned long gcr3 = hvm_get_guest_ctrl_reg(v, 3); 30.8 + unsigned long gcr3 = v->arch.hvm_vcpu.guest_cr[3]; 30.9 int mode = GUEST_PAGING_LEVELS; 30.10 int lev, index; 30.11 paddr_t gpa = 0;
31.1 --- a/xen/arch/x86/mm/hap/hap.c Thu Aug 16 10:03:26 2007 -0600 31.2 +++ b/xen/arch/x86/mm/hap/hap.c Thu Aug 16 10:47:33 2007 -0600 31.3 @@ -603,48 +603,37 @@ static int hap_invlpg(struct vcpu *v, un 31.4 return 0; 31.5 } 31.6 31.7 -/* 31.8 - * HAP guests do not need to take any action on CR3 writes (they are still 31.9 - * intercepted, so that Xen's copy of the guest's CR3 can be kept in sync.) 31.10 - */ 31.11 static void hap_update_cr3(struct vcpu *v, int do_locking) 31.12 { 31.13 + v->arch.hvm_vcpu.hw_cr[3] = v->arch.hvm_vcpu.guest_cr[3]; 31.14 + hvm_update_guest_cr(v, 3); 31.15 } 31.16 31.17 static void hap_update_paging_modes(struct vcpu *v) 31.18 { 31.19 - struct domain *d; 31.20 + struct domain *d = v->domain; 31.21 31.22 - d = v->domain; 31.23 hap_lock(d); 31.24 31.25 - /* update guest paging mode. Note that we rely on hvm functions to detect 31.26 - * guest's paging mode. So, make sure the shadow registers (CR0, CR4, EFER) 31.27 - * reflect guest's status correctly. 31.28 - */ 31.29 - if ( hvm_paging_enabled(v) ) 31.30 - { 31.31 - if ( hvm_long_mode_enabled(v) ) 31.32 - v->arch.paging.mode = &hap_paging_long_mode; 31.33 - else if ( hvm_pae_enabled(v) ) 31.34 - v->arch.paging.mode = &hap_paging_pae_mode; 31.35 - else 31.36 - v->arch.paging.mode = &hap_paging_protected_mode; 31.37 - } 31.38 - else 31.39 - { 31.40 - v->arch.paging.mode = &hap_paging_real_mode; 31.41 - } 31.42 + v->arch.paging.mode = 31.43 + !hvm_paging_enabled(v) ? &hap_paging_real_mode : 31.44 + hvm_long_mode_enabled(v) ? &hap_paging_long_mode : 31.45 + hvm_pae_enabled(v) ? &hap_paging_pae_mode : 31.46 + &hap_paging_protected_mode; 31.47 31.48 - v->arch.paging.translate_enabled = !!hvm_paging_enabled(v); 31.49 + v->arch.paging.translate_enabled = hvm_paging_enabled(v); 31.50 31.51 if ( pagetable_is_null(v->arch.monitor_table) ) 31.52 { 31.53 mfn_t mmfn = hap_make_monitor_table(v); 31.54 v->arch.monitor_table = pagetable_from_mfn(mmfn); 31.55 make_cr3(v, mfn_x(mmfn)); 31.56 + hvm_update_host_cr3(v); 31.57 } 31.58 31.59 + /* CR3 is effectively updated by a mode change. Flush ASIDs, etc. */ 31.60 + hap_update_cr3(v, 0); 31.61 + 31.62 hap_unlock(d); 31.63 } 31.64
32.1 --- a/xen/arch/x86/mm/shadow/common.c Thu Aug 16 10:03:26 2007 -0600 32.2 +++ b/xen/arch/x86/mm/shadow/common.c Thu Aug 16 10:47:33 2007 -0600 32.3 @@ -2266,7 +2266,7 @@ static void sh_update_paging_modes(struc 32.4 ASSERT(shadow_mode_translate(d)); 32.5 ASSERT(shadow_mode_external(d)); 32.6 32.7 - v->arch.paging.translate_enabled = !!hvm_paging_enabled(v); 32.8 + v->arch.paging.translate_enabled = hvm_paging_enabled(v); 32.9 if ( !v->arch.paging.translate_enabled ) 32.10 { 32.11 /* Set v->arch.guest_table to use the p2m map, and choose 32.12 @@ -2347,7 +2347,7 @@ static void sh_update_paging_modes(struc 32.13 SHADOW_PRINTK("new paging mode: d=%u v=%u pe=%d g=%u s=%u " 32.14 "(was g=%u s=%u)\n", 32.15 d->domain_id, v->vcpu_id, 32.16 - is_hvm_domain(d) ? !!hvm_paging_enabled(v) : 1, 32.17 + is_hvm_domain(d) ? hvm_paging_enabled(v) : 1, 32.18 v->arch.paging.mode->guest_levels, 32.19 v->arch.paging.mode->shadow.shadow_levels, 32.20 old_mode ? old_mode->guest_levels : 0,
33.1 --- a/xen/arch/x86/mm/shadow/multi.c Thu Aug 16 10:03:26 2007 -0600 33.2 +++ b/xen/arch/x86/mm/shadow/multi.c Thu Aug 16 10:47:33 2007 -0600 33.3 @@ -175,7 +175,7 @@ guest_supports_superpages(struct vcpu *v 33.4 /* The _PAGE_PSE bit must be honoured in HVM guests, whenever 33.5 * CR4.PSE is set or the guest is in PAE or long mode */ 33.6 return (is_hvm_vcpu(v) && (GUEST_PAGING_LEVELS != 2 33.7 - || (hvm_get_guest_ctrl_reg(v, 4) & X86_CR4_PSE))); 33.8 + || (v->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PSE))); 33.9 } 33.10 33.11 static inline int 33.12 @@ -3483,7 +3483,7 @@ sh_update_cr3(struct vcpu *v, int do_loc 33.13 * Paravirtual guests should set v->arch.guest_table (and guest_table_user, 33.14 * if appropriate). 33.15 * HVM guests should also make sure hvm_get_guest_cntl_reg(v, 3) works; 33.16 - * this function will call hvm_update_guest_cr3() to tell them where the 33.17 + * this function will call hvm_update_guest_cr(v, 3) to tell them where the 33.18 * shadow tables are. 33.19 * If do_locking != 0, assume we are being called from outside the 33.20 * shadow code, and must take and release the shadow lock; otherwise 33.21 @@ -3525,7 +3525,7 @@ sh_update_cr3(struct vcpu *v, int do_loc 33.22 // Is paging enabled on this vcpu? 33.23 if ( paging_vcpu_mode_translate(v) ) 33.24 { 33.25 - gfn = _gfn(paddr_to_pfn(hvm_get_guest_ctrl_reg(v, 3))); 33.26 + gfn = _gfn(paddr_to_pfn(v->arch.hvm_vcpu.guest_cr[3])); 33.27 gmfn = vcpu_gfn_to_mfn(v, gfn); 33.28 ASSERT(mfn_valid(gmfn)); 33.29 ASSERT(pagetable_get_pfn(v->arch.guest_table) == mfn_x(gmfn)); 33.30 @@ -3576,11 +3576,11 @@ sh_update_cr3(struct vcpu *v, int do_loc 33.31 33.32 if ( shadow_mode_external(d) && paging_vcpu_mode_translate(v) ) 33.33 /* Paging enabled: find where in the page the l3 table is */ 33.34 - guest_idx = guest_index((void *)hvm_get_guest_ctrl_reg(v, 3)); 33.35 - else 33.36 - /* Paging disabled or PV: l3 is at the start of a page */ 33.37 - guest_idx = 0; 33.38 - 33.39 + guest_idx = guest_index((void *)v->arch.hvm_vcpu.guest_cr[3]); 33.40 + else 33.41 + /* Paging disabled or PV: l3 is at the start of a page */ 33.42 + guest_idx = 0; 33.43 + 33.44 // Ignore the low 2 bits of guest_idx -- they are really just 33.45 // cache control. 33.46 guest_idx &= ~3; 33.47 @@ -3718,18 +3718,21 @@ sh_update_cr3(struct vcpu *v, int do_loc 33.48 33.49 33.50 /// 33.51 - /// v->arch.hvm_vcpu.hw_cr3 33.52 + /// v->arch.hvm_vcpu.hw_cr[3] 33.53 /// 33.54 if ( shadow_mode_external(d) ) 33.55 { 33.56 ASSERT(is_hvm_domain(d)); 33.57 #if SHADOW_PAGING_LEVELS == 3 33.58 /* 2-on-3 or 3-on-3: Use the PAE shadow l3 table we just fabricated */ 33.59 - hvm_update_guest_cr3(v, virt_to_maddr(&v->arch.paging.shadow.l3table)); 33.60 + v->arch.hvm_vcpu.hw_cr[3] = 33.61 + virt_to_maddr(&v->arch.paging.shadow.l3table); 33.62 #else 33.63 /* 2-on-2 or 4-on-4: Just use the shadow top-level directly */ 33.64 - hvm_update_guest_cr3(v, pagetable_get_paddr(v->arch.shadow_table[0])); 33.65 + v->arch.hvm_vcpu.hw_cr[3] = 33.66 + pagetable_get_paddr(v->arch.shadow_table[0]); 33.67 #endif 33.68 + hvm_update_guest_cr(v, 3); 33.69 } 33.70 33.71 /* Fix up the linear pagetable mappings */
34.1 --- a/xen/arch/x86/physdev.c Thu Aug 16 10:03:26 2007 -0600 34.2 +++ b/xen/arch/x86/physdev.c Thu Aug 16 10:47:33 2007 -0600 34.3 @@ -28,6 +28,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.4 { 34.5 int irq; 34.6 ret_t ret; 34.7 + struct vcpu *v = current; 34.8 34.9 switch ( cmd ) 34.10 { 34.11 @@ -36,13 +37,13 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.12 ret = -EFAULT; 34.13 if ( copy_from_guest(&eoi, arg, 1) != 0 ) 34.14 break; 34.15 - ret = pirq_guest_eoi(current->domain, eoi.irq); 34.16 + ret = pirq_guest_eoi(v->domain, eoi.irq); 34.17 break; 34.18 } 34.19 34.20 /* Legacy since 0x00030202. */ 34.21 case PHYSDEVOP_IRQ_UNMASK_NOTIFY: { 34.22 - ret = pirq_guest_unmask(current->domain); 34.23 + ret = pirq_guest_unmask(v->domain); 34.24 break; 34.25 } 34.26 34.27 @@ -70,7 +71,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.28 if ( copy_from_guest(&apic, arg, 1) != 0 ) 34.29 break; 34.30 ret = -EPERM; 34.31 - if ( !IS_PRIV(current->domain) ) 34.32 + if ( !IS_PRIV(v->domain) ) 34.33 break; 34.34 ret = ioapic_guest_read(apic.apic_physbase, apic.reg, &apic.value); 34.35 if ( copy_to_guest(arg, &apic, 1) != 0 ) 34.36 @@ -84,7 +85,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.37 if ( copy_from_guest(&apic, arg, 1) != 0 ) 34.38 break; 34.39 ret = -EPERM; 34.40 - if ( !IS_PRIV(current->domain) ) 34.41 + if ( !IS_PRIV(v->domain) ) 34.42 break; 34.43 ret = ioapic_guest_write(apic.apic_physbase, apic.reg, apic.value); 34.44 break; 34.45 @@ -98,7 +99,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.46 break; 34.47 34.48 ret = -EPERM; 34.49 - if ( !IS_PRIV(current->domain) ) 34.50 + if ( !IS_PRIV(v->domain) ) 34.51 break; 34.52 34.53 irq = irq_op.irq; 34.54 @@ -120,7 +121,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.55 if ( set_iopl.iopl > 3 ) 34.56 break; 34.57 ret = 0; 34.58 - current->arch.iopl = set_iopl.iopl; 34.59 + v->arch.iopl = set_iopl.iopl; 34.60 break; 34.61 } 34.62 34.63 @@ -135,11 +136,11 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H 34.64 break; 34.65 ret = 0; 34.66 #ifndef COMPAT 34.67 - current->arch.iobmp = set_iobitmap.bitmap; 34.68 + v->arch.iobmp = set_iobitmap.bitmap; 34.69 #else 34.70 - guest_from_compat_handle(current->arch.iobmp, set_iobitmap.bitmap); 34.71 + guest_from_compat_handle(v->arch.iobmp, set_iobitmap.bitmap); 34.72 #endif 34.73 - current->arch.iobmp_limit = set_iobitmap.nr_ports; 34.74 + v->arch.iobmp_limit = set_iobitmap.nr_ports; 34.75 break; 34.76 } 34.77
35.1 --- a/xen/arch/x86/setup.c Thu Aug 16 10:03:26 2007 -0600 35.2 +++ b/xen/arch/x86/setup.c Thu Aug 16 10:47:33 2007 -0600 35.3 @@ -106,6 +106,8 @@ extern void init_IRQ(void); 35.4 extern void trap_init(void); 35.5 extern void early_time_init(void); 35.6 extern void early_cpu_init(void); 35.7 +extern void vesa_init(void); 35.8 +extern void vesa_mtrr_init(void); 35.9 35.10 struct tss_struct init_tss[NR_CPUS]; 35.11 35.12 @@ -282,9 +284,28 @@ static void __init srat_detect_node(int 35.13 printk(KERN_INFO "CPU %d APIC %d -> Node %d\n", cpu, apicid, node); 35.14 } 35.15 35.16 +/* 35.17 + * Ensure a given physical memory range is present in the bootstrap mappings. 35.18 + * Use superpage mappings to ensure that pagetable memory needn't be allocated. 35.19 + */ 35.20 +static void __init bootstrap_map(unsigned long start, unsigned long end) 35.21 +{ 35.22 + unsigned long mask = (1UL << L2_PAGETABLE_SHIFT) - 1; 35.23 + start = start & ~mask; 35.24 + end = (end + mask) & ~mask; 35.25 + if ( end > BOOTSTRAP_DIRECTMAP_END ) 35.26 + panic("Cannot access memory beyond end of " 35.27 + "bootstrap direct-map area\n"); 35.28 + map_pages_to_xen( 35.29 + (unsigned long)maddr_to_bootstrap_virt(start), 35.30 + start >> PAGE_SHIFT, (end-start) >> PAGE_SHIFT, PAGE_HYPERVISOR); 35.31 +} 35.32 + 35.33 static void __init move_memory( 35.34 unsigned long dst, unsigned long src_start, unsigned long src_end) 35.35 { 35.36 + bootstrap_map(src_start, src_end); 35.37 + bootstrap_map(dst, dst + src_end - src_start); 35.38 memmove(maddr_to_bootstrap_virt(dst), 35.39 maddr_to_bootstrap_virt(src_start), 35.40 src_end - src_start); 35.41 @@ -882,6 +903,7 @@ void __init __start_xen(unsigned long mb 35.42 #ifdef __x86_64__ 35.43 init_xenheap_pages(xen_phys_start, __pa(&_start)); 35.44 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT; 35.45 + vesa_init(); 35.46 #endif 35.47 xenheap_phys_start = xen_phys_start; 35.48 printk("Xen heap: %luMB (%lukB)\n", 35.49 @@ -947,6 +969,9 @@ void __init __start_xen(unsigned long mb 35.50 set_in_cr4(X86_CR4_OSFXSR); 35.51 if ( cpu_has_xmm ) 35.52 set_in_cr4(X86_CR4_OSXMMEXCPT); 35.53 +#ifdef CONFIG_X86_64 35.54 + vesa_mtrr_init(); 35.55 +#endif 35.56 35.57 if ( opt_nosmp ) 35.58 max_cpus = 0;
36.1 --- a/xen/arch/x86/string.c Thu Aug 16 10:03:26 2007 -0600 36.2 +++ b/xen/arch/x86/string.c Thu Aug 16 10:47:33 2007 -0600 36.3 @@ -11,10 +11,18 @@ 36.4 #undef memcpy 36.5 void *memcpy(void *dest, const void *src, size_t n) 36.6 { 36.7 - int d0, d1, d2; 36.8 + long d0, d1, d2; 36.9 36.10 __asm__ __volatile__ ( 36.11 - " rep ; movsl ; " 36.12 +#ifdef __i386__ 36.13 + " rep movsl ; " 36.14 +#else 36.15 + " rep movsq ; " 36.16 + " testb $4,%b4 ; " 36.17 + " je 0f ; " 36.18 + " movsl ; " 36.19 + "0: ; " 36.20 +#endif 36.21 " testb $2,%b4 ; " 36.22 " je 1f ; " 36.23 " movsw ; " 36.24 @@ -23,7 +31,7 @@ void *memcpy(void *dest, const void *src 36.25 " movsb ; " 36.26 "2: " 36.27 : "=&c" (d0), "=&D" (d1), "=&S" (d2) 36.28 - : "0" (n/4), "q" (n), "1" (dest), "2" (src) 36.29 + : "0" (n/sizeof(long)), "q" (n), "1" (dest), "2" (src) 36.30 : "memory"); 36.31 36.32 return dest; 36.33 @@ -32,10 +40,10 @@ void *memcpy(void *dest, const void *src 36.34 #undef memset 36.35 void *memset(void *s, int c, size_t n) 36.36 { 36.37 - int d0, d1; 36.38 + long d0, d1; 36.39 36.40 __asm__ __volatile__ ( 36.41 - "rep ; stosb" 36.42 + "rep stosb" 36.43 : "=&c" (d0), "=&D" (d1) 36.44 : "a" (c), "1" (s), "0" (n) 36.45 : "memory"); 36.46 @@ -46,14 +54,14 @@ void *memset(void *s, int c, size_t n) 36.47 #undef memmove 36.48 void *memmove(void *dest, const void *src, size_t n) 36.49 { 36.50 - int d0, d1, d2; 36.51 + long d0, d1, d2; 36.52 36.53 if ( dest < src ) 36.54 return memcpy(dest, src, n); 36.55 36.56 __asm__ __volatile__ ( 36.57 " std ; " 36.58 - " rep ; movsb ; " 36.59 + " rep movsb ; " 36.60 " cld " 36.61 : "=&c" (d0), "=&S" (d1), "=&D" (d2) 36.62 : "0" (n), "1" (n-1+(const char *)src), "2" (n-1+(char *)dest)
37.1 --- a/xen/arch/x86/traps.c Thu Aug 16 10:03:26 2007 -0600 37.2 +++ b/xen/arch/x86/traps.c Thu Aug 16 10:47:33 2007 -0600 37.3 @@ -1219,7 +1219,7 @@ static int emulate_privileged_op(struct 37.4 unsigned long code_base, code_limit; 37.5 char io_emul_stub[16]; 37.6 void (*io_emul)(struct cpu_user_regs *) __attribute__((__regparm__(1))); 37.7 - u32 l, h; 37.8 + u32 l, h, eax, edx; 37.9 37.10 if ( !read_descriptor(regs->cs, v, regs, 37.11 &code_base, &code_limit, &ar, 37.12 @@ -1696,43 +1696,43 @@ static int emulate_privileged_op(struct 37.13 break; 37.14 37.15 case 0x30: /* WRMSR */ 37.16 + eax = regs->eax; 37.17 + edx = regs->edx; 37.18 + res = ((u64)edx << 32) | eax; 37.19 switch ( regs->ecx ) 37.20 { 37.21 #ifdef CONFIG_X86_64 37.22 case MSR_FS_BASE: 37.23 if ( is_pv_32on64_vcpu(v) ) 37.24 goto fail; 37.25 - if ( wrmsr_safe(MSR_FS_BASE, regs->eax, regs->edx) ) 37.26 + if ( wrmsr_safe(MSR_FS_BASE, eax, edx) ) 37.27 goto fail; 37.28 - v->arch.guest_context.fs_base = 37.29 - ((u64)regs->edx << 32) | regs->eax; 37.30 + v->arch.guest_context.fs_base = res; 37.31 break; 37.32 case MSR_GS_BASE: 37.33 if ( is_pv_32on64_vcpu(v) ) 37.34 goto fail; 37.35 - if ( wrmsr_safe(MSR_GS_BASE, regs->eax, regs->edx) ) 37.36 + if ( wrmsr_safe(MSR_GS_BASE, eax, edx) ) 37.37 goto fail; 37.38 - v->arch.guest_context.gs_base_kernel = 37.39 - ((u64)regs->edx << 32) | regs->eax; 37.40 + v->arch.guest_context.gs_base_kernel = res; 37.41 break; 37.42 case MSR_SHADOW_GS_BASE: 37.43 if ( is_pv_32on64_vcpu(v) ) 37.44 goto fail; 37.45 - if ( wrmsr_safe(MSR_SHADOW_GS_BASE, regs->eax, regs->edx) ) 37.46 + if ( wrmsr_safe(MSR_SHADOW_GS_BASE, eax, edx) ) 37.47 goto fail; 37.48 - v->arch.guest_context.gs_base_user = 37.49 - ((u64)regs->edx << 32) | regs->eax; 37.50 + v->arch.guest_context.gs_base_user = res; 37.51 break; 37.52 #endif 37.53 default: 37.54 - if ( wrmsr_hypervisor_regs(regs->ecx, regs->eax, regs->edx) ) 37.55 + if ( wrmsr_hypervisor_regs(regs->ecx, eax, edx) ) 37.56 break; 37.57 37.58 if ( (rdmsr_safe(regs->ecx, l, h) != 0) || 37.59 - (regs->eax != l) || (regs->edx != h) ) 37.60 + (eax != l) || (edx != h) ) 37.61 gdprintk(XENLOG_WARNING, "Domain attempted WRMSR %p from " 37.62 - "%08x:%08x to %08lx:%08lx.\n", 37.63 - _p(regs->ecx), h, l, (long)regs->edx, (long)regs->eax); 37.64 + "%08x:%08x to %08x:%08x.\n", 37.65 + _p(regs->ecx), h, l, edx, eax); 37.66 break; 37.67 } 37.68 break;
38.1 --- a/xen/arch/x86/x86_32/asm-offsets.c Thu Aug 16 10:03:26 2007 -0600 38.2 +++ b/xen/arch/x86/x86_32/asm-offsets.c Thu Aug 16 10:47:33 2007 -0600 38.3 @@ -85,7 +85,7 @@ void __dummy__(void) 38.4 BLANK(); 38.5 38.6 OFFSET(VCPU_vmx_launched, struct vcpu, arch.hvm_vmx.launched); 38.7 - OFFSET(VCPU_vmx_cr2, struct vcpu, arch.hvm_vmx.cpu_cr2); 38.8 + OFFSET(VCPU_hvm_guest_cr2, struct vcpu, arch.hvm_vcpu.guest_cr[2]); 38.9 BLANK(); 38.10 38.11 OFFSET(VMCB_rax, struct vmcb_struct, rax);
39.1 --- a/xen/arch/x86/x86_32/traps.c Thu Aug 16 10:03:26 2007 -0600 39.2 +++ b/xen/arch/x86/x86_32/traps.c Thu Aug 16 10:47:33 2007 -0600 39.3 @@ -172,6 +172,7 @@ asmlinkage void do_double_fault(void) 39.4 unsigned long do_iret(void) 39.5 { 39.6 struct cpu_user_regs *regs = guest_cpu_user_regs(); 39.7 + struct vcpu *v = current; 39.8 u32 eflags; 39.9 39.10 /* Check worst-case stack frame for overlap with Xen protected area. */ 39.11 @@ -215,10 +216,10 @@ unsigned long do_iret(void) 39.12 } 39.13 39.14 /* No longer in NMI context. */ 39.15 - current->nmi_masked = 0; 39.16 + v->nmi_masked = 0; 39.17 39.18 /* Restore upcall mask from supplied EFLAGS.IF. */ 39.19 - current->vcpu_info->evtchn_upcall_mask = !(eflags & X86_EFLAGS_IF); 39.20 + vcpu_info(v, evtchn_upcall_mask) = !(eflags & X86_EFLAGS_IF); 39.21 39.22 /* 39.23 * The hypercall exit path will overwrite EAX with this return 39.24 @@ -228,7 +229,7 @@ unsigned long do_iret(void) 39.25 39.26 exit_and_crash: 39.27 gdprintk(XENLOG_ERR, "Fatal error\n"); 39.28 - domain_crash(current->domain); 39.29 + domain_crash(v->domain); 39.30 return 0; 39.31 } 39.32
40.1 --- a/xen/arch/x86/x86_64/asm-offsets.c Thu Aug 16 10:03:26 2007 -0600 40.2 +++ b/xen/arch/x86/x86_64/asm-offsets.c Thu Aug 16 10:47:33 2007 -0600 40.3 @@ -88,7 +88,7 @@ void __dummy__(void) 40.4 BLANK(); 40.5 40.6 OFFSET(VCPU_vmx_launched, struct vcpu, arch.hvm_vmx.launched); 40.7 - OFFSET(VCPU_vmx_cr2, struct vcpu, arch.hvm_vmx.cpu_cr2); 40.8 + OFFSET(VCPU_hvm_guest_cr2, struct vcpu, arch.hvm_vcpu.guest_cr[2]); 40.9 BLANK(); 40.10 40.11 OFFSET(DOMAIN_is_32bit_pv, struct domain, arch.is_32bit_pv);
41.1 --- a/xen/arch/x86/x86_64/compat/traps.c Thu Aug 16 10:03:26 2007 -0600 41.2 +++ b/xen/arch/x86/x86_64/compat/traps.c Thu Aug 16 10:47:33 2007 -0600 41.3 @@ -37,6 +37,7 @@ void compat_show_guest_stack(struct cpu_ 41.4 unsigned int compat_iret(void) 41.5 { 41.6 struct cpu_user_regs *regs = guest_cpu_user_regs(); 41.7 + struct vcpu *v = current; 41.8 u32 eflags; 41.9 41.10 /* Trim stack pointer to 32 bits. */ 41.11 @@ -70,7 +71,7 @@ unsigned int compat_iret(void) 41.12 * mode frames). 41.13 */ 41.14 const struct trap_info *ti; 41.15 - u32 x, ksp = current->arch.guest_context.kernel_sp - 40; 41.16 + u32 x, ksp = v->arch.guest_context.kernel_sp - 40; 41.17 unsigned int i; 41.18 int rc = 0; 41.19 41.20 @@ -95,9 +96,9 @@ unsigned int compat_iret(void) 41.21 if ( rc ) 41.22 goto exit_and_crash; 41.23 regs->_esp = ksp; 41.24 - regs->ss = current->arch.guest_context.kernel_ss; 41.25 + regs->ss = v->arch.guest_context.kernel_ss; 41.26 41.27 - ti = ¤t->arch.guest_context.trap_ctxt[13]; 41.28 + ti = &v->arch.guest_context.trap_ctxt[13]; 41.29 if ( TI_GET_IF(ti) ) 41.30 eflags &= ~X86_EFLAGS_IF; 41.31 regs->_eflags = eflags & ~(X86_EFLAGS_VM|X86_EFLAGS_RF| 41.32 @@ -121,10 +122,10 @@ unsigned int compat_iret(void) 41.33 regs->_esp += 16; 41.34 41.35 /* No longer in NMI context. */ 41.36 - current->nmi_masked = 0; 41.37 + v->nmi_masked = 0; 41.38 41.39 /* Restore upcall mask from supplied EFLAGS.IF. */ 41.40 - vcpu_info(current, evtchn_upcall_mask) = !(eflags & X86_EFLAGS_IF); 41.41 + vcpu_info(v, evtchn_upcall_mask) = !(eflags & X86_EFLAGS_IF); 41.42 41.43 /* 41.44 * The hypercall exit path will overwrite EAX with this return 41.45 @@ -134,11 +135,12 @@ unsigned int compat_iret(void) 41.46 41.47 exit_and_crash: 41.48 gdprintk(XENLOG_ERR, "Fatal error\n"); 41.49 - domain_crash(current->domain); 41.50 + domain_crash(v->domain); 41.51 return 0; 41.52 } 41.53 41.54 -static long compat_register_guest_callback(struct compat_callback_register *reg) 41.55 +static long compat_register_guest_callback( 41.56 + struct compat_callback_register *reg) 41.57 { 41.58 long ret = 0; 41.59 struct vcpu *v = current; 41.60 @@ -175,7 +177,8 @@ static long compat_register_guest_callba 41.61 return ret; 41.62 } 41.63 41.64 -static long compat_unregister_guest_callback(struct compat_callback_unregister *unreg) 41.65 +static long compat_unregister_guest_callback( 41.66 + struct compat_callback_unregister *unreg) 41.67 { 41.68 long ret; 41.69
42.1 --- a/xen/arch/x86/x86_64/traps.c Thu Aug 16 10:03:26 2007 -0600 42.2 +++ b/xen/arch/x86/x86_64/traps.c Thu Aug 16 10:47:33 2007 -0600 42.3 @@ -235,10 +235,10 @@ unsigned long do_iret(void) 42.4 } 42.5 42.6 /* No longer in NMI context. */ 42.7 - current->nmi_masked = 0; 42.8 + v->nmi_masked = 0; 42.9 42.10 /* Restore upcall mask from supplied EFLAGS.IF. */ 42.11 - vcpu_info(current, evtchn_upcall_mask) = !(iret_saved.rflags & EF_IE); 42.12 + vcpu_info(v, evtchn_upcall_mask) = !(iret_saved.rflags & EF_IE); 42.13 42.14 /* Saved %rax gets written back to regs->rax in entry.S. */ 42.15 return iret_saved.rax;
43.1 --- a/xen/common/domctl.c Thu Aug 16 10:03:26 2007 -0600 43.2 +++ b/xen/common/domctl.c Thu Aug 16 10:47:33 2007 -0600 43.3 @@ -463,19 +463,13 @@ long do_domctl(XEN_GUEST_HANDLE(xen_domc 43.4 case XEN_DOMCTL_getdomaininfo: 43.5 { 43.6 struct domain *d; 43.7 - domid_t dom; 43.8 - 43.9 - dom = op->domain; 43.10 - if ( dom == DOMID_SELF ) 43.11 - dom = current->domain->domain_id; 43.12 + domid_t dom = op->domain; 43.13 43.14 rcu_read_lock(&domlist_read_lock); 43.15 43.16 for_each_domain ( d ) 43.17 - { 43.18 if ( d->domain_id >= dom ) 43.19 break; 43.20 - } 43.21 43.22 if ( d == NULL ) 43.23 {
44.1 --- a/xen/common/page_alloc.c Thu Aug 16 10:03:26 2007 -0600 44.2 +++ b/xen/common/page_alloc.c Thu Aug 16 10:47:33 2007 -0600 44.3 @@ -54,21 +54,14 @@ boolean_param("bootscrub", opt_bootscrub 44.4 /* 44.5 * Bit width of the DMA heap. 44.6 */ 44.7 -static unsigned int dma_bitsize = CONFIG_DMA_BITSIZE; 44.8 -static unsigned long max_dma_mfn = (1UL<<(CONFIG_DMA_BITSIZE-PAGE_SHIFT))-1; 44.9 +static unsigned int dma_bitsize = CONFIG_DMA_BITSIZE; 44.10 static void __init parse_dma_bits(char *s) 44.11 { 44.12 unsigned int v = simple_strtol(s, NULL, 0); 44.13 if ( v >= (BITS_PER_LONG + PAGE_SHIFT) ) 44.14 - { 44.15 dma_bitsize = BITS_PER_LONG + PAGE_SHIFT; 44.16 - max_dma_mfn = ~0UL; 44.17 - } 44.18 else if ( v > PAGE_SHIFT + 1 ) 44.19 - { 44.20 dma_bitsize = v; 44.21 - max_dma_mfn = (1UL << (dma_bitsize - PAGE_SHIFT)) - 1; 44.22 - } 44.23 else 44.24 printk("Invalid dma_bits value of %u ignored.\n", v); 44.25 }
45.1 --- a/xen/common/xencomm.c Thu Aug 16 10:03:26 2007 -0600 45.2 +++ b/xen/common/xencomm.c Thu Aug 16 10:47:33 2007 -0600 45.3 @@ -26,35 +26,36 @@ 45.4 #include <public/xen.h> 45.5 #include <public/xencomm.h> 45.6 45.7 - 45.8 #undef DEBUG 45.9 #ifdef DEBUG 45.10 -static int xencomm_debug = 1; /* extremely verbose */ 45.11 +#define xc_dprintk(f, a...) printk("[xencomm]" f , ## a) 45.12 #else 45.13 -#define xencomm_debug 0 45.14 +#define xc_dprintk(f, a...) ((void)0) 45.15 #endif 45.16 45.17 +static void* 45.18 +xencomm_maddr_to_vaddr(unsigned long maddr) 45.19 +{ 45.20 + return maddr ? maddr_to_virt(maddr) : NULL; 45.21 +} 45.22 + 45.23 static unsigned long 45.24 -xencomm_inline_from_guest(void *to, const void *from, unsigned int n, 45.25 - unsigned int skip) 45.26 +xencomm_inline_from_guest( 45.27 + void *to, const void *from, unsigned int n, unsigned int skip) 45.28 { 45.29 - unsigned long src_paddr = xencomm_inline_addr(from); 45.30 + unsigned long src_paddr = xencomm_inline_addr(from) + skip; 45.31 45.32 - src_paddr += skip; 45.33 - 45.34 - while (n > 0) { 45.35 - unsigned int chunksz; 45.36 + while ( n > 0 ) 45.37 + { 45.38 + unsigned int chunksz, bytes; 45.39 unsigned long src_maddr; 45.40 - unsigned int bytes; 45.41 45.42 chunksz = PAGE_SIZE - (src_paddr % PAGE_SIZE); 45.43 - 45.44 - bytes = min(chunksz, n); 45.45 + bytes = min(chunksz, n); 45.46 45.47 src_maddr = paddr_to_maddr(src_paddr); 45.48 - if (xencomm_debug) 45.49 - printk("%lx[%d] -> %lx\n", src_maddr, bytes, (unsigned long)to); 45.50 - memcpy(to, (void *)src_maddr, bytes); 45.51 + xc_dprintk("%lx[%d] -> %lx\n", src_maddr, bytes, (unsigned long)to); 45.52 + memcpy(to, maddr_to_virt(src_maddr), bytes); 45.53 src_paddr += bytes; 45.54 to += bytes; 45.55 n -= bytes; 45.56 @@ -77,36 +78,40 @@ xencomm_inline_from_guest(void *to, cons 45.57 * On success, this will be zero. 45.58 */ 45.59 unsigned long 45.60 -xencomm_copy_from_guest(void *to, const void *from, unsigned int n, 45.61 - unsigned int skip) 45.62 +xencomm_copy_from_guest( 45.63 + void *to, const void *from, unsigned int n, unsigned int skip) 45.64 { 45.65 struct xencomm_desc *desc; 45.66 unsigned int from_pos = 0; 45.67 unsigned int to_pos = 0; 45.68 unsigned int i = 0; 45.69 45.70 - if (xencomm_is_inline(from)) 45.71 + if ( xencomm_is_inline(from) ) 45.72 return xencomm_inline_from_guest(to, from, n, skip); 45.73 45.74 - /* first we need to access the descriptor */ 45.75 - desc = (struct xencomm_desc *)paddr_to_maddr((unsigned long)from); 45.76 - if (desc == NULL) 45.77 + /* First we need to access the descriptor. */ 45.78 + desc = (struct xencomm_desc *) 45.79 + xencomm_maddr_to_vaddr(paddr_to_maddr((unsigned long)from)); 45.80 + if ( desc == NULL ) 45.81 return n; 45.82 45.83 - if (desc->magic != XENCOMM_MAGIC) { 45.84 + if ( desc->magic != XENCOMM_MAGIC ) 45.85 + { 45.86 printk("%s: error: %p magic was 0x%x\n", 45.87 __func__, desc, desc->magic); 45.88 return n; 45.89 } 45.90 45.91 - /* iterate through the descriptor, copying up to a page at a time */ 45.92 - while ((to_pos < n) && (i < desc->nr_addrs)) { 45.93 + /* Iterate through the descriptor, copying up to a page at a time. */ 45.94 + while ( (to_pos < n) && (i < desc->nr_addrs) ) 45.95 + { 45.96 unsigned long src_paddr = desc->address[i]; 45.97 unsigned int pgoffset; 45.98 unsigned int chunksz; 45.99 unsigned int chunk_skip; 45.100 45.101 - if (src_paddr == XENCOMM_INVALID) { 45.102 + if ( src_paddr == XENCOMM_INVALID ) 45.103 + { 45.104 i++; 45.105 continue; 45.106 } 45.107 @@ -119,18 +124,18 @@ xencomm_copy_from_guest(void *to, const 45.108 chunksz -= chunk_skip; 45.109 skip -= chunk_skip; 45.110 45.111 - if (skip == 0 && chunksz > 0) { 45.112 + if ( (skip == 0) && (chunksz > 0) ) 45.113 + { 45.114 unsigned long src_maddr; 45.115 unsigned long dest = (unsigned long)to + to_pos; 45.116 unsigned int bytes = min(chunksz, n - to_pos); 45.117 45.118 src_maddr = paddr_to_maddr(src_paddr + chunk_skip); 45.119 - if (src_maddr == 0) 45.120 + if ( src_maddr == 0 ) 45.121 return n - to_pos; 45.122 45.123 - if (xencomm_debug) 45.124 - printk("%lx[%d] -> %lx\n", src_maddr, bytes, dest); 45.125 - memcpy((void *)dest, (void *)src_maddr, bytes); 45.126 + xc_dprintk("%lx[%d] -> %lx\n", src_maddr, bytes, dest); 45.127 + memcpy((void *)dest, maddr_to_virt(src_maddr), bytes); 45.128 from_pos += bytes; 45.129 to_pos += bytes; 45.130 } 45.131 @@ -142,32 +147,28 @@ xencomm_copy_from_guest(void *to, const 45.132 } 45.133 45.134 static unsigned long 45.135 -xencomm_inline_to_guest(void *to, const void *from, unsigned int n, 45.136 - unsigned int skip) 45.137 +xencomm_inline_to_guest( 45.138 + void *to, const void *from, unsigned int n, unsigned int skip) 45.139 { 45.140 - unsigned long dest_paddr = xencomm_inline_addr(to); 45.141 + unsigned long dest_paddr = xencomm_inline_addr(to) + skip; 45.142 45.143 - dest_paddr += skip; 45.144 - 45.145 - while (n > 0) { 45.146 - unsigned int chunksz; 45.147 + while ( n > 0 ) 45.148 + { 45.149 + unsigned int chunksz, bytes; 45.150 unsigned long dest_maddr; 45.151 - unsigned int bytes; 45.152 45.153 chunksz = PAGE_SIZE - (dest_paddr % PAGE_SIZE); 45.154 - 45.155 - bytes = min(chunksz, n); 45.156 + bytes = min(chunksz, n); 45.157 45.158 dest_maddr = paddr_to_maddr(dest_paddr); 45.159 - if (xencomm_debug) 45.160 - printk("%lx[%d] -> %lx\n", (unsigned long)from, bytes, dest_maddr); 45.161 - memcpy((void *)dest_maddr, (void *)from, bytes); 45.162 + xc_dprintk("%lx[%d] -> %lx\n", (unsigned long)from, bytes, dest_maddr); 45.163 + memcpy(maddr_to_virt(dest_maddr), (void *)from, bytes); 45.164 dest_paddr += bytes; 45.165 from += bytes; 45.166 n -= bytes; 45.167 } 45.168 45.169 - /* Always successful. */ 45.170 + /* Always successful. */ 45.171 return 0; 45.172 } 45.173 45.174 @@ -184,35 +185,37 @@ xencomm_inline_to_guest(void *to, const 45.175 * On success, this will be zero. 45.176 */ 45.177 unsigned long 45.178 -xencomm_copy_to_guest(void *to, const void *from, unsigned int n, 45.179 - unsigned int skip) 45.180 +xencomm_copy_to_guest( 45.181 + void *to, const void *from, unsigned int n, unsigned int skip) 45.182 { 45.183 struct xencomm_desc *desc; 45.184 unsigned int from_pos = 0; 45.185 unsigned int to_pos = 0; 45.186 unsigned int i = 0; 45.187 45.188 - if (xencomm_is_inline(to)) 45.189 + if ( xencomm_is_inline(to) ) 45.190 return xencomm_inline_to_guest(to, from, n, skip); 45.191 45.192 - /* first we need to access the descriptor */ 45.193 - desc = (struct xencomm_desc *)paddr_to_maddr((unsigned long)to); 45.194 - if (desc == NULL) 45.195 + /* First we need to access the descriptor. */ 45.196 + desc = (struct xencomm_desc *) 45.197 + xencomm_maddr_to_vaddr(paddr_to_maddr((unsigned long)to)); 45.198 + if ( desc == NULL ) 45.199 return n; 45.200 45.201 - if (desc->magic != XENCOMM_MAGIC) { 45.202 + if ( desc->magic != XENCOMM_MAGIC ) 45.203 + { 45.204 printk("%s error: %p magic was 0x%x\n", __func__, desc, desc->magic); 45.205 return n; 45.206 } 45.207 45.208 - /* iterate through the descriptor, copying up to a page at a time */ 45.209 - while ((from_pos < n) && (i < desc->nr_addrs)) { 45.210 + /* Iterate through the descriptor, copying up to a page at a time. */ 45.211 + while ( (from_pos < n) && (i < desc->nr_addrs) ) 45.212 + { 45.213 unsigned long dest_paddr = desc->address[i]; 45.214 - unsigned int pgoffset; 45.215 - unsigned int chunksz; 45.216 - unsigned int chunk_skip; 45.217 + unsigned int pgoffset, chunksz, chunk_skip; 45.218 45.219 - if (dest_paddr == XENCOMM_INVALID) { 45.220 + if ( dest_paddr == XENCOMM_INVALID ) 45.221 + { 45.222 i++; 45.223 continue; 45.224 } 45.225 @@ -225,18 +228,18 @@ xencomm_copy_to_guest(void *to, const vo 45.226 chunksz -= chunk_skip; 45.227 skip -= chunk_skip; 45.228 45.229 - if (skip == 0 && chunksz > 0) { 45.230 + if ( (skip == 0) && (chunksz > 0) ) 45.231 + { 45.232 unsigned long dest_maddr; 45.233 unsigned long source = (unsigned long)from + from_pos; 45.234 unsigned int bytes = min(chunksz, n - from_pos); 45.235 45.236 dest_maddr = paddr_to_maddr(dest_paddr + chunk_skip); 45.237 - if (dest_maddr == 0) 45.238 - return -1; 45.239 + if ( dest_maddr == 0 ) 45.240 + return n - from_pos; 45.241 45.242 - if (xencomm_debug) 45.243 - printk("%lx[%d] -> %lx\n", source, bytes, dest_maddr); 45.244 - memcpy((void *)dest_maddr, (void *)source, bytes); 45.245 + xc_dprintk("%lx[%d] -> %lx\n", source, bytes, dest_maddr); 45.246 + memcpy(maddr_to_virt(dest_maddr), (void *)source, bytes); 45.247 from_pos += bytes; 45.248 to_pos += bytes; 45.249 } 45.250 @@ -260,38 +263,46 @@ int xencomm_add_offset(void **handle, un 45.251 struct xencomm_desc *desc; 45.252 int i = 0; 45.253 45.254 - if (xencomm_is_inline(*handle)) 45.255 + if ( xencomm_is_inline(*handle) ) 45.256 return xencomm_inline_add_offset(handle, bytes); 45.257 45.258 - /* first we need to access the descriptor */ 45.259 - desc = (struct xencomm_desc *)paddr_to_maddr((unsigned long)*handle); 45.260 - if (desc == NULL) 45.261 + /* First we need to access the descriptor. */ 45.262 + desc = (struct xencomm_desc *) 45.263 + xencomm_maddr_to_vaddr(paddr_to_maddr((unsigned long)*handle)); 45.264 + if ( desc == NULL ) 45.265 return -1; 45.266 45.267 - if (desc->magic != XENCOMM_MAGIC) { 45.268 + if ( desc->magic != XENCOMM_MAGIC ) 45.269 + { 45.270 printk("%s error: %p magic was 0x%x\n", __func__, desc, desc->magic); 45.271 return -1; 45.272 } 45.273 45.274 - /* iterate through the descriptor incrementing addresses */ 45.275 - while ((bytes > 0) && (i < desc->nr_addrs)) { 45.276 + /* Iterate through the descriptor incrementing addresses. */ 45.277 + while ( (bytes > 0) && (i < desc->nr_addrs) ) 45.278 + { 45.279 unsigned long dest_paddr = desc->address[i]; 45.280 - unsigned int pgoffset; 45.281 - unsigned int chunksz; 45.282 - unsigned int chunk_skip; 45.283 + unsigned int pgoffset, chunksz, chunk_skip; 45.284 + 45.285 + if ( dest_paddr == XENCOMM_INVALID ) 45.286 + { 45.287 + i++; 45.288 + continue; 45.289 + } 45.290 45.291 pgoffset = dest_paddr % PAGE_SIZE; 45.292 chunksz = PAGE_SIZE - pgoffset; 45.293 45.294 chunk_skip = min(chunksz, bytes); 45.295 - if (chunk_skip == chunksz) { 45.296 - /* exhausted this page */ 45.297 - desc->address[i] = XENCOMM_INVALID; 45.298 - } else { 45.299 + if ( chunk_skip == chunksz ) 45.300 + desc->address[i] = XENCOMM_INVALID; /* exchausted this page */ 45.301 + else 45.302 desc->address[i] += chunk_skip; 45.303 - } 45.304 bytes -= chunk_skip; 45.305 + 45.306 + i++; 45.307 } 45.308 + 45.309 return 0; 45.310 } 45.311 45.312 @@ -300,17 +311,17 @@ int xencomm_handle_is_null(void *handle) 45.313 struct xencomm_desc *desc; 45.314 int i; 45.315 45.316 - if (xencomm_is_inline(handle)) 45.317 + if ( xencomm_is_inline(handle) ) 45.318 return xencomm_inline_addr(handle) == 0; 45.319 45.320 - desc = (struct xencomm_desc *)paddr_to_maddr((unsigned long)handle); 45.321 - if (desc == NULL) 45.322 + desc = (struct xencomm_desc *) 45.323 + xencomm_maddr_to_vaddr(paddr_to_maddr((unsigned long)handle)); 45.324 + if ( desc == NULL ) 45.325 return 1; 45.326 45.327 - for (i = 0; i < desc->nr_addrs; i++) 45.328 - if (desc->address[i] != XENCOMM_INVALID) 45.329 + for ( i = 0; i < desc->nr_addrs; i++ ) 45.330 + if ( desc->address[i] != XENCOMM_INVALID ) 45.331 return 0; 45.332 45.333 return 1; 45.334 } 45.335 -
46.1 --- a/xen/drivers/char/console.c Thu Aug 16 10:03:26 2007 -0600 46.2 +++ b/xen/drivers/char/console.c Thu Aug 16 10:47:33 2007 -0600 46.3 @@ -331,13 +331,11 @@ static long guest_console_write(XEN_GUES 46.4 kbuf[kcount] = '\0'; 46.5 46.6 sercon_puts(kbuf); 46.7 + vga_puts(kbuf); 46.8 46.9 - for ( kptr = kbuf; *kptr != '\0'; kptr++ ) 46.10 - { 46.11 - vga_putchar(*kptr); 46.12 - if ( opt_console_to_ring ) 46.13 + if ( opt_console_to_ring ) 46.14 + for ( kptr = kbuf; *kptr != '\0'; kptr++ ) 46.15 putchar_console_ring(*kptr); 46.16 - } 46.17 46.18 if ( opt_console_to_ring ) 46.19 send_guest_global_virq(dom0, VIRQ_CON_RING); 46.20 @@ -404,12 +402,10 @@ static void __putstr(const char *str) 46.21 int c; 46.22 46.23 sercon_puts(str); 46.24 + vga_puts(str); 46.25 46.26 while ( (c = *str++) != '\0' ) 46.27 - { 46.28 - vga_putchar(c); 46.29 putchar_console_ring(c); 46.30 - } 46.31 46.32 send_guest_global_virq(dom0, VIRQ_CON_RING); 46.33 }
47.1 --- a/xen/drivers/video/Makefile Thu Aug 16 10:03:26 2007 -0600 47.2 +++ b/xen/drivers/video/Makefile Thu Aug 16 10:47:33 2007 -0600 47.3 @@ -1,4 +1,8 @@ 47.4 -obj-y += font_8x14.o 47.5 -obj-y += font_8x16.o 47.6 -obj-y += font_8x8.o 47.7 -obj-y += vga.o 47.8 +obj-y := vga.o 47.9 +obj-$(CONFIG_X86_64) += font_8x14.o 47.10 +obj-$(CONFIG_X86_64) += font_8x16.o 47.11 +obj-$(CONFIG_X86_64) += font_8x8.o 47.12 +obj-$(CONFIG_X86_64) += vesa.o 47.13 + 47.14 +# extra dependencies 47.15 +vesa.o: font.h
48.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 48.2 +++ b/xen/drivers/video/vesa.c Thu Aug 16 10:47:33 2007 -0600 48.3 @@ -0,0 +1,307 @@ 48.4 +/****************************************************************************** 48.5 + * vesa.c 48.6 + * 48.7 + * VESA linear frame buffer handling. 48.8 + */ 48.9 + 48.10 +#include <xen/config.h> 48.11 +#include <xen/compile.h> 48.12 +#include <xen/init.h> 48.13 +#include <xen/lib.h> 48.14 +#include <xen/mm.h> 48.15 +#include <xen/errno.h> 48.16 +#include <xen/console.h> 48.17 +#include <xen/vga.h> 48.18 +#include "font.h" 48.19 + 48.20 +#define vlfb_info vga_console_info.u.vesa_lfb 48.21 +#define text_columns (vlfb_info.width / font->width) 48.22 +#define text_rows (vlfb_info.height / font->height) 48.23 + 48.24 +static void vesa_redraw_puts(const char *s); 48.25 +static void vesa_scroll_puts(const char *s); 48.26 + 48.27 +static unsigned char *lfb, *lbuf, *text_buf; 48.28 +static const struct font_desc *font; 48.29 +static bool_t vga_compat; 48.30 +static unsigned int pixel_on; 48.31 +static unsigned int xpos, ypos; 48.32 + 48.33 +static unsigned int vram_total; 48.34 +integer_param("vesa-ram", vram_total); 48.35 + 48.36 +static unsigned int vram_remap; 48.37 +integer_param("vesa-map", vram_remap); 48.38 + 48.39 +static int font_height; 48.40 +static void __init parse_font_height(const char *s) 48.41 +{ 48.42 + if ( simple_strtoul(s, &s, 10) == 8 && (*s++ == 'x') ) 48.43 + font_height = simple_strtoul(s, &s, 10); 48.44 + if ( *s != '\0' ) 48.45 + font_height = 0; 48.46 +} 48.47 +custom_param("font", parse_font_height); 48.48 + 48.49 +void __init vesa_early_init(void) 48.50 +{ 48.51 + unsigned int vram_vmode; 48.52 + 48.53 + /* XXX vga_compat = !(boot_video_info.capabilities & 2); */ 48.54 + 48.55 + if ( (vlfb_info.bits_per_pixel < 8) || (vlfb_info.bits_per_pixel > 32) ) 48.56 + return; 48.57 + 48.58 + if ( font_height == 0 ) /* choose a sensible default */ 48.59 + font = ((vlfb_info.height <= 600) ? &font_vga_8x8 : 48.60 + (vlfb_info.height <= 768) ? &font_vga_8x14 : &font_vga_8x16); 48.61 + else if ( font_height <= 8 ) 48.62 + font = &font_vga_8x8; 48.63 + else if ( font_height <= 14 ) 48.64 + font = &font_vga_8x14; 48.65 + else 48.66 + font = &font_vga_8x16; 48.67 + 48.68 + /* vram_vmode -- that is the amount of memory needed for the 48.69 + * used video mode, i.e. the minimum amount of 48.70 + * memory we need. */ 48.71 + vram_vmode = vlfb_info.height * vlfb_info.bytes_per_line; 48.72 + 48.73 + /* vram_total -- all video memory we have. Used for mtrr 48.74 + * entries. */ 48.75 + vram_total = vram_total ? (vram_total << 20) : (vlfb_info.lfb_size << 16); 48.76 + vram_total = max_t(unsigned int, vram_total, vram_vmode); 48.77 + 48.78 + /* vram_remap -- the amount of video memory we are going to 48.79 + * use for vesafb. With modern cards it is no 48.80 + * option to simply use vram_total as that 48.81 + * wastes plenty of kernel address space. */ 48.82 + vram_remap = (vram_remap ? 48.83 + (vram_remap << 20) : 48.84 + ((vram_vmode + (1 << L2_PAGETABLE_SHIFT) - 1) & 48.85 + ~((1 << L2_PAGETABLE_SHIFT) - 1))); 48.86 + vram_remap = max_t(unsigned int, vram_remap, vram_vmode); 48.87 + vram_remap = min_t(unsigned int, vram_remap, vram_total); 48.88 +} 48.89 + 48.90 +void __init vesa_init(void) 48.91 +{ 48.92 + if ( !font ) 48.93 + goto fail; 48.94 + 48.95 + lbuf = xmalloc_bytes(vlfb_info.bytes_per_line); 48.96 + if ( !lbuf ) 48.97 + goto fail; 48.98 + 48.99 + text_buf = xmalloc_bytes(text_columns * text_rows); 48.100 + if ( !text_buf ) 48.101 + goto fail; 48.102 + 48.103 + if ( map_pages_to_xen(IOREMAP_VIRT_START, 48.104 + vlfb_info.lfb_base >> PAGE_SHIFT, 48.105 + vram_remap >> PAGE_SHIFT, 48.106 + PAGE_HYPERVISOR_NOCACHE) ) 48.107 + goto fail; 48.108 + 48.109 + lfb = memset((void *)IOREMAP_VIRT_START, 0, vram_remap); 48.110 + memset(text_buf, 0, text_columns * text_rows); 48.111 + 48.112 + vga_puts = vesa_redraw_puts; 48.113 + 48.114 + printk(XENLOG_INFO "vesafb: framebuffer at 0x%x, mapped to 0x%p, " 48.115 + "using %uk, total %uk\n", 48.116 + vlfb_info.lfb_base, lfb, 48.117 + vram_remap >> 10, vram_total >> 10); 48.118 + printk(XENLOG_INFO "vesafb: mode is %dx%dx%u, linelength=%d, font %ux%u\n", 48.119 + vlfb_info.width, vlfb_info.height, 48.120 + vlfb_info.bits_per_pixel, vlfb_info.bytes_per_line, 48.121 + font->width, font->height); 48.122 + printk(XENLOG_INFO "vesafb: %scolor: size=%d:%d:%d:%d, " 48.123 + "shift=%d:%d:%d:%d\n", 48.124 + vlfb_info.bits_per_pixel > 8 ? "True" : 48.125 + vga_compat ? "Pseudo" : "Static Pseudo", 48.126 + vlfb_info.rsvd_size, vlfb_info.red_size, 48.127 + vlfb_info.green_size, vlfb_info.blue_size, 48.128 + vlfb_info.rsvd_pos, vlfb_info.red_pos, 48.129 + vlfb_info.green_pos, vlfb_info.blue_pos); 48.130 + 48.131 + if ( vlfb_info.bits_per_pixel > 8 ) 48.132 + { 48.133 + /* Light grey in truecolor. */ 48.134 + unsigned int grey = 0xaaaaaaaa; 48.135 + pixel_on = 48.136 + ((grey >> (32 - vlfb_info. red_size)) << vlfb_info. red_pos) | 48.137 + ((grey >> (32 - vlfb_info.green_size)) << vlfb_info.green_pos) | 48.138 + ((grey >> (32 - vlfb_info. blue_size)) << vlfb_info. blue_pos); 48.139 + } 48.140 + else 48.141 + { 48.142 + /* White(ish) in default pseudocolor palette. */ 48.143 + pixel_on = 7; 48.144 + } 48.145 + 48.146 + return; 48.147 + 48.148 + fail: 48.149 + xfree(lbuf); 48.150 + xfree(text_buf); 48.151 +} 48.152 + 48.153 +void __init vesa_endboot(void) 48.154 +{ 48.155 + xpos = 0; 48.156 + vga_puts = vesa_scroll_puts; 48.157 +} 48.158 + 48.159 +#if defined(CONFIG_X86) 48.160 + 48.161 +#include <asm/mtrr.h> 48.162 + 48.163 +static unsigned int vesa_mtrr; 48.164 +integer_param("vesa-mtrr", vesa_mtrr); 48.165 + 48.166 +void __init vesa_mtrr_init(void) 48.167 +{ 48.168 + static const int mtrr_types[] = { 48.169 + 0, MTRR_TYPE_UNCACHABLE, MTRR_TYPE_WRBACK, 48.170 + MTRR_TYPE_WRCOMB, MTRR_TYPE_WRTHROUGH }; 48.171 + unsigned int size_total; 48.172 + int rc, type; 48.173 + 48.174 + if ( !lfb || (vesa_mtrr == 0) || (vesa_mtrr >= ARRAY_SIZE(mtrr_types)) ) 48.175 + return; 48.176 + 48.177 + type = mtrr_types[vesa_mtrr]; 48.178 + if ( !type ) 48.179 + return; 48.180 + 48.181 + /* Find the largest power-of-two */ 48.182 + size_total = vram_total; 48.183 + while ( size_total & (size_total - 1) ) 48.184 + size_total &= size_total - 1; 48.185 + 48.186 + /* Try and find a power of two to add */ 48.187 + do { 48.188 + rc = mtrr_add(vlfb_info.lfb_base, size_total, type, 1); 48.189 + size_total >>= 1; 48.190 + } while ( (size_total >= PAGE_SIZE) && (rc == -EINVAL) ); 48.191 +} 48.192 + 48.193 +static void lfb_flush(void) 48.194 +{ 48.195 + if ( vesa_mtrr == 3 ) 48.196 + __asm__ __volatile__ ("sfence" : : : "memory"); 48.197 +} 48.198 + 48.199 +#else /* !defined(CONFIG_X86) */ 48.200 + 48.201 +#define lfb_flush() ((void)0) 48.202 + 48.203 +#endif 48.204 + 48.205 +/* Render one line of text to given linear framebuffer line. */ 48.206 +static void vesa_show_line( 48.207 + const unsigned char *text_line, 48.208 + unsigned char *video_line, 48.209 + unsigned int nr_chars) 48.210 +{ 48.211 + unsigned int i, j, b, bpp, pixel; 48.212 + 48.213 + bpp = (vlfb_info.bits_per_pixel + 7) >> 3; 48.214 + 48.215 + for ( i = 0; i < font->height; i++ ) 48.216 + { 48.217 + unsigned char *ptr = lbuf; 48.218 + 48.219 + for ( j = 0; j < nr_chars; j++ ) 48.220 + { 48.221 + const unsigned char *bits = font->data; 48.222 + bits += ((text_line[j] * font->height + i) * 48.223 + ((font->width + 7) >> 3)); 48.224 + for ( b = font->width; b--; ) 48.225 + { 48.226 + pixel = test_bit(b, bits) ? pixel_on : 0; 48.227 + memcpy(ptr, &pixel, bpp); 48.228 + ptr += bpp; 48.229 + } 48.230 + } 48.231 + 48.232 + memset(ptr, 0, (vlfb_info.width - nr_chars * font->width) * bpp); 48.233 + memcpy(video_line, lbuf, vlfb_info.width * bpp); 48.234 + video_line += vlfb_info.bytes_per_line; 48.235 + } 48.236 +} 48.237 + 48.238 +/* Fast mode which redraws all modified parts of a 2D text buffer. */ 48.239 +static void vesa_redraw_puts(const char *s) 48.240 +{ 48.241 + unsigned int i, min_redraw_y = ypos; 48.242 + char c; 48.243 + 48.244 + /* Paste characters into text buffer. */ 48.245 + while ( (c = *s++) != '\0' ) 48.246 + { 48.247 + if ( (c == '\n') || (xpos >= text_columns) ) 48.248 + { 48.249 + if ( ++ypos >= text_rows ) 48.250 + { 48.251 + min_redraw_y = 0; 48.252 + ypos = text_rows - 1; 48.253 + memmove(text_buf, text_buf + text_columns, 48.254 + ypos * text_columns); 48.255 + memset(text_buf + ypos * text_columns, 0, xpos); 48.256 + } 48.257 + xpos = 0; 48.258 + } 48.259 + 48.260 + if ( c != '\n' ) 48.261 + text_buf[xpos++ + ypos * text_columns] = c; 48.262 + } 48.263 + 48.264 + /* Render modified section of text buffer to VESA linear framebuffer. */ 48.265 + for ( i = min_redraw_y; i <= ypos; i++ ) 48.266 + vesa_show_line(text_buf + i * text_columns, 48.267 + lfb + i * font->height * vlfb_info.bytes_per_line, 48.268 + text_columns); 48.269 + 48.270 + lfb_flush(); 48.271 +} 48.272 + 48.273 +/* Slower line-based scroll mode which interacts better with dom0. */ 48.274 +static void vesa_scroll_puts(const char *s) 48.275 +{ 48.276 + unsigned int i; 48.277 + char c; 48.278 + 48.279 + while ( (c = *s++) != '\0' ) 48.280 + { 48.281 + if ( (c == '\n') || (xpos >= text_columns) ) 48.282 + { 48.283 + unsigned int bytes = (vlfb_info.width * 48.284 + ((vlfb_info.bits_per_pixel + 7) >> 3)); 48.285 + unsigned char *src = lfb + font->height * vlfb_info.bytes_per_line; 48.286 + unsigned char *dst = lfb; 48.287 + 48.288 + /* New line: scroll all previous rows up one line. */ 48.289 + for ( i = font->height; i < vlfb_info.height; i++ ) 48.290 + { 48.291 + memcpy(dst, src, bytes); 48.292 + src += vlfb_info.bytes_per_line; 48.293 + dst += vlfb_info.bytes_per_line; 48.294 + } 48.295 + 48.296 + /* Render new line. */ 48.297 + vesa_show_line( 48.298 + text_buf, 48.299 + lfb + (text_rows-1) * font->height * vlfb_info.bytes_per_line, 48.300 + xpos); 48.301 + 48.302 + xpos = 0; 48.303 + } 48.304 + 48.305 + if ( c != '\n' ) 48.306 + text_buf[xpos++] = c; 48.307 + } 48.308 + 48.309 + lfb_flush(); 48.310 +}
49.1 --- a/xen/drivers/video/vga.c Thu Aug 16 10:03:26 2007 -0600 49.2 +++ b/xen/drivers/video/vga.c Thu Aug 16 10:47:33 2007 -0600 49.3 @@ -10,22 +10,20 @@ 49.4 #include <xen/lib.h> 49.5 #include <xen/mm.h> 49.6 #include <xen/errno.h> 49.7 -#include <xen/event.h> 49.8 -#include <xen/spinlock.h> 49.9 #include <xen/console.h> 49.10 #include <xen/vga.h> 49.11 #include <asm/io.h> 49.12 -#include "font.h" 49.13 49.14 /* Filled in by arch boot code. */ 49.15 struct xen_vga_console_info vga_console_info; 49.16 49.17 -static int vgacon_enabled = 0; 49.18 -static int vgacon_keep = 0; 49.19 -/*static const struct font_desc *font;*/ 49.20 +static int vgacon_keep; 49.21 +static unsigned int xpos, ypos; 49.22 +static unsigned char *video; 49.23 49.24 -static int xpos, ypos; 49.25 -static unsigned char *video; 49.26 +static void vga_text_puts(const char *s); 49.27 +static void vga_noop_puts(const char *s) {} 49.28 +void (*vga_puts)(const char *) = vga_noop_puts; 49.29 49.30 /* 49.31 * 'vga=<mode-specifier>[,keep]' where <mode-specifier> is one of: 49.32 @@ -55,10 +53,16 @@ static char opt_vga[30] = ""; 49.33 string_param("vga", opt_vga); 49.34 49.35 /* VGA text-mode definitions. */ 49.36 -#define COLUMNS vga_console_info.u.text_mode_3.columns 49.37 -#define LINES vga_console_info.u.text_mode_3.rows 49.38 +static unsigned int columns, lines; 49.39 #define ATTRIBUTE 7 49.40 -#define VIDEO_SIZE (COLUMNS * LINES * 2) 49.41 + 49.42 +#ifdef CONFIG_X86_64 49.43 +void vesa_early_init(void); 49.44 +void vesa_endboot(void); 49.45 +#else 49.46 +#define vesa_early_init() ((void)0) 49.47 +#define vesa_endboot() ((void)0) 49.48 +#endif 49.49 49.50 void __init vga_init(void) 49.51 { 49.52 @@ -76,77 +80,61 @@ void __init vga_init(void) 49.53 switch ( vga_console_info.video_type ) 49.54 { 49.55 case XEN_VGATYPE_TEXT_MODE_3: 49.56 - if ( memory_is_conventional_ram(0xB8000) ) 49.57 + if ( memory_is_conventional_ram(0xB8000) || 49.58 + ((video = ioremap(0xB8000, 0x8000)) == NULL) ) 49.59 return; 49.60 - video = ioremap(0xB8000, 0x8000); 49.61 - if ( video == NULL ) 49.62 - return; 49.63 - /* Disable cursor. */ 49.64 - outw(0x200a, 0x3d4); 49.65 - memset(video, 0, VIDEO_SIZE); 49.66 + outw(0x200a, 0x3d4); /* disable cursor */ 49.67 + columns = vga_console_info.u.text_mode_3.columns; 49.68 + lines = vga_console_info.u.text_mode_3.rows; 49.69 + memset(video, 0, columns * lines * 2); 49.70 + vga_puts = vga_text_puts; 49.71 break; 49.72 case XEN_VGATYPE_VESA_LFB: 49.73 -#if 0 49.74 - /* XXX Implement me! */ 49.75 - video = ioremap(vga_console_info.u.vesa_lfb.lfb_base, 49.76 - vga_console_info.u.vesa_lfb.lfb_size); 49.77 - if ( video == NULL ) 49.78 - return; 49.79 - memset(video, 0, vga_console_info.u.vesa_lfb.lfb_size); 49.80 + vesa_early_init(); 49.81 break; 49.82 -#else 49.83 - return; 49.84 -#endif 49.85 default: 49.86 memset(&vga_console_info, 0, sizeof(vga_console_info)); 49.87 - return; 49.88 + break; 49.89 } 49.90 - 49.91 - vgacon_enabled = 1; 49.92 } 49.93 49.94 void __init vga_endboot(void) 49.95 { 49.96 - if ( !vgacon_enabled ) 49.97 + if ( vga_puts == vga_noop_puts ) 49.98 return; 49.99 49.100 printk("Xen is %s VGA console.\n", 49.101 vgacon_keep ? "keeping" : "relinquishing"); 49.102 49.103 - vgacon_enabled = vgacon_keep; 49.104 + vesa_endboot(); 49.105 + 49.106 + if ( !vgacon_keep ) 49.107 + vga_puts = vga_noop_puts; 49.108 } 49.109 49.110 - 49.111 -static void put_newline(void) 49.112 +static void vga_text_puts(const char *s) 49.113 { 49.114 - xpos = 0; 49.115 - ypos++; 49.116 - 49.117 - if ( ypos >= LINES ) 49.118 - { 49.119 - ypos = LINES-1; 49.120 - memmove((char*)video, 49.121 - (char*)video + 2*COLUMNS, (LINES-1)*2*COLUMNS); 49.122 - memset((char*)video + (LINES-1)*2*COLUMNS, 0, 2*COLUMNS); 49.123 - } 49.124 -} 49.125 + char c; 49.126 49.127 -void vga_putchar(int c) 49.128 -{ 49.129 - if ( !vgacon_enabled ) 49.130 - return; 49.131 - 49.132 - if ( c == '\n' ) 49.133 + while ( (c = *s++) != '\0' ) 49.134 { 49.135 - put_newline(); 49.136 - } 49.137 - else 49.138 - { 49.139 - if ( xpos >= COLUMNS ) 49.140 - put_newline(); 49.141 - video[(xpos + ypos * COLUMNS) * 2] = c & 0xFF; 49.142 - video[(xpos + ypos * COLUMNS) * 2 + 1] = ATTRIBUTE; 49.143 - ++xpos; 49.144 + if ( (c == '\n') || (xpos >= columns) ) 49.145 + { 49.146 + if ( ++ypos >= lines ) 49.147 + { 49.148 + ypos = lines - 1; 49.149 + memmove(video, video + 2 * columns, ypos * 2 * columns); 49.150 + memset(video + ypos * 2 * columns, 0, 2 * xpos); 49.151 + } 49.152 + xpos = 0; 49.153 + } 49.154 + 49.155 + if ( c != '\n' ) 49.156 + { 49.157 + video[(xpos + ypos * columns) * 2] = c; 49.158 + video[(xpos + ypos * columns) * 2 + 1] = ATTRIBUTE; 49.159 + xpos++; 49.160 + } 49.161 } 49.162 } 49.163
50.1 --- a/xen/include/asm-x86/hvm/hvm.h Thu Aug 16 10:03:26 2007 -0600 50.2 +++ b/xen/include/asm-x86/hvm/hvm.h Thu Aug 16 10:47:33 2007 -0600 50.3 @@ -95,36 +95,27 @@ struct hvm_function_table { 50.4 50.5 /* 50.6 * Examine specifics of the guest state: 50.7 - * 1) determine whether paging is enabled, 50.8 - * 2) determine whether long mode is enabled, 50.9 - * 3) determine whether PAE paging is enabled, 50.10 - * 4) determine whether NX is enabled, 50.11 - * 5) determine whether interrupts are enabled or not, 50.12 - * 6) determine the mode the guest is running in, 50.13 - * 7) return the current guest control-register value 50.14 - * 8) return the current guest segment descriptor base 50.15 - * 9) return the current guest segment descriptor 50.16 + * 1) determine whether interrupts are enabled or not 50.17 + * 2) determine the mode the guest is running in 50.18 + * 3) return the current guest segment descriptor base 50.19 + * 4) return the current guest segment descriptor 50.20 */ 50.21 - int (*paging_enabled)(struct vcpu *v); 50.22 - int (*long_mode_enabled)(struct vcpu *v); 50.23 - int (*pae_enabled)(struct vcpu *v); 50.24 - int (*nx_enabled)(struct vcpu *v); 50.25 int (*interrupts_enabled)(struct vcpu *v, enum hvm_intack); 50.26 int (*guest_x86_mode)(struct vcpu *v); 50.27 - unsigned long (*get_guest_ctrl_reg)(struct vcpu *v, unsigned int num); 50.28 unsigned long (*get_segment_base)(struct vcpu *v, enum x86_segment seg); 50.29 void (*get_segment_register)(struct vcpu *v, enum x86_segment seg, 50.30 struct segment_register *reg); 50.31 50.32 /* 50.33 - * Re-set the value of CR3 that Xen runs on when handling VM exits 50.34 + * Re-set the value of CR3 that Xen runs on when handling VM exits. 50.35 */ 50.36 void (*update_host_cr3)(struct vcpu *v); 50.37 50.38 /* 50.39 - * Called to inform HVM layer that a guest cr3 has changed 50.40 + * Called to inform HVM layer that a guest CRn or EFER has changed. 50.41 */ 50.42 - void (*update_guest_cr3)(struct vcpu *v); 50.43 + void (*update_guest_cr)(struct vcpu *v, unsigned int cr); 50.44 + void (*update_guest_efer)(struct vcpu *v); 50.45 50.46 /* 50.47 * Called to ensure than all guest-specific mappings in a tagged TLB 50.48 @@ -189,41 +180,27 @@ hvm_load_cpu_guest_regs(struct vcpu *v, 50.49 void hvm_set_guest_time(struct vcpu *v, u64 gtime); 50.50 u64 hvm_get_guest_time(struct vcpu *v); 50.51 50.52 -static inline int 50.53 -hvm_paging_enabled(struct vcpu *v) 50.54 -{ 50.55 - return hvm_funcs.paging_enabled(v); 50.56 -} 50.57 +#define hvm_paging_enabled(v) \ 50.58 + (!!((v)->arch.hvm_vcpu.guest_cr[0] & X86_CR0_PG)) 50.59 +#define hvm_pae_enabled(v) \ 50.60 + (hvm_paging_enabled(v) && ((v)->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PAE)) 50.61 +#define hvm_nx_enabled(v) \ 50.62 + (!!((v)->arch.hvm_vcpu.guest_efer & EFER_NX)) 50.63 50.64 #ifdef __x86_64__ 50.65 -static inline int 50.66 -hvm_long_mode_enabled(struct vcpu *v) 50.67 -{ 50.68 - return hvm_funcs.long_mode_enabled(v); 50.69 -} 50.70 +#define hvm_long_mode_enabled(v) \ 50.71 + ((v)->arch.hvm_vcpu.guest_efer & EFER_LMA) 50.72 #else 50.73 #define hvm_long_mode_enabled(v) (v,0) 50.74 #endif 50.75 50.76 static inline int 50.77 -hvm_pae_enabled(struct vcpu *v) 50.78 -{ 50.79 - return hvm_funcs.pae_enabled(v); 50.80 -} 50.81 - 50.82 -static inline int 50.83 hvm_interrupts_enabled(struct vcpu *v, enum hvm_intack type) 50.84 { 50.85 return hvm_funcs.interrupts_enabled(v, type); 50.86 } 50.87 50.88 static inline int 50.89 -hvm_nx_enabled(struct vcpu *v) 50.90 -{ 50.91 - return hvm_funcs.nx_enabled(v); 50.92 -} 50.93 - 50.94 -static inline int 50.95 hvm_guest_x86_mode(struct vcpu *v) 50.96 { 50.97 return hvm_funcs.guest_x86_mode(v); 50.98 @@ -244,7 +221,15 @@ hvm_update_vtpr(struct vcpu *v, unsigned 50.99 hvm_funcs.update_vtpr(v, value); 50.100 } 50.101 50.102 -void hvm_update_guest_cr3(struct vcpu *v, unsigned long guest_cr3); 50.103 +static inline void hvm_update_guest_cr(struct vcpu *v, unsigned int cr) 50.104 +{ 50.105 + hvm_funcs.update_guest_cr(v, cr); 50.106 +} 50.107 + 50.108 +static inline void hvm_update_guest_efer(struct vcpu *v) 50.109 +{ 50.110 + hvm_funcs.update_guest_efer(v); 50.111 +} 50.112 50.113 static inline void 50.114 hvm_flush_guest_tlbs(void) 50.115 @@ -257,12 +242,6 @@ void hvm_hypercall_page_initialise(struc 50.116 void *hypercall_page); 50.117 50.118 static inline unsigned long 50.119 -hvm_get_guest_ctrl_reg(struct vcpu *v, unsigned int num) 50.120 -{ 50.121 - return hvm_funcs.get_guest_ctrl_reg(v, num); 50.122 -} 50.123 - 50.124 -static inline unsigned long 50.125 hvm_get_segment_base(struct vcpu *v, enum x86_segment seg) 50.126 { 50.127 return hvm_funcs.get_segment_base(v, seg); 50.128 @@ -277,7 +256,6 @@ hvm_get_segment_register(struct vcpu *v, 50.129 50.130 void hvm_cpuid(unsigned int input, unsigned int *eax, unsigned int *ebx, 50.131 unsigned int *ecx, unsigned int *edx); 50.132 -void hvm_stts(struct vcpu *v); 50.133 void hvm_migrate_timers(struct vcpu *v); 50.134 void hvm_do_resume(struct vcpu *v); 50.135
51.1 --- a/xen/include/asm-x86/hvm/support.h Thu Aug 16 10:03:26 2007 -0600 51.2 +++ b/xen/include/asm-x86/hvm/support.h Thu Aug 16 10:47:33 2007 -0600 51.3 @@ -234,4 +234,8 @@ int hvm_do_hypercall(struct cpu_user_reg 51.4 void hvm_hlt(unsigned long rflags); 51.5 void hvm_triple_fault(void); 51.6 51.7 +int hvm_set_cr0(unsigned long value); 51.8 +int hvm_set_cr3(unsigned long value); 51.9 +int hvm_set_cr4(unsigned long value); 51.10 + 51.11 #endif /* __ASM_X86_HVM_SUPPORT_H__ */
52.1 --- a/xen/include/asm-x86/hvm/svm/asid.h Thu Aug 16 10:03:26 2007 -0600 52.2 +++ b/xen/include/asm-x86/hvm/svm/asid.h Thu Aug 16 10:47:33 2007 -0600 52.3 @@ -32,20 +32,6 @@ void svm_asid_init_vcpu(struct vcpu *v); 52.4 void svm_asid_inv_asid(struct vcpu *v); 52.5 void svm_asid_inc_generation(void); 52.6 52.7 -/* 52.8 - * ASID related, guest triggered events. 52.9 - */ 52.10 - 52.11 -static inline void svm_asid_g_update_paging(struct vcpu *v) 52.12 -{ 52.13 - svm_asid_inv_asid(v); 52.14 -} 52.15 - 52.16 -static inline void svm_asid_g_mov_to_cr3(struct vcpu *v) 52.17 -{ 52.18 - svm_asid_inv_asid(v); 52.19 -} 52.20 - 52.21 static inline void svm_asid_g_invlpg(struct vcpu *v, unsigned long g_vaddr) 52.22 { 52.23 #if 0
53.1 --- a/xen/include/asm-x86/hvm/svm/vmcb.h Thu Aug 16 10:03:26 2007 -0600 53.2 +++ b/xen/include/asm-x86/hvm/svm/vmcb.h Thu Aug 16 10:47:33 2007 -0600 53.3 @@ -440,11 +440,6 @@ struct arch_svm_struct { 53.4 u32 *msrpm; 53.5 int launch_core; 53.6 bool_t vmcb_in_sync; /* VMCB sync'ed with VMSAVE? */ 53.7 - unsigned long cpu_shadow_cr0; /* Guest value for CR0 */ 53.8 - unsigned long cpu_shadow_cr4; /* Guest value for CR4 */ 53.9 - unsigned long cpu_shadow_efer; /* Guest value for EFER */ 53.10 - unsigned long cpu_cr2; 53.11 - unsigned long cpu_cr3; 53.12 }; 53.13 53.14 struct vmcb_struct *alloc_vmcb(void);
54.1 --- a/xen/include/asm-x86/hvm/vcpu.h Thu Aug 16 10:03:26 2007 -0600 54.2 +++ b/xen/include/asm-x86/hvm/vcpu.h Thu Aug 16 10:47:33 2007 -0600 54.3 @@ -29,7 +29,18 @@ 54.4 #define HVM_VCPU_INIT_SIPI_SIPI_STATE_WAIT_SIPI 1 54.5 54.6 struct hvm_vcpu { 54.7 - unsigned long hw_cr3; /* value we give to HW to use */ 54.8 + /* Guest control-register and EFER values, just as the guest sees them. */ 54.9 + unsigned long guest_cr[5]; 54.10 + unsigned long guest_efer; 54.11 + 54.12 + /* 54.13 + * Processor-visible control-register values, while guest executes. 54.14 + * CR0, CR4: Used as a cache of VMCS contents by VMX only. 54.15 + * CR1, CR2: Never used (guest_cr[2] is always processor-visible CR2). 54.16 + * CR3: Always used and kept up to date by paging subsystem. 54.17 + */ 54.18 + unsigned long hw_cr[5]; 54.19 + 54.20 struct hvm_io_op io_op; 54.21 struct vlapic vlapic; 54.22 s64 cache_tsc_offset;
55.1 --- a/xen/include/asm-x86/hvm/vmx/vmcs.h Thu Aug 16 10:03:26 2007 -0600 55.2 +++ b/xen/include/asm-x86/hvm/vmx/vmcs.h Thu Aug 16 10:47:33 2007 -0600 55.3 @@ -67,17 +67,11 @@ struct arch_vmx_struct { 55.4 /* Cache of cpu execution control. */ 55.5 u32 exec_control; 55.6 55.7 - unsigned long cpu_cr0; /* copy of guest CR0 */ 55.8 - unsigned long cpu_shadow_cr0; /* copy of guest read shadow CR0 */ 55.9 - unsigned long cpu_shadow_cr4; /* copy of guest read shadow CR4 */ 55.10 - unsigned long cpu_cr2; /* save CR2 */ 55.11 - unsigned long cpu_cr3; 55.12 #ifdef __x86_64__ 55.13 struct vmx_msr_state msr_state; 55.14 unsigned long shadow_gs; 55.15 unsigned long cstar; 55.16 #endif 55.17 - unsigned long efer; 55.18 55.19 /* Following fields are all specific to vmxassist. */ 55.20 unsigned long vmxassist_enabled:1;
56.1 --- a/xen/include/asm-x86/hvm/vmx/vmx.h Thu Aug 16 10:03:26 2007 -0600 56.2 +++ b/xen/include/asm-x86/hvm/vmx/vmx.h Thu Aug 16 10:47:33 2007 -0600 56.3 @@ -279,8 +279,8 @@ static inline void __vmx_inject_exceptio 56.4 56.5 __vmwrite(VM_ENTRY_INTR_INFO, intr_fields); 56.6 56.7 - if (trap == TRAP_page_fault) 56.8 - HVMTRACE_2D(PF_INJECT, v, v->arch.hvm_vmx.cpu_cr2, error_code); 56.9 + if ( trap == TRAP_page_fault ) 56.10 + HVMTRACE_2D(PF_INJECT, v, v->arch.hvm_vcpu.guest_cr[2], error_code); 56.11 else 56.12 HVMTRACE_2D(INJ_EXC, v, trap, error_code); 56.13 }
57.1 --- a/xen/include/public/arch-x86/xen-x86_32.h Thu Aug 16 10:03:26 2007 -0600 57.2 +++ b/xen/include/public/arch-x86/xen-x86_32.h Thu Aug 16 10:47:33 2007 -0600 57.3 @@ -64,18 +64,34 @@ 57.4 #define FLAT_USER_DS FLAT_RING3_DS 57.5 #define FLAT_USER_SS FLAT_RING3_SS 57.6 57.7 -/* 57.8 - * Virtual addresses beyond this are not modifiable by guest OSes. The 57.9 - * machine->physical mapping table starts at this address, read-only. 57.10 - */ 57.11 +#define __HYPERVISOR_VIRT_START_PAE 0xF5800000 57.12 +#define __MACH2PHYS_VIRT_START_PAE 0xF5800000 57.13 +#define __MACH2PHYS_VIRT_END_PAE 0xF6800000 57.14 +#define HYPERVISOR_VIRT_START_PAE \ 57.15 + mk_unsigned_long(__HYPERVISOR_VIRT_START_PAE) 57.16 +#define MACH2PHYS_VIRT_START_PAE \ 57.17 + mk_unsigned_long(__MACH2PHYS_VIRT_START_PAE) 57.18 +#define MACH2PHYS_VIRT_END_PAE \ 57.19 + mk_unsigned_long(__MACH2PHYS_VIRT_END_PAE) 57.20 + 57.21 +#define __HYPERVISOR_VIRT_START_NONPAE 0xFC000000 57.22 +#define __MACH2PHYS_VIRT_START_NONPAE 0xFC000000 57.23 +#define __MACH2PHYS_VIRT_END_NONPAE 0xFC400000 57.24 +#define HYPERVISOR_VIRT_START_NONPAE \ 57.25 + mk_unsigned_long(__HYPERVISOR_VIRT_START_NONPAE) 57.26 +#define MACH2PHYS_VIRT_START_NONPAE \ 57.27 + mk_unsigned_long(__MACH2PHYS_VIRT_START_NONPAE) 57.28 +#define MACH2PHYS_VIRT_END_NONPAE \ 57.29 + mk_unsigned_long(__MACH2PHYS_VIRT_END_NONPAE) 57.30 + 57.31 #ifdef CONFIG_X86_PAE 57.32 -#define __HYPERVISOR_VIRT_START 0xF5800000 57.33 -#define __MACH2PHYS_VIRT_START 0xF5800000 57.34 -#define __MACH2PHYS_VIRT_END 0xF6800000 57.35 +#define __HYPERVISOR_VIRT_START __HYPERVISOR_VIRT_START_PAE 57.36 +#define __MACH2PHYS_VIRT_START __MACH2PHYS_VIRT_START_PAE 57.37 +#define __MACH2PHYS_VIRT_END __MACH2PHYS_VIRT_END_PAE 57.38 #else 57.39 -#define __HYPERVISOR_VIRT_START 0xFC000000 57.40 -#define __MACH2PHYS_VIRT_START 0xFC000000 57.41 -#define __MACH2PHYS_VIRT_END 0xFC400000 57.42 +#define __HYPERVISOR_VIRT_START __HYPERVISOR_VIRT_START_NONPAE 57.43 +#define __MACH2PHYS_VIRT_START __MACH2PHYS_VIRT_START_NONPAE 57.44 +#define __MACH2PHYS_VIRT_END __MACH2PHYS_VIRT_END_NONPAE 57.45 #endif 57.46 57.47 #ifndef HYPERVISOR_VIRT_START
58.1 --- a/xen/include/xen/vga.h Thu Aug 16 10:03:26 2007 -0600 58.2 +++ b/xen/include/xen/vga.h Thu Aug 16 10:47:33 2007 -0600 58.3 @@ -15,11 +15,11 @@ 58.4 extern struct xen_vga_console_info vga_console_info; 58.5 void vga_init(void); 58.6 void vga_endboot(void); 58.7 -void vga_putchar(int c); 58.8 +extern void (*vga_puts)(const char *); 58.9 #else 58.10 -#define vga_init() ((void)0) 58.11 -#define vga_endboot() ((void)0) 58.12 -#define vga_putchar(c) ((void)0) 58.13 +#define vga_init() ((void)0) 58.14 +#define vga_endboot() ((void)0) 58.15 +#define vga_puts(s) ((void)0) 58.16 #endif 58.17 58.18 #endif /* _XEN_VGA_H */
59.1 --- a/xen/include/xen/xencomm.h Thu Aug 16 10:03:26 2007 -0600 59.2 +++ b/xen/include/xen/xencomm.h Thu Aug 16 10:47:33 2007 -0600 59.3 @@ -23,13 +23,12 @@ 59.4 59.5 #include <public/xen.h> 59.6 59.7 -extern unsigned long xencomm_copy_to_guest(void *to, const void *from, 59.8 - unsigned int len, unsigned int skip); 59.9 -extern unsigned long xencomm_copy_from_guest(void *to, const void *from, 59.10 - unsigned int len, unsigned int skip); 59.11 -extern int xencomm_add_offset(void **handle, unsigned int bytes); 59.12 -extern int xencomm_handle_is_null(void *ptr); 59.13 - 59.14 +unsigned long xencomm_copy_to_guest( 59.15 + void *to, const void *from, unsigned int len, unsigned int skip); 59.16 +unsigned long xencomm_copy_from_guest( 59.17 + void *to, const void *from, unsigned int len, unsigned int skip); 59.18 +int xencomm_add_offset(void **handle, unsigned int bytes); 59.19 +int xencomm_handle_is_null(void *ptr); 59.20 59.21 static inline int xencomm_is_inline(const void *handle) 59.22 { 59.23 @@ -39,7 +38,7 @@ static inline int xencomm_is_inline(cons 59.24 59.25 static inline unsigned long xencomm_inline_addr(const void *handle) 59.26 { 59.27 - return (unsigned long)handle & ~XENCOMM_INLINE_FLAG; 59.28 + return (unsigned long)handle & ~XENCOMM_INLINE_FLAG; 59.29 } 59.30 59.31 /* Is the guest handle a NULL reference? */