2 * This file is part of the Palacios Virtual Machine Monitor developed
3 * by the V3VEE Project with funding from the United States National
4 * Science Foundation and the Department of Energy.
6 * The V3VEE Project is a joint project between Northwestern University
7 * and the University of New Mexico. You can find out more at
10 * Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu>
11 * Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org>
12 * All rights reserved.
14 * Author: Jack Lange <jarusl@cs.northwestern.edu>
16 * This is free software. You are permitted to use,
17 * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
23 static inline int activate_shadow_pt_32(struct guest_info * info) {
24 struct cr3_32 * shadow_cr3 = (struct cr3_32 *)&(info->ctrl_regs.cr3);
25 struct cr3_32 * guest_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.guest_cr3);
26 struct shadow_page_data * shdw_page = create_new_shadow_pt(info);
28 shdw_page->cr3 = shdw_page->page_pa;
30 shadow_cr3->pdt_base_addr = PAGE_BASE_ADDR_4KB(shdw_page->page_pa);
31 PrintDebug( "Created new shadow page table %p\n", (void *)BASE_TO_PAGE_ADDR(shadow_cr3->pdt_base_addr));
33 shadow_cr3->pwt = guest_cr3->pwt;
34 shadow_cr3->pcd = guest_cr3->pcd;
36 #ifdef CONFIG_SYMBIOTIC_SWAP
49 * * 32 bit Page table fault handlers
53 static int handle_4MB_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
54 pte32_t * shadow_pt, pde32_4MB_t * large_guest_pde);
56 static int handle_pte_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
57 pte32_t * shadow_pt, pte32_t * guest_pt);
60 static inline int handle_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
61 pde32_t * guest_pd = NULL;
62 pde32_t * shadow_pd = CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
63 addr_t guest_cr3 = CR3_TO_PDE32_PA(info->shdw_pg_state.guest_cr3);
64 pt_access_status_t guest_pde_access;
65 pt_access_status_t shadow_pde_access;
66 pde32_t * guest_pde = NULL;
67 pde32_t * shadow_pde = (pde32_t *)&(shadow_pd[PDE32_INDEX(fault_addr)]);
69 PrintDebug("Shadow page fault handler: %p\n", (void*) fault_addr );
70 PrintDebug("Handling PDE32 Fault\n");
72 if (guest_pa_to_host_va(info, guest_cr3, (addr_t*)&guest_pd) == -1) {
73 PrintError("Invalid Guest PDE Address: 0x%p\n", (void *)guest_cr3);
77 guest_pde = (pde32_t *)&(guest_pd[PDE32_INDEX(fault_addr)]);
80 // Check the guest page permissions
81 guest_pde_access = v3_can_access_pde32(guest_pd, fault_addr, error_code);
83 // Check the shadow page permissions
84 shadow_pde_access = v3_can_access_pde32(shadow_pd, fault_addr, error_code);
86 /* Was the page fault caused by the Guest's page tables? */
87 if (is_guest_pf(guest_pde_access, shadow_pde_access) == 1) {
88 PrintDebug("Injecting PDE pf to guest: (guest access error=%d) (shdw access error=%d) (pf error code=%d)\n",
89 *(uint_t *)&guest_pde_access, *(uint_t *)&shadow_pde_access, *(uint_t *)&error_code);
90 if (inject_guest_pf(info, fault_addr, error_code) == -1) {
91 PrintError("Could not inject guest page fault\n");
99 if (shadow_pde_access == PT_ACCESS_USER_ERROR) {
101 // PDE Entry marked non user
103 PrintDebug("Shadow Paging User access error (shadow_pde_access=0x%x, guest_pde_access=0x%x)\n",
104 shadow_pde_access, guest_pde_access);
106 if (inject_guest_pf(info, fault_addr, error_code) == -1) {
107 PrintError("Could not inject guest page fault\n");
111 } else if ((shadow_pde_access == PT_ACCESS_WRITE_ERROR) &&
112 (guest_pde->large_page == 1)) {
114 ((pde32_4MB_t *)guest_pde)->dirty = 1;
115 shadow_pde->writable = guest_pde->writable;
117 } else if ((shadow_pde_access != PT_ACCESS_NOT_PRESENT) &&
118 (shadow_pde_access != PT_ACCESS_OK)) {
119 // inject page fault in guest
120 if (inject_guest_pf(info, fault_addr, error_code) == -1) {
121 PrintError("Could not inject guest page fault\n");
124 PrintDebug("Unknown Error occurred (shadow_pde_access=%d)\n", shadow_pde_access);
125 PrintDebug("Manual Says to inject page fault into guest\n");
130 pte32_t * shadow_pt = NULL;
131 pte32_t * guest_pt = NULL;
133 // Get the next shadow page level, allocate if not present
135 if (shadow_pde_access == PT_ACCESS_NOT_PRESENT) {
136 struct shadow_page_data * shdw_page = create_new_shadow_pt(info);
137 shadow_pt = (pte32_t *)V3_VAddr((void *)shdw_page->page_pa);
139 shadow_pde->present = 1;
140 shadow_pde->user_page = guest_pde->user_page;
143 if (guest_pde->large_page == 0) {
144 shadow_pde->writable = guest_pde->writable;
146 // This large page flag is temporary until we can get a working cache....
147 ((pde32_4MB_t *)guest_pde)->vmm_info = V3_LARGE_PG;
149 if (error_code.write) {
150 shadow_pde->writable = guest_pde->writable;
151 ((pde32_4MB_t *)guest_pde)->dirty = 1;
153 shadow_pde->writable = 0;
154 ((pde32_4MB_t *)guest_pde)->dirty = 0;
159 // VMM Specific options
160 shadow_pde->write_through = guest_pde->write_through;
161 shadow_pde->cache_disable = guest_pde->cache_disable;
162 shadow_pde->global_page = guest_pde->global_page;
165 guest_pde->accessed = 1;
170 shadow_pde->pt_base_addr = PAGE_BASE_ADDR(shdw_page->page_pa);
172 shadow_pt = (pte32_t *)V3_VAddr((void *)BASE_TO_PAGE_ADDR(shadow_pde->pt_base_addr));
177 if (guest_pde->large_page == 0) {
178 if (guest_pa_to_host_va(info, BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr), (addr_t*)&guest_pt) == -1) {
179 // Machine check the guest
180 PrintDebug("Invalid Guest PTE Address: 0x%p\n", (void *)BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr));
181 v3_raise_exception(info, MC_EXCEPTION);
185 if (handle_pte_shadow_pagefault_32(info, fault_addr, error_code, shadow_pt, guest_pt) == -1) {
186 PrintError("Error handling Page fault caused by PTE\n");
190 if (handle_4MB_shadow_pagefault_32(info, fault_addr, error_code, shadow_pt, (pde32_4MB_t *)guest_pde) == -1) {
191 PrintError("Error handling large pagefault\n");
201 static int handle_pte_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
202 pte32_t * shadow_pt, pte32_t * guest_pt) {
204 pt_access_status_t guest_pte_access;
205 pt_access_status_t shadow_pte_access;
206 pte32_t * guest_pte = (pte32_t *)&(guest_pt[PTE32_INDEX(fault_addr)]);;
207 pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(fault_addr)]);
208 addr_t guest_pa = BASE_TO_PAGE_ADDR((addr_t)(guest_pte->page_base_addr)) + PAGE_OFFSET(fault_addr);
210 struct v3_shadow_region * shdw_reg = v3_get_shadow_region(info, guest_pa);
212 if (shdw_reg == NULL) {
213 // Inject a machine check in the guest
214 PrintDebug("Invalid Guest Address in page table (0x%p)\n", (void *)guest_pa);
215 v3_raise_exception(info, MC_EXCEPTION);
219 // Check the guest page permissions
220 guest_pte_access = v3_can_access_pte32(guest_pt, fault_addr, error_code);
222 // Check the shadow page permissions
223 shadow_pte_access = v3_can_access_pte32(shadow_pt, fault_addr, error_code);
226 /* Was the page fault caused by the Guest's page tables? */
227 if (is_guest_pf(guest_pte_access, shadow_pte_access) == 1) {
229 PrintDebug("Access error injecting pf to guest (guest access error=%d) (pf error code=%d)\n",
230 guest_pte_access, *(uint_t*)&error_code);
232 #ifdef CONFIG_SYMBIOTIC_SWAP
233 if (is_swapped_pte32(guest_pte)) {
235 #ifdef CONFIG_SYMBIOTIC_SWAP_TELEMETRY
236 if (error_code.write == 0) {
237 info->swap_state.read_faults++;
239 info->swap_state.write_faults++;
242 if (error_code.write == 0) {
243 V3_Print("Page fault on swapped out page (vaddr=%p) (pte=%x) (error_code=%x)\n",
244 (void *)fault_addr, *(uint32_t *)guest_pte, *(uint32_t *)&error_code);
246 addr_t swp_pg_addr = v3_get_swapped_pg_addr(info, shadow_pte, guest_pte);
248 if (swp_pg_addr == 0) {
249 if (inject_guest_pf(info, fault_addr, error_code) == -1) {
250 PrintError("Could not inject guest page fault\n");
255 * Setup shadow paging state
258 /* We need some way to check permissions.... */
260 shadow_pte->accessed = 1;
261 shadow_pte->writable = 0;
263 if ((fault_addr & 0xc0000000) == 0xc0000000) {
264 shadow_pte->user_page = 0;
266 shadow_pte->user_page = 1;
269 shadow_pte->write_through = 0;
270 shadow_pte->cache_disable = 0;
271 shadow_pte->global_page = 0;
273 shadow_pte->present = 1;
275 shadow_pte->page_base_addr = swp_pg_addr;
284 if (inject_guest_pf(info, fault_addr, error_code) == -1) {
285 PrintError("Could not inject guest page fault\n");
294 if (shadow_pte_access == PT_ACCESS_OK) {
295 // Inconsistent state...
296 // Guest Re-Entry will flush page tables and everything should now work
297 PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
302 if (shadow_pte_access == PT_ACCESS_NOT_PRESENT) {
303 // Page Table Entry Not Present
304 PrintDebug("guest_pa =%p\n", (void *)guest_pa);
306 if ((shdw_reg->host_type == SHDW_REGION_ALLOCATED) ||
307 (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK)) {
308 addr_t shadow_pa = v3_get_shadow_addr(shdw_reg, guest_pa);
310 shadow_pte->page_base_addr = PAGE_BASE_ADDR(shadow_pa);
312 PrintDebug("\tMapping shadow page (%p)\n", (void *)BASE_TO_PAGE_ADDR(shadow_pte->page_base_addr));
314 shadow_pte->present = guest_pte->present;
315 shadow_pte->user_page = guest_pte->user_page;
317 //set according to VMM policy
318 shadow_pte->write_through = guest_pte->write_through;
319 shadow_pte->cache_disable = guest_pte->cache_disable;
320 shadow_pte->global_page = guest_pte->global_page;
323 guest_pte->accessed = 1;
325 if (guest_pte->dirty == 1) {
326 shadow_pte->writable = guest_pte->writable;
327 } else if ((guest_pte->dirty == 0) && (error_code.write == 1)) {
328 shadow_pte->writable = guest_pte->writable;
329 guest_pte->dirty = 1;
330 } else if ((guest_pte->dirty == 0) && (error_code.write == 0)) {
331 shadow_pte->writable = 0;
336 // Write hooks trump all, and are set Read Only
337 if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
338 shadow_pte->writable = 0;
342 // Page fault handled by hook functions
344 if (v3_handle_mem_full_hook(info, fault_addr, guest_pa, shdw_reg, error_code) == -1) {
345 PrintError("Special Page fault handler returned error for address: %p\n", (void *)fault_addr);
349 } else if (shadow_pte_access == PT_ACCESS_WRITE_ERROR) {
350 guest_pte->dirty = 1;
352 if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
353 if (v3_handle_mem_wr_hook(info, fault_addr, guest_pa, shdw_reg, error_code) == -1) {
354 PrintError("Special Page fault handler returned error for address: %p\n", (void *)fault_addr);
358 PrintDebug("Shadow PTE Write Error\n");
359 shadow_pte->writable = guest_pte->writable;
366 // Inject page fault into the guest
367 if (inject_guest_pf(info, fault_addr, error_code) == -1) {
368 PrintError("Could not inject guest page fault\n");
372 PrintError("PTE Page fault fell through... Not sure if this should ever happen\n");
373 PrintError("Manual Says to inject page fault into guest\n");
382 static int handle_4MB_shadow_pagefault_32(struct guest_info * info,
383 addr_t fault_addr, pf_error_t error_code,
384 pte32_t * shadow_pt, pde32_4MB_t * large_guest_pde)
386 pt_access_status_t shadow_pte_access = v3_can_access_pte32(shadow_pt, fault_addr, error_code);
387 pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(fault_addr)]);
388 addr_t guest_fault_pa = BASE_TO_PAGE_ADDR_4MB(large_guest_pde->page_base_addr) + PAGE_OFFSET_4MB(fault_addr);
391 PrintDebug("Handling 4MB fault (guest_fault_pa=%p) (error_code=%x)\n", (void *)guest_fault_pa, *(uint_t*)&error_code);
392 PrintDebug("ShadowPT=%p, LargeGuestPDE=%p\n", shadow_pt, large_guest_pde);
394 struct v3_shadow_region * shdw_reg = v3_get_shadow_region(info, guest_fault_pa);
397 if (shdw_reg == NULL) {
398 // Inject a machine check in the guest
399 PrintDebug("Invalid Guest Address in page table (0x%p)\n", (void *)guest_fault_pa);
400 v3_raise_exception(info, MC_EXCEPTION);
404 if (shadow_pte_access == PT_ACCESS_OK) {
405 // Inconsistent state...
406 // Guest Re-Entry will flush tables and everything should now workd
407 PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
412 if (shadow_pte_access == PT_ACCESS_NOT_PRESENT) {
413 // Get the guest physical address of the fault
415 if ((shdw_reg->host_type == SHDW_REGION_ALLOCATED) ||
416 (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK)) {
417 addr_t shadow_pa = v3_get_shadow_addr(shdw_reg, guest_fault_pa);
419 shadow_pte->page_base_addr = PAGE_BASE_ADDR(shadow_pa);
421 PrintDebug("\tMapping shadow page (%p)\n", (void *)BASE_TO_PAGE_ADDR(shadow_pte->page_base_addr));
423 shadow_pte->present = 1;
425 /* We are assuming that the PDE entry has precedence
426 * so the Shadow PDE will mirror the guest PDE settings,
427 * and we don't have to worry about them here
430 shadow_pte->user_page = 1;
432 if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
433 shadow_pte->writable = 0;
435 shadow_pte->writable = 1;
438 //set according to VMM policy
439 shadow_pte->write_through = large_guest_pde->write_through;
440 shadow_pte->cache_disable = large_guest_pde->cache_disable;
441 shadow_pte->global_page = large_guest_pde->global_page;
445 if (v3_handle_mem_full_hook(info, fault_addr, guest_fault_pa, shdw_reg, error_code) == -1) {
446 PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
450 } else if (shadow_pte_access == PT_ACCESS_WRITE_ERROR) {
452 if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
454 if (v3_handle_mem_wr_hook(info, fault_addr, guest_fault_pa, shdw_reg, error_code) == -1) {
455 PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
461 PrintError("Error in large page fault handler...\n");
462 PrintError("This case should have been handled at the top level handler\n");
466 PrintDebug("Returning from large page fault handler\n");
481 /* If we start to optimize we should look up the guest pages in the cache... */
482 static inline int handle_shadow_invlpg_32(struct guest_info * info, addr_t vaddr) {
483 pde32_t * shadow_pd = (pde32_t *)CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
484 pde32_t * shadow_pde = (pde32_t *)&shadow_pd[PDE32_INDEX(vaddr)];
486 addr_t guest_cr3 = CR3_TO_PDE32_PA(info->shdw_pg_state.guest_cr3);
487 pde32_t * guest_pd = NULL;
490 if (guest_pa_to_host_va(info, guest_cr3, (addr_t*)&guest_pd) == -1) {
491 PrintError("Invalid Guest PDE Address: 0x%p\n", (void *)guest_cr3);
495 guest_pde = (pde32_t *)&(guest_pd[PDE32_INDEX(vaddr)]);
497 if (guest_pde->large_page == 1) {
498 shadow_pde->present = 0;
499 PrintDebug("Invalidating Large Page\n");
500 } else if (shadow_pde->present == 1) {
501 pte32_t * shadow_pt = (pte32_t *)(addr_t)BASE_TO_PAGE_ADDR_4KB(shadow_pde->pt_base_addr);
502 pte32_t * shadow_pte = (pte32_t *) V3_VAddr( (void*) &shadow_pt[PTE32_INDEX(vaddr)] );
504 PrintDebug("Setting not present\n");
506 shadow_pte->present = 0;