Palacios Public Git Repository

To checkout Palacios execute

  git clone http://v3vee.org/palacios/palacios.web/palacios.git
This will give you the master branch. You probably want the devel branch or one of the release branches. To switch to the devel branch, simply execute
  cd palacios
  git checkout --track -b devel origin/devel
The other branches are similar.


Added initial cut at support with using large pages in shadow
Patrick Bridges [Tue, 10 Aug 2010 20:53:06 +0000 (14:53 -0600)]
page tables to back large pages in the guest, assuming the
guest memory is appropraitely aligned.

palacios/src/palacios/mmu/vmm_shdw_pg_tlb_64.h

index 34102ac..db59d7d 100644 (file)
@@ -52,8 +52,11 @@ static inline int activate_shadow_pt_64(struct guest_info * info) {
  * *
  */
 
-static int handle_2MB_shadow_pagefault_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
-                                         pte64_t * shadow_pt, pde64_2MB_t * large_guest_pde);
+static int handle_2MB_shadow_pagefault_pde_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
+                                             pt_access_status_t shadow_pde_access, pde64_2MB_t * shadow_pt, 
+                                             pde64_2MB_t * large_guest_pde);
+static int handle_2MB_shadow_pagefault_pte_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
+                                             pte64_t * shadow_pt, pde64_2MB_t * large_guest_pde);
 
 static int handle_pte_shadow_pagefault_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code,
                                          pte64_t * shadow_pt, pte64_t * guest_pt);
@@ -329,12 +332,22 @@ static int handle_pde_shadow_pagefault_64(struct guest_info * info, addr_t fault
        return 0;
     }
 
+    // Handle as a shadow large page if possible
+    if (guest_pde->large_page 
+       && (info->vm_info->mem_align >= PAGE_SIZE_2MB)) {
+       if (handle_2MB_shadow_pagefault_pde_64(info, fault_addr, error_code, shadow_pde_access,
+                                              (pde64_2MB_t *)shadow_pde, (pde64_2MB_t *)guest_pde) == -1) {
+           PrintError("Error handling large pagefault with large page\n");
+           return -1;
+       } else {
+           return 0;
+       }
+    } 
 
     pte64_t * shadow_pt = NULL;
     pte64_t * guest_pt = NULL;
 
-    // Get the next shadow page level, allocate if not present
-
+    // get the next shadow page level, allocate if not present
     if (shadow_pde_access == PT_ACCESS_NOT_PRESENT) {
        struct shadow_page_data * shdw_page = create_new_shadow_pt(info);
        shadow_pt = (pte64_t *)V3_VAddr((void *)shdw_page->page_pa);
@@ -387,8 +400,8 @@ static int handle_pde_shadow_pagefault_64(struct guest_info * info, addr_t fault
            return -1;
        }
     } else {
-       if (handle_2MB_shadow_pagefault_64(info, fault_addr, error_code, shadow_pt, (pde64_2MB_t *)guest_pde) == -1) {
-           PrintError("Error handling large pagefault\n");
+       if (handle_2MB_shadow_pagefault_pte_64(info, fault_addr, error_code, shadow_pt, (pde64_2MB_t *)guest_pde) == -1) {
+           PrintError("Error handling large pagefault with small page\n");
            return -1;
        } 
     }
@@ -529,10 +542,90 @@ static int handle_pte_shadow_pagefault_64(struct guest_info * info, addr_t fault
 }
 
 
+static int handle_2MB_shadow_pagefault_pde_64(struct guest_info * info, 
+                                             addr_t fault_addr, pf_error_t error_code, 
+                                             pt_access_status_t shadow_pde_access,
+                                             pde64_2MB_t * large_shadow_pde, pde64_2MB_t * large_guest_pde) 
+{
+    addr_t guest_fault_pa = BASE_TO_PAGE_ADDR_2MB(large_guest_pde->page_base_addr) + PAGE_OFFSET_2MB(fault_addr);
+    //  struct shadow_page_state * state = &(info->shdw_pg_state);
+
+    PrintDebug("Handling 2MB fault with large page (guest_fault_pa=%p) (error_code=%x)\n", (void *)guest_fault_pa, *(uint_t*)&error_code);
+    PrintDebug("LargeShadowPDE=%p, LargeGuestPDE=%p\n", large_shadow_pde, large_guest_pde);
+
+    struct v3_mem_region * shdw_reg = v3_get_mem_region(info->vm_info, info->cpu_id, guest_fault_pa);
+    if (shdw_reg == NULL) {
+       // Inject a machine check in the guest
+       PrintError("Invalid Guest Address in page table (0x%p)\n", (void *)guest_fault_pa);
+       v3_raise_exception(info, MC_EXCEPTION);
+       return 0;
+    }
+
+    if (shadow_pde_access == PT_ACCESS_OK) {
+       // Inconsistent state...
+       // Guest Re-Entry will flush tables and everything should now workd
+       PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
+       //PrintHostPageTree(info, fault_addr, info->ctrl_regs.cr3);
+       return 0;
+    }
+
+  
+    if (shadow_pde_access == PT_ACCESS_NOT_PRESENT) {
+       // Get the guest physical address of the fault
+
+       if ((shdw_reg->flags.alloced == 1) || 
+           (shdw_reg->flags.read == 1)) {
+           addr_t shadow_pa = 0;
+
+           if (v3_gpa_to_hpa(info, guest_fault_pa, &shadow_pa) == -1) {
+               PrintError("could not translate page fault address (%p)\n", (void *)guest_fault_pa);
+               return -1;
+           }
+
+           large_guest_pde->vmm_info = V3_LARGE_PG; /* For invalidations */
+           large_shadow_pde->page_base_addr = PAGE_BASE_ADDR_2MB(shadow_pa);
+           large_shadow_pde->large_page = 1;
+           large_shadow_pde->present = 1;
+           large_shadow_pde->user_page = 1;
+
+           if (shdw_reg->flags.write == 0) {
+               large_shadow_pde->writable = 0;
+           } else {
+               large_shadow_pde->writable = 1;
+           }
+
+           //set according to VMM policy
+           large_shadow_pde->write_through = large_guest_pde->write_through;
+           large_shadow_pde->cache_disable = large_guest_pde->cache_disable;
+           large_shadow_pde->global_page = large_guest_pde->global_page;
+           //
+      
+       } else {
+           if (shdw_reg->unhandled(info, fault_addr, guest_fault_pa, shdw_reg, error_code) == -1) {
+               PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
+               return -1;
+           }
+       }
+    } else if (shadow_pde_access == PT_ACCESS_WRITE_ERROR) {
+       if (shdw_reg->unhandled(info, fault_addr, guest_fault_pa, shdw_reg, error_code) == -1) {
+           PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
+           return -1;
+       }
+    } else {
+       PrintError("Error in large page fault handler...\n");
+       PrintError("This case should have been handled at the top level handler\n");
+       return -1;
+    }
+
+    //  PrintHostPageTree(info, fault_addr, info->ctrl_regs.cr3);
+    PrintDebug("Returning from large page->large page fault handler\n");
+    return 0;
+}
 
-static int handle_2MB_shadow_pagefault_64(struct guest_info * info, 
-                                         addr_t fault_addr, pf_error_t error_code, 
-                                         pte64_t * shadow_pt, pde64_2MB_t * large_guest_pde) 
+static int handle_2MB_shadow_pagefault_pte_64(struct guest_info * info, 
+                                             addr_t fault_addr, pf_error_t error_code, 
+                                             pte64_t * shadow_pt, pde64_2MB_t * large_guest_pde) 
 {
     pt_access_status_t shadow_pte_access = v3_can_access_pte64(shadow_pt, fault_addr, error_code);
     pte64_t * shadow_pte = (pte64_t *)&(shadow_pt[PTE64_INDEX(fault_addr)]);
@@ -614,7 +707,7 @@ static int handle_2MB_shadow_pagefault_64(struct guest_info * info,
     }
 
     //  PrintHostPageTree(info, fault_addr, info->ctrl_regs.cr3);
-    PrintDebug("Returning from large page fault handler\n");
+    PrintDebug("Returning from large page->small page fault handler\n");
     return 0;
 }