Use page allocation for base memory region array

[palacios.git] / palacios / src / palacios / vmm_direct_paging_64.h

diff --git a/palacios/src/palacios/vmm_direct_paging_64.h b/palacios/src/palacios/vmm_direct_paging_64.h
index 217373a..6afb9fb 100644 (file)
--- a/palacios/src/palacios/vmm_direct_paging_64.h
+++ b/palacios/src/palacios/vmm_direct_paging_64.h
@@ -27,9 +27,12 @@
 #include <palacios/vm_guest_mem.h>
 #include <palacios/vm_guest.h>
 
+/* this always builds 4 level page tables, but large pages are allowed */
+
 // Reference: AMD Software Developer Manual Vol.2 Ch.5 "Page Translation and Protection"
 
-static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr_t fault_addr, pf_error_t error_code) {
+static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr_t fault_addr, pf_error_t error_code,
+                                                 addr_t *actual_start, addr_t *actual_end) {
     pml4e64_t * pml      = NULL;
     pdpe64_t * pdpe      = NULL;
     pde64_t * pde        = NULL;
@@ -46,7 +49,7 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
     int page_size = PAGE_SIZE_4KB;
 
     if (region == NULL) {
-       PrintError("%s: invalid region, addr=%p\n", __FUNCTION__, (void *)fault_addr);
+       PrintError(core->vm_info, core, "%s: invalid region, addr=%p\n", __FUNCTION__, (void *)fault_addr);
        return -1;
     }
 
@@ -58,7 +61,7 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
        page_size = v3_get_max_page_size(core, fault_addr, LONG);
     }
 
-    PrintDebug("Using page size of %dKB\n", page_size / 1024);
+    PrintDebug(core->vm_info, core, "Using page size of %dKB\n", page_size / 1024);
 
  
     // Lookup the correct PML address based on the PAGING MODE
@@ -70,7 +73,7 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
 
     //Fix up the PML entry
     if (pml[pml_index].present == 0) {
-       pdpe = (pdpe64_t *)create_generic_pt_page();
+       pdpe = (pdpe64_t *)create_generic_pt_page(core);
    
        // Set default PML Flags...
        pml[pml_index].present = 1;
@@ -84,7 +87,7 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
 
     // Fix up the PDPE entry
     if (pdpe[pdpe_index].present == 0) {
-       pde = (pde64_t *)create_generic_pt_page();
+       pde = (pde64_t *)create_generic_pt_page(core);
        
        // Set default PDPE Flags...
        pdpe[pdpe_index].present = 1;
@@ -101,6 +104,9 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
        pde2mb = (pde64_2MB_t *)pde; // all but these two lines are the same for PTE
        pde2mb[pde_index].large_page = 1;
 
+       *actual_start = BASE_TO_PAGE_ADDR_2MB(PAGE_BASE_ADDR_2MB(fault_addr));
+       *actual_end = BASE_TO_PAGE_ADDR_2MB(PAGE_BASE_ADDR_2MB(fault_addr)+1)-1;
+
        if (pde2mb[pde_index].present == 0) {
            pde2mb[pde_index].user_page = 1;
 
@@ -116,7 +122,7 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
                }
 
                if (v3_gpa_to_hpa(core, fault_addr, &host_addr) == -1) {
-                   PrintError("Error Could not translate fault addr (%p)\n", (void *)fault_addr);
+                   PrintError(core->vm_info, core, "Error Could not translate fault addr (%p)\n", (void *)fault_addr);
                    return -1;
                }
 
@@ -137,9 +143,12 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
 
     // Continue with the 4KiB page heirarchy
     
+    *actual_start = BASE_TO_PAGE_ADDR_4KB(PAGE_BASE_ADDR_4KB(fault_addr));
+    *actual_end = BASE_TO_PAGE_ADDR_4KB(PAGE_BASE_ADDR_4KB(fault_addr)+1)-1;
+
     // Fix up the PDE entry
     if (pde[pde_index].present == 0) {
-       pte = (pte64_t *)create_generic_pt_page();
+       pte = (pte64_t *)create_generic_pt_page(core);
        
        pde[pde_index].present = 1;
        pde[pde_index].writable = 1;
@@ -166,7 +175,7 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
            }
 
            if (v3_gpa_to_hpa(core, fault_addr, &host_addr) == -1) {
-               PrintError("Error Could not translate fault addr (%p)\n", (void *)fault_addr);
+               PrintError(core->vm_info, core, "Error Could not translate fault addr (%p)\n", (void *)fault_addr);
                return -1;
            }
 
@@ -184,7 +193,8 @@ static inline int handle_passthrough_pagefault_64(struct guest_info * core, addr
     return 0;
 }
 
-static inline int invalidate_addr_64(struct guest_info * core, addr_t inv_addr) {
+static inline int invalidate_addr_64_internal(struct guest_info * core, addr_t inv_addr,
+                                             addr_t *actual_start, uint64_t *actual_size) {
     pml4e64_t * pml = NULL;
     pdpe64_t * pdpe = NULL;
     pde64_t * pde = NULL;
@@ -209,28 +219,38 @@ static inline int invalidate_addr_64(struct guest_info * core, addr_t inv_addr)
     }
 
     if (pml[pml_index].present == 0) {
-       return 0;
+        *actual_start = BASE_TO_PAGE_ADDR_512GB(PAGE_BASE_ADDR_512GB(inv_addr));
+        *actual_size = PAGE_SIZE_512GB;
+       return 0;
     }
 
     pdpe = V3_VAddr((void*)BASE_TO_PAGE_ADDR(pml[pml_index].pdp_base_addr));
 
     if (pdpe[pdpe_index].present == 0) {
+        *actual_start = BASE_TO_PAGE_ADDR_1GB(PAGE_BASE_ADDR_1GB(inv_addr));
+        *actual_size = PAGE_SIZE_1GB;
        return 0;
     } else if (pdpe[pdpe_index].large_page == 1) { // 1GiB
        pdpe[pdpe_index].present = 0;
        pdpe[pdpe_index].writable = 0;
        pdpe[pdpe_index].user_page = 0;
+        *actual_start = BASE_TO_PAGE_ADDR_1GB(PAGE_BASE_ADDR_1GB(inv_addr));
+        *actual_size = PAGE_SIZE_1GB;
        return 0;
     }
 
     pde = V3_VAddr((void*)BASE_TO_PAGE_ADDR(pdpe[pdpe_index].pd_base_addr));
 
     if (pde[pde_index].present == 0) {
+        *actual_start = BASE_TO_PAGE_ADDR_2MB(PAGE_BASE_ADDR_2MB(inv_addr));
+        *actual_size = PAGE_SIZE_2MB;
        return 0;
     } else if (pde[pde_index].large_page == 1) { // 2MiB
        pde[pde_index].present = 0;
        pde[pde_index].writable = 0;
        pde[pde_index].user_page = 0;
+        *actual_start = BASE_TO_PAGE_ADDR_2MB(PAGE_BASE_ADDR_2MB(inv_addr));
+        *actual_size = PAGE_SIZE_2MB;
        return 0;
     }
 
@@ -240,9 +260,50 @@ static inline int invalidate_addr_64(struct guest_info * core, addr_t inv_addr)
     pte[pte_index].writable = 0;
     pte[pte_index].user_page = 0;
 
+    *actual_start = BASE_TO_PAGE_ADDR_4KB(PAGE_BASE_ADDR_4KB(inv_addr));
+    *actual_size = PAGE_SIZE_4KB;
+
     return 0;
 }
 
+static inline int invalidate_addr_64(struct guest_info * core, addr_t inv_addr, 
+                                    addr_t *actual_start, addr_t *actual_end)
+{
+  uint64_t len;
+  int rc;
+  
+  rc = invalidate_addr_64_internal(core,inv_addr,actual_start,&len);
+
+  *actual_end = *actual_start + len - 1;
+
+  return rc;
+}
+   
+static inline int invalidate_addr_64_range(struct guest_info * core, addr_t inv_addr_start, addr_t inv_addr_end, 
+                                          addr_t *actual_start, addr_t *actual_end)
+{
+  addr_t next;
+  addr_t start;
+  uint64_t len;
+  int rc;
+  
+  for (next=inv_addr_start; next<=inv_addr_end; ) {
+    rc = invalidate_addr_64_internal(core,next,&start, &len);
+    if (next==inv_addr_start) { 
+      // first iteration, capture where we start invalidating
+      *actual_start = start;
+    }
+    if (rc) { 
+      return rc;
+    }
+    next = start + len;
+    *actual_end = next;
+  }
+  // last iteration, actual_end is off by one
+  (*actual_end)--;
+  return 0;
+}
+
 
 
 #endif