basic process support and some syscalls

[lunaix-os.git] / lunaix-os / kernel / mm / vmm.c

diff --git a/lunaix-os/kernel/mm/vmm.c b/lunaix-os/kernel/mm/vmm.c
index 364332e186ee7621ecea955a90f64030c21c5dbc..f791cc38969f340d49438fc748f32790ec7b2609 100644 (file)
--- a/lunaix-os/kernel/mm/vmm.c
+++ b/lunaix-os/kernel/mm/vmm.c
@@ -1,143 +1,258 @@
-#include <lunaix/mm/page.h>
+#include <hal/cpu.h>
+#include <klibc/string.h>

 #include <lunaix/mm/vmm.h>
 #include <lunaix/mm/vmm.h>

-#include <lunaix/mm/pmm.h>
-#include <libc/string.h>
-
-// TODO: Move these nasty inline asm stuff into hal
-//      These should be arch dependent
-ptd_t* get_pd() {
-    ptd_t* pd;
-    #ifdef __ARCH_IA32
-    __asm__(
-        "movl %%cr3, %0\n"
-        "andl $0xfffff000, %0"
-        : "=r"(pd)
-    );
-    #endif
-    return P2V(pd);
-}
+#include <lunaix/spike.h>

 
 

-void set_pd(ptd_t* pd) {
-    #ifdef __ARCH_IA32
-    __asm__(
-        "movl %0, %%eax\n"
-        "andl $0xfffff000, %%eax\n"
-        "movl %%eax, %%cr3\n"
-        :
-        : "r" (pd)
-    );
-    #endif
-}
+#include <stdbool.h>

 
 

-void vmm_init() {
-    // TODO: something here?
+void
+vmm_init()
+{
+    // XXX: something here?

 }
 
 }
 

-ptd_t* vmm_init_pd() {
-    ptd_t* dir = pmm_alloc_page();
-    for (size_t i = 0; i < 1024; i++)
-    {
-        dir[i] = 0;
+x86_page_table*
+vmm_init_pd()
+{
+    x86_page_table* dir = (x86_page_table*)pmm_alloc_page(KERNEL_PID, PP_FGPERSIST);
+    for (size_t i = 0; i < PG_MAX_ENTRIES; i++) {
+        dir->entry[i] = PTE_NULL;

     }
     }

-    
-    // 自己映射自己，方便我们在软件层面进行查表地址转换
-    dir[1023] = PDE(T_SELF_REF_PERM, dir);
+
+    // 递归映射，方便我们在软件层面进行查表地址转换
+    dir->entry[PG_MAX_ENTRIES - 1] = NEW_L1_ENTRY(T_SELF_REF_PERM, dir);

 
     return dir;
 }
 
 
     return dir;
 }
 

-void* vmm_map_page(void* va, void* pa, pt_attr dattr, pt_attr tattr) {
+int
+__vmm_map_internal(pid_t pid, 
+                   uint32_t l1_inx,
+                   uint32_t l2_inx,
+                   uintptr_t pa,
+                   pt_attr attr,
+                   int forced)
+{
+    x86_page_table* l1pt = (x86_page_table*)L1_BASE_VADDR;
+    x86_page_table* l2pt = (x86_page_table*)L2_VADDR(l1_inx);
+
+    // See if attr make sense
+    assert(attr <= 128);
+
+    if (!l1pt->entry[l1_inx]) {
+        x86_page_table* new_l1pt_pa = pmm_alloc_page(pid, PP_FGPERSIST);
+
+        // 物理内存已满！
+        if (!new_l1pt_pa) {
+            return 0;
+        }
+
+        // This must be writable
+        l1pt->entry[l1_inx] = NEW_L1_ENTRY(attr | PG_WRITE, new_l1pt_pa);
+        memset((void*)L2_VADDR(l1_inx), 0, PG_SIZE);
+    }
+
+    x86_pte_t l2pte = l2pt->entry[l2_inx];
+    if (l2pte) {
+        if (!forced) {
+            return 0;
+        }
+    }
+
+    if ((HAS_FLAGS(attr, PG_PRESENT))) {
+        // add one on reference count, regardless of existence.
+        pmm_ref_page(pid, pa);
+    }
+
+    l2pt->entry[l2_inx] = NEW_L2_ENTRY(attr, pa);
+
+    return 1;
+}
+
+void*
+vmm_map_page(pid_t pid, void* va, void* pa, pt_attr tattr)
+{

     // 显然，对空指针进行映射没有意义。
     if (!pa || !va) {
         return NULL;
     }
 
     // 显然，对空指针进行映射没有意义。
     if (!pa || !va) {
         return NULL;
     }
 

-    uintptr_t pd_offset = PD_INDEX(va);
-    uintptr_t pt_offset = PT_INDEX(va);
-    ptd_t* ptd = (ptd_t*)PTD_BASE_VADDR;
+    assert(((uintptr_t)va & 0xFFFU) == 0) assert(((uintptr_t)pa & 0xFFFU) == 0);
+
+    uint32_t l1_index = L1_INDEX(va);
+    uint32_t l2_index = L2_INDEX(va);
+    x86_page_table* l1pt = (x86_page_table*)L1_BASE_VADDR;

 
     // 在页表与页目录中找到一个可用的空位进行映射（位于va或其附近）
 
     // 在页表与页目录中找到一个可用的空位进行映射（位于va或其附近）

-    ptd_t* pde = ptd[pd_offset];
-    pt_t* pt = (uintptr_t)PT_VADDR(pd_offset);
-    while (pde && pd_offset < 1024) {
-        if (pt_offset == 1024) {
-            pd_offset++;
-            pt_offset = 0;
-            pde = ptd[pd_offset];
-            pt = (pt_t*)PT_VADDR(pd_offset);
+    x86_pte_t l1pte = l1pt->entry[l1_index];
+    x86_page_table* l2pt = (x86_page_table*)L2_VADDR(l1_index);
+    while (l1pte && l1_index < PG_MAX_ENTRIES) {
+        if (l2_index == PG_MAX_ENTRIES) {
+            l1_index++;
+            l2_index = 0;
+            l1pte = l1pt->entry[l1_index];
+            l2pt = (x86_page_table*)L2_VADDR(l1_index);

         }
         }

-        // 页表有空位，只需要开辟一个新的 PTE
-        if (pt && !pt[pt_offset]) {
-            pt[pt_offset] = PTE(tattr, pa);
-            return V_ADDR(pd_offset, pt_offset, PG_OFFSET(va));
+        // 页表有空位，只需要开辟一个新的 PTE (Level 2)
+        if (__vmm_map_internal(pid, l1_index, l2_index, pa, tattr, false)) {
+            return (void*)V_ADDR(l1_index, l2_index, PG_OFFSET(va));

         }
         }

-        pt_offset++;
+        l2_index++;

     }
     }

-    
+

     // 页目录与所有页表已满！
     // 页目录与所有页表已满！

-    if (pd_offset > 1024) {
+    if (l1_index > PG_MAX_ENTRIES) {

         return NULL;
     }
 
         return NULL;
     }
 

-    // 页目录有空位，需要开辟一个新的 PDE
-    uint8_t* new_pt_pa = pmm_alloc_page();
-    
-    // 物理内存已满！
-    if (!new_pt_pa) {
+    if (!__vmm_map_internal(pid, l1_index, l2_index, (uintptr_t)pa, tattr, false)) {

         return NULL;
     }
         return NULL;
     }

-    
-    ptd[pd_offset] = PDE(dattr, new_pt_pa);
-    
-    memset((void*)PT_VADDR(pd_offset), 0, PM_PAGE_SIZE);
-    pt[pt_offset] = PTE(tattr, pa);

 
 

-    return V_ADDR(pd_offset, pt_offset, PG_OFFSET(va));
+    return (void*)V_ADDR(l1_index, l2_index, PG_OFFSET(va));

 }
 
 }
 

-void* vmm_alloc_page(void* vpn, pt_attr dattr, pt_attr tattr) {
-    void* pp = pmm_alloc_page();
-    void* result = vmm_map_page(vpn, pp, dattr, tattr);
+void*
+vmm_fmap_page(pid_t pid, void* va, void* pa, pt_attr tattr)
+{
+    if (!pa || !va) {
+        return NULL;
+    }
+
+    assert(((uintptr_t)va & 0xFFFU) == 0) assert(((uintptr_t)pa & 0xFFFU) == 0);
+
+    uint32_t l1_index = L1_INDEX(va);
+    uint32_t l2_index = L2_INDEX(va);
+
+    if (!__vmm_map_internal(pid, l1_index, l2_index, (uintptr_t)pa, tattr, true)) {
+        return NULL;
+    }
+
+    cpu_invplg(va);
+
+    return va;
+}
+
+void*
+vmm_alloc_page(pid_t pid, void* vpn, void** pa, pt_attr tattr, pp_attr_t pattr)
+{
+    void* pp = pmm_alloc_page(pid, pattr);
+    void* result = vmm_map_page(pid, vpn, pp, tattr);

     if (!result) {
     if (!result) {

-        pmm_free_page(pp);
+        pmm_free_page(pp, pid);

     }
     }

+    pa ? (*pa = pp) : 0;

     return result;
 }
 
     return result;
 }
 

-void vmm_unmap_page(void* vpn) {
-    uintptr_t pd_offset = PD_INDEX(vpn);
-    uintptr_t pt_offset = PT_INDEX(vpn);
-    ptd_t* self_pde = PTD_BASE_VADDR;
-
-    ptd_t pde = self_pde[pd_offset];
-
-    if (pde) {
-        pt_t* pt = (pt_t*)PT_VADDR(pd_offset);
-        uint32_t pte = pt[pt_offset];
-        if (IS_CACHED(pte) && pmm_free_page(pte)) {
-            // 刷新TLB
-            #ifdef __ARCH_IA32
-            __asm__("invlpg (%0)" :: "r"((uintptr_t)vpn) : "memory");
-            #endif
+int
+vmm_alloc_pages(pid_t pid, void* va, size_t sz, pt_attr tattr, pp_attr_t pattr)
+{
+    assert((uintptr_t)va % PG_SIZE == 0) assert(sz % PG_SIZE == 0);
+
+    void* va_ = va;
+    for (size_t i = 0; i < (sz >> PG_SIZE_BITS); i++, va_ += PG_SIZE) {
+        void* pp = pmm_alloc_page(pid, pattr);
+        uint32_t l1_index = L1_INDEX(va_);
+        uint32_t l2_index = L2_INDEX(va_);
+        if (!pp || !__vmm_map_internal(
+                     pid,
+                     l1_index, l2_index, (uintptr_t)pp, tattr, false)) {
+            // if one failed, release previous allocated pages.
+            va_ = va;
+            for (size_t j = 0; j < i; j++, va_ += PG_SIZE) {
+                vmm_unmap_page(pid, va_);
+            }
+
+            return false;
+        }
+    }
+
+    return true;
+}
+
+int
+vmm_set_mapping(pid_t pid, void* va, void* pa, pt_attr attr) {
+    assert(((uintptr_t)va & 0xFFFU) == 0);
+
+    uint32_t l1_index = L1_INDEX(va);
+    uint32_t l2_index = L2_INDEX(va);
+
+    // prevent map of recursive mapping region
+    if (l1_index == 1023) {
+        return 0;
+    }
+    
+    __vmm_map_internal(pid, l1_index, l2_index, (uintptr_t)pa, attr, false);
+    return 1;
+}
+
+void
+__vmm_unmap_internal(pid_t pid, void* va, int free_ppage) {
+    assert(((uintptr_t)va & 0xFFFU) == 0);
+
+    uint32_t l1_index = L1_INDEX(va);
+    uint32_t l2_index = L2_INDEX(va);
+
+    // prevent unmap of recursive mapping region
+    if (l1_index == 1023) {
+        return;
+    }
+
+    x86_page_table* l1pt = (x86_page_table*)L1_BASE_VADDR;
+
+    x86_pte_t l1pte = l1pt->entry[l1_index];
+
+    if (l1pte) {
+        x86_page_table* l2pt = (x86_page_table*)L2_VADDR(l1_index);
+        x86_pte_t l2pte = l2pt->entry[l2_index];
+        if (IS_CACHED(l2pte) && free_ppage) {
+            pmm_free_page(pid, (void*)l2pte);

         }
         }

-        pt[pt_offset] = 0;
+        cpu_invplg(va);
+        l2pt->entry[l2_index] = PTE_NULL;

     }
 }
 
     }
 }
 

-void* vmm_v2p(void* va) {
-    uintptr_t pd_offset = PD_INDEX(va);
-    uintptr_t pt_offset = PT_INDEX(va);
-    uintptr_t po = PG_OFFSET(va);
-    ptd_t* self_pde = PTD_BASE_VADDR;
-
-    ptd_t pde = self_pde[pd_offset];
-    if (pde) {
-        pt_t pte = ((pt_t*)PT_VADDR(pd_offset))[pt_offset];
-        if (pte) {
-            uintptr_t ppn = pte >> 12;
-            return (void*)P_ADDR(ppn, po);
+void
+vmm_unset_mapping(void* va) {
+    __vmm_unmap_internal(0, va, false);
+}
+
+void
+vmm_unmap_page(pid_t pid, void* va)
+{
+    __vmm_unmap_internal(pid, va, true);
+}
+
+v_mapping
+vmm_lookup(void* va)
+{
+    assert(((uintptr_t)va & 0xFFFU) == 0);
+
+    uint32_t l1_index = L1_INDEX(va);
+    uint32_t l2_index = L2_INDEX(va);
+
+    x86_page_table* l1pt = (x86_page_table*)L1_BASE_VADDR;
+    x86_pte_t l1pte = l1pt->entry[l1_index];
+
+    v_mapping mapping = { .flags = 0, .pa = 0, .pn = 0 };
+    if (l1pte) {
+        x86_pte_t* l2pte =
+          &((x86_page_table*)L2_VADDR(l1_index))->entry[l2_index];
+        if (l2pte) {
+            mapping.flags = PG_ENTRY_FLAGS(*l2pte);
+            mapping.pa = PG_ENTRY_ADDR(*l2pte);
+            mapping.pn = mapping.pa >> PG_SIZE_BITS;
+            mapping.pte = l2pte;

         }
     }
 
         }
     }
 

-    return NULL;
+    return mapping;
+}
+
+void*
+vmm_v2p(void* va)
+{
+    return (void*)vmm_lookup(va).pa;

 }
\ No newline at end of file
 }
\ No newline at end of file