refactor: decouple i386 specific instruction invocation

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

diff --git a/lunaix-os/kernel/mm/mmap.c b/lunaix-os/kernel/mm/mmap.c
index 7b8ce8fb81f79f810ab6aacdadafa297ef4603c2..77e1264d2926661d78ea1ff335e030a250f5ccc5 100644 (file)
--- a/lunaix-os/kernel/mm/mmap.c
+++ b/lunaix-os/kernel/mm/mmap.c
@@ -5,79 +5,253 @@
 #include <lunaix/spike.h>
 
 #include <lunaix/syscall.h>
 #include <lunaix/spike.h>
 
 #include <lunaix/syscall.h>

+#include <lunaix/syscall_utils.h>

 
 

-void*
-mem_map(ptr_t mnt,
-        vm_regions_t* regions,
-        void* addr,
-        struct v_file* file,
-        off_t offset,
-        size_t length,
-        u32_t proct,
-        u32_t options)
+// any size beyond this is bullshit
+#define BS_SIZE (KERNEL_MM_BASE - UMMAP_START)
+
+int
+mem_has_overlap(vm_regions_t* regions, ptr_t start, ptr_t end)
+{
+    struct mm_region *pos, *n;
+    llist_for_each(pos, n, regions, head)
+    {
+        if (pos->end >= start && pos->start < start) {
+            return 1;
+        }
+
+        if (pos->end <= end && pos->start >= start) {
+            return 1;
+        }
+
+        if (pos->end >= end && pos->start < end) {
+            return 1;
+        }
+    }
+
+    return 0;
+}
+
+int
+mem_adjust_inplace(vm_regions_t* regions,
+                   struct mm_region* region,
+                   ptr_t newend)

 {
 {

-    if (!length || (length & (PG_SIZE - 1)) || (offset & (PG_SIZE - 1))) {
-        __current->k_status = EINVAL;
-        return (void*)-1;
+    ssize_t len = newend - region->start;
+    if (len == 0) {
+        return 0;

     }
 
     }
 

-    // read_page is not supported
-    if (!file->ops->read_page) {
-        __current->k_status = ENODEV;
-        return (void*)-1;
+    if (len < 0) {
+        return EINVAL;

     }
 
     }
 

-    ptr_t last_end = USER_START;
+    if (mem_has_overlap(regions, region->start, newend)) {
+        return ENOMEM;
+    }
+
+    region->end = newend;
+
+    return 0;
+}
+
+int
+mem_map(void** addr_out,
+        struct mm_region** created,
+        ptr_t addr,
+        struct v_file* file,
+        struct mmap_param* param)
+{
+    assert_msg(addr, "addr can not be NULL");
+
+    ptr_t last_end = USER_START, found_loc = addr;

     struct mm_region *pos, *n;
     struct mm_region *pos, *n;

-    llist_for_each(pos, n, regions, head)
+
+    vm_regions_t* vm_regions = &param->pvms->regions;
+
+    if ((param->flags & MAP_FIXED_NOREPLACE)) {
+        if (mem_has_overlap(vm_regions, found_loc, param->mlen + found_loc)) {
+            return EEXIST;
+        }
+        goto found;
+    }
+
+    if ((param->flags & MAP_FIXED)) {
+        int status =
+          mem_unmap(param->vms_mnt, vm_regions, found_loc, param->mlen);
+        if (status) {
+            return status;
+        }
+        goto found;
+    }
+
+    llist_for_each(pos, n, vm_regions, head)

     {
     {

-        if (pos->start - last_end >= length && last_end >= addr) {
-            goto found;
+        if (last_end < found_loc) {
+            size_t avail_space = pos->start - found_loc;
+            if (pos->start > found_loc && avail_space > param->mlen) {
+                goto found;
+            }
+            found_loc = pos->end + PG_SIZE;

         }
         }

+

         last_end = pos->end;
     }
 
         last_end = pos->end;
     }
 

-    __current->k_status = ENOMEM;
-    return (void*)-1;
+    return ENOMEM;

 
 found:
 
 found:

-    addr = last_end;
-    ptr_t end = addr + length;
+    if (found_loc >= KERNEL_MM_BASE || found_loc < USER_START) {
+        return ENOMEM;
+    }
+
+    struct mm_region* region = region_create_range(
+      found_loc,
+      param->mlen,
+      ((param->proct | param->flags) & 0x3f) | (param->type & ~0xffff));

 
 

-    struct mm_region* region =
-      region_create(addr, end, proct | (options & 0x1f));

     region->mfile = file;
     region->mfile = file;

-    region->offset = offset;
+    region->foff = param->offset;
+    region->flen = param->flen;
+    region->proc_vms = param->pvms;

 
 

-    region_add(regions, region);
+    region_add(vm_regions, region);

 
     u32_t attr = PG_ALLOW_USER;
 
     u32_t attr = PG_ALLOW_USER;

-    if ((proct & REGION_WRITE)) {
+    if ((param->proct & REGION_WRITE)) {

         attr |= PG_WRITE;
     }
 
         attr |= PG_WRITE;
     }
 

-    for (u32_t i = 0; i < length; i += PG_SIZE) {
-        vmm_set_mapping(mnt, addr + i, 0, attr, 0);
+    for (u32_t i = 0; i < param->mlen; i += PG_SIZE) {
+        vmm_set_mapping(param->vms_mnt, found_loc + i, 0, attr, 0);
+    }
+
+    if (file) {
+        vfs_ref_file(file);
+    }
+
+    if (addr_out) {
+        *addr_out = (void*)found_loc;
+    }
+    if (created) {
+        *created = region;
+    }
+    return 0;
+}
+
+int
+mem_remap(void** addr_out,
+          struct mm_region** remapped,
+          void* addr,
+          struct v_file* file,
+          struct mmap_param* param)
+{
+    // TODO
+
+    return EINVAL;
+}
+
+void
+mem_sync_pages(ptr_t mnt,
+               struct mm_region* region,
+               ptr_t start,
+               ptr_t length,
+               int options)
+{
+    if (!region->mfile || !(region->attr & REGION_WSHARED)) {
+        return;
+    }
+
+    v_mapping mapping;
+    for (size_t i = 0; i < length; i += PG_SIZE) {
+        if (!vmm_lookupat(mnt, start + i, &mapping)) {
+            continue;
+        }
+
+        if (PG_IS_DIRTY(*mapping.pte)) {
+            size_t offset = mapping.va - region->start + region->foff;
+            struct v_inode* inode = region->mfile->inode;
+
+            region->mfile->ops->write_page(
+              inode, (void*)mapping.va, PG_SIZE, offset);
+
+            *mapping.pte &= ~PG_DIRTY;
+
+            cpu_flush_page((ptr_t)mapping.pte);
+        } else if ((options & MS_INVALIDATE)) {
+            goto invalidate;
+        }
+
+        if (options & MS_INVALIDATE_ALL) {
+            goto invalidate;
+        }
+
+        continue;
+
+    invalidate:
+        *mapping.pte &= ~PG_PRESENT;
+        pmm_free_page(KERNEL_PID, mapping.pa);
+        cpu_flush_page((ptr_t)mapping.pte);
+    }
+}
+
+int
+mem_msync(ptr_t mnt,
+          vm_regions_t* regions,
+          ptr_t addr,
+          size_t length,
+          int options)
+{
+    struct mm_region* pos = list_entry(regions->next, struct mm_region, head);
+    while (length && (ptr_t)&pos->head != (ptr_t)regions) {
+        if (pos->end >= addr && pos->start <= addr) {
+            size_t l = MIN(length, pos->end - addr);
+            mem_sync_pages(mnt, pos, addr, l, options);
+
+            addr += l;
+            length -= l;
+        }
+        pos = list_entry(pos->head.next, struct mm_region, head);
+    }
+
+    if (length) {
+        return ENOMEM;

     }
 
     }
 

-    return addr;
+    return 0;
+}
+
+void
+mem_unmap_region(ptr_t mnt, struct mm_region* region)
+{
+    size_t len = ROUNDUP(region->end - region->start, PG_SIZE);
+    mem_sync_pages(mnt, region, region->start, len, 0);
+
+    for (size_t i = region->start; i <= region->end; i += PG_SIZE) {
+        ptr_t pa = vmm_del_mapping(mnt, i);
+        if (pa) {
+            pmm_free_page(__current->pid, pa);
+        }
+    }
+    llist_delete(&region->head);
+    region_release(region);

 }
 
 }
 

-void*
-mem_unmap(ptr_t mnt, vm_regions_t* regions, void* addr, size_t length)
+int
+mem_unmap(ptr_t mnt, vm_regions_t* regions, ptr_t addr, size_t length)

 {
     length = ROUNDUP(length, PG_SIZE);
 {
     length = ROUNDUP(length, PG_SIZE);

-    ptr_t cur_addr = ROUNDDOWN((ptr_t)addr, PG_SIZE);
+    ptr_t cur_addr = PG_ALIGN(addr);

     struct mm_region *pos, *n;
 
     llist_for_each(pos, n, regions, head)
     {
     struct mm_region *pos, *n;
 
     llist_for_each(pos, n, regions, head)
     {

-        if (pos->start <= cur_addr) {
+        if (pos->start <= cur_addr && pos->end >= cur_addr) {

             break;
         }
     }
 
             break;
         }
     }
 

-    while (&pos->head != regions && cur_addr > pos->start) {
+    while (&pos->head != regions && cur_addr >= pos->start) {

         u32_t l = pos->end - cur_addr;
         pos->end = cur_addr;
 
         u32_t l = pos->end - cur_addr;
         pos->end = cur_addr;
 


@@ -90,26 +264,27 @@ mem_unmap(ptr_t mnt, vm_regions_t* regions, void* addr, size_t length)
             l = length;
         }
 
             l = length;
         }
 

-        // TODO for shared mappings, sync page content if modified. (also
-        // implement msync)
+        mem_sync_pages(mnt, pos, cur_addr, l, 0);

 
         for (size_t i = 0; i < l; i += PG_SIZE) {
             ptr_t pa = vmm_del_mapping(mnt, cur_addr + i);
             if (pa) {
 
         for (size_t i = 0; i < l; i += PG_SIZE) {
             ptr_t pa = vmm_del_mapping(mnt, cur_addr + i);
             if (pa) {

-                pmm_free_page(__current->pid, pa);
+                pmm_free_page(pos->proc_vms->pid, pa);

             }
         }
 
         n = container_of(pos->head.next, typeof(*pos), head);
         if (pos->end == pos->start) {
             llist_delete(&pos->head);
             }
         }
 
         n = container_of(pos->head.next, typeof(*pos), head);
         if (pos->end == pos->start) {
             llist_delete(&pos->head);

-            vfree(pos);
+            region_release(pos);

         }
 
         pos = n;
         length -= l;
         cur_addr += length;
     }
         }
 
         pos = n;
         length -= l;
         cur_addr += length;
     }

+
+    return 0;

 }
 
 __DEFINE_LXSYSCALL3(void*, sys_mmap, void*, addr, size_t, length, va_list, lst)
 }
 
 __DEFINE_LXSYSCALL3(void*, sys_mmap, void*, addr, size_t, length, va_list, lst)


@@ -118,27 +293,72 @@ __DEFINE_LXSYSCALL3(void*, sys_mmap, void*, addr, size_t, length, va_list, lst)
     int fd = va_arg(lst, u32_t);
     off_t offset = va_arg(lst, off_t);
     int options = va_arg(lst, int);
     int fd = va_arg(lst, u32_t);
     off_t offset = va_arg(lst, off_t);
     int options = va_arg(lst, int);

-

     int errno = 0;
     int errno = 0;

-    struct v_fd* vfd;
-    if ((errno = vfs_getfd(fd, &vfd))) {
-        __current->k_status = errno;
-        return (void*)-1;
+    void* result = (void*)-1;
+
+    ptr_t addr_ptr = (ptr_t)addr;
+
+    if (!length || length > BS_SIZE || !PG_ALIGNED(addr_ptr)) {
+        errno = EINVAL;
+        goto done;

     }
 
     }
 

-    length = ROUNDUP(length, PG_SIZE);
+    if (!addr_ptr) {
+        addr_ptr = UMMAP_START;
+    } else if (addr_ptr < UMMAP_START || addr_ptr + length >= UMMAP_END) {
+        if (!(options & (MAP_FIXED | MAP_FIXED_NOREPLACE))) {
+            errno = ENOMEM;
+            goto done;
+        }
+    }
+
+    struct v_file* file = NULL;
+
+    if (!(options & MAP_ANON)) {
+        struct v_fd* vfd;
+        if ((errno = vfs_getfd(fd, &vfd))) {
+            goto done;
+        }
+
+        file = vfd->file;
+        if (!file->ops->read_page) {
+            errno = ENODEV;
+            goto done;
+        }
+    }
+
+    struct mmap_param param = { .flags = options,
+                                .mlen = ROUNDUP(length, PG_SIZE),
+                                .offset = offset,
+                                .type = REGION_TYPE_GENERAL,
+                                .proct = proct,
+                                .pvms = (struct proc_mm*)&__current->mm,
+                                .vms_mnt = VMS_SELF };
+
+    errno = mem_map(&result, NULL, addr_ptr, file, &param);

 
 

-    return mem_map(PD_REFERENCED,
-                   &__current->mm.regions,
-                   addr,
-                   vfd->file,
-                   offset,
-                   length,
-                   proct,
-                   options);
+done:
+    __current->k_status = errno;
+    return result;

 }
 
 }
 

-__DEFINE_LXSYSCALL2(void, munmap, void*, addr, size_t, length)
+__DEFINE_LXSYSCALL2(int, munmap, void*, addr, size_t, length)

 {
 {

-    return mem_unmap(PD_REFERENCED, &__current->mm.regions, addr, length);
+    return mem_unmap(
+      VMS_SELF, (vm_regions_t*)&__current->mm.regions, (ptr_t)addr, length);
+}
+
+__DEFINE_LXSYSCALL3(int, msync, void*, addr, size_t, length, int, flags)
+{
+    if (!PG_ALIGNED(addr) || ((flags & MS_ASYNC) && (flags & MS_SYNC))) {
+        return DO_STATUS(EINVAL);
+    }
+
+    int status = mem_msync(VMS_SELF,
+                           (vm_regions_t*)&__current->mm.regions,
+                           (ptr_t)addr,
+                           length,
+                           flags);
+
+    return DO_STATUS(status);

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