Multiuser, Capabilities and Access Controls (#54)

[lunaix-os.git] / lunaix-os / kernel / fs / mount.c

diff --git a/lunaix-os/kernel/fs/mount.c b/lunaix-os/kernel/fs/mount.c
index c5b7344be4f2723f45961f53705a10c395f67fe0..b62b77faa8788d273888a7828933a56b0c1ffa13 100644 (file)
--- a/lunaix-os/kernel/fs/mount.c
+++ b/lunaix-os/kernel/fs/mount.c
@@ -11,8 +11,16 @@ LOG_MODULE("fs")
 
 struct llist_header all_mnts = { .next = &all_mnts, .prev = &all_mnts };
 
 
 struct llist_header all_mnts = { .next = &all_mnts, .prev = &all_mnts };
 

-struct v_mount*
-vfs_create_mount(struct v_mount* parent, struct v_dnode* mnt_point)
+static void
+__vfs_attach_vmnt(struct v_dnode* mnt_point, struct v_mount* vmnt)
+{
+    vmnt->mnt_point = mnt_point;
+    vfs_d_assign_vmnt(mnt_point, vmnt);
+    vfs_ref_dnode(mnt_point);
+}
+
+static struct v_mount*
+__vfs_create_mount(struct v_mount* parent, struct v_superblock* mnt_sb)

 {
     struct v_mount* mnt = vzalloc(sizeof(struct v_mount));
     if (!mnt) {
 {
     struct v_mount* mnt = vzalloc(sizeof(struct v_mount));
     if (!mnt) {


@@ -20,12 +28,12 @@ vfs_create_mount(struct v_mount* parent, struct v_dnode* mnt_point)
     }
 
     llist_init_head(&mnt->submnts);
     }
 
     llist_init_head(&mnt->submnts);

+    llist_init_head(&mnt->sibmnts);

     llist_append(&all_mnts, &mnt->list);
     mutex_init(&mnt->lock);
 
     mnt->parent = parent;
     llist_append(&all_mnts, &mnt->list);
     mutex_init(&mnt->lock);
 
     mnt->parent = parent;

-    mnt->mnt_point = mnt_point;
-    vfs_vmnt_assign_sb(mnt, mnt_point->super_block);
+    vfs_vmnt_assign_sb(mnt, mnt_sb);

 
     if (parent) {
         mnt_mkbusy(parent);
 
     if (parent) {
         mnt_mkbusy(parent);


@@ -33,28 +41,40 @@ vfs_create_mount(struct v_mount* parent, struct v_dnode* mnt_point)
         llist_append(&parent->submnts, &mnt->sibmnts);
         mutex_unlock(&mnt->parent->lock);
     }
         llist_append(&parent->submnts, &mnt->sibmnts);
         mutex_unlock(&mnt->parent->lock);
     }

-    
-    atomic_fetch_add(&mnt_point->ref_count, 1);

 
     return mnt;
 }
 
 
     return mnt;
 }
 

-void
-__vfs_release_vmnt(struct v_mount* mnt)
+static void
+__vfs_detach_vmnt(struct v_mount* mnt)

 {
     assert(llist_empty(&mnt->submnts));
 
 {
     assert(llist_empty(&mnt->submnts));
 

+    vfs_unref_dnode(mnt->mnt_point);
+    assert(!mnt->busy_counter);
+

     if (mnt->parent) {
         mnt_chillax(mnt->parent);
     }
 
     if (mnt->parent) {
         mnt_chillax(mnt->parent);
     }
 

+    mnt->mnt_point->mnt = NULL;
+

     llist_delete(&mnt->sibmnts);
     llist_delete(&mnt->list);
     llist_delete(&mnt->sibmnts);
     llist_delete(&mnt->list);

-    atomic_fetch_sub(&mnt->mnt_point->ref_count, 1);

     vfree(mnt);
 }
 
     vfree(mnt);
 }
 

-int
+static inline void
+__detach_node_cache_ref(struct hbucket* bucket)
+{
+    if (!bucket->head) {
+        return;
+    }
+
+    bucket->head->pprev = 0;
+}
+
+static int

 __vfs_do_unmount(struct v_mount* mnt)
 {
     int errno = 0;
 __vfs_do_unmount(struct v_mount* mnt)
 {
     int errno = 0;


@@ -66,17 +86,18 @@ __vfs_do_unmount(struct v_mount* mnt)
 
     // detached the inodes from cache, and let lru policy to recycle them
     for (size_t i = 0; i < VFS_HASHTABLE_SIZE; i++) {
 
     // detached the inodes from cache, and let lru policy to recycle them
     for (size_t i = 0; i < VFS_HASHTABLE_SIZE; i++) {

-        struct hbucket* bucket = &sb->i_cache[i];
-        if (!bucket) {
-            continue;
-        }
-        bucket->head->pprev = 0;
+        __detach_node_cache_ref(&sb->i_cache[i]);
+        __detach_node_cache_ref(&sb->d_cache[i]);

     }
 
     }
 

-    mnt->mnt_point->mnt = mnt->parent;
+    struct v_dnode *pos, *next;
+    llist_for_each(pos, next, &mnt->mnt_point->children, siblings)
+    {
+        vfs_d_free(pos);
+    }

 
 

-    vfs_sb_free(sb);
-    __vfs_release_vmnt(mnt);
+    __vfs_detach_vmnt(mnt);
+    vfs_d_assign_vmnt(mnt->mnt_point, mnt->parent);

 
     return errno;
 }
 
     return errno;
 }


@@ -138,12 +159,13 @@ vfs_unmount(const char* target)
     return errno;
 }
 
     return errno;
 }
 

-int
-vfs_mount_at(const char* fs_name,
-             struct device* device,
-             struct v_dnode* mnt_point,
-             int options)
+static int
+vfs_mount_fsat(struct filesystem* fs,
+               struct device* device,
+               struct v_dnode* mnt_point,
+               int options)

 {
 {

+

     if (device && device->dev_type != DEV_IFVOL) {
         return ENOTBLK;
     }
     if (device && device->dev_type != DEV_IFVOL) {
         return ENOTBLK;
     }


@@ -152,11 +174,6 @@ vfs_mount_at(const char* fs_name,
         return ENOTDIR;
     }
 
         return ENOTDIR;
     }
 

-    struct filesystem* fs = fsm_get(fs_name);
-    if (!fs) {
-        return ENODEV;
-    }
-

     if ((fs->types & FSTYPE_ROFS)) {
         options |= MNT_RO;
     }
     if ((fs->types & FSTYPE_ROFS)) {
         options |= MNT_RO;
     }


@@ -166,45 +183,84 @@ vfs_mount_at(const char* fs_name,
     }
 
     int errno = 0;
     }
 
     int errno = 0;

-    char* dev_name = "sys";
-    struct v_mount* parent_mnt = mnt_point->mnt;
-    struct v_superblock *sb = vfs_sb_alloc(), 
-                        *old_sb = mnt_point->super_block;
+    char* dev_name;
+    char* fsname;
+    struct v_mount *parent_mnt, *vmnt;
+    struct v_superblock *sb;

 
 

-    if (device) {
-        dev_name = device->name_val;
-    }
+    fsname = HSTR_VAL(fs->fs_name);
+    parent_mnt = mnt_point->mnt;
+    sb = vfs_sb_alloc();
+
+    dev_name = device ? device->name_val : "sys";

 
     // prepare v_superblock for fs::mount invoke
     sb->dev = device;
     sb->fs = fs;
     sb->root = mnt_point;
 
     // prepare v_superblock for fs::mount invoke
     sb->dev = device;
     sb->fs = fs;
     sb->root = mnt_point;

-    vfs_d_assign_sb(mnt_point, sb);
+    
+    vmnt = __vfs_create_mount(parent_mnt, sb);

 
 

-    if (!(mnt_point->mnt = vfs_create_mount(parent_mnt, mnt_point))) {
+    if (!vmnt) {

         errno = ENOMEM;
         goto cleanup;
     }
 
         errno = ENOMEM;
         goto cleanup;
     }
 

+    __vfs_attach_vmnt(mnt_point, vmnt);
+

     mnt_point->mnt->flags = options;
     if (!(errno = fs->mount(sb, mnt_point))) {
     mnt_point->mnt->flags = options;
     if (!(errno = fs->mount(sb, mnt_point))) {

-        kprintf("mount: dev=%s, fs=%s, mode=%d", dev_name, fs_name, options);
+        kprintf("mount: dev=%s, fs=%s, mode=%d", 
+                    dev_name, fsname, options);

     } else {
         goto cleanup;
     }
 
     } else {
         goto cleanup;
     }
 

-    vfs_sb_free(old_sb);

     return errno;
 
 cleanup:
     ERROR("failed mount: dev=%s, fs=%s, mode=%d, err=%d",
     return errno;
 
 cleanup:
     ERROR("failed mount: dev=%s, fs=%s, mode=%d, err=%d",

-          dev_name,
-          fs_name,
-          options,
-          errno);
-    vfs_d_assign_sb(mnt_point, old_sb);
-    vfs_sb_free(sb);
-    __vfs_release_vmnt(mnt_point->mnt);
+            dev_name, fsname, options, errno);
+
+    __vfs_detach_vmnt(mnt_point->mnt);
+    vfs_d_assign_vmnt(mnt_point, parent_mnt);
+
+    mnt_point->mnt = parent_mnt;
+
+    return errno;
+}
+
+int
+vfs_mount_at(const char* fs_name,
+             struct device* device,
+             struct v_dnode* mnt_point,
+             int options)
+{
+    if (fs_name) {
+        struct filesystem* fs = fsm_get(fs_name);
+        if (!fs) {
+            return ENODEV;
+        }
+
+        return vfs_mount_fsat(fs, device, mnt_point, options);
+    }
+
+    int errno = ENODEV;
+    struct fs_iter fsi;
+
+    fsm_itbegin(&fsi);
+    while (fsm_itnext(&fsi))
+    {
+        if ((fsi.fs->types & FSTYPE_PSEUDO)) {
+            continue;
+        }
+
+        INFO("mount attempt: %s", HSTR_VAL(fsi.fs->fs_name));
+        errno = vfs_mount_fsat(fsi.fs, device, mnt_point, options);
+        if (!errno) {
+            break;
+        }
+    }

 
     return errno;
 }
 
     return errno;
 }


@@ -213,7 +269,9 @@ int
 vfs_unmount_at(struct v_dnode* mnt_point)
 {
     int errno = 0;
 vfs_unmount_at(struct v_dnode* mnt_point)
 {
     int errno = 0;

-    struct v_superblock* sb = mnt_point->super_block;
+    struct v_superblock* sb;
+
+    sb = mnt_point->super_block;

     if (!sb) {
         return EINVAL;
     }
     if (!sb) {
         return EINVAL;
     }


@@ -222,15 +280,11 @@ vfs_unmount_at(struct v_dnode* mnt_point)
         return EINVAL;
     }
 
         return EINVAL;
     }
 

-    if (mnt_point->mnt->busy_counter) {
+    if (mnt_check_busy(mnt_point->mnt)) {

         return EBUSY;
     }
 
         return EBUSY;
     }
 

-    if (!(errno = __vfs_do_unmount(mnt_point->mnt))) {
-        atomic_fetch_sub(&mnt_point->ref_count, 1);
-    }
-
-    return errno;
+    return __vfs_do_unmount(mnt_point->mnt);

 }
 
 int
 }
 
 int


@@ -239,6 +293,11 @@ vfs_check_writable(struct v_dnode* dnode)
     if ((dnode->mnt->flags & MNT_RO)) {
         return EROFS;
     }
     if ((dnode->mnt->flags & MNT_RO)) {
         return EROFS;
     }

+
+    if (!check_allow_write(dnode->inode)) {
+        return EPERM;
+    }
+

     return 0;
 }
 
     return 0;
 }
 


@@ -253,6 +312,7 @@ __DEFINE_LXSYSCALL4(int,
                     int,
                     options)
 {
                     int,
                     options)
 {

+    struct device* device = NULL;

     struct v_dnode *dev = NULL, *mnt = NULL;
     int errno = 0;
 
     struct v_dnode *dev = NULL, *mnt = NULL;
     int errno = 0;
 


@@ -273,16 +333,14 @@ __DEFINE_LXSYSCALL4(int,
         goto done;
     }
 
         goto done;
     }
 

-    // By our convention.
-    // XXX could we do better?
-    struct device* device = NULL;
-

     if (dev) {
         if (!check_voldev_node(dev->inode)) {
             errno = ENOTDEV;
             goto done;
         }
     if (dev) {
         if (!check_voldev_node(dev->inode)) {
             errno = ENOTDEV;
             goto done;
         }

-        device = (struct device*)dev->inode->data;
+
+        device = resolve_device(dev->inode->data);
+        assert(device);

     }
 
     errno = vfs_mount_at(fstype, device, mnt, options);
     }
 
     errno = vfs_mount_at(fstype, device, mnt, options);