X-Git-Url: https://scm.lunaixsky.com/lunaix-os.git/blobdiff_plain/270869139db617e29a35bb9ded41087bb702f9ac..0fd474df7001837bde53da0e42e83081827c9641:/lunaix-os/kernel/fs/vfs.c

diff --git a/lunaix-os/kernel/fs/vfs.c b/lunaix-os/kernel/fs/vfs.c
index 415e36b..32014d3 100644
--- a/lunaix-os/kernel/fs/vfs.c
+++ b/lunaix-os/kernel/fs/vfs.c
@@ -57,14 +57,16 @@
 
 #include <usr/lunaix/dirent_defs.h>
 
+#define INODE_ACCESSED  0
+#define INODE_MODIFY    1
+
 static struct cake_pile* dnode_pile;
 static struct cake_pile* inode_pile;
 static struct cake_pile* file_pile;
 static struct cake_pile* superblock_pile;
 static struct cake_pile* fd_pile;
 
-struct v_dnode* vfs_sysroot;
-static struct hbucket* dnode_cache;
+struct v_dnode* vfs_sysroot = NULL;
 
 struct lru_zone *dnode_lru, *inode_lru;
 
@@ -89,8 +91,6 @@ vfs_init()
     superblock_pile =
       cake_new_pile("sb_cache", sizeof(struct v_superblock), 1, 0);
 
-    dnode_cache = vzalloc(VFS_HASHTABLE_SIZE * sizeof(struct hbucket));
-
     dnode_lru = lru_new_zone("vfs_dnode", __vfs_try_evict_dnode);
     inode_lru = lru_new_zone("vfs_inode", __vfs_try_evict_inode);
 
@@ -100,19 +100,23 @@ vfs_init()
     // 创建一个根dnode。
     vfs_sysroot = vfs_d_alloc(NULL, &vfs_empty);
     vfs_sysroot->parent = vfs_sysroot;
-    atomic_fetch_add(&vfs_sysroot->ref_count, 1);
+    
+    vfs_ref_dnode(vfs_sysroot);
 }
 
 static inline struct hbucket*
 __dcache_hash(struct v_dnode* parent, u32_t* hash)
 {
-    u32_t _hash = *hash;
-    // 确保低位更加随机
+    struct hbucket* d_cache;
+    u32_t _hash;
+    
+    d_cache = parent->super_block->d_cache;
+    _hash = *hash;
     _hash = _hash ^ (_hash >> VFS_HASHBITS);
-    // 与parent的指针值做加法，来减小碰撞的可能性。
     _hash += (u32_t)__ptr(parent);
+
     *hash = _hash;
-    return &dnode_cache[_hash & VFS_HASH_MASK];
+    return &d_cache[_hash & VFS_HASH_MASK];
 }
 
 static inline int
@@ -151,12 +155,26 @@ vfs_dcache_lookup(struct v_dnode* parent, struct hstr* str)
     return NULL;
 }
 
+static void
+__vfs_touch_inode(struct v_inode* inode, const int type)
+{
+    if (type == INODE_MODIFY) {
+        inode->mtime = clock_unixtime();
+    }
+
+    else if (type == INODE_ACCESSED) {
+        inode->atime = clock_unixtime();
+    }
+
+    lru_use_one(inode_lru, &inode->lru);
+}
+
 void
 vfs_dcache_add(struct v_dnode* parent, struct v_dnode* dnode)
 {
     assert(parent);
 
-    atomic_fetch_add(&dnode->ref_count, 1);
+    dnode->ref_count = 1;
     dnode->parent = parent;
     llist_append(&parent->children, &dnode->siblings);
 
@@ -175,7 +193,7 @@ vfs_dcache_remove(struct v_dnode* dnode)
     hlist_delete(&dnode->hash_list);
 
     dnode->parent = NULL;
-    atomic_fetch_sub(&dnode->ref_count, 1);
+    dnode->ref_count = 0;
 }
 
 void
@@ -191,12 +209,12 @@ vfs_dcache_rehash(struct v_dnode* new_parent, struct v_dnode* dnode)
 int
 vfs_open(struct v_dnode* dnode, struct v_file** file)
 {
-    if (!dnode->inode || !dnode->inode->ops->open) {
+    struct v_inode* inode = dnode->inode;
+    
+    if (!inode || !inode->ops->open) {
         return ENOTSUP;
     }
 
-    struct v_inode* inode = dnode->inode;
-
     lock_inode(inode);
 
     struct v_file* vfile = cake_grab(file_pile);
@@ -204,10 +222,10 @@ vfs_open(struct v_dnode* dnode, struct v_file** file)
 
     vfile->dnode = dnode;
     vfile->inode = inode;
-    vfile->ref_count = ATOMIC_VAR_INIT(1);
+    vfile->ref_count = 1;
     vfile->ops = inode->default_fops;
 
-    if (check_file_node(inode) && !inode->pg_cache) {
+    if (check_regfile_node(inode) && !inode->pg_cache) {
         struct pcache* pcache = vzalloc(sizeof(struct pcache));
         pcache_init(pcache);
         pcache->master = inode;
@@ -218,9 +236,8 @@ vfs_open(struct v_dnode* dnode, struct v_file** file)
     if (errno) {
         cake_release(file_pile, vfile);
     } else {
-        atomic_fetch_add(&dnode->ref_count, 1);
+        vfs_ref_dnode(dnode);
         inode->open_count++;
-        mnt_mkbusy(dnode->mnt);
 
         *file = vfile;
     }
@@ -270,12 +287,7 @@ vfs_pclose(struct v_file* file, pid_t pid)
 {
     struct v_inode* inode;
     int errno = 0;
-    
-    if (file->ref_count > 1) {
-        atomic_fetch_sub(&file->ref_count, 1);
-        return 0;
-    } 
-    
+
     inode = file->inode;
 
     /*
@@ -296,29 +308,41 @@ vfs_pclose(struct v_file* file, pid_t pid)
      * process is writing to this file later after B exit.
     */
 
-    if (mutex_on_hold(&inode->lock)) {
-        mutex_unlock_for(&inode->lock, pid);
+    mutex_unlock_for(&inode->lock, pid);
+    
+    if (vfs_check_duped_file(file)) {
+        vfs_unref_file(file);
+        return 0;
     }
 
-    lock_inode(inode);
-
-    pcache_commit_all(inode);
     if ((errno = file->ops->close(file))) {
-        goto unlock;
+        goto done;
     }
 
-    atomic_fetch_sub(&file->dnode->ref_count, 1);
+    vfs_unref_dnode(file->dnode);
+    cake_release(file_pile, file);
+
+    /*
+        if the current inode is not being locked by other 
+        threads that does not share same open context,
+        then we can try to do sync opportunistically
+    */
+    if (mutex_on_hold(&inode->lock)) {
+        goto done;
+    }
+    
+    lock_inode(inode);
+
+    pcache_commit_all(inode);
     inode->open_count--;
 
     if (!inode->open_count) {
         __sync_inode_nolock(inode);
     }
 
-    mnt_chillax(file->dnode->mnt);
-    cake_release(file_pile, file);
-
-unlock:
     unlock_inode(inode);
+
+done:
     return errno;
 }
 
@@ -375,7 +399,10 @@ vfs_sb_alloc()
     struct v_superblock* sb = cake_grab(superblock_pile);
     memset(sb, 0, sizeof(*sb));
     llist_init_head(&sb->sb_list);
+    
     sb->i_cache = vzalloc(VFS_HASHTABLE_SIZE * sizeof(struct hbucket));
+    sb->d_cache = vzalloc(VFS_HASHTABLE_SIZE * sizeof(struct hbucket));
+
     sb->ref_count = 1;
     return sb;
 }
@@ -387,12 +414,12 @@ vfs_sb_ref(struct v_superblock* sb)
 }
 
 void
-vfs_sb_free(struct v_superblock* sb)
+vfs_sb_unref(struct v_superblock* sb)
 {
     assert(sb->ref_count);
 
     sb->ref_count--;
-    if (sb->ref_count) {
+    if (likely(sb->ref_count)) {
         return;
     }
 
@@ -401,6 +428,8 @@ vfs_sb_free(struct v_superblock* sb)
     }
 
     vfree(sb->i_cache);
+    vfree(sb->d_cache);
+    
     cake_release(superblock_pile, sb);
 }
 
@@ -446,7 +475,6 @@ vfs_d_alloc(struct v_dnode* parent, struct hstr* name)
     llist_init_head(&dnode->aka_list);
     mutex_init(&dnode->lock);
 
-    dnode->ref_count = ATOMIC_VAR_INIT(0);
     dnode->name = HHSTR(vzalloc(VFS_NAME_MAXLEN), 0, 0);
 
     hstrcpy(&dnode->name, name);
@@ -485,7 +513,7 @@ vfs_d_free(struct v_dnode* dnode)
         dnode->destruct(dnode);
     }
 
-    vfs_sb_free(dnode->super_block);
+    vfs_sb_unref(dnode->super_block);
     vfree((void*)dnode->name.value);
     cake_release(dnode_pile, dnode);
 }
@@ -558,7 +586,7 @@ vfs_i_free(struct v_inode* inode)
         inode->destruct(inode);
     }
 
-    vfs_sb_free(inode->sb);
+    vfs_sb_unref(inode->sb);
     hlist_delete(&inode->hash_list);
     cake_release(inode_pile, inode);
 }
@@ -685,6 +713,31 @@ done:
     return errno;
 }
 
+
+static bool
+__check_unlinkable(struct v_dnode* dnode)
+{
+    int acl;
+    bool wr_self, wr_parent;
+    struct v_dnode* parent;
+
+    parent = dnode->parent;
+    acl = dnode->inode->acl;
+
+    wr_self = check_allow_write(dnode->inode);
+    wr_parent = check_allow_write(parent->inode);
+
+    if (!fsacl_test(acl, svtx)) {
+        return wr_self;
+    }
+
+    if (current_euid() == dnode->inode->uid) {
+        return true;
+    }
+
+    return wr_self && wr_parent;
+}
+
 int
 vfs_do_open(const char* path, int options)
 {
@@ -803,6 +856,11 @@ __DEFINE_LXSYSCALL2(int, sys_readdir, int, fd, struct lx_dirent*, dent)
         goto unlock;
     }
 
+    if (!check_allow_read(inode)) {
+        errno = EPERM;
+        goto unlock;
+    }
+
     struct dir_context dctx = (struct dir_context) {
         .cb_data = dent,
         .read_complete_callback = __vfs_readdir_callback
@@ -825,6 +883,8 @@ __DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
 {
     int errno = 0;
     struct v_fd* fd_s;
+    struct v_inode* inode;
+
     if ((errno = vfs_getfd(fd, &fd_s))) {
         goto done;
     }
@@ -835,23 +895,29 @@ __DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
         goto done;
     }
 
-    lock_inode(file->inode);
+    if (!check_allow_read(file->inode)) {
+        errno = EPERM;
+        goto done;
+    }
+
+    inode = file->inode;
+    lock_inode(inode);
 
-    file->inode->atime = clock_unixtime();
+    __vfs_touch_inode(inode, INODE_ACCESSED);
 
-    if (check_seqdev_node(file->inode) || (fd_s->flags & FO_DIRECT)) {
-        errno = file->ops->read(file->inode, buf, count, file->f_pos);
+    if (check_seqdev_node(inode) || (fd_s->flags & FO_DIRECT)) {
+        errno = file->ops->read(inode, buf, count, file->f_pos);
     } else {
-        errno = pcache_read(file->inode, buf, count, file->f_pos);
+        errno = pcache_read(inode, buf, count, file->f_pos);
     }
 
     if (errno > 0) {
         file->f_pos += errno;
-        unlock_inode(file->inode);
+        unlock_inode(inode);
         return errno;
     }
 
-    unlock_inode(file->inode);
+    unlock_inode(inode);
 
 done:
     return DO_STATUS(errno);
@@ -880,7 +946,7 @@ __DEFINE_LXSYSCALL3(int, write, int, fd, void*, buf, size_t, count)
     inode = file->inode;
     lock_inode(inode);
 
-    inode->mtime = clock_unixtime();
+    __vfs_touch_inode(inode, INODE_MODIFY);
     if ((fd_s->flags & O_APPEND)) {
         file->f_pos = inode->fsize;
     }
@@ -921,6 +987,11 @@ __DEFINE_LXSYSCALL3(int, lseek, int, fd, int, offset, int, options)
         goto done;
     }
 
+    if (!check_allow_read(inode)) {
+        errno = EPERM;
+        goto done;
+    }
+
     lock_inode(inode);
 
     int overflow = 0;
@@ -1001,6 +1072,10 @@ vfs_readlink(struct v_dnode* dnode, char* buf, size_t size)
         return ENOTSUP;
     }
 
+    if (!check_allow_read(inode)) {
+        return EPERM;
+    }
+
     lock_inode(inode);
 
     int errno = inode->ops->read_symlink(inode, &link);
@@ -1107,6 +1182,11 @@ __DEFINE_LXSYSCALL1(int, rmdir, const char*, pathname)
 
     lock_dnode(dnode);
 
+    if (!__check_unlinkable(dnode)) {
+        errno = EPERM;
+        goto done;
+    } 
+
     if ((errno = vfs_check_writable(dnode))) {
         goto done;
     }
@@ -1203,7 +1283,7 @@ done:
     return DO_STATUS(errno);
 }
 
-int
+static int
 __vfs_do_unlink(struct v_dnode* dnode)
 {
     int errno;
@@ -1213,6 +1293,10 @@ __vfs_do_unlink(struct v_dnode* dnode)
         return EBUSY;
     }
 
+    if (!__check_unlinkable(dnode)) {
+        return EPERM;
+    } 
+
     if ((errno = vfs_check_writable(dnode))) {
         return errno;
     }
@@ -1316,7 +1400,7 @@ vfs_dup_fd(struct v_fd* old, struct v_fd** new)
 
     memcpy(copied, old, sizeof(struct v_fd));
 
-    atomic_fetch_add(&old->file->ref_count, 1);
+    vfs_ref_file(old->file);
 
     *new = copied;
 
@@ -1423,53 +1507,34 @@ done:
     return DO_STATUS(errno);
 }
 
-void
-vfs_ref_file(struct v_file* file)
-{
-    atomic_fetch_add(&file->ref_count, 1);
-}
-
-void
-vfs_ref_dnode(struct v_dnode* dnode)
+static int
+vfs_do_chdir_nolock(struct proc_info* proc, struct v_dnode* dnode)
 {
-    atomic_fetch_add(&dnode->ref_count, 1);
-    
-    if (dnode->mnt) {
-        mnt_mkbusy(dnode->mnt);
+    if (!check_directory_node(dnode->inode)) {
+        return ENOTDIR;
     }
-}
 
-void
-vfs_unref_dnode(struct v_dnode* dnode)
-{
-    atomic_fetch_sub(&dnode->ref_count, 1);
-    if (dnode->mnt) {
-        mnt_chillax(dnode->mnt);
+    if (proc->cwd) {
+        vfs_unref_dnode(proc->cwd);
     }
+
+    vfs_ref_dnode(dnode);
+    proc->cwd = dnode;
+
+    return 0;
 }
 
-int
+static int
 vfs_do_chdir(struct proc_info* proc, struct v_dnode* dnode)
 {
     int errno = 0;
 
     lock_dnode(dnode);
 
-    if (!check_directory_node(dnode->inode)) {
-        errno = ENOTDIR;
-        goto done;
-    }
-
-    if (proc->cwd) {
-        vfs_unref_dnode(proc->cwd);
-    }
-
-    vfs_ref_dnode(dnode);
-    proc->cwd = dnode;
+    errno = vfs_do_chdir_nolock(proc, dnode);
 
     unlock_dnode(dnode);
 
-done:
     return errno;
 }
 
@@ -1503,6 +1568,31 @@ done:
     return DO_STATUS(errno);
 }
 
+
+__DEFINE_LXSYSCALL1(int, chroot, const char*, path)
+{
+    int errno;
+    struct v_dnode* dnode;
+    if ((errno = vfs_walk_proc(path, &dnode, NULL, 0))) {
+        return errno;
+    }
+
+    lock_dnode(dnode);
+
+    errno = vfs_do_chdir_nolock(__current, dnode);
+    if (errno) {
+        unlock_dnode(dnode);
+        goto done;
+    }
+
+    __current->root = dnode;
+    
+    unlock_dnode(dnode);
+
+done:
+    return DO_STATUS(errno);
+}
+
 __DEFINE_LXSYSCALL2(char*, getcwd, char*, buf, size_t, size)
 {
     int errno = 0;
@@ -1643,29 +1733,140 @@ __DEFINE_LXSYSCALL2(int, fstat, int, fd, struct file_stat*, stat)
     struct v_inode* vino = fds->file->inode;
     struct device* fdev = vino->sb->dev;
 
-    *stat = (struct file_stat){.st_ino = vino->id,
-                               .st_blocks = vino->lb_usage,
-                               .st_size = vino->fsize,
-                               .mode = vino->itype,
-                               .st_ioblksize = PAGE_SIZE,
-                               .st_blksize = vino->sb->blksize};
+    stat->st_ino     = vino->id;
+    stat->st_blocks  = vino->lb_usage;
+    stat->st_size    = vino->fsize;
+    stat->st_blksize = vino->sb->blksize;
+    stat->st_nlink   = vino->link_count;
+    stat->st_uid     = vino->uid;
+    stat->st_gid     = vino->gid;
+
+    stat->st_ctim    = vino->ctime;
+    stat->st_atim    = vino->atime;
+    stat->st_mtim    = vino->mtime;
+
+    stat->st_mode    = (vino->itype << 16) | vino->acl;
+
+    stat->st_ioblksize = PAGE_SIZE;
 
     if (check_device_node(vino)) {
         struct device* rdev = resolve_device(vino->data);
-        if (!rdev || rdev->magic != DEV_STRUCT_MAGIC) {
+        if (!rdev) {
             errno = EINVAL;
             goto done;
         }
 
         stat->st_rdev = (dev_t){.meta = rdev->ident.fn_grp,
                                 .unique = rdev->ident.unique,
-                                .index = rdev->dev_uid};
+                                .index = dev_uid(rdev) };
     }
 
     if (fdev) {
         stat->st_dev = (dev_t){.meta = fdev->ident.fn_grp,
                                .unique = fdev->ident.unique,
-                               .index = fdev->dev_uid};
+                               .index = dev_uid(fdev) };
+    }
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL4(int, fchmodat, int, fd, 
+                    const char*, path, int, mode, int, flags)
+{
+    int errno;
+    struct v_dnode *dnode;
+    struct v_inode* inode;
+
+    errno = vfs_walkat(fd, path, flags, &dnode);
+    if (errno) {
+        goto done;
+    }
+
+    errno = vfs_check_writable(dnode);
+    if (errno) {
+        return errno;
+    }
+
+    inode = dnode->inode;
+    lock_inode(inode);
+
+    if (!current_is_root()) {
+        mode = mode & FSACL_RWXMASK;
+    }
+
+    inode->acl = mode;
+    __vfs_touch_inode(inode, INODE_MODIFY);
+
+    unlock_inode(inode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL5(int, fchownat, int, fd, 
+                    const char*, path, uid_t, uid, gid_t, gid, int, flags)
+{
+    int errno;
+    struct v_dnode *dnode;
+    struct v_inode *inode;
+
+    errno = vfs_walkat(fd, path, flags, &dnode);
+    if (errno) {
+        goto done;
+    }
+
+    errno = vfs_check_writable(dnode);
+    if (errno) {
+        return errno;
+    }
+
+    inode = dnode->inode;
+    lock_inode(inode);
+
+    inode->uid = uid;
+    inode->gid = gid;
+    __vfs_touch_inode(inode, INODE_MODIFY);
+
+    unlock_inode(inode);
+
+done:
+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL4(int, faccessat, int, fd, 
+                    const char*, path, int, amode, int, flags)
+{
+    int errno, acl;
+    struct v_dnode *dnode;
+    struct v_inode *inode;
+    struct user_scope* uscope;
+
+    uid_t tuid;
+    gid_t tgid;
+
+    errno = vfs_walkat(fd, path, flags, &dnode);
+    if (errno) {
+        goto done;
+    }
+
+    if ((flags & AT_EACCESS)) {
+        tuid = current_euid();
+        tgid = current_egid();
+    }
+    else {
+        uscope = current_user_scope();
+        tuid = uscope->ruid;
+        tgid = uscope->rgid;
+    }
+
+    inode = dnode->inode;
+
+    acl  = inode->acl;
+    acl &= amode;
+    acl &= check_acl_between(inode->uid, inode->gid, tuid, tgid);
+    if (!acl) {
+        errno = EACCESS;
     }
 
 done: