fix: separate any i/o to sequential device from caching layer

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

diff --git a/lunaix-os/kernel/fs/vfs.c b/lunaix-os/kernel/fs/vfs.c
index 05741d846f15df903cca7e9b524acce835eb5112..53d21d8b026baa7f07ce38bc34a12cda78c0ee90 100644 (file)
--- a/lunaix-os/kernel/fs/vfs.c
+++ b/lunaix-os/kernel/fs/vfs.c
@@ -72,6 +72,7 @@ vfs_init()
     // 创建一个根superblock，用来蕴含我们的根目录。
     root_sb = vfs_sb_alloc();
     root_sb->root = vfs_d_alloc();
     // 创建一个根superblock，用来蕴含我们的根目录。
     root_sb = vfs_sb_alloc();
     root_sb->root = vfs_d_alloc();

+    root_sb->root->inode = vfs_i_alloc();

 }
 
 inline struct hbucket*
 }
 
 inline struct hbucket*


@@ -114,11 +115,11 @@ vfs_dcache_add(struct v_dnode* parent, struct v_dnode* dnode)
 }
 
 int
 }
 
 int

-vfs_walk(struct v_dnode* start,
-         const char* path,
-         struct v_dnode** dentry,
-         struct hstr* component,
-         int walk_options)
+__vfs_walk(struct v_dnode* start,
+           const char* path,
+           struct v_dnode** dentry,
+           struct hstr* component,
+           int walk_options)

 {
     int errno = 0;
     int i = 0, j = 0;
 {
     int errno = 0;
     int i = 0, j = 0;


@@ -146,7 +147,7 @@ vfs_walk(struct v_dnode* start,
                 return ENAMETOOLONG;
             }
             if (!VFS_VALID_CHAR(current)) {
                 return ENAMETOOLONG;
             }
             if (!VFS_VALID_CHAR(current)) {

-                return VFS_EINVLD;
+                return EINVAL;

             }
             name_content[j++] = current;
             if (lookahead) {
             }
             name_content[j++] = current;
             if (lookahead) {


@@ -218,6 +219,44 @@ error:
     return errno;
 }
 
     return errno;
 }
 

+#define VFS_MAX_SYMLINK 16
+
+int
+vfs_walk(struct v_dnode* start,
+         const char* path,
+         struct v_dnode** dentry,
+         struct hstr* component,
+         int options)
+{
+    struct v_dnode* interim;
+    char* pathname = path;
+    int errno = __vfs_walk(start, path, &interim, component, options);
+    int counter = 0;
+
+    while (!errno) {
+        if (counter >= VFS_MAX_SYMLINK) {
+            errno = ELOOP;
+            continue;
+        }
+        if ((interim->inode->itype & VFS_IFSYMLINK) &&
+            !(options & VFS_WALK_NOFOLLOW) &&
+            interim->inode->ops.read_symlink) {
+            errno = interim->inode->ops.read_symlink(interim->inode, &pathname);
+            if (errno) {
+                break;
+            }
+        } else {
+            break;
+        }
+        errno = __vfs_walk(start, pathname, &interim, component, options);
+        counter++;
+    }
+
+    *dentry = errno ? 0 : interim;
+
+    return errno;
+}
+

 int
 vfs_mount(const char* target, const char* fs_name, bdev_t device)
 {
 int
 vfs_mount(const char* target, const char* fs_name, bdev_t device)
 {


@@ -249,7 +288,7 @@ vfs_mount_at(const char* fs_name, bdev_t device, struct v_dnode* mnt_point)
 {
     struct filesystem* fs = fsm_get(fs_name);
     if (!fs)
 {
     struct filesystem* fs = fsm_get(fs_name);
     if (!fs)

-        return VFS_ENOFS;
+        return ENODEV;

     struct v_superblock* sb = vfs_sb_alloc();
     sb->dev = device;
     sb->fs_id = fs_id++;
     struct v_superblock* sb = vfs_sb_alloc();
     sb->dev = device;
     sb->fs_id = fs_id++;


@@ -271,7 +310,7 @@ vfs_unmount_at(struct v_dnode* mnt_point)
     int errno = 0;
     struct v_superblock* sb = mnt_point->super_block;
     if (!sb) {
     int errno = 0;
     struct v_superblock* sb = mnt_point->super_block;
     if (!sb) {

-        return VFS_EBADMNT;
+        return EINVAL;

     }
     if (!(errno = sb->fs->unmount(sb))) {
         struct v_dnode* fs_root = sb->root;
     }
     if (!(errno = sb->fs->unmount(sb))) {
         struct v_dnode* fs_root = sb->root;


@@ -289,14 +328,23 @@ vfs_open(struct v_dnode* dnode, struct v_file** file)
         return ENOTSUP;
     }
 
         return ENOTSUP;
     }
 

+    struct v_inode* inode = dnode->inode;

     struct v_file* vfile = cake_grab(file_pile);
     memset(vfile, 0, sizeof(*vfile));
 
     vfile->dnode = dnode;
     struct v_file* vfile = cake_grab(file_pile);
     memset(vfile, 0, sizeof(*vfile));
 
     vfile->dnode = dnode;

-    vfile->inode = dnode->inode;
-    dnode->inode->open_count++;
+    vfile->inode = inode;
+    vfile->ref_count = 1;
+    vfile->ops = inode->default_fops;
+    inode->open_count++;
+
+    if ((inode->itype & VFS_IFFILE) && !inode->pg_cache) {
+        struct pcache* pcache = vzalloc(sizeof(struct pcache));
+        pcache_init(pcache);
+        inode->pg_cache = pcache;
+    }

 
 

-    int errno = dnode->inode->ops.open(dnode->inode, vfile);
+    int errno = inode->ops.open(inode, vfile);

     if (errno) {
         cake_release(file_pile, vfile);
     } else {
     if (errno) {
         cake_release(file_pile, vfile);
     } else {


@@ -324,15 +372,12 @@ vfs_link(struct v_dnode* to_link, struct v_dnode* name)
 int
 vfs_close(struct v_file* file)
 {
 int
 vfs_close(struct v_file* file)
 {

-    if (!file->ops.close) {
-        return ENOTSUP;
-    }
-
-    int errno = file->ops.close(file);
-    if (!errno) {
+    int errno = 0;
+    if (!file->ops.close || !(errno = file->ops.close(file))) {

         if (file->inode->open_count) {
             file->inode->open_count--;
         }
         if (file->inode->open_count) {
             file->inode->open_count--;
         }

+        pcache_commit_all(file);

         cake_release(file_pile, file);
     }
     return errno;
         cake_release(file_pile, file);
     }
     return errno;


@@ -342,6 +387,7 @@ int
 vfs_fsync(struct v_file* file)
 {
     int errno = ENOTSUP;
 vfs_fsync(struct v_file* file)
 {
     int errno = ENOTSUP;

+    pcache_commit_all(file);

     if (file->ops.sync) {
         errno = file->ops.sync(file);
     }
     if (file->ops.sync) {
         errno = file->ops.sync(file);
     }


@@ -402,7 +448,6 @@ vfs_i_alloc()
 {
     struct v_inode* inode = cake_grab(inode_pile);
     memset(inode, 0, sizeof(*inode));
 {
     struct v_inode* inode = cake_grab(inode_pile);
     memset(inode, 0, sizeof(*inode));

-

     return inode;
 }
 
     return inode;
 }
 


@@ -417,6 +462,7 @@ vfs_i_free(struct v_inode* inode)
 #define FLOCATE_CREATE_EMPTY 1
 
 #define DO_STATUS(errno) SYSCALL_ESTATUS(__current->k_status = errno)
 #define FLOCATE_CREATE_EMPTY 1
 
 #define DO_STATUS(errno) SYSCALL_ESTATUS(__current->k_status = errno)

+#define DO_STATUS_OR_RETURN(errno) ({ errno < 0 ? DO_STATUS(errno) : errno; })

 
 int
 __vfs_try_locate_file(const char* path,
 
 int
 __vfs_try_locate_file(const char* path,


@@ -427,71 +473,98 @@ __vfs_try_locate_file(const char* path,
     char name_str[VFS_NAME_MAXLEN];
     struct hstr name = HSTR(name_str, 0);
     int errno;
     char name_str[VFS_NAME_MAXLEN];
     struct hstr name = HSTR(name_str, 0);
     int errno;

-    if (!(errno = vfs_walk(NULL, path, fdir, &name, VFS_WALK_PARENT))) {
-        errno = vfs_walk(*fdir, name.value, file, NULL, 0);
-        if (errno == ENOENT && (options & FLOCATE_CREATE_EMPTY)) {
-            struct v_dnode* file_new;
-            file_new = vfs_d_alloc();
-            file_new->name =
-              HHSTR(valloc(VFS_NAME_MAXLEN), name.len, name.hash);
-            strcpy(file_new->name.value, name_str);
-            *file = file_new;
-
-            llist_append(&(*fdir)->children, &file_new->siblings);
-        }
+    if ((errno = vfs_walk(NULL, path, fdir, &name, VFS_WALK_PARENT))) {
+        return errno;
+    }
+
+    errno = vfs_walk(*fdir, name.value, file, NULL, 0);
+    if (errno != ENOENT || !(options & FLOCATE_CREATE_EMPTY)) {
+        return errno;
+    }
+
+    if (!(errno = (*fdir)->inode->ops.create((*fdir)->inode))) {
+        struct v_dnode* file_new;
+        file_new = vfs_d_alloc();
+        file_new->name = HHSTR(valloc(VFS_NAME_MAXLEN), name.len, name.hash);
+        strcpy(file_new->name.value, name_str);
+        *file = file_new;
+
+        llist_append(&(*fdir)->children, &file_new->siblings);

     }
 
     return errno;
 }
 
 int
     }
 
     return errno;
 }
 
 int

-__vfs_do_open(struct v_file** file_out, const char* path, int options)
+__file_cached_read(struct v_file* file, void* buf, size_t len, size_t fpos)

 {
 {

-    int errno;
+    return pcache_read(file, buf, len, fpos);
+}
+
+int
+vfs_do_open(const char* path, int options)
+{
+    int errno, fd;

     struct v_dnode *dentry, *file;
     struct v_dnode *dentry, *file;

-    struct v_file* opened_file = 0;
+    struct v_file* ofile = 0;

 
     errno = __vfs_try_locate_file(path, &dentry, &file, 0);
 
     if (errno != ENOENT && (options & FO_CREATE)) {
 
     errno = __vfs_try_locate_file(path, &dentry, &file, 0);
 
     if (errno != ENOENT && (options & FO_CREATE)) {

-        errno = dentry->inode->ops.create(dentry->inode, opened_file);
-    } else if (!errno) {
-        errno = vfs_open(file, &opened_file);
+        errno = dentry->inode->ops.create(dentry->inode);

     }
 
     }
 

-    *file_out = opened_file;
-    return errno;
-}
+    if (!errno && (errno = vfs_open(file, &ofile))) {
+        return errno;
+    }

 
 

-__DEFINE_LXSYSCALL2(int, open, const char*, path, int, options)
-{
-    struct v_file* opened_file;
-    int errno = __vfs_do_open(&opened_file, path, options), fd;
+    struct v_inode* o_inode = ofile->inode;
+    if (!(o_inode->itype & VFS_IFSEQDEV) && !(options & FO_DIRECT)) {
+        // XXX Change here accordingly when signature of pcache_r/w changed.
+        ofile->ops.read = pcache_read;
+        ofile->ops.write = pcache_write;
+    }

 
     if (!errno && !(errno = vfs_alloc_fdslot(&fd))) {
         struct v_fd* fd_s = vzalloc(sizeof(*fd_s));
 
     if (!errno && !(errno = vfs_alloc_fdslot(&fd))) {
         struct v_fd* fd_s = vzalloc(sizeof(*fd_s));

-        fd_s->file = opened_file;
-        fd_s->pos = opened_file->inode->fsize & -((options & FO_APPEND) != 0);
+        ofile->f_pos = ofile->inode->fsize & -((options & FO_APPEND) != 0);
+        fd_s->file = ofile;
+        fd_s->flags = options;

         __current->fdtable->fds[fd] = fd_s;
         return fd;
     }
 
         __current->fdtable->fds[fd] = fd_s;
         return fd;
     }
 

-    return DO_STATUS(errno);
+    return errno;
+}
+
+__DEFINE_LXSYSCALL2(int, open, const char*, path, int, options)
+{
+    int errno = vfs_do_open(path, options);
+    return DO_STATUS_OR_RETURN(errno);

 }
 }

+

 #define TEST_FD(fd) (fd >= 0 && fd < VFS_MAX_FD)
 #define GET_FD(fd, fd_s) (TEST_FD(fd) && (fd_s = __current->fdtable->fds[fd]))
 
 __DEFINE_LXSYSCALL1(int, close, int, fd)
 {
     struct v_fd* fd_s;
 #define TEST_FD(fd) (fd >= 0 && fd < VFS_MAX_FD)
 #define GET_FD(fd, fd_s) (TEST_FD(fd) && (fd_s = __current->fdtable->fds[fd]))
 
 __DEFINE_LXSYSCALL1(int, close, int, fd)
 {
     struct v_fd* fd_s;

-    int errno;
+    int errno = 0;

     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;

-    } else if (!(errno = vfs_close(fd_s->file))) {
-        vfree(fd_s);
-        __current->fdtable->fds[fd] = 0;
+        goto done_err;
+    }
+
+    if (fd_s->file->ref_count > 1) {
+        fd_s->file->ref_count--;
+    } else if ((errno = vfs_close(fd_s->file))) {
+        goto done_err;

     }
 
     }
 

+    vfree(fd_s);
+    __current->fdtable->fds[fd] = 0;
+
+done_err:

     return DO_STATUS(errno);
 }
 
     return DO_STATUS(errno);
 }
 


@@ -513,7 +586,7 @@ __DEFINE_LXSYSCALL2(int, readdir, int, fd, struct dirent*, dent)
     int errno;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;
     int errno;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;

-    } else if (!(fd_s->file->inode->itype & VFS_INODE_TYPE_DIR)) {
+    } else if (!(fd_s->file->inode->itype & VFS_IFDIR)) {

         errno = ENOTDIR;
     } else {
         struct dir_context dctx =
         errno = ENOTDIR;
     } else {
         struct dir_context dctx =


@@ -552,7 +625,7 @@ __DEFINE_LXSYSCALL1(int, mkdir, const char*, path)
         errno = ENOTSUP;
     } else if (!parent->inode->ops.mkdir) {
         errno = ENOTSUP;
         errno = ENOTSUP;
     } else if (!parent->inode->ops.mkdir) {
         errno = ENOTSUP;

-    } else if (!(parent->inode->itype & VFS_INODE_TYPE_DIR)) {
+    } else if (!(parent->inode->itype & VFS_IFDIR)) {

         errno = ENOTDIR;
     } else {
         dir = vfs_d_alloc();
         errno = ENOTDIR;
     } else {
         dir = vfs_d_alloc();


@@ -575,14 +648,23 @@ __DEFINE_LXSYSCALL3(int, read, int, fd, void*, buf, size_t, count)
     struct v_fd* fd_s;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;
     struct v_fd* fd_s;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;

-    } else {
-        struct v_file* file = fd_s->file;
-        file->f_pos = fd_s->pos;
-        if ((errno = file->ops.read(file, buf, count)) >= 0) {
-            fd_s->pos += errno;
-        }
+        goto done;

     }
 
     }
 

+    struct v_file* file = fd_s->file;
+    if ((file->inode->itype & VFS_IFDIR)) {
+        errno = EISDIR;
+        goto done;
+    }
+
+    errno = file->ops.read(file, buf, count, file->f_pos);
+
+    if (errno > 0) {
+        file->f_pos += errno;
+        return errno;
+    }
+
+done:

     return DO_STATUS(errno);
 }
 
     return DO_STATUS(errno);
 }
 


@@ -592,14 +674,23 @@ __DEFINE_LXSYSCALL3(int, write, int, fd, void*, buf, size_t, count)
     struct v_fd* fd_s;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;
     struct v_fd* fd_s;
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;

-    } else {
-        struct v_file* file = fd_s->file;
-        file->f_pos = fd_s->pos;
-        if ((errno = file->ops.write(file, buf, count)) >= 0) {
-            fd_s->pos += errno;
-        }
+        goto done;
+    }
+
+    struct v_file* file = fd_s->file;
+    if ((file->inode->itype & VFS_IFDIR)) {
+        errno = EISDIR;
+        goto done;
+    }
+
+    errno = file->ops.write(file, buf, count, file->f_pos);
+
+    if (errno > 0) {
+        file->f_pos += errno;
+        return errno;

     }
 
     }
 

+done:

     return DO_STATUS(errno);
 }
 
     return DO_STATUS(errno);
 }
 


@@ -610,29 +701,29 @@ __DEFINE_LXSYSCALL3(int, lseek, int, fd, int, offset, int, options)
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;
     } else {
     if (!GET_FD(fd, fd_s)) {
         errno = EBADF;
     } else {

-        size_t fpos = fd_s->file->f_pos;
+        struct v_file* file = fd_s->file;
+        size_t fpos = file->f_pos;

         switch (options) {
             case FSEEK_CUR:
         switch (options) {
             case FSEEK_CUR:

-                fpos = (size_t)((int)fd_s->file->f_pos + offset);
+                fpos = (size_t)((int)file->f_pos + offset);

                 break;
             case FSEEK_END:
                 break;
             case FSEEK_END:

-                fpos = (size_t)((int)fd_s->file->inode->fsize + offset);
+                fpos = (size_t)((int)file->inode->fsize + offset);

                 break;
             case FSEEK_SET:
                 fpos = offset;
                 break;
                 break;
             case FSEEK_SET:
                 fpos = offset;
                 break;

-
-            default:
-                break;

         }
         }

-        fd_s->pos = fpos;
+        if (!file->ops.seek || !(errno = file->ops.seek(file, fpos))) {
+            file->f_pos = fpos;
+        }

     }
 
     return DO_STATUS(errno);
 }
 
 int
     }
 
     return DO_STATUS(errno);
 }
 
 int

-vfs_readlink(struct v_dnode* dnode, char* buf, size_t size, int depth)
+vfs_get_path(struct v_dnode* dnode, char* buf, size_t size, int depth)

 {
     if (!dnode) {
         return 0;
 {
     if (!dnode) {
         return 0;


@@ -642,7 +733,7 @@ vfs_readlink(struct v_dnode* dnode, char* buf, size_t size, int depth)
         return ELOOP;
     }
 
         return ELOOP;
     }
 

-    size_t len = vfs_readlink(dnode->parent, buf, size, depth + 1);
+    size_t len = vfs_get_path(dnode->parent, buf, size, depth + 1);

 
     if (len >= size) {
         return len;
 
     if (len >= size) {
         return len;


@@ -659,12 +750,42 @@ vfs_readlink(struct v_dnode* dnode, char* buf, size_t size, int depth)
     return len;
 }
 
     return len;
 }
 

+int
+vfs_readlink(struct v_dnode* dnode, char* buf, size_t size)
+{
+    char* link;
+    if (dnode->inode->ops.read_symlink) {
+        int errno = dnode->inode->ops.read_symlink(dnode->inode, &link);
+        strncpy(buf, link, size);
+        return errno;
+    }
+    return 0;
+}
+
+__DEFINE_LXSYSCALL3(int, realpathat, int, fd, char*, buf, size_t, size)
+{
+    int errno;
+    struct v_fd* fd_s;
+    if (!GET_FD(fd, fd_s)) {
+        errno = EBADF;
+    } else {
+        struct v_dnode* dnode;
+        errno = vfs_get_path(fd_s->file->dnode, buf, size, 0);
+    }
+
+    if (errno >= 0) {
+        return errno;
+    }
+
+    return DO_STATUS(errno);
+}
+

 __DEFINE_LXSYSCALL3(int, readlink, const char*, path, char*, buf, size_t, size)
 {
     int errno;
     struct v_dnode* dnode;
 __DEFINE_LXSYSCALL3(int, readlink, const char*, path, char*, buf, size_t, size)
 {
     int errno;
     struct v_dnode* dnode;

-    if (!(errno = vfs_walk(NULL, path, &dnode, NULL, 0))) {
-        errno = vfs_readlink(dnode, buf, size, 0);
+    if (!(errno = vfs_walk(NULL, path, &dnode, NULL, VFS_WALK_NOFOLLOW))) {
+        errno = vfs_readlink(dnode, buf, size);

     }
 
     if (errno >= 0) {
     }
 
     if (errno >= 0) {


@@ -691,8 +812,12 @@ __DEFINE_LXSYSCALL4(int,
         errno = EBADF;
     } else {
         struct v_dnode* dnode;
         errno = EBADF;
     } else {
         struct v_dnode* dnode;

-        if (!(errno = vfs_walk(fd_s->file->dnode, pathname, &dnode, NULL, 0))) {
-            errno = vfs_readlink(fd_s->file->dnode, buf, size, 0);
+        if (!(errno = vfs_walk(fd_s->file->dnode,
+                               pathname,
+                               &dnode,
+                               NULL,
+                               VFS_WALK_NOFOLLOW))) {
+            errno = vfs_readlink(fd_s->file->dnode, buf, size);

         }
     }
 
         }
     }
 


@@ -720,7 +845,7 @@ __DEFINE_LXSYSCALL1(int, rmdir, const char*, pathname)
         goto done;
     }
 
         goto done;
     }
 

-    if ((dnode->inode->itype & VFS_INODE_TYPE_DIR)) {
+    if ((dnode->inode->itype & VFS_IFDIR)) {

         errno = dnode->inode->ops.rmdir(dnode->inode);
     } else {
         errno = ENOTDIR;
         errno = dnode->inode->ops.rmdir(dnode->inode);
     } else {
         errno = ENOTDIR;


@@ -736,7 +861,7 @@ __vfs_do_unlink(struct v_inode* inode)
     int errno;
     if (inode->open_count) {
         errno = EBUSY;
     int errno;
     if (inode->open_count) {
         errno = EBUSY;

-    } else if (!(inode->itype & VFS_INODE_TYPE_DIR)) {
+    } else if (!(inode->itype & VFS_IFDIR)) {

         // TODO handle symbolic link and type other than regular file
         errno = inode->ops.unlink(inode);
         if (!errno) {
         // TODO handle symbolic link and type other than regular file
         errno = inode->ops.unlink(inode);
         if (!errno) {


@@ -819,13 +944,10 @@ int
 __vfs_dup_fd(struct v_fd* old, struct v_fd** new)
 {
     int errno = 0;
 __vfs_dup_fd(struct v_fd* old, struct v_fd** new)
 {
     int errno = 0;

-    struct v_file* newopened;
-    if (!(errno = vfs_open(old->file->dnode, &newopened))) {
-        *new = cake_grab(fd_pile);
-        **new = (struct v_fd){ .file = newopened,
-                               .pos = old->pos,
-                               .flags = old->flags };
-    }
+    struct v_fd* copied = cake_grab(fd_pile);
+
+    memcpy(copied, old, sizeof(struct v_fd));
+    old->file->ref_count++;

 
     return errno;
 }
 
     return errno;
 }


@@ -868,5 +990,32 @@ __DEFINE_LXSYSCALL1(int, dup, int, oldfd)
         return newfd;
     }
 
         return newfd;
     }
 

+    return DO_STATUS(errno);
+}
+
+__DEFINE_LXSYSCALL2(int,
+                    symlink,
+                    const char*,
+                    pathname,
+                    const char*,
+                    link_target)
+{
+    int errno;
+    struct v_dnode* dnode;
+    if ((errno = vfs_walk(NULL, pathname, &dnode, NULL, 0))) {
+        goto done;
+    }
+    if ((dnode->super_block->fs->types & FSTYPE_ROFS)) {
+        errno = EROFS;
+        goto done;
+    }
+    if (!dnode->inode->ops.symlink) {
+        errno = ENOTSUP;
+        goto done;
+    }
+
+    errno = dnode->inode->ops.symlink(dnode->inode, link_target);
+
+done:

     return DO_STATUS(errno);
 }
\ No newline at end of file
     return DO_STATUS(errno);
 }
\ No newline at end of file