circuitpython/extmod/vfs_fat.c

550 lines
19 KiB
C
Raw Normal View History

/*
* This file is part of the MicroPython project, http://micropython.org/
*
* The MIT License (MIT)
*
* Copyright (c) 2014 Damien P. George
* Copyright (c) 2016 Paul Sokolovsky
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "py/mpconfig.h"
#if MICROPY_VFS_FAT
#if !MICROPY_ENABLE_FINALISER
#error "MICROPY_VFS_FAT requires MICROPY_ENABLE_FINALISER"
#endif
#if !MICROPY_VFS
#error "with MICROPY_VFS_FAT enabled, must also enable MICROPY_VFS"
#endif
#include <string.h>
#include "py/obj.h"
#include "py/objproperty.h"
#include "py/runtime.h"
#include "py/mperrno.h"
#include "lib/oofatfs/ff.h"
#include "extmod/vfs_fat.h"
#include "shared/timeutils/timeutils.h"
#include "supervisor/filesystem.h"
#if FF_MAX_SS == FF_MIN_SS
#define SECSIZE(fs) (FF_MIN_SS)
#else
#define SECSIZE(fs) ((fs)->ssize)
#endif
#define mp_obj_fat_vfs_t fs_user_mount_t
// Factoring this common call saves about 90 bytes.
STATIC NORETURN void mp_raise_OSError_fresult(FRESULT res) {
mp_raise_OSError(fresult_to_errno_table[res]);
}
STATIC mp_import_stat_t fat_vfs_import_stat(void *vfs_in, const char *path) {
fs_user_mount_t *vfs = vfs_in;
FILINFO fno;
assert(vfs != NULL);
FRESULT res = f_stat(&vfs->fatfs, path, &fno);
if (res == FR_OK) {
if ((fno.fattrib & AM_DIR) != 0) {
return MP_IMPORT_STAT_DIR;
} else {
return MP_IMPORT_STAT_FILE;
}
}
return MP_IMPORT_STAT_NO_EXIST;
}
STATIC mp_obj_t fat_vfs_make_new(const mp_obj_type_t *type, size_t n_args, size_t n_kw, const mp_obj_t *args) {
mp_arg_check_num(n_args, n_kw, 1, 1, false);
// create new object
fs_user_mount_t *vfs = mp_obj_malloc(fs_user_mount_t, type);
vfs->fatfs.drv = vfs;
// Initialise underlying block device
vfs->blockdev.flags = MP_BLOCKDEV_FLAG_FREE_OBJ;
vfs->blockdev.block_size = FF_MIN_SS; // default, will be populated by call to MP_BLOCKDEV_IOCTL_BLOCK_SIZE
mp_vfs_blockdev_init(&vfs->blockdev, args[0]);
// mount the block device so the VFS methods can be used
FRESULT res = f_mount(&vfs->fatfs);
if (res == FR_NO_FILESYSTEM) {
// don't error out if no filesystem, to let mkfs()/mount() create one if wanted
vfs->blockdev.flags |= MP_BLOCKDEV_FLAG_NO_FILESYSTEM;
} else if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return MP_OBJ_FROM_PTR(vfs);
}
STATIC void verify_fs_writable(fs_user_mount_t *vfs) {
if (!filesystem_is_writable_by_python(vfs)) {
mp_raise_OSError(MP_EROFS);
}
}
#if FF_FS_REENTRANT
STATIC mp_obj_t fat_vfs_del(mp_obj_t self_in) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(self_in);
// f_umount only needs to be called to release the sync object
f_umount(&self->fatfs);
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(fat_vfs_del_obj, fat_vfs_del);
#endif
STATIC mp_obj_t fat_vfs_mkfs(mp_obj_t bdev_in) {
// create new object
fs_user_mount_t *vfs = MP_OBJ_TO_PTR(fat_vfs_make_new(&mp_fat_vfs_type, 1, 0, &bdev_in));
// make the filesystem
uint8_t working_buf[FF_MAX_SS];
FRESULT res = f_mkfs(&vfs->fatfs, FM_FAT | FM_SFD, 0, working_buf, sizeof(working_buf));
if (res == FR_MKFS_ABORTED) { // Probably doesn't support FAT16
res = f_mkfs(&vfs->fatfs, FM_FAT32, 0, working_buf, sizeof(working_buf));
}
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(fat_vfs_mkfs_fun_obj, fat_vfs_mkfs);
STATIC MP_DEFINE_CONST_STATICMETHOD_OBJ(fat_vfs_mkfs_obj, MP_ROM_PTR(&fat_vfs_mkfs_fun_obj));
typedef struct _mp_vfs_fat_ilistdir_it_t {
mp_obj_base_t base;
mp_fun_1_t iternext;
2023-09-22 14:38:46 -04:00
mp_fun_1_t finaliser;
bool is_str;
FF_DIR dir;
} mp_vfs_fat_ilistdir_it_t;
STATIC mp_obj_t mp_vfs_fat_ilistdir_it_iternext(mp_obj_t self_in) {
mp_vfs_fat_ilistdir_it_t *self = MP_OBJ_TO_PTR(self_in);
for (;;) {
FILINFO fno;
FRESULT res = f_readdir(&self->dir, &fno);
char *fn = fno.fname;
if (res != FR_OK || fn[0] == 0) {
// stop on error or end of dir
break;
}
// Note that FatFS already filters . and .., so we don't need to
// make 4-tuple with info about this entry
mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(4, NULL));
if (self->is_str) {
t->items[0] = mp_obj_new_str(fn, strlen(fn));
} else {
2021-03-15 09:57:36 -04:00
t->items[0] = mp_obj_new_bytes((const byte *)fn, strlen(fn));
}
if (fno.fattrib & AM_DIR) {
// dir
t->items[1] = MP_OBJ_NEW_SMALL_INT(MP_S_IFDIR);
} else {
// file
t->items[1] = MP_OBJ_NEW_SMALL_INT(MP_S_IFREG);
}
t->items[2] = MP_OBJ_NEW_SMALL_INT(0); // no inode number
t->items[3] = mp_obj_new_int_from_uint(fno.fsize);
return MP_OBJ_FROM_PTR(t);
}
// ignore error because we may be closing a second time
f_closedir(&self->dir);
return MP_OBJ_STOP_ITERATION;
}
STATIC mp_obj_t mp_vfs_fat_ilistdir_it_del(mp_obj_t self_in) {
mp_vfs_fat_ilistdir_it_t *self = MP_OBJ_TO_PTR(self_in);
// ignore result / error because we may be closing a second time.
f_closedir(&self->dir);
return mp_const_none;
}
STATIC mp_obj_t fat_vfs_ilistdir_func(size_t n_args, const mp_obj_t *args) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(args[0]);
bool is_str_type = true;
const char *path;
if (n_args == 2) {
if (mp_obj_get_type(args[1]) == &mp_type_bytes) {
is_str_type = false;
}
path = mp_obj_str_get_str(args[1]);
} else {
path = "";
}
// Create a new iterator object to list the dir
mp_vfs_fat_ilistdir_it_t *iter = m_new_obj_with_finaliser(mp_vfs_fat_ilistdir_it_t);
iter->base.type = &mp_type_polymorph_iter_with_finaliser;
iter->iternext = mp_vfs_fat_ilistdir_it_iternext;
iter->finaliser = mp_vfs_fat_ilistdir_it_del;
iter->is_str = is_str_type;
FRESULT res = f_opendir(&self->fatfs, &iter->dir, path);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return MP_OBJ_FROM_PTR(iter);
}
STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(fat_vfs_ilistdir_obj, 1, 2, fat_vfs_ilistdir_func);
STATIC mp_obj_t fat_vfs_remove_internal(mp_obj_t vfs_in, mp_obj_t path_in, mp_int_t attr) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
verify_fs_writable(self);
const char *path = mp_obj_str_get_str(path_in);
FILINFO fno;
FRESULT res = f_stat(&self->fatfs, path, &fno);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
// check if path is a file or directory
if ((fno.fattrib & AM_DIR) == attr) {
res = f_unlink(&self->fatfs, path);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return mp_const_none;
} else {
mp_raise_OSError(attr ? MP_ENOTDIR : MP_EISDIR);
}
}
STATIC mp_obj_t fat_vfs_remove(mp_obj_t vfs_in, mp_obj_t path_in) {
return fat_vfs_remove_internal(vfs_in, path_in, 0); // 0 == file attribute
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_remove_obj, fat_vfs_remove);
STATIC mp_obj_t fat_vfs_rmdir(mp_obj_t vfs_in, mp_obj_t path_in) {
return fat_vfs_remove_internal(vfs_in, path_in, AM_DIR);
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_rmdir_obj, fat_vfs_rmdir);
2016-02-28 18:22:38 -05:00
STATIC mp_obj_t fat_vfs_rename(mp_obj_t vfs_in, mp_obj_t path_in, mp_obj_t path_out) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
verify_fs_writable(self);
2016-02-28 18:22:38 -05:00
const char *old_path = mp_obj_str_get_str(path_in);
const char *new_path = mp_obj_str_get_str(path_out);
FRESULT res = f_rename(&self->fatfs, old_path, new_path);
if (res == FR_EXIST) {
// if new_path exists then try removing it (but only if it's a file)
fat_vfs_remove_internal(vfs_in, path_out, 0); // 0 == file attribute
// try to rename again
res = f_rename(&self->fatfs, old_path, new_path);
}
if (res == FR_OK) {
return mp_const_none;
} else {
mp_raise_OSError_fresult(res);
2016-02-28 18:22:38 -05:00
}
}
STATIC MP_DEFINE_CONST_FUN_OBJ_3(fat_vfs_rename_obj, fat_vfs_rename);
2016-02-28 17:02:50 -05:00
STATIC mp_obj_t fat_vfs_mkdir(mp_obj_t vfs_in, mp_obj_t path_o) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
verify_fs_writable(self);
2016-02-28 17:02:50 -05:00
const char *path = mp_obj_str_get_str(path_o);
FRESULT res = f_mkdir(&self->fatfs, path);
if (res == FR_OK) {
return mp_const_none;
} else {
mp_raise_OSError_fresult(res);
2016-02-28 17:02:50 -05:00
}
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_mkdir_obj, fat_vfs_mkdir);
// Change current directory.
STATIC mp_obj_t fat_vfs_chdir(mp_obj_t vfs_in, mp_obj_t path_in) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
const char *path;
path = mp_obj_str_get_str(path_in);
FRESULT res = f_chdir(&self->fatfs, path);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_chdir_obj, fat_vfs_chdir);
// Get the current directory.
STATIC mp_obj_t fat_vfs_getcwd(mp_obj_t vfs_in) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
char buf[MICROPY_ALLOC_PATH_MAX + 1];
FRESULT res = f_getcwd(&self->fatfs, buf, sizeof(buf));
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return mp_obj_new_str(buf, strlen(buf));
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(fat_vfs_getcwd_obj, fat_vfs_getcwd);
// Get the status of a file or directory.
STATIC mp_obj_t fat_vfs_stat(mp_obj_t vfs_in, mp_obj_t path_in) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
const char *path = mp_obj_str_get_str(path_in);
FILINFO fno;
if (path[0] == 0 || (path[0] == '/' && path[1] == 0)) {
// stat root directory
fno.fsize = 0;
fno.fdate = 0x2821; // Jan 1, 2000
fno.ftime = 0;
fno.fattrib = AM_DIR;
} else {
FRESULT res = f_stat(&self->fatfs, path, &fno);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
}
mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(10, NULL));
mp_int_t mode = 0;
if (fno.fattrib & AM_DIR) {
mode |= MP_S_IFDIR;
} else {
mode |= MP_S_IFREG;
}
2021-03-15 09:57:36 -04:00
#if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_NONE
// On non-longint builds, the number of seconds since 1970 (epoch) is too
// large to fit in a smallint, so just return 31-DEC-1999 (0).
mp_obj_t seconds = MP_OBJ_NEW_SMALL_INT(946684800);
2021-03-15 09:57:36 -04:00
#else
mp_obj_t seconds = mp_obj_new_int_from_uint(
timeutils_seconds_since_epoch(
1980 + ((fno.fdate >> 9) & 0x7f),
(fno.fdate >> 5) & 0x0f,
fno.fdate & 0x1f,
(fno.ftime >> 11) & 0x1f,
(fno.ftime >> 5) & 0x3f,
2 * (fno.ftime & 0x1f)
));
2021-03-15 09:57:36 -04:00
#endif
t->items[0] = MP_OBJ_NEW_SMALL_INT(mode); // st_mode
t->items[1] = MP_OBJ_NEW_SMALL_INT(0); // st_ino
t->items[2] = MP_OBJ_NEW_SMALL_INT(0); // st_dev
t->items[3] = MP_OBJ_NEW_SMALL_INT(0); // st_nlink
t->items[4] = MP_OBJ_NEW_SMALL_INT(0); // st_uid
t->items[5] = MP_OBJ_NEW_SMALL_INT(0); // st_gid
t->items[6] = mp_obj_new_int_from_uint(fno.fsize); // st_size
t->items[7] = seconds; // st_atime
t->items[8] = seconds; // st_mtime
t->items[9] = seconds; // st_ctime
return MP_OBJ_FROM_PTR(t);
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_stat_obj, fat_vfs_stat);
// Get the status of a VFS.
STATIC mp_obj_t fat_vfs_statvfs(mp_obj_t vfs_in, mp_obj_t path_in) {
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
(void)path_in;
DWORD nclst;
FATFS *fatfs = &self->fatfs;
FRESULT res = f_getfree(fatfs, &nclst);
if (FR_OK != res) {
mp_raise_OSError_fresult(res);
}
mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(10, NULL));
t->items[0] = MP_OBJ_NEW_SMALL_INT(fatfs->csize * SECSIZE(fatfs)); // f_bsize
t->items[1] = t->items[0]; // f_frsize
t->items[2] = MP_OBJ_NEW_SMALL_INT((fatfs->n_fatent - 2)); // f_blocks
t->items[3] = MP_OBJ_NEW_SMALL_INT(nclst); // f_bfree
t->items[4] = t->items[3]; // f_bavail
t->items[5] = MP_OBJ_NEW_SMALL_INT(0); // f_files
t->items[6] = MP_OBJ_NEW_SMALL_INT(0); // f_ffree
t->items[7] = MP_OBJ_NEW_SMALL_INT(0); // f_favail
t->items[8] = MP_OBJ_NEW_SMALL_INT(0); // f_flags
t->items[9] = MP_OBJ_NEW_SMALL_INT(FF_MAX_LFN); // f_namemax
return MP_OBJ_FROM_PTR(t);
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_statvfs_obj, fat_vfs_statvfs);
STATIC mp_obj_t vfs_fat_mount(mp_obj_t self_in, mp_obj_t readonly, mp_obj_t mkfs) {
fs_user_mount_t *self = MP_OBJ_TO_PTR(self_in);
// Read-only device indicated by writeblocks[0] == MP_OBJ_NULL.
// User can specify read-only device by:
// 1. readonly=True keyword argument
// 2. nonexistent writeblocks method (then writeblocks[0] == MP_OBJ_NULL already)
if (mp_obj_is_true(readonly)) {
self->blockdev.writeblocks[0] = MP_OBJ_NULL;
}
// check if we need to make the filesystem
FRESULT res = (self->blockdev.flags & MP_BLOCKDEV_FLAG_NO_FILESYSTEM) ? FR_NO_FILESYSTEM : FR_OK;
if (res == FR_NO_FILESYSTEM && mp_obj_is_true(mkfs)) {
uint8_t working_buf[FF_MAX_SS];
res = f_mkfs(&self->fatfs, FM_FAT | FM_SFD, 0, working_buf, sizeof(working_buf));
}
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
self->blockdev.flags &= ~MP_BLOCKDEV_FLAG_NO_FILESYSTEM;
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_3(vfs_fat_mount_obj, vfs_fat_mount);
STATIC mp_obj_t vfs_fat_umount(mp_obj_t self_in) {
(void)self_in;
// keep the FAT filesystem mounted internally so the VFS methods can still be used
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(fat_vfs_umount_obj, vfs_fat_umount);
STATIC mp_obj_t vfs_fat_utime(mp_obj_t vfs_in, mp_obj_t path_in, mp_obj_t times_in) {
2022-09-18 21:43:27 -04:00
mp_obj_fat_vfs_t *self = MP_OBJ_TO_PTR(vfs_in);
const char *path = mp_obj_str_get_str(path_in);
if (!mp_obj_is_tuple_compatible(times_in)) {
mp_raise_type_arg(&mp_type_TypeError, times_in);
}
2022-09-18 21:43:27 -04:00
2022-09-22 12:06:02 -04:00
mp_obj_t *otimes;
mp_obj_get_array_fixed_n(times_in, 2, &otimes);
// Validate that both elements of the tuple are int and discard the second one
int time[2];
time[0] = mp_obj_get_int(otimes[0]);
time[1] = mp_obj_get_int(otimes[1]);
2022-09-18 21:43:27 -04:00
timeutils_struct_time_t tm;
2022-09-22 12:06:02 -04:00
timeutils_seconds_since_epoch_to_struct_time(time[0], &tm);
2022-09-18 21:43:27 -04:00
FILINFO fno;
fno.fdate = (WORD)(((tm.tm_year - 1980) * 512U) | tm.tm_mon * 32U | tm.tm_mday);
fno.ftime = (WORD)(tm.tm_hour * 2048U | tm.tm_min * 32U | tm.tm_sec / 2U);
FRESULT res = f_utime(&self->fatfs, path, &fno);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_3(fat_vfs_utime_obj, vfs_fat_utime);
STATIC mp_obj_t vfs_fat_getreadonly(mp_obj_t self_in) {
fs_user_mount_t *self = MP_OBJ_TO_PTR(self_in);
return mp_obj_new_bool(!filesystem_is_writable_by_python(self));
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(fat_vfs_getreadonly_obj, vfs_fat_getreadonly);
STATIC const mp_obj_property_t fat_vfs_readonly_obj = {
.base.type = &mp_type_property,
.proxy = {(mp_obj_t)&fat_vfs_getreadonly_obj,
MP_ROM_NONE,
MP_ROM_NONE},
};
#if MICROPY_FATFS_USE_LABEL
STATIC mp_obj_t vfs_fat_getlabel(mp_obj_t self_in) {
fs_user_mount_t *self = MP_OBJ_TO_PTR(self_in);
char working_buf[12];
FRESULT res = f_getlabel(&self->fatfs, working_buf, NULL);
if (res != FR_OK) {
mp_raise_OSError_fresult(res);
}
return mp_obj_new_str(working_buf, strlen(working_buf));
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(fat_vfs_getlabel_obj, vfs_fat_getlabel);
STATIC mp_obj_t vfs_fat_setlabel(mp_obj_t self_in, mp_obj_t label_in) {
fs_user_mount_t *self = MP_OBJ_TO_PTR(self_in);
verify_fs_writable(self);
const char *label_str = mp_obj_str_get_str(label_in);
FRESULT res = f_setlabel(&self->fatfs, label_str);
if (res != FR_OK) {
2021-03-15 09:57:36 -04:00
if (res == FR_WRITE_PROTECTED) {
mp_raise_msg(&mp_type_OSError, MP_ERROR_TEXT("Read-only filesystem"));
}
mp_raise_OSError_fresult(res);
}
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(fat_vfs_setlabel_obj, vfs_fat_setlabel);
STATIC const mp_obj_property_t fat_vfs_label_obj = {
.base.type = &mp_type_property,
.proxy = {(mp_obj_t)&fat_vfs_getlabel_obj,
(mp_obj_t)&fat_vfs_setlabel_obj,
MP_ROM_NONE},
};
#endif
STATIC const mp_rom_map_elem_t fat_vfs_locals_dict_table[] = {
#if FF_FS_REENTRANT
{ MP_ROM_QSTR(MP_QSTR___del__), MP_ROM_PTR(&fat_vfs_del_obj) },
#endif
{ MP_ROM_QSTR(MP_QSTR_mkfs), MP_ROM_PTR(&fat_vfs_mkfs_obj) },
{ MP_ROM_QSTR(MP_QSTR_open), MP_ROM_PTR(&fat_vfs_open_obj) },
{ MP_ROM_QSTR(MP_QSTR_ilistdir), MP_ROM_PTR(&fat_vfs_ilistdir_obj) },
2016-02-28 17:02:50 -05:00
{ MP_ROM_QSTR(MP_QSTR_mkdir), MP_ROM_PTR(&fat_vfs_mkdir_obj) },
{ MP_ROM_QSTR(MP_QSTR_rmdir), MP_ROM_PTR(&fat_vfs_rmdir_obj) },
{ MP_ROM_QSTR(MP_QSTR_chdir), MP_ROM_PTR(&fat_vfs_chdir_obj) },
{ MP_ROM_QSTR(MP_QSTR_getcwd), MP_ROM_PTR(&fat_vfs_getcwd_obj) },
{ MP_ROM_QSTR(MP_QSTR_remove), MP_ROM_PTR(&fat_vfs_remove_obj) },
2016-02-28 18:22:38 -05:00
{ MP_ROM_QSTR(MP_QSTR_rename), MP_ROM_PTR(&fat_vfs_rename_obj) },
{ MP_ROM_QSTR(MP_QSTR_stat), MP_ROM_PTR(&fat_vfs_stat_obj) },
{ MP_ROM_QSTR(MP_QSTR_statvfs), MP_ROM_PTR(&fat_vfs_statvfs_obj) },
{ MP_ROM_QSTR(MP_QSTR_mount), MP_ROM_PTR(&vfs_fat_mount_obj) },
{ MP_ROM_QSTR(MP_QSTR_umount), MP_ROM_PTR(&fat_vfs_umount_obj) },
2022-09-18 21:43:27 -04:00
{ MP_ROM_QSTR(MP_QSTR_utime), MP_ROM_PTR(&fat_vfs_utime_obj) },
{ MP_ROM_QSTR(MP_QSTR_readonly), MP_ROM_PTR(&fat_vfs_readonly_obj) },
2021-03-15 09:57:36 -04:00
#if MICROPY_FATFS_USE_LABEL
{ MP_ROM_QSTR(MP_QSTR_label), MP_ROM_PTR(&fat_vfs_label_obj) },
2021-03-15 09:57:36 -04:00
#endif
};
STATIC MP_DEFINE_CONST_DICT(fat_vfs_locals_dict, fat_vfs_locals_dict_table);
STATIC const mp_vfs_proto_t fat_vfs_proto = {
protocols: Allow them to be (optionally) type-safe Protocols are nice, but there is no way for C code to verify whether a type's "protocol" structure actually implements some particular protocol. As a result, you can pass an object that implements the "vfs" protocol to one that expects the "stream" protocol, and the opposite of awesomeness ensues. This patch adds an OPTIONAL (but enabled by default) protocol identifier as the first member of any protocol structure. This identifier is simply a unique QSTR chosen by the protocol designer and used by each protocol implementer. When checking for protocol support, instead of just checking whether the object's type has a non-NULL protocol field, use `mp_proto_get` which implements the protocol check when possible. The existing protocols are now named: protocol_framebuf protocol_i2c protocol_pin protocol_stream protocol_spi protocol_vfs (most of these are unused in CP and are just inherited from MP; vfs and stream are definitely used though) I did not find any crashing examples, but here's one to give a flavor of what is improved, using `micropython_coverage`. Before the change, the vfs "ioctl" protocol is invoked, and the result is not intelligible as json (but it could have resulted in a hard fault, potentially): >>> import uos, ujson >>> u = uos.VfsPosix('/tmp') >>> ujson.load(u) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: syntax error in JSON After the change, the vfs object is correctly detected as not supporting the stream protocol: >>> ujson.load(p) Traceback (most recent call last): File "<stdin>", line 1, in <module> OSError: stream operation not supported
2019-12-03 15:50:37 -05:00
MP_PROTO_IMPLEMENT(MP_QSTR_protocol_vfs)
.import_stat = fat_vfs_import_stat,
};
MP_DEFINE_CONST_OBJ_TYPE(
mp_fat_vfs_type,
MP_QSTR_VfsFat,
MP_TYPE_FLAG_HAS_SPECIAL_ACCESSORS,
make_new, fat_vfs_make_new,
protocol, &fat_vfs_proto,
locals_dict, &fat_vfs_locals_dict
);
#endif // MICROPY_VFS_FAT