minix/servers/iso9660fs/inode.c
Thomas Veerman a9d0ffbd0d Define protocol version of {mode,ino,uid,gid}_t
The next commit will make mode_t 32-bits, ino_t 64-bits, nlink_t
32-bits, and off_t 64-bits. However, in the protocol there is no
room for that.

When analyzing the usage of these data types we can see that of
mode_t we currently use 15 bits and that future flags (if copied
from NetBSD) are only of importance to VFS and not to the actual
file system on disk. Therefore we can get away with keeping the
protocol version of mode_t (pmode_t) 16 bits.

At this point it is unlikely that MINIX will support file systems
that can hold more than 2^32 files. Keeping the protocol version of
ino_t (pino_t) 32-bits and exposing it as 64-bits to the rest of the
system is fine.

uid_t and gid_t are already 32-bits, but in behind the scenes they're
16-bits. This is sufficient since we don't support uid/gids larger
than 2^16-1.

off_t will become 64-bits, but the protocol had room reserved
for that since it was revised years ago.

So to make the difference clear between what the protocol supports
and what is exposed to the system, we define protocol version of
these data types.

Change-Id: Ifb883aafb222142e69d6dcc3049a47170b78b3ee
2013-03-28 19:39:21 +00:00

297 lines
8.8 KiB
C

/* This file contains all the function that handle the dir records
* (inodes) for the ISO9660 filesystem.*/
#include "inc.h"
#include "buf.h"
#include <minix/vfsif.h>
/*===========================================================================*
* fs_putnode *
*===========================================================================*/
int fs_putnode()
{
/* Find the inode specified by the request message and decrease its counter. */
int count;
struct dir_record *dir = NULL;
dir = get_dir_record(fs_m_in.REQ_INODE_NR);
release_dir_record(dir);
count = fs_m_in.REQ_COUNT;
if (count <= 0) return(EINVAL);
if (count > dir->d_count) {
printf("put_inode: count too high: %d > %d\n", count, dir->d_count);
return(EINVAL);
}
if (dir->d_count > 1)
dir->d_count = dir->d_count - count + 1;/*Keep at least one reference*/
release_dir_record(dir); /* Actual inode release, might be last reference */
return(OK);
}
/*===========================================================================*
* release_dir_record *
*===========================================================================*/
int release_dir_record(dir)
struct dir_record *dir;
{
/* Release a dir record (decrement the counter) */
if (dir == NULL)
return(EINVAL);
if (--dir->d_count == 0) {
if (dir->ext_attr != NULL)
dir->ext_attr->count = 0;
dir->ext_attr = NULL;
dir->d_mountpoint = FALSE;
dir->d_prior = NULL;
if (dir->d_next != NULL)
release_dir_record(dir);
dir->d_next = NULL;
}
return(OK);
}
/*===========================================================================*
* get_free_dir_record *
*===========================================================================*/
struct dir_record *get_free_dir_record(void)
{
/* Get a free dir record */
struct dir_record *dir;
for(dir = dir_records; dir < &dir_records[NR_ATTR_RECS]; dir++) {
if (dir->d_count == 0) { /* The record is free */
dir->d_count = 1; /* Set count to 1 */
dir->ext_attr = NULL;
return(dir);
}
}
return(NULL);
}
/*===========================================================================*
* get_dir_record *
*===========================================================================*/
struct dir_record *get_dir_record(id_dir_record)
pino_t id_dir_record;
{
struct dir_record *dir = NULL;
u32_t address;
int i;
/* Search through the cache if the inode is still present */
for(i = 0; i < NR_DIR_RECORDS && dir == NULL; ++i) {
if (dir_records[i].d_ino_nr == id_dir_record
&& dir_records[i].d_count > 0) {
dir = dir_records + i;
dir->d_count++;
}
}
if (dir == NULL) {
address = (u32_t)id_dir_record;
dir = load_dir_record_from_disk(address);
}
if (dir == NULL) return(NULL);
return(dir);
}
/*===========================================================================*
* get_free_ext_attr *
*===========================================================================*/
struct ext_attr_rec *get_free_ext_attr(void) {
/* Get a free extended attribute structure */
struct ext_attr_rec *dir;
for(dir = ext_attr_recs; dir < &ext_attr_recs[NR_ATTR_RECS]; dir++) {
if (dir->count == 0) { /* The record is free */
dir->count = 1;
return(dir);
}
}
return(NULL);
}
/*===========================================================================*
* create_ext_attr *
*===========================================================================*/
int create_ext_attr(struct ext_attr_rec *ext,char *buffer)
{
/* Fill an extent structure from the data read on the device */
if (ext == NULL) return(EINVAL);
/* In input we have a stream of bytes that are physically read from the
* device. This stream of data is copied to the data structure. */
memcpy(&ext->own_id,buffer,sizeof(u32_t));
memcpy(&ext->group_id,buffer + 4,sizeof(u32_t));
memcpy(&ext->permissions,buffer + 8,sizeof(u16_t));
memcpy(&ext->file_cre_date,buffer + 10,ISO9660_SIZE_VOL_CRE_DATE);
memcpy(&ext->file_mod_date,buffer + 27,ISO9660_SIZE_VOL_MOD_DATE);
memcpy(&ext->file_exp_date,buffer + 44,ISO9660_SIZE_VOL_EXP_DATE);
memcpy(&ext->file_eff_date,buffer + 61,ISO9660_SIZE_VOL_EFF_DATE);
memcpy(&ext->rec_format,buffer + 78,sizeof(u8_t));
memcpy(&ext->rec_attrs,buffer + 79,sizeof(u8_t));
memcpy(&ext->rec_length,buffer + 80,sizeof(u32_t));
memcpy(&ext->system_id,buffer + 84,ISO9660_SIZE_SYS_ID);
memcpy(&ext->system_use,buffer + 116,ISO9660_SIZE_SYSTEM_USE);
memcpy(&ext->ext_attr_rec_ver,buffer + 180,sizeof(u8_t));
memcpy(&ext->len_esc_seq,buffer + 181,sizeof(u8_t));
return(OK);
}
/*===========================================================================*
* create_ext_attr *
*===========================================================================*/
int create_dir_record(dir,buffer,address)
struct dir_record *dir;
char *buffer;
u32_t address;
{
/* Fills a dir record structure from the data read on the device */
/* If the flag assign id is active it will return the id associated;
* otherwise it will return OK. */
short size;
size = buffer[0];
if (dir == NULL) return(EINVAL);
/* The data structure dir record is filled with the stream of data
* that is read. */
dir->length = size;
dir->ext_attr_rec_length = *((u8_t*)buffer + 1);
memcpy(&dir->loc_extent_l,buffer + 2,sizeof(u32_t));
memcpy(&dir->loc_extent_m,buffer + 6,sizeof(u32_t));
memcpy(&dir->data_length_l,buffer + 10,sizeof(u32_t));
memcpy(&dir->data_length_m,buffer + 14,sizeof(u32_t));
memcpy(dir->rec_date,buffer + 18, sizeof(dir->rec_date));
dir->file_flags = *((u8_t*)buffer + 25);
dir->file_unit_size = *((u8_t*)buffer + 26);
dir->inter_gap_size = *((u8_t*)buffer + 27);
dir->vol_seq_number = *((u8_t*)buffer + 28);
dir->length_file_id = *((u8_t*)buffer + 32);
memcpy(dir->file_id,buffer + 33,dir->length_file_id);
dir->ext_attr = NULL;
/* set memory attrs */
if ((dir->file_flags & D_TYPE) == D_DIRECTORY)
dir->d_mode = I_DIRECTORY;
else
dir->d_mode = I_REGULAR;
/* Set permission to read only. Equal for all users. */
dir->d_mode |= R_BIT | X_BIT;
dir->d_mode |= R_BIT << 3 | X_BIT << 3;
dir->d_mode |= R_BIT << 6 | X_BIT << 6;
dir->d_mountpoint = FALSE;
dir->d_next = NULL;
dir->d_prior = NULL;
dir->d_file_size = dir->data_length_l;
/* Set physical address of the dir record */
dir->d_phy_addr = address;
dir->d_ino_nr = (pino_t) address; /* u32_t e ino_t are the same datatype so
* the cast is safe */
return(OK);
}
/*===========================================================================*
* load_dir_record_from_disk *
*===========================================================================*/
struct dir_record *load_dir_record_from_disk(address)
u32_t address;
{
/* This function load a particular dir record from a specific address
* on the device */
int block_nr, offset, block_size, new_pos;
struct buf *bp;
struct dir_record *dir, *dir_next, *dir_parent, *dir_tmp;
char name[NAME_MAX + 1];
char old_name[NAME_MAX + 1];
u32_t new_address, size;
block_size = v_pri.logical_block_size_l; /* Block size */
block_nr = address / block_size; /* Block number from the address */
offset = address % block_size; /* Offset starting from the block */
bp = get_block(block_nr); /* Read the block from the device */
if (bp == NULL)
return(NULL);
dir = get_free_dir_record(); /* Get a free record */
if (dir == NULL)
return(NULL);
/* Fill the dir record with the data read from the device */
create_dir_record(dir,b_data(bp) + offset, address);
/* In case the file is composed of more file sections, load also the
* next section into the structure */
new_pos = offset + dir->length;
dir_parent = dir;
new_address = address + dir->length;
while (new_pos < block_size) {
dir_next = get_free_dir_record();
create_dir_record(dir_next, b_data(bp) + new_pos, new_address);
if (dir_next->length > 0) {
strncpy(name,dir_next->file_id,dir_next->length_file_id);
name[dir_next->length_file_id] = '\0';
strncpy(old_name, dir_parent->file_id,
dir_parent->length_file_id);
old_name[dir_parent->length_file_id] = '\0';
if (strcmp(name, old_name) == 0) {
dir_parent->d_next = dir_next;
dir_next->d_prior = dir_parent;
/* Link the dir records */
dir_tmp = dir_next;
size = dir_tmp->data_length_l;
/* Update the file size */
while (dir_tmp->d_prior != NULL) {
dir_tmp = dir_tmp->d_prior;
size += dir_tmp->data_length_l;
dir_tmp->d_file_size = size;
}
new_pos += dir_parent->length;
new_address += dir_next->length;
dir_parent = dir_next;
} else { /* This is another inode. */
release_dir_record(dir_next);
new_pos = block_size;
}
} else { /* record not valid */
release_dir_record(dir_next);
new_pos = block_size; /* Exit from the while */
}
}
put_block(bp); /* Release the block read. */
return(dir);
}