minix/servers/mfs/read.c

1046 lines
32 KiB
C
Raw Normal View History

2005-04-21 16:53:53 +02:00
#include "fs.h"
#include <fcntl.h>
#include <stddef.h>
#include <string.h>
endpoint-aware conversion of servers. 'who', indicating caller number in pm and fs and some other servers, has been removed in favour of 'who_e' (endpoint) and 'who_p' (proc nr.). In both PM and FS, isokendpt() convert endpoints to process slot numbers, returning OK if it was a valid and consistent endpoint number. okendpt() does the same but panic()s if it doesn't succeed. (In PM, this is pm_isok..) pm and fs keep their own records of process endpoints in their proc tables, which are needed to make kernel calls about those processes. message field names have changed. fs drivers are endpoints. fs now doesn't try to get out of driver deadlock, as the protocol isn't supposed to let that happen any more. (A warning is printed if ELOCKED is detected though.) fproc[].fp_task (indicating which driver the process is suspended on) became an int. PM and FS now get endpoint numbers of initial boot processes from the kernel. These happen to be the same as the old proc numbers, to let user processes reach them with the old numbers, but FS and PM don't know that. All new processes after INIT, even after the generation number wraps around, get endpoint numbers with generation 1 and higher, so the first instances of the boot processes are the only processes ever to have endpoint numbers in the old proc number range. More return code checks of sys_* functions have been added. IS has become endpoint-aware. Ditched the 'text' and 'data' fields in the kernel dump (which show locations, not sizes, so aren't terribly useful) in favour of the endpoint number. Proc number is still visible. Some other dumps (e.g. dmap, rs) show endpoint numbers now too which got the formatting changed. PM reading segments using rw_seg() has changed - it uses other fields in the message now instead of encoding the segment and process number and fd in the fd field. For that it uses _read_pm() and _write_pm() which to _taskcall()s directly in pm/misc.c. PM now sys_exit()s itself on panic(), instead of sys_abort(). RS also talks in endpoints instead of process numbers.
2006-03-03 11:20:58 +01:00
#include <unistd.h>
2005-04-21 16:53:53 +02:00
#include <minix/com.h>
#include <minix/u64.h>
2005-04-21 16:53:53 +02:00
#include "buf.h"
#include "inode.h"
#include "super.h"
#include <minix/vfsif.h>
FORWARD _PROTOTYPE( int rw_chunk, (struct inode *rip, u64_t position,
2005-08-29 18:47:18 +02:00
unsigned off, int chunk, unsigned left, int rw_flag,
char *buff, int seg, int usr, int block_size, int *completed));
2007-08-07 14:38:35 +02:00
FORWARD _PROTOTYPE( int rw_chunk_s, (struct inode *rip, u64_t position,
unsigned off, int chunk, unsigned left, int rw_flag,
cp_grant_id_t gid, unsigned buf_off, int block_size, int *completed));
2005-04-21 16:53:53 +02:00
/*===========================================================================*
2007-08-07 14:38:35 +02:00
* fs_readwrite_o *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
2007-08-07 14:38:35 +02:00
PUBLIC int fs_readwrite_o(void)
2005-04-21 16:53:53 +02:00
{
int r, usr, seg, rw_flag, chunk, block_size, block_spec;
int partial_cnt, regular, partial_pipe, nrbytes;
off_t position, f_size, bytes_left;
2005-04-21 16:53:53 +02:00
unsigned int off, cum_io;
mode_t mode_word;
int completed, r2 = OK;
char *user_addr;
struct inode *rip;
partial_pipe = 0;
2005-04-21 16:53:53 +02:00
r = OK;
/* Try to get inode according to its index */
if (fs_m_in.REQ_FD_INODE_INDEX >= 0 &&
fs_m_in.REQ_FD_INODE_INDEX < NR_INODES &&
inode[fs_m_in.REQ_FD_INODE_INDEX].i_num == fs_m_in.REQ_FD_INODE_NR) {
rip = &inode[fs_m_in.REQ_FD_INODE_INDEX];
2005-04-21 16:53:53 +02:00
}
else {
/* Find the inode referred */
rip = find_inode(fs_dev, fs_m_in.REQ_FD_INODE_NR);
if (!rip) {
printf("FS: unavaliable inode by fs_readwrite(), nr: %d\n",
fs_m_in.REQ_FD_INODE_NR);
return EINVAL;
}
2005-04-21 16:53:53 +02:00
}
mode_word = rip->i_mode & I_TYPE;
regular = (mode_word == I_REGULAR || mode_word == I_NAMED_PIPE);
block_spec = (mode_word == I_BLOCK_SPECIAL ? 1 : 0);
/* Determine blocksize */
block_size = (block_spec ? get_block_size(rip->i_zone[0])
: rip->i_sp->s_block_size);
2005-04-21 16:53:53 +02:00
f_size = (block_spec ? ULONG_MAX : rip->i_size);
/* Get the values from the request message */
2007-08-07 14:38:35 +02:00
rw_flag = (fs_m_in.m_type == REQ_READ_O ? READING : WRITING);
usr = fs_m_in.REQ_FD_WHO_E;
seg = fs_m_in.REQ_FD_SEG;
position = fs_m_in.REQ_FD_POS;
nrbytes = (unsigned) fs_m_in.REQ_FD_NBYTES;
/*partial_cnt = fs_m_in.REQ_FD_PARTIAL;*/
user_addr = fs_m_in.REQ_FD_USER_ADDR;
/*if (partial_cnt > 0) partial_pipe = 1;*/
2005-04-21 16:53:53 +02:00
rdwt_err = OK; /* set to EIO if disk error occurs */
if (rw_flag == WRITING && block_spec == 0) {
/* Clear the zone containing present EOF if hole about
* to be created. This is necessary because all unwritten
* blocks prior to the EOF must read as zeros.
*/
if (position > f_size) clear_zone(rip, f_size, 0);
}
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes != 0) {
off = (unsigned int) (position % block_size);/* offset in blk*/
chunk = MIN(nrbytes, block_size - off);
if (chunk < 0) chunk = block_size - off;
if (rw_flag == READING) {
bytes_left = f_size - position;
if (position >= f_size) break; /* we are beyond EOF */
if (chunk > bytes_left) chunk = (int) bytes_left;
}
/* Read or write 'chunk' bytes. */
r = rw_chunk(rip, cvul64(position), off, chunk, (unsigned) nrbytes,
rw_flag, user_addr, seg, usr, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
user_addr += chunk; /* user buffer address */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position += chunk; /* position within the file */
2005-04-21 16:53:53 +02:00
}
fs_m_out.RES_FD_POS = position; /* It might change later and the VFS has
to know this value */
2005-04-21 16:53:53 +02:00
/* On write, update file size and access time. */
if (rw_flag == WRITING) {
if (regular || mode_word == I_DIRECTORY) {
if (position > f_size) rip->i_size = position;
}
}
else {
if (rip->i_pipe == I_PIPE) {
2005-09-11 18:45:46 +02:00
if ( position >= rip->i_size) {
/* Reset pipe pointers. */
rip->i_size = 0; /* no data left */
position = 0; /* reset reader(s) */
}
2005-04-21 16:53:53 +02:00
}
}
/* Check to see if read-ahead is called for, and if so, set it up. */
if (rw_flag == READING && rip->i_seek == NO_SEEK && position % block_size == 0
2005-04-21 16:53:53 +02:00
&& (regular || mode_word == I_DIRECTORY)) {
rdahed_inode = rip;
rdahedpos = position;
}
rip->i_seek = NO_SEEK;
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
/* if user-space copying failed, read/write failed. */
if (r == OK && r2 != OK) {
r = r2;
}
2005-04-21 16:53:53 +02:00
if (r == OK) {
if (rw_flag == READING) rip->i_update |= ATIME;
if (rw_flag == WRITING) rip->i_update |= CTIME | MTIME;
rip->i_dirt = DIRTY; /* inode is thus now dirty */
}
fs_m_out.RES_FD_CUM_IO = cum_io;
fs_m_out.RES_FD_SIZE = rip->i_size;
return(r);
}
/*===========================================================================*
2007-08-07 14:38:35 +02:00
* fs_readwrite_s *
*===========================================================================*/
2007-08-07 14:38:35 +02:00
PUBLIC int fs_readwrite_s(void)
{
int r, rw_flag, chunk, block_size, block_spec;
int partial_cnt, regular, partial_pipe, nrbytes;
cp_grant_id_t gid;
off_t position, f_size, bytes_left;
unsigned int off, cum_io;
mode_t mode_word;
int completed, r2 = OK;
struct inode *rip;
partial_pipe = 0;
r = OK;
/* Try to get inode according to its index */
if (fs_m_in.REQ_FD_INODE_INDEX >= 0 &&
fs_m_in.REQ_FD_INODE_INDEX < NR_INODES &&
inode[fs_m_in.REQ_FD_INODE_INDEX].i_num == fs_m_in.REQ_FD_INODE_NR) {
rip = &inode[fs_m_in.REQ_FD_INODE_INDEX];
}
else {
/* Find the inode referred */
rip = find_inode(fs_dev, fs_m_in.REQ_FD_INODE_NR);
if (!rip) {
printf("FS: unavaliable inode by fs_readwrite(), nr: %d\n",
fs_m_in.REQ_FD_INODE_NR);
return EINVAL;
}
}
mode_word = rip->i_mode & I_TYPE;
regular = (mode_word == I_REGULAR || mode_word == I_NAMED_PIPE);
block_spec = (mode_word == I_BLOCK_SPECIAL ? 1 : 0);
/* Determine blocksize */
block_size = (block_spec ? get_block_size(rip->i_zone[0])
: rip->i_sp->s_block_size);
f_size = (block_spec ? ULONG_MAX : rip->i_size);
/* Get the values from the request message */
rw_flag = (fs_m_in.m_type == REQ_READ_S ? READING : WRITING);
gid = fs_m_in.REQ_FD_GID;
position = fs_m_in.REQ_FD_POS;
nrbytes = (unsigned) fs_m_in.REQ_FD_NBYTES;
/*partial_cnt = fs_m_in.REQ_FD_PARTIAL;*/
/*if (partial_cnt > 0) partial_pipe = 1;*/
rdwt_err = OK; /* set to EIO if disk error occurs */
if (rw_flag == WRITING && block_spec == 0) {
/* Check in advance to see if file will grow too big. */
if (position > rip->i_sp->s_max_size - nrbytes)
return(EFBIG);
/* Clear the zone containing present EOF if hole about
* to be created. This is necessary because all unwritten
* blocks prior to the EOF must read as zeros.
*/
if (position > f_size) clear_zone(rip, f_size, 0);
}
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes != 0) {
off = (unsigned int) (position % block_size);/* offset in blk*/
chunk = MIN(nrbytes, block_size - off);
if (chunk < 0) chunk = block_size - off;
if (rw_flag == READING) {
bytes_left = f_size - position;
if (position >= f_size) break; /* we are beyond EOF */
if (chunk > bytes_left) chunk = (int) bytes_left;
}
/* Read or write 'chunk' bytes. */
r = rw_chunk_s(rip, cvul64(position), off, chunk, (unsigned) nrbytes,
rw_flag, gid, cum_io, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position += chunk; /* position within the file */
}
fs_m_out.RES_FD_POS = position; /* It might change later and the VFS has
to know this value */
/* On write, update file size and access time. */
if (rw_flag == WRITING) {
if (regular || mode_word == I_DIRECTORY) {
if (position > f_size) rip->i_size = position;
}
}
else {
if (rip->i_pipe == I_PIPE) {
if ( position >= rip->i_size) {
/* Reset pipe pointers. */
rip->i_size = 0; /* no data left */
position = 0; /* reset reader(s) */
}
}
}
/* Check to see if read-ahead is called for, and if so, set it up. */
if (rw_flag == READING && rip->i_seek == NO_SEEK && position % block_size == 0
&& (regular || mode_word == I_DIRECTORY)) {
rdahed_inode = rip;
rdahedpos = position;
}
rip->i_seek = NO_SEEK;
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
/* if user-space copying failed, read/write failed. */
if (r == OK && r2 != OK) {
r = r2;
}
if (r == OK) {
if (rw_flag == READING) rip->i_update |= ATIME;
if (rw_flag == WRITING) rip->i_update |= CTIME | MTIME;
rip->i_dirt = DIRTY; /* inode is thus now dirty */
}
fs_m_out.RES_FD_CUM_IO = cum_io;
fs_m_out.RES_FD_SIZE = rip->i_size;
return(r);
}
/*===========================================================================*
* fs_breadwrite_o *
*===========================================================================*/
PUBLIC int fs_breadwrite_o(void)
{
int r, usr, rw_flag, chunk, block_size;
int nrbytes;
u64_t position;
unsigned int off, cum_io;
mode_t mode_word;
int completed, r2 = OK;
char *user_addr;
/* Pseudo inode for rw_chunk */
struct inode rip;
r = OK;
/* Get the values from the request message */
2007-08-07 14:38:35 +02:00
rw_flag = (fs_m_in.m_type == REQ_BREAD_O ? READING : WRITING);
usr = fs_m_in.REQ_XFD_WHO_E;
position = make64(fs_m_in.REQ_XFD_POS_LO, fs_m_in.REQ_XFD_POS_HI);
nrbytes = (unsigned) fs_m_in.REQ_XFD_NBYTES;
user_addr = fs_m_in.REQ_XFD_USER_ADDR;
block_size = get_block_size(fs_m_in.REQ_XFD_BDEV);
rip.i_zone[0] = fs_m_in.REQ_XFD_BDEV;
rip.i_mode = I_BLOCK_SPECIAL;
rip.i_size = 0;
rdwt_err = OK; /* set to EIO if disk error occurs */
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes != 0) {
off = rem64u(position, block_size); /* offset in blk*/
chunk = MIN(nrbytes, block_size - off);
if (chunk < 0) chunk = block_size - off;
/* Read or write 'chunk' bytes. */
r = rw_chunk(&rip, position, off, chunk, (unsigned) nrbytes,
rw_flag, user_addr, D, usr, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
user_addr += chunk; /* user buffer address */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position= add64ul(position, chunk); /* position within the file */
}
fs_m_out.RES_XFD_POS_LO = ex64lo(position);
fs_m_out.RES_XFD_POS_HI = ex64hi(position);
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
fs_m_out.RES_XFD_CUM_IO = cum_io;
return(r);
2005-04-21 16:53:53 +02:00
}
2007-08-07 14:38:35 +02:00
/*===========================================================================*
* fs_breadwrite_s *
*===========================================================================*/
PUBLIC int fs_breadwrite_s(void)
{
int r, rw_flag, chunk, block_size;
cp_grant_id_t gid;
int nrbytes;
u64_t position;
unsigned int off, cum_io;
mode_t mode_word;
int completed, r2 = OK;
/* Pseudo inode for rw_chunk */
struct inode rip;
r = OK;
/* Get the values from the request message */
rw_flag = (fs_m_in.m_type == REQ_BREAD_S ? READING : WRITING);
gid = fs_m_in.REQ_XFD_GID;
position = make64(fs_m_in.REQ_XFD_POS_LO, fs_m_in.REQ_XFD_POS_HI);
nrbytes = (unsigned) fs_m_in.REQ_XFD_NBYTES;
block_size = get_block_size(fs_m_in.REQ_XFD_BDEV);
rip.i_zone[0] = fs_m_in.REQ_XFD_BDEV;
rip.i_mode = I_BLOCK_SPECIAL;
rip.i_size = 0;
rdwt_err = OK; /* set to EIO if disk error occurs */
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes != 0) {
off = rem64u(position, block_size); /* offset in blk*/
chunk = MIN(nrbytes, block_size - off);
if (chunk < 0) chunk = block_size - off;
/* Read or write 'chunk' bytes. */
r = rw_chunk_s(&rip, position, off, chunk, (unsigned) nrbytes,
rw_flag, gid, cum_io, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position= add64ul(position, chunk); /* position within the file */
}
fs_m_out.RES_XFD_POS_LO = ex64lo(position);
fs_m_out.RES_XFD_POS_HI = ex64hi(position);
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
fs_m_out.RES_XFD_CUM_IO = cum_io;
return(r);
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* rw_chunk *
*===========================================================================*/
2005-08-29 18:47:18 +02:00
PRIVATE int rw_chunk(rip, position, off, chunk, left, rw_flag, buff,
seg, usr, block_size, completed)
2005-04-21 16:53:53 +02:00
register struct inode *rip; /* pointer to inode for file to be rd/wr */
u64_t position; /* position within file to read or write */
2005-04-21 16:53:53 +02:00
unsigned off; /* off within the current block */
int chunk; /* number of bytes to read or write */
unsigned left; /* max number of bytes wanted after position */
int rw_flag; /* READING or WRITING */
char *buff; /* virtual address of the user buffer */
int seg; /* T or D segment in user space */
int usr; /* which user process */
int block_size; /* block size of FS operating on */
int *completed; /* number of bytes copied */
2005-04-21 16:53:53 +02:00
{
/* Read or write (part of) a block. */
register struct buf *bp;
register int r = OK;
2005-04-21 16:53:53 +02:00
int n, block_spec;
block_t b;
dev_t dev;
*completed = 0;
2005-04-21 16:53:53 +02:00
block_spec = (rip->i_mode & I_TYPE) == I_BLOCK_SPECIAL;
2005-04-21 16:53:53 +02:00
if (block_spec) {
b = div64u(position, block_size);
2005-04-21 16:53:53 +02:00
dev = (dev_t) rip->i_zone[0];
}
else {
if (ex64hi(position) != 0)
panic(__FILE__, "rw_chunk: position too high", NO_NUM);
b = read_map(rip, ex64lo(position));
2005-04-21 16:53:53 +02:00
dev = rip->i_dev;
}
if (!block_spec && b == NO_BLOCK) {
if (rw_flag == READING) {
/* Reading from a nonexistent block. Must read as all zeros.*/
bp = get_block(NO_DEV, NO_BLOCK, NORMAL); /* get a buffer */
zero_block(bp);
}
else {
2005-04-21 16:53:53 +02:00
/* Writing to a nonexistent block. Create and enter in inode.*/
if ((bp= new_block(rip, ex64lo(position))) == NIL_BUF)
return(err_code);
2005-04-21 16:53:53 +02:00
}
}
else if (rw_flag == READING) {
2005-04-21 16:53:53 +02:00
/* Read and read ahead if convenient. */
bp = rahead(rip, b, position, left);
}
else {
2005-04-21 16:53:53 +02:00
/* Normally an existing block to be partially overwritten is first read
* in. However, a full block need not be read in. If it is already in
* the cache, acquire it, otherwise just acquire a free buffer.
*/
n = (chunk == block_size ? NO_READ : NORMAL);
if (!block_spec && off == 0 && ex64lo(position) >= rip->i_size)
n = NO_READ;
2005-04-21 16:53:53 +02:00
bp = get_block(dev, b, n);
}
/* In all cases, bp now points to a valid buffer. */
2005-09-11 18:45:46 +02:00
if (bp == NIL_BUF) {
panic(__FILE__,"bp not valid in rw_chunk, this can't happen", NO_NUM);
}
2005-04-21 16:53:53 +02:00
if (rw_flag == WRITING && chunk != block_size && !block_spec &&
ex64lo(position) >= rip->i_size && off == 0) {
2005-04-21 16:53:53 +02:00
zero_block(bp);
}
2005-04-21 16:53:53 +02:00
if (rw_flag == READING) {
/* Copy a chunk from the block buffer to user space. */
r = sys_vircopy(SELF_E, D, (phys_bytes) (bp->b_data+off),
2005-04-21 16:53:53 +02:00
usr, seg, (phys_bytes) buff,
(phys_bytes) chunk);
}
else {
2005-04-21 16:53:53 +02:00
/* Copy a chunk from user space to the block buffer. */
r = sys_vircopy(usr, seg, (phys_bytes) buff,
SELF_E, D, (phys_bytes) (bp->b_data+off),
2005-04-21 16:53:53 +02:00
(phys_bytes) chunk);
bp->b_dirt = DIRTY;
}
n = (off + chunk == block_size ? FULL_DATA_BLOCK : PARTIAL_DATA_BLOCK);
put_block(bp, n);
2005-04-21 16:53:53 +02:00
return(r);
}
2007-08-07 14:38:35 +02:00
/*===========================================================================*
* rw_chunk_s *
*===========================================================================*/
PRIVATE int rw_chunk_s(rip, position, off, chunk, left, rw_flag, gid,
buf_off, block_size, completed)
register struct inode *rip; /* pointer to inode for file to be rd/wr */
u64_t position; /* position within file to read or write */
unsigned off; /* off within the current block */
int chunk; /* number of bytes to read or write */
unsigned left; /* max number of bytes wanted after position */
int rw_flag; /* READING or WRITING */
cp_grant_id_t gid; /* grant */
unsigned buf_off; /* offset in grant */
int block_size; /* block size of FS operating on */
int *completed; /* number of bytes copied */
{
/* Read or write (part of) a block. */
register struct buf *bp;
register int r = OK;
int n, block_spec;
block_t b;
dev_t dev;
*completed = 0;
block_spec = (rip->i_mode & I_TYPE) == I_BLOCK_SPECIAL;
if (block_spec) {
b = div64u(position, block_size);
dev = (dev_t) rip->i_zone[0];
}
else {
if (ex64hi(position) != 0)
panic(__FILE__, "rw_chunk: position too high", NO_NUM);
b = read_map(rip, ex64lo(position));
dev = rip->i_dev;
}
if (!block_spec && b == NO_BLOCK) {
if (rw_flag == READING) {
/* Reading from a nonexistent block. Must read as all zeros.*/
bp = get_block(NO_DEV, NO_BLOCK, NORMAL); /* get a buffer */
zero_block(bp);
}
else {
/* Writing to a nonexistent block. Create and enter in inode.*/
if ((bp= new_block(rip, ex64lo(position))) == NIL_BUF)
return(err_code);
}
}
else if (rw_flag == READING) {
/* Read and read ahead if convenient. */
bp = rahead(rip, b, position, left);
}
else {
/* Normally an existing block to be partially overwritten is first read
* in. However, a full block need not be read in. If it is already in
* the cache, acquire it, otherwise just acquire a free buffer.
*/
n = (chunk == block_size ? NO_READ : NORMAL);
if (!block_spec && off == 0 && ex64lo(position) >= rip->i_size)
n = NO_READ;
bp = get_block(dev, b, n);
}
/* In all cases, bp now points to a valid buffer. */
if (bp == NIL_BUF) {
panic(__FILE__,"bp not valid in rw_chunk, this can't happen", NO_NUM);
}
if (rw_flag == WRITING && chunk != block_size && !block_spec &&
ex64lo(position) >= rip->i_size && off == 0) {
zero_block(bp);
}
if (rw_flag == READING) {
/* Copy a chunk from the block buffer to user space. */
#if 0
printf("sys_safecopyto: proc %d, gid %d, off %d, size %d\n",
FS_PROC_NR, gid, buf_off, chunk);
#endif
r = sys_safecopyto(FS_PROC_NR, gid, buf_off,
(vir_bytes) (bp->b_data+off), (phys_bytes) chunk, D);
}
else {
/* Copy a chunk from user space to the block buffer. */
r = sys_safecopyfrom(FS_PROC_NR, gid, buf_off,
(vir_bytes) (bp->b_data+off), (phys_bytes) chunk, D);
bp->b_dirt = DIRTY;
}
n = (off + chunk == block_size ? FULL_DATA_BLOCK : PARTIAL_DATA_BLOCK);
put_block(bp, n);
return(r);
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* read_map *
*===========================================================================*/
PUBLIC block_t read_map(rip, position)
register struct inode *rip; /* ptr to inode to map from */
off_t position; /* position in file whose blk wanted */
{
/* Given an inode and a position within the corresponding file, locate the
* block (not zone) number in which that position is to be found and return it.
*/
register struct buf *bp;
register zone_t z;
int scale, boff, dzones, nr_indirects, index, zind, ex;
block_t b;
long excess, zone, block_pos;
scale = rip->i_sp->s_log_zone_size; /* for block-zone conversion */
block_pos = position/rip->i_sp->s_block_size; /* relative blk # in file */
zone = block_pos >> scale; /* position's zone */
boff = (int) (block_pos - (zone << scale) ); /* relative blk # within zone */
dzones = rip->i_ndzones;
nr_indirects = rip->i_nindirs;
/* Is 'position' to be found in the inode itself? */
if (zone < dzones) {
zind = (int) zone; /* index should be an int */
z = rip->i_zone[zind];
if (z == NO_ZONE) return(NO_BLOCK);
b = ((block_t) z << scale) + boff;
return(b);
}
/* It is not in the inode, so it must be single or double indirect. */
excess = zone - dzones; /* first Vx_NR_DZONES don't count */
if (excess < nr_indirects) {
/* 'position' can be located via the single indirect block. */
z = rip->i_zone[dzones];
} else {
/* 'position' can be located via the double indirect block. */
if ( (z = rip->i_zone[dzones+1]) == NO_ZONE) return(NO_BLOCK);
excess -= nr_indirects; /* single indir doesn't count*/
b = (block_t) z << scale;
ASSERT(rip->i_dev != NO_DEV);
2005-04-21 16:53:53 +02:00
bp = get_block(rip->i_dev, b, NORMAL); /* get double indirect block */
index = (int) (excess/nr_indirects);
ASSERT(bp->b_dev != NO_DEV);
ASSERT(bp->b_dev == rip->i_dev);
2005-04-21 16:53:53 +02:00
z = rd_indir(bp, index); /* z= zone for single*/
put_block(bp, INDIRECT_BLOCK); /* release double ind block */
excess = excess % nr_indirects; /* index into single ind blk */
}
/* 'z' is zone num for single indirect block; 'excess' is index into it. */
if (z == NO_ZONE) return(NO_BLOCK);
b = (block_t) z << scale; /* b is blk # for single ind */
bp = get_block(rip->i_dev, b, NORMAL); /* get single indirect block */
ex = (int) excess; /* need an integer */
z = rd_indir(bp, ex); /* get block pointed to */
put_block(bp, INDIRECT_BLOCK); /* release single indir blk */
if (z == NO_ZONE) return(NO_BLOCK);
b = ((block_t) z << scale) + boff;
return(b);
}
/*===========================================================================*
* rd_indir *
*===========================================================================*/
PUBLIC zone_t rd_indir(bp, index)
struct buf *bp; /* pointer to indirect block */
int index; /* index into *bp */
{
/* Given a pointer to an indirect block, read one entry. The reason for
* making a separate routine out of this is that there are four cases:
* V1 (IBM and 68000), and V2 (IBM and 68000).
*/
struct super_block *sp;
zone_t zone; /* V2 zones are longs (shorts in V1) */
if(bp == NIL_BUF)
panic(__FILE__, "rd_indir() on NIL_BUF", NO_NUM);
2005-04-21 16:53:53 +02:00
sp = get_super(bp->b_dev); /* need super block to find file sys type */
/* read a zone from an indirect block */
if (sp->s_version == V1)
zone = (zone_t) conv2(sp->s_native, (int) bp->b_v1_ind[index]);
else
zone = (zone_t) conv4(sp->s_native, (long) bp->b_v2_ind[index]);
if (zone != NO_ZONE &&
(zone < (zone_t) sp->s_firstdatazone || zone >= sp->s_zones)) {
printf("Illegal zone number %ld in indirect block, index %d\n",
(long) zone, index);
panic(__FILE__,"check file system", NO_NUM);
2005-04-21 16:53:53 +02:00
}
return(zone);
}
/*===========================================================================*
* read_ahead *
*===========================================================================*/
PUBLIC void read_ahead()
{
/* Read a block into the cache before it is needed. */
int block_size;
register struct inode *rip;
struct buf *bp;
block_t b;
rip = rdahed_inode; /* pointer to inode to read ahead from */
block_size = get_block_size(rip->i_dev);
rdahed_inode = NIL_INODE; /* turn off read ahead */
if ( (b = read_map(rip, rdahedpos)) == NO_BLOCK) return; /* at EOF */
bp = rahead(rip, b, cvul64(rdahedpos), block_size);
2005-04-21 16:53:53 +02:00
put_block(bp, PARTIAL_DATA_BLOCK);
}
/*===========================================================================*
* rahead *
*===========================================================================*/
PUBLIC struct buf *rahead(rip, baseblock, position, bytes_ahead)
register struct inode *rip; /* pointer to inode for file to be read */
block_t baseblock; /* block at current position */
u64_t position; /* position within file */
2005-04-21 16:53:53 +02:00
unsigned bytes_ahead; /* bytes beyond position for immediate use */
{
/* Fetch a block from the cache or the device. If a physical read is
* required, prefetch as many more blocks as convenient into the cache.
* This usually covers bytes_ahead and is at least BLOCKS_MINIMUM.
* The device driver may decide it knows better and stop reading at a
* cylinder boundary (or after an error). Rw_scattered() puts an optional
* flag on all reads to allow this.
*/
int block_size;
/* Minimum number of blocks to prefetch. */
# define BLOCKS_MINIMUM (NR_BUFS < 50 ? 18 : 32)
int block_spec, scale, read_q_size;
unsigned int blocks_ahead, fragment;
block_t block, blocks_left;
off_t ind1_pos;
dev_t dev;
struct buf *bp;
static struct buf **read_q;
STATICINIT(read_q, NR_BUFS);
2005-04-21 16:53:53 +02:00
block_spec = (rip->i_mode & I_TYPE) == I_BLOCK_SPECIAL;
if (block_spec) {
dev = (dev_t) rip->i_zone[0];
} else {
dev = rip->i_dev;
}
block_size = get_block_size(dev);
block = baseblock;
bp = get_block(dev, block, PREFETCH);
if (bp->b_dev != NO_DEV) return(bp);
/* The best guess for the number of blocks to prefetch: A lot.
* It is impossible to tell what the device looks like, so we don't even
* try to guess the geometry, but leave it to the driver.
*
* The floppy driver can read a full track with no rotational delay, and it
* avoids reading partial tracks if it can, so handing it enough buffers to
* read two tracks is perfect. (Two, because some diskette types have
* an odd number of sectors per track, so a block may span tracks.)
*
* The disk drivers don't try to be smart. With todays disks it is
* impossible to tell what the real geometry looks like, so it is best to
* read as much as you can. With luck the caching on the drive allows
* for a little time to start the next read.
*
* The current solution below is a bit of a hack, it just reads blocks from
* the current file position hoping that more of the file can be found. A
* better solution must look at the already available zone pointers and
* indirect blocks (but don't call read_map!).
*/
fragment = rem64u(position, block_size);
position= sub64u(position, fragment);
2005-04-21 16:53:53 +02:00
bytes_ahead += fragment;
blocks_ahead = (bytes_ahead + block_size - 1) / block_size;
if (block_spec && rip->i_size == 0) {
blocks_left = NR_IOREQS;
} else {
blocks_left = (rip->i_size - ex64lo(position) + block_size - 1) /
block_size;
2005-04-21 16:53:53 +02:00
/* Go for the first indirect block if we are in its neighborhood. */
if (!block_spec) {
scale = rip->i_sp->s_log_zone_size;
ind1_pos = (off_t) rip->i_ndzones * (block_size << scale);
if (ex64lo(position) <= ind1_pos && rip->i_size > ind1_pos) {
2005-04-21 16:53:53 +02:00
blocks_ahead++;
blocks_left++;
}
}
}
/* No more than the maximum request. */
if (blocks_ahead > NR_IOREQS) blocks_ahead = NR_IOREQS;
/* Read at least the minimum number of blocks, but not after a seek. */
if (blocks_ahead < BLOCKS_MINIMUM && rip->i_seek == NO_SEEK)
blocks_ahead = BLOCKS_MINIMUM;
/* Can't go past end of file. */
if (blocks_ahead > blocks_left) blocks_ahead = blocks_left;
read_q_size = 0;
/* Acquire block buffers. */
for (;;) {
read_q[read_q_size++] = bp;
if (--blocks_ahead == 0) break;
/* Don't trash the cache, leave 4 free. */
if (bufs_in_use >= NR_BUFS - 4) break;
block++;
bp = get_block(dev, block, PREFETCH);
if (bp->b_dev != NO_DEV) {
/* Oops, block already in the cache, get out. */
put_block(bp, FULL_DATA_BLOCK);
break;
}
}
rw_scattered(dev, read_q, read_q_size, READING);
return(get_block(dev, baseblock, NORMAL));
}
#define GETDENTS_BUFSIZ 257
PRIVATE char getdents_buf[GETDENTS_BUFSIZ];
/*===========================================================================*
* fs_getdents *
*===========================================================================*/
PUBLIC int fs_getdents(void)
{
register struct inode *rip;
int o, r, block_size, len, reclen, done;
ino_t ino;
block_t b;
cp_grant_id_t gid;
size_t size, tmpbuf_off, userbuf_off;
off_t pos, off, block_pos, new_pos, ent_pos;
struct buf *bp;
struct direct *dp;
struct dirent *dep;
char *cp;
ino= fs_m_in.REQ_GDE_INODE;
gid= fs_m_in.REQ_GDE_GRANT;
size= fs_m_in.REQ_GDE_SIZE;
pos= fs_m_in.REQ_GDE_POS;
/* Check whether the position is properly aligned */
if (pos % DIR_ENTRY_SIZE)
return ENOENT;
if ( (rip = get_inode(fs_dev, ino)) == NIL_INODE) {
printf("MFS(%d) get_inode by fs_getdents() failed\n", SELF_E);
return(EINVAL);
}
block_size= rip->i_sp->s_block_size;
off= (pos % block_size); /* Offset in block */
block_pos= pos-off;
done= FALSE; /* Stop processing directory blocks
* when done is set.
*/
tmpbuf_off= 0; /* Offset in getdents_buf */
memset(getdents_buf, '\0', GETDENTS_BUFSIZ); /* Avoid leaking any data */
userbuf_off= 0; /* Offset in the user's buffer */
/* The default position for the next request is EOF. If the user's buffer
* fills up before EOF, new_pos will be modified.
*/
new_pos= rip->i_size;
for (; block_pos < rip->i_size; block_pos += block_size) {
b = read_map(rip, block_pos); /* get block number */
/* Since directories don't have holes, 'b' cannot be NO_BLOCK. */
bp = get_block(rip->i_dev, b, NORMAL); /* get a dir block */
if (bp == NO_BLOCK)
panic(__FILE__,"get_block returned NO_BLOCK", NO_NUM);
/* Search a directory block. */
if (block_pos < pos)
dp = &bp->b_dir[off / DIR_ENTRY_SIZE];
else
dp = &bp->b_dir[0];
for (; dp < &bp->b_dir[NR_DIR_ENTRIES(block_size)]; dp++) {
if (dp->d_ino == 0)
continue; /* Entry is not in use */
/* Compute the length of the name */
cp= memchr(dp->d_name, '\0', NAME_MAX);
if (cp == NULL)
len= NAME_MAX;
else
len= cp-dp->d_name;
/* Compute record length */
reclen= offsetof(struct dirent, d_name) + len + 1;
o= (reclen % sizeof(long));
if (o != 0)
reclen += sizeof(long)-o;
/* Need the postition of this entry in the directory */
ent_pos= block_pos + ((char *)dp - bp->b_data);
if (tmpbuf_off + reclen > GETDENTS_BUFSIZ)
{
r= sys_safecopyto(FS_PROC_NR, gid, userbuf_off,
(vir_bytes)getdents_buf, tmpbuf_off, D);
if (r != OK)
{
panic(__FILE__,
"fs_getdents: sys_safecopyto failed\n",
r);
}
userbuf_off += tmpbuf_off;
tmpbuf_off= 0;
}
if (userbuf_off + tmpbuf_off + reclen > size)
{
/* The user has no space for one more record */
done= TRUE;
/* Record the postion of this entry, it is the
* starting point of the next request (unless the
* postion is modified with lseek).
*/
new_pos= ent_pos;
break;
}
dep= (struct dirent *)&getdents_buf[tmpbuf_off];
dep->d_ino= dp->d_ino;
dep->d_off= ent_pos;
dep->d_reclen= reclen;
memcpy(dep->d_name, dp->d_name, len);
dep->d_name[len]= '\0';
tmpbuf_off += reclen;
}
put_block(bp, DIRECTORY_BLOCK);
if (done)
break;
}
if (tmpbuf_off != 0)
{
r= sys_safecopyto(FS_PROC_NR, gid, userbuf_off,
(vir_bytes)getdents_buf, tmpbuf_off, D);
if (r != OK)
panic(__FILE__, "fs_getdents: sys_safecopyto failed\n", r);
userbuf_off += tmpbuf_off;
}
if (done && userbuf_off == 0)
r= EINVAL; /* The user's buffer is too small */
else
{
fs_m_out.RES_GDE_CUM_IO= userbuf_off;
if (new_pos >= pos)
fs_m_out.RES_GDE_POS_CHANGE= new_pos-pos;
else
fs_m_out.RES_GDE_POS_CHANGE= 0;
Various VFS and MFS fixes to improve correctness, consistency and POSIX compliance. VFS changes: * truncate() on a file system mounted read-only no longer panics MFS. * ftruncate() and fcntl(F_FREESP) now check for write permission on the file descriptor instead of the file, write(). * utime(), chown() and fchown() now check for file system read-only status. MFS changes: * link() and rename() no longer return the internal EENTERMOUNT and ELEAVEMOUNT errors to the application as part of a check on the source path. * rename() now treats EENTERMOUNT from the destination path check as an error, preventing file system corruption from renaming a normal directory to an existing mountpoint directory. * mountpoints (mounted-on dirs) are hidden better during lookups: - if a lookup starts from a mountpoint, the first component has to be ".." (anything else being a VFS-FS protocol violation). - in that case, the permissions of the mountpoint are not checked. - in all other cases, visiting a mountpoint always results in EENTERMOUNT. * a lookup on ".." from a mount root or chroot(2) root no longer succeeds if the caller does not have search permission on that directory. * POSIX: getdents() now updates directory access times. * POSIX: readlink() now returns partial results instead of ERANGE. Miscellaneous changes: * semaphore file handling bug (leading to hangs) fixed in test 32. The VFS changes should now put the burden of checking for read-only status of file systems entirely on VFS, and limit the access permission checks that file systems have to perform, to checking search permission on directories during lookups. From this point on, any deviation from that spceification should be considered a bug. Note that for legacy reasons, the root partition is assumed to be mounted read-write.
2009-05-18 13:27:12 +02:00
rip->i_update |= ATIME;
rip->i_dirt = DIRTY;
r= OK;
}
put_inode(rip); /* release the inode */
return(r);
}
/*===========================================================================*
* fs_getdents_o *
*===========================================================================*/
PUBLIC int fs_getdents_o(void)
{
/* Legacy support: wrapper around new getdents, returning the resulting number
* of bytes in the m_type field of the reply message instead.
*/
int r;
r = fs_getdents();
if (r == OK)
r = fs_m_out.RES_GDE_CUM_IO;
return(r);
}