minix/servers/fs/read.c

561 lines
18 KiB
C
Raw Normal View History

2005-04-21 16:53:53 +02:00
/* This file contains the heart of the mechanism used to read (and write)
* files. Read and write requests are split up into chunks that do not cross
* block boundaries. Each chunk is then processed in turn. Reads on special
* files are also detected and handled.
*
* The entry points into this file are
* do_read: perform the READ system call by calling read_write
* read_write: actually do the work of READ and WRITE
* read_map: given an inode and file position, look up its zone number
* rd_indir: read an entry in an indirect block
* read_ahead: manage the block read ahead business
*/
#include "fs.h"
#include <fcntl.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 "buf.h"
#include "file.h"
#include "fproc.h"
#include "inode.h"
#include "param.h"
#include "super.h"
FORWARD _PROTOTYPE( int rw_chunk, (struct inode *rip, off_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));
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* do_read *
*===========================================================================*/
PUBLIC int do_read()
{
return(read_write(READING));
}
/*===========================================================================*
* read_write *
*===========================================================================*/
PUBLIC int read_write(rw_flag)
int rw_flag; /* READING or WRITING */
{
/* Perform read(fd, buffer, nbytes) or write(fd, buffer, nbytes) call. */
register struct inode *rip;
register struct filp *f;
off_t bytes_left, f_size, position;
unsigned int off, cum_io;
int op, oflags, r, chunk, usr, seg, block_spec, char_spec;
int regular, partial_pipe = 0, partial_cnt = 0;
mode_t mode_word;
struct filp *wf;
int block_size = 0;
int completed, r2 = OK;
phys_bytes p;
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
/* PM loads segments by putting funny things in other bits of the
* message, indicated by a high bit in fd.
*/
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
if (who_e == PM_PROC_NR && (m_in.fd & _PM_SEG_FLAG)) {
seg = (int) m_in.m1_p2;
usr = (int) m_in.m1_p3;
m_in.fd &= ~(_PM_SEG_FLAG); /* get rid of flag bit */
2005-04-21 16:53:53 +02:00
} else {
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
usr = who_e; /* normal case */
2005-04-21 16:53:53 +02:00
seg = D;
}
/* If the file descriptor is valid, get the inode, size and mode. */
if (m_in.nbytes < 0) return(EINVAL);
if ((f = get_filp(m_in.fd)) == NIL_FILP) return(err_code);
if (((f->filp_mode) & (rw_flag == READING ? R_BIT : W_BIT)) == 0) {
return(f->filp_mode == FILP_CLOSED ? EIO : EBADF);
}
2005-08-29 18:47:18 +02:00
if (m_in.nbytes == 0)
return(0); /* so char special files need not check for 0*/
/* check if user process has the memory it needs.
* if not, copying will fail later.
* do this after 0-check above because umap doesn't want to map 0 bytes.
*/
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
if ((r = sys_umap(usr, seg, (vir_bytes) m_in.buffer, m_in.nbytes, &p)) != OK) {
printf("FS: read_write: umap failed for process %d\n", usr);
return r;
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
}
2005-04-21 16:53:53 +02:00
position = f->filp_pos;
oflags = f->filp_flags;
rip = f->filp_ino;
f_size = rip->i_size;
r = OK;
if (rip->i_pipe == I_PIPE) {
/* fp->fp_cum_io_partial is only nonzero when doing partial writes */
cum_io = fp->fp_cum_io_partial;
} else {
cum_io = 0;
}
op = (rw_flag == READING ? DEV_READ : DEV_WRITE);
mode_word = rip->i_mode & I_TYPE;
regular = mode_word == I_REGULAR || mode_word == I_NAMED_PIPE;
2005-09-11 18:45:46 +02:00
if ((char_spec = (mode_word == I_CHAR_SPECIAL ? 1 : 0))) {
if (rip->i_zone[0] == NO_DEV)
2005-08-29 18:47:18 +02:00
panic(__FILE__,"read_write tries to read from "
"character device NO_DEV", NO_NUM);
2005-04-21 16:53:53 +02:00
block_size = get_block_size(rip->i_zone[0]);
}
2005-09-11 18:45:46 +02:00
if ((block_spec = (mode_word == I_BLOCK_SPECIAL ? 1 : 0))) {
f_size = ULONG_MAX;
2005-09-11 18:45:46 +02:00
if (rip->i_zone[0] == NO_DEV)
2005-08-29 18:47:18 +02:00
panic(__FILE__,"read_write tries to read from "
" block device NO_DEV", NO_NUM);
2005-04-21 16:53:53 +02:00
block_size = get_block_size(rip->i_zone[0]);
}
2005-09-11 18:45:46 +02:00
if (!char_spec && !block_spec)
2005-04-21 16:53:53 +02:00
block_size = rip->i_sp->s_block_size;
rdwt_err = OK; /* set to EIO if disk error occurs */
/* Check for character special files. */
if (char_spec) {
dev_t dev;
dev = (dev_t) rip->i_zone[0];
r = dev_io(op, dev, usr, m_in.buffer, position, m_in.nbytes, oflags);
if (r >= 0) {
cum_io = r;
position += r;
r = OK;
}
} else {
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 - m_in.nbytes)
return(EFBIG);
/* Check for O_APPEND flag. */
if (oflags & O_APPEND) position = f_size;
/* 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);
}
/* Pipes are a little different. Check. */
if (rip->i_pipe == I_PIPE) {
r = pipe_check(rip, rw_flag, oflags,
m_in.nbytes, position, &partial_cnt, 0);
2005-04-21 16:53:53 +02:00
if (r <= 0) return(r);
}
if (partial_cnt > 0) partial_pipe = 1;
/* Split the transfer into chunks that don't span two blocks. */
while (m_in.nbytes != 0) {
2005-04-21 16:53:53 +02:00
off = (unsigned int) (position % block_size);/* offset in blk*/
if (partial_pipe) { /* pipes only */
chunk = MIN(partial_cnt, block_size - off);
} else
chunk = MIN(m_in.nbytes, 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, position, off, chunk, (unsigned) m_in.nbytes,
rw_flag, m_in.buffer, seg, usr, block_size, &completed);
2005-04-21 16:53:53 +02:00
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
m_in.buffer += chunk; /* user buffer address */
m_in.nbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position += chunk; /* position within the file */
if (partial_pipe) {
partial_cnt -= chunk;
if (partial_cnt <= 0) break;
}
}
}
/* 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) */
wf = find_filp(rip, W_BIT);
if (wf != NIL_FILP) wf->filp_pos = 0;
}
2005-04-21 16:53:53 +02:00
}
}
f->filp_pos = position;
/* 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;
}
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 */
if (partial_pipe) {
partial_pipe = 0;
/* partial write on pipe with */
/* O_NONBLOCK, return write count */
2005-04-21 16:53:53 +02:00
if (!(oflags & O_NONBLOCK)) {
fp->fp_cum_io_partial = cum_io;
suspend(XPIPE); /* partial write on pipe with */
return(SUSPEND); /* nbyte > PIPE_SIZE - non-atomic */
}
}
fp->fp_cum_io_partial = 0;
return(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 */
off_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 */
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;
if (block_spec) {
b = position/block_size;
dev = (dev_t) rip->i_zone[0];
} else {
b = read_map(rip, 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, 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 && position >= rip->i_size) n = NO_READ;
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 &&
position >= rip->i_size && off == 0) {
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(FS_PROC_NR, D, (phys_bytes) (bp->b_data+off),
usr, seg, (phys_bytes) buff,
(phys_bytes) chunk);
} else {
/* Copy a chunk from user space to the block buffer. */
r = sys_vircopy(usr, seg, (phys_bytes) buff,
FS_PROC_NR, D, (phys_bytes) (bp->b_data+off),
(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);
}
/*===========================================================================*
* 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;
bp = get_block(rip->i_dev, b, NORMAL); /* get double indirect block */
index = (int) (excess/nr_indirects);
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, rdahedpos, block_size);
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 */
off_t position; /* position within file */
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[NR_BUFS];
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 = position % block_size;
position -= fragment;
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 - position + block_size - 1) / block_size;
/* 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 (position <= ind1_pos && rip->i_size > ind1_pos) {
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));
}