summaryrefslogtreecommitdiffstats
path: root/fs/jffs/intrep.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 15:20:36 -0700
committerLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 15:20:36 -0700
commit1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch)
tree0bba044c4ce775e45a88a51686b5d9f90697ea9d /fs/jffs/intrep.c
downloadtalos-obmc-linux-1da177e4c3f41524e886b7f1b8a0c1fc7321cac2.tar.gz
talos-obmc-linux-1da177e4c3f41524e886b7f1b8a0c1fc7321cac2.zip
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
Diffstat (limited to 'fs/jffs/intrep.c')
-rw-r--r--fs/jffs/intrep.c3457
1 files changed, 3457 insertions, 0 deletions
diff --git a/fs/jffs/intrep.c b/fs/jffs/intrep.c
new file mode 100644
index 000000000000..8cc6893fc56c
--- /dev/null
+++ b/fs/jffs/intrep.c
@@ -0,0 +1,3457 @@
+/*
+ * JFFS -- Journaling Flash File System, Linux implementation.
+ *
+ * Copyright (C) 1999, 2000 Axis Communications, Inc.
+ *
+ * Created by Finn Hakansson <finn@axis.com>.
+ *
+ * This is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * $Id: intrep.c,v 1.102 2001/09/23 23:28:36 dwmw2 Exp $
+ *
+ * Ported to Linux 2.3.x and MTD:
+ * Copyright (C) 2000 Alexander Larsson (alex@cendio.se), Cendio Systems AB
+ *
+ */
+
+/* This file contains the code for the internal structure of the
+ Journaling Flash File System, JFFS. */
+
+/*
+ * Todo list:
+ *
+ * memcpy_to_flash() and memcpy_from_flash() functions.
+ *
+ * Implementation of hard links.
+ *
+ * Organize the source code in a better way. Against the VFS we could
+ * have jffs_ext.c, and against the block device jffs_int.c.
+ * A better file-internal organization too.
+ *
+ * A better checksum algorithm.
+ *
+ * Consider endianness stuff. ntohl() etc.
+ *
+ * Are we handling the atime, mtime, ctime members of the inode right?
+ *
+ * Remove some duplicated code. Take a look at jffs_write_node() and
+ * jffs_rewrite_data() for instance.
+ *
+ * Implement more meaning of the nlink member in various data structures.
+ * nlink could be used in conjunction with hard links for instance.
+ *
+ * Better memory management. Allocate data structures in larger chunks
+ * if possible.
+ *
+ * If too much meta data is stored, a garbage collect should be issued.
+ * We have experienced problems with too much meta data with for instance
+ * log files.
+ *
+ * Improve the calls to jffs_ioctl(). We would like to retrieve more
+ * information to be able to debug (or to supervise) JFFS during run-time.
+ *
+ */
+
+#include <linux/config.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/jffs.h>
+#include <linux/fs.h>
+#include <linux/stat.h>
+#include <linux/pagemap.h>
+#include <asm/semaphore.h>
+#include <asm/byteorder.h>
+#include <linux/smp_lock.h>
+#include <linux/time.h>
+#include <linux/ctype.h>
+
+#include "intrep.h"
+#include "jffs_fm.h"
+
+long no_jffs_node = 0;
+static long no_jffs_file = 0;
+#if defined(JFFS_MEMORY_DEBUG) && JFFS_MEMORY_DEBUG
+long no_jffs_control = 0;
+long no_jffs_raw_inode = 0;
+long no_jffs_node_ref = 0;
+long no_jffs_fm = 0;
+long no_jffs_fmcontrol = 0;
+long no_hash = 0;
+long no_name = 0;
+#endif
+
+static int jffs_scan_flash(struct jffs_control *c);
+static int jffs_update_file(struct jffs_file *f, struct jffs_node *node);
+static int jffs_build_file(struct jffs_file *f);
+static int jffs_free_file(struct jffs_file *f);
+static int jffs_free_node_list(struct jffs_file *f);
+static int jffs_garbage_collect_now(struct jffs_control *c);
+static int jffs_insert_file_into_hash(struct jffs_file *f);
+static int jffs_remove_redundant_nodes(struct jffs_file *f);
+
+/* Is there enough space on the flash? */
+static inline int JFFS_ENOUGH_SPACE(struct jffs_control *c, __u32 space)
+{
+ struct jffs_fmcontrol *fmc = c->fmc;
+
+ while (1) {
+ if ((fmc->flash_size - (fmc->used_size + fmc->dirty_size))
+ >= fmc->min_free_size + space) {
+ return 1;
+ }
+ if (fmc->dirty_size < fmc->sector_size)
+ return 0;
+
+ if (jffs_garbage_collect_now(c)) {
+ D1(printk("JFFS_ENOUGH_SPACE: jffs_garbage_collect_now() failed.\n"));
+ return 0;
+ }
+ }
+}
+
+#if CONFIG_JFFS_FS_VERBOSE > 0
+static __u8
+flash_read_u8(struct mtd_info *mtd, loff_t from)
+{
+ size_t retlen;
+ __u8 ret;
+ int res;
+
+ res = MTD_READ(mtd, from, 1, &retlen, &ret);
+ if (retlen != 1) {
+ printk("Didn't read a byte in flash_read_u8(). Returned %d\n", res);
+ return 0;
+ }
+
+ return ret;
+}
+
+static void
+jffs_hexdump(struct mtd_info *mtd, loff_t pos, int size)
+{
+ char line[16];
+ int j = 0;
+
+ while (size > 0) {
+ int i;
+
+ printk("%ld:", (long) pos);
+ for (j = 0; j < 16; j++) {
+ line[j] = flash_read_u8(mtd, pos++);
+ }
+ for (i = 0; i < j; i++) {
+ if (!(i & 1)) {
+ printk(" %.2x", line[i] & 0xff);
+ }
+ else {
+ printk("%.2x", line[i] & 0xff);
+ }
+ }
+
+ /* Print empty space */
+ for (; i < 16; i++) {
+ if (!(i & 1)) {
+ printk(" ");
+ }
+ else {
+ printk(" ");
+ }
+ }
+ printk(" ");
+
+ for (i = 0; i < j; i++) {
+ if (isgraph(line[i])) {
+ printk("%c", line[i]);
+ }
+ else {
+ printk(".");
+ }
+ }
+ printk("\n");
+ size -= 16;
+ }
+}
+
+#endif
+
+#define flash_safe_acquire(arg)
+#define flash_safe_release(arg)
+
+
+static int
+flash_safe_read(struct mtd_info *mtd, loff_t from,
+ u_char *buf, size_t count)
+{
+ size_t retlen;
+ int res;
+
+ D3(printk(KERN_NOTICE "flash_safe_read(%p, %08x, %p, %08x)\n",
+ mtd, (unsigned int) from, buf, count));
+
+ res = MTD_READ(mtd, from, count, &retlen, buf);
+ if (retlen != count) {
+ panic("Didn't read all bytes in flash_safe_read(). Returned %d\n", res);
+ }
+ return res?res:retlen;
+}
+
+
+static __u32
+flash_read_u32(struct mtd_info *mtd, loff_t from)
+{
+ size_t retlen;
+ __u32 ret;
+ int res;
+
+ res = MTD_READ(mtd, from, 4, &retlen, (unsigned char *)&ret);
+ if (retlen != 4) {
+ printk("Didn't read all bytes in flash_read_u32(). Returned %d\n", res);
+ return 0;
+ }
+
+ return ret;
+}
+
+
+static int
+flash_safe_write(struct mtd_info *mtd, loff_t to,
+ const u_char *buf, size_t count)
+{
+ size_t retlen;
+ int res;
+
+ D3(printk(KERN_NOTICE "flash_safe_write(%p, %08x, %p, %08x)\n",
+ mtd, (unsigned int) to, buf, count));
+
+ res = MTD_WRITE(mtd, to, count, &retlen, buf);
+ if (retlen != count) {
+ printk("Didn't write all bytes in flash_safe_write(). Returned %d\n", res);
+ }
+ return res?res:retlen;
+}
+
+
+static int
+flash_safe_writev(struct mtd_info *mtd, const struct kvec *vecs,
+ unsigned long iovec_cnt, loff_t to)
+{
+ size_t retlen, retlen_a;
+ int i;
+ int res;
+
+ D3(printk(KERN_NOTICE "flash_safe_writev(%p, %08x, %p)\n",
+ mtd, (unsigned int) to, vecs));
+
+ if (mtd->writev) {
+ res = MTD_WRITEV(mtd, vecs, iovec_cnt, to, &retlen);
+ return res ? res : retlen;
+ }
+ /* Not implemented writev. Repeatedly use write - on the not so
+ unreasonable assumption that the mtd driver doesn't care how
+ many write cycles we use. */
+ res=0;
+ retlen=0;
+
+ for (i=0; !res && i<iovec_cnt; i++) {
+ res = MTD_WRITE(mtd, to, vecs[i].iov_len, &retlen_a, vecs[i].iov_base);
+ if (retlen_a != vecs[i].iov_len) {
+ printk("Didn't write all bytes in flash_safe_writev(). Returned %d\n", res);
+ if (i != iovec_cnt-1)
+ return -EIO;
+ }
+ /* If res is non-zero, retlen_a is undefined, but we don't
+ care because in that case it's not going to be
+ returned anyway.
+ */
+ to += retlen_a;
+ retlen += retlen_a;
+ }
+ return res?res:retlen;
+}
+
+
+static int
+flash_memset(struct mtd_info *mtd, loff_t to,
+ const u_char c, size_t size)
+{
+ static unsigned char pattern[64];
+ int i;
+
+ /* fill up pattern */
+
+ for(i = 0; i < 64; i++)
+ pattern[i] = c;
+
+ /* write as many 64-byte chunks as we can */
+
+ while (size >= 64) {
+ flash_safe_write(mtd, to, pattern, 64);
+ size -= 64;
+ to += 64;
+ }
+
+ /* and the rest */
+
+ if(size)
+ flash_safe_write(mtd, to, pattern, size);
+
+ return size;
+}
+
+
+static void
+intrep_erase_callback(struct erase_info *done)
+{
+ wait_queue_head_t *wait_q;
+
+ wait_q = (wait_queue_head_t *)done->priv;
+
+ wake_up(wait_q);
+}
+
+
+static int
+flash_erase_region(struct mtd_info *mtd, loff_t start,
+ size_t size)
+{
+ struct erase_info *erase;
+ DECLARE_WAITQUEUE(wait, current);
+ wait_queue_head_t wait_q;
+
+ erase = kmalloc(sizeof(struct erase_info), GFP_KERNEL);
+ if (!erase)
+ return -ENOMEM;
+
+ init_waitqueue_head(&wait_q);
+
+ erase->mtd = mtd;
+ erase->callback = intrep_erase_callback;
+ erase->addr = start;
+ erase->len = size;
+ erase->priv = (u_long)&wait_q;
+
+ /* FIXME: Use TASK_INTERRUPTIBLE and deal with being interrupted */
+ set_current_state(TASK_UNINTERRUPTIBLE);
+ add_wait_queue(&wait_q, &wait);
+
+ if (MTD_ERASE(mtd, erase) < 0) {
+ set_current_state(TASK_RUNNING);
+ remove_wait_queue(&wait_q, &wait);
+ kfree(erase);
+
+ printk(KERN_WARNING "flash: erase of region [0x%lx, 0x%lx] "
+ "totally failed\n", (long)start, (long)start + size);
+
+ return -1;
+ }
+
+ schedule(); /* Wait for flash to finish. */
+ remove_wait_queue(&wait_q, &wait);
+
+ kfree(erase);
+
+ return 0;
+}
+
+/* This routine calculates checksums in JFFS. */
+static __u32
+jffs_checksum(const void *data, int size)
+{
+ __u32 sum = 0;
+ __u8 *ptr = (__u8 *)data;
+ while (size-- > 0) {
+ sum += *ptr++;
+ }
+ D3(printk(", result: 0x%08x\n", sum));
+ return sum;
+}
+
+
+static int
+jffs_checksum_flash(struct mtd_info *mtd, loff_t start, int size, __u32 *result)
+{
+ __u32 sum = 0;
+ loff_t ptr = start;
+ __u8 *read_buf;
+ int i, length;
+
+ /* Allocate read buffer */
+ read_buf = (__u8 *) kmalloc (sizeof(__u8) * 4096, GFP_KERNEL);
+ if (!read_buf) {
+ printk(KERN_NOTICE "kmalloc failed in jffs_checksum_flash()\n");
+ return -ENOMEM;
+ }
+ /* Loop until checksum done */
+ while (size) {
+ /* Get amount of data to read */
+ if (size < 4096)
+ length = size;
+ else
+ length = 4096;
+
+ /* Perform flash read */
+ D3(printk(KERN_NOTICE "jffs_checksum_flash\n"));
+ flash_safe_read(mtd, ptr, &read_buf[0], length);
+
+ /* Compute checksum */
+ for (i=0; i < length ; i++)
+ sum += read_buf[i];
+
+ /* Update pointer and size */
+ size -= length;
+ ptr += length;
+ }
+
+ /* Free read buffer */
+ kfree (read_buf);
+
+ /* Return result */
+ D3(printk("checksum result: 0x%08x\n", sum));
+ *result = sum;
+ return 0;
+}
+
+static __inline__ void jffs_fm_write_lock(struct jffs_fmcontrol *fmc)
+{
+ // down(&fmc->wlock);
+}
+
+static __inline__ void jffs_fm_write_unlock(struct jffs_fmcontrol *fmc)
+{
+ // up(&fmc->wlock);
+}
+
+
+/* Create and initialize a new struct jffs_file. */
+static struct jffs_file *
+jffs_create_file(struct jffs_control *c,
+ const struct jffs_raw_inode *raw_inode)
+{
+ struct jffs_file *f;
+
+ if (!(f = (struct jffs_file *)kmalloc(sizeof(struct jffs_file),
+ GFP_KERNEL))) {
+ D(printk("jffs_create_file(): Failed!\n"));
+ return NULL;
+ }
+ no_jffs_file++;
+ memset(f, 0, sizeof(struct jffs_file));
+ f->ino = raw_inode->ino;
+ f->pino = raw_inode->pino;
+ f->nlink = raw_inode->nlink;
+ f->deleted = raw_inode->deleted;
+ f->c = c;
+
+ return f;
+}
+
+
+/* Build a control block for the file system. */
+static struct jffs_control *
+jffs_create_control(struct super_block *sb)
+{
+ struct jffs_control *c;
+ register int s = sizeof(struct jffs_control);
+ int i;
+ D(char *t = 0);
+
+ D2(printk("jffs_create_control()\n"));
+
+ if (!(c = (struct jffs_control *)kmalloc(s, GFP_KERNEL))) {
+ goto fail_control;
+ }
+ DJM(no_jffs_control++);
+ c->root = NULL;
+ c->gc_task = NULL;
+ c->hash_len = JFFS_HASH_SIZE;
+ s = sizeof(struct list_head) * c->hash_len;
+ if (!(c->hash = (struct list_head *)kmalloc(s, GFP_KERNEL))) {
+ goto fail_hash;
+ }
+ DJM(no_hash++);
+ for (i = 0; i < c->hash_len; i++)
+ INIT_LIST_HEAD(&c->hash[i]);
+ if (!(c->fmc = jffs_build_begin(c, MINOR(sb->s_dev)))) {
+ goto fail_fminit;
+ }
+ c->next_ino = JFFS_MIN_INO + 1;
+ c->delete_list = (struct jffs_delete_list *) 0;
+ return c;
+
+fail_fminit:
+ D(t = "c->fmc");
+fail_hash:
+ kfree(c);
+ DJM(no_jffs_control--);
+ D(t = t ? t : "c->hash");
+fail_control:
+ D(t = t ? t : "control");
+ D(printk("jffs_create_control(): Allocation failed: (%s)\n", t));
+ return (struct jffs_control *)0;
+}
+
+
+/* Clean up all data structures associated with the file system. */
+void
+jffs_cleanup_control(struct jffs_control *c)
+{
+ D2(printk("jffs_cleanup_control()\n"));
+
+ if (!c) {
+ D(printk("jffs_cleanup_control(): c == NULL !!!\n"));
+ return;
+ }
+
+ while (c->delete_list) {
+ struct jffs_delete_list *delete_list_element;
+ delete_list_element = c->delete_list;
+ c->delete_list = c->delete_list->next;
+ kfree(delete_list_element);
+ }
+
+ /* Free all files and nodes. */
+ if (c->hash) {
+ jffs_foreach_file(c, jffs_free_node_list);
+ jffs_foreach_file(c, jffs_free_file);
+ kfree(c->hash);
+ DJM(no_hash--);
+ }
+ jffs_cleanup_fmcontrol(c->fmc);
+ kfree(c);
+ DJM(no_jffs_control--);
+ D3(printk("jffs_cleanup_control(): Leaving...\n"));
+}
+
+
+/* This function adds a virtual root node to the in-RAM representation.
+ Called by jffs_build_fs(). */
+static int
+jffs_add_virtual_root(struct jffs_control *c)
+{
+ struct jffs_file *root;
+ struct jffs_node *node;
+
+ D2(printk("jffs_add_virtual_root(): "
+ "Creating a virtual root directory.\n"));
+
+ if (!(root = (struct jffs_file *)kmalloc(sizeof(struct jffs_file),
+ GFP_KERNEL))) {
+ return -ENOMEM;
+ }
+ no_jffs_file++;
+ if (!(node = jffs_alloc_node())) {
+ kfree(root);
+ no_jffs_file--;
+ return -ENOMEM;
+ }
+ DJM(no_jffs_node++);
+ memset(node, 0, sizeof(struct jffs_node));
+ node->ino = JFFS_MIN_INO;
+ memset(root, 0, sizeof(struct jffs_file));
+ root->ino = JFFS_MIN_INO;
+ root->mode = S_IFDIR | S_IRWXU | S_IRGRP
+ | S_IXGRP | S_IROTH | S_IXOTH;
+ root->atime = root->mtime = root->ctime = get_seconds();
+ root->nlink = 1;
+ root->c = c;
+ root->version_head = root->version_tail = node;
+ jffs_insert_file_into_hash(root);
+ return 0;
+}
+
+
+/* This is where the file system is built and initialized. */
+int
+jffs_build_fs(struct super_block *sb)
+{
+ struct jffs_control *c;
+ int err = 0;
+
+ D2(printk("jffs_build_fs()\n"));
+
+ if (!(c = jffs_create_control(sb))) {
+ return -ENOMEM;
+ }
+ c->building_fs = 1;
+ c->sb = sb;
+ if ((err = jffs_scan_flash(c)) < 0) {
+ if(err == -EAGAIN){
+ /* scan_flash() wants us to try once more. A flipping
+ bits sector was detect in the middle of the scan flash.
+ Clean up old allocated memory before going in.
+ */
+ D1(printk("jffs_build_fs: Cleaning up all control structures,"
+ " reallocating them and trying mount again.\n"));
+ jffs_cleanup_control(c);
+ if (!(c = jffs_create_control(sb))) {
+ return -ENOMEM;
+ }
+ c->building_fs = 1;
+ c->sb = sb;
+
+ if ((err = jffs_scan_flash(c)) < 0) {
+ goto jffs_build_fs_fail;
+ }
+ }else{
+ goto jffs_build_fs_fail;
+ }
+ }
+
+ /* Add a virtual root node if no one exists. */
+ if (!jffs_find_file(c, JFFS_MIN_INO)) {
+ if ((err = jffs_add_virtual_root(c)) < 0) {
+ goto jffs_build_fs_fail;
+ }
+ }
+
+ while (c->delete_list) {
+ struct jffs_file *f;
+ struct jffs_delete_list *delete_list_element;
+
+ if ((f = jffs_find_file(c, c->delete_list->ino))) {
+ f->deleted = 1;
+ }
+ delete_list_element = c->delete_list;
+ c->delete_list = c->delete_list->next;
+ kfree(delete_list_element);
+ }
+
+ /* Remove deleted nodes. */
+ if ((err = jffs_foreach_file(c, jffs_possibly_delete_file)) < 0) {
+ printk(KERN_ERR "JFFS: Failed to remove deleted nodes.\n");
+ goto jffs_build_fs_fail;
+ }
+ /* Remove redundant nodes. (We are not interested in the
+ return value in this case.) */
+ jffs_foreach_file(c, jffs_remove_redundant_nodes);
+ /* Try to build a tree from all the nodes. */
+ if ((err = jffs_foreach_file(c, jffs_insert_file_into_tree)) < 0) {
+ printk("JFFS: Failed to build tree.\n");
+ goto jffs_build_fs_fail;
+ }
+ /* Compute the sizes of all files in the filesystem. Adjust if
+ necessary. */
+ if ((err = jffs_foreach_file(c, jffs_build_file)) < 0) {
+ printk("JFFS: Failed to build file system.\n");
+ goto jffs_build_fs_fail;
+ }
+ sb->s_fs_info = (void *)c;
+ c->building_fs = 0;
+
+ D1(jffs_print_hash_table(c));
+ D1(jffs_print_tree(c->root, 0));
+
+ return 0;
+
+jffs_build_fs_fail:
+ jffs_cleanup_control(c);
+ return err;
+} /* jffs_build_fs() */
+
+
+/*
+ This checks for sectors that were being erased in their previous
+ lifetimes and for some reason or the other (power fail etc.),
+ the erase cycles never completed.
+ As the flash array would have reverted back to read status,
+ these sectors are detected by the symptom of the "flipping bits",
+ i.e. bits being read back differently from the same location in
+ flash if read multiple times.
+ The only solution to this is to re-erase the entire
+ sector.
+ Unfortunately detecting "flipping bits" is not a simple exercise
+ as a bit may be read back at 1 or 0 depending on the alignment
+ of the stars in the universe.
+ The level of confidence is in direct proportion to the number of
+ scans done. By power fail testing I (Vipin) have been able to
+ proove that reading twice is not enough.
+ Maybe 4 times? Change NUM_REREADS to a higher number if you want
+ a (even) higher degree of confidence in your mount process.
+ A higher number would of course slow down your mount.
+*/
+static int check_partly_erased_sectors(struct jffs_fmcontrol *fmc){
+
+#define NUM_REREADS 4 /* see note above */
+#define READ_AHEAD_BYTES 4096 /* must be a multiple of 4,
+ usually set to kernel page size */
+
+ __u8 *read_buf1;
+ __u8 *read_buf2;
+
+ int err = 0;
+ int retlen;
+ int i;
+ int cnt;
+ __u32 offset;
+ loff_t pos = 0;
+ loff_t end = fmc->flash_size;
+
+
+ /* Allocate read buffers */
+ read_buf1 = (__u8 *) kmalloc (sizeof(__u8) * READ_AHEAD_BYTES, GFP_KERNEL);
+ if (!read_buf1)
+ return -ENOMEM;
+
+ read_buf2 = (__u8 *) kmalloc (sizeof(__u8) * READ_AHEAD_BYTES, GFP_KERNEL);
+ if (!read_buf2) {
+ kfree(read_buf1);
+ return -ENOMEM;
+ }
+
+ CHECK_NEXT:
+ while(pos < end){
+
+ D1(printk("check_partly_erased_sector():checking sector which contains"
+ " offset 0x%x for flipping bits..\n", (__u32)pos));
+
+ retlen = flash_safe_read(fmc->mtd, pos,
+ &read_buf1[0], READ_AHEAD_BYTES);
+ retlen &= ~3;
+
+ for(cnt = 0; cnt < NUM_REREADS; cnt++){
+ (void)flash_safe_read(fmc->mtd, pos,
+ &read_buf2[0], READ_AHEAD_BYTES);
+
+ for (i=0 ; i < retlen ; i+=4) {
+ /* buffers MUST match, double word for word! */
+ if(*((__u32 *) &read_buf1[i]) !=
+ *((__u32 *) &read_buf2[i])
+ ){
+ /* flipping bits detected, time to erase sector */
+ /* This will help us log some statistics etc. */
+ D1(printk("Flipping bits detected in re-read round:%i of %i\n",
+ cnt, NUM_REREADS));
+ D1(printk("check_partly_erased_sectors:flipping bits detected"
+ " @offset:0x%x(0x%x!=0x%x)\n",
+ (__u32)pos+i, *((__u32 *) &read_buf1[i]),
+ *((__u32 *) &read_buf2[i])));
+
+ /* calculate start of present sector */
+ offset = (((__u32)pos+i)/(__u32)fmc->sector_size) * (__u32)fmc->sector_size;
+
+ D1(printk("check_partly_erased_sector():erasing sector starting 0x%x.\n",
+ offset));
+
+ if (flash_erase_region(fmc->mtd,
+ offset, fmc->sector_size) < 0) {
+ printk(KERN_ERR "JFFS: Erase of flash failed. "
+ "offset = %u, erase_size = %d\n",
+ offset , fmc->sector_size);
+
+ err = -EIO;
+ goto returnBack;
+
+ }else{
+ D1(printk("JFFS: Erase of flash sector @0x%x successful.\n",
+ offset));
+ /* skip ahead to the next sector */
+ pos = (((__u32)pos+i)/(__u32)fmc->sector_size) * (__u32)fmc->sector_size;
+ pos += fmc->sector_size;
+ goto CHECK_NEXT;
+ }
+ }
+ }
+ }
+ pos += READ_AHEAD_BYTES;
+ }
+
+ returnBack:
+ kfree(read_buf1);
+ kfree(read_buf2);
+
+ D2(printk("check_partly_erased_sector():Done checking all sectors till offset 0x%x for flipping bits.\n",
+ (__u32)pos));
+
+ return err;
+
+}/* end check_partly_erased_sectors() */
+
+
+
+/* Scan the whole flash memory in order to find all nodes in the
+ file systems. */
+static int
+jffs_scan_flash(struct jffs_control *c)
+{
+ char name[JFFS_MAX_NAME_LEN + 2];
+ struct jffs_raw_inode raw_inode;
+ struct jffs_node *node = NULL;
+ struct jffs_fmcontrol *fmc = c->fmc;
+ __u32 checksum;
+ __u8 tmp_accurate;
+ __u16 tmp_chksum;
+ __u32 deleted_file;
+ loff_t pos = 0;
+ loff_t start;
+ loff_t test_start;
+ loff_t end = fmc->flash_size;
+ __u8 *read_buf;
+ int i, len, retlen;
+ __u32 offset;
+
+ __u32 free_chunk_size1;
+ __u32 free_chunk_size2;
+
+
+#define NUMFREEALLOWED 2 /* 2 chunks of at least erase size space allowed */
+ int num_free_space = 0; /* Flag err if more than TWO
+ free blocks found. This is NOT allowed
+ by the current jffs design.
+ */
+ int num_free_spc_not_accp = 0; /* For debugging purposed keep count
+ of how much free space was rejected and
+ marked dirty
+ */
+
+ D1(printk("jffs_scan_flash(): start pos = 0x%lx, end = 0x%lx\n",
+ (long)pos, (long)end));
+
+ flash_safe_acquire(fmc->mtd);
+
+ /*
+ check and make sure that any sector does not suffer
+ from the "partly erased, bit flipping syndrome" (TM Vipin :)
+ If so, offending sectors will be erased.
+ */
+ if(check_partly_erased_sectors(fmc) < 0){
+
+ flash_safe_release(fmc->mtd);
+ return -EIO; /* bad, bad, bad error. Cannot continue.*/
+ }
+
+ /* Allocate read buffer */
+ read_buf = (__u8 *) kmalloc (sizeof(__u8) * 4096, GFP_KERNEL);
+ if (!read_buf) {
+ flash_safe_release(fmc->mtd);
+ return -ENOMEM;
+ }
+
+ /* Start the scan. */
+ while (pos < end) {
+ deleted_file = 0;
+
+ /* Remember the position from where we started this scan. */
+ start = pos;
+
+ switch (flash_read_u32(fmc->mtd, pos)) {
+ case JFFS_EMPTY_BITMASK:
+ /* We have found 0xffffffff at this position. We have to
+ scan the rest of the flash till the end or till
+ something else than 0xffffffff is found.
+ Keep going till we do not find JFFS_EMPTY_BITMASK
+ anymore */
+
+ D1(printk("jffs_scan_flash(): 0xffffffff at pos 0x%lx.\n",
+ (long)pos));
+
+ while(pos < end){
+
+ len = end - pos < 4096 ? end - pos : 4096;
+
+ retlen = flash_safe_read(fmc->mtd, pos,
+ &read_buf[0], len);
+
+ retlen &= ~3;
+
+ for (i=0 ; i < retlen ; i+=4, pos += 4) {
+ if(*((__u32 *) &read_buf[i]) !=
+ JFFS_EMPTY_BITMASK)
+ break;
+ }
+ if (i == retlen)
+ continue;
+ else
+ break;
+ }
+
+ D1(printk("jffs_scan_flash():0xffffffff ended at pos 0x%lx.\n",
+ (long)pos));
+
+ /* If some free space ends in the middle of a sector,
+ treat it as dirty rather than clean.
+ This is to handle the case where one thread
+ allocated space for a node, but didn't get to
+ actually _write_ it before power was lost, leaving
+ a gap in the log. Shifting all node writes into
+ a single kernel thread will fix the original problem.
+ */
+ if ((__u32) pos % fmc->sector_size) {
+ /* If there was free space in previous
+ sectors, don't mark that dirty too -
+ only from the beginning of this sector
+ (or from start)
+ */
+
+ test_start = pos & ~(fmc->sector_size-1); /* end of last sector */
+
+ if (start < test_start) {
+
+ /* free space started in the previous sector! */
+
+ if((num_free_space < NUMFREEALLOWED) &&
+ ((unsigned int)(test_start - start) >= fmc->sector_size)){
+
+ /*
+ Count it in if we are still under NUMFREEALLOWED *and* it is
+ at least 1 erase sector in length. This will keep us from
+ picking any little ole' space as "free".
+ */
+
+ D1(printk("Reducing end of free space to 0x%x from 0x%x\n",
+ (unsigned int)test_start, (unsigned int)pos));
+
+ D1(printk("Free space accepted: Starting 0x%x for 0x%x bytes\n",
+ (unsigned int) start,
+ (unsigned int)(test_start - start)));
+
+ /* below, space from "start" to "pos" will be marked dirty. */
+ start = test_start;
+
+ /* Being in here means that we have found at least an entire
+ erase sector size of free space ending on a sector boundary.
+ Keep track of free spaces accepted.
+ */
+ num_free_space++;
+ }else{
+ num_free_spc_not_accp++;
+ D1(printk("Free space (#%i) found but *Not* accepted: Starting"
+ " 0x%x for 0x%x bytes\n",
+ num_free_spc_not_accp, (unsigned int)start,
+ (unsigned int)((unsigned int)(pos & ~(fmc->sector_size-1)) - (unsigned int)start)));
+
+ }
+
+ }
+ if((((__u32)(pos - start)) != 0)){
+
+ D1(printk("Dirty space: Starting 0x%x for 0x%x bytes\n",
+ (unsigned int) start, (unsigned int) (pos - start)));
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ }else{
+ /* "Flipping bits" detected. This means that our scan for them
+ did not catch this offset. See check_partly_erased_sectors() for
+ more info.
+ */
+
+ D1(printk("jffs_scan_flash():wants to allocate dirty flash "
+ "space for 0 bytes.\n"));
+ D1(printk("jffs_scan_flash(): Flipping bits! We will free "
+ "all allocated memory, erase this sector and remount\n"));
+
+ /* calculate start of present sector */
+ offset = (((__u32)pos)/(__u32)fmc->sector_size) * (__u32)fmc->sector_size;
+
+ D1(printk("jffs_scan_flash():erasing sector starting 0x%x.\n",
+ offset));
+
+ if (flash_erase_region(fmc->mtd,
+ offset, fmc->sector_size) < 0) {
+ printk(KERN_ERR "JFFS: Erase of flash failed. "
+ "offset = %u, erase_size = %d\n",
+ offset , fmc->sector_size);
+
+ flash_safe_release(fmc->mtd);
+ kfree (read_buf);
+ return -1; /* bad, bad, bad! */
+
+ }
+ flash_safe_release(fmc->mtd);
+ kfree (read_buf);
+
+ return -EAGAIN; /* erased offending sector. Try mount one more time please. */
+ }
+ }else{
+ /* Being in here means that we have found free space that ends on an erase sector
+ boundary.
+ Count it in if we are still under NUMFREEALLOWED *and* it is at least 1 erase
+ sector in length. This will keep us from picking any little ole' space as "free".
+ */
+ if((num_free_space < NUMFREEALLOWED) &&
+ ((unsigned int)(pos - start) >= fmc->sector_size)){
+ /* We really don't do anything to mark space as free, except *not*
+ mark it dirty and just advance the "pos" location pointer.
+ It will automatically be picked up as free space.
+ */
+ num_free_space++;
+ D1(printk("Free space accepted: Starting 0x%x for 0x%x bytes\n",
+ (unsigned int) start, (unsigned int) (pos - start)));
+ }else{
+ num_free_spc_not_accp++;
+ D1(printk("Free space (#%i) found but *Not* accepted: Starting "
+ "0x%x for 0x%x bytes\n", num_free_spc_not_accp,
+ (unsigned int) start,
+ (unsigned int) (pos - start)));
+
+ /* Mark this space as dirty. We already have our free space. */
+ D1(printk("Dirty space: Starting 0x%x for 0x%x bytes\n",
+ (unsigned int) start, (unsigned int) (pos - start)));
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ }
+
+ }
+ if(num_free_space > NUMFREEALLOWED){
+ printk(KERN_WARNING "jffs_scan_flash(): Found free space "
+ "number %i. Only %i free space is allowed.\n",
+ num_free_space, NUMFREEALLOWED);
+ }
+ continue;
+
+ case JFFS_DIRTY_BITMASK:
+ /* We have found 0x00000000 at this position. Scan as far
+ as possible to find out how much is dirty. */
+ D1(printk("jffs_scan_flash(): 0x00000000 at pos 0x%lx.\n",
+ (long)pos));
+ for (; pos < end
+ && JFFS_DIRTY_BITMASK == flash_read_u32(fmc->mtd, pos);
+ pos += 4);
+ D1(printk("jffs_scan_flash(): 0x00 ended at "
+ "pos 0x%lx.\n", (long)pos));
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ continue;
+
+ case JFFS_MAGIC_BITMASK:
+ /* We have probably found a new raw inode. */
+ break;
+
+ default:
+ bad_inode:
+ /* We're f*cked. This is not solved yet. We have
+ to scan for the magic pattern. */
+ D1(printk("*************** Dirty flash memory or "
+ "bad inode: "
+ "hexdump(pos = 0x%lx, len = 128):\n",
+ (long)pos));
+ D1(jffs_hexdump(fmc->mtd, pos, 128));
+
+ for (pos += 4; pos < end; pos += 4) {
+ switch (flash_read_u32(fmc->mtd, pos)) {
+ case JFFS_MAGIC_BITMASK:
+ case JFFS_EMPTY_BITMASK:
+ /* handle these in the main switch() loop */
+ goto cont_scan;
+
+ default:
+ break;
+ }
+ }
+
+ cont_scan:
+ /* First, mark as dirty the region
+ which really does contain crap. */
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start),
+ NULL);
+
+ continue;
+ }/* switch */
+
+ /* We have found the beginning of an inode. Create a
+ node for it unless there already is one available. */
+ if (!node) {
+ if (!(node = jffs_alloc_node())) {
+ /* Free read buffer */
+ kfree (read_buf);
+
+ /* Release the flash device */
+ flash_safe_release(fmc->mtd);
+
+ return -ENOMEM;
+ }
+ DJM(no_jffs_node++);
+ }
+
+ /* Read the next raw inode. */
+
+ flash_safe_read(fmc->mtd, pos, (u_char *) &raw_inode,
+ sizeof(struct jffs_raw_inode));
+
+ /* When we compute the checksum for the inode, we never
+ count the 'accurate' or the 'checksum' fields. */
+ tmp_accurate = raw_inode.accurate;
+ tmp_chksum = raw_inode.chksum;
+ raw_inode.accurate = 0;
+ raw_inode.chksum = 0;
+ checksum = jffs_checksum(&raw_inode,
+ sizeof(struct jffs_raw_inode));
+ raw_inode.accurate = tmp_accurate;
+ raw_inode.chksum = tmp_chksum;
+
+ D3(printk("*** We have found this raw inode at pos 0x%lx "
+ "on the flash:\n", (long)pos));
+ D3(jffs_print_raw_inode(&raw_inode));
+
+ if (checksum != raw_inode.chksum) {
+ D1(printk("jffs_scan_flash(): Bad checksum: "
+ "checksum = %u, "
+ "raw_inode.chksum = %u\n",
+ checksum, raw_inode.chksum));
+ pos += sizeof(struct jffs_raw_inode);
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ /* Reuse this unused struct jffs_node. */
+ continue;
+ }
+
+ /* Check the raw inode read so far. Start with the
+ maximum length of the filename. */
+ if (raw_inode.nsize > JFFS_MAX_NAME_LEN) {
+ printk(KERN_WARNING "jffs_scan_flash: Found a "
+ "JFFS node with name too large\n");
+ goto bad_inode;
+ }
+
+ if (raw_inode.rename && raw_inode.dsize != sizeof(__u32)) {
+ printk(KERN_WARNING "jffs_scan_flash: Found a "
+ "rename node with dsize %u.\n",
+ raw_inode.dsize);
+ jffs_print_raw_inode(&raw_inode);
+ goto bad_inode;
+ }
+
+ /* The node's data segment should not exceed a
+ certain length. */
+ if (raw_inode.dsize > fmc->max_chunk_size) {
+ printk(KERN_WARNING "jffs_scan_flash: Found a "
+ "JFFS node with dsize (0x%x) > max_chunk_size (0x%x)\n",
+ raw_inode.dsize, fmc->max_chunk_size);
+ goto bad_inode;
+ }
+
+ pos += sizeof(struct jffs_raw_inode);
+
+ /* This shouldn't be necessary because a node that
+ violates the flash boundaries shouldn't be written
+ in the first place. */
+ if (pos >= end) {
+ goto check_node;
+ }
+
+ /* Read the name. */
+ *name = 0;
+ if (raw_inode.nsize) {
+ flash_safe_read(fmc->mtd, pos, name, raw_inode.nsize);
+ name[raw_inode.nsize] = '\0';
+ pos += raw_inode.nsize
+ + JFFS_GET_PAD_BYTES(raw_inode.nsize);
+ D3(printk("name == \"%s\"\n", name));
+ checksum = jffs_checksum(name, raw_inode.nsize);
+ if (checksum != raw_inode.nchksum) {
+ D1(printk("jffs_scan_flash(): Bad checksum: "
+ "checksum = %u, "
+ "raw_inode.nchksum = %u\n",
+ checksum, raw_inode.nchksum));
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ /* Reuse this unused struct jffs_node. */
+ continue;
+ }
+ if (pos >= end) {
+ goto check_node;
+ }
+ }
+
+ /* Read the data, if it exists, in order to be sure it
+ matches the checksum. */
+ if (raw_inode.dsize) {
+ if (raw_inode.rename) {
+ deleted_file = flash_read_u32(fmc->mtd, pos);
+ }
+ if (jffs_checksum_flash(fmc->mtd, pos, raw_inode.dsize, &checksum)) {
+ printk("jffs_checksum_flash() failed to calculate a checksum\n");
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ /* Reuse this unused struct jffs_node. */
+ continue;
+ }
+ pos += raw_inode.dsize
+ + JFFS_GET_PAD_BYTES(raw_inode.dsize);
+
+ if (checksum != raw_inode.dchksum) {
+ D1(printk("jffs_scan_flash(): Bad checksum: "
+ "checksum = %u, "
+ "raw_inode.dchksum = %u\n",
+ checksum, raw_inode.dchksum));
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ /* Reuse this unused struct jffs_node. */
+ continue;
+ }
+ }
+
+ check_node:
+
+ /* Remember the highest inode number in the whole file
+ system. This information will be used when assigning
+ new files new inode numbers. */
+ if (c->next_ino <= raw_inode.ino) {
+ c->next_ino = raw_inode.ino + 1;
+ }
+
+ if (raw_inode.accurate) {
+ int err;
+ node->data_offset = raw_inode.offset;
+ node->data_size = raw_inode.dsize;
+ node->removed_size = raw_inode.rsize;
+ /* Compute the offset to the actual data in the
+ on-flash node. */
+ node->fm_offset
+ = sizeof(struct jffs_raw_inode)
+ + raw_inode.nsize
+ + JFFS_GET_PAD_BYTES(raw_inode.nsize);
+ node->fm = jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start),
+ node);
+ if (!node->fm) {
+ D(printk("jffs_scan_flash(): !node->fm\n"));
+ jffs_free_node(node);
+ DJM(no_jffs_node--);
+
+ /* Free read buffer */
+ kfree (read_buf);
+
+ /* Release the flash device */
+ flash_safe_release(fmc->mtd);
+
+ return -ENOMEM;
+ }
+ if ((err = jffs_insert_node(c, NULL, &raw_inode,
+ name, node)) < 0) {
+ printk("JFFS: Failed to handle raw inode. "
+ "(err = %d)\n", err);
+ break;
+ }
+ if (raw_inode.rename) {
+ struct jffs_delete_list *dl
+ = (struct jffs_delete_list *)
+ kmalloc(sizeof(struct jffs_delete_list),
+ GFP_KERNEL);
+ if (!dl) {
+ D(printk("jffs_scan_flash: !dl\n"));
+ jffs_free_node(node);
+ DJM(no_jffs_node--);
+
+ /* Release the flash device */
+ flash_safe_release(fmc->flash_part);
+
+ /* Free read buffer */
+ kfree (read_buf);
+
+ return -ENOMEM;
+ }
+ dl->ino = deleted_file;
+ dl->next = c->delete_list;
+ c->delete_list = dl;
+ node->data_size = 0;
+ }
+ D3(jffs_print_node(node));
+ node = NULL; /* Don't free the node! */
+ }
+ else {
+ jffs_fmalloced(fmc, (__u32) start,
+ (__u32) (pos - start), NULL);
+ D3(printk("jffs_scan_flash(): Just found an obsolete "
+ "raw_inode. Continuing the scan...\n"));
+ /* Reuse this unused struct jffs_node. */
+ }
+ }
+
+ if (node) {
+ jffs_free_node(node);
+ DJM(no_jffs_node--);
+ }
+ jffs_build_end(fmc);
+
+ /* Free read buffer */
+ kfree (read_buf);
+
+ if(!num_free_space){
+ printk(KERN_WARNING "jffs_scan_flash(): Did not find even a single "
+ "chunk of free space. This is BAD!\n");
+ }
+
+ /* Return happy */
+ D3(printk("jffs_scan_flash(): Leaving...\n"));
+ flash_safe_release(fmc->mtd);
+
+ /* This is to trap the "free size accounting screwed error. */
+ free_chunk_size1 = jffs_free_size1(fmc);
+ free_chunk_size2 = jffs_free_size2(fmc);
+
+ if (free_chunk_size1 + free_chunk_size2 != fmc->free_size) {
+
+ printk(KERN_WARNING "jffs_scan_falsh():Free size accounting screwed\n");
+ printk(KERN_WARNING "jfffs_scan_flash():free_chunk_size1 == 0x%x, "
+ "free_chunk_size2 == 0x%x, fmc->free_size == 0x%x\n",
+ free_chunk_size1, free_chunk_size2, fmc->free_size);
+
+ return -1; /* Do NOT mount f/s so that we can inspect what happened.
+ Mounting this screwed up f/s will screw us up anyway.
+ */
+ }
+
+ return 0; /* as far as we are concerned, we are happy! */
+} /* jffs_scan_flash() */
+
+
+/* Insert any kind of node into the file system. Take care of data
+ insertions and deletions. Also remove redundant information. The
+ memory allocated for the `name' is regarded as "given away" in the
+ caller's perspective. */
+int
+jffs_insert_node(struct jffs_control *c, struct jffs_file *f,
+ const struct jffs_raw_inode *raw_inode,
+ const char *name, struct jffs_node *node)
+{
+ int update_name = 0;
+ int insert_into_tree = 0;
+
+ D2(printk("jffs_insert_node(): ino = %u, version = %u, "
+ "name = \"%s\", deleted = %d\n",
+ raw_inode->ino, raw_inode->version,
+ ((name && *name) ? name : ""), raw_inode->deleted));
+
+ /* If there doesn't exist an associated jffs_file, then
+ create, initialize and insert one into the file system. */
+ if (!f && !(f = jffs_find_file(c, raw_inode->ino))) {
+ if (!(f = jffs_create_file(c, raw_inode))) {
+ return -ENOMEM;
+ }
+ jffs_insert_file_into_hash(f);
+ insert_into_tree = 1;
+ }
+ node->ino = raw_inode->ino;
+ node->version = raw_inode->version;
+ node->data_size = raw_inode->dsize;
+ node->fm_offset = sizeof(struct jffs_raw_inode) + raw_inode->nsize
+ + JFFS_GET_PAD_BYTES(raw_inode->nsize);
+ node->name_size = raw_inode->nsize;
+
+ /* Now insert the node at the correct position into the file's
+ version list. */
+ if (!f->version_head) {
+ /* This is the first node. */
+ f->version_head = node;
+ f->version_tail = node;
+ node->version_prev = NULL;
+ node->version_next = NULL;
+ f->highest_version = node->version;
+ update_name = 1;
+ f->mode = raw_inode->mode;
+ f->uid = raw_inode->uid;
+ f->gid = raw_inode->gid;
+ f->atime = raw_inode->atime;
+ f->mtime = raw_inode->mtime;
+ f->ctime = raw_inode->ctime;
+ }
+ else if ((f->highest_version < node->version)
+ || (node->version == 0)) {
+ /* Insert at the end of the list. I.e. this node is the
+ newest one so far. */
+ node->version_prev = f->version_tail;
+ node->version_next = NULL;
+ f->version_tail->version_next = node;
+ f->version_tail = node;
+ f->highest_version = node->version;
+ update_name = 1;
+ f->pino = raw_inode->pino;
+ f->mode = raw_inode->mode;
+ f->uid = raw_inode->uid;
+ f->gid = raw_inode->gid;
+ f->atime = raw_inode->atime;
+ f->mtime = raw_inode->mtime;
+ f->ctime = raw_inode->ctime;
+ }
+ else if (f->version_head->version > node->version) {
+ /* Insert at the bottom of the list. */
+ node->version_prev = NULL;
+ node->version_next = f->version_head;
+ f->version_head->version_prev = node;
+ f->version_head = node;
+ if (!f->name) {
+ update_name = 1;
+ }
+ }
+ else {
+ struct jffs_node *n;
+ int newer_name = 0;
+ /* Search for the insertion position starting from
+ the tail (newest node). */
+ for (n = f->version_tail; n; n = n->version_prev) {
+ if (n->version < node->version) {
+ node->version_prev = n;
+ node->version_next = n->version_next;
+ node->version_next->version_prev = node;
+ n->version_next = node;
+ if (!newer_name) {
+ update_name = 1;
+ }
+ break;
+ }
+ if (n->name_size) {
+ newer_name = 1;
+ }
+ }
+ }
+
+ /* Deletion is irreversible. If any 'deleted' node is ever
+ written, the file is deleted */
+ if (raw_inode->deleted)
+ f->deleted = raw_inode->deleted;
+
+ /* Perhaps update the name. */
+ if (raw_inode->nsize && update_name && name && *name && (name != f->name)) {
+ if (f->name) {
+ kfree(f->name);
+ DJM(no_name--);
+ }
+ if (!(f->name = (char *) kmalloc(raw_inode->nsize + 1,
+ GFP_KERNEL))) {
+ return -ENOMEM;
+ }
+ DJM(no_name++);
+ memcpy(f->name, name, raw_inode->nsize);
+ f->name[raw_inode->nsize] = '\0';
+ f->nsize = raw_inode->nsize;
+ D3(printk("jffs_insert_node(): Updated the name of "
+ "the file to \"%s\".\n", name));
+ }
+
+ if (!c->building_fs) {
+ D3(printk("jffs_insert_node(): ---------------------------"
+ "------------------------------------------- 1\n"));
+ if (insert_into_tree) {
+ jffs_insert_file_into_tree(f);
+ }
+ /* Once upon a time, we would call jffs_possibly_delete_file()
+ here. That causes an oops if someone's still got the file
+ open, so now we only do it in jffs_delete_inode()
+ -- dwmw2
+ */
+ if (node->data_size || node->removed_size) {
+ jffs_update_file(f, node);
+ }
+ jffs_remove_redundant_nodes(f);
+
+ jffs_garbage_collect_trigger(c);
+
+ D3(printk("jffs_insert_node(): ---------------------------"
+ "------------------------------------------- 2\n"));
+ }
+
+ return 0;
+} /* jffs_insert_node() */
+
+
+/* Unlink a jffs_node from the version list it is in. */
+static inline void
+jffs_unlink_node_from_version_list(struct jffs_file *f,
+ struct jffs_node *node)
+{
+ if (node->version_prev) {
+ node->version_prev->version_next = node->version_next;
+ } else {
+ f->version_head = node->version_next;
+ }
+ if (node->version_next) {
+ node->version_next->version_prev = node->version_prev;
+ } else {
+ f->version_tail = node->version_prev;
+ }
+}
+
+
+/* Unlink a jffs_node from the range list it is in. */
+static inline void
+jffs_unlink_node_from_range_list(struct jffs_file *f, struct jffs_node *node)
+{
+ if (node->range_prev) {
+ node->range_prev->range_next = node->range_next;
+ }
+ else {
+ f->range_head = node->range_next;
+ }
+ if (node->range_next) {
+ node->range_next->range_prev = node->range_prev;
+ }
+ else {
+ f->range_tail = node->range_prev;
+ }
+}
+
+
+/* Function used by jffs_remove_redundant_nodes() below. This function
+ classifies what kind of information a node adds to a file. */
+static inline __u8
+jffs_classify_node(struct jffs_node *node)
+{
+ __u8 mod_type = JFFS_MODIFY_INODE;
+
+ if (node->name_size) {
+ mod_type |= JFFS_MODIFY_NAME;
+ }
+ if (node->data_size || node->removed_size) {
+ mod_type |= JFFS_MODIFY_DATA;
+ }
+ return mod_type;
+}
+
+
+/* Remove redundant nodes from a file. Mark the on-flash memory
+ as dirty. */
+static int
+jffs_remove_redundant_nodes(struct jffs_file *f)
+{
+ struct jffs_node *newest_node;
+ struct jffs_node *cur;
+ struct jffs_node *prev;
+ __u8 newest_type;
+ __u8 mod_type;
+ __u8 node_with_name_later = 0;
+
+ if (!(newest_node = f->version_tail)) {
+ return 0;
+ }
+
+ /* What does the `newest_node' modify? */
+ newest_type = jffs_classify_node(newest_node);
+ node_with_name_later = newest_type & JFFS_MODIFY_NAME;
+
+ D3(printk("jffs_remove_redundant_nodes(): ino: %u, name: \"%s\", "
+ "newest_type: %u\n", f->ino, (f->name ? f->name : ""),
+ newest_type));
+
+ /* Traverse the file's nodes and determine which of them that are
+ superfluous. Yeah, this might look very complex at first
+ glance but it is actually very simple. */
+ for (cur = newest_node->version_prev; cur; cur = prev) {
+ prev = cur->version_prev;
+ mod_type = jffs_classify_node(cur);
+ if ((mod_type <= JFFS_MODIFY_INODE)
+ || ((newest_type & JFFS_MODIFY_NAME)
+ && (mod_type
+ <= (JFFS_MODIFY_INODE + JFFS_MODIFY_NAME)))
+ || (cur->data_size == 0 && cur->removed_size
+ && !cur->version_prev && node_with_name_later)) {
+ /* Yes, this node is redundant. Remove it. */
+ D2(printk("jffs_remove_redundant_nodes(): "
+ "Removing node: ino: %u, version: %u, "
+ "mod_type: %u\n", cur->ino, cur->version,
+ mod_type));
+ jffs_unlink_node_from_version_list(f, cur);
+ jffs_fmfree(f->c->fmc, cur->fm, cur);
+ jffs_free_node(cur);
+ DJM(no_jffs_node--);
+ }
+ else {
+ node_with_name_later |= (mod_type & JFFS_MODIFY_NAME);
+ }
+ }
+
+ return 0;
+}
+
+
+/* Insert a file into the hash table. */
+static int
+jffs_insert_file_into_hash(struct jffs_file *f)
+{
+ int i = f->ino % f->c->hash_len;
+
+ D3(printk("jffs_insert_file_into_hash(): f->ino: %u\n", f->ino));
+
+ list_add(&f->hash, &f->c->hash[i]);
+ return 0;
+}
+
+
+/* Insert a file into the file system tree. */
+int
+jffs_insert_file_into_tree(struct jffs_file *f)
+{
+ struct jffs_file *parent;
+
+ D3(printk("jffs_insert_file_into_tree(): name: \"%s\"\n",
+ (f->name ? f->name : "")));
+
+ if (!(parent = jffs_find_file(f->c, f->pino))) {
+ if (f->pino == 0) {
+ f->c->root = f;
+ f->parent = NULL;
+ f->sibling_prev = NULL;
+ f->sibling_next = NULL;
+ return 0;
+ }
+ else {
+ D1(printk("jffs_insert_file_into_tree(): Found "
+ "inode with no parent and pino == %u\n",
+ f->pino));
+ return -1;
+ }
+ }
+ f->parent = parent;
+ f->sibling_next = parent->children;
+ if (f->sibling_next) {
+ f->sibling_next->sibling_prev = f;
+ }
+ f->sibling_prev = NULL;
+ parent->children = f;
+ return 0;
+}
+
+
+/* Remove a file from the hash table. */
+static int
+jffs_unlink_file_from_hash(struct jffs_file *f)
+{
+ D3(printk("jffs_unlink_file_from_hash(): f: 0x%p, "
+ "ino %u\n", f, f->ino));
+
+ list_del(&f->hash);
+ return 0;
+}
+
+
+/* Just remove the file from the parent's children. Don't free
+ any memory. */
+int
+jffs_unlink_file_from_tree(struct jffs_file *f)
+{
+ D3(printk("jffs_unlink_file_from_tree(): ino: %d, pino: %d, name: "
+ "\"%s\"\n", f->ino, f->pino, (f->name ? f->name : "")));
+
+ if (f->sibling_prev) {
+ f->sibling_prev->sibling_next = f->sibling_next;
+ }
+ else if (f->parent) {
+ D3(printk("f->parent=%p\n", f->parent));
+ f->parent->children = f->sibling_next;
+ }
+ if (f->sibling_next) {
+ f->sibling_next->sibling_prev = f->sibling_prev;
+ }
+ return 0;
+}
+
+
+/* Find a file with its inode number. */
+struct jffs_file *
+jffs_find_file(struct jffs_control *c, __u32 ino)
+{
+ struct jffs_file *f;
+ int i = ino % c->hash_len;
+ struct list_head *tmp;
+
+ D3(printk("jffs_find_file(): ino: %u\n", ino));
+
+ for (tmp = c->hash[i].next; tmp != &c->hash[i]; tmp = tmp->next) {
+ f = list_entry(tmp, struct jffs_file, hash);
+ if (ino != f->ino)
+ continue;
+ D3(printk("jffs_find_file(): Found file with ino "
+ "%u. (name: \"%s\")\n",
+ ino, (f->name ? f->name : ""));
+ );
+ return f;
+ }
+ D3(printk("jffs_find_file(): Didn't find file "
+ "with ino %u.\n", ino);
+ );
+ return NULL;
+}
+
+
+/* Find a file in a directory. We are comparing the names. */
+struct jffs_file *
+jffs_find_child(struct jffs_file *dir, const char *name, int len)
+{
+ struct jffs_file *f;
+
+ D3(printk("jffs_find_child()\n"));
+
+ for (f = dir->children; f; f = f->sibling_next) {
+ if (!f->deleted && f->name
+ && !strncmp(f->name, name, len)
+ && f->name[len] == '\0') {
+ break;
+ }
+ }
+
+ D3(if (f) {
+ printk("jffs_find_child(): Found \"%s\".\n", f->name);
+ }
+ else {
+ char *copy = (char *) kmalloc(len + 1, GFP_KERNEL);
+ if (copy) {
+ memcpy(copy, name, len);
+ copy[len] = '\0';
+ }
+ printk("jffs_find_child(): Didn't find the file \"%s\".\n",
+ (copy ? copy : ""));
+ if (copy) {
+ kfree(copy);
+ }
+ });
+
+ return f;
+}
+
+
+/* Write a raw inode that takes up a certain amount of space in the flash
+ memory. At the end of the flash device, there is often space that is
+ impossible to use. At these times we want to mark this space as not
+ used. In the cases when the amount of space is greater or equal than
+ a struct jffs_raw_inode, we write a "dummy node" that takes up this
+ space. The space after the raw inode, if it exists, is left as it is.
+ Since this space after the raw inode contains JFFS_EMPTY_BITMASK bytes,
+ we can compute the checksum of it; we don't have to manipulate it any
+ further.
+
+ If the space left on the device is less than the size of a struct
+ jffs_raw_inode, this space is filled with JFFS_DIRTY_BITMASK bytes.
+ No raw inode is written this time. */
+static int
+jffs_write_dummy_node(struct jffs_control *c, struct jffs_fm *dirty_fm)
+{
+ struct jffs_fmcontrol *fmc = c->fmc;
+ int err;
+
+ D1(printk("jffs_write_dummy_node(): dirty_fm->offset = 0x%08x, "
+ "dirty_fm->size = %u\n",
+ dirty_fm->offset, dirty_fm->size));
+
+ if (dirty_fm->size >= sizeof(struct jffs_raw_inode)) {
+ struct jffs_raw_inode raw_inode;
+ memset(&raw_inode, 0, sizeof(struct jffs_raw_inode));
+ raw_inode.magic = JFFS_MAGIC_BITMASK;
+ raw_inode.dsize = dirty_fm->size
+ - sizeof(struct jffs_raw_inode);
+ raw_inode.dchksum = raw_inode.dsize * 0xff;
+ raw_inode.chksum
+ = jffs_checksum(&raw_inode, sizeof(struct jffs_raw_inode));
+
+ if ((err = flash_safe_write(fmc->mtd,
+ dirty_fm->offset,
+ (u_char *)&raw_inode,
+ sizeof(struct jffs_raw_inode)))
+ < 0) {
+ printk(KERN_ERR "JFFS: jffs_write_dummy_node: "
+ "flash_safe_write failed!\n");
+ return err;
+ }
+ }
+ else {
+ flash_safe_acquire(fmc->mtd);
+ flash_memset(fmc->mtd, dirty_fm->offset, 0, dirty_fm->size);
+ flash_safe_release(fmc->mtd);
+ }
+
+ D3(printk("jffs_write_dummy_node(): Leaving...\n"));
+ return 0;
+}
+
+
+/* Write a raw inode, possibly its name and possibly some data. */
+int
+jffs_write_node(struct jffs_control *c, struct jffs_node *node,
+ struct jffs_raw_inode *raw_inode,
+ const char *name, const unsigned char *data,
+ int recoverable,
+ struct jffs_file *f)
+{
+ struct jffs_fmcontrol *fmc = c->fmc;
+ struct jffs_fm *fm;
+ struct kvec node_iovec[4];
+ unsigned long iovec_cnt;
+
+ __u32 pos;
+ int err;
+ __u32 slack = 0;
+
+ __u32 total_name_size = raw_inode->nsize
+ + JFFS_GET_PAD_BYTES(raw_inode->nsize);
+ __u32 total_data_size = raw_inode->dsize
+ + JFFS_GET_PAD_BYTES(raw_inode->dsize);
+ __u32 total_size = sizeof(struct jffs_raw_inode)
+ + total_name_size + total_data_size;
+
+ /* If this node isn't something that will eventually let
+ GC free even more space, then don't allow it unless
+ there's at least max_chunk_size space still available
+ */
+ if (!recoverable)
+ slack = fmc->max_chunk_size;
+
+
+ /* Fire the retrorockets and shoot the fruiton torpedoes, sir! */
+
+ ASSERT(if (!node) {
+ printk("jffs_write_node(): node == NULL\n");
+ return -EINVAL;
+ });
+ ASSERT(if (raw_inode && raw_inode->nsize && !name) {
+ printk("*** jffs_write_node(): nsize = %u but name == NULL\n",
+ raw_inode->nsize);
+ return -EINVAL;
+ });
+
+ D1(printk("jffs_write_node(): filename = \"%s\", ino = %u, "
+ "total_size = %u\n",
+ (name ? name : ""), raw_inode->ino,
+ total_size));
+
+ jffs_fm_write_lock(fmc);
+
+retry:
+ fm = NULL;
+ err = 0;
+ while (!fm) {
+
+ /* Deadlocks suck. */
+ while(fmc->free_size < fmc->min_free_size + total_size + slack) {
+ jffs_fm_write_unlock(fmc);
+ if (!JFFS_ENOUGH_SPACE(c, total_size + slack))
+ return -ENOSPC;
+ jffs_fm_write_lock(fmc);
+ }
+
+ /* First try to allocate some flash memory. */
+ err = jffs_fmalloc(fmc, total_size, node, &fm);
+
+ if (err == -ENOSPC) {
+ /* Just out of space. GC and try again */
+ if (fmc->dirty_size < fmc->sector_size) {
+ D(printk("jffs_write_node(): jffs_fmalloc(0x%p, %u) "
+ "failed, no dirty space to GC\n", fmc,
+ total_size));
+ return err;
+ }
+
+ D1(printk(KERN_INFO "jffs_write_node(): Calling jffs_garbage_collect_now()\n"));
+ jffs_fm_write_unlock(fmc);
+ if ((err = jffs_garbage_collect_now(c))) {
+ D(printk("jffs_write_node(): jffs_garbage_collect_now() failed\n"));
+ return err;
+ }
+ jffs_fm_write_lock(fmc);
+ continue;
+ }
+
+ if (err < 0) {
+ jffs_fm_write_unlock(fmc);
+
+ D(printk("jffs_write_node(): jffs_fmalloc(0x%p, %u) "
+ "failed!\n", fmc, total_size));
+ return err;
+ }
+
+ if (!fm->nodes) {
+ /* The jffs_fm struct that we got is not good enough.
+ Make that space dirty and try again */
+ if ((err = jffs_write_dummy_node(c, fm)) < 0) {
+ kfree(fm);
+ DJM(no_jffs_fm--);
+ jffs_fm_write_unlock(fmc);
+ D(printk("jffs_write_node(): "
+ "jffs_write_dummy_node(): Failed!\n"));
+ return err;
+ }
+ fm = NULL;
+ }
+ } /* while(!fm) */
+ node->fm = fm;
+
+ ASSERT(if (fm->nodes == 0) {
+ printk(KERN_ERR "jffs_write_node(): fm->nodes == 0\n");
+ });
+
+ pos = node->fm->offset;
+
+ /* Increment the version number here. We can't let the caller
+ set it beforehand, because we might have had to do GC on a node
+ of this file - and we'd end up reusing version numbers.
+ */
+ if (f) {
+ raw_inode->version = f->highest_version + 1;
+ D1(printk (KERN_NOTICE "jffs_write_node(): setting version of %s to %d\n", f->name, raw_inode->version));
+
+ /* if the file was deleted, set the deleted bit in the raw inode */
+ if (f->deleted)
+ raw_inode->deleted = 1;
+ }
+
+ /* Compute the checksum for the data and name chunks. */
+ raw_inode->dchksum = jffs_checksum(data, raw_inode->dsize);
+ raw_inode->nchksum = jffs_checksum(name, raw_inode->nsize);
+
+ /* The checksum is calculated without the chksum and accurate
+ fields so set them to zero first. */
+ raw_inode->accurate = 0;
+ raw_inode->chksum = 0;
+ raw_inode->chksum = jffs_checksum(raw_inode,
+ sizeof(struct jffs_raw_inode));
+ raw_inode->accurate = 0xff;
+
+ D3(printk("jffs_write_node(): About to write this raw inode to the "
+ "flash at pos 0x%lx:\n", (long)pos));
+ D3(jffs_print_raw_inode(raw_inode));
+
+ /* The actual raw JFFS node */
+ node_iovec[0].iov_base = (void *) raw_inode;
+ node_iovec[0].iov_len = (size_t) sizeof(struct jffs_raw_inode);
+ iovec_cnt = 1;
+
+ /* Get name and size if there is one */
+ if (raw_inode->nsize) {
+ node_iovec[iovec_cnt].iov_base = (void *) name;
+ node_iovec[iovec_cnt].iov_len = (size_t) raw_inode->nsize;
+ iovec_cnt++;
+
+ if (JFFS_GET_PAD_BYTES(raw_inode->nsize)) {
+ static char allff[3]={255,255,255};
+ /* Add some extra padding if necessary */
+ node_iovec[iovec_cnt].iov_base = allff;
+ node_iovec[iovec_cnt].iov_len =
+ JFFS_GET_PAD_BYTES(raw_inode->nsize);
+ iovec_cnt++;
+ }
+ }
+
+ /* Get data and size if there is any */
+ if (raw_inode->dsize) {
+ node_iovec[iovec_cnt].iov_base = (void *) data;
+ node_iovec[iovec_cnt].iov_len = (size_t) raw_inode->dsize;
+ iovec_cnt++;
+ /* No need to pad this because we're not actually putting
+ anything after it.
+ */
+ }
+
+ if ((err = flash_safe_writev(fmc->mtd, node_iovec, iovec_cnt,
+ pos)) < 0) {
+ jffs_fmfree_partly(fmc, fm, 0);
+ jffs_fm_write_unlock(fmc);
+ printk(KERN_ERR "JFFS: jffs_write_node: Failed to write, "
+ "requested %i, wrote %i\n", total_size, err);
+ goto retry;
+ }
+ if (raw_inode->deleted)
+ f->deleted = 1;
+
+ jffs_fm_write_unlock(fmc);
+ D3(printk("jffs_write_node(): Leaving...\n"));
+ return raw_inode->dsize;
+} /* jffs_write_node() */
+
+
+/* Read data from the node and write it to the buffer. 'node_offset'
+ is how much we have read from this particular node before and which
+ shouldn't be read again. 'max_size' is how much space there is in
+ the buffer. */
+static int
+jffs_get_node_data(struct jffs_file *f, struct jffs_node *node,
+ unsigned char *buf,__u32 node_offset, __u32 max_size)
+{
+ struct jffs_fmcontrol *fmc = f->c->fmc;
+ __u32 pos = node->fm->offset + node->fm_offset + node_offset;
+ __u32 avail = node->data_size - node_offset;
+ __u32 r;
+
+ D2(printk(" jffs_get_node_data(): file: \"%s\", ino: %u, "
+ "version: %u, node_offset: %u\n",
+ f->name, node->ino, node->version, node_offset));
+
+ r = min(avail, max_size);
+ D3(printk(KERN_NOTICE "jffs_get_node_data\n"));
+ flash_safe_read(fmc->mtd, pos, buf, r);
+
+ D3(printk(" jffs_get_node_data(): Read %u byte%s.\n",
+ r, (r == 1 ? "" : "s")));
+
+ return r;
+}
+
+
+/* Read data from the file's nodes. Write the data to the buffer
+ 'buf'. 'read_offset' tells how much data we should skip. */
+int
+jffs_read_data(struct jffs_file *f, unsigned char *buf, __u32 read_offset,
+ __u32 size)
+{
+ struct jffs_node *node;
+ __u32 read_data = 0; /* Total amount of read data. */
+ __u32 node_offset = 0;
+ __u32 pos = 0; /* Number of bytes traversed. */
+
+ D2(printk("jffs_read_data(): file = \"%s\", read_offset = %d, "
+ "size = %u\n",
+ (f->name ? f->name : ""), read_offset, size));
+
+ if (read_offset >= f->size) {
+ D(printk(" f->size: %d\n", f->size));
+ return 0;
+ }
+
+ /* First find the node to read data from. */
+ node = f->range_head;
+ while (pos <= read_offset) {
+ node_offset = read_offset - pos;
+ if (node_offset >= node->data_size) {
+ pos += node->data_size;
+ node = node->range_next;
+ }
+ else {
+ break;
+ }
+ }
+
+ /* "Cats are living proof that not everything in nature
+ has to be useful."
+ - Garrison Keilor ('97) */
+
+ /* Fill the buffer. */
+ while (node && (read_data < size)) {
+ int r;
+ if (!node->fm) {
+ /* This node does not refer to real data. */
+ r = min(size - read_data,
+ node->data_size - node_offset);
+ memset(&buf[read_data], 0, r);
+ }
+ else if ((r = jffs_get_node_data(f, node, &buf[read_data],
+ node_offset,
+ size - read_data)) < 0) {
+ return r;
+ }
+ read_data += r;
+ node_offset = 0;
+ node = node->range_next;
+ }
+ D3(printk(" jffs_read_data(): Read %u bytes.\n", read_data));
+ return read_data;
+}
+
+
+/* Used for traversing all nodes in the hash table. */
+int
+jffs_foreach_file(struct jffs_control *c, int (*func)(struct jffs_file *))
+{
+ int pos;
+ int r;
+ int result = 0;
+
+ for (pos = 0; pos < c->hash_len; pos++) {
+ struct list_head *p, *next;
+ for (p = c->hash[pos].next; p != &c->hash[pos]; p = next) {
+ /* We need a reference to the next file in the
+ list because `func' might remove the current
+ file `f'. */
+ next = p->next;
+ r = func(list_entry(p, struct jffs_file, hash));
+ if (r < 0)
+ return r;
+ result += r;
+ }
+ }
+
+ return result;
+}
+
+
+/* Free all nodes associated with a file. */
+static int
+jffs_free_node_list(struct jffs_file *f)
+{
+ struct jffs_node *node;
+ struct jffs_node *p;
+
+ D3(printk("jffs_free_node_list(): f #%u, \"%s\"\n",
+ f->ino, (f->name ? f->name : "")));
+ node = f->version_head;
+ while (node) {
+ p = node;
+ node = node->version_next;
+ jffs_free_node(p);
+ DJM(no_jffs_node--);
+ }
+ return 0;
+}
+
+
+/* Free a file and its name. */
+static int
+jffs_free_file(struct jffs_file *f)
+{
+ D3(printk("jffs_free_file: f #%u, \"%s\"\n",
+ f->ino, (f->name ? f->name : "")));
+
+ if (f->name) {
+ kfree(f->name);
+ DJM(no_name--);
+ }
+ kfree(f);
+ no_jffs_file--;
+ return 0;
+}
+
+static long
+jffs_get_file_count(void)
+{
+ return no_jffs_file;
+}
+
+/* See if a file is deleted. If so, mark that file's nodes as obsolete. */
+int
+jffs_possibly_delete_file(struct jffs_file *f)
+{
+ struct jffs_node *n;
+
+ D3(printk("jffs_possibly_delete_file(): ino: %u\n",
+ f->ino));
+
+ ASSERT(if (!f) {
+ printk(KERN_ERR "jffs_possibly_delete_file(): f == NULL\n");
+ return -1;
+ });
+
+ if (f->deleted) {
+ /* First try to remove all older versions. Commence with
+ the oldest node. */
+ for (n = f->version_head; n; n = n->version_next) {
+ if (!n->fm) {
+ continue;
+ }
+ if (jffs_fmfree(f->c->fmc, n->fm, n) < 0) {
+ break;
+ }
+ }
+ /* Unlink the file from the filesystem. */
+ if (!f->c->building_fs) {
+ jffs_unlink_file_from_tree(f);
+ }
+ jffs_unlink_file_from_hash(f);
+ jffs_free_node_list(f);
+ jffs_free_file(f);
+ }
+ return 0;
+}
+
+
+/* Used in conjunction with jffs_foreach_file() to count the number
+ of files in the file system. */
+int
+jffs_file_count(struct jffs_file *f)
+{
+ return 1;
+}
+
+
+/* Build up a file's range list from scratch by going through the
+ version list. */
+static int
+jffs_build_file(struct jffs_file *f)
+{
+ struct jffs_node *n;
+
+ D3(printk("jffs_build_file(): ino: %u, name: \"%s\"\n",
+ f->ino, (f->name ? f->name : "")));
+
+ for (n = f->version_head; n; n = n->version_next) {
+ jffs_update_file(f, n);
+ }
+ return 0;
+}
+
+
+/* Remove an amount of data from a file. If this amount of data is
+ zero, that could mean that a node should be split in two parts.
+ We remove or change the appropriate nodes in the lists.
+
+ Starting offset of area to be removed is node->data_offset,
+ and the length of the area is in node->removed_size. */
+static int
+jffs_delete_data(struct jffs_file *f, struct jffs_node *node)
+{
+ struct jffs_node *n;
+ __u32 offset = node->data_offset;
+ __u32 remove_size = node->removed_size;
+
+ D3(printk("jffs_delete_data(): offset = %u, remove_size = %u\n",
+ offset, remove_size));
+
+ if (remove_size == 0
+ && f->range_tail
+ && f->range_tail->data_offset + f->range_tail->data_size
+ == offset) {
+ /* A simple append; nothing to remove or no node to split. */
+ return 0;
+ }
+
+ /* Find the node where we should begin the removal. */
+ for (n = f->range_head; n; n = n->range_next) {
+ if (n->data_offset + n->data_size > offset) {
+ break;
+ }
+ }
+ if (!n) {
+ /* If there's no data in the file there's no data to
+ remove either. */
+ return 0;
+ }
+
+ if (n->data_offset > offset) {
+ /* XXX: Not implemented yet. */
+ printk(KERN_WARNING "JFFS: An unexpected situation "
+ "occurred in jffs_delete_data.\n");
+ }
+ else if (n->data_offset < offset) {
+ /* See if the node has to be split into two parts. */
+ if (n->data_offset + n->data_size > offset + remove_size) {
+ /* Do the split. */
+ struct jffs_node *new_node;
+ D3(printk("jffs_delete_data(): Split node with "
+ "version number %u.\n", n->version));
+
+ if (!(new_node = jffs_alloc_node())) {
+ D(printk("jffs_delete_data(): -ENOMEM\n"));
+ return -ENOMEM;
+ }
+ DJM(no_jffs_node++);
+
+ new_node->ino = n->ino;
+ new_node->version = n->version;
+ new_node->data_offset = offset;
+ new_node->data_size = n->data_size - (remove_size + (offset - n->data_offset));
+ new_node->fm_offset = n->fm_offset + (remove_size + (offset - n->data_offset));
+ new_node->name_size = n->name_size;
+ new_node->fm = n->fm;
+ new_node->version_prev = n;
+ new_node->version_next = n->version_next;
+ if (new_node->version_next) {
+ new_node->version_next->version_prev
+ = new_node;
+ }
+ else {
+ f->version_tail = new_node;
+ }
+ n->version_next = new_node;
+ new_node->range_prev = n;
+ new_node->range_next = n->range_next;
+ if (new_node->range_next) {
+ new_node->range_next->range_prev = new_node;
+ }
+ else {
+ f->range_tail = new_node;
+ }
+ /* A very interesting can of worms. */
+ n->range_next = new_node;
+ n->data_size = offset - n->data_offset;
+ if (new_node->fm)
+ jffs_add_node(new_node);
+ else {
+ D1(printk(KERN_WARNING "jffs_delete_data(): Splitting an empty node (file hold).\n!"));
+ D1(printk(KERN_WARNING "FIXME: Did dwmw2 do the right thing here?\n"));
+ }
+ n = new_node->range_next;
+ remove_size = 0;
+ }
+ else {
+ /* No. No need to split the node. Just remove
+ the end of the node. */
+ int r = min(n->data_offset + n->data_size
+ - offset, remove_size);
+ n->data_size -= r;
+ remove_size -= r;
+ n = n->range_next;
+ }
+ }
+
+ /* Remove as many nodes as necessary. */
+ while (n && remove_size) {
+ if (n->data_size <= remove_size) {
+ struct jffs_node *p = n;
+ remove_size -= n->data_size;
+ n = n->range_next;
+ D3(printk("jffs_delete_data(): Removing node: "
+ "ino: %u, version: %u%s\n",
+ p->ino, p->version,
+ (p->fm ? "" : " (virtual)")));
+ if (p->fm) {
+ jffs_fmfree(f->c->fmc, p->fm, p);
+ }
+ jffs_unlink_node_from_range_list(f, p);
+ jffs_unlink_node_from_version_list(f, p);
+ jffs_free_node(p);
+ DJM(no_jffs_node--);
+ }
+ else {
+ n->data_size -= remove_size;
+ n->fm_offset += remove_size;
+ n->data_offset -= (node->removed_size - remove_size);
+ n = n->range_next;
+ break;
+ }
+ }
+
+ /* Adjust the following nodes' information about offsets etc. */
+ while (n && node->removed_size) {
+ n->data_offset -= node->removed_size;
+ n = n->range_next;
+ }
+
+ if (node->removed_size > (f->size - node->data_offset)) {
+ /* It's possible that the removed_size is in fact
+ * greater than the amount of data we actually thought
+ * were present in the first place - some of the nodes
+ * which this node originally obsoleted may already have
+ * been deleted from the flash by subsequent garbage
+ * collection.
+ *
+ * If this is the case, don't let f->size go negative.
+ * Bad things would happen :)
+ */
+ f->size = node->data_offset;
+ } else {
+ f->size -= node->removed_size;
+ }
+ D3(printk("jffs_delete_data(): f->size = %d\n", f->size));
+ return 0;
+} /* jffs_delete_data() */
+
+
+/* Insert some data into a file. Prior to the call to this function,
+ jffs_delete_data should be called. */
+static int
+jffs_insert_data(struct jffs_file *f, struct jffs_node *node)
+{
+ D3(printk("jffs_insert_data(): node->data_offset = %u, "
+ "node->data_size = %u, f->size = %u\n",
+ node->data_offset, node->data_size, f->size));
+
+ /* Find the position where we should insert data. */
+ retry:
+ if (node->data_offset == f->size) {
+ /* A simple append. This is the most common operation. */
+ node->range_next = NULL;
+ node->range_prev = f->range_tail;
+ if (node->range_prev) {
+ node->range_prev->range_next = node;
+ }
+ f->range_tail = node;
+ f->size += node->data_size;
+ if (!f->range_head) {
+ f->range_head = node;
+ }
+ }
+ else if (node->data_offset < f->size) {
+ /* Trying to insert data into the middle of the file. This
+ means no problem because jffs_delete_data() has already
+ prepared the range list for us. */
+ struct jffs_node *n;
+
+ /* Find the correct place for the insertion and then insert
+ the node. */
+ for (n = f->range_head; n; n = n->range_next) {
+ D2(printk("Cool stuff's happening!\n"));
+
+ if (n->data_offset == node->data_offset) {
+ node->range_prev = n->range_prev;
+ if (node->range_prev) {
+ node->range_prev->range_next = node;
+ }
+ else {
+ f->range_head = node;
+ }
+ node->range_next = n;
+ n->range_prev = node;
+ break;
+ }
+ ASSERT(else if (n->data_offset + n->data_size >
+ node->data_offset) {
+ printk(KERN_ERR "jffs_insert_data(): "
+ "Couldn't find a place to insert "
+ "the data!\n");
+ return -1;
+ });
+ }
+
+ /* Adjust later nodes' offsets etc. */
+ n = node->range_next;
+ while (n) {
+ n->data_offset += node->data_size;
+ n = n->range_next;
+ }
+ f->size += node->data_size;
+ }
+ else if (node->data_offset > f->size) {
+ /* Okay. This is tricky. This means that we want to insert
+ data at a place that is beyond the limits of the file as
+ it is constructed right now. This is actually a common
+ event that for instance could occur during the mounting
+ of the file system if a large file have been truncated,
+ rewritten and then only partially garbage collected. */
+
+ struct jffs_node *n;
+
+ /* We need a place holder for the data that is missing in
+ front of this insertion. This "virtual node" will not
+ be associated with any space on the flash device. */
+ struct jffs_node *virtual_node;
+ if (!(virtual_node = jffs_alloc_node())) {
+ return -ENOMEM;
+ }
+
+ D(printk("jffs_insert_data: Inserting a virtual node.\n"));
+ D(printk(" node->data_offset = %u\n", node->data_offset));
+ D(printk(" f->size = %u\n", f->size));
+
+ virtual_node->ino = node->ino;
+ virtual_node->version = node->version;
+ virtual_node->removed_size = 0;
+ virtual_node->fm_offset = 0;
+ virtual_node->name_size = 0;
+ virtual_node->fm = NULL; /* This is a virtual data holder. */
+ virtual_node->version_prev = NULL;
+ virtual_node->version_next = NULL;
+ virtual_node->range_next = NULL;
+
+ /* Are there any data at all in the file yet? */
+ if (f->range_head) {
+ virtual_node->data_offset
+ = f->range_tail->data_offset
+ + f->range_tail->data_size;
+ virtual_node->data_size
+ = node->data_offset - virtual_node->data_offset;
+ virtual_node->range_prev = f->range_tail;
+ f->range_tail->range_next = virtual_node;
+ }
+ else {
+ virtual_node->data_offset = 0;
+ virtual_node->data_size = node->data_offset;
+ virtual_node->range_prev = NULL;
+ f->range_head = virtual_node;
+ }
+
+ f->range_tail = virtual_node;
+ f->size += virtual_node->data_size;
+
+ /* Insert this virtual node in the version list as well. */
+ for (n = f->version_head; n ; n = n->version_next) {
+ if (n->version == virtual_node->version) {
+ virtual_node->version_prev = n->version_prev;
+ n->version_prev = virtual_node;
+ if (virtual_node->version_prev) {
+ virtual_node->version_prev
+ ->version_next = virtual_node;
+ }
+ else {
+ f->version_head = virtual_node;
+ }
+ virtual_node->version_next = n;
+ break;
+ }
+ }
+
+ D(jffs_print_node(virtual_node));
+
+ /* Make a new try to insert the node. */
+ goto retry;
+ }
+
+ D3(printk("jffs_insert_data(): f->size = %d\n", f->size));
+ return 0;
+}
+
+
+/* A new node (with data) has been added to the file and now the range
+ list has to be modified. */
+static int
+jffs_update_file(struct jffs_file *f, struct jffs_node *node)
+{
+ int err;
+
+ D3(printk("jffs_update_file(): ino: %u, version: %u\n",
+ f->ino, node->version));
+
+ if (node->data_size == 0) {
+ if (node->removed_size == 0) {
+ /* data_offset == X */
+ /* data_size == 0 */
+ /* remove_size == 0 */
+ }
+ else {
+ /* data_offset == X */
+ /* data_size == 0 */
+ /* remove_size != 0 */
+ if ((err = jffs_delete_data(f, node)) < 0) {
+ return err;
+ }
+ }
+ }
+ else {
+ /* data_offset == X */
+ /* data_size != 0 */
+ /* remove_size == Y */
+ if ((err = jffs_delete_data(f, node)) < 0) {
+ return err;
+ }
+ if ((err = jffs_insert_data(f, node)) < 0) {
+ return err;
+ }
+ }
+ return 0;
+}
+
+/* Print the contents of a node. */
+void
+jffs_print_node(struct jffs_node *n)
+{
+ D(printk("jffs_node: 0x%p\n", n));
+ D(printk("{\n"));
+ D(printk(" 0x%08x, /* version */\n", n->version));
+ D(printk(" 0x%08x, /* data_offset */\n", n->data_offset));
+ D(printk(" 0x%08x, /* data_size */\n", n->data_size));
+ D(printk(" 0x%08x, /* removed_size */\n", n->removed_size));
+ D(printk(" 0x%08x, /* fm_offset */\n", n->fm_offset));
+ D(printk(" 0x%02x, /* name_size */\n", n->name_size));
+ D(printk(" 0x%p, /* fm, fm->offset: %u */\n",
+ n->fm, (n->fm ? n->fm->offset : 0)));
+ D(printk(" 0x%p, /* version_prev */\n", n->version_prev));
+ D(printk(" 0x%p, /* version_next */\n", n->version_next));
+ D(printk(" 0x%p, /* range_prev */\n", n->range_prev));
+ D(printk(" 0x%p, /* range_next */\n", n->range_next));
+ D(printk("}\n"));
+}
+
+
+/* Print the contents of a raw inode. */
+void
+jffs_print_raw_inode(struct jffs_raw_inode *raw_inode)
+{
+ D(printk("jffs_raw_inode: inode number: %u\n", raw_inode->ino));
+ D(printk("{\n"));
+ D(printk(" 0x%08x, /* magic */\n", raw_inode->magic));
+ D(printk(" 0x%08x, /* ino */\n", raw_inode->ino));
+ D(printk(" 0x%08x, /* pino */\n", raw_inode->pino));
+ D(printk(" 0x%08x, /* version */\n", raw_inode->version));
+ D(printk(" 0x%08x, /* mode */\n", raw_inode->mode));
+ D(printk(" 0x%04x, /* uid */\n", raw_inode->uid));
+ D(printk(" 0x%04x, /* gid */\n", raw_inode->gid));
+ D(printk(" 0x%08x, /* atime */\n", raw_inode->atime));
+ D(printk(" 0x%08x, /* mtime */\n", raw_inode->mtime));
+ D(printk(" 0x%08x, /* ctime */\n", raw_inode->ctime));
+ D(printk(" 0x%08x, /* offset */\n", raw_inode->offset));
+ D(printk(" 0x%08x, /* dsize */\n", raw_inode->dsize));
+ D(printk(" 0x%08x, /* rsize */\n", raw_inode->rsize));
+ D(printk(" 0x%02x, /* nsize */\n", raw_inode->nsize));
+ D(printk(" 0x%02x, /* nlink */\n", raw_inode->nlink));
+ D(printk(" 0x%02x, /* spare */\n",
+ raw_inode->spare));
+ D(printk(" %u, /* rename */\n",
+ raw_inode->rename));
+ D(printk(" %u, /* deleted */\n",
+ raw_inode->deleted));
+ D(printk(" 0x%02x, /* accurate */\n",
+ raw_inode->accurate));
+ D(printk(" 0x%08x, /* dchksum */\n", raw_inode->dchksum));
+ D(printk(" 0x%04x, /* nchksum */\n", raw_inode->nchksum));
+ D(printk(" 0x%04x, /* chksum */\n", raw_inode->chksum));
+ D(printk("}\n"));
+}
+
+
+/* Print the contents of a file. */
+#if 0
+int
+jffs_print_file(struct jffs_file *f)
+{
+ D(int i);
+ D(printk("jffs_file: 0x%p\n", f));
+ D(printk("{\n"));
+ D(printk(" 0x%08x, /* ino */\n", f->ino));
+ D(printk(" 0x%08x, /* pino */\n", f->pino));
+ D(printk(" 0x%08x, /* mode */\n", f->mode));
+ D(printk(" 0x%04x, /* uid */\n", f->uid));
+ D(printk(" 0x%04x, /* gid */\n", f->gid));
+ D(printk(" 0x%08x, /* atime */\n", f->atime));
+ D(printk(" 0x%08x, /* mtime */\n", f->mtime));
+ D(printk(" 0x%08x, /* ctime */\n", f->ctime));
+ D(printk(" 0x%02x, /* nsize */\n", f->nsize));
+ D(printk(" 0x%02x, /* nlink */\n", f->nlink));
+ D(printk(" 0x%02x, /* deleted */\n", f->deleted));
+ D(printk(" \"%s\", ", (f->name ? f->name : "")));
+ D(for (i = strlen(f->name ? f->name : ""); i < 8; ++i) {
+ printk(" ");
+ });
+ D(printk("/* name */\n"));
+ D(printk(" 0x%08x, /* size */\n", f->size));
+ D(printk(" 0x%08x, /* highest_version */\n",
+ f->highest_version));
+ D(printk(" 0x%p, /* c */\n", f->c));
+ D(printk(" 0x%p, /* parent */\n", f->parent));
+ D(printk(" 0x%p, /* children */\n", f->children));
+ D(printk(" 0x%p, /* sibling_prev */\n", f->sibling_prev));
+ D(printk(" 0x%p, /* sibling_next */\n", f->sibling_next));
+ D(printk(" 0x%p, /* hash_prev */\n", f->hash.prev));
+ D(printk(" 0x%p, /* hash_next */\n", f->hash.next));
+ D(printk(" 0x%p, /* range_head */\n", f->range_head));
+ D(printk(" 0x%p, /* range_tail */\n", f->range_tail));
+ D(printk(" 0x%p, /* version_head */\n", f->version_head));
+ D(printk(" 0x%p, /* version_tail */\n", f->version_tail));
+ D(printk("}\n"));
+ return 0;
+}
+#endif /* 0 */
+
+void
+jffs_print_hash_table(struct jffs_control *c)
+{
+ int i;
+
+ printk("JFFS: Dumping the file system's hash table...\n");
+ for (i = 0; i < c->hash_len; i++) {
+ struct list_head *p;
+ for (p = c->hash[i].next; p != &c->hash[i]; p = p->next) {
+ struct jffs_file *f=list_entry(p,struct jffs_file,hash);
+ printk("*** c->hash[%u]: \"%s\" "
+ "(ino: %u, pino: %u)\n",
+ i, (f->name ? f->name : ""),
+ f->ino, f->pino);
+ }
+ }
+}
+
+
+void
+jffs_print_tree(struct jffs_file *first_file, int indent)
+{
+ struct jffs_file *f;
+ char *space;
+ int dir;
+
+ if (!first_file) {
+ return;
+ }
+
+ if (!(space = (char *) kmalloc(indent + 1, GFP_KERNEL))) {
+ printk("jffs_print_tree(): Out of memory!\n");
+ return;
+ }
+
+ memset(space, ' ', indent);
+ space[indent] = '\0';
+
+ for (f = first_file; f; f = f->sibling_next) {
+ dir = S_ISDIR(f->mode);
+ printk("%s%s%s (ino: %u, highest_version: %u, size: %u)\n",
+ space, (f->name ? f->name : ""), (dir ? "/" : ""),
+ f->ino, f->highest_version, f->size);
+ if (dir) {
+ jffs_print_tree(f->children, indent + 2);
+ }
+ }
+
+ kfree(space);
+}
+
+
+#if defined(JFFS_MEMORY_DEBUG) && JFFS_MEMORY_DEBUG
+void
+jffs_print_memory_allocation_statistics(void)
+{
+ static long printout;
+ printk("________ Memory printout #%ld ________\n", ++printout);
+ printk("no_jffs_file = %ld\n", no_jffs_file);
+ printk("no_jffs_node = %ld\n", no_jffs_node);
+ printk("no_jffs_control = %ld\n", no_jffs_control);
+ printk("no_jffs_raw_inode = %ld\n", no_jffs_raw_inode);
+ printk("no_jffs_node_ref = %ld\n", no_jffs_node_ref);
+ printk("no_jffs_fm = %ld\n", no_jffs_fm);
+ printk("no_jffs_fmcontrol = %ld\n", no_jffs_fmcontrol);
+ printk("no_hash = %ld\n", no_hash);
+ printk("no_name = %ld\n", no_name);
+ printk("\n");
+}
+#endif
+
+
+/* Rewrite `size' bytes, and begin at `node'. */
+static int
+jffs_rewrite_data(struct jffs_file *f, struct jffs_node *node, __u32 size)
+{
+ struct jffs_control *c = f->c;
+ struct jffs_fmcontrol *fmc = c->fmc;
+ struct jffs_raw_inode raw_inode;
+ struct jffs_node *new_node;
+ struct jffs_fm *fm;
+ __u32 pos;
+ __u32 pos_dchksum;
+ __u32 total_name_size;
+ __u32 total_data_size;
+ __u32 total_size;
+ int err;
+
+ D1(printk("***jffs_rewrite_data(): node: %u, name: \"%s\", size: %u\n",
+ f->ino, (f->name ? f->name : "(null)"), size));
+
+ /* Create and initialize the new node. */
+ if (!(new_node = jffs_alloc_node())) {
+ D(printk("jffs_rewrite_data(): "
+ "Failed to allocate node.\n"));
+ return -ENOMEM;
+ }
+ DJM(no_jffs_node++);
+ new_node->data_offset = node->data_offset;
+ new_node->removed_size = size;
+ total_name_size = JFFS_PAD(f->nsize);
+ total_data_size = JFFS_PAD(size);
+ total_size = sizeof(struct jffs_raw_inode)
+ + total_name_size + total_data_size;
+ new_node->fm_offset = sizeof(struct jffs_raw_inode)
+ + total_name_size;
+
+retry:
+ jffs_fm_write_lock(fmc);
+ err = 0;
+
+ if ((err = jffs_fmalloc(fmc, total_size, new_node, &fm)) < 0) {
+ DJM(no_jffs_node--);
+ jffs_fm_write_unlock(fmc);
+ D(printk("jffs_rewrite_data(): Failed to allocate fm.\n"));
+ jffs_free_node(new_node);
+ return err;
+ }
+ else if (!fm->nodes) {
+ /* The jffs_fm struct that we got is not big enough. */
+ /* This should never happen, because we deal with this case
+ in jffs_garbage_collect_next().*/
+ printk(KERN_WARNING "jffs_rewrite_data(): Allocated node is too small (%d bytes of %d)\n", fm->size, total_size);
+ if ((err = jffs_write_dummy_node(c, fm)) < 0) {
+ D(printk("jffs_rewrite_data(): "
+ "jffs_write_dummy_node() Failed!\n"));
+ } else {
+ err = -ENOSPC;
+ }
+ DJM(no_jffs_fm--);
+ jffs_fm_write_unlock(fmc);
+ kfree(fm);
+
+ return err;
+ }
+ new_node->fm = fm;
+
+ /* Initialize the raw inode. */
+ raw_inode.magic = JFFS_MAGIC_BITMASK;
+ raw_inode.ino = f->ino;
+ raw_inode.pino = f->pino;
+ raw_inode.version = f->highest_version + 1;
+ raw_inode.mode = f->mode;
+ raw_inode.uid = f->uid;
+ raw_inode.gid = f->gid;
+ raw_inode.atime = f->atime;
+ raw_inode.mtime = f->mtime;
+ raw_inode.ctime = f->ctime;
+ raw_inode.offset = node->data_offset;
+ raw_inode.dsize = size;
+ raw_inode.rsize = size;
+ raw_inode.nsize = f->nsize;
+ raw_inode.nlink = f->nlink;
+ raw_inode.spare = 0;
+ raw_inode.rename = 0;
+ raw_inode.deleted = f->deleted;
+ raw_inode.accurate = 0xff;
+ raw_inode.dchksum = 0;
+ raw_inode.nchksum = 0;
+
+ pos = new_node->fm->offset;
+ pos_dchksum = pos +JFFS_RAW_INODE_DCHKSUM_OFFSET;
+
+ D3(printk("jffs_rewrite_data(): Writing this raw inode "
+ "to pos 0x%ul.\n", pos));
+ D3(jffs_print_raw_inode(&raw_inode));
+
+ if ((err = flash_safe_write(fmc->mtd, pos,
+ (u_char *) &raw_inode,
+ sizeof(struct jffs_raw_inode)
+ - sizeof(__u32)
+ - sizeof(__u16) - sizeof(__u16))) < 0) {
+ jffs_fmfree_partly(fmc, fm,
+ total_name_size + total_data_size);
+ jffs_fm_write_unlock(fmc);
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: Write error during "
+ "rewrite. (raw inode)\n");
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: Now retrying "
+ "rewrite. (raw inode)\n");
+ goto retry;
+ }
+ pos += sizeof(struct jffs_raw_inode);
+
+ /* Write the name to the flash memory. */
+ if (f->nsize) {
+ D3(printk("jffs_rewrite_data(): Writing name \"%s\" to "
+ "pos 0x%ul.\n", f->name, (unsigned int) pos));
+ if ((err = flash_safe_write(fmc->mtd, pos,
+ (u_char *)f->name,
+ f->nsize)) < 0) {
+ jffs_fmfree_partly(fmc, fm, total_data_size);
+ jffs_fm_write_unlock(fmc);
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: Write "
+ "error during rewrite. (name)\n");
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: Now retrying "
+ "rewrite. (name)\n");
+ goto retry;
+ }
+ pos += total_name_size;
+ raw_inode.nchksum = jffs_checksum(f->name, f->nsize);
+ }
+
+ /* Write the data. */
+ if (size) {
+ int r;
+ unsigned char *page;
+ __u32 offset = node->data_offset;
+
+ if (!(page = (unsigned char *)__get_free_page(GFP_KERNEL))) {
+ jffs_fmfree_partly(fmc, fm, 0);
+ return -1;
+ }
+
+ while (size) {
+ __u32 s = min(size, (__u32)PAGE_SIZE);
+ if ((r = jffs_read_data(f, (char *)page,
+ offset, s)) < s) {
+ free_page((unsigned long)page);
+ jffs_fmfree_partly(fmc, fm, 0);
+ jffs_fm_write_unlock(fmc);
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: "
+ "jffs_read_data() "
+ "failed! (r = %d)\n", r);
+ return -1;
+ }
+ if ((err = flash_safe_write(fmc->mtd,
+ pos, page, r)) < 0) {
+ free_page((unsigned long)page);
+ jffs_fmfree_partly(fmc, fm, 0);
+ jffs_fm_write_unlock(fmc);
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: "
+ "Write error during rewrite. "
+ "(data)\n");
+ goto retry;
+ }
+ pos += r;
+ size -= r;
+ offset += r;
+ raw_inode.dchksum += jffs_checksum(page, r);
+ }
+
+ free_page((unsigned long)page);
+ }
+
+ raw_inode.accurate = 0;
+ raw_inode.chksum = jffs_checksum(&raw_inode,
+ sizeof(struct jffs_raw_inode)
+ - sizeof(__u16));
+
+ /* Add the checksum. */
+ if ((err
+ = flash_safe_write(fmc->mtd, pos_dchksum,
+ &((u_char *)
+ &raw_inode)[JFFS_RAW_INODE_DCHKSUM_OFFSET],
+ sizeof(__u32) + sizeof(__u16)
+ + sizeof(__u16))) < 0) {
+ jffs_fmfree_partly(fmc, fm, 0);
+ jffs_fm_write_unlock(fmc);
+ printk(KERN_ERR "JFFS: jffs_rewrite_data: Write error during "
+ "rewrite. (checksum)\n");
+ goto retry;
+ }
+
+ /* Now make the file system aware of the newly written node. */
+ jffs_insert_node(c, f, &raw_inode, f->name, new_node);
+ jffs_fm_write_unlock(fmc);
+
+ D3(printk("jffs_rewrite_data(): Leaving...\n"));
+ return 0;
+} /* jffs_rewrite_data() */
+
+
+/* jffs_garbage_collect_next implements one step in the garbage collect
+ process and is often called multiple times at each occasion of a
+ garbage collect. */
+
+static int
+jffs_garbage_collect_next(struct jffs_control *c)
+{
+ struct jffs_fmcontrol *fmc = c->fmc;
+ struct jffs_node *node;
+ struct jffs_file *f;
+ int err = 0;
+ __u32 size;
+ __u32 data_size;
+ __u32 total_name_size;
+ __u32 extra_available;
+ __u32 space_needed;
+ __u32 free_chunk_size1 = jffs_free_size1(fmc);
+ D2(__u32 free_chunk_size2 = jffs_free_size2(fmc));
+
+ /* Get the oldest node in the flash. */
+ node = jffs_get_oldest_node(fmc);
+ ASSERT(if (!node) {
+ printk(KERN_ERR "JFFS: jffs_garbage_collect_next: "
+ "No oldest node found!\n");
+ err = -1;
+ goto jffs_garbage_collect_next_end;
+
+
+ });
+
+ /* Find its corresponding file too. */
+ f = jffs_find_file(c, node->ino);
+
+ if (!f) {
+ printk (KERN_ERR "JFFS: jffs_garbage_collect_next: "
+ "No file to garbage collect! "
+ "(ino = 0x%08x)\n", node->ino);
+ /* FIXME: Free the offending node and recover. */
+ err = -1;
+ goto jffs_garbage_collect_next_end;
+ }
+
+ /* We always write out the name. Theoretically, we don't need
+ to, but for now it's easier - because otherwise we'd have
+ to keep track of how many times the current name exists on
+ the flash and make sure it never reaches zero.
+
+ The current approach means that would be possible to cause
+ the GC to end up eating its tail by writing lots of nodes
+ with no name for it to garbage-collect. Hence the change in
+ inode.c to write names with _every_ node.
+
+ It sucks, but it _should_ work.
+ */
+ total_name_size = JFFS_PAD(f->nsize);
+
+ D1(printk("jffs_garbage_collect_next(): \"%s\", "
+ "ino: %u, version: %u, location 0x%x, dsize %u\n",
+ (f->name ? f->name : ""), node->ino, node->version,
+ node->fm->offset, node->data_size));
+
+ /* Compute how many data it's possible to rewrite at the moment. */
+ data_size = f->size - node->data_offset;
+
+ /* And from that, the total size of the chunk we want to write */
+ size = sizeof(struct jffs_raw_inode) + total_name_size
+ + data_size + JFFS_GET_PAD_BYTES(data_size);
+
+ /* If that's more than max_chunk_size, reduce it accordingly */
+ if (size > fmc->max_chunk_size) {
+ size = fmc->max_chunk_size;
+ data_size = size - sizeof(struct jffs_raw_inode)
+ - total_name_size;
+ }
+
+ /* If we're asking to take up more space than free_chunk_size1
+ but we _could_ fit in it, shrink accordingly.
+ */
+ if (size > free_chunk_size1) {
+
+ if (free_chunk_size1 <
+ (sizeof(struct jffs_raw_inode) + total_name_size + BLOCK_SIZE)){
+ /* The space left is too small to be of any
+ use really. */
+ struct jffs_fm *dirty_fm
+ = jffs_fmalloced(fmc,
+ fmc->tail->offset + fmc->tail->size,
+ free_chunk_size1, NULL);
+ if (!dirty_fm) {
+ printk(KERN_ERR "JFFS: "
+ "jffs_garbage_collect_next: "
+ "Failed to allocate `dirty' "
+ "flash memory!\n");
+ err = -1;
+ goto jffs_garbage_collect_next_end;
+ }
+ D1(printk("Dirtying end of flash - too small\n"));
+ jffs_write_dummy_node(c, dirty_fm);
+ err = 0;
+ goto jffs_garbage_collect_next_end;
+ }
+ D1(printk("Reducing size of new node from %d to %d to avoid "
+ " exceeding free_chunk_size1\n",
+ size, free_chunk_size1));
+
+ size = free_chunk_size1;
+ data_size = size - sizeof(struct jffs_raw_inode)
+ - total_name_size;
+ }
+
+
+ /* Calculate the amount of space needed to hold the nodes
+ which are remaining in the tail */
+ space_needed = fmc->min_free_size - (node->fm->offset % fmc->sector_size);
+
+ /* From that, calculate how much 'extra' space we can use to
+ increase the size of the node we're writing from the size
+ of the node we're obsoleting
+ */
+ if (space_needed > fmc->free_size) {
+ /* If we've gone below min_free_size for some reason,
+ don't fuck up. This is why we have
+ min_free_size > sector_size. Whinge about it though,
+ just so I can convince myself my maths is right.
+ */
+ D1(printk(KERN_WARNING "jffs_garbage_collect_next(): "
+ "space_needed %d exceeded free_size %d\n",
+ space_needed, fmc->free_size));
+ extra_available = 0;
+ } else {
+ extra_available = fmc->free_size - space_needed;
+ }
+
+ /* Check that we don't use up any more 'extra' space than
+ what's available */
+ if (size > JFFS_PAD(node->data_size) + total_name_size +
+ sizeof(struct jffs_raw_inode) + extra_available) {
+ D1(printk("Reducing size of new node from %d to %ld to avoid "
+ "catching our tail\n", size,
+ (long) (JFFS_PAD(node->data_size) + JFFS_PAD(node->name_size) +
+ sizeof(struct jffs_raw_inode) + extra_available)));
+ D1(printk("space_needed = %d, extra_available = %d\n",
+ space_needed, extra_available));
+
+ size = JFFS_PAD(node->data_size) + total_name_size +
+ sizeof(struct jffs_raw_inode) + extra_available;
+ data_size = size - sizeof(struct jffs_raw_inode)
+ - total_name_size;
+ };
+
+ D2(printk(" total_name_size: %u\n", total_name_size));
+ D2(printk(" data_size: %u\n", data_size));
+ D2(printk(" size: %u\n", size));
+ D2(printk(" f->nsize: %u\n", f->nsize));
+ D2(printk(" f->size: %u\n", f->size));
+ D2(printk(" node->data_offset: %u\n", node->data_offset));
+ D2(printk(" free_chunk_size1: %u\n", free_chunk_size1));
+ D2(printk(" free_chunk_size2: %u\n", free_chunk_size2));
+ D2(printk(" node->fm->offset: 0x%08x\n", node->fm->offset));
+
+ if ((err = jffs_rewrite_data(f, node, data_size))) {
+ printk(KERN_WARNING "jffs_rewrite_data() failed: %d\n", err);
+ return err;
+ }
+
+jffs_garbage_collect_next_end:
+ D3(printk("jffs_garbage_collect_next: Leaving...\n"));
+ return err;
+} /* jffs_garbage_collect_next */
+
+
+/* If an obsolete node is partly going to be erased due to garbage
+ collection, the part that isn't going to be erased must be filled
+ with zeroes so that the scan of the flash will work smoothly next
+ time. (The data in the file could for instance be a JFFS image
+ which could cause enormous confusion during a scan of the flash
+ device if we didn't do this.)
+ There are two phases in this procedure: First, the clearing of
+ the name and data parts of the node. Second, possibly also clearing
+ a part of the raw inode as well. If the box is power cycled during
+ the first phase, only the checksum of this node-to-be-cleared-at-
+ the-end will be wrong. If the box is power cycled during, or after,
+ the clearing of the raw inode, the information like the length of
+ the name and data parts are zeroed. The next time the box is
+ powered up, the scanning algorithm manages this faulty data too
+ because:
+
+ - The checksum is invalid and thus the raw inode must be discarded
+ in any case.
+ - If the lengths of the data part or the name part are zeroed, the
+ scanning just continues after the raw inode. But after the inode
+ the scanning procedure just finds zeroes which is the same as
+ dirt.
+
+ So, in the end, this could never fail. :-) Even if it does fail,
+ the scanning algorithm should manage that too. */
+
+static int
+jffs_clear_end_of_node(struct jffs_control *c, __u32 erase_size)
+{
+ struct jffs_fm *fm;
+ struct jffs_fmcontrol *fmc = c->fmc;
+ __u32 zero_offset;
+ __u32 zero_size;
+ __u32 zero_offset_data;
+ __u32 zero_size_data;
+ __u32 cutting_raw_inode = 0;
+
+ if (!(fm = jffs_cut_node(fmc, erase_size))) {
+ D3(printk("jffs_clear_end_of_node(): fm == NULL\n"));
+ return 0;
+ }
+
+ /* Where and how much shall we clear? */
+ zero_offset = fmc->head->offset + erase_size;
+ zero_size = fm->offset + fm->size - zero_offset;
+
+ /* Do we have to clear the raw_inode explicitly? */
+ if (fm->size - zero_size < sizeof(struct jffs_raw_inode)) {
+ cutting_raw_inode = sizeof(struct jffs_raw_inode)
+ - (fm->size - zero_size);
+ }
+
+ /* First, clear the name and data fields. */
+ zero_offset_data = zero_offset + cutting_raw_inode;
+ zero_size_data = zero_size - cutting_raw_inode;
+ flash_safe_acquire(fmc->mtd);
+ flash_memset(fmc->mtd, zero_offset_data, 0, zero_size_data);
+ flash_safe_release(fmc->mtd);
+
+ /* Should we clear a part of the raw inode? */
+ if (cutting_raw_inode) {
+ /* I guess it is ok to clear the raw inode in this order. */
+ flash_safe_acquire(fmc->mtd);
+ flash_memset(fmc->mtd, zero_offset, 0,
+ cutting_raw_inode);
+ flash_safe_release(fmc->mtd);
+ }
+
+ return 0;
+} /* jffs_clear_end_of_node() */
+
+/* Try to erase as much as possible of the dirt in the flash memory. */
+static long
+jffs_try_to_erase(struct jffs_control *c)
+{
+ struct jffs_fmcontrol *fmc = c->fmc;
+ long erase_size;
+ int err;
+ __u32 offset;
+
+ D3(printk("jffs_try_to_erase()\n"));
+
+ erase_size = jffs_erasable_size(fmc);
+
+ D2(printk("jffs_try_to_erase(): erase_size = %ld\n", erase_size));
+
+ if (erase_size == 0) {
+ return 0;
+ }
+ else if (erase_size < 0) {
+ printk(KERN_ERR "JFFS: jffs_try_to_erase: "
+ "jffs_erasable_size returned %ld.\n", erase_size);
+ return erase_size;
+ }
+
+ if ((err = jffs_clear_end_of_node(c, erase_size)) < 0) {
+ printk(KERN_ERR "JFFS: jffs_try_to_erase: "
+ "Clearing of node failed.\n");
+ return err;
+ }
+
+ offset = fmc->head->offset;
+
+ /* Now, let's try to do the erase. */
+ if ((err = flash_erase_region(fmc->mtd,
+ offset, erase_size)) < 0) {
+ printk(KERN_ERR "JFFS: Erase of flash failed. "
+ "offset = %u, erase_size = %ld\n",
+ offset, erase_size);
+ /* XXX: Here we should allocate this area as dirty
+ with jffs_fmalloced or something similar. Now
+ we just report the error. */
+ return err;
+ }
+
+#if 0
+ /* Check if the erased sectors really got erased. */
+ {
+ __u32 pos;
+ __u32 end;
+
+ pos = (__u32)flash_get_direct_pointer(to_kdev_t(c->sb->s_dev), offset);
+ end = pos + erase_size;
+
+ D2(printk("JFFS: Checking erased sector(s)...\n"));
+
+ flash_safe_acquire(fmc->mtd);
+
+ for (; pos < end; pos += 4) {
+ if (*(__u32 *)pos != JFFS_EMPTY_BITMASK) {
+ printk("JFFS: Erase failed! pos = 0x%lx\n",
+ (long)pos);
+ jffs_hexdump(fmc->mtd, pos,
+ jffs_min(256, end - pos));
+ err = -1;
+ break;
+ }
+ }
+
+ flash_safe_release(fmc->mtd);
+
+ if (!err) {
+ D2(printk("JFFS: Erase succeeded.\n"));
+ }
+ else {
+ /* XXX: Here we should allocate the memory
+ with jffs_fmalloced() in order to prevent
+ JFFS from using this area accidentally. */
+ return err;
+ }
+ }
+#endif
+
+ /* Update the flash memory data structures. */
+ jffs_sync_erase(fmc, erase_size);
+
+ return erase_size;
+}
+
+
+/* There are different criteria that should trigger a garbage collect:
+
+ 1. There is too much dirt in the memory.
+ 2. The free space is becoming small.
+ 3. There are many versions of a node.
+
+ The garbage collect should always be done in a manner that guarantees
+ that future garbage collects cannot be locked. E.g. Rewritten chunks
+ should not be too large (span more than one sector in the flash memory
+ for exemple). Of course there is a limit on how intelligent this garbage
+ collection can be. */
+
+
+static int
+jffs_garbage_collect_now(struct jffs_control *c)
+{
+ struct jffs_fmcontrol *fmc = c->fmc;
+ long erased = 0;
+ int result = 0;
+ D1(int i = 1);
+ D2(printk("***jffs_garbage_collect_now(): fmc->dirty_size = %u, fmc->free_size = 0x%x\n, fcs1=0x%x, fcs2=0x%x",
+ fmc->dirty_size, fmc->free_size, jffs_free_size1(fmc), jffs_free_size2(fmc)));
+ D2(jffs_print_fmcontrol(fmc));
+
+ // down(&fmc->gclock);
+
+ /* If it is possible to garbage collect, do so. */
+
+ while (erased == 0) {
+ D1(printk("***jffs_garbage_collect_now(): round #%u, "
+ "fmc->dirty_size = %u\n", i++, fmc->dirty_size));
+ D2(jffs_print_fmcontrol(fmc));
+
+ if ((erased = jffs_try_to_erase(c)) < 0) {
+ printk(KERN_WARNING "JFFS: Error in "
+ "garbage collector.\n");
+ result = erased;
+ goto gc_end;
+ }
+ if (erased)
+ break;
+
+ if (fmc->free_size == 0) {
+ /* Argh */
+ printk(KERN_ERR "jffs_garbage_collect_now(): free_size == 0. This is BAD.\n");
+ result = -ENOSPC;
+ break;
+ }
+
+ if (fmc->dirty_size < fmc->sector_size) {
+ /* Actually, we _may_ have been able to free some,
+ * if there are many overlapping nodes which aren't
+ * actually marked dirty because they still have
+ * some valid data in each.
+ */
+ result = -ENOSPC;
+ break;
+ }
+
+ /* Let's dare to make a garbage collect. */
+ if ((result = jffs_garbage_collect_next(c)) < 0) {
+ printk(KERN_ERR "JFFS: Something "
+ "has gone seriously wrong "
+ "with a garbage collect.\n");
+ goto gc_end;
+ }
+
+ D1(printk(" jffs_garbage_collect_now(): erased: %ld\n", erased));
+ DJM(jffs_print_memory_allocation_statistics());
+ }
+
+gc_end:
+ // up(&fmc->gclock);
+
+ D3(printk(" jffs_garbage_collect_now(): Leaving...\n"));
+ D1(if (erased) {
+ printk("jffs_g_c_now(): erased = %ld\n", erased);
+ jffs_print_fmcontrol(fmc);
+ });
+
+ if (!erased && !result)
+ return -ENOSPC;
+
+ return result;
+} /* jffs_garbage_collect_now() */
+
+
+/* Determine if it is reasonable to start garbage collection.
+ We start a gc pass if either:
+ - The number of free bytes < MIN_FREE_BYTES && at least one
+ block is dirty, OR
+ - The number of dirty bytes > MAX_DIRTY_BYTES
+*/
+static inline int thread_should_wake (struct jffs_control *c)
+{
+ D1(printk (KERN_NOTICE "thread_should_wake(): free=%d, dirty=%d, blocksize=%d.\n",
+ c->fmc->free_size, c->fmc->dirty_size, c->fmc->sector_size));
+
+ /* If there's not enough dirty space to free a block, there's no point. */
+ if (c->fmc->dirty_size < c->fmc->sector_size) {
+ D2(printk(KERN_NOTICE "thread_should_wake(): Not waking. Insufficient dirty space\n"));
+ return 0;
+ }
+#if 1
+ /* If there is too much RAM used by the various structures, GC */
+ if (jffs_get_node_inuse() > (c->fmc->used_size/c->fmc->max_chunk_size * 5 + jffs_get_file_count() * 2 + 50)) {
+ /* FIXME: Provide proof that this test can be satisfied. We
+ don't want a filesystem doing endless GC just because this
+ condition cannot ever be false.
+ */
+ D2(printk(KERN_NOTICE "thread_should_wake(): Waking due to number of nodes\n"));
+ return 1;
+ }
+#endif
+ /* If there are fewer free bytes than the threshold, GC */
+ if (c->fmc->free_size < c->gc_minfree_threshold) {
+ D2(printk(KERN_NOTICE "thread_should_wake(): Waking due to insufficent free space\n"));
+ return 1;
+ }
+ /* If there are more dirty bytes than the threshold, GC */
+ if (c->fmc->dirty_size > c->gc_maxdirty_threshold) {
+ D2(printk(KERN_NOTICE "thread_should_wake(): Waking due to excessive dirty space\n"));
+ return 1;
+ }
+ /* FIXME: What about the "There are many versions of a node" condition? */
+
+ return 0;
+}
+
+
+void jffs_garbage_collect_trigger(struct jffs_control *c)
+{
+ /* NOTE: We rely on the fact that we have the BKL here.
+ * Otherwise, the gc_task could go away between the check
+ * and the wake_up_process()
+ */
+ if (c->gc_task && thread_should_wake(c))
+ send_sig(SIGHUP, c->gc_task, 1);
+}
+
+
+/* Kernel threads take (void *) as arguments. Thus we pass
+ the jffs_control data as a (void *) and then cast it. */
+int
+jffs_garbage_collect_thread(void *ptr)
+{
+ struct jffs_control *c = (struct jffs_control *) ptr;
+ struct jffs_fmcontrol *fmc = c->fmc;
+ long erased;
+ int result = 0;
+ D1(int i = 1);
+
+ daemonize("jffs_gcd");
+
+ c->gc_task = current;
+
+ lock_kernel();
+ init_completion(&c->gc_thread_comp); /* barrier */
+ spin_lock_irq(&current->sighand->siglock);
+ siginitsetinv (&current->blocked, sigmask(SIGHUP) | sigmask(SIGKILL) | sigmask(SIGSTOP) | sigmask(SIGCONT));
+ recalc_sigpending();
+ spin_unlock_irq(&current->sighand->siglock);
+
+ D1(printk (KERN_NOTICE "jffs_garbage_collect_thread(): Starting infinite loop.\n"));
+
+ for (;;) {
+
+ /* See if we need to start gc. If we don't, go to sleep.
+
+ Current implementation is a BAD THING(tm). If we try
+ to unmount the FS, the unmount operation will sleep waiting
+ for this thread to exit. We need to arrange to send it a
+ sig before the umount process sleeps.
+ */
+
+ if (!thread_should_wake(c))
+ set_current_state (TASK_INTERRUPTIBLE);
+
+ schedule(); /* Yes, we do this even if we want to go
+ on immediately - we're a low priority
+ background task. */
+
+ /* Put_super will send a SIGKILL and then wait on the sem.
+ */
+ while (signal_pending(current)) {
+ siginfo_t info;
+ unsigned long signr = 0;
+
+ spin_lock_irq(&current->sighand->siglock);
+ signr = dequeue_signal(current, &current->blocked, &info);
+ spin_unlock_irq(&current->sighand->siglock);
+
+ switch(signr) {
+ case SIGSTOP:
+ D1(printk("jffs_garbage_collect_thread(): SIGSTOP received.\n"));
+ set_current_state(TASK_STOPPED);
+ schedule();
+ break;
+
+ case SIGKILL:
+ D1(printk("jffs_garbage_collect_thread(): SIGKILL received.\n"));
+ c->gc_task = NULL;
+ complete_and_exit(&c->gc_thread_comp, 0);
+ }
+ }
+
+
+ D1(printk (KERN_NOTICE "jffs_garbage_collect_thread(): collecting.\n"));
+
+ D3(printk (KERN_NOTICE "g_c_thread(): down biglock\n"));
+ down(&fmc->biglock);
+
+ D1(printk("***jffs_garbage_collect_thread(): round #%u, "
+ "fmc->dirty_size = %u\n", i++, fmc->dirty_size));
+ D2(jffs_print_fmcontrol(fmc));
+
+ if ((erased = jffs_try_to_erase(c)) < 0) {
+ printk(KERN_WARNING "JFFS: Error in "
+ "garbage collector: %ld.\n", erased);
+ }
+
+ if (erased)
+ goto gc_end;
+
+ if (fmc->free_size == 0) {
+ /* Argh. Might as well commit suicide. */
+ printk(KERN_ERR "jffs_garbage_collect_thread(): free_size == 0. This is BAD.\n");
+ send_sig(SIGQUIT, c->gc_task, 1);
+ // panic()
+ goto gc_end;
+ }
+
+ /* Let's dare to make a garbage collect. */
+ if ((result = jffs_garbage_collect_next(c)) < 0) {
+ printk(KERN_ERR "JFFS: Something "
+ "has gone seriously wrong "
+ "with a garbage collect: %d\n", result);
+ }
+
+ gc_end:
+ D3(printk (KERN_NOTICE "g_c_thread(): up biglock\n"));
+ up(&fmc->biglock);
+ } /* for (;;) */
+} /* jffs_garbage_collect_thread() */
OpenPOWER on IntegriCloud