of this option is that corruption of the contents
of a file can go unnoticed.
chk_data_crc (*) do not skip checking CRCs on data nodes
+compr=none override default compressor and set it to "none"
+compr=lzo override default compressor and set it to "lzo"
+compr=zlib override default compressor and set it to "zlib"
Quick usage instructions
#include "ubifs.h"
#include <linux/writeback.h>
-#include <asm/div64.h>
+#include <linux/math64.h>
/*
* When pessimistic budget calculations say that there is no enough space,
* UBIFS starts writing back dirty inodes and pages, doing garbage collection,
- * or committing. The below constants define maximum number of times UBIFS
+ * or committing. The below constant defines maximum number of times UBIFS
* repeats the operations.
*/
-#define MAX_SHRINK_RETRIES 8
-#define MAX_GC_RETRIES 4
-#define MAX_CMT_RETRIES 2
-#define MAX_NOSPC_RETRIES 1
+#define MAX_MKSPC_RETRIES 3
/*
* The below constant defines amount of dirty pages which should be written
*/
#define NR_TO_WRITE 16
-/**
- * struct retries_info - information about re-tries while making free space.
- * @prev_liability: previous liability
- * @shrink_cnt: how many times the liability was shrinked
- * @shrink_retries: count of liability shrink re-tries (increased when
- * liability does not shrink)
- * @try_gc: GC should be tried first
- * @gc_retries: how many times GC was run
- * @cmt_retries: how many times commit has been done
- * @nospc_retries: how many times GC returned %-ENOSPC
- *
- * Since we consider budgeting to be the fast-path, and this structure has to
- * be allocated on stack and zeroed out, we make it smaller using bit-fields.
- */
-struct retries_info {
- long long prev_liability;
- unsigned int shrink_cnt;
- unsigned int shrink_retries:5;
- unsigned int try_gc:1;
- unsigned int gc_retries:4;
- unsigned int cmt_retries:3;
- unsigned int nospc_retries:1;
-};
-
/**
* shrink_liability - write-back some dirty pages/inodes.
* @c: UBIFS file-system description object
return 0;
}
+/**
+ * get_liability - calculate current liability.
+ * @c: UBIFS file-system description object
+ *
+ * This function calculates and returns current UBIFS liability, i.e. the
+ * amount of bytes UBIFS has "promised" to write to the media.
+ */
+static long long get_liability(struct ubifs_info *c)
+{
+ long long liab;
+
+ spin_lock(&c->space_lock);
+ liab = c->budg_idx_growth + c->budg_data_growth + c->budg_dd_growth;
+ spin_unlock(&c->space_lock);
+ return liab;
+}
+
/**
* make_free_space - make more free space on the file-system.
* @c: UBIFS file-system description object
- * @ri: information about previous invocations of this function
*
* This function is called when an operation cannot be budgeted because there
* is supposedly no free space. But in most cases there is some free space:
* Returns %-ENOSPC if it couldn't do more free space, and other negative error
* codes on failures.
*/
-static int make_free_space(struct ubifs_info *c, struct retries_info *ri)
+static int make_free_space(struct ubifs_info *c)
{
- int err;
-
- /*
- * If we have some dirty pages and inodes (liability), try to write
- * them back unless this was tried too many times without effect
- * already.
- */
- if (ri->shrink_retries < MAX_SHRINK_RETRIES && !ri->try_gc) {
- long long liability;
-
- spin_lock(&c->space_lock);
- liability = c->budg_idx_growth + c->budg_data_growth +
- c->budg_dd_growth;
- spin_unlock(&c->space_lock);
+ int err, retries = 0;
+ long long liab1, liab2;
- if (ri->prev_liability >= liability) {
- /* Liability does not shrink, next time try GC then */
- ri->shrink_retries += 1;
- if (ri->gc_retries < MAX_GC_RETRIES)
- ri->try_gc = 1;
- dbg_budg("liability did not shrink: retries %d of %d",
- ri->shrink_retries, MAX_SHRINK_RETRIES);
- }
+ do {
+ liab1 = get_liability(c);
+ /*
+ * We probably have some dirty pages or inodes (liability), try
+ * to write them back.
+ */
+ dbg_budg("liability %lld, run write-back", liab1);
+ shrink_liability(c, NR_TO_WRITE);
- dbg_budg("force write-back (count %d)", ri->shrink_cnt);
- shrink_liability(c, NR_TO_WRITE + ri->shrink_cnt);
+ liab2 = get_liability(c);
+ if (liab2 < liab1)
+ return -EAGAIN;
- ri->prev_liability = liability;
- ri->shrink_cnt += 1;
- return -EAGAIN;
- }
+ dbg_budg("new liability %lld (not shrinked)", liab2);
- /*
- * Try to run garbage collector unless it was already tried too many
- * times.
- */
- if (ri->gc_retries < MAX_GC_RETRIES) {
- ri->gc_retries += 1;
- dbg_budg("run GC, retries %d of %d",
- ri->gc_retries, MAX_GC_RETRIES);
-
- ri->try_gc = 0;
+ /* Liability did not shrink again, try GC */
+ dbg_budg("Run GC");
err = run_gc(c);
if (!err)
return -EAGAIN;
- if (err == -EAGAIN) {
- dbg_budg("GC asked to commit");
- err = ubifs_run_commit(c);
- if (err)
- return err;
- return -EAGAIN;
- }
-
- if (err != -ENOSPC)
- return err;
-
- /*
- * GC could not make any progress. If this is the first time,
- * then it makes sense to try to commit, because it might make
- * some dirty space.
- */
- dbg_budg("GC returned -ENOSPC, retries %d",
- ri->nospc_retries);
- if (ri->nospc_retries >= MAX_NOSPC_RETRIES)
+ if (err != -EAGAIN && err != -ENOSPC)
+ /* Some real error happened */
return err;
- ri->nospc_retries += 1;
- }
- /* Neither GC nor write-back helped, try to commit */
- if (ri->cmt_retries < MAX_CMT_RETRIES) {
- ri->cmt_retries += 1;
- dbg_budg("run commit, retries %d of %d",
- ri->cmt_retries, MAX_CMT_RETRIES);
+ dbg_budg("Run commit (retries %d)", retries);
err = ubifs_run_commit(c);
if (err)
return err;
- return -EAGAIN;
- }
+ } while (retries++ < MAX_MKSPC_RETRIES);
+
return -ENOSPC;
}
*/
int ubifs_calc_min_idx_lebs(struct ubifs_info *c)
{
- int ret;
- uint64_t idx_size;
+ int idx_lebs, eff_leb_size = c->leb_size - c->max_idx_node_sz;
+ long long idx_size;
idx_size = c->old_idx_sz + c->budg_idx_growth + c->budg_uncommitted_idx;
* pair, nor similarly the two variables for the new index size, so we
* have to do this costly 64-bit division on fast-path.
*/
- if (do_div(idx_size, c->leb_size - c->max_idx_node_sz))
- ret = idx_size + 1;
- else
- ret = idx_size;
+ idx_size += eff_leb_size - 1;
+ idx_lebs = div_u64(idx_size, eff_leb_size);
/*
* The index head is not available for the in-the-gaps method, so add an
* extra LEB to compensate.
*/
- ret += 1;
- /*
- * At present the index needs at least 2 LEBs: one for the index head
- * and one for in-the-gaps method (which currently does not cater for
- * the index head and so excludes it from consideration).
- */
- if (ret < 2)
- ret = 2;
- return ret;
+ idx_lebs += 1;
+ if (idx_lebs < MIN_INDEX_LEBS)
+ idx_lebs = MIN_INDEX_LEBS;
+ return idx_lebs;
}
/**
int ubifs_budget_space(struct ubifs_info *c, struct ubifs_budget_req *req)
{
int uninitialized_var(cmt_retries), uninitialized_var(wb_retries);
- int err, idx_growth, data_growth, dd_growth;
- struct retries_info ri;
+ int err, idx_growth, data_growth, dd_growth, retried = 0;
ubifs_assert(req->new_page <= 1);
ubifs_assert(req->dirtied_page <= 1);
if (!data_growth && !dd_growth)
return 0;
idx_growth = calc_idx_growth(c, req);
- memset(&ri, 0, sizeof(struct retries_info));
again:
spin_lock(&c->space_lock);
return err;
}
- err = make_free_space(c, &ri);
+ err = make_free_space(c);
+ cond_resched();
if (err == -EAGAIN) {
dbg_budg("try again");
- cond_resched();
goto again;
} else if (err == -ENOSPC) {
+ if (!retried) {
+ retried = 1;
+ dbg_budg("-ENOSPC, but anyway try once again");
+ goto again;
+ }
dbg_budg("FS is full, -ENOSPC");
c->nospace = 1;
if (can_use_rp(c) || c->rp_size == 0)
* user-space. User-space application tend to expect that if the file-system
* (e.g., via the 'statfs()' call) reports that it has N bytes available, they
* are able to write a file of size N. UBIFS attaches node headers to each data
- * node and it has to write indexind nodes as well. This introduces additional
- * overhead, and UBIFS it has to report sligtly less free space to meet the
- * above expectetion.
+ * node and it has to write indexing nodes as well. This introduces additional
+ * overhead, and UBIFS has to report slightly less free space to meet the above
+ * expectations.
*
* This function assumes free space is made up of uncompressed data nodes and
* full index nodes (one per data node, tripled because we always allow enough
* Note, the calculation is pessimistic, which means that most of the time
* UBIFS reports less space than it actually has.
*/
-long long ubifs_reported_space(const struct ubifs_info *c, uint64_t free)
+long long ubifs_reported_space(const struct ubifs_info *c, long long free)
{
int divisor, factor, f;
* of data nodes, f - fanout. Because effective UBIFS fanout is twice
* as less than maximum fanout, we assume that each data node
* introduces 3 * @c->max_idx_node_sz / (@c->fanout/2 - 1) bytes.
- * Note, the multiplier 3 is because UBIFS reseves thrice as more space
+ * Note, the multiplier 3 is because UBIFS reserves thrice as more space
* for the index.
*/
f = c->fanout > 3 ? c->fanout >> 1 : 2;
divisor = UBIFS_MAX_DATA_NODE_SZ;
divisor += (c->max_idx_node_sz * 3) / (f - 1);
free *= factor;
- do_div(free, divisor);
- return free;
+ return div_u64(free, divisor);
}
/**
* This function calculates amount of free space to report to user-space.
*
* Because UBIFS may introduce substantial overhead (the index, node headers,
- * alighment, wastage at the end of eraseblocks, etc), it cannot report real
+ * alignment, wastage at the end of eraseblocks, etc), it cannot report real
* amount of free flash space it has (well, because not all dirty space is
- * reclamable, UBIFS does not actually know the real amount). If UBIFS did so,
- * it would bread user expectetion about what free space is. Users seem to
+ * reclaimable, UBIFS does not actually know the real amount). If UBIFS did so,
+ * it would bread user expectations about what free space is. Users seem to
* accustomed to assume that if the file-system reports N bytes of free space,
* they would be able to fit a file of N bytes to the FS. This almost works for
* traditional file-systems, because they have way less overhead than UBIFS.
long long available, outstanding, free;
spin_lock(&c->space_lock);
- min_idx_lebs = ubifs_calc_min_idx_lebs(c);
+ min_idx_lebs = c->min_idx_lebs;
+ ubifs_assert(min_idx_lebs == ubifs_calc_min_idx_lebs(c));
outstanding = c->budg_data_growth + c->budg_dd_growth;
-
- /*
- * Force the amount available to the total size reported if the used
- * space is zero.
- */
- if (c->lst.total_used <= UBIFS_INO_NODE_SZ && !outstanding) {
- spin_unlock(&c->space_lock);
- return (long long)c->block_cnt << UBIFS_BLOCK_SHIFT;
- }
-
available = ubifs_calc_available(c, min_idx_lebs);
/*
{
struct ubifs_idx_node *idx;
int lnum, offs, len, err = 0;
+ struct ubifs_debug_info *d = c->dbg;
- c->old_zroot = *zroot;
-
- lnum = c->old_zroot.lnum;
- offs = c->old_zroot.offs;
- len = c->old_zroot.len;
+ d->old_zroot = *zroot;
+ lnum = d->old_zroot.lnum;
+ offs = d->old_zroot.offs;
+ len = d->old_zroot.len;
idx = kmalloc(c->max_idx_node_sz, GFP_NOFS);
if (!idx)
if (err)
goto out;
- c->old_zroot_level = le16_to_cpu(idx->level);
- c->old_zroot_sqnum = le64_to_cpu(idx->ch.sqnum);
+ d->old_zroot_level = le16_to_cpu(idx->level);
+ d->old_zroot_sqnum = le64_to_cpu(idx->ch.sqnum);
out:
kfree(idx);
return err;
{
int lnum, offs, len, err = 0, uninitialized_var(last_level), child_cnt;
int first = 1, iip;
+ struct ubifs_debug_info *d = c->dbg;
union ubifs_key lower_key, upper_key, l_key, u_key;
unsigned long long uninitialized_var(last_sqnum);
struct ubifs_idx_node *idx;
UBIFS_IDX_NODE_SZ;
/* Start at the old zroot */
- lnum = c->old_zroot.lnum;
- offs = c->old_zroot.offs;
- len = c->old_zroot.len;
+ lnum = d->old_zroot.lnum;
+ offs = d->old_zroot.offs;
+ len = d->old_zroot.len;
iip = 0;
/*
if (first) {
first = 0;
/* Check root level and sqnum */
- if (le16_to_cpu(idx->level) != c->old_zroot_level) {
+ if (le16_to_cpu(idx->level) != d->old_zroot_level) {
err = 2;
goto out_dump;
}
- if (le64_to_cpu(idx->ch.sqnum) != c->old_zroot_sqnum) {
+ if (le64_to_cpu(idx->ch.sqnum) != d->old_zroot_sqnum) {
err = 3;
goto out_dump;
}
/* Fake description object for the "none" compressor */
static struct ubifs_compressor none_compr = {
.compr_type = UBIFS_COMPR_NONE,
- .name = "no compression",
+ .name = "none",
.capi_name = "",
};
static struct ubifs_compressor lzo_compr = {
.compr_type = UBIFS_COMPR_LZO,
.comp_mutex = &lzo_mutex,
- .name = "LZO",
+ .name = "lzo",
.capi_name = "lzo",
};
#else
static struct ubifs_compressor lzo_compr = {
.compr_type = UBIFS_COMPR_LZO,
- .name = "LZO",
+ .name = "lzo",
};
#endif
if (compr->comp_mutex)
mutex_lock(compr->comp_mutex);
err = crypto_comp_compress(compr->cc, in_buf, in_len, out_buf,
- out_len);
+ (unsigned int *)out_len);
if (compr->comp_mutex)
mutex_unlock(compr->comp_mutex);
if (unlikely(err)) {
}
/*
- * Presently, we just require that compression results in less data,
- * rather than any defined minimum compression ratio or amount.
+ * If the data compressed only slightly, it is better to leave it
+ * uncompressed to improve read speed.
*/
- if (ALIGN(*out_len, 8) >= ALIGN(in_len, 8))
+ if (in_len - *out_len < UBIFS_MIN_COMPRESS_DIFF)
goto no_compr;
return;
if (compr->decomp_mutex)
mutex_lock(compr->decomp_mutex);
err = crypto_comp_decompress(compr->cc, in_buf, in_len, out_buf,
- out_len);
+ (unsigned int *)out_len);
if (compr->decomp_mutex)
mutex_unlock(compr->decomp_mutex);
if (err)
/**
* ubifs_compressors_exit - de-initialize UBIFS compressors.
*/
-void __exit ubifs_compressors_exit(void)
+void ubifs_compressors_exit(void)
{
compr_exit(&lzo_compr);
compr_exit(&zlib_compr);
#include "ubifs.h"
#include <linux/module.h>
#include <linux/moduleparam.h>
+#include <linux/debugfs.h>
+#include <linux/math64.h>
#ifdef CONFIG_UBIFS_FS_DEBUG
struct rb_node *rb;
struct ubifs_bud *bud;
struct ubifs_gced_idx_leb *idx_gc;
+ long long available, outstanding, free;
+ ubifs_assert(spin_is_locked(&c->space_lock));
spin_lock(&dbg_lock);
printk(KERN_DEBUG "(pid %d) Budgeting info: budg_data_growth %lld, "
"budg_dd_growth %lld, budg_idx_growth %lld\n", current->pid,
printk(KERN_DEBUG "\tGC'ed idx LEB %d unmap %d\n",
idx_gc->lnum, idx_gc->unmap);
printk(KERN_DEBUG "\tcommit state %d\n", c->cmt_state);
+
+ /* Print budgeting predictions */
+ available = ubifs_calc_available(c, c->min_idx_lebs);
+ outstanding = c->budg_data_growth + c->budg_dd_growth;
+ if (available > outstanding)
+ free = ubifs_reported_space(c, available - outstanding);
+ else
+ free = 0;
+ printk(KERN_DEBUG "Budgeting predictions:\n");
+ printk(KERN_DEBUG "\tavailable: %lld, outstanding %lld, free %lld\n",
+ available, outstanding, free);
spin_unlock(&dbg_lock);
}
struct ubifs_lprops lp;
struct ubifs_lp_stats lst;
- printk(KERN_DEBUG "(pid %d) Dumping LEB properties\n", current->pid);
+ printk(KERN_DEBUG "(pid %d) start dumping LEB properties\n",
+ current->pid);
ubifs_get_lp_stats(c, &lst);
dbg_dump_lstats(&lst);
dbg_dump_lprop(c, &lp);
}
+ printk(KERN_DEBUG "(pid %d) finish dumping LEB properties\n",
+ current->pid);
}
void dbg_dump_lpt_info(struct ubifs_info *c)
int i;
spin_lock(&dbg_lock);
+ printk(KERN_DEBUG "(pid %d) dumping LPT information\n", current->pid);
printk(KERN_DEBUG "\tlpt_sz: %lld\n", c->lpt_sz);
printk(KERN_DEBUG "\tpnode_sz: %d\n", c->pnode_sz);
printk(KERN_DEBUG "\tnnode_sz: %d\n", c->nnode_sz);
printk(KERN_DEBUG "\tLPT root is at %d:%d\n", c->lpt_lnum, c->lpt_offs);
printk(KERN_DEBUG "\tLPT head is at %d:%d\n",
c->nhead_lnum, c->nhead_offs);
- printk(KERN_DEBUG "\tLPT ltab is at %d:%d\n", c->ltab_lnum, c->ltab_offs);
+ printk(KERN_DEBUG "\tLPT ltab is at %d:%d\n",
+ c->ltab_lnum, c->ltab_offs);
if (c->big_lpt)
printk(KERN_DEBUG "\tLPT lsave is at %d:%d\n",
c->lsave_lnum, c->lsave_offs);
if (dbg_failure_mode)
return;
- printk(KERN_DEBUG "(pid %d) Dumping LEB %d\n", current->pid, lnum);
-
- sleb = ubifs_scan(c, lnum, 0, c->dbg_buf);
+ printk(KERN_DEBUG "(pid %d) start dumping LEB %d\n",
+ current->pid, lnum);
+ sleb = ubifs_scan(c, lnum, 0, c->dbg->buf);
if (IS_ERR(sleb)) {
ubifs_err("scan error %d", (int)PTR_ERR(sleb));
return;
dbg_dump_node(c, snod->node);
}
+ printk(KERN_DEBUG "(pid %d) finish dumping LEB %d\n",
+ current->pid, lnum);
ubifs_scan_destroy(sleb);
return;
}
{
int i;
- printk(KERN_DEBUG "(pid %d) Dumping heap cat %d (%d elements)\n",
+ printk(KERN_DEBUG "(pid %d) start dumping heap cat %d (%d elements)\n",
current->pid, cat, heap->cnt);
for (i = 0; i < heap->cnt; i++) {
struct ubifs_lprops *lprops = heap->arr[i];
"flags %d\n", i, lprops->lnum, lprops->hpos,
lprops->free, lprops->dirty, lprops->flags);
}
+ printk(KERN_DEBUG "(pid %d) finish dumping heap\n", current->pid);
}
void dbg_dump_pnode(struct ubifs_info *c, struct ubifs_pnode *pnode,
{
int i;
- printk(KERN_DEBUG "(pid %d) Dumping pnode:\n", current->pid);
+ printk(KERN_DEBUG "(pid %d) dumping pnode:\n", current->pid);
printk(KERN_DEBUG "\taddress %zx parent %zx cnext %zx\n",
(size_t)pnode, (size_t)parent, (size_t)pnode->cnext);
printk(KERN_DEBUG "\tflags %lu iip %d level %d num %d\n",
int level;
printk(KERN_DEBUG "\n");
- printk(KERN_DEBUG "(pid %d) Dumping the TNC tree\n", current->pid);
+ printk(KERN_DEBUG "(pid %d) start dumping TNC tree\n", current->pid);
znode = ubifs_tnc_levelorder_next(c->zroot.znode, NULL);
level = znode->level;
printk(KERN_DEBUG "== Level %d ==\n", level);
dbg_dump_znode(c, znode);
znode = ubifs_tnc_levelorder_next(c->zroot.znode, znode);
}
-
- printk(KERN_DEBUG "\n");
+ printk(KERN_DEBUG "(pid %d) finish dumping TNC tree\n", current->pid);
}
static int dump_znode(struct ubifs_info *c, struct ubifs_znode *znode,
zbr1->offs, DBGKEY(&key));
dbg_err("but it should have key %s according to tnc",
DBGKEY(&zbr1->key));
- dbg_dump_node(c, dent1);
- goto out_free;
+ dbg_dump_node(c, dent1);
+ goto out_free;
}
key_read(c, &dent2->key, &key);
zbr1->offs, DBGKEY(&key));
dbg_err("but it should have key %s according to tnc",
DBGKEY(&zbr2->key));
- dbg_dump_node(c, dent2);
- goto out_free;
+ dbg_dump_node(c, dent2);
+ goto out_free;
}
nlen1 = le16_to_cpu(dent1->nlen);
dbg_err("bad order of colliding key %s",
DBGKEY(&key));
- dbg_msg("first node at %d:%d\n", zbr1->lnum, zbr1->offs);
+ ubifs_msg("first node at %d:%d\n", zbr1->lnum, zbr1->offs);
dbg_dump_node(c, dent1);
- dbg_msg("second node at %d:%d\n", zbr2->lnum, zbr2->offs);
+ ubifs_msg("second node at %d:%d\n", zbr2->lnum, zbr2->offs);
dbg_dump_node(c, dent2);
out_free:
return (next >> 16) & 32767;
}
-void dbg_failure_mode_registration(struct ubifs_info *c)
+static void failure_mode_init(struct ubifs_info *c)
{
struct failure_mode_info *fmi;
fmi = kmalloc(sizeof(struct failure_mode_info), GFP_NOFS);
if (!fmi) {
- dbg_err("Failed to register failure mode - no memory");
+ ubifs_err("Failed to register failure mode - no memory");
return;
}
fmi->c = c;
spin_unlock(&fmi_lock);
}
-void dbg_failure_mode_deregistration(struct ubifs_info *c)
+static void failure_mode_exit(struct ubifs_info *c)
{
struct failure_mode_info *fmi, *tmp;
struct ubifs_info *c = dbg_find_info(desc);
if (c && dbg_failure_mode)
- return c->failure_mode;
+ return c->dbg->failure_mode;
return 0;
}
static int do_fail(struct ubi_volume_desc *desc, int lnum, int write)
{
struct ubifs_info *c = dbg_find_info(desc);
+ struct ubifs_debug_info *d;
if (!c || !dbg_failure_mode)
return 0;
- if (c->failure_mode)
+ d = c->dbg;
+ if (d->failure_mode)
return 1;
- if (!c->fail_cnt) {
+ if (!d->fail_cnt) {
/* First call - decide delay to failure */
if (chance(1, 2)) {
unsigned int delay = 1 << (simple_rand() >> 11);
if (chance(1, 2)) {
- c->fail_delay = 1;
- c->fail_timeout = jiffies +
+ d->fail_delay = 1;
+ d->fail_timeout = jiffies +
msecs_to_jiffies(delay);
dbg_rcvry("failing after %ums", delay);
} else {
- c->fail_delay = 2;
- c->fail_cnt_max = delay;
+ d->fail_delay = 2;
+ d->fail_cnt_max = delay;
dbg_rcvry("failing after %u calls", delay);
}
}
- c->fail_cnt += 1;
+ d->fail_cnt += 1;
}
/* Determine if failure delay has expired */
- if (c->fail_delay == 1) {
- if (time_before(jiffies, c->fail_timeout))
+ if (d->fail_delay == 1) {
+ if (time_before(jiffies, d->fail_timeout))
return 0;
- } else if (c->fail_delay == 2)
- if (c->fail_cnt++ < c->fail_cnt_max)
+ } else if (d->fail_delay == 2)
+ if (d->fail_cnt++ < d->fail_cnt_max)
return 0;
if (lnum == UBIFS_SB_LNUM) {
if (write) {
dbg_rcvry("failing in bud LEB %d commit not running", lnum);
}
ubifs_err("*** SETTING FAILURE MODE ON (LEB %d) ***", lnum);
- c->failure_mode = 1;
+ d->failure_mode = 1;
dump_stack();
return 1;
}
return 0;
}
+/**
+ * ubifs_debugging_init - initialize UBIFS debugging.
+ * @c: UBIFS file-system description object
+ *
+ * This function initializes debugging-related data for the file system.
+ * Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubifs_debugging_init(struct ubifs_info *c)
+{
+ c->dbg = kzalloc(sizeof(struct ubifs_debug_info), GFP_KERNEL);
+ if (!c->dbg)
+ return -ENOMEM;
+
+ c->dbg->buf = vmalloc(c->leb_size);
+ if (!c->dbg->buf)
+ goto out;
+
+ failure_mode_init(c);
+ return 0;
+
+out:
+ kfree(c->dbg);
+ return -ENOMEM;
+}
+
+/**
+ * ubifs_debugging_exit - free debugging data.
+ * @c: UBIFS file-system description object
+ */
+void ubifs_debugging_exit(struct ubifs_info *c)
+{
+ failure_mode_exit(c);
+ vfree(c->dbg->buf);
+ kfree(c->dbg);
+}
+
+/*
+ * Root directory for UBIFS stuff in debugfs. Contains sub-directories which
+ * contain the stuff specific to particular file-system mounts.
+ */
+static struct dentry *debugfs_rootdir;
+
+/**
+ * dbg_debugfs_init - initialize debugfs file-system.
+ *
+ * UBIFS uses debugfs file-system to expose various debugging knobs to
+ * user-space. This function creates "ubifs" directory in the debugfs
+ * file-system. Returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+int dbg_debugfs_init(void)
+{
+ debugfs_rootdir = debugfs_create_dir("ubifs", NULL);
+ if (IS_ERR(debugfs_rootdir)) {
+ int err = PTR_ERR(debugfs_rootdir);
+ ubifs_err("cannot create \"ubifs\" debugfs directory, "
+ "error %d\n", err);
+ return err;
+ }
+
+ return 0;
+}
+
+/**
+ * dbg_debugfs_exit - remove the "ubifs" directory from debugfs file-system.
+ */
+void dbg_debugfs_exit(void)
+{
+ debugfs_remove(debugfs_rootdir);
+}
+
+static int open_debugfs_file(struct inode *inode, struct file *file)
+{
+ file->private_data = inode->i_private;
+ return 0;
+}
+
+static ssize_t write_debugfs_file(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ struct ubifs_info *c = file->private_data;
+ struct ubifs_debug_info *d = c->dbg;
+
+ if (file->f_path.dentry == d->dump_lprops)
+ dbg_dump_lprops(c);
+ else if (file->f_path.dentry == d->dump_budg) {
+ spin_lock(&c->space_lock);
+ dbg_dump_budg(c);
+ spin_unlock(&c->space_lock);
+ } else if (file->f_path.dentry == d->dump_tnc) {
+ mutex_lock(&c->tnc_mutex);
+ dbg_dump_tnc(c);
+ mutex_unlock(&c->tnc_mutex);
+ } else
+ return -EINVAL;
+
+ *ppos += count;
+ return count;
+}
+
+static const struct file_operations debugfs_fops = {
+ .open = open_debugfs_file,
+ .write = write_debugfs_file,
+ .owner = THIS_MODULE,
+};
+
+/**
+ * dbg_debugfs_init_fs - initialize debugfs for UBIFS instance.
+ * @c: UBIFS file-system description object
+ *
+ * This function creates all debugfs files for this instance of UBIFS. Returns
+ * zero in case of success and a negative error code in case of failure.
+ *
+ * Note, the only reason we have not merged this function with the
+ * 'ubifs_debugging_init()' function is because it is better to initialize
+ * debugfs interfaces at the very end of the mount process, and remove them at
+ * the very beginning of the mount process.
+ */
+int dbg_debugfs_init_fs(struct ubifs_info *c)
+{
+ int err;
+ const char *fname;
+ struct dentry *dent;
+ struct ubifs_debug_info *d = c->dbg;
+
+ sprintf(d->debugfs_dir_name, "ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
+ d->debugfs_dir = debugfs_create_dir(d->debugfs_dir_name,
+ debugfs_rootdir);
+ if (IS_ERR(d->debugfs_dir)) {
+ err = PTR_ERR(d->debugfs_dir);
+ ubifs_err("cannot create \"%s\" debugfs directory, error %d\n",
+ d->debugfs_dir_name, err);
+ goto out;
+ }
+
+ fname = "dump_lprops";
+ dent = debugfs_create_file(fname, S_IWUGO, d->debugfs_dir, c,
+ &debugfs_fops);
+ if (IS_ERR(dent))
+ goto out_remove;
+ d->dump_lprops = dent;
+
+ fname = "dump_budg";
+ dent = debugfs_create_file(fname, S_IWUGO, d->debugfs_dir, c,
+ &debugfs_fops);
+ if (IS_ERR(dent))
+ goto out_remove;
+ d->dump_budg = dent;
+
+ fname = "dump_tnc";
+ dent = debugfs_create_file(fname, S_IWUGO, d->debugfs_dir, c,
+ &debugfs_fops);
+ if (IS_ERR(dent))
+ goto out_remove;
+ d->dump_tnc = dent;
+
+ return 0;
+
+out_remove:
+ err = PTR_ERR(dent);
+ ubifs_err("cannot create \"%s\" debugfs directory, error %d\n",
+ fname, err);
+ debugfs_remove_recursive(d->debugfs_dir);
+out:
+ return err;
+}
+
+/**
+ * dbg_debugfs_exit_fs - remove all debugfs files.
+ * @c: UBIFS file-system description object
+ */
+void dbg_debugfs_exit_fs(struct ubifs_info *c)
+{
+ debugfs_remove_recursive(c->dbg->debugfs_dir);
+}
+
#endif /* CONFIG_UBIFS_FS_DEBUG */
#ifdef CONFIG_UBIFS_FS_DEBUG
-#define UBIFS_DBG(op) op
+/**
+ * ubifs_debug_info - per-FS debugging information.
+ * @buf: a buffer of LEB size, used for various purposes
+ * @old_zroot: old index root - used by 'dbg_check_old_index()'
+ * @old_zroot_level: old index root level - used by 'dbg_check_old_index()'
+ * @old_zroot_sqnum: old index root sqnum - used by 'dbg_check_old_index()'
+ * @failure_mode: failure mode for recovery testing
+ * @fail_delay: 0=>don't delay, 1=>delay a time, 2=>delay a number of calls
+ * @fail_timeout: time in jiffies when delay of failure mode expires
+ * @fail_cnt: current number of calls to failure mode I/O functions
+ * @fail_cnt_max: number of calls by which to delay failure mode
+ * @chk_lpt_sz: used by LPT tree size checker
+ * @chk_lpt_sz2: used by LPT tree size checker
+ * @chk_lpt_wastage: used by LPT tree size checker
+ * @chk_lpt_lebs: used by LPT tree size checker
+ * @new_nhead_offs: used by LPT tree size checker
+ * @new_ihead_lnum: used by debugging to check ihead_lnum
+ * @new_ihead_offs: used by debugging to check ihead_offs
+ *
+ * debugfs_dir_name: name of debugfs directory containing this file-system's
+ * files
+ * debugfs_dir: direntry object of the file-system debugfs directory
+ * dump_lprops: "dump lprops" debugfs knob
+ * dump_budg: "dump budgeting information" debugfs knob
+ * dump_tnc: "dump TNC" debugfs knob
+ */
+struct ubifs_debug_info {
+ void *buf;
+ struct ubifs_zbranch old_zroot;
+ int old_zroot_level;
+ unsigned long long old_zroot_sqnum;
+ int failure_mode;
+ int fail_delay;
+ unsigned long fail_timeout;
+ unsigned int fail_cnt;
+ unsigned int fail_cnt_max;
+ long long chk_lpt_sz;
+ long long chk_lpt_sz2;
+ long long chk_lpt_wastage;
+ int chk_lpt_lebs;
+ int new_nhead_offs;
+ int new_ihead_lnum;
+ int new_ihead_offs;
+
+ char debugfs_dir_name[100];
+ struct dentry *debugfs_dir;
+ struct dentry *dump_lprops;
+ struct dentry *dump_budg;
+ struct dentry *dump_tnc;
+};
#define ubifs_assert(expr) do { \
if (unlikely(!(expr))) { \
extern unsigned int ubifs_chk_flags;
extern unsigned int ubifs_tst_flags;
-/* Dump functions */
+int ubifs_debugging_init(struct ubifs_info *c);
+void ubifs_debugging_exit(struct ubifs_info *c);
+/* Dump functions */
const char *dbg_ntype(int type);
const char *dbg_cstate(int cmt_state);
const char *dbg_get_key_dump(const struct ubifs_info *c,
const union ubifs_key *key);
void dbg_dump_inode(const struct ubifs_info *c, const struct inode *inode);
void dbg_dump_node(const struct ubifs_info *c, const void *node);
+void dbg_dump_lpt_node(const struct ubifs_info *c, void *node, int lnum,
+ int offs);
void dbg_dump_budget_req(const struct ubifs_budget_req *req);
void dbg_dump_lstats(const struct ubifs_lp_stats *lst);
void dbg_dump_budg(struct ubifs_info *c);
struct ubifs_nnode *parent, int iip);
void dbg_dump_tnc(struct ubifs_info *c);
void dbg_dump_index(struct ubifs_info *c);
+void dbg_dump_lpt_lebs(const struct ubifs_info *c);
/* Checking helper functions */
-
typedef int (*dbg_leaf_callback)(struct ubifs_info *c,
struct ubifs_zbranch *zbr, void *priv);
typedef int (*dbg_znode_callback)(struct ubifs_info *c,
#define dbg_failure_mode (ubifs_tst_flags & UBIFS_TST_RCVRY)
-void dbg_failure_mode_registration(struct ubifs_info *c);
-void dbg_failure_mode_deregistration(struct ubifs_info *c);
-
#ifndef UBIFS_DBG_PRESERVE_UBI
#define ubi_leb_read dbg_leb_read
return dbg_leb_change(desc, lnum, buf, len, UBI_UNKNOWN);
}
-#else /* !CONFIG_UBIFS_FS_DEBUG */
+/* Debugfs-related stuff */
+int dbg_debugfs_init(void);
+void dbg_debugfs_exit(void);
+int dbg_debugfs_init_fs(struct ubifs_info *c);
+void dbg_debugfs_exit_fs(struct ubifs_info *c);
-#define UBIFS_DBG(op)
+#else /* !CONFIG_UBIFS_FS_DEBUG */
/* Use "if (0)" to make compiler check arguments even if debugging is off */
#define ubifs_assert(expr) do { \
#define DBGKEY(key) ((char *)(key))
#define DBGKEY1(key) ((char *)(key))
-#define dbg_ntype(type) ""
-#define dbg_cstate(cmt_state) ""
-#define dbg_get_key_dump(c, key) ({})
-#define dbg_dump_inode(c, inode) ({})
-#define dbg_dump_node(c, node) ({})
-#define dbg_dump_budget_req(req) ({})
-#define dbg_dump_lstats(lst) ({})
-#define dbg_dump_budg(c) ({})
-#define dbg_dump_lprop(c, lp) ({})
-#define dbg_dump_lprops(c) ({})
-#define dbg_dump_lpt_info(c) ({})
-#define dbg_dump_leb(c, lnum) ({})
-#define dbg_dump_znode(c, znode) ({})
-#define dbg_dump_heap(c, heap, cat) ({})
-#define dbg_dump_pnode(c, pnode, parent, iip) ({})
-#define dbg_dump_tnc(c) ({})
-#define dbg_dump_index(c) ({})
+#define ubifs_debugging_init(c) 0
+#define ubifs_debugging_exit(c) ({})
+
+#define dbg_ntype(type) ""
+#define dbg_cstate(cmt_state) ""
+#define dbg_get_key_dump(c, key) ({})
+#define dbg_dump_inode(c, inode) ({})
+#define dbg_dump_node(c, node) ({})
+#define dbg_dump_lpt_node(c, node, lnum, offs) ({})
+#define dbg_dump_budget_req(req) ({})
+#define dbg_dump_lstats(lst) ({})
+#define dbg_dump_budg(c) ({})
+#define dbg_dump_lprop(c, lp) ({})
+#define dbg_dump_lprops(c) ({})
+#define dbg_dump_lpt_info(c) ({})
+#define dbg_dump_leb(c, lnum) ({})
+#define dbg_dump_znode(c, znode) ({})
+#define dbg_dump_heap(c, heap, cat) ({})
+#define dbg_dump_pnode(c, pnode, parent, iip) ({})
+#define dbg_dump_tnc(c) ({})
+#define dbg_dump_index(c) ({})
+#define dbg_dump_lpt_lebs(c) ({})
#define dbg_walk_index(c, leaf_cb, znode_cb, priv) 0
#define dbg_old_index_check_init(c, zroot) 0
#define dbg_force_in_the_gaps_enabled 0
#define dbg_force_in_the_gaps() 0
#define dbg_failure_mode 0
-#define dbg_failure_mode_registration(c) ({})
-#define dbg_failure_mode_deregistration(c) ({})
-#endif /* !CONFIG_UBIFS_FS_DEBUG */
+#define dbg_debugfs_init() 0
+#define dbg_debugfs_exit()
+#define dbg_debugfs_init_fs(c) 0
+#define dbg_debugfs_exit_fs(c) 0
+#endif /* !CONFIG_UBIFS_FS_DEBUG */
#endif /* !__UBIFS_DEBUG_H__ */
return err;
}
- ubifs_assert(le64_to_cpu(dn->ch.sqnum) > ubifs_inode(inode)->creat_sqnum);
-
+ ubifs_assert(le64_to_cpu(dn->ch.sqnum) >
+ ubifs_inode(inode)->creat_sqnum);
len = le32_to_cpu(dn->size);
if (len <= 0 || len > UBIFS_BLOCK_SIZE)
goto dump;
}
if (!PageUptodate(page)) {
- if (!(pos & PAGE_CACHE_MASK) && len == PAGE_CACHE_SIZE)
+ if (!(pos & ~PAGE_CACHE_MASK) && len == PAGE_CACHE_SIZE)
SetPageChecked(page);
else {
err = do_readpage(page);
if (!PageUptodate(page)) {
/* The page is not loaded from the flash */
- if (!(pos & PAGE_CACHE_MASK) && len == PAGE_CACHE_SIZE)
+ if (!(pos & ~PAGE_CACHE_MASK) && len == PAGE_CACHE_SIZE)
/*
* We change whole page so no need to load it. But we
* have to set the @PG_checked flag to make the further
case FS_IOC_GETFLAGS:
flags = ubifs2ioctl(ubifs_inode(inode)->flags);
+ dbg_gen("get flags: %#x, i_flags %#x", flags, inode->i_flags);
return put_user(flags, (int __user *) arg);
case FS_IOC_SETFLAGS: {
err = mnt_want_write(file->f_path.mnt);
if (err)
return err;
+ dbg_gen("set flags: %#x, i_flags %#x", flags, inode->i_flags);
err = setflags(inode, flags);
mnt_drop_write(file->f_path.mnt);
return err;
data->size = cpu_to_le32(len);
zero_data_node_unused(data);
- if (!(ui->flags && UBIFS_COMPR_FL))
+ if (!(ui->flags & UBIFS_COMPR_FL))
/* Compression is disabled for this inode */
compr_type = UBIFS_COMPR_NONE;
else
data_key_init(c, &key, inum, blk);
bit = old_size & (UBIFS_BLOCK_SIZE - 1);
- blk = (old_size >> UBIFS_BLOCK_SHIFT) - (bit ? 0: 1);
+ blk = (old_size >> UBIFS_BLOCK_SHIFT) - (bit ? 0 : 1);
data_key_init(c, &to_key, inum, blk);
err = ubifs_tnc_remove_range(c, &key, &to_key);
#ifndef __UBIFS_KEY_H__
#define __UBIFS_KEY_H__
+/**
+ * key_mask_hash - mask a valid hash value.
+ * @val: value to be masked
+ *
+ * We use hash values as offset in directories, so values %0 and %1 are
+ * reserved for "." and "..". %2 is reserved for "end of readdir" marker. This
+ * function makes sure the reserved values are not used.
+ */
+static inline uint32_t key_mask_hash(uint32_t hash)
+{
+ hash &= UBIFS_S_KEY_HASH_MASK;
+ if (unlikely(hash <= 2))
+ hash += 3;
+ return hash;
+}
+
/**
* key_r5_hash - R5 hash function (borrowed from reiserfs).
* @s: direntry name
str++;
}
- a &= UBIFS_S_KEY_HASH_MASK;
-
- /*
- * We use hash values as offset in directories, so values %0 and %1 are
- * reserved for "." and "..". %2 is reserved for "end of readdir"
- * marker.
- */
- if (unlikely(a >= 0 && a <= 2))
- a += 3;
- return a;
+ return key_mask_hash(a);
}
/**
len = min_t(uint32_t, len, 4);
memcpy(&a, str, len);
- a &= UBIFS_S_KEY_HASH_MASK;
- if (unlikely(a >= 0 && a <= 2))
- a += 3;
- return a;
+ return key_mask_hash(a);
}
/**
* @flags: new flags
* @idx_gc_cnt: change to the count of idx_gc list
*
- * This function changes LEB properties. This function does not change a LEB
- * property (@free, @dirty or @flag) if the value passed is %LPROPS_NC.
+ * This function changes LEB properties (@free, @dirty or @flag). However, the
+ * property which has the %LPROPS_NC value is not changed. Returns a pointer to
+ * the updated LEB properties on success and a negative error code on failure.
*
- * This function returns a pointer to the updated LEB properties on success
- * and a negative error code on failure. N.B. the LEB properties may have had to
- * be copied (due to COW) and consequently the pointer returned may not be the
- * same as the pointer passed.
+ * Note, the LEB properties may have had to be copied (due to COW) and
+ * consequently the pointer returned may not be the same as the pointer
+ * passed.
*/
const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
const struct ubifs_lprops *lp,
}
}
- sleb = ubifs_scan(c, lnum, 0, c->dbg_buf);
+ sleb = ubifs_scan(c, lnum, 0, c->dbg->buf);
if (IS_ERR(sleb)) {
/*
* After an unclean unmount, empty and freeable LEBs
* can be written into a single eraseblock. In that case, garbage collection
* consists of just writing the whole table, which therefore makes all other
* eraseblocks reusable. In the case of the big model, dirty eraseblocks are
- * selected for garbage collection, which consists are marking the nodes in
+ * selected for garbage collection, which consists of marking the clean nodes in
* that LEB as dirty, and then only the dirty nodes are written out. Also, in
* the case of the big model, a table of LEB numbers is saved so that the entire
* LPT does not to be scanned looking for empty eraseblocks when UBIFS is first
* mounted.
*/
-#include <linux/crc16.h>
#include "ubifs.h"
+#include <linux/crc16.h>
+#include <linux/math64.h>
/**
* do_calc_lpt_geom - calculate sizes for the LPT area.
int ubifs_calc_lpt_geom(struct ubifs_info *c)
{
int lebs_needed;
- uint64_t sz;
+ long long sz;
do_calc_lpt_geom(c);
/* Verify that lpt_lebs is big enough */
sz = c->lpt_sz * 2; /* Must have at least 2 times the size */
- sz += c->leb_size - 1;
- do_div(sz, c->leb_size);
- lebs_needed = sz;
+ lebs_needed = div_u64(sz + c->leb_size - 1, c->leb_size);
if (lebs_needed > c->lpt_lebs) {
ubifs_err("too few LPT LEBs");
return -EINVAL;
}
c->check_lpt_free = c->big_lpt;
-
return 0;
}
int *big_lpt)
{
int i, lebs_needed;
- uint64_t sz;
+ long long sz;
/* Start by assuming the minimum number of LPT LEBs */
c->lpt_lebs = UBIFS_MIN_LPT_LEBS;
/* Now check there are enough LPT LEBs */
for (i = 0; i < 64 ; i++) {
sz = c->lpt_sz * 4; /* Allow 4 times the size */
- sz += c->leb_size - 1;
- do_div(sz, c->leb_size);
- lebs_needed = sz;
+ lebs_needed = div_u64(sz + c->leb_size - 1, c->leb_size);
if (lebs_needed > c->lpt_lebs) {
/* Not enough LPT LEBs so try again with more */
c->lpt_lebs = lebs_needed;
* This function calculates and returns the nnode number based on the parent's
* nnode number and the index in parent.
*/
-static int calc_nnode_num_from_parent(struct ubifs_info *c,
+static int calc_nnode_num_from_parent(const struct ubifs_info *c,
struct ubifs_nnode *parent, int iip)
{
int num, shft;
* This function calculates and returns the pnode number based on the parent's
* nnode number and the index in parent.
*/
-static int calc_pnode_num_from_parent(struct ubifs_info *c,
+static int calc_pnode_num_from_parent(const struct ubifs_info *c,
struct ubifs_nnode *parent, int iip)
{
int i, n = c->lpt_hght - 1, pnum = parent->num, num = 0;
*
* This function returns %0 on success and a negative error code on failure.
*/
-static int unpack_pnode(struct ubifs_info *c, void *buf,
+static int unpack_pnode(const struct ubifs_info *c, void *buf,
struct ubifs_pnode *pnode)
{
uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
}
/**
- * unpack_nnode - unpack a nnode.
+ * ubifs_unpack_nnode - unpack a nnode.
* @c: UBIFS file-system description object
* @buf: buffer containing packed nnode to unpack
* @nnode: nnode structure to fill
*
* This function returns %0 on success and a negative error code on failure.
*/
-static int unpack_nnode(struct ubifs_info *c, void *buf,
- struct ubifs_nnode *nnode)
+int ubifs_unpack_nnode(const struct ubifs_info *c, void *buf,
+ struct ubifs_nnode *nnode)
{
uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
int i, pos = 0, err;
*
* This function returns %0 on success and a negative error code on failure.
*/
-static int unpack_ltab(struct ubifs_info *c, void *buf)
+static int unpack_ltab(const struct ubifs_info *c, void *buf)
{
uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
int i, pos = 0, err;
*
* This function returns %0 on success and a negative error code on failure.
*/
-static int unpack_lsave(struct ubifs_info *c, void *buf)
+static int unpack_lsave(const struct ubifs_info *c, void *buf)
{
uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
int i, pos = 0, err;
*
* This function returns %0 on success and a negative error code on failure.
*/
-static int validate_nnode(struct ubifs_info *c, struct ubifs_nnode *nnode,
+static int validate_nnode(const struct ubifs_info *c, struct ubifs_nnode *nnode,
struct ubifs_nnode *parent, int iip)
{
int i, lvl, max_offs;
*
* This function returns %0 on success and a negative error code on failure.
*/
-static int validate_pnode(struct ubifs_info *c, struct ubifs_pnode *pnode,
+static int validate_pnode(const struct ubifs_info *c, struct ubifs_pnode *pnode,
struct ubifs_nnode *parent, int iip)
{
int i;
* This function calculates the LEB numbers for the LEB properties it contains
* based on the pnode number.
*/
-static void set_pnode_lnum(struct ubifs_info *c, struct ubifs_pnode *pnode)
+static void set_pnode_lnum(const struct ubifs_info *c,
+ struct ubifs_pnode *pnode)
{
int i, lnum;
err = ubi_read(c->ubi, lnum, buf, offs, c->nnode_sz);
if (err)
goto out;
- err = unpack_nnode(c, buf, nnode);
+ err = ubifs_unpack_nnode(c, buf, nnode);
if (err)
goto out;
}
c->nnode_sz);
if (err)
return ERR_PTR(err);
- err = unpack_nnode(c, buf, nnode);
+ err = ubifs_unpack_nnode(c, buf, nnode);
if (err)
return ERR_PTR(err);
}
dbg_err("LPT out of space at LEB %d:%d needing %d, done_ltab %d, "
"done_lsave %d", lnum, offs, len, done_ltab, done_lsave);
dbg_dump_lpt_info(c);
+ dbg_dump_lpt_lebs(c);
+ dump_stack();
return err;
}
no_space:
ubifs_err("LPT out of space mismatch");
dbg_err("LPT out of space mismatch at LEB %d:%d needing %d, done_ltab "
- "%d, done_lsave %d", lnum, offs, len, done_ltab, done_lsave);
+ "%d, done_lsave %d", lnum, offs, len, done_ltab, done_lsave);
dbg_dump_lpt_info(c);
+ dbg_dump_lpt_lebs(c);
+ dump_stack();
return err;
}
* LPT trivial garbage collection is where a LPT LEB contains only dirty and
* free space and so may be reused as soon as the next commit is completed.
* This function is called after the commit is completed (master node has been
- * written) and unmaps LPT LEBs that were marked for trivial GC.
+ * written) and un-maps LPT LEBs that were marked for trivial GC.
*/
static int lpt_tgc_end(struct ubifs_info *c)
{
* @c: UBIFS file-system description object
* @node_type: LPT node type
*/
-static int get_lpt_node_len(struct ubifs_info *c, int node_type)
+static int get_lpt_node_len(const struct ubifs_info *c, int node_type)
{
switch (node_type) {
case UBIFS_LPT_NNODE:
* @buf: buffer
* @len: length of buffer
*/
-static int get_pad_len(struct ubifs_info *c, uint8_t *buf, int len)
+static int get_pad_len(const struct ubifs_info *c, uint8_t *buf, int len)
{
int offs, pad_len;
* @buf: buffer
* @node_num: node number is returned here
*/
-static int get_lpt_node_type(struct ubifs_info *c, uint8_t *buf, int *node_num)
+static int get_lpt_node_type(const struct ubifs_info *c, uint8_t *buf,
+ int *node_num)
{
uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
int pos = 0, node_type;
*
* This function returns %1 if the buffer contains a node or %0 if it does not.
*/
-static int is_a_node(struct ubifs_info *c, uint8_t *buf, int len)
+static int is_a_node(const struct ubifs_info *c, uint8_t *buf, int len)
{
uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
int pos = 0, node_type, node_len;
return 1;
}
-
/**
* lpt_gc_lnum - garbage collect a LPT LEB.
* @c: UBIFS file-system description object
#ifdef CONFIG_UBIFS_FS_DEBUG
/**
- * dbg_is_all_ff - determine if a buffer contains only 0xff bytes.
+ * dbg_is_all_ff - determine if a buffer contains only 0xFF bytes.
* @buf: buffer
* @len: buffer length
*/
struct ubifs_nnode *nnode;
int hght;
- /* Entire tree is in memory so first_nnode / next_nnode are ok */
+ /* Entire tree is in memory so first_nnode / next_nnode are OK */
nnode = first_nnode(c, &hght);
for (; nnode; nnode = next_nnode(c, nnode, &hght)) {
struct ubifs_nbranch *branch;
{
int err, len = c->leb_size, dirty = 0, node_type, node_num, node_len;
int ret;
- void *buf = c->dbg_buf;
+ void *buf = c->dbg->buf;
+
+ if (!(ubifs_chk_flags & UBIFS_CHK_LPROPS))
+ return 0;
dbg_lp("LEB %d", lnum);
err = ubi_read(c->ubi, lnum, buf, 0, c->leb_size);
long long free = 0;
int i;
+ if (!(ubifs_chk_flags & UBIFS_CHK_LPROPS))
+ return 0;
+
for (i = 0; i < c->lpt_lebs; i++) {
if (c->ltab[i].tgc || c->ltab[i].cmt)
continue;
dbg_err("LPT space error: free %lld lpt_sz %lld",
free, c->lpt_sz);
dbg_dump_lpt_info(c);
+ dbg_dump_lpt_lebs(c);
+ dump_stack();
return -EINVAL;
}
return 0;
*/
int dbg_chk_lpt_sz(struct ubifs_info *c, int action, int len)
{
+ struct ubifs_debug_info *d = c->dbg;
long long chk_lpt_sz, lpt_sz;
int err = 0;
+ if (!(ubifs_chk_flags & UBIFS_CHK_LPROPS))
+ return 0;
+
switch (action) {
case 0:
- c->chk_lpt_sz = 0;
- c->chk_lpt_sz2 = 0;
- c->chk_lpt_lebs = 0;
- c->chk_lpt_wastage = 0;
+ d->chk_lpt_sz = 0;
+ d->chk_lpt_sz2 = 0;
+ d->chk_lpt_lebs = 0;
+ d->chk_lpt_wastage = 0;
if (c->dirty_pn_cnt > c->pnode_cnt) {
dbg_err("dirty pnodes %d exceed max %d",
c->dirty_pn_cnt, c->pnode_cnt);
}
return err;
case 1:
- c->chk_lpt_sz += len;
+ d->chk_lpt_sz += len;
return 0;
case 2:
- c->chk_lpt_sz += len;
- c->chk_lpt_wastage += len;
- c->chk_lpt_lebs += 1;
+ d->chk_lpt_sz += len;
+ d->chk_lpt_wastage += len;
+ d->chk_lpt_lebs += 1;
return 0;
case 3:
chk_lpt_sz = c->leb_size;
- chk_lpt_sz *= c->chk_lpt_lebs;
+ chk_lpt_sz *= d->chk_lpt_lebs;
chk_lpt_sz += len - c->nhead_offs;
- if (c->chk_lpt_sz != chk_lpt_sz) {
+ if (d->chk_lpt_sz != chk_lpt_sz) {
dbg_err("LPT wrote %lld but space used was %lld",
- c->chk_lpt_sz, chk_lpt_sz);
+ d->chk_lpt_sz, chk_lpt_sz);
err = -EINVAL;
}
- if (c->chk_lpt_sz > c->lpt_sz) {
+ if (d->chk_lpt_sz > c->lpt_sz) {
dbg_err("LPT wrote %lld but lpt_sz is %lld",
- c->chk_lpt_sz, c->lpt_sz);
+ d->chk_lpt_sz, c->lpt_sz);
err = -EINVAL;
}
- if (c->chk_lpt_sz2 && c->chk_lpt_sz != c->chk_lpt_sz2) {
+ if (d->chk_lpt_sz2 && d->chk_lpt_sz != d->chk_lpt_sz2) {
dbg_err("LPT layout size %lld but wrote %lld",
- c->chk_lpt_sz, c->chk_lpt_sz2);
+ d->chk_lpt_sz, d->chk_lpt_sz2);
err = -EINVAL;
}
- if (c->chk_lpt_sz2 && c->new_nhead_offs != len) {
+ if (d->chk_lpt_sz2 && d->new_nhead_offs != len) {
dbg_err("LPT new nhead offs: expected %d was %d",
- c->new_nhead_offs, len);
+ d->new_nhead_offs, len);
err = -EINVAL;
}
lpt_sz = (long long)c->pnode_cnt * c->pnode_sz;
lpt_sz += c->ltab_sz;
if (c->big_lpt)
lpt_sz += c->lsave_sz;
- if (c->chk_lpt_sz - c->chk_lpt_wastage > lpt_sz) {
+ if (d->chk_lpt_sz - d->chk_lpt_wastage > lpt_sz) {
dbg_err("LPT chk_lpt_sz %lld + waste %lld exceeds %lld",
- c->chk_lpt_sz, c->chk_lpt_wastage, lpt_sz);
+ d->chk_lpt_sz, d->chk_lpt_wastage, lpt_sz);
err = -EINVAL;
}
- if (err)
+ if (err) {
dbg_dump_lpt_info(c);
- c->chk_lpt_sz2 = c->chk_lpt_sz;
- c->chk_lpt_sz = 0;
- c->chk_lpt_wastage = 0;
- c->chk_lpt_lebs = 0;
- c->new_nhead_offs = len;
+ dbg_dump_lpt_lebs(c);
+ dump_stack();
+ }
+ d->chk_lpt_sz2 = d->chk_lpt_sz;
+ d->chk_lpt_sz = 0;
+ d->chk_lpt_wastage = 0;
+ d->chk_lpt_lebs = 0;
+ d->new_nhead_offs = len;
return err;
case 4:
- c->chk_lpt_sz += len;
- c->chk_lpt_wastage += len;
+ d->chk_lpt_sz += len;
+ d->chk_lpt_wastage += len;
return 0;
default:
return -EINVAL;
}
}
+/**
+ * dbg_dump_lpt_leb - dump an LPT LEB.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number to dump
+ *
+ * This function dumps an LEB from LPT area. Nodes in this area are very
+ * different to nodes in the main area (e.g., they do not have common headers,
+ * they do not have 8-byte alignments, etc), so we have a separate function to
+ * dump LPT area LEBs. Note, LPT has to be locked by the caller.
+ */
+static void dump_lpt_leb(const struct ubifs_info *c, int lnum)
+{
+ int err, len = c->leb_size, node_type, node_num, node_len, offs;
+ void *buf = c->dbg->buf;
+
+ printk(KERN_DEBUG "(pid %d) start dumping LEB %d\n",
+ current->pid, lnum);
+ err = ubi_read(c->ubi, lnum, buf, 0, c->leb_size);
+ if (err) {
+ ubifs_err("cannot read LEB %d, error %d", lnum, err);
+ return;
+ }
+ while (1) {
+ offs = c->leb_size - len;
+ if (!is_a_node(c, buf, len)) {
+ int pad_len;
+
+ pad_len = get_pad_len(c, buf, len);
+ if (pad_len) {
+ printk(KERN_DEBUG "LEB %d:%d, pad %d bytes\n",
+ lnum, offs, pad_len);
+ buf += pad_len;
+ len -= pad_len;
+ continue;
+ }
+ if (len)
+ printk(KERN_DEBUG "LEB %d:%d, free %d bytes\n",
+ lnum, offs, len);
+ break;
+ }
+
+ node_type = get_lpt_node_type(c, buf, &node_num);
+ switch (node_type) {
+ case UBIFS_LPT_PNODE:
+ {
+ node_len = c->pnode_sz;
+ if (c->big_lpt)
+ printk(KERN_DEBUG "LEB %d:%d, pnode num %d\n",
+ lnum, offs, node_num);
+ else
+ printk(KERN_DEBUG "LEB %d:%d, pnode\n",
+ lnum, offs);
+ break;
+ }
+ case UBIFS_LPT_NNODE:
+ {
+ int i;
+ struct ubifs_nnode nnode;
+
+ node_len = c->nnode_sz;
+ if (c->big_lpt)
+ printk(KERN_DEBUG "LEB %d:%d, nnode num %d, ",
+ lnum, offs, node_num);
+ else
+ printk(KERN_DEBUG "LEB %d:%d, nnode, ",
+ lnum, offs);
+ err = ubifs_unpack_nnode(c, buf, &nnode);
+ for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+ printk("%d:%d", nnode.nbranch[i].lnum,
+ nnode.nbranch[i].offs);
+ if (i != UBIFS_LPT_FANOUT - 1)
+ printk(", ");
+ }
+ printk("\n");
+ break;
+ }
+ case UBIFS_LPT_LTAB:
+ node_len = c->ltab_sz;
+ printk(KERN_DEBUG "LEB %d:%d, ltab\n",
+ lnum, offs);
+ break;
+ case UBIFS_LPT_LSAVE:
+ node_len = c->lsave_sz;
+ printk(KERN_DEBUG "LEB %d:%d, lsave len\n", lnum, offs);
+ break;
+ default:
+ ubifs_err("LPT node type %d not recognized", node_type);
+ return;
+ }
+
+ buf += node_len;
+ len -= node_len;
+ }
+
+ printk(KERN_DEBUG "(pid %d) finish dumping LEB %d\n",
+ current->pid, lnum);
+}
+
+/**
+ * dbg_dump_lpt_lebs - dump LPT lebs.
+ * @c: UBIFS file-system description object
+ *
+ * This function dumps all LPT LEBs. The caller has to make sure the LPT is
+ * locked.
+ */
+void dbg_dump_lpt_lebs(const struct ubifs_info *c)
+{
+ int i;
+
+ printk(KERN_DEBUG "(pid %d) start dumping all LPT LEBs\n",
+ current->pid);
+ for (i = 0; i < c->lpt_lebs; i++)
+ dump_lpt_leb(c, i + c->lpt_first);
+ printk(KERN_DEBUG "(pid %d) finish dumping all LPT LEBs\n",
+ current->pid);
+}
+
#endif /* CONFIG_UBIFS_FS_DEBUG */
for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
struct ubifs_scan_leb *sleb;
- sleb = ubifs_scan(c, lnum, 0, c->dbg_buf);
+ sleb = ubifs_scan(c, lnum, 0, c->dbg->buf);
if (IS_ERR(sleb)) {
err = PTR_ERR(sleb);
break;
/*
* If the replay order was perfect the dirty space would now be
* zero. The order is not perfect because the the journal heads
- * race with eachother. This is not a problem but is does mean
+ * race with each other. This is not a problem but is does mean
* that the dirty space may temporarily exceed c->leb_size
* during the replay.
*/
* @dirty: amount of dirty space from padding and deletion nodes
*
* This function inserts a reference node to the replay tree and returns zero
- * in case of success ort a negative error code in case of failure.
+ * in case of success or a negative error code in case of failure.
*/
static int insert_ref_node(struct ubifs_info *c, int lnum, int offs,
unsigned long long sqnum, int free, int dirty)
* This means that we reached end of log and now
* look to the older log data, which was already
* committed but the eraseblock was not erased (UBIFS
- * only unmaps it). So this basically means we have to
+ * only un-maps it). So this basically means we have to
* exit with "end of log" code.
*/
err = 1;
if (err)
goto out;
+ /*
+ * UBIFS budgeting calculations use @c->budg_uncommitted_idx variable
+ * to roughly estimate index growth. Things like @c->min_idx_lebs
+ * depend on it. This means we have to initialize it to make sure
+ * budgeting works properly.
+ */
+ c->budg_uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
+ c->budg_uncommitted_idx *= c->max_idx_node_sz;
+
ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery);
dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, "
"highest_inum %lu", c->lhead_lnum, c->lhead_offs, c->max_sqnum,
#include "ubifs.h"
#include <linux/random.h>
+#include <linux/math64.h>
/*
* Default journal size in logical eraseblocks as a percent of total
int err, tmp, jnl_lebs, log_lebs, max_buds, main_lebs, main_first;
int lpt_lebs, lpt_first, orph_lebs, big_lpt, ino_waste, sup_flags = 0;
int min_leb_cnt = UBIFS_MIN_LEB_CNT;
- uint64_t tmp64, main_bytes;
+ long long tmp64, main_bytes;
__le64 tmp_le64;
/* Some functions called from here depend on the @c->key_len filed */
if (!sup)
return -ENOMEM;
- tmp64 = (uint64_t)max_buds * c->leb_size;
+ tmp64 = (long long)max_buds * c->leb_size;
if (big_lpt)
sup_flags |= UBIFS_FLG_BIGLPT;
sup->fanout = cpu_to_le32(DEFAULT_FANOUT);
sup->lsave_cnt = cpu_to_le32(c->lsave_cnt);
sup->fmt_version = cpu_to_le32(UBIFS_FORMAT_VERSION);
- sup->default_compr = cpu_to_le16(UBIFS_COMPR_LZO);
sup->time_gran = cpu_to_le32(DEFAULT_TIME_GRAN);
+ if (c->mount_opts.override_compr)
+ sup->default_compr = cpu_to_le16(c->mount_opts.compr_type);
+ else
+ sup->default_compr = cpu_to_le16(UBIFS_COMPR_LZO);
generate_random_uuid(sup->uuid);
- main_bytes = (uint64_t)main_lebs * c->leb_size;
- tmp64 = main_bytes * DEFAULT_RP_PERCENT;
- do_div(tmp64, 100);
+ main_bytes = (long long)main_lebs * c->leb_size;
+ tmp64 = div_u64(main_bytes * DEFAULT_RP_PERCENT, 100);
if (tmp64 > DEFAULT_MAX_RP_SIZE)
tmp64 = DEFAULT_MAX_RP_SIZE;
sup->rp_size = cpu_to_le64(tmp64);
c->jhead_cnt = le32_to_cpu(sup->jhead_cnt) + NONDATA_JHEADS_CNT;
c->fanout = le32_to_cpu(sup->fanout);
c->lsave_cnt = le32_to_cpu(sup->lsave_cnt);
- c->default_compr = le16_to_cpu(sup->default_compr);
c->rp_size = le64_to_cpu(sup->rp_size);
c->rp_uid = le32_to_cpu(sup->rp_uid);
c->rp_gid = le32_to_cpu(sup->rp_gid);
sup_flags = le32_to_cpu(sup->flags);
+ if (!c->mount_opts.override_compr)
+ c->default_compr = le16_to_cpu(sup->default_compr);
c->vfs_sb->s_time_gran = le32_to_cpu(sup->time_gran);
-
memcpy(&c->uuid, &sup->uuid, 16);
-
c->big_lpt = !!(sup_flags & UBIFS_FLG_BIGLPT);
/* Automatically increase file system size to the maximum size */
#include <linux/parser.h>
#include <linux/seq_file.h>
#include <linux/mount.h>
+#include <linux/math64.h>
+#include <linux/writeback.h>
#include "ubifs.h"
/*
else if (c->mount_opts.chk_data_crc == 1)
seq_printf(s, ",no_chk_data_crc");
+ if (c->mount_opts.override_compr) {
+ seq_printf(s, ",compr=");
+ seq_printf(s, ubifs_compr_name(c->mount_opts.compr_type));
+ }
+
return 0;
}
static int ubifs_sync_fs(struct super_block *sb, int wait)
{
+ int i, err;
struct ubifs_info *c = sb->s_fs_info;
- int i, ret = 0, err;
- long long bud_bytes;
-
- if (c->jheads) {
- for (i = 0; i < c->jhead_cnt; i++) {
- err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
- if (err && !ret)
- ret = err;
- }
+ struct writeback_control wbc = {
+ .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_HOLD,
+ .range_start = 0,
+ .range_end = LLONG_MAX,
+ .nr_to_write = LONG_MAX,
+ };
+
+ if (sb->s_flags & MS_RDONLY)
+ return 0;
- /* Commit the journal unless it has too little data */
- spin_lock(&c->buds_lock);
- bud_bytes = c->bud_bytes;
- spin_unlock(&c->buds_lock);
- if (bud_bytes > c->leb_size) {
- err = ubifs_run_commit(c);
- if (err)
- return err;
- }
+ /*
+ * Synchronize write buffers, because 'ubifs_run_commit()' does not
+ * do this if it waits for an already running commit.
+ */
+ for (i = 0; i < c->jhead_cnt; i++) {
+ err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
+ if (err)
+ return err;
}
/*
- * We ought to call sync for c->ubi but it does not have one. If it had
- * it would in turn call mtd->sync, however mtd operations are
- * synchronous anyway, so we don't lose any sleep here.
+ * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
+ * pages, so synchronize them first, then commit the journal. Strictly
+ * speaking, it is not necessary to commit the journal here,
+ * synchronizing write-buffers would be enough. But committing makes
+ * UBIFS free space predictions much more accurate, so we want to let
+ * the user be able to get more accurate results of 'statfs()' after
+ * they synchronize the file system.
*/
- return ret;
+ generic_sync_sb_inodes(sb, &wbc);
+
+ err = ubifs_run_commit(c);
+ if (err)
+ return err;
+
+ return ubi_sync(c->vi.ubi_num);
}
/**
}
/*
- * init_constants_late - initialize UBIFS constants.
+ * init_constants_sb - initialize UBIFS constants.
* @c: UBIFS file-system description object
*
* This is a helper function which initializes various UBIFS constants after
* makes sure they are all right. Returns zero in case of success and a
* negative error code in case of failure.
*/
-static int init_constants_late(struct ubifs_info *c)
+static int init_constants_sb(struct ubifs_info *c)
{
int tmp, err;
- uint64_t tmp64;
+ long long tmp64;
c->main_bytes = (long long)c->main_lebs * c->leb_size;
c->max_znode_sz = sizeof(struct ubifs_znode) +
* Make sure that the log is large enough to fit reference nodes for
* all buds plus one reserved LEB.
*/
- tmp64 = c->max_bud_bytes;
- tmp = do_div(tmp64, c->leb_size);
- c->max_bud_cnt = tmp64 + !!tmp;
+ tmp64 = c->max_bud_bytes + c->leb_size - 1;
+ c->max_bud_cnt = div_u64(tmp64, c->leb_size);
tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
tmp /= c->leb_size;
tmp += 1;
* Consequently, if the journal is too small, UBIFS will treat it as
* always full.
*/
- tmp64 = (uint64_t)(c->jhead_cnt + 1) * c->leb_size + 1;
+ tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
if (c->bg_bud_bytes < tmp64)
c->bg_bud_bytes = tmp64;
if (c->max_bud_bytes < tmp64 + c->leb_size)
if (err)
return err;
+ return 0;
+}
+
+/*
+ * init_constants_master - initialize UBIFS constants.
+ * @c: UBIFS file-system description object
+ *
+ * This is a helper function which initializes various UBIFS constants after
+ * the master node has been read. It also checks various UBIFS parameters and
+ * makes sure they are all right.
+ */
+static void init_constants_master(struct ubifs_info *c)
+{
+ long long tmp64;
+
c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
/*
* necessary to report something for the 'statfs()' call.
*
* Subtract the LEB reserved for GC, the LEB which is reserved for
- * deletions, and assume only one journal head is available.
+ * deletions, minimum LEBs for the index, and assume only one journal
+ * head is available.
*/
- tmp64 = c->main_lebs - 2 - c->jhead_cnt + 1;
- tmp64 *= (uint64_t)c->leb_size - c->leb_overhead;
+ tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
+ tmp64 *= (long long)c->leb_size - c->leb_overhead;
tmp64 = ubifs_reported_space(c, tmp64);
c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
-
- return 0;
}
/**
* Opt_no_bulk_read: disable bulk-reads
* Opt_chk_data_crc: check CRCs when reading data nodes
* Opt_no_chk_data_crc: do not check CRCs when reading data nodes
+ * Opt_override_compr: override default compressor
* Opt_err: just end of array marker
*/
enum {
Opt_no_bulk_read,
Opt_chk_data_crc,
Opt_no_chk_data_crc,
+ Opt_override_compr,
Opt_err,
};
{Opt_no_bulk_read, "no_bulk_read"},
{Opt_chk_data_crc, "chk_data_crc"},
{Opt_no_chk_data_crc, "no_chk_data_crc"},
+ {Opt_override_compr, "compr=%s"},
{Opt_err, NULL},
};
c->mount_opts.chk_data_crc = 1;
c->no_chk_data_crc = 1;
break;
+ case Opt_override_compr:
+ {
+ char *name = match_strdup(&args[0]);
+
+ if (!name)
+ return -ENOMEM;
+ if (!strcmp(name, "none"))
+ c->mount_opts.compr_type = UBIFS_COMPR_NONE;
+ else if (!strcmp(name, "lzo"))
+ c->mount_opts.compr_type = UBIFS_COMPR_LZO;
+ else if (!strcmp(name, "zlib"))
+ c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
+ else {
+ ubifs_err("unknown compressor \"%s\"", name);
+ kfree(name);
+ return -EINVAL;
+ }
+ kfree(name);
+ c->mount_opts.override_compr = 1;
+ c->default_compr = c->mount_opts.compr_type;
+ break;
+ }
default:
ubifs_err("unrecognized mount option \"%s\" "
"or missing value", p);
}
}
+/**
+ * check_free_space - check if there is enough free space to mount.
+ * @c: UBIFS file-system description object
+ *
+ * This function makes sure UBIFS has enough free space to be mounted in
+ * read/write mode. UBIFS must always have some free space to allow deletions.
+ */
+static int check_free_space(struct ubifs_info *c)
+{
+ ubifs_assert(c->dark_wm > 0);
+ if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
+ ubifs_err("insufficient free space to mount in read/write mode");
+ dbg_dump_budg(c);
+ dbg_dump_lprops(c);
+ /*
+ * We return %-EINVAL instead of %-ENOSPC because it seems to
+ * be the closest error code mentioned in the mount function
+ * documentation.
+ */
+ return -EINVAL;
+ }
+ return 0;
+}
+
/**
* mount_ubifs - mount UBIFS file-system.
* @c: UBIFS file-system description object
if (err)
return err;
-#ifdef CONFIG_UBIFS_FS_DEBUG
- c->dbg_buf = vmalloc(c->leb_size);
- if (!c->dbg_buf)
- return -ENOMEM;
-#endif
+ err = ubifs_debugging_init(c);
+ if (err)
+ return err;
err = check_volume_empty(c);
if (err)
goto out_free;
/*
- * Make sure the compressor which is set as the default on in the
- * superblock was actually compiled in.
+ * Make sure the compressor which is set as default in the superblock
+ * or overridden by mount options is actually compiled in.
*/
if (!ubifs_compr_present(c->default_compr)) {
- ubifs_warn("'%s' compressor is set by superblock, but not "
- "compiled in", ubifs_compr_name(c->default_compr));
- c->default_compr = UBIFS_COMPR_NONE;
+ ubifs_err("'compressor \"%s\" is not compiled in",
+ ubifs_compr_name(c->default_compr));
+ goto out_free;
}
- dbg_failure_mode_registration(c);
-
- err = init_constants_late(c);
+ err = init_constants_sb(c);
if (err)
- goto out_dereg;
+ goto out_free;
sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
c->cbuf = kmalloc(sz, GFP_NOFS);
if (!c->cbuf) {
err = -ENOMEM;
- goto out_dereg;
+ goto out_free;
}
sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
if (err)
goto out_master;
+ init_constants_master(c);
+
if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
ubifs_msg("recovery needed");
c->need_recovery = 1;
if (!mounted_read_only) {
int lnum;
- /* Check for enough free space */
- if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
- ubifs_err("insufficient available space");
- err = -EINVAL;
+ err = check_free_space(c);
+ if (err)
goto out_orphans;
- }
/* Check for enough log space */
lnum = c->lhead_lnum + 1;
}
}
+ err = dbg_debugfs_init_fs(c);
+ if (err)
+ goto out_infos;
+
err = dbg_check_filesystem(c);
if (err)
goto out_infos;
dbg_msg("tree fanout: %d", c->fanout);
dbg_msg("reserved GC LEB: %d", c->gc_lnum);
dbg_msg("first main LEB: %d", c->main_first);
+ dbg_msg("max. znode size %d", c->max_znode_sz);
+ dbg_msg("max. index node size %d", c->max_idx_node_sz);
+ dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
+ UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
+ dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
+ UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
+ dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
+ UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
+ dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
+ UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
+ UBIFS_MAX_DENT_NODE_SZ);
dbg_msg("dead watermark: %d", c->dead_wm);
dbg_msg("dark watermark: %d", c->dark_wm);
+ dbg_msg("LEB overhead: %d", c->leb_overhead);
x = (long long)c->main_lebs * c->dark_wm;
dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
x, x >> 10, x >> 20);
free_wbufs(c);
out_cbuf:
kfree(c->cbuf);
-out_dereg:
- dbg_failure_mode_deregistration(c);
out_free:
kfree(c->bu.buf);
vfree(c->ileb_buf);
vfree(c->sbuf);
kfree(c->bottom_up_buf);
- UBIFS_DBG(vfree(c->dbg_buf));
+ ubifs_debugging_exit(c);
return err;
}
dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
c->vi.vol_id);
+ dbg_debugfs_exit_fs(c);
spin_lock(&ubifs_infos_lock);
list_del(&c->infos_list);
spin_unlock(&ubifs_infos_lock);
vfree(c->ileb_buf);
vfree(c->sbuf);
kfree(c->bottom_up_buf);
- UBIFS_DBG(vfree(c->dbg_buf));
- dbg_failure_mode_deregistration(c);
+ ubifs_debugging_exit(c);
}
/**
c->remounting_rw = 1;
c->always_chk_crc = 1;
- /* Check for enough free space */
- if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
- ubifs_err("insufficient available space");
- err = -EINVAL;
+ err = check_free_space(c);
+ if (err)
goto out;
- }
if (c->old_leb_cnt != c->leb_cnt) {
struct ubifs_sb_node *sup;
* @c: UBIFS file-system description object
*
* This function is called during un-mounting and re-mounting, and it commits
- * the journal unless the "fast unmount" mode is enabled. It also avoids
- * committing the journal if it contains too few data.
+ * the journal unless the "fast unmount" mode is enabled.
*/
static void commit_on_unmount(struct ubifs_info *c)
{
- if (!c->fast_unmount) {
- long long bud_bytes;
+ struct super_block *sb = c->vfs_sb;
+ long long bud_bytes;
- spin_lock(&c->buds_lock);
- bud_bytes = c->bud_bytes;
- spin_unlock(&c->buds_lock);
- if (bud_bytes > c->leb_size)
- ubifs_run_commit(c);
- }
+ /*
+ * This function is called before the background thread is stopped, so
+ * we may race with ongoing commit, which means we have to take
+ * @c->bud_lock to access @c->bud_bytes.
+ */
+ spin_lock(&c->buds_lock);
+ bud_bytes = c->bud_bytes;
+ spin_unlock(&c->buds_lock);
+
+ if (!c->fast_unmount && !(sb->s_flags & MS_RDONLY) && bud_bytes)
+ ubifs_run_commit(c);
}
/**
goto out_iput;
mutex_unlock(&c->umount_mutex);
-
return 0;
out_iput:
* We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
* in order to be outside BKL.
*/
- if (sb->s_root && !(sb->s_flags & MS_RDONLY))
+ if (sb->s_root)
commit_on_unmount(c);
/* The un-mount routine is actually done in put_super() */
generic_shutdown_super(sb);
BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
+ /*
+ * We use 2 bit wide bit-fields to store compression type, which should
+ * be amended if more compressors are added. The bit-fields are:
+ * @compr_type in 'struct ubifs_inode', @default_compr in
+ * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
+ */
+ BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
+
/*
* We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
* UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
register_shrinker(&ubifs_shrinker_info);
err = ubifs_compressors_init();
+ if (err)
+ goto out_shrinker;
+
+ err = dbg_debugfs_init();
if (err)
goto out_compr;
return 0;
out_compr:
+ ubifs_compressors_exit();
+out_shrinker:
unregister_shrinker(&ubifs_shrinker_info);
kmem_cache_destroy(ubifs_inode_slab);
out_reg:
ubifs_assert(list_empty(&ubifs_infos));
ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
+ dbg_debugfs_exit();
ubifs_compressors_exit();
unregister_shrinker(&ubifs_shrinker_info);
kmem_cache_destroy(ubifs_inode_slab);
if (found) {
/* Ensure the znode is dirtied */
if (znode->cnext || !ubifs_zn_dirty(znode)) {
- znode = dirty_cow_bottom_up(c,
- znode);
- if (IS_ERR(znode)) {
- err = PTR_ERR(znode);
- goto out_unlock;
- }
+ znode = dirty_cow_bottom_up(c, znode);
+ if (IS_ERR(znode)) {
+ err = PTR_ERR(znode);
+ goto out_unlock;
+ }
}
zbr = &znode->zbranch[n];
lnc_free(zbr);
/* Ensure the znode is dirtied */
if (znode->cnext || !ubifs_zn_dirty(znode)) {
- znode = dirty_cow_bottom_up(c, znode);
- if (IS_ERR(znode)) {
- err = PTR_ERR(znode);
- goto out_unlock;
- }
+ znode = dirty_cow_bottom_up(c, znode);
+ if (IS_ERR(znode)) {
+ err = PTR_ERR(znode);
+ goto out_unlock;
+ }
}
if (found == 1) {
/* Ensure the znode is dirtied */
if (znode->cnext || !ubifs_zn_dirty(znode)) {
- znode = dirty_cow_bottom_up(c, znode);
- if (IS_ERR(znode)) {
- err = PTR_ERR(znode);
- goto out_unlock;
- }
+ znode = dirty_cow_bottom_up(c, znode);
+ if (IS_ERR(znode)) {
+ err = PTR_ERR(znode);
+ goto out_unlock;
+ }
}
/* Remove all keys in range except the first */
}
#ifdef CONFIG_UBIFS_FS_DEBUG
- c->new_ihead_lnum = lnum;
- c->new_ihead_offs = buf_offs;
+ c->dbg->new_ihead_lnum = lnum;
+ c->dbg->new_ihead_offs = buf_offs;
#endif
return 0;
* budgeting subsystem to assume the index is already committed,
* even though it is not.
*/
+ ubifs_assert(c->min_idx_lebs == ubifs_calc_min_idx_lebs(c));
c->old_idx_sz = c->calc_idx_sz;
c->budg_uncommitted_idx = 0;
+ c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
spin_unlock(&c->space_lock);
mutex_unlock(&c->tnc_mutex);
}
#ifdef CONFIG_UBIFS_FS_DEBUG
- if (lnum != c->new_ihead_lnum || buf_offs != c->new_ihead_offs) {
+ if (lnum != c->dbg->new_ihead_lnum ||
+ buf_offs != c->dbg->new_ihead_offs) {
ubifs_err("inconsistent ihead");
return -EINVAL;
}
*/
#define UBIFS_MIN_COMPR_LEN 128
+/*
+ * If compressed data length is less than %UBIFS_MIN_COMPRESS_DIFF bytes
+ * shorter than uncompressed data length, UBIFS preferes to leave this data
+ * node uncompress, because it'll be read faster.
+ */
+#define UBIFS_MIN_COMPRESS_DIFF 64
+
/* Root inode number */
#define UBIFS_ROOT_INO 1
#define SQNUM_WARN_WATERMARK 0xFFFFFFFF00000000ULL
#define SQNUM_WATERMARK 0xFFFFFFFFFF000000ULL
+/*
+ * Minimum amount of LEBs reserved for the index. At present the index needs at
+ * least 2 LEBs: one for the index head and one for in-the-gaps method (which
+ * currently does not cater for the index head and so excludes it from
+ * consideration).
+ */
+#define MIN_INDEX_LEBS 2
+
/* Minimum amount of data UBIFS writes to the flash */
#define MIN_WRITE_SZ (UBIFS_DATA_NODE_SZ + 8)
unsigned int dirty:1;
unsigned int xattr:1;
unsigned int bulk_read:1;
+ unsigned int compr_type:2;
struct mutex ui_mutex;
spinlock_t ui_lock;
loff_t synced_i_size;
loff_t ui_size;
int flags;
- int compr_type;
pgoff_t last_page_read;
pgoff_t read_in_a_row;
int data_len;
*
* LPROPS_UNCAT: not categorized
* LPROPS_DIRTY: dirty > 0, not index
- * LPROPS_DIRTY_IDX: dirty + free > UBIFS_CH_SZ and index
+ * LPROPS_DIRTY_IDX: dirty + free > @c->min_idx_node_sze and index
* LPROPS_FREE: free > 0, not empty, not index
* LPROPS_HEAP_CNT: number of heaps used for storing categorized LEBs
* LPROPS_EMPTY: LEB is empty, not taken
struct ubifs_lpt_lprops {
int free;
int dirty;
- unsigned tgc : 1;
- unsigned cmt : 1;
+ unsigned tgc:1;
+ unsigned cmt:1;
};
/**
* @empty_lebs: number of empty LEBs
* @taken_empty_lebs: number of taken LEBs
* @idx_lebs: number of indexing LEBs
- * @total_free: total free space in bytes
- * @total_dirty: total dirty space in bytes
- * @total_used: total used space in bytes (includes only data LEBs)
- * @total_dead: total dead space in bytes (includes only data LEBs)
- * @total_dark: total dark space in bytes (includes only data LEBs)
+ * @total_free: total free space in bytes (includes all LEBs)
+ * @total_dirty: total dirty space in bytes (includes all LEBs)
+ * @total_used: total used space in bytes (does not include index LEBs)
+ * @total_dead: total dead space in bytes (does not include index LEBs)
+ * @total_dark: total dark space in bytes (does not include index LEBs)
+ *
+ * The @taken_empty_lebs field counts the LEBs that are in the transient state
+ * of having been "taken" for use but not yet written to. @taken_empty_lebs is
+ * needed to account correctly for @gc_lnum, otherwise @empty_lebs could be
+ * used by itself (in which case 'unused_lebs' would be a better name). In the
+ * case of @gc_lnum, it is "taken" at mount time or whenever a LEB is retained
+ * by GC, but unlike other empty LEBs that are "taken", it may not be written
+ * straight away (i.e. before the next commit start or unmount), so either
+ * @gc_lnum must be specially accounted for, or the current approach followed
+ * i.e. count it under @taken_empty_lebs.
*
- * N.B. total_dirty and total_used are different to other total_* fields,
- * because they account _all_ LEBs, not just data LEBs.
+ * @empty_lebs includes @taken_empty_lebs.
*
- * 'taken_empty_lebs' counts the LEBs that are in the transient state of having
- * been 'taken' for use but not yet written to. 'taken_empty_lebs' is needed
- * to account correctly for gc_lnum, otherwise 'empty_lebs' could be used
- * by itself (in which case 'unused_lebs' would be a better name). In the case
- * of gc_lnum, it is 'taken' at mount time or whenever a LEB is retained by GC,
- * but unlike other empty LEBs that are 'taken', it may not be written straight
- * away (i.e. before the next commit start or unmount), so either gc_lnum must
- * be specially accounted for, or the current approach followed i.e. count it
- * under 'taken_empty_lebs'.
+ * @total_used, @total_dead and @total_dark fields do not account indexing
+ * LEBs.
*/
struct ubifs_lp_stats {
int empty_lebs;
/**
* struct ubifs_mount_opts - UBIFS-specific mount options information.
* @unmount_mode: selected unmount mode (%0 default, %1 normal, %2 fast)
- * @bulk_read: enable bulk-reads
- * @chk_data_crc: check CRCs when reading data nodes
+ * @bulk_read: enable/disable bulk-reads (%0 default, %1 disabe, %2 enable)
+ * @chk_data_crc: enable/disable CRC data checking when reading data nodes
+ * (%0 default, %1 disabe, %2 enable)
+ * @override_compr: override default compressor (%0 - do not override and use
+ * superblock compressor, %1 - override and use compressor
+ * specified in @compr_type)
+ * @compr_type: compressor type to override the superblock compressor with
+ * (%UBIFS_COMPR_NONE, etc)
*/
struct ubifs_mount_opts {
unsigned int unmount_mode:2;
unsigned int bulk_read:2;
unsigned int chk_data_crc:2;
+ unsigned int override_compr:1;
+ unsigned int compr_type:2;
};
+struct ubifs_debug_info;
+
/**
* struct ubifs_info - UBIFS file-system description data structure
* (per-superblock).
* @no_chk_data_crc: do not check CRCs when reading data nodes (except during
* recovery)
* @bulk_read: enable bulk-reads
+ * @default_compr: default compression algorithm (%UBIFS_COMPR_LZO, etc)
*
* @tnc_mutex: protects the Tree Node Cache (TNC), @zroot, @cnext, @enext, and
* @calc_idx_sz
* @ileb_nxt: next pre-allocated index LEBs
* @old_idx: tree of index nodes obsoleted since the last commit start
* @bottom_up_buf: a buffer which is used by 'dirty_cow_bottom_up()' in tnc.c
- * @new_ihead_lnum: used by debugging to check ihead_lnum
- * @new_ihead_offs: used by debugging to check ihead_offs
*
* @mst_node: master node
* @mst_offs: offset of valid master node
* @main_lebs: count of LEBs in the main area
* @main_first: first LEB of the main area
* @main_bytes: main area size in bytes
- * @default_compr: default compression algorithm (%UBIFS_COMPR_LZO, etc)
*
* @key_hash_type: type of the key hash
* @key_hash: direntry key hash function
* @always_chk_crc: always check CRCs (while mounting and remounting rw)
* @mount_opts: UBIFS-specific mount options
*
- * @dbg_buf: a buffer of LEB size used for debugging purposes
- * @old_zroot: old index root - used by 'dbg_check_old_index()'
- * @old_zroot_level: old index root level - used by 'dbg_check_old_index()'
- * @old_zroot_sqnum: old index root sqnum - used by 'dbg_check_old_index()'
- * @failure_mode: failure mode for recovery testing
- * @fail_delay: 0=>don't delay, 1=>delay a time, 2=>delay a number of calls
- * @fail_timeout: time in jiffies when delay of failure mode expires
- * @fail_cnt: current number of calls to failure mode I/O functions
- * @fail_cnt_max: number of calls by which to delay failure mode
+ * @dbg: debugging-related information
*/
struct ubifs_info {
struct super_block *vfs_sb;
unsigned int big_lpt:1;
unsigned int no_chk_data_crc:1;
unsigned int bulk_read:1;
+ unsigned int default_compr:2;
struct mutex tnc_mutex;
struct ubifs_zbranch zroot;
int ileb_nxt;
struct rb_root old_idx;
int *bottom_up_buf;
-#ifdef CONFIG_UBIFS_FS_DEBUG
- int new_ihead_lnum;
- int new_ihead_offs;
-#endif
struct ubifs_mst_node *mst_node;
int mst_offs;
int main_lebs;
int main_first;
long long main_bytes;
- int default_compr;
uint8_t key_hash_type;
uint32_t (*key_hash)(const char *str, int len);
void *sbuf;
struct list_head idx_gc;
int idx_gc_cnt;
- volatile int gc_seq;
- volatile int gced_lnum;
+ int gc_seq;
+ int gced_lnum;
struct list_head infos_list;
struct mutex umount_mutex;
struct ubifs_mount_opts mount_opts;
#ifdef CONFIG_UBIFS_FS_DEBUG
- void *dbg_buf;
- struct ubifs_zbranch old_zroot;
- int old_zroot_level;
- unsigned long long old_zroot_sqnum;
- int failure_mode;
- int fail_delay;
- unsigned long fail_timeout;
- unsigned int fail_cnt;
- unsigned int fail_cnt_max;
- long long chk_lpt_sz;
- long long chk_lpt_sz2;
- long long chk_lpt_wastage;
- int chk_lpt_lebs;
- int new_nhead_lnum;
- int new_nhead_offs;
+ struct ubifs_debug_info *dbg;
#endif
};
long long ubifs_get_free_space(struct ubifs_info *c);
int ubifs_calc_min_idx_lebs(struct ubifs_info *c);
void ubifs_convert_page_budget(struct ubifs_info *c);
-long long ubifs_reported_space(const struct ubifs_info *c, uint64_t free);
+long long ubifs_reported_space(const struct ubifs_info *c, long long free);
long long ubifs_calc_available(const struct ubifs_info *c, int min_idx_lebs);
/* find.c */
void ubifs_add_nnode_dirt(struct ubifs_info *c, struct ubifs_nnode *nnode);
uint32_t ubifs_unpack_bits(uint8_t **addr, int *pos, int nrbits);
struct ubifs_nnode *ubifs_first_nnode(struct ubifs_info *c, int *hght);
+/* Needed only in debugging code in lpt_commit.c */
+int ubifs_unpack_nnode(const struct ubifs_info *c, void *buf,
+ struct ubifs_nnode *nnode);
/* lpt_commit.c */
int ubifs_lpt_start_commit(struct ubifs_info *c);
/* compressor.c */
int __init ubifs_compressors_init(void);
-void __exit ubifs_compressors_exit(void);
+void ubifs_compressors_exit(void);
void ubifs_compress(const void *in_buf, int in_len, void *out_buf, int *out_len,
int *compr_type);
int ubifs_decompress(const void *buf, int len, void *out, int *out_len,
projects / linux-drm-fsl-dcu.git / commitdiff