linux/fs/jbd2/revoke.c
Michal Hocko 7b506b1035 jbd2: get rid of open coded allocation retry loop
insert_revoke_hash does an open coded endless allocation loop if
journal_oom_retry is true. It doesn't implement any allocation fallback
strategy between the retries, though. The memory allocator doesn't know
about the never fail requirement so it cannot potentially help to move
on with the allocation (e.g. use memory reserves).

Get rid of the retry loop and use __GFP_NOFAIL instead. We will lose the
debugging message but I am not sure it is anyhow helpful.

Do the same for journal_alloc_journal_head which is doing a similar
thing.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2015-06-15 15:45:58 -04:00

759 lines
22 KiB
C

/*
* linux/fs/jbd2/revoke.c
*
* Written by Stephen C. Tweedie <sct@redhat.com>, 2000
*
* Copyright 2000 Red Hat corp --- All Rights Reserved
*
* This file is part of the Linux kernel and is made available under
* the terms of the GNU General Public License, version 2, or at your
* option, any later version, incorporated herein by reference.
*
* Journal revoke routines for the generic filesystem journaling code;
* part of the ext2fs journaling system.
*
* Revoke is the mechanism used to prevent old log records for deleted
* metadata from being replayed on top of newer data using the same
* blocks. The revoke mechanism is used in two separate places:
*
* + Commit: during commit we write the entire list of the current
* transaction's revoked blocks to the journal
*
* + Recovery: during recovery we record the transaction ID of all
* revoked blocks. If there are multiple revoke records in the log
* for a single block, only the last one counts, and if there is a log
* entry for a block beyond the last revoke, then that log entry still
* gets replayed.
*
* We can get interactions between revokes and new log data within a
* single transaction:
*
* Block is revoked and then journaled:
* The desired end result is the journaling of the new block, so we
* cancel the revoke before the transaction commits.
*
* Block is journaled and then revoked:
* The revoke must take precedence over the write of the block, so we
* need either to cancel the journal entry or to write the revoke
* later in the log than the log block. In this case, we choose the
* latter: journaling a block cancels any revoke record for that block
* in the current transaction, so any revoke for that block in the
* transaction must have happened after the block was journaled and so
* the revoke must take precedence.
*
* Block is revoked and then written as data:
* The data write is allowed to succeed, but the revoke is _not_
* cancelled. We still need to prevent old log records from
* overwriting the new data. We don't even need to clear the revoke
* bit here.
*
* We cache revoke status of a buffer in the current transaction in b_states
* bits. As the name says, revokevalid flag indicates that the cached revoke
* status of a buffer is valid and we can rely on the cached status.
*
* Revoke information on buffers is a tri-state value:
*
* RevokeValid clear: no cached revoke status, need to look it up
* RevokeValid set, Revoked clear:
* buffer has not been revoked, and cancel_revoke
* need do nothing.
* RevokeValid set, Revoked set:
* buffer has been revoked.
*
* Locking rules:
* We keep two hash tables of revoke records. One hashtable belongs to the
* running transaction (is pointed to by journal->j_revoke), the other one
* belongs to the committing transaction. Accesses to the second hash table
* happen only from the kjournald and no other thread touches this table. Also
* journal_switch_revoke_table() which switches which hashtable belongs to the
* running and which to the committing transaction is called only from
* kjournald. Therefore we need no locks when accessing the hashtable belonging
* to the committing transaction.
*
* All users operating on the hash table belonging to the running transaction
* have a handle to the transaction. Therefore they are safe from kjournald
* switching hash tables under them. For operations on the lists of entries in
* the hash table j_revoke_lock is used.
*
* Finally, also replay code uses the hash tables but at this moment no one else
* can touch them (filesystem isn't mounted yet) and hence no locking is
* needed.
*/
#ifndef __KERNEL__
#include "jfs_user.h"
#else
#include <linux/time.h>
#include <linux/fs.h>
#include <linux/jbd2.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <linux/init.h>
#include <linux/bio.h>
#include <linux/log2.h>
#include <linux/hash.h>
#endif
static struct kmem_cache *jbd2_revoke_record_cache;
static struct kmem_cache *jbd2_revoke_table_cache;
/* Each revoke record represents one single revoked block. During
journal replay, this involves recording the transaction ID of the
last transaction to revoke this block. */
struct jbd2_revoke_record_s
{
struct list_head hash;
tid_t sequence; /* Used for recovery only */
unsigned long long blocknr;
};
/* The revoke table is just a simple hash table of revoke records. */
struct jbd2_revoke_table_s
{
/* It is conceivable that we might want a larger hash table
* for recovery. Must be a power of two. */
int hash_size;
int hash_shift;
struct list_head *hash_table;
};
#ifdef __KERNEL__
static void write_one_revoke_record(journal_t *, transaction_t *,
struct list_head *,
struct buffer_head **, int *,
struct jbd2_revoke_record_s *, int);
static void flush_descriptor(journal_t *, struct buffer_head *, int, int);
#endif
/* Utility functions to maintain the revoke table */
static inline int hash(journal_t *journal, unsigned long long block)
{
return hash_64(block, journal->j_revoke->hash_shift);
}
static int insert_revoke_hash(journal_t *journal, unsigned long long blocknr,
tid_t seq)
{
struct list_head *hash_list;
struct jbd2_revoke_record_s *record;
gfp_t gfp_mask = GFP_NOFS;
if (journal_oom_retry)
gfp_mask |= __GFP_NOFAIL;
record = kmem_cache_alloc(jbd2_revoke_record_cache, gfp_mask);
if (!record)
return -ENOMEM;
record->sequence = seq;
record->blocknr = blocknr;
hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
spin_lock(&journal->j_revoke_lock);
list_add(&record->hash, hash_list);
spin_unlock(&journal->j_revoke_lock);
return 0;
}
/* Find a revoke record in the journal's hash table. */
static struct jbd2_revoke_record_s *find_revoke_record(journal_t *journal,
unsigned long long blocknr)
{
struct list_head *hash_list;
struct jbd2_revoke_record_s *record;
hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
spin_lock(&journal->j_revoke_lock);
record = (struct jbd2_revoke_record_s *) hash_list->next;
while (&(record->hash) != hash_list) {
if (record->blocknr == blocknr) {
spin_unlock(&journal->j_revoke_lock);
return record;
}
record = (struct jbd2_revoke_record_s *) record->hash.next;
}
spin_unlock(&journal->j_revoke_lock);
return NULL;
}
void jbd2_journal_destroy_revoke_caches(void)
{
if (jbd2_revoke_record_cache) {
kmem_cache_destroy(jbd2_revoke_record_cache);
jbd2_revoke_record_cache = NULL;
}
if (jbd2_revoke_table_cache) {
kmem_cache_destroy(jbd2_revoke_table_cache);
jbd2_revoke_table_cache = NULL;
}
}
int __init jbd2_journal_init_revoke_caches(void)
{
J_ASSERT(!jbd2_revoke_record_cache);
J_ASSERT(!jbd2_revoke_table_cache);
jbd2_revoke_record_cache = KMEM_CACHE(jbd2_revoke_record_s,
SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY);
if (!jbd2_revoke_record_cache)
goto record_cache_failure;
jbd2_revoke_table_cache = KMEM_CACHE(jbd2_revoke_table_s,
SLAB_TEMPORARY);
if (!jbd2_revoke_table_cache)
goto table_cache_failure;
return 0;
table_cache_failure:
jbd2_journal_destroy_revoke_caches();
record_cache_failure:
return -ENOMEM;
}
static struct jbd2_revoke_table_s *jbd2_journal_init_revoke_table(int hash_size)
{
int shift = 0;
int tmp = hash_size;
struct jbd2_revoke_table_s *table;
table = kmem_cache_alloc(jbd2_revoke_table_cache, GFP_KERNEL);
if (!table)
goto out;
while((tmp >>= 1UL) != 0UL)
shift++;
table->hash_size = hash_size;
table->hash_shift = shift;
table->hash_table =
kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL);
if (!table->hash_table) {
kmem_cache_free(jbd2_revoke_table_cache, table);
table = NULL;
goto out;
}
for (tmp = 0; tmp < hash_size; tmp++)
INIT_LIST_HEAD(&table->hash_table[tmp]);
out:
return table;
}
static void jbd2_journal_destroy_revoke_table(struct jbd2_revoke_table_s *table)
{
int i;
struct list_head *hash_list;
for (i = 0; i < table->hash_size; i++) {
hash_list = &table->hash_table[i];
J_ASSERT(list_empty(hash_list));
}
kfree(table->hash_table);
kmem_cache_free(jbd2_revoke_table_cache, table);
}
/* Initialise the revoke table for a given journal to a given size. */
int jbd2_journal_init_revoke(journal_t *journal, int hash_size)
{
J_ASSERT(journal->j_revoke_table[0] == NULL);
J_ASSERT(is_power_of_2(hash_size));
journal->j_revoke_table[0] = jbd2_journal_init_revoke_table(hash_size);
if (!journal->j_revoke_table[0])
goto fail0;
journal->j_revoke_table[1] = jbd2_journal_init_revoke_table(hash_size);
if (!journal->j_revoke_table[1])
goto fail1;
journal->j_revoke = journal->j_revoke_table[1];
spin_lock_init(&journal->j_revoke_lock);
return 0;
fail1:
jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
fail0:
return -ENOMEM;
}
/* Destroy a journal's revoke table. The table must already be empty! */
void jbd2_journal_destroy_revoke(journal_t *journal)
{
journal->j_revoke = NULL;
if (journal->j_revoke_table[0])
jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
if (journal->j_revoke_table[1])
jbd2_journal_destroy_revoke_table(journal->j_revoke_table[1]);
}
#ifdef __KERNEL__
/*
* jbd2_journal_revoke: revoke a given buffer_head from the journal. This
* prevents the block from being replayed during recovery if we take a
* crash after this current transaction commits. Any subsequent
* metadata writes of the buffer in this transaction cancel the
* revoke.
*
* Note that this call may block --- it is up to the caller to make
* sure that there are no further calls to journal_write_metadata
* before the revoke is complete. In ext3, this implies calling the
* revoke before clearing the block bitmap when we are deleting
* metadata.
*
* Revoke performs a jbd2_journal_forget on any buffer_head passed in as a
* parameter, but does _not_ forget the buffer_head if the bh was only
* found implicitly.
*
* bh_in may not be a journalled buffer - it may have come off
* the hash tables without an attached journal_head.
*
* If bh_in is non-zero, jbd2_journal_revoke() will decrement its b_count
* by one.
*/
int jbd2_journal_revoke(handle_t *handle, unsigned long long blocknr,
struct buffer_head *bh_in)
{
struct buffer_head *bh = NULL;
journal_t *journal;
struct block_device *bdev;
int err;
might_sleep();
if (bh_in)
BUFFER_TRACE(bh_in, "enter");
journal = handle->h_transaction->t_journal;
if (!jbd2_journal_set_features(journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)){
J_ASSERT (!"Cannot set revoke feature!");
return -EINVAL;
}
bdev = journal->j_fs_dev;
bh = bh_in;
if (!bh) {
bh = __find_get_block(bdev, blocknr, journal->j_blocksize);
if (bh)
BUFFER_TRACE(bh, "found on hash");
}
#ifdef JBD2_EXPENSIVE_CHECKING
else {
struct buffer_head *bh2;
/* If there is a different buffer_head lying around in
* memory anywhere... */
bh2 = __find_get_block(bdev, blocknr, journal->j_blocksize);
if (bh2) {
/* ... and it has RevokeValid status... */
if (bh2 != bh && buffer_revokevalid(bh2))
/* ...then it better be revoked too,
* since it's illegal to create a revoke
* record against a buffer_head which is
* not marked revoked --- that would
* risk missing a subsequent revoke
* cancel. */
J_ASSERT_BH(bh2, buffer_revoked(bh2));
put_bh(bh2);
}
}
#endif
/* We really ought not ever to revoke twice in a row without
first having the revoke cancelled: it's illegal to free a
block twice without allocating it in between! */
if (bh) {
if (!J_EXPECT_BH(bh, !buffer_revoked(bh),
"inconsistent data on disk")) {
if (!bh_in)
brelse(bh);
return -EIO;
}
set_buffer_revoked(bh);
set_buffer_revokevalid(bh);
if (bh_in) {
BUFFER_TRACE(bh_in, "call jbd2_journal_forget");
jbd2_journal_forget(handle, bh_in);
} else {
BUFFER_TRACE(bh, "call brelse");
__brelse(bh);
}
}
jbd_debug(2, "insert revoke for block %llu, bh_in=%p\n",blocknr, bh_in);
err = insert_revoke_hash(journal, blocknr,
handle->h_transaction->t_tid);
BUFFER_TRACE(bh_in, "exit");
return err;
}
/*
* Cancel an outstanding revoke. For use only internally by the
* journaling code (called from jbd2_journal_get_write_access).
*
* We trust buffer_revoked() on the buffer if the buffer is already
* being journaled: if there is no revoke pending on the buffer, then we
* don't do anything here.
*
* This would break if it were possible for a buffer to be revoked and
* discarded, and then reallocated within the same transaction. In such
* a case we would have lost the revoked bit, but when we arrived here
* the second time we would still have a pending revoke to cancel. So,
* do not trust the Revoked bit on buffers unless RevokeValid is also
* set.
*/
int jbd2_journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
{
struct jbd2_revoke_record_s *record;
journal_t *journal = handle->h_transaction->t_journal;
int need_cancel;
int did_revoke = 0; /* akpm: debug */
struct buffer_head *bh = jh2bh(jh);
jbd_debug(4, "journal_head %p, cancelling revoke\n", jh);
/* Is the existing Revoke bit valid? If so, we trust it, and
* only perform the full cancel if the revoke bit is set. If
* not, we can't trust the revoke bit, and we need to do the
* full search for a revoke record. */
if (test_set_buffer_revokevalid(bh)) {
need_cancel = test_clear_buffer_revoked(bh);
} else {
need_cancel = 1;
clear_buffer_revoked(bh);
}
if (need_cancel) {
record = find_revoke_record(journal, bh->b_blocknr);
if (record) {
jbd_debug(4, "cancelled existing revoke on "
"blocknr %llu\n", (unsigned long long)bh->b_blocknr);
spin_lock(&journal->j_revoke_lock);
list_del(&record->hash);
spin_unlock(&journal->j_revoke_lock);
kmem_cache_free(jbd2_revoke_record_cache, record);
did_revoke = 1;
}
}
#ifdef JBD2_EXPENSIVE_CHECKING
/* There better not be one left behind by now! */
record = find_revoke_record(journal, bh->b_blocknr);
J_ASSERT_JH(jh, record == NULL);
#endif
/* Finally, have we just cleared revoke on an unhashed
* buffer_head? If so, we'd better make sure we clear the
* revoked status on any hashed alias too, otherwise the revoke
* state machine will get very upset later on. */
if (need_cancel) {
struct buffer_head *bh2;
bh2 = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
if (bh2) {
if (bh2 != bh)
clear_buffer_revoked(bh2);
__brelse(bh2);
}
}
return did_revoke;
}
/*
* journal_clear_revoked_flag clears revoked flag of buffers in
* revoke table to reflect there is no revoked buffers in the next
* transaction which is going to be started.
*/
void jbd2_clear_buffer_revoked_flags(journal_t *journal)
{
struct jbd2_revoke_table_s *revoke = journal->j_revoke;
int i = 0;
for (i = 0; i < revoke->hash_size; i++) {
struct list_head *hash_list;
struct list_head *list_entry;
hash_list = &revoke->hash_table[i];
list_for_each(list_entry, hash_list) {
struct jbd2_revoke_record_s *record;
struct buffer_head *bh;
record = (struct jbd2_revoke_record_s *)list_entry;
bh = __find_get_block(journal->j_fs_dev,
record->blocknr,
journal->j_blocksize);
if (bh) {
clear_buffer_revoked(bh);
__brelse(bh);
}
}
}
}
/* journal_switch_revoke table select j_revoke for next transaction
* we do not want to suspend any processing until all revokes are
* written -bzzz
*/
void jbd2_journal_switch_revoke_table(journal_t *journal)
{
int i;
if (journal->j_revoke == journal->j_revoke_table[0])
journal->j_revoke = journal->j_revoke_table[1];
else
journal->j_revoke = journal->j_revoke_table[0];
for (i = 0; i < journal->j_revoke->hash_size; i++)
INIT_LIST_HEAD(&journal->j_revoke->hash_table[i]);
}
/*
* Write revoke records to the journal for all entries in the current
* revoke hash, deleting the entries as we go.
*/
void jbd2_journal_write_revoke_records(journal_t *journal,
transaction_t *transaction,
struct list_head *log_bufs,
int write_op)
{
struct buffer_head *descriptor;
struct jbd2_revoke_record_s *record;
struct jbd2_revoke_table_s *revoke;
struct list_head *hash_list;
int i, offset, count;
descriptor = NULL;
offset = 0;
count = 0;
/* select revoke table for committing transaction */
revoke = journal->j_revoke == journal->j_revoke_table[0] ?
journal->j_revoke_table[1] : journal->j_revoke_table[0];
for (i = 0; i < revoke->hash_size; i++) {
hash_list = &revoke->hash_table[i];
while (!list_empty(hash_list)) {
record = (struct jbd2_revoke_record_s *)
hash_list->next;
write_one_revoke_record(journal, transaction, log_bufs,
&descriptor, &offset,
record, write_op);
count++;
list_del(&record->hash);
kmem_cache_free(jbd2_revoke_record_cache, record);
}
}
if (descriptor)
flush_descriptor(journal, descriptor, offset, write_op);
jbd_debug(1, "Wrote %d revoke records\n", count);
}
/*
* Write out one revoke record. We need to create a new descriptor
* block if the old one is full or if we have not already created one.
*/
static void write_one_revoke_record(journal_t *journal,
transaction_t *transaction,
struct list_head *log_bufs,
struct buffer_head **descriptorp,
int *offsetp,
struct jbd2_revoke_record_s *record,
int write_op)
{
int csum_size = 0;
struct buffer_head *descriptor;
int sz, offset;
journal_header_t *header;
/* If we are already aborting, this all becomes a noop. We
still need to go round the loop in
jbd2_journal_write_revoke_records in order to free all of the
revoke records: only the IO to the journal is omitted. */
if (is_journal_aborted(journal))
return;
descriptor = *descriptorp;
offset = *offsetp;
/* Do we need to leave space at the end for a checksum? */
if (jbd2_journal_has_csum_v2or3(journal))
csum_size = sizeof(struct jbd2_journal_revoke_tail);
if (JBD2_HAS_INCOMPAT_FEATURE(journal, JBD2_FEATURE_INCOMPAT_64BIT))
sz = 8;
else
sz = 4;
/* Make sure we have a descriptor with space left for the record */
if (descriptor) {
if (offset + sz > journal->j_blocksize - csum_size) {
flush_descriptor(journal, descriptor, offset, write_op);
descriptor = NULL;
}
}
if (!descriptor) {
descriptor = jbd2_journal_get_descriptor_buffer(journal);
if (!descriptor)
return;
header = (journal_header_t *)descriptor->b_data;
header->h_magic = cpu_to_be32(JBD2_MAGIC_NUMBER);
header->h_blocktype = cpu_to_be32(JBD2_REVOKE_BLOCK);
header->h_sequence = cpu_to_be32(transaction->t_tid);
/* Record it so that we can wait for IO completion later */
BUFFER_TRACE(descriptor, "file in log_bufs");
jbd2_file_log_bh(log_bufs, descriptor);
offset = sizeof(jbd2_journal_revoke_header_t);
*descriptorp = descriptor;
}
if (JBD2_HAS_INCOMPAT_FEATURE(journal, JBD2_FEATURE_INCOMPAT_64BIT))
* ((__be64 *)(&descriptor->b_data[offset])) =
cpu_to_be64(record->blocknr);
else
* ((__be32 *)(&descriptor->b_data[offset])) =
cpu_to_be32(record->blocknr);
offset += sz;
*offsetp = offset;
}
static void jbd2_revoke_csum_set(journal_t *j, struct buffer_head *bh)
{
struct jbd2_journal_revoke_tail *tail;
__u32 csum;
if (!jbd2_journal_has_csum_v2or3(j))
return;
tail = (struct jbd2_journal_revoke_tail *)(bh->b_data + j->j_blocksize -
sizeof(struct jbd2_journal_revoke_tail));
tail->r_checksum = 0;
csum = jbd2_chksum(j, j->j_csum_seed, bh->b_data, j->j_blocksize);
tail->r_checksum = cpu_to_be32(csum);
}
/*
* Flush a revoke descriptor out to the journal. If we are aborting,
* this is a noop; otherwise we are generating a buffer which needs to
* be waited for during commit, so it has to go onto the appropriate
* journal buffer list.
*/
static void flush_descriptor(journal_t *journal,
struct buffer_head *descriptor,
int offset, int write_op)
{
jbd2_journal_revoke_header_t *header;
if (is_journal_aborted(journal)) {
put_bh(descriptor);
return;
}
header = (jbd2_journal_revoke_header_t *)descriptor->b_data;
header->r_count = cpu_to_be32(offset);
jbd2_revoke_csum_set(journal, descriptor);
set_buffer_jwrite(descriptor);
BUFFER_TRACE(descriptor, "write");
set_buffer_dirty(descriptor);
write_dirty_buffer(descriptor, write_op);
}
#endif
/*
* Revoke support for recovery.
*
* Recovery needs to be able to:
*
* record all revoke records, including the tid of the latest instance
* of each revoke in the journal
*
* check whether a given block in a given transaction should be replayed
* (ie. has not been revoked by a revoke record in that or a subsequent
* transaction)
*
* empty the revoke table after recovery.
*/
/*
* First, setting revoke records. We create a new revoke record for
* every block ever revoked in the log as we scan it for recovery, and
* we update the existing records if we find multiple revokes for a
* single block.
*/
int jbd2_journal_set_revoke(journal_t *journal,
unsigned long long blocknr,
tid_t sequence)
{
struct jbd2_revoke_record_s *record;
record = find_revoke_record(journal, blocknr);
if (record) {
/* If we have multiple occurrences, only record the
* latest sequence number in the hashed record */
if (tid_gt(sequence, record->sequence))
record->sequence = sequence;
return 0;
}
return insert_revoke_hash(journal, blocknr, sequence);
}
/*
* Test revoke records. For a given block referenced in the log, has
* that block been revoked? A revoke record with a given transaction
* sequence number revokes all blocks in that transaction and earlier
* ones, but later transactions still need replayed.
*/
int jbd2_journal_test_revoke(journal_t *journal,
unsigned long long blocknr,
tid_t sequence)
{
struct jbd2_revoke_record_s *record;
record = find_revoke_record(journal, blocknr);
if (!record)
return 0;
if (tid_gt(sequence, record->sequence))
return 0;
return 1;
}
/*
* Finally, once recovery is over, we need to clear the revoke table so
* that it can be reused by the running filesystem.
*/
void jbd2_journal_clear_revoke(journal_t *journal)
{
int i;
struct list_head *hash_list;
struct jbd2_revoke_record_s *record;
struct jbd2_revoke_table_s *revoke;
revoke = journal->j_revoke;
for (i = 0; i < revoke->hash_size; i++) {
hash_list = &revoke->hash_table[i];
while (!list_empty(hash_list)) {
record = (struct jbd2_revoke_record_s*) hash_list->next;
list_del(&record->hash);
kmem_cache_free(jbd2_revoke_record_cache, record);
}
}
}