/*
* The ARC has requested that the filesystem drop entries from the dentry
* and inode caches. This can occur when the ARC needs to free meta data
* blocks but can't because they are all pinned by entries in these caches.
*/
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
int error = 0;
#if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK)
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
#endif
ZFS_ENTER(zsb);
#if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \
defined(SHRINK_CONTROL_HAS_NID) && \
defined(SHRINKER_NUMA_AWARE)
if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) {
*objects = 0;
for_each_online_node(sc.nid)
*objects += (*shrinker->scan_objects)(shrinker, &sc);
} else {
*objects = (*shrinker->scan_objects)(shrinker, &sc);
}
#elif defined(HAVE_SPLIT_SHRINKER_CALLBACK)
*objects = (*shrinker->scan_objects)(shrinker, &sc);
#elif defined(HAVE_SHRINK)
*objects = (*shrinker->shrink)(shrinker, &sc);
#elif defined(HAVE_D_PRUNE_ALIASES)
#define D_PRUNE_ALIASES_IS_DEFAULT
*objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
#else
#error "No available dentry and inode cache pruning mechanism."
#endif
#if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT)
#undef D_PRUNE_ALIASES_IS_DEFAULT
/*
* Fall back to zfs_sb_prune_aliases if the kernel's per-superblock
* shrinker couldn't free anything, possibly due to the inodes being
* allocated in a different memcg.
*/
if (*objects == 0)
*objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
#endif
ZFS_EXIT(zsb);
dprintf_ds(zsb->z_os->os_dsl_dataset,
"pruning, nr_to_scan=%lu objects=%d error=%d\n",
nr_to_scan, *objects, error);
return (error);
}
/*
* The ARC has requested that the filesystem drop entries from the dentry
* and inode caches. This can occur when the ARC needs to free meta data
* blocks but can't because they are all pinned by entries in these caches.
*/
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
int error = 0;
#if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK)
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
#endif
ZFS_ENTER(zsb);
#if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \
defined(SHRINK_CONTROL_HAS_NID) && \
defined(SHRINKER_NUMA_AWARE)
if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) {
*objects = 0;
for_each_online_node(sc.nid)
*objects += (*shrinker->scan_objects)(shrinker, &sc);
} else {
*objects = (*shrinker->scan_objects)(shrinker, &sc);
}
#elif defined(HAVE_SPLIT_SHRINKER_CALLBACK)
*objects = (*shrinker->scan_objects)(shrinker, &sc);
#elif defined(HAVE_SHRINK)
*objects = (*shrinker->shrink)(shrinker, &sc);
#elif defined(HAVE_D_PRUNE_ALIASES)
*objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
#else
#error "No available dentry and inode cache pruning mechanism."
#endif
ZFS_EXIT(zsb);
dprintf_ds(zsb->z_os->os_dsl_dataset,
"pruning, nr_to_scan=%lu objects=%d error=%d\n",
nr_to_scan, *objects, error);
return (error);
}
/*
* The ARC has requested that the filesystem drop entries from the dentry
* and inode caches. This can occur when the ARC needs to free meta data
* blocks but can't because they are all pinned by entries in these caches.
*/
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
int error = 0;
#if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK)
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
#endif
ZFS_ENTER(zsb);
#if defined(HAVE_SPLIT_SHRINKER_CALLBACK)
*objects = (*shrinker->scan_objects)(shrinker, &sc);
#elif defined(HAVE_SHRINK)
*objects = (*shrinker->shrink)(shrinker, &sc);
#else
/*
* Linux kernels older than 3.1 do not support a per-filesystem
* shrinker. Therefore, we must fall back to the only available
* interface which is to discard all unused dentries and inodes.
* This behavior clearly isn't ideal but it's required so the ARC
* may free memory. The performance impact is mitigated by the
* fact that the frequently accessed dentry and inode buffers will
* still be in the ARC making them relatively cheap to recreate.
*/
*objects = 0;
shrink_dcache_parent(sb->s_root);
#endif
ZFS_EXIT(zsb);
dprintf_ds(zsb->z_os->os_dsl_dataset,
"pruning, nr_to_scan=%lu objects=%d error=%d\n",
nr_to_scan, *objects, error);
return (error);
}
/* called from dsl */
void
dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx)
{
int txgoff;
zbookmark_t zb;
zio_prop_t zp;
zio_t *zio;
list_t *list;
list_t *newlist = NULL;
dbuf_dirty_record_t *dr;
dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
ASSERT(dmu_tx_is_syncing(tx));
/* XXX the write_done callback should really give us the tx... */
os->os_synctx = tx;
if (os->os_dsl_dataset == NULL) {
/*
* This is the MOS. If we have upgraded,
* spa_max_replication() could change, so reset
* os_copies here.
*/
os->os_copies = spa_max_replication(os->os_spa);
}
/*
* Create the root block IO
*/
SET_BOOKMARK(&zb, os->os_dsl_dataset ?
os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
arc_release(os->os_phys_buf, &os->os_phys_buf);
dmu_write_policy(os, NULL, 0, 0, &zp);
zio = arc_write(pio, os->os_spa, tx->tx_txg,
os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), &zp,
dmu_objset_write_ready, dmu_objset_write_done, os,
ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
/*
* Sync special dnodes - the parent IO for the sync is the root block
*/
DMU_META_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_META_DNODE(os), tx);
os->os_phys->os_flags = os->os_flags;
if (DMU_USERUSED_DNODE(os) &&
DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) {
DMU_USERUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_USERUSED_DNODE(os), tx);
DMU_GROUPUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_GROUPUSED_DNODE(os), tx);
}
txgoff = tx->tx_txg & TXG_MASK;
if (dmu_objset_userused_enabled(os)) {
newlist = &os->os_synced_dnodes;
/*
* We must create the list here because it uses the
* dn_dirty_link[] of this txg.
*/
list_create(newlist, sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[txgoff]));
}
dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx);
dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx);
list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff];
while (dr = list_head(list)) {
ASSERT0(dr->dr_dbuf->db_level);
list_remove(list, dr);
if (dr->dr_zio)
zio_nowait(dr->dr_zio);
}
/*
* Free intent log blocks up to this tx.
*/
zil_sync(os->os_zil, tx);
os->os_phys->os_zil_header = os->os_zil_header;
zio_nowait(zio);
}
/* called from dsl */
void
dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
{
int txgoff;
zbookmark_t zb;
writeprops_t wp = { 0 };
zio_t *zio;
list_t *list;
dbuf_dirty_record_t *dr;
dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
ASSERT(dmu_tx_is_syncing(tx));
/* XXX the write_done callback should really give us the tx... */
os->os_synctx = tx;
if (os->os_dsl_dataset == NULL) {
/*
* This is the MOS. If we have upgraded,
* spa_max_replication() could change, so reset
* os_copies here.
*/
os->os_copies = spa_max_replication(os->os_spa);
}
/*
* Create the root block IO
*/
zb.zb_objset = os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : 0;
zb.zb_object = 0;
zb.zb_level = -1;
zb.zb_blkid = 0;
if (BP_IS_OLDER(os->os_rootbp, tx->tx_txg)) {
(void) dsl_dataset_block_kill(os->os_dsl_dataset,
os->os_rootbp, pio, tx);
}
wp.wp_type = DMU_OT_OBJSET;
wp.wp_copies = os->os_copies;
wp.wp_level = (uint8_t)-1;
wp.wp_oschecksum = os->os_checksum;
wp.wp_oscompress = os->os_compress;
arc_release(os->os_phys_buf, &os->os_phys_buf);
zio = arc_write(pio, os->os_spa, &wp,
tx->tx_txg, os->os_rootbp, os->os_phys_buf, ready, NULL, os,
ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED | ZIO_FLAG_METADATA,
&zb);
/*
* Sync meta-dnode - the parent IO for the sync is the root block
*/
os->os_meta_dnode->dn_zio = zio;
dnode_sync(os->os_meta_dnode, tx);
txgoff = tx->tx_txg & TXG_MASK;
dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], tx);
dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], tx);
list = &os->os_meta_dnode->dn_dirty_records[txgoff];
while (dr = list_head(list)) {
ASSERT(dr->dr_dbuf->db_level == 0);
list_remove(list, dr);
if (dr->dr_zio)
zio_nowait(dr->dr_zio);
}
/*
* Free intent log blocks up to this tx.
*/
zil_sync(os->os_zil, tx);
os->os_phys->os_zil_header = os->os_zil_header;
zio_nowait(zio);
}
/* called from dsl */
void
dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
{
int txgoff;
zbookmark_t zb;
writeprops_t wp = { 0 };
zio_t *zio;
list_t *list;
list_t *newlist = NULL;
dbuf_dirty_record_t *dr;
dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
ASSERT(dmu_tx_is_syncing(tx));
/* XXX the write_done callback should really give us the tx... */
os->os_synctx = tx;
if (os->os_dsl_dataset == NULL) {
/*
* This is the MOS. If we have upgraded,
* spa_max_replication() could change, so reset
* os_copies here.
*/
os->os_copies = spa_max_replication(os->os_spa);
}
/*
* Create the root block IO
*/
zb.zb_objset = os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : 0;
zb.zb_object = 0;
zb.zb_level = -1; /* for block ordering; it's level 0 on disk */
zb.zb_blkid = 0;
wp.wp_type = DMU_OT_OBJSET;
wp.wp_level = 0; /* on-disk BP level; see above */
wp.wp_copies = os->os_copies;
wp.wp_oschecksum = os->os_checksum;
wp.wp_oscompress = os->os_compress;
if (BP_IS_OLDER(os->os_rootbp, tx->tx_txg)) {
(void) dsl_dataset_block_kill(os->os_dsl_dataset,
os->os_rootbp, pio, tx);
}
arc_release(os->os_phys_buf, &os->os_phys_buf);
zio = arc_write(pio, os->os_spa, &wp, DMU_OS_IS_L2CACHEABLE(os),
tx->tx_txg, os->os_rootbp, os->os_phys_buf, ready, NULL, os,
ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
/*
* Sync special dnodes - the parent IO for the sync is the root block
*/
os->os_meta_dnode->dn_zio = zio;
dnode_sync(os->os_meta_dnode, tx);
os->os_phys->os_flags = os->os_flags;
if (os->os_userused_dnode &&
os->os_userused_dnode->dn_type != DMU_OT_NONE) {
os->os_userused_dnode->dn_zio = zio;
dnode_sync(os->os_userused_dnode, tx);
os->os_groupused_dnode->dn_zio = zio;
dnode_sync(os->os_groupused_dnode, tx);
}
txgoff = tx->tx_txg & TXG_MASK;
if (dmu_objset_userused_enabled(os)) {
newlist = &os->os_synced_dnodes;
/*
* We must create the list here because it uses the
* dn_dirty_link[] of this txg.
*/
list_create(newlist, sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[txgoff]));
}
dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx);
dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx);
list = &os->os_meta_dnode->dn_dirty_records[txgoff];
while ((dr = list_head(list))) {
ASSERT(dr->dr_dbuf->db_level == 0);
list_remove(list, dr);
if (dr->dr_zio)
zio_nowait(dr->dr_zio);
}
/*
* Free intent log blocks up to this tx.
*/
zil_sync(os->os_zil, tx);
os->os_phys->os_zil_header = os->os_zil_header;
zio_nowait(zio);
}
