diff options
author | Benjamin Marzinski <bmarzins@redhat.com> | 2014-11-13 21:42:04 -0500 |
---|---|---|
committer | Steven Whitehouse <swhiteho@redhat.com> | 2014-11-17 05:36:39 -0500 |
commit | 2e60d7683c8d2ea21317f6d9f4cd3bf5428ce162 (patch) | |
tree | fc7900ad18814d1ea46879f93eab063ff5754d1e /fs/gfs2/super.c | |
parent | 48b6bca6b7b8309697fc8a101793befe92d249d9 (diff) |
GFS2: update freeze code to use freeze/thaw_super on all nodes
The current gfs2 freezing code is considerably more complicated than it
should be because it doesn't use the vfs freezing code on any node except
the one that begins the freeze. This is because it needs to acquire a
cluster glock before calling the vfs code to prevent a deadlock, and
without the new freeze_super and thaw_super hooks, that was impossible. To
deal with the issue, gfs2 had to do some hacky locking tricks to make sure
that a frozen node couldn't be holding on a lock it needed to do the
unfreeze ioctl.
This patch makes use of the new hooks to simply the gfs2 locking code. Now,
all the nodes in the cluster freeze and thaw in exactly the same way. Every
node in the cluster caches the freeze glock in the shared state. The new
freeze_super hook allows the freezing node to grab this freeze glock in
the exclusive state without first calling the vfs freeze_super function.
All the nodes in the cluster see this lock change, and call the vfs
freeze_super function. The vfs locking code guarantees that the nodes can't
get stuck holding the glocks necessary to unfreeze the system. To
unfreeze, the freezing node uses the new thaw_super hook to drop the freeze
glock. Again, all the nodes notice this, reacquire the glock in shared mode
and call the vfs thaw_super function.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Diffstat (limited to 'fs/gfs2/super.c')
-rw-r--r-- | fs/gfs2/super.c | 112 |
1 files changed, 75 insertions, 37 deletions
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c index a346f56c4c6d..5b327f837de7 100644 --- a/fs/gfs2/super.c +++ b/fs/gfs2/super.c | |||
@@ -26,6 +26,7 @@ | |||
26 | #include <linux/wait.h> | 26 | #include <linux/wait.h> |
27 | #include <linux/writeback.h> | 27 | #include <linux/writeback.h> |
28 | #include <linux/backing-dev.h> | 28 | #include <linux/backing-dev.h> |
29 | #include <linux/kernel.h> | ||
29 | 30 | ||
30 | #include "gfs2.h" | 31 | #include "gfs2.h" |
31 | #include "incore.h" | 32 | #include "incore.h" |
@@ -399,7 +400,7 @@ int gfs2_make_fs_rw(struct gfs2_sbd *sdp) | |||
399 | { | 400 | { |
400 | struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode); | 401 | struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
401 | struct gfs2_glock *j_gl = ip->i_gl; | 402 | struct gfs2_glock *j_gl = ip->i_gl; |
402 | struct gfs2_holder thaw_gh; | 403 | struct gfs2_holder freeze_gh; |
403 | struct gfs2_log_header_host head; | 404 | struct gfs2_log_header_host head; |
404 | int error; | 405 | int error; |
405 | 406 | ||
@@ -408,7 +409,7 @@ int gfs2_make_fs_rw(struct gfs2_sbd *sdp) | |||
408 | return error; | 409 | return error; |
409 | 410 | ||
410 | error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, 0, | 411 | error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, 0, |
411 | &thaw_gh); | 412 | &freeze_gh); |
412 | if (error) | 413 | if (error) |
413 | goto fail_threads; | 414 | goto fail_threads; |
414 | 415 | ||
@@ -434,13 +435,13 @@ int gfs2_make_fs_rw(struct gfs2_sbd *sdp) | |||
434 | 435 | ||
435 | set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); | 436 | set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); |
436 | 437 | ||
437 | gfs2_glock_dq_uninit(&thaw_gh); | 438 | gfs2_glock_dq_uninit(&freeze_gh); |
438 | 439 | ||
439 | return 0; | 440 | return 0; |
440 | 441 | ||
441 | fail: | 442 | fail: |
442 | thaw_gh.gh_flags |= GL_NOCACHE; | 443 | freeze_gh.gh_flags |= GL_NOCACHE; |
443 | gfs2_glock_dq_uninit(&thaw_gh); | 444 | gfs2_glock_dq_uninit(&freeze_gh); |
444 | fail_threads: | 445 | fail_threads: |
445 | kthread_stop(sdp->sd_quotad_process); | 446 | kthread_stop(sdp->sd_quotad_process); |
446 | kthread_stop(sdp->sd_logd_process); | 447 | kthread_stop(sdp->sd_logd_process); |
@@ -580,14 +581,15 @@ int gfs2_statfs_sync(struct super_block *sb, int type) | |||
580 | struct buffer_head *m_bh, *l_bh; | 581 | struct buffer_head *m_bh, *l_bh; |
581 | int error; | 582 | int error; |
582 | 583 | ||
584 | sb_start_write(sb); | ||
583 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE, | 585 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE, |
584 | &gh); | 586 | &gh); |
585 | if (error) | 587 | if (error) |
586 | return error; | 588 | goto out; |
587 | 589 | ||
588 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); | 590 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); |
589 | if (error) | 591 | if (error) |
590 | goto out; | 592 | goto out_unlock; |
591 | 593 | ||
592 | spin_lock(&sdp->sd_statfs_spin); | 594 | spin_lock(&sdp->sd_statfs_spin); |
593 | gfs2_statfs_change_in(m_sc, m_bh->b_data + | 595 | gfs2_statfs_change_in(m_sc, m_bh->b_data + |
@@ -615,8 +617,10 @@ out_bh2: | |||
615 | brelse(l_bh); | 617 | brelse(l_bh); |
616 | out_bh: | 618 | out_bh: |
617 | brelse(m_bh); | 619 | brelse(m_bh); |
618 | out: | 620 | out_unlock: |
619 | gfs2_glock_dq_uninit(&gh); | 621 | gfs2_glock_dq_uninit(&gh); |
622 | out: | ||
623 | sb_end_write(sb); | ||
620 | return error; | 624 | return error; |
621 | } | 625 | } |
622 | 626 | ||
@@ -643,14 +647,8 @@ static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp, | |||
643 | struct lfcc *lfcc; | 647 | struct lfcc *lfcc; |
644 | LIST_HEAD(list); | 648 | LIST_HEAD(list); |
645 | struct gfs2_log_header_host lh; | 649 | struct gfs2_log_header_host lh; |
646 | struct gfs2_inode *dip = GFS2_I(sdp->sd_root_dir->d_inode); | ||
647 | int error; | 650 | int error; |
648 | 651 | ||
649 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, | ||
650 | &sdp->sd_freeze_root_gh); | ||
651 | if (error) | ||
652 | return error; | ||
653 | atomic_set(&sdp->sd_frozen_root, 1); | ||
654 | list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { | 652 | list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { |
655 | lfcc = kmalloc(sizeof(struct lfcc), GFP_KERNEL); | 653 | lfcc = kmalloc(sizeof(struct lfcc), GFP_KERNEL); |
656 | if (!lfcc) { | 654 | if (!lfcc) { |
@@ -692,11 +690,6 @@ out: | |||
692 | gfs2_glock_dq_uninit(&lfcc->gh); | 690 | gfs2_glock_dq_uninit(&lfcc->gh); |
693 | kfree(lfcc); | 691 | kfree(lfcc); |
694 | } | 692 | } |
695 | if (error) { | ||
696 | atomic_dec(&sdp->sd_frozen_root); | ||
697 | wait_event(sdp->sd_frozen_root_wait, atomic_read(&sdp->sd_frozen_root) == 0); | ||
698 | gfs2_glock_dq_uninit(&sdp->sd_freeze_root_gh); | ||
699 | } | ||
700 | return error; | 693 | return error; |
701 | } | 694 | } |
702 | 695 | ||
@@ -834,18 +827,14 @@ out: | |||
834 | 827 | ||
835 | static int gfs2_make_fs_ro(struct gfs2_sbd *sdp) | 828 | static int gfs2_make_fs_ro(struct gfs2_sbd *sdp) |
836 | { | 829 | { |
837 | struct gfs2_holder thaw_gh; | 830 | struct gfs2_holder freeze_gh; |
838 | int error; | 831 | int error; |
839 | 832 | ||
840 | error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, GL_NOCACHE, | 833 | error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, GL_NOCACHE, |
841 | &thaw_gh); | 834 | &freeze_gh); |
842 | if (error && !test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) | 835 | if (error && !test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) |
843 | return error; | 836 | return error; |
844 | 837 | ||
845 | down_write(&sdp->sd_log_flush_lock); | ||
846 | clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); | ||
847 | up_write(&sdp->sd_log_flush_lock); | ||
848 | |||
849 | kthread_stop(sdp->sd_quotad_process); | 838 | kthread_stop(sdp->sd_quotad_process); |
850 | kthread_stop(sdp->sd_logd_process); | 839 | kthread_stop(sdp->sd_logd_process); |
851 | 840 | ||
@@ -853,11 +842,16 @@ static int gfs2_make_fs_ro(struct gfs2_sbd *sdp) | |||
853 | gfs2_quota_sync(sdp->sd_vfs, 0); | 842 | gfs2_quota_sync(sdp->sd_vfs, 0); |
854 | gfs2_statfs_sync(sdp->sd_vfs, 0); | 843 | gfs2_statfs_sync(sdp->sd_vfs, 0); |
855 | 844 | ||
845 | down_write(&sdp->sd_log_flush_lock); | ||
846 | clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); | ||
847 | up_write(&sdp->sd_log_flush_lock); | ||
848 | |||
856 | gfs2_log_flush(sdp, NULL, SHUTDOWN_FLUSH); | 849 | gfs2_log_flush(sdp, NULL, SHUTDOWN_FLUSH); |
850 | wait_event(sdp->sd_reserving_log_wait, atomic_read(&sdp->sd_reserving_log) == 0); | ||
857 | gfs2_assert_warn(sdp, atomic_read(&sdp->sd_log_blks_free) == sdp->sd_jdesc->jd_blocks); | 851 | gfs2_assert_warn(sdp, atomic_read(&sdp->sd_log_blks_free) == sdp->sd_jdesc->jd_blocks); |
858 | 852 | ||
859 | if (thaw_gh.gh_gl) | 853 | if (freeze_gh.gh_gl) |
860 | gfs2_glock_dq_uninit(&thaw_gh); | 854 | gfs2_glock_dq_uninit(&freeze_gh); |
861 | 855 | ||
862 | gfs2_quota_cleanup(sdp); | 856 | gfs2_quota_cleanup(sdp); |
863 | 857 | ||
@@ -943,11 +937,41 @@ static int gfs2_sync_fs(struct super_block *sb, int wait) | |||
943 | struct gfs2_sbd *sdp = sb->s_fs_info; | 937 | struct gfs2_sbd *sdp = sb->s_fs_info; |
944 | 938 | ||
945 | gfs2_quota_sync(sb, -1); | 939 | gfs2_quota_sync(sb, -1); |
946 | if (wait && sdp && !atomic_read(&sdp->sd_log_freeze)) | 940 | if (wait && sdp) |
947 | gfs2_log_flush(sdp, NULL, NORMAL_FLUSH); | 941 | gfs2_log_flush(sdp, NULL, NORMAL_FLUSH); |
948 | return 0; | 942 | return 0; |
949 | } | 943 | } |
950 | 944 | ||
945 | void gfs2_freeze_func(struct work_struct *work) | ||
946 | { | ||
947 | int error; | ||
948 | struct gfs2_holder freeze_gh; | ||
949 | struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_freeze_work); | ||
950 | struct super_block *sb = sdp->sd_vfs; | ||
951 | |||
952 | atomic_inc(&sb->s_active); | ||
953 | error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, 0, | ||
954 | &freeze_gh); | ||
955 | if (error) { | ||
956 | printk(KERN_INFO "GFS2: couln't get freeze lock : %d\n", error); | ||
957 | gfs2_assert_withdraw(sdp, 0); | ||
958 | } | ||
959 | else { | ||
960 | atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN); | ||
961 | error = thaw_super(sb); | ||
962 | if (error) { | ||
963 | printk(KERN_INFO "GFS2: couldn't thaw filesystem: %d\n", | ||
964 | error); | ||
965 | gfs2_assert_withdraw(sdp, 0); | ||
966 | } | ||
967 | if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) | ||
968 | freeze_gh.gh_flags |= GL_NOCACHE; | ||
969 | gfs2_glock_dq_uninit(&freeze_gh); | ||
970 | } | ||
971 | deactivate_super(sb); | ||
972 | return; | ||
973 | } | ||
974 | |||
951 | /** | 975 | /** |
952 | * gfs2_freeze - prevent further writes to the filesystem | 976 | * gfs2_freeze - prevent further writes to the filesystem |
953 | * @sb: the VFS structure for the filesystem | 977 | * @sb: the VFS structure for the filesystem |
@@ -957,10 +981,16 @@ static int gfs2_sync_fs(struct super_block *sb, int wait) | |||
957 | static int gfs2_freeze(struct super_block *sb) | 981 | static int gfs2_freeze(struct super_block *sb) |
958 | { | 982 | { |
959 | struct gfs2_sbd *sdp = sb->s_fs_info; | 983 | struct gfs2_sbd *sdp = sb->s_fs_info; |
960 | int error; | 984 | int error = 0; |
961 | 985 | ||
962 | if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) | 986 | mutex_lock(&sdp->sd_freeze_mutex); |
963 | return -EINVAL; | 987 | if (atomic_read(&sdp->sd_freeze_state) != SFS_UNFROZEN) |
988 | goto out; | ||
989 | |||
990 | if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) { | ||
991 | error = -EINVAL; | ||
992 | goto out; | ||
993 | } | ||
964 | 994 | ||
965 | for (;;) { | 995 | for (;;) { |
966 | error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh); | 996 | error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh); |
@@ -980,7 +1010,10 @@ static int gfs2_freeze(struct super_block *sb) | |||
980 | fs_err(sdp, "retrying...\n"); | 1010 | fs_err(sdp, "retrying...\n"); |
981 | msleep(1000); | 1011 | msleep(1000); |
982 | } | 1012 | } |
983 | return 0; | 1013 | error = 0; |
1014 | out: | ||
1015 | mutex_unlock(&sdp->sd_freeze_mutex); | ||
1016 | return error; | ||
984 | } | 1017 | } |
985 | 1018 | ||
986 | /** | 1019 | /** |
@@ -993,10 +1026,15 @@ static int gfs2_unfreeze(struct super_block *sb) | |||
993 | { | 1026 | { |
994 | struct gfs2_sbd *sdp = sb->s_fs_info; | 1027 | struct gfs2_sbd *sdp = sb->s_fs_info; |
995 | 1028 | ||
1029 | mutex_lock(&sdp->sd_freeze_mutex); | ||
1030 | if (atomic_read(&sdp->sd_freeze_state) != SFS_FROZEN || | ||
1031 | sdp->sd_freeze_gh.gh_gl == NULL) { | ||
1032 | mutex_unlock(&sdp->sd_freeze_mutex); | ||
1033 | return 0; | ||
1034 | } | ||
1035 | |||
996 | gfs2_glock_dq_uninit(&sdp->sd_freeze_gh); | 1036 | gfs2_glock_dq_uninit(&sdp->sd_freeze_gh); |
997 | atomic_dec(&sdp->sd_frozen_root); | 1037 | mutex_unlock(&sdp->sd_freeze_mutex); |
998 | wait_event(sdp->sd_frozen_root_wait, atomic_read(&sdp->sd_frozen_root) == 0); | ||
999 | gfs2_glock_dq_uninit(&sdp->sd_freeze_root_gh); | ||
1000 | return 0; | 1038 | return 0; |
1001 | } | 1039 | } |
1002 | 1040 | ||
@@ -1618,8 +1656,8 @@ const struct super_operations gfs2_super_ops = { | |||
1618 | .evict_inode = gfs2_evict_inode, | 1656 | .evict_inode = gfs2_evict_inode, |
1619 | .put_super = gfs2_put_super, | 1657 | .put_super = gfs2_put_super, |
1620 | .sync_fs = gfs2_sync_fs, | 1658 | .sync_fs = gfs2_sync_fs, |
1621 | .freeze_fs = gfs2_freeze, | 1659 | .freeze_super = gfs2_freeze, |
1622 | .unfreeze_fs = gfs2_unfreeze, | 1660 | .thaw_super = gfs2_unfreeze, |
1623 | .statfs = gfs2_statfs, | 1661 | .statfs = gfs2_statfs, |
1624 | .remount_fs = gfs2_remount_fs, | 1662 | .remount_fs = gfs2_remount_fs, |
1625 | .drop_inode = gfs2_drop_inode, | 1663 | .drop_inode = gfs2_drop_inode, |