]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/log
mirror_ubuntu-eoan-kernel.git
7 years agobtrfs: Make btrfs_extent_item_to_extent_map take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:51:02 +0000 (13:51 +0200)]
btrfs: Make btrfs_extent_item_to_extent_map take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_log_inode_parent take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:51:01 +0000 (13:51 +0200)]
btrfs: make btrfs_log_inode_parent take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make check_parent_dirs_for_sync take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:51:00 +0000 (13:51 +0200)]
btrfs: Make check_parent_dirs_for_sync take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_orphan_add take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:59 +0000 (13:50 +0200)]
btrfs: Make btrfs_orphan_add take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_orphan_del take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:58 +0000 (13:50 +0200)]
btrfs: make btrfs_orphan_del take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_free_io_failure_record take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:57 +0000 (13:50 +0200)]
btrfs: make btrfs_free_io_failure_record take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make clean_io_failure take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:56 +0000 (13:50 +0200)]
btrfs: make clean_io_failure take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make repair_io_failure take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:55 +0000 (13:50 +0200)]
btrfs: make repair_io_failure take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make check_compressed_csum take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:54 +0000 (13:50 +0200)]
btrfs: make check_compressed_csum take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_print_data_csum_error take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:53 +0000 (13:50 +0200)]
btrfs: make btrfs_print_data_csum_error take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make free_io_failure take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:52 +0000 (13:50 +0200)]
btrfs: make free_io_failure take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make lock_and_cleanup_extent_if_need take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:51 +0000 (13:50 +0200)]
btrfs: Make lock_and_cleanup_extent_if_need take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make check_can_nocow take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:50 +0000 (13:50 +0200)]
btrfs: Make check_can_nocow take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_lookup_ordered_range take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:49 +0000 (13:50 +0200)]
btrfs: Make btrfs_lookup_ordered_range take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_mark_extent_written take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:48 +0000 (13:50 +0200)]
btrfs: Make btrfs_mark_extent_written take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make fill_holes take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:47 +0000 (13:50 +0200)]
btrfs: Make fill_holes take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make hole_mergeable take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:46 +0000 (13:50 +0200)]
btrfs: Make hole_mergeable take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_drop_extent_cache take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:45 +0000 (13:50 +0200)]
btrfs: Make btrfs_drop_extent_cache take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_requeue_inode_defrag take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:44 +0000 (13:50 +0200)]
btrfs: Make btrfs_requeue_inode_defrag take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make (__)btrfs_add_inode_defrag take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:43 +0000 (13:50 +0200)]
btrfs: Make (__)btrfs_add_inode_defrag take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: all btrfs_delalloc_release_metadata take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:42 +0000 (13:50 +0200)]
btrfs: all btrfs_delalloc_release_metadata take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_delalloc_reserve_metadata take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:41 +0000 (13:50 +0200)]
btrfs: Make btrfs_delalloc_reserve_metadata take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_orphan_release_metadata take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:40 +0000 (13:50 +0200)]
btrfs: Make btrfs_orphan_release_metadata take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_orphan_reserve_metadata take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:39 +0000 (13:50 +0200)]
btrfs: Make btrfs_orphan_reserve_metadata take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make calc_csum_metadata_size take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:38 +0000 (13:50 +0200)]
btrfs: Make calc_csum_metadata_size take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make drop_outstanding_extent take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:37 +0000 (13:50 +0200)]
btrfs: Make drop_outstanding_extent take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_alloc_data_chunk_ondemand take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:36 +0000 (13:50 +0200)]
btrfs: make btrfs_alloc_data_chunk_ondemand take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_is_free_space_inode take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:35 +0000 (13:50 +0200)]
btrfs: make btrfs_is_free_space_inode take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_i_size_write take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:34 +0000 (13:50 +0200)]
btrfs: Make btrfs_i_size_write take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_set_inode_index take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:33 +0000 (13:50 +0200)]
btrfs: Make btrfs_set_inode_index take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make btrfs_set_inode_index_count take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:32 +0000 (13:50 +0200)]
btrfs: make btrfs_set_inode_index_count take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_insert_dir_item take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:31 +0000 (13:50 +0200)]
btrfs: Make btrfs_insert_dir_item take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Make btrfs_log_all_parents take btrfs_inode
Nikolay Borisov [Mon, 20 Feb 2017 11:50:30 +0000 (13:50 +0200)]
btrfs: Make btrfs_log_all_parents take btrfs_inode

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: use the correct type when creating cow dio extent
Liu Bo [Tue, 21 Feb 2017 20:12:58 +0000 (12:12 -0800)]
Btrfs: use the correct type when creating cow dio extent

'BTRFS_ORDERED_REGULAR' was introduced for the cow case in patch
'Btrfs: specify a new ordered extent type for create_io_em',
but it missed the directIO cow case.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Chris Mason <clm@fb.com>
7 years agoBtrfs: fix deadlock between dedup on same file and starting writeback
Filipe Manana [Tue, 21 Feb 2017 17:14:52 +0000 (17:14 +0000)]
Btrfs: fix deadlock between dedup on same file and starting writeback

If we are deduping two ranges of the same file we need to make sure that
we lock all pages in ascending order, that is, lock first the pages from
the range with lower offset and then the pages from the other range, as
otherwise we can deadlock with a concurrent task that is starting delalloc
(writeback). Example trace:

[74073.052218] INFO: task kworker/u32:10:17997 blocked for more than 120 seconds.
[74073.053889]       Tainted: G        W       4.9.0-rc7-btrfs-next-36+ #1
[74073.055071] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[74073.056696] kworker/u32:10  D    0 17997      2 0x00000000
[74073.058606] Workqueue: writeback wb_workfn (flush-btrfs-53176)
[74073.061370]  ffff880031e79858 ffff8802159d2580 ffff880237004580 ffff880031e79240
[74073.064784]  ffff88023f4978c0 ffffc9000817b638 ffffffff814c15e1 0000000000000000
[74073.068386]  ffff88023f4978d8 ffff88023f4978c0 000000000017b620 ffff880031e79240
[74073.071712] Call Trace:
[74073.072884]  [<ffffffff814c15e1>] ? __schedule+0x48f/0x6f4
[74073.075395]  [<ffffffff814c1c8b>] ? bit_wait+0x2f/0x2f
[74073.077511]  [<ffffffff814c18d2>] schedule+0x8c/0xa0
[74073.079440]  [<ffffffff814c4b36>] schedule_timeout+0x43/0xff
[74073.081637]  [<ffffffff8110953e>] ? time_hardirqs_on+0x9/0x14
[74073.083809]  [<ffffffff81095c67>] ? trace_hardirqs_on_caller+0x16/0x197
[74073.086314]  [<ffffffff810bde98>] ? timekeeping_get_ns+0x1e/0x32
[74073.100654]  [<ffffffff810be048>] ? ktime_get+0x41/0x52
[74073.102619]  [<ffffffff814c10f0>] io_schedule_timeout+0xa0/0x102
[74073.104771]  [<ffffffff814c10f0>] ? io_schedule_timeout+0xa0/0x102
[74073.106969]  [<ffffffff814c1ca6>] bit_wait_io+0x1b/0x39
[74073.108954]  [<ffffffff814c1fb8>] __wait_on_bit_lock+0x4f/0x99
[74073.110981]  [<ffffffff8112b692>] __lock_page+0x6b/0x6d
[74073.112833]  [<ffffffff8108ceb4>] ? autoremove_wake_function+0x3a/0x3a
[74073.115010]  [<ffffffffa031178b>] lock_page+0x2f/0x32 [btrfs]
[74073.116999]  [<ffffffffa0311d9f>] lock_delalloc_pages+0xc7/0x1a0 [btrfs]
[74073.119243]  [<ffffffffa0313d15>] find_lock_delalloc_range+0xc3/0x1a4 [btrfs]
[74073.121636]  [<ffffffffa0313e81>] writepage_delalloc.isra.31+0x8b/0x134 [btrfs]
[74073.124229]  [<ffffffffa0315d69>] __extent_writepage+0x1c1/0x2bf [btrfs]
[74073.126372]  [<ffffffffa03160f2>] extent_write_cache_pages.isra.30.constprop.49+0x28b/0x36c [btrfs]
[74073.129371]  [<ffffffffa03165b9>] extent_writepages+0x4b/0x5c [btrfs]
[74073.131440]  [<ffffffffa02fcb59>] ? insert_reserved_file_extent.constprop.42+0x261/0x261 [btrfs]
[74073.134303]  [<ffffffff811b4ce4>] ? writeback_sb_inodes+0xe0/0x4a1
[74073.136298]  [<ffffffffa02fab7f>] btrfs_writepages+0x28/0x2a [btrfs]
[74073.138248]  [<ffffffff81138200>] do_writepages+0x23/0x2c
[74073.139910]  [<ffffffff811b3cab>] __writeback_single_inode+0x105/0x6d2
[74073.142003]  [<ffffffff811b4e96>] writeback_sb_inodes+0x292/0x4a1
[74073.136298]  [<ffffffffa02fab7f>] btrfs_writepages+0x28/0x2a [btrfs]
[74073.138248]  [<ffffffff81138200>] do_writepages+0x23/0x2c
[74073.139910]  [<ffffffff811b3cab>] __writeback_single_inode+0x105/0x6d2
[74073.142003]  [<ffffffff811b4e96>] writeback_sb_inodes+0x292/0x4a1
[74073.143911]  [<ffffffff811b511b>] __writeback_inodes_wb+0x76/0xae
[74073.145787]  [<ffffffff811b53ca>] wb_writeback+0x1cc/0x4d7
[74073.147452]  [<ffffffff811b60cd>] wb_workfn+0x194/0x37d
[74073.149084]  [<ffffffff811b60cd>] ? wb_workfn+0x194/0x37d
[74073.150726]  [<ffffffff8106ce77>] ? process_one_work+0x154/0x4e4
[74073.152694]  [<ffffffff8106cf96>] process_one_work+0x273/0x4e4
[74073.154452]  [<ffffffff8106d6db>] worker_thread+0x1eb/0x2ca
[74073.156138]  [<ffffffff8106d4f0>] ? rescuer_thread+0x2b6/0x2b6
[74073.157837]  [<ffffffff81072a81>] kthread+0xd5/0xdd
[74073.159339]  [<ffffffff810729ac>] ? __kthread_unpark+0x5a/0x5a
[74073.161088]  [<ffffffff814c6257>] ret_from_fork+0x27/0x40
[74073.162680] INFO: lockdep is turned off.
[74073.163855] INFO: task do-dedup:30264 blocked for more than 120 seconds.
[74073.181180]       Tainted: G        W       4.9.0-rc7-btrfs-next-36+ #1
[74073.181180] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[74073.185296] fdm-stress      D    0 30264  29974 0x00000000
[74073.186810]  ffff880089595118 ffff880211b8eac0 ffff880237030380 ffff880089594b00
[74073.188998]  ffff88023f2978c0 ffffc900063abb68 ffffffff814c15e1 0000000000000000
[74073.191070]  ffff88023f2978d8 ffff88023f2978c0 00000000003abb50 ffff880089594b00
[74073.193286] Call Trace:
[74073.193990]  [<ffffffff814c15e1>] ? __schedule+0x48f/0x6f4
[74073.195418]  [<ffffffff814c1c8b>] ? bit_wait+0x2f/0x2f
[74073.196796]  [<ffffffff814c18d2>] schedule+0x8c/0xa0
[74073.198163]  [<ffffffff814c4b36>] schedule_timeout+0x43/0xff
[74073.199621]  [<ffffffff81095df5>] ? trace_hardirqs_on+0xd/0xf
[74073.201100]  [<ffffffff810bde98>] ? timekeeping_get_ns+0x1e/0x32
[74073.202686]  [<ffffffff810be048>] ? ktime_get+0x41/0x52
[74073.204051]  [<ffffffff814c10f0>] io_schedule_timeout+0xa0/0x102
[74073.205585]  [<ffffffff814c10f0>] ? io_schedule_timeout+0xa0/0x102
[74073.207123]  [<ffffffff814c1ca6>] bit_wait_io+0x1b/0x39
[74073.208238]  [<ffffffff814c1fb8>] __wait_on_bit_lock+0x4f/0x99
[74073.208871]  [<ffffffff8112b692>] __lock_page+0x6b/0x6d
[74073.209430]  [<ffffffff8108ceb4>] ? autoremove_wake_function+0x3a/0x3a
[74073.210101]  [<ffffffff8112b800>] lock_page+0x2f/0x32
[74073.210636]  [<ffffffff8112c502>] pagecache_get_page+0x5e/0x153
[74073.211270]  [<ffffffffa03257eb>] gather_extent_pages+0x4e/0x109 [btrfs]
[74073.212166]  [<ffffffffa032a04c>] btrfs_dedupe_file_range+0x1e1/0x4dd [btrfs]
[74073.213257]  [<ffffffff8118d9b5>] vfs_dedupe_file_range+0x1c1/0x221
[74073.214086]  [<ffffffff8119e0c4>] do_vfs_ioctl+0x442/0x600
[74073.214767]  [<ffffffff811a7874>] ? rcu_read_unlock+0x5b/0x5d
[74073.215619]  [<ffffffff811a7953>] ? __fget+0x6b/0x77
[74073.216338]  [<ffffffff8119e2d9>] SyS_ioctl+0x57/0x79
[74073.217149]  [<ffffffff814c5fea>] entry_SYSCALL_64_fastpath+0x18/0xad
[74073.218102]  [<ffffffff81109552>] ? time_hardirqs_off+0x9/0x14
[74073.218968]  [<ffffffff810938ce>] ? trace_hardirqs_off_caller+0x1f/0xaa
[74073.219938] INFO: lockdep is turned off.

What happened was the following:

      CPU 1                                       CPU 2

                                             btrfs_dedupe_file_range()
                                               --> using same inode as source
                                                   and target
                                               --> src range is [768K, 1Mb[
                                               --> dst range is [0, 256K[
                                              btrfs_cmp_data_prepare()
                                               --> calls gather_extent_pages()
                                                   for range [768K, 1Mb[ and
                                                   locks all pages in that range

 do_writepages()
  btrfs_writepages()
   extent_writepages()
    extent_write_cache_pages()
     __extent_writepage()
      writepage_delalloc()
       find_lock_delalloc_range()
         --> finds range [0, 1Mb[
         lock_delalloc_pages()
          --> locks all pages in the
              range [0, 768K[
          --> tries to lock page at
              offset 768K
                --> deadlock

                                               --> calls gather_extent_pages()
                                                   to lock pages in the range
                                                   [0, 256K[
                                                    --> deadlock, task at CPU 1
                                                        already locked that
                                                        range and it's trying
                                                        to lock the range we
                                                        locked previously

So fix this by making sure that during a dedup we always lock first the
pages from the range with lower offset.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Chris Mason <clm@fb.com>
7 years agobtrfs: use btrfs_debug instead of pr_debug in transaction abort
Jeff Mahoney [Wed, 15 Feb 2017 21:28:34 +0000 (16:28 -0500)]
btrfs: use btrfs_debug instead of pr_debug in transaction abort

Commit e5d6b12fe14 (Btrfs: don't WARN() in btrfs_transaction_abort() for
IO errors) added a pr_debug call to be printed when a transaction is
aborted with -EIO instead of WARN.  btrfs_debug prints which file system
the message is associated with so let's use that instead.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: btrfs_truncate_free_space_cache always allocates path
Jeff Mahoney [Wed, 15 Feb 2017 21:28:32 +0000 (16:28 -0500)]
btrfs: btrfs_truncate_free_space_cache always allocates path

btrfs_truncate_free_space_cache always allocates a btrfs_path structure
but only uses it when the caller passes a block group.  Let's move the
allocation and free into the conditional.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: free-space-cache, clean up unnecessary root arguments
Jeff Mahoney [Wed, 15 Feb 2017 21:28:30 +0000 (16:28 -0500)]
btrfs: free-space-cache, clean up unnecessary root arguments

The free space cache APIs accept a root but always use the tree root.

Also, btrfs_truncate_free_space_cache accepts a root AND an inode but
the inode always points to the root anyway, so let's just pass the inode.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: convert btrfs_inc_block_group_ro to accept fs_info
Jeff Mahoney [Wed, 15 Feb 2017 21:28:29 +0000 (16:28 -0500)]
btrfs: convert btrfs_inc_block_group_ro to accept fs_info

btrfs_inc_block_group_ro is either passed the extent root or the dev
root, but it doesn't do anything with the dev tree.  Let's convert
to passing an fs_info and using the extent root.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: flush_space always takes fs_info->fs_root
Jeff Mahoney [Wed, 15 Feb 2017 21:28:28 +0000 (16:28 -0500)]
btrfs: flush_space always takes fs_info->fs_root

We don't need to pass a root to flush_space since it always uses
the fs_root.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: pass fs_info to (more) routines that are only called with extent_root
Jeff Mahoney [Wed, 15 Feb 2017 21:28:27 +0000 (16:28 -0500)]
btrfs: pass fs_info to (more) routines that are only called with extent_root

Outside of interactions with qgroups, the roots passed in extent-tree.c
are usually passed to ensure that we don't do refcounts on log trees or
to get the allocation profile for an allocation request.  Otherwise, it
operates on the extent root.  This patch converts some more routines in
extent-tree.c that are always called with the extent root to accept
an fs_info instead.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: qgroup: Move half of the qgroup accounting time out of commit trans
Qu Wenruo [Wed, 15 Feb 2017 02:43:03 +0000 (10:43 +0800)]
btrfs: qgroup: Move half of the qgroup accounting time out of commit trans

Just as Filipe pointed out, the most time consuming parts of qgroup are
btrfs_qgroup_account_extents() and
btrfs_qgroup_prepare_account_extents().
Which both call btrfs_find_all_roots() to get old_roots and new_roots
ulist.

What makes things worse is, we're calling that expensive
btrfs_find_all_roots() at transaction committing time with
TRANS_STATE_COMMIT_DOING, which will blocks all incoming transaction.

Such behavior is necessary for @new_roots search as current
btrfs_find_all_roots() can't do it correctly so we do call it just
before switch commit roots.

However for @old_roots search, it's not necessary as such search is
based on commit_root, so it will always be correct and we can move it
out of transaction committing.

This patch moves the @old_roots search part out of
commit_transaction(), so in theory we can half the time qgroup time
consumption at commit_transaction().

But please note that, this won't speedup qgroup overall, the total time
consumption is still the same, just reduce the performance stall.

Cc: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from adjust_slots_upwards
David Sterba [Fri, 10 Feb 2017 19:30:23 +0000 (20:30 +0100)]
btrfs: remove unused parameter from adjust_slots_upwards

Never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameters from __btrfs_write_out_cache
David Sterba [Fri, 10 Feb 2017 19:26:24 +0000 (20:26 +0100)]
btrfs: remove unused parameters from __btrfs_write_out_cache

Both unused after the call to update_cache_item has been moved to
__btrfs_wait_cache_io.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from cleanup_write_cache_enospc
David Sterba [Fri, 10 Feb 2017 19:23:00 +0000 (20:23 +0100)]
btrfs: remove unused parameter from cleanup_write_cache_enospc

bitmap_list is unused since the io_ctl framework.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from __add_inode_ref
David Sterba [Fri, 10 Feb 2017 19:20:19 +0000 (20:20 +0100)]
btrfs: remove unused parameter from __add_inode_ref

Unused since the helper has been split, eb used in the caller.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from clone_copy_inline_extent
David Sterba [Fri, 10 Feb 2017 19:18:49 +0000 (20:18 +0100)]
btrfs: remove unused parameter from clone_copy_inline_extent

Never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameters from btrfs_cmp_data
David Sterba [Fri, 10 Feb 2017 19:15:10 +0000 (20:15 +0100)]
btrfs: remove unused parameters from btrfs_cmp_data

After the page locking has been reworked, we get all pages prepared via
cmp_pages.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from __add_inline_refs
David Sterba [Fri, 10 Feb 2017 18:57:27 +0000 (19:57 +0100)]
btrfs: remove unused parameter from __add_inline_refs

Never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameters from scrub_setup_wr_ctx
David Sterba [Fri, 10 Feb 2017 18:55:54 +0000 (19:55 +0100)]
btrfs: remove unused parameters from scrub_setup_wr_ctx

Never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from create_snapshot
David Sterba [Fri, 10 Feb 2017 18:54:06 +0000 (19:54 +0100)]
btrfs: remove unused parameter from create_snapshot

The name parameters have never been used, as the name is passed via the
dentry.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from init_first_rw_device
David Sterba [Fri, 10 Feb 2017 18:49:01 +0000 (19:49 +0100)]
btrfs: remove unused parameter from init_first_rw_device

The 'device' used to be added in that function, but now it's done by the
caller.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from __btrfs_alloc_chunk
David Sterba [Fri, 10 Feb 2017 18:46:27 +0000 (19:46 +0100)]
btrfs: remove unused parameter from __btrfs_alloc_chunk

We grab fs_info from other parameters.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from btrfs_fill_super
David Sterba [Fri, 10 Feb 2017 18:44:31 +0000 (19:44 +0100)]
btrfs: remove unused parameter from btrfs_fill_super

Never used for anything meaningful since we have our own superblock
filler.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from extent_write_cache_pages
David Sterba [Fri, 10 Feb 2017 18:38:24 +0000 (19:38 +0100)]
btrfs: remove unused parameter from extent_write_cache_pages

The 'tree' was used to call locking hook that does not exist anymore.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from add_pending_csums
David Sterba [Fri, 10 Feb 2017 18:35:37 +0000 (19:35 +0100)]
btrfs: remove unused parameter from add_pending_csums

Never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from update_nr_written
David Sterba [Fri, 10 Feb 2017 18:33:41 +0000 (19:33 +0100)]
btrfs: remove unused parameter from update_nr_written

The logic has been updated in "Btrfs: make mapping->writeback_index
point to the last written page" (a91326679f2a0a4c2) and page is not
needed anymore.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from submit_extent_page
David Sterba [Fri, 10 Feb 2017 18:29:38 +0000 (19:29 +0100)]
btrfs: remove unused parameter from submit_extent_page

This used to hold number of maximum pages to allocate, but this is now
limited by BIO_MAX_PAGES. The local are now unused and removed as well.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from tree_move_next_or_upnext
David Sterba [Fri, 10 Feb 2017 18:25:51 +0000 (19:25 +0100)]
btrfs: remove unused parameter from tree_move_next_or_upnext

Not needed.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from tree_move_down
David Sterba [Fri, 10 Feb 2017 18:24:53 +0000 (19:24 +0100)]
btrfs: remove unused parameter from tree_move_down

Never needed.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from btrfs_check_super_valid
David Sterba [Fri, 10 Feb 2017 18:23:20 +0000 (19:23 +0100)]
btrfs: remove unused parameter from btrfs_check_super_valid

None of the checks need to know the ro/rw status as they're all not
changing the superblock. Moreover, we can access the sb flags directly
if we'd need to decide by the ro/rw status.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from btrfs_prepare_extent_commit
David Sterba [Fri, 10 Feb 2017 18:20:56 +0000 (19:20 +0100)]
btrfs: remove unused parameter from btrfs_prepare_extent_commit

Added but never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from btrfs_subvolume_release_metadata
David Sterba [Fri, 10 Feb 2017 18:18:18 +0000 (19:18 +0100)]
btrfs: remove unused parameter from btrfs_subvolume_release_metadata

Unused since qgroup refactoring that split data and metadata accounting,
the btrfs_qgroup_free helper.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from __push_leaf_left
David Sterba [Fri, 10 Feb 2017 18:14:36 +0000 (19:14 +0100)]
btrfs: remove unused parameter from __push_leaf_left

Unused since long ago.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from __push_leaf_right
David Sterba [Fri, 10 Feb 2017 18:13:06 +0000 (19:13 +0100)]
btrfs: remove unused parameter from __push_leaf_right

Unused since long ago.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: merge two superblock writing helpers
David Sterba [Fri, 10 Feb 2017 18:04:32 +0000 (19:04 +0100)]
btrfs: merge two superblock writing helpers

write_all_supers and write_ctree_super are almost equal, the parameter
'trans' is unused so we can drop it and have just one helper.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from write_dev_supers
David Sterba [Fri, 10 Feb 2017 17:52:06 +0000 (18:52 +0100)]
btrfs: remove unused parameter from write_dev_supers

The barriers are handled by the caller.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from split_item
David Sterba [Fri, 10 Feb 2017 17:49:53 +0000 (18:49 +0100)]
btrfs: remove unused parameter from split_item

Never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from clean_tree_block
David Sterba [Fri, 10 Feb 2017 17:47:57 +0000 (18:47 +0100)]
btrfs: remove unused parameter from clean_tree_block

Added but never needed.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from check_async_write
David Sterba [Fri, 10 Feb 2017 17:46:09 +0000 (18:46 +0100)]
btrfs: remove unused parameter from check_async_write

Added but never used.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused parameter from read_block_for_search
David Sterba [Fri, 10 Feb 2017 17:44:32 +0000 (18:44 +0100)]
btrfs: remove unused parameter from read_block_for_search

Never used in that function.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: ulist: rename ulist_fini to ulist_release
David Sterba [Wed, 15 Feb 2017 15:47:36 +0000 (16:47 +0100)]
btrfs: ulist: rename ulist_fini to ulist_release

Change the name so it matches the naming we already use eg. for
btrfs_path.

Suggested-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove pointless rcu protection from btrfs_qgroup_inherit
David Sterba [Mon, 13 Feb 2017 14:05:33 +0000 (15:05 +0100)]
btrfs: remove pointless rcu protection from btrfs_qgroup_inherit

There was never need for RCU protection around reading nodesize or other
fairly constant filesystem data.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: qgroups: opencode qgroup_free helper
David Sterba [Mon, 13 Feb 2017 13:24:35 +0000 (14:24 +0100)]
btrfs: qgroups: opencode qgroup_free helper

The helper name is not too helpful and is just wrapping a simple call.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unnecessary mutex lock in qgroup_account_snapshot
David Sterba [Mon, 13 Feb 2017 13:07:02 +0000 (14:07 +0100)]
btrfs: remove unnecessary mutex lock in qgroup_account_snapshot

The quota status used to be tracked as a variable, so the mutex was
needed (until "Btrfs: add a flags field to btrfs_fs_info" afcdd129e05a9).
Since the status is a bit modified atomically and we don't hold the
mutex beyond the check, we can drop it.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: check quota status earlier and don't do unnecessary frees
David Sterba [Mon, 13 Feb 2017 13:05:24 +0000 (14:05 +0100)]
btrfs: check quota status earlier and don't do unnecessary frees

Status of quotas should be the first check in
btrfs_qgroup_account_extent and we can return immediatelly, no need to
do no-op ulist frees.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: embed extent_changeset::range_changed to the structure
David Sterba [Mon, 13 Feb 2017 12:42:29 +0000 (13:42 +0100)]
btrfs: embed extent_changeset::range_changed to the structure

We can embed range_changed to the extent changeset to address following
problems:

- no need to allocate ulist dynamically, we also get rid of the GFP_NOFS
  for free
- fix lack of allocation failure checking in btrfs_qgroup_reserve_data

The stack consuption where extent_changeset is used slightly increases:

before: 16
after: 16 - 8 (for pointer) + 32 (sizeof ulist) = 40

Which is bearable.

Reviewed-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: ulist: make the finalization function public
David Sterba [Mon, 13 Feb 2017 12:40:16 +0000 (13:40 +0100)]
btrfs: ulist: make the finalization function public

Make ulist_fini externally visible so the ulist API is complete.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: qgroups: make __del_qgroup_relation static
David Sterba [Mon, 13 Feb 2017 12:00:51 +0000 (13:00 +0100)]
btrfs: qgroups: make __del_qgroup_relation static

Internal helper.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: make space cache inode readahead failure nonfatal
David Sterba [Mon, 23 Jan 2017 16:28:19 +0000 (17:28 +0100)]
btrfs: make space cache inode readahead failure nonfatal

We do a readahead of the free space cache inode to speed things up but
the failure is not fatal, like in other readahead cases. Proper reads
would need to happen anyway and any errors would be caught there.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: use GFP_KERNEL in btrfs_add/del_qgroup_relation
David Sterba [Mon, 13 Feb 2017 11:41:02 +0000 (12:41 +0100)]
btrfs: use GFP_KERNEL in btrfs_add/del_qgroup_relation

Qgroup relations are added/deleted from ioctl, we hold the high level
qgroup lock, no deadlocks or recursion from the allocation possible
here.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: use GFP_KERNEL in btrfs_quota_enable
David Sterba [Mon, 13 Feb 2017 10:03:44 +0000 (11:03 +0100)]
btrfs: use GFP_KERNEL in btrfs_quota_enable

We don't need to use GFP_NOFS here as this is called from ioctls an the
only lock held is the subvol_sem, which is of a high level and protects
creation/renames/deletion and is never held in the writeout paths.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: use GFP_KERNEL in btrfs_read_qgroup_config
David Sterba [Mon, 13 Feb 2017 11:10:20 +0000 (12:10 +0100)]
btrfs: use GFP_KERNEL in btrfs_read_qgroup_config

The qgroup config is read during mount, we do not have to use NOFS.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: use GFP_KERNEL in create_snapshot
David Sterba [Mon, 13 Feb 2017 10:03:44 +0000 (11:03 +0100)]
btrfs: use GFP_KERNEL in create_snapshot

We don't need to use GFP_NOFS here as this is called from ioctls an the
only lock held is the subvol_sem, which is of a high level and protects
creation/renames/deletion and is never held in the writeout paths.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: specify a new ordered extent type for create_io_em
Liu Bo [Mon, 13 Feb 2017 23:35:09 +0000 (15:35 -0800)]
Btrfs: specify a new ordered extent type for create_io_em

As 0 refers to an existing type BTRFS_ORDERED_IO_DONE, this specifies a
new type 'REGULAR' for regular IO.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: create a helper to create em for IO
Liu Bo [Tue, 31 Jan 2017 15:50:22 +0000 (07:50 -0800)]
Btrfs: create a helper to create em for IO

We have similar codes to create and insert extent mapping around IO path,
this merges them into a single helper.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: use helper to get used bytes of space_info
Liu Bo [Mon, 13 Feb 2017 23:42:21 +0000 (15:42 -0800)]
Btrfs: use helper to get used bytes of space_info

This uses a helper instead of open code around used byte of space_info
everywhere.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: try to avoid acquiring free space ctl's lock
Liu Bo [Mon, 13 Feb 2017 23:42:30 +0000 (15:42 -0800)]
Btrfs: try to avoid acquiring free space ctl's lock

We don't need to take the lock if the block group has not been cached.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: Better csum error message for data csum mismatch
Qu Wenruo [Thu, 9 Feb 2017 02:45:06 +0000 (10:45 +0800)]
btrfs: Better csum error message for data csum mismatch

The original csum error message only outputs inode number, offset, check
sum and expected check sum.

However no root objectid is outputted, which sometimes makes debugging
quite painful under multi-subvolume case (including relocation).

Also the checksum output is decimal, which seldom makes sense for
users/developers and is hard to read in most time.

This patch will add root objectid, which will be %lld for rootid larger
than LAST_FREE_OBJECTID, and hex csum output for better readability.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: add another missing end_page_writeback on submit_extent_page failure
Takafumi Kubota [Thu, 9 Feb 2017 08:24:33 +0000 (17:24 +0900)]
Btrfs: add another missing end_page_writeback on submit_extent_page failure

If btrfs_bio_alloc fails in submit_extent_page, submit_extent_page returns
without clearing the writeback bit of the failed page.

__extent_writepage_io, that is a caller of submit_extent_page,
does not clear the remaining writeback bit anywhere.
As a result, this will cause the hang at filemap_fdatawait_range,
because it waits the writeback bit to be cleared from the failed page.
So, we have to call end_page_writeback to clear the writeback bit.

For reproducing the hang, we inject a fault like

   if (should_failtest()) { // I define should_failtest()
        bio = NULL;
   }
   else {
        bio = btrfs_bio_alloc(...);
   }

in submit_extent_page.

We should also check whether page has the bit before end_page_writeback,
to avoid the conflict against the other end_page_writeback in bio_endio.
Thus, we add PageWriteback checks not only in __extent_writepage_io,
but also in write_one_eb too, because it misses the check.

Signed-off-by: Takafumi Kubota <takafumi.kubota1012@sslab.ics.keio.ac.jp>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Cc: David Sterba <dsterba@suse.cz>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: remove unused ulist members
David Sterba [Thu, 9 Feb 2017 15:47:43 +0000 (16:47 +0100)]
btrfs: remove unused ulist members

Commit "btrfs: ulist: Add ulist_del() function" (d4b804045924d7f8)
removed some debugging code but left the structure defintions.

Reviewed-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: use helper to simplify lock/unlock pages
Liu Bo [Fri, 10 Feb 2017 15:42:14 +0000 (16:42 +0100)]
Btrfs: use helper to simplify lock/unlock pages

Since we have a helper to set page bits, let lock_delalloc_pages and
__unlock_for_delalloc use it.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: teach __process_pages_contig about PAGE_LOCK operation
Liu Bo [Fri, 10 Feb 2017 15:41:05 +0000 (16:41 +0100)]
btrfs: teach __process_pages_contig about PAGE_LOCK operation

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ changes to the helper separated from the following patch ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: create helper for processing bits on contiguous pages
Liu Bo [Fri, 3 Feb 2017 01:49:22 +0000 (17:49 -0800)]
Btrfs: create helper for processing bits on contiguous pages

This introduces a new helper which can be used to process pages bits.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist
Liu Bo [Mon, 30 Jan 2017 20:25:28 +0000 (12:25 -0800)]
Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist

run_delalloc_nocow has used trans in two places where they don't
actually need @trans.

For btrfs_lookup_file_extent, we search for file extents without COWing
anything, and for btrfs_cross_ref_exist, the only place where we need
@trans is deferencing it in order to get running_transaction which we
could easily get from the global fs_info.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: pass delayed_refs directly to btrfs_find_delayed_ref_head
Liu Bo [Mon, 30 Jan 2017 20:24:37 +0000 (12:24 -0800)]
Btrfs: pass delayed_refs directly to btrfs_find_delayed_ref_head

All we need is @delayed_refs, all callers have get it ahead of calling
btrfs_find_delayed_ref_head since lock needs to be acquired firstly,
there is no reason to deference it again inside the function.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: remove unused trans in read_block_for_search
Liu Bo [Mon, 30 Jan 2017 20:23:42 +0000 (12:23 -0800)]
Btrfs: remove unused trans in read_block_for_search

@trans is not used at all, this removes it.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: cleanup unused cached_state in __extent_writepage_io
Liu Bo [Thu, 26 Jan 2017 01:15:54 +0000 (17:15 -0800)]
Btrfs: cleanup unused cached_state in __extent_writepage_io

@cached_state is no more required in __extent_writepage_io, also remove
the goto label.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs: allow unlink to exceed subvolume quota
Jeff Mahoney [Wed, 25 Jan 2017 14:50:33 +0000 (09:50 -0500)]
btrfs: allow unlink to exceed subvolume quota

Once a qgroup limit is exceeded, it's impossible to restore normal
operation to the subvolume without modifying the limit or removing
the subvolume.  This is a surprising situation for many users used
to the typical workflow with quotas on other file systems where it's
possible to remove files until the used space is back under the limit.

When we go to unlink a file and start the transaction, we'll hit
the qgroup limit while trying to reserve space for the items we'll
modify while removing the file.  We discussed last month how best
to handle this situation and agreed that there is no perfect solution.
The best principle-of-least-surprise solution is to handle it similarly
to how we already handle ENOSPC when unlinking, which is to allow
the operation to succeed with the expectation that it will ultimately
release space under most circumstances.

This patch modifies the transaction start path to select whether to
honor the qgroups limits.  btrfs_start_transaction_fallback_global_rsv
is the only caller that skips enforcement.  The reservation and tracking
still happens normally -- it just skips the enforcement step.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs: fix wrong argument for btrfs_lookup_ordered_range
Liu Bo [Tue, 24 Jan 2017 23:58:51 +0000 (15:58 -0800)]
Btrfs: fix wrong argument for btrfs_lookup_ordered_range

Commit Btrfs: btrfs_page_mkwrite: Reserve space in sectorsized units"
(d0b7da88) did this, but btrfs_lookup_ordered_range expects a 'length'
rather than a 'page_end'.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: David Sterba <dsterba@suse.com>