- 09 Sep, 2019 40 commits
-
-
Johannes Thumshirn authored
Create a structure to encode the type and length for the known on-disk checksums. This makes it easier to add new checksums later. The structure and helpers are moved from ctree.h so they don't occupy space in all headers including ctree.h. This save some space in the final object. Reviewed-by:
Nikolay Borisov <nborisov@suse.com> Signed-off-by:
Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by:
David Sterba <dsterba@suse.com> Signed-off-by:
-
Johannes Thumshirn authored
Turn the checksum type definition into a enum. This eases later addition of new checksums. Reviewed-by:
-
Josef Bacik authored
When debugging weird enospc problems it's handy to be able to dump the space info when we wake up all tickets, and see what the ticket values are. This helped me figure out cases where we were enospc'ing when we shouldn't have been. Reviewed-by:
Josef Bacik <josef@toxicpanda.com> Signed-off-by:
-
Josef Bacik authored
We ran into a problem in production where a box with plenty of space was getting wedged doing ENOSPC flushing. These boxes only had 20% of the disk allocated, but their metadata space + global reserve was right at the size of their metadata chunk. In this case can_overcommit should be allowing allocations without problem, but there's logic in can_overcommit that doesn't allow us to overcommit if there's not enough real space to satisfy the global reserve. This is for historical reasons. Before there were only certain places we could allocate chunks. We could go to commit the transaction and not have enough space for our pending delayed refs and such and be unable to allocate a new chunk. This would result in a abort because of ENOSPC. This code was added to solve this problem. However since then we've gained the ability to always be able to allocate a chunk. So we can easily overcommit in these cases without risking a transaction abort because of ENOSPC. Also prior to now the global reserve really would be used because that's the space we relied on for delayed refs. With delayed refs being tracked separately we no longer have to worry about running out of delayed refs space while committing. We are much less likely to exhaust our global reserve space during transaction commit. Fix the can_overcommit code to simply see if our current usage + what we want is less than our current free space plus whatever slack space we have in the disk is. This solves the problem we were seeing in production and keeps us from flushing as aggressively as we approach our actual metadata size usage. Reviewed-by:
-
Josef Bacik authored
We have some annoying xfstests tests that will create a very small fs, fill it up, delete it, and repeat to make sure everything works right. This trips btrfs up sometimes because we may commit a transaction to free space, but most of the free metadata space was being reserved by the global reserve. So we commit and update the global reserve, but the space is simply added to bytes_may_use directly, instead of trying to add it to existing tickets. This results in ENOSPC when we really did have space. Fix this by calling btrfs_try_granting_tickets once we add back our excess space to wake any pending tickets. Reviewed-by:
-
Josef Bacik authored
While messing with the overcommit logic I noticed that sometimes we'd ENOSPC out when really we should have run out of space much earlier. It turns out it's because we'll only reserve up to the free amount left in the space info for the global reserve, but that doesn't make sense with overcommit because we could be well above our actual size. This results in the global reserve not carving out it's entire reservation, and thus not putting enough pressure on the rest of the infrastructure to do the right thing and ENOSPC out at a convenient time. Fix this by always taking our full reservation amount for the global reserve. Reviewed-by:
-
Josef Bacik authored
It made sense to have the global reserve set at 16M in the past, but since it is used less nowadays set the minimum size to the number of items we'll need to update the main trees we update during a transaction commit, plus some slop area so we can do unlinks if we need to. In practice this doesn't affect normal file systems, but for xfstests where we do things like fill up a fs and then rm * it can fall over in weird ways. This enables us for more sane behavior at extremely small file system sizes. Signed-off-by:
-
Josef Bacik authored
This name doesn't really fit with how the space reservation stuff works now, rename it to btrfs_space_info_free_bytes_may_use so it's clear what the function is doing. Reviewed-by:
-
Josef Bacik authored
Now that we do not do partial filling of tickets simply remove orig_bytes, it is no longer needed. Reviewed-by:
-
Josef Bacik authored
Now that we aren't partially filling tickets we may have some slack space left in the space_info. We need to account for this in may_commit_transaction, otherwise we may choose to not commit the transaction despite it actually having enough space to satisfy our ticket. Calculate the free space we have in the space_info, if any, and subtract this from the ticket we have and use that amount to determine if we will need to commit to reclaim enough space. Reviewed-by:
-
Josef Bacik authored
Now that we no longer partially fill tickets we need to rework wake_all_tickets to call btrfs_try_to_wakeup_tickets() in order to see if any subsequent tickets are able to be satisfied. If our tickets_id changes we know something happened and we can keep flushing. Also if we find a ticket that is smaller than the first ticket in our queue then we want to retry the flushing loop again in case may_commit_transaction() decides we could satisfy the ticket by committing the transaction. Rename this to maybe_fail_all_tickets() while we're at it, to better reflect what the function is actually doing. Signed-off-by:
-
Josef Bacik authored
Now that btrfs_space_info_add_old_bytes simply checks if we can make the reservation and updates bytes_may_use, there's no reason to have both helpers in place. Factor out the ticket wakeup logic into it's own helper, make btrfs_space_info_add_old_bytes() update bytes_may_use and then call the wakeup helper, and replace all calls to btrfs_space_info_add_new_bytes() with the wakeup helper. Reviewed-by:
-
Josef Bacik authored
btrfs_space_info_add_old_bytes is used when adding the extra space from an existing reservation back into the space_info to be used by any waiting tickets. In order to keep us from overcommitting we check to make sure that we can still use this space for our reserve ticket, and if we cannot we'll simply subtract it from space_info->bytes_may_use. However this is problematic, because it assumes that only changes to bytes_may_use would affect our ability to make reservations. Any changes to bytes_reserved would be missed. If we were unable to make a reservation prior because of reserved space, but that reserved space was free'd due to unlink or truncate and we were allowed to immediately reclaim that metadata space we would still ENOSPC. Consider the example where we create a file with a bunch of extents, using up 2MiB of actual space for the new tree blocks. Then we try to make a reservation of 2MiB but we do not have enough space to make this reservation. The iput() occurs in another thread and we remove this space, and since we did not write the blocks we simply do space_info->bytes_reserved -= 2MiB. We would never see this because we do not check our space info used, we just try to re-use the freed reservations. To fix this problem, and to greatly simplify the wakeup code, do away with this partial refilling nonsense. Use btrfs_space_info_add_old_bytes to subtract the reservation from space_info->bytes_may_use, and then check the ticket against the total used of the space_info the same way we do with the initial reservation attempt. This keeps the reservation logic consistent and solves the problem of early ENOSPC in the case that we free up space in places other than bytes_may_use and bytes_pinned. Thanks, Signed-off-by:
-
Josef Bacik authored
I noticed when folding the trace_btrfs_space_reservation() tracepoint into the btrfs_space_info_update_* helpers that we didn't emit a tracepoint when doing btrfs_add_reserved_bytes(). I know this is because we were swapping bytes_may_use for bytes_reserved, so in my mind there was no reason to have the tracepoint there. But now there is because we always emit the unreserve for the bytes_may_use side, and this would have broken if compression was on anyway. Add a tracepoint to cover the bytes_reserved counter so the math still comes out right. Signed-off-by:
-
Josef Bacik authored
We duplicate this tracepoint everywhere we call these helpers, so update the helper to have the tracepoint as well. Signed-off-by:
-
Josef Bacik authored
If we already have tickets on the list we don't want to steal their reservations. This is a preparation patch for upcoming changes, technically this shouldn't happen today because of the way we add bytes to tickets before adding them to the space_info in most cases. This does not change the FIFO nature of reserve tickets, it simply allows us to enforce it in a different way. Previously it was enforced because any new space would be added to the first ticket on the list, which would result in new reservations getting a reserve ticket. This replaces that mechanism by simply checking to see if we have outstanding reserve tickets and skipping straight to adding a ticket for our reservation. Reviewed-by:
-
Omar Sandoval authored
Since commit fee187d9 ("Btrfs: do not set EXTENT_DIRTY along with EXTENT_DELALLOC"), we never set EXTENT_DIRTY in inode->io_tree, so we can simplify and stop trying to clear it. Reviewed-by:
Omar Sandoval <osandov@fb.com> Reviewed-by:
-
Omar Sandoval authored
The VFS indicates a synchronous write to ->write_iter() via iocb->ki_flags. The IOCB_{,D}SYNC flags may be set based on the file (see iocb_flags()) or the RWF_* flags passed to a syscall like pwritev2() (see kiocb_set_rw_flags()). However, in btrfs_file_write_iter(), we're checking if a write is synchronous based only on the file; we use this to decide when to bump the sync_writers counter and thus do CRCs synchronously. Make sure we do this for all synchronous writes as determined by the VFS. Reviewed-by: -
Omar Sandoval authored
generic_write_checks() may modify iov_iter_count(), so we must get the count after the call, not before. Using the wrong one has a couple of consequences: 1. We check a longer range in check_can_nocow() for nowait than we're actually writing. 2. We create extra hole extent maps in btrfs_cont_expand(). As far as I can tell, this is harmless, but I might be missing something. These issues are pretty minor, but let's fix it before something more important trips on it. Fixes: edf064e7 ("btrfs: nowait aio support") Reviewed-by:
-
David Sterba authored
Further simplifaction of the get/set helpers is possible when the token is uniquely tied to an extent buffer. A condition and an assignment can be avoided. The initializations are moved closer to the first use when the extent buffer is valid. There's one exception in __push_leaf_left where the token is reused. Signed-off-by: -
David Sterba authored
Now that we can safely assume that the token is always a valid pointer, remove the branches that check that. Signed-off-by: -
David Sterba authored
There are helpers for all type widths defined via macro and optionally can use a token which is a cached pointer to avoid repeated mapping of the extent buffer. The token value is known at compile time, when it's valid it's always address of a local variable, otherwise it's NULL passed by the token-less helpers. This can be utilized to remove some branching as the helpers are used frequenlty. Signed-off-by: -
Nikolay Borisov authored
btrfs_find_name_in_ext_backref returns either 0/1 depending on whether it found a backref for the given name. If it returns true then the actual inode_ref struct is returned in one of its parameters. That's pointless, instead refactor the function such that it returns either a pointer to the btrfs_inode_extref or NULL it it didn't find anything. This streamlines the function calling convention. Signed-off-by:
-
Nikolay Borisov authored
btrfs_find_name_in_backref returns either 0/1 depending on whether it found a backref for the given name. If it returns true then the actual inode_ref struct is returned in one of its parameters. That's pointless, instead refactor the function such that it returns either a pointer to the btrfs_inode_ref or NULL it it didn't find anything. This streamlines the function calling convention. Signed-off-by:
-
David Sterba authored
The other dev stats functions are already there and the helpers are not used by anything else. Reviewed-by:
-
David Sterba authored
The io_ctl structure is used for free space management, and used only by the v1 space cache code, but unfortunatlly the full definition is required by block-group.h so it can't be moved to free-space-cache.c without additional changes. Reviewed-by:
-
David Sterba authored
Send is the only user of tree_compare, we can move it there along with the other helpers and definitions. Reviewed-by:
-
David Sterba authored
Preparatory work for code that will be moved out of ctree and uses this function. Reviewed-by:
-
David Sterba authored
Reviewed-by:
-
David Sterba authored
Reviewed-by:
-
David Sterba authored
The file ctree.h serves as a header for everything and has become quite bloated. Split some helpers that are generic and create a new file that should be the catch-all for code that's not btrfs-specific. Reviewed-by:
-
Anand Jain authored
Fix the fake ENOMEM return error code to the actual error in clone_fs_devices(). Reviewed-by:
Anand Jain <anand.jain@oracle.com> Reviewed-by:
-
Anand Jain authored
In a corrupted tree, if search for next devid finds the device with devid = -1, then report the error -EUCLEAN back to the parent function to fail gracefully. The tree checker will not catch this in case the devids are created using the following script: umount /btrfs dev1=/dev/sdb dev2=/dev/sdc mkfs.btrfs -fq -dsingle -msingle $dev1 mount $dev1 /btrfs _fail() { echo $1 exit 1 } while true; do btrfs dev add -f $dev2 /btrfs || _fail "add failed" btrfs dev del $dev1 /btrfs || _fail "del failed" dev_tmp=$dev1 dev1=$dev2 dev2=$dev_tmp done With output: BTRFS critical (device sdb): corrupt leaf: root=3 block=313739198464 slot=1 devid=1 invalid devid: has=507 expect=[0, 506] BTRFS error (device sdb): block=313739198464 write time tree block corruption detected BTRFS: error (device sdb) in btrfs_commit_transaction:2268: errno=-5 IO failure (Error while writing out transaction) BTRFS warning (device sdb): Skipping commit of aborted transaction. BTRFS: error (device sdb) in cleanup_transaction:1827: errno=-5 IO failure Reviewed-by:Qu Wenruo <wqu@suse.com> Reviewed-by:
-
Christophe Leroy authored
Various notifications of type "BUG kmalloc-4096 () : Redzone overwritten" have been observed recently in various parts of the kernel. After some time, it has been made a relation with the use of BTRFS filesystem and with SLUB_DEBUG turned on. [ 22.809700] BUG kmalloc-4096 (Tainted: G W ): Redzone overwritten [ 22.810286] INFO: 0xbe1a5921-0xfbfc06cd. First byte 0x0 instead of 0xcc [ 22.810866] INFO: Allocated in __load_free_space_cache+0x588/0x780 [btrfs] age=22 cpu=0 pid=224 [ 22.811193] __slab_alloc.constprop.26+0x44/0x70 [ 22.811345] kmem_cache_alloc_trace+0xf0/0x2ec [ 22.811588] __load_free_space_cache+0x588/0x780 [btrfs] [ 22.811848] load_free_space_cache+0xf4/0x1b0 [btrfs] [ 22.812090] cache_block_group+0x1d0/0x3d0 [btrfs] [ 22.812321] find_free_extent+0x680/0x12a4 [btrfs] [ 22.812549] btrfs_reserve_extent+0xec/0x220 [btrfs] [ 22.812785] btrfs_alloc_tree_block+0x178/0x5f4 [btrfs] [ 22.813032] __btrfs_cow_block+0x150/0x5d4 [btrfs] [ 22.813262] btrfs_cow_block+0x194/0x298 [btrfs] [ 22.813484] commit_cowonly_roots+0x44/0x294 [btrfs] [ 22.813718] btrfs_commit_transaction+0x63c/0xc0c [btrfs] [ 22.813973] close_ctree+0xf8/0x2a4 [btrfs] [ 22.814107] generic_shutdown_super+0x80/0x110 [ 22.814250] kill_anon_super+0x18/0x30 [ 22.814437] btrfs_kill_super+0x18/0x90 [btrfs] [ 22.814590] INFO: Freed in proc_cgroup_show+0xc0/0x248 age=41 cpu=0 pid=83 [ 22.814841] proc_cgroup_show+0xc0/0x248 [ 22.814967] proc_single_show+0x54/0x98 [ 22.815086] seq_read+0x278/0x45c [ 22.815190] __vfs_read+0x28/0x17c [ 22.815289] vfs_read+0xa8/0x14c [ 22.815381] ksys_read+0x50/0x94 [ 22.815475] ret_from_syscall+0x0/0x38 Commit 69d24804 ("btrfs: use copy_page for copying pages instead of memcpy") changed the way bitmap blocks are copied. But allthough bitmaps have the size of a page, they were allocated with kzalloc(). Most of the time, kzalloc() allocates aligned blocks of memory, so copy_page() can be used. But when some debug options like SLAB_DEBUG are activated, kzalloc() may return unaligned pointer. On powerpc, memcpy(), copy_page() and other copying functions use 'dcbz' instruction which provides an entire zeroed cacheline to avoid memory read when the intention is to overwrite a full line. Functions like memcpy() are writen to care about partial cachelines at the start and end of the destination, but copy_page() assumes it gets pages. As pages are naturally cache aligned, copy_page() doesn't care about partial lines. This means that when copy_page() is called with a misaligned pointer, a few leading bytes are zeroed. To fix it, allocate bitmaps through kmem_cache instead of using kzalloc() The cache pool is created with PAGE_SIZE alignment constraint. Reported-by:
Erhard F. <erhard_f@mailbox.org> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=204371 Fixes: 69d24804 ("btrfs: use copy_page for copying pages instead of memcpy") Cc: stable@vger.kernel.org # 4.19+ Signed-off-by:
Christophe Leroy <christophe.leroy@c-s.fr> Reviewed-by:
-
Qu Wenruo authored
[BUG] With crafted image, btrfs will panic at btree operations: kernel BUG at fs/btrfs/ctree.c:3894! invalid opcode: 0000 [#1] SMP PTI CPU: 0 PID: 1138 Comm: btrfs-transacti Not tainted 5.0.0-rc8+ #9 RIP: 0010:__push_leaf_left+0x6b6/0x6e0 RSP: 0018:ffffc0bd4128b990 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffffa0a4ab8f0e38 RCX: 0000000000000000 RDX: ffffa0a280000000 RSI: 0000000000000000 RDI: ffffa0a4b3814000 RBP: ffffc0bd4128ba38 R08: 0000000000001000 R09: ffffc0bd4128b948 R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000240 R13: ffffa0a4b556fb60 R14: ffffa0a4ab8f0af0 R15: ffffa0a4ab8f0af0 FS: 0000000000000000(0000) GS:ffffa0a4b7a00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f2461c80020 CR3: 000000022b32a006 CR4: 00000000000206f0 Call Trace: ? _cond_resched+0x1a/0x50 push_leaf_left+0x179/0x190 btrfs_del_items+0x316/0x470 btrfs_del_csums+0x215/0x3a0 __btrfs_free_extent.isra.72+0x5a7/0xbe0 __btrfs_run_delayed_refs+0x539/0x1120 btrfs_run_delayed_refs+0xdb/0x1b0 btrfs_commit_transaction+0x52/0x950 ? start_transaction+0x94/0x450 transaction_kthread+0x163/0x190 kthread+0x105/0x140 ? btrfs_cleanup_transaction+0x560/0x560 ? kthread_destroy_worker+0x50/0x50 ret_from_fork+0x35/0x40 Modules linked in: ---[ end trace c2425e6e89b5558f ]--- [CAUSE] The offending csum tree looks like this: checksum tree key (CSUM_TREE ROOT_ITEM 0) node 29741056 level 1 items 14 free 107 generation 19 owner CSUM_TREE ... key (EXTENT_CSUM EXTENT_CSUM 85975040) block 29630464 gen 17 key (EXTENT_CSUM EXTENT_CSUM 89911296) block 29642752 gen 17 <<< key (EXTENT_CSUM EXTENT_CSUM 92274688) block 29646848 gen 17 ... leaf 29630464 items 6 free space 1 generation 17 owner CSUM_TREE item 0 key (EXTENT_CSUM EXTENT_CSUM 85975040) itemoff 3987 itemsize 8 range start 85975040 end 85983232 length 8192 ... leaf 29642752 items 0 free space 3995 generation 17 owner 0 ^ empty leaf invalid owner ^ leaf 29646848 items 1 free space 602 generation 17 owner CSUM_TREE item 0 key (EXTENT_CSUM EXTENT_CSUM 92274688) itemoff 627 itemsize 3368 range start 92274688 end 95723520 length 3448832 So we have a corrupted csum tree where one tree leaf is completely empty, causing unbalanced btree, thus leading to unexpected btree balance error. [FIX] For this particular case, we handle it in two directions to catch it: - Check if the tree block is empty through btrfs_verify_level_key() So that invalid tree blocks won't be read out through btrfs_search_slot() and its variants. - Check 0 tree owner in tree checker NO tree is using 0 as its tree owner, detect it and reject at tree block read time. Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=202821Reviewed-by:
-
Nikolay Borisov authored
Support for asynchronous snapshot creation was originally added in 72fd032e ("Btrfs: add SNAP_CREATE_ASYNC ioctl") to cater for ceph's backend needs. However, since Ceph has deprecated support for btrfs there is no longer need for that support in btrfs. Additionally, this was never supported by btrfs-progs, the official userspace tools. Reviewed-by:
-
Nikolay Borisov authored
Correctly handle failure cases when adding an ordered extents in case of REGULAR or PREALLOC extents. Remove the BUG_ON. Signed-off-by:
-
Nikolay Borisov authored
Add a comment explaining why we keep the BUG also use the already read and cached value of extent ram bytes stored in 'ram_bytes'. Signed-off-by:
-
Nikolay Borisov authored
The extent range check right after the "out_check" label is redundant, because the only way it can trigger is if we have an inline extent. In this case it makes more sense to actually move it in the branch explictly dealing with inlines extents. What's more, the nested 'if (nocow)' can never be true because for inline extents we always do COW and there is no chance 'nocow' can be true, just remove that check. Signed-off-by:
-
Nikolay Borisov authored
There is no point in checking the type of the extent again just to set the 'type' variable, when this check has already been performed before. Instead, extend the original if branch with an 'else' clause. This allows to remove one local variable and make it obvious how the code flow differs for prealloc/regular extents. Reviewed-by:
-
