diff options
Diffstat (limited to 'fs/btrfs/space-info.c')
| -rw-r--r-- | fs/btrfs/space-info.c | 233 |
1 files changed, 50 insertions, 183 deletions
diff --git a/fs/btrfs/space-info.c b/fs/btrfs/space-info.c index 2dc674b7c3b1..f79bf85f2439 100644 --- a/fs/btrfs/space-info.c +++ b/fs/btrfs/space-info.c @@ -133,18 +133,13 @@ * operations, however they won't be usable until the transaction commits. * * COMMIT_TRANS - * may_commit_transaction() is the ultimate arbiter on whether we commit the - * transaction or not. In order to avoid constantly churning we do all the - * above flushing first and then commit the transaction as the last resort. - * However we need to take into account things like pinned space that would - * be freed, plus any delayed work we may not have gotten rid of in the case - * of metadata. - * - * FORCE_COMMIT_TRANS - * For use by the preemptive flusher. We use this to bypass the ticketing - * checks in may_commit_transaction, as we have more information about the - * overall state of the system and may want to commit the transaction ahead - * of actual ENOSPC conditions. + * This will commit the transaction. Historically we had a lot of logic + * surrounding whether or not we'd commit the transaction, but this waits born + * out of a pre-tickets era where we could end up committing the transaction + * thousands of times in a row without making progress. Now thanks to our + * ticketing system we know if we're not making progress and can error + * everybody out after a few commits rather than burning the disk hoping for + * a different answer. * * OVERCOMMIT * @@ -197,13 +192,6 @@ static int create_space_info(struct btrfs_fs_info *info, u64 flags) if (!space_info) return -ENOMEM; - ret = percpu_counter_init(&space_info->total_bytes_pinned, 0, - GFP_KERNEL); - if (ret) { - kfree(space_info); - return ret; - } - for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) INIT_LIST_HEAD(&space_info->block_groups[i]); init_rwsem(&space_info->groups_sem); @@ -389,7 +377,7 @@ again: ticket = list_first_entry(head, struct reserve_ticket, list); - /* Check and see if our ticket can be satisified now. */ + /* Check and see if our ticket can be satisfied now. */ if ((used + ticket->bytes <= space_info->total_bytes) || btrfs_can_overcommit(fs_info, space_info, ticket->bytes, flush)) { @@ -495,7 +483,8 @@ static inline u64 calc_reclaim_items_nr(struct btrfs_fs_info *fs_info, */ static void shrink_delalloc(struct btrfs_fs_info *fs_info, struct btrfs_space_info *space_info, - u64 to_reclaim, bool wait_ordered) + u64 to_reclaim, bool wait_ordered, + bool for_preempt) { struct btrfs_trans_handle *trans; u64 delalloc_bytes; @@ -532,7 +521,7 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info, * ordered extents, otherwise we'll waste time trying to flush delalloc * that likely won't give us the space back we need. */ - if (ordered_bytes > delalloc_bytes) + if (ordered_bytes > delalloc_bytes && !for_preempt) wait_ordered = true; loops = 0; @@ -551,6 +540,14 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info, break; } + /* + * If we are for preemption we just want a one-shot of delalloc + * flushing so we can stop flushing if we decide we don't need + * to anymore. + */ + if (for_preempt) + break; + spin_lock(&space_info->lock); if (list_empty(&space_info->tickets) && list_empty(&space_info->priority_tickets)) { @@ -566,109 +563,6 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info, } } -/** - * Possibly commit the transaction if its ok to - * - * @fs_info: the filesystem - * @space_info: space_info we are checking for commit, either data or metadata - * - * This will check to make sure that committing the transaction will actually - * get us somewhere and then commit the transaction if it does. Otherwise it - * will return -ENOSPC. - */ -static int may_commit_transaction(struct btrfs_fs_info *fs_info, - struct btrfs_space_info *space_info) -{ - struct reserve_ticket *ticket = NULL; - struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_block_rsv; - struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv; - struct btrfs_block_rsv *trans_rsv = &fs_info->trans_block_rsv; - struct btrfs_trans_handle *trans; - u64 reclaim_bytes = 0; - u64 bytes_needed = 0; - u64 cur_free_bytes = 0; - - trans = (struct btrfs_trans_handle *)current->journal_info; - if (trans) - return -EAGAIN; - - spin_lock(&space_info->lock); - cur_free_bytes = btrfs_space_info_used(space_info, true); - if (cur_free_bytes < space_info->total_bytes) - cur_free_bytes = space_info->total_bytes - cur_free_bytes; - else - cur_free_bytes = 0; - - if (!list_empty(&space_info->priority_tickets)) - ticket = list_first_entry(&space_info->priority_tickets, - struct reserve_ticket, list); - else if (!list_empty(&space_info->tickets)) - ticket = list_first_entry(&space_info->tickets, - struct reserve_ticket, list); - if (ticket) - bytes_needed = ticket->bytes; - - if (bytes_needed > cur_free_bytes) - bytes_needed -= cur_free_bytes; - else - bytes_needed = 0; - spin_unlock(&space_info->lock); - - if (!bytes_needed) - return 0; - - trans = btrfs_join_transaction(fs_info->extent_root); - if (IS_ERR(trans)) - return PTR_ERR(trans); - - /* - * See if there is enough pinned space to make this reservation, or if - * we have block groups that are going to be freed, allowing us to - * possibly do a chunk allocation the next loop through. - */ - if (test_bit(BTRFS_TRANS_HAVE_FREE_BGS, &trans->transaction->flags) || - __percpu_counter_compare(&space_info->total_bytes_pinned, - bytes_needed, - BTRFS_TOTAL_BYTES_PINNED_BATCH) >= 0) - goto commit; - - /* - * See if there is some space in the delayed insertion reserve for this - * reservation. If the space_info's don't match (like for DATA or - * SYSTEM) then just go enospc, reclaiming this space won't recover any - * space to satisfy those reservations. - */ - if (space_info != delayed_rsv->space_info) - goto enospc; - - spin_lock(&delayed_rsv->lock); - reclaim_bytes += delayed_rsv->reserved; - spin_unlock(&delayed_rsv->lock); - - spin_lock(&delayed_refs_rsv->lock); - reclaim_bytes += delayed_refs_rsv->reserved; - spin_unlock(&delayed_refs_rsv->lock); - - spin_lock(&trans_rsv->lock); - reclaim_bytes += trans_rsv->reserved; - spin_unlock(&trans_rsv->lock); - - if (reclaim_bytes >= bytes_needed) - goto commit; - bytes_needed -= reclaim_bytes; - - if (__percpu_counter_compare(&space_info->total_bytes_pinned, - bytes_needed, - BTRFS_TOTAL_BYTES_PINNED_BATCH) < 0) - goto enospc; - -commit: - return btrfs_commit_transaction(trans); -enospc: - btrfs_end_transaction(trans); - return -ENOSPC; -} - /* * Try to flush some data based on policy set by @state. This is only advisory * and may fail for various reasons. The caller is supposed to examine the @@ -702,7 +596,7 @@ static void flush_space(struct btrfs_fs_info *fs_info, case FLUSH_DELALLOC: case FLUSH_DELALLOC_WAIT: shrink_delalloc(fs_info, space_info, num_bytes, - state == FLUSH_DELALLOC_WAIT); + state == FLUSH_DELALLOC_WAIT, for_preempt); break; case FLUSH_DELAYED_REFS_NR: case FLUSH_DELAYED_REFS: @@ -743,9 +637,7 @@ static void flush_space(struct btrfs_fs_info *fs_info, btrfs_wait_on_delayed_iputs(fs_info); break; case COMMIT_TRANS: - ret = may_commit_transaction(fs_info, space_info); - break; - case FORCE_COMMIT_TRANS: + ASSERT(current->journal_info == NULL); trans = btrfs_join_transaction(root); if (IS_ERR(trans)) { ret = PTR_ERR(trans); @@ -792,12 +684,14 @@ btrfs_calc_reclaim_metadata_size(struct btrfs_fs_info *fs_info, static bool need_preemptive_reclaim(struct btrfs_fs_info *fs_info, struct btrfs_space_info *space_info) { + u64 global_rsv_size = fs_info->global_block_rsv.reserved; u64 ordered, delalloc; u64 thresh = div_factor_fine(space_info->total_bytes, 98); u64 used; /* If we're just plain full then async reclaim just slows us down. */ - if ((space_info->bytes_used + space_info->bytes_reserved) >= thresh) + if ((space_info->bytes_used + space_info->bytes_reserved + + global_rsv_size) >= thresh) return false; /* @@ -838,8 +732,10 @@ static bool need_preemptive_reclaim(struct btrfs_fs_info *fs_info, thresh = calc_available_free_space(fs_info, space_info, BTRFS_RESERVE_FLUSH_ALL); - thresh += (space_info->total_bytes - space_info->bytes_used - - space_info->bytes_reserved - space_info->bytes_readonly); + used = space_info->bytes_used + space_info->bytes_reserved + + space_info->bytes_readonly + global_rsv_size; + if (used < space_info->total_bytes) + thresh += space_info->total_bytes - used; thresh >>= space_info->clamp; used = space_info->bytes_pinned; @@ -860,14 +756,20 @@ static bool need_preemptive_reclaim(struct btrfs_fs_info *fs_info, * clearly be heavy enough to warrant preemptive flushing. In the case * of heavy DIO or ordered reservations, preemptive flushing will just * waste time and cause us to slow down. + * + * We want to make sure we truly are maxed out on ordered however, so + * cut ordered in half, and if it's still higher than delalloc then we + * can keep flushing. This is to avoid the case where we start + * flushing, and now delalloc == ordered and we stop preemptively + * flushing when we could still have several gigs of delalloc to flush. */ - ordered = percpu_counter_read_positive(&fs_info->ordered_bytes); + ordered = percpu_counter_read_positive(&fs_info->ordered_bytes) >> 1; delalloc = percpu_counter_read_positive(&fs_info->delalloc_bytes); if (ordered >= delalloc) used += fs_info->delayed_refs_rsv.reserved + fs_info->delayed_block_rsv.reserved; else - used += space_info->bytes_may_use; + used += space_info->bytes_may_use - global_rsv_size; return (used >= thresh && !btrfs_fs_closing(fs_info) && !test_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state)); @@ -921,7 +823,6 @@ static bool maybe_fail_all_tickets(struct btrfs_fs_info *fs_info, { struct reserve_ticket *ticket; u64 tickets_id = space_info->tickets_id; - u64 first_ticket_bytes = 0; if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) { btrfs_info(fs_info, "cannot satisfy tickets, dumping space info"); @@ -937,21 +838,6 @@ static bool maybe_fail_all_tickets(struct btrfs_fs_info *fs_info, steal_from_global_rsv(fs_info, space_info, ticket)) return true; - /* - * may_commit_transaction will avoid committing the transaction - * if it doesn't feel like the space reclaimed by the commit - * would result in the ticket succeeding. However if we have a - * smaller ticket in the queue it may be small enough to be - * satisified by committing the transaction, so if any - * subsequent ticket is smaller than the first ticket go ahead - * and send us back for another loop through the enospc flushing - * code. - */ - if (first_ticket_bytes == 0) - first_ticket_bytes = ticket->bytes; - else if (first_ticket_bytes > ticket->bytes) - return true; - if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) btrfs_info(fs_info, "failing ticket with %llu bytes", ticket->bytes); @@ -1117,7 +1003,7 @@ static void btrfs_preempt_reclaim_metadata_space(struct work_struct *work) (delayed_block_rsv->reserved + delayed_refs_rsv->reserved)) { to_reclaim = space_info->bytes_pinned; - flush = FORCE_COMMIT_TRANS; + flush = COMMIT_TRANS; } else if (delayed_block_rsv->reserved > delayed_refs_rsv->reserved) { to_reclaim = delayed_block_rsv->reserved; @@ -1171,28 +1057,9 @@ static void btrfs_preempt_reclaim_metadata_space(struct work_struct *work) * immediately re-usable, it comes in the form of a delayed ref, which must be * run and then the transaction must be committed. * - * FLUSH_DELAYED_REFS - * The above two cases generate delayed refs that will affect - * ->total_bytes_pinned. However this counter can be inconsistent with - * reality if there are outstanding delayed refs. This is because we adjust - * the counter based solely on the current set of delayed refs and disregard - * any on-disk state which might include more refs. So for example, if we - * have an extent with 2 references, but we only drop 1, we'll see that there - * is a negative delayed ref count for the extent and assume that the space - * will be freed, and thus increase ->total_bytes_pinned. - * - * Running the delayed refs gives us the actual real view of what will be - * freed at the transaction commit time. This stage will not actually free - * space for us, it just makes sure that may_commit_transaction() has all of - * the information it needs to make the right decision. - * * COMMIT_TRANS - * This is where we reclaim all of the pinned space generated by the previous - * two stages. We will not commit the transaction if we don't think we're - * likely to satisfy our request, which means if our current free space + - * total_bytes_pinned < reservation we will not commit. This is why the - * previous states are actually important, to make sure we know for sure - * whether committing the transaction will allow us to make progress. + * This is where we reclaim all of the pinned space generated by running the + * iputs * * ALLOC_CHUNK_FORCE * For data we start with alloc chunk force, however we could have been full @@ -1202,7 +1069,6 @@ static void btrfs_preempt_reclaim_metadata_space(struct work_struct *work) static const enum btrfs_flush_state data_flush_states[] = { FLUSH_DELALLOC_WAIT, RUN_DELAYED_IPUTS, - FLUSH_DELAYED_REFS, COMMIT_TRANS, ALLOC_CHUNK_FORCE, }; @@ -1561,6 +1427,15 @@ static int __reserve_bytes(struct btrfs_fs_info *fs_info, flush == BTRFS_RESERVE_FLUSH_DATA) { list_add_tail(&ticket.list, &space_info->tickets); if (!space_info->flush) { + /* + * We were forced to add a reserve ticket, so + * our preemptive flushing is unable to keep + * up. Clamp down on the threshold for the + * preemptive flushing in order to keep up with + * the workload. + */ + maybe_clamp_preempt(fs_info, space_info); + space_info->flush = 1; trace_btrfs_trigger_flush(fs_info, space_info->flags, @@ -1572,14 +1447,6 @@ static int __reserve_bytes(struct btrfs_fs_info *fs_info, list_add_tail(&ticket.list, &space_info->priority_tickets); } - - /* - * We were forced to add a reserve ticket, so our preemptive - * flushing is unable to keep up. Clamp down on the threshold - * for the preemptive flushing in order to keep up with the - * workload. - */ - maybe_clamp_preempt(fs_info, space_info); } else if (!ret && space_info->flags & BTRFS_BLOCK_GROUP_METADATA) { used += orig_bytes; /* @@ -1588,8 +1455,8 @@ static int __reserve_bytes(struct btrfs_fs_info *fs_info, * the async reclaim as we will panic. */ if (!test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags) && - need_preemptive_reclaim(fs_info, space_info) && - !work_busy(&fs_info->preempt_reclaim_work)) { + !work_busy(&fs_info->preempt_reclaim_work) && + need_preemptive_reclaim(fs_info, space_info)) { trace_btrfs_trigger_flush(fs_info, space_info->flags, orig_bytes, flush, "preempt"); queue_work(system_unbound_wq, |