- 27 Jul, 2012 40 commits
-
-
Joe Thornber authored
Remove some duplicate initialisation of struct dm_pool_metadata. These pmd fields are initialised by both: __format_metadata's calls to dm_btree_empty __write_initial_superblock + __begin_transaction Signed-off-by:
Joe Thornber <ejt@redhat.com> Signed-off-by:
Mike Snitzer <snitzer@redhat.com> Signed-off-by:
Alasdair G Kergon <agk@redhat.com>
-
Joe Thornber authored
Remove 'create' parameter from __create_persistent_data_objects() in dm-thin-metadata. Signed-off-by:
-
Joe Thornber authored
Move the check for __superblock_all_zeroes from __create_persistent_data_objects() down to __open_or_format_metadata in dm-thin-metadata. Signed-off-by:
-
Joe Thornber authored
Remove nr_blocks arg from __create_persistent_data_objects in dm-thin-metadata. It was always passed as zero. Signed-off-by:
-
Joe Thornber authored
Split __open_or_format_metadata into __format_metadata and __open_metadata in dm-thin-metadata. Signed-off-by:
-
Joe Thornber authored
Clean up __open_or_format_metadata in dm-thin-metadata by using struct dm_pool_metadata members to replace local variables. Signed-off-by:
-
Joe Thornber authored
Zero the unused uuid when initialising the metadata superblock. Signed-off-by:
-
Joe Thornber authored
Lift the call to __begin_transaction out of __write_initial_superblock in dm-thin-metadata. Called higher up the call chain now. Signed-off-by:
-
Joe Thornber authored
Move dm_commit_pool_metadata inline into __write_initial_superblock in dm-thin-metadata. Signed-off-by:
-
Joe Thornber authored
Factor out __write_initial_superblock and also pull some other initial creation code out of dm_pool_metadata_open. Signed-off-by:
-
Joe Thornber authored
Lift some initialisation out of __open_or_format_metadata in dm-thin-metadata. Signed-off-by:
-
Joe Thornber authored
Factor __destroy_persistent_data_objects out of dm_pool_metadata_close. Signed-off-by:
-
Joe Thornber authored
Move block manager creation and the check for unformatted metadata into __create_persistent_data_objects(). Signed-off-by:
-
Joe Thornber authored
Rename init_pmd to __create_persistent_data_objects in dm-thin-metadata. Signed-off-by:
-
Joe Thornber authored
Introduce wrappers to handle write locking the superblock appropriately. Signed-off-by:
-
Joe Thornber authored
Stop using dm_bm_unlock_move when shadowing blocks in the transaction manager as an optimisation and remove the function as it is then no longer used. Some code, such as the space maps, keeps using on-disk data structures from the previous transaction. It can do this because blocks won't be reallocated until the subsequent transaction. Using dm_bm_unlock_move to copy blocks sounds like a win, but it forces a synchronous read should the old block be accessed. Signed-off-by:
-
Joe Thornber authored
Tidy the transaction manager creation functions. They no longer lock the superblock. Superblock locking is pulled out to the caller. Also export dm_bm_write_lock_zero. Signed-off-by:
-
Joe Thornber authored
Remove an optimisation that tracks whether or not a thin metadata commit is needed. If dm_pool_commit_metadata() is called and no changes have been made to the metadata then this optimisation avoided writing to disk. Removing because we're going to do something better later. Signed-off-by:
-
Joe Thornber authored
This patch introduces a separate struct for the block_manager. It also uses IS_ERR to check the return value of dm_bufio_client_create instead of testing incorrectly for NULL. Prior to this patch a struct dm_block_manager was really an alias for a struct dm_bufio_client. We want to add some functionality to the block manager that will require extra fields, so this one to one mapping is no longer valid. Signed-off-by:
-
Joe Thornber authored
Factor __setup_btree_details out of init_pmd in dm-thin-metadata. Signed-off-by:
-
Alasdair G Kergon authored
Use boolean bit fields for flags in struct dm_target. Signed-off-by: -
Joe Thornber authored
The thin provisioning target commits internal metadata on flush. So it should receive flushes regardless of whether the underlying devices support them. Signed-off-by:
-
Joe Thornber authored
Allow targets to override the 'supports flush' calculation. Set 'flush_supported' if a target needs to receive flushes regardless of whether or not its underlying devices have support. Signed-off-by:
-
Joe Thornber authored
Introduce bitmap_index_changed to track whether or not the index changed then only commit a space map if it did. Signed-off-by:
-
Joe Thornber authored
Unlock the superblock even if initial dm_bufio_write_dirty_buffers fails. Also, remove redundant flush calls. dm_bm_flush_and_unlock's calls to dm_bufio_write_dirty_buffers already result in dm_bufio_issue_flush being called. This avoids warnings about unflushed dirty buffers from bufio. Signed-off-by:
-
Joe Thornber authored
There's no need to break sharing, triggering a copy, for a write that has no data (i.e. a flush). Signed-off-by:
-
Joe Thornber authored
Fix memory leak in process_prepared_mapping by always freeing the dm_thin_new_mapping structs from the mapping_pool mempool on the error paths. Signed-off-by:
-
Mikulas Patocka authored
Rename sector to cc_sector in dm-crypt's convert_context struct. This is preparation for a future patch that merges dm_io and convert_context which both have a "sector" field. Signed-off-by:
Mikulas Patocka <mpatocka@redhat.com> Signed-off-by:
-
Alasdair G Kergon authored
Store the crypt_config struct pointer directly in struct dm_crypt_io instead of the dm_target struct pointer. Target information is never used - only target->private is referenced, thus we can change it to point directly to struct crypt_config. Signed-off-by:
-
Mikulas Patocka authored
Move static dm-crypt cipher data out of per-cpu structure. Cipher information is static, so it does not have to be in a per-cpu structure. Signed-off-by:
-
Mikulas Patocka authored
There are two dm crypt structures that have a field called "pending". This patch renames them to "cc_pending" and "io_pending" to reduce confusion and ease searching the code. Also remove unnecessary initialisation of r in crypt_convert_block(). Signed-off-by:
-
Jonathan E Brassow authored
In preparation for RAID10 inclusion in dm-raid, we move the sectors_per_dev calculation later in the device creation process. This is because we won't know up-front how many stripes vs how many mirrors there are which will change the calculation. Signed-off-by:
Jonathan Brassow <jbrassow@redhat.com> Signed-off-by:
-
Jonathan E Brassow authored
In preparation for RAID10 addition to dm-raid, we change an 'if' conditional to a 'switch' conditional to make it easier to see what is being checked for each RAID type. Signed-off-by:
-
Mike Snitzer authored
A SCSI device handler might get attached to a device during the initial device scan. We do not necessarily want to override this when loading a multipath table, so this patch adds a new multipath feature argument "retain_attached_hw_handler". During SCSI device scan all loaded SCSI device handlers will be consulted for a match (via scsi_dh's provided .match). If a match is found that device handler will be attached. We need a way to have userspace multipathd's provided 'hw_handler' not override the already attached hardware handler. When specifying the new feature 'retain_attached_hw_handler' multipath will use the currently attached hardware handler instead of trying to attach the one specified during table load. If no hardware handler is attached the specified hardware handler will still be used. Leverages scsi_dh_attach's ability to increment the scsi_dh's reference count if the same scsi_dh name is provided when attaching - currently attached scsi_dh name is determined with scsi_dh_attached_handler_name. Depends upon commit 7e8a74b1 ("[SCSI] scsi_dh: add scsi_dh_attached_handler_name"). Signed-off-by:
Babu Moger <babu.moger@netapp.com> Reviewed-by:
Chandra Seetharaman <sekharan@us.ibm.com> Acked-by:
Hannes Reinecke <hare@suse.de> Signed-off-by:
-
Mikulas Patocka authored
dm-thin will be most likely used with a block size that is a power of two. So it should be optimized for this case. This patch changes division and modulo operations to shifts and bit masks if block size is a power of two. A test that bi_sector is divisible by a block size is removed from io_overlaps_block. Device mapper never sends bios that span a block boundary. Consequently, if we tested that bi_size is equivalent to block size, bi_sector must already be on a block boundary. Signed-off-by:
-
Mikulas Patocka authored
This patch sets the variable "ti->split_discard_requests" for the dm thin target so that device mapper core splits discard requests on a block boundary. Consequently, a discard request that spans multiple blocks is never sent to dm-thin. The patch also removes some code in process_discard that deals with discards that span multiple blocks. Signed-off-by:
-
Mikulas Patocka authored
This patch introduces a new variable split_discard_requests. It can be set by targets so that discard requests are split on max_io_len boundaries. When split_discard_requests is not set, discard requests are only split on boundaries between targets, as was the case before this patch. Signed-off-by:
-
Mike Snitzer authored
Non power of 2 blocksize support is needed to properly align thinp IO on storage that has non power of 2 optimal IO sizes (e.g. RAID6 10+2). Use sector_div to support non power of 2 blocksize for the pool's data device. This provides comparable performance to the power of 2 math that was performed until now (as tested on modern x86_64 hardware). The kernel currently assumes that limits->discard_granularity is a power of two so the thin target only enables discard support if the block size is a power of two. Eliminate pool structure's 'block_shift', 'offset_mask' and remaining 4 byte holes. Signed-off-by:
-
Mikulas Patocka authored
dm-stripe is usually used with a chunk size that is a power of two. Use faster shifts and bit masks in such cases. stripe_width is already optimized in a similar way. Signed-off-by:
-
Mikulas Patocka authored
There is no technical limitation in device mapper that would prevent the dm-stripe target from using a stripe size smaller than page size. This patch removes the limit and makes stripe volumes portable across architectures with different page size. Signed-off-by:
-
