Hi Jens,
Please pull drbd-8.4.2 into your for-3.8/drivers branch.
The most noticeable change is the support for multiple replicated
volumes in a single DRBD connection.
This release requires new drbd userland tools >= 8.4.0 (available
since July 2011).
8.4.x is network protocol compatible with all previous releaes.
This release brings a new meta-data format. Forward (8.3 -> 8.4)
conversion happens complete seamless. Backward conversion
is done by a single command (drbdadm apply-al res).
The "recent changes" chapter of our user's guide describes all
changes in detail:
http://www.drbd.org/users-guide/ap-recent-changes.html
Changelog
8.4.2 (api:genl1/proto:86-100)
--------
* Go into inconsistent disk state with on-io-error=pass-on policy
* Timeouts for requests processing on the peer (previously that
worked only if the data socket was congested)
* Conflicting write detection is now based on an interval tree,
removed the hash-tables (necessary for the unlimited BIO sizes)
* Support for multiple volumes (minors, block devices) per connection;
up to 65536 volumes per connection supported
* Reduced IO latencies during some state changes (esp. start resync)
* New on disk format for the AL: double capacity; 4k aligned IO; same space
* Multiple AL changes in a single transaction (precondition for
unlimited BIO sizes)
* DRBD no longer imposes any limit on BIO sizes
* Removed DRBD's limits on the number of minor devices
* DRBD's minors can now be removed (not only unconfigured)
* Switched the user space interface form connector to generic netlink
* The wire-protocol is now a regular connection option, which can be
changed while the device is online
* IO freezing/thawing is done on connection (all volumes) level
* fencing is done on connection (all volumes) level
* Enforce application of activity log after primary crash in user space
* New default values (compared to drbd-8.3) for: minor-count, ko-count, al-extents,
c-plan-ahead, c-fill-target, c-min-rate, use-rle, on-io-error
* Optional load balancing for read requests: new keyword "read-balance"
* New option 'al-updates no' to disable writing transactions into the
activity log. It is use full if you prefer a full sync after a primary
crash, for improved performance of a spread out random write work load
* Expose the data generation identifies via sysfs
Jens, regarding the code:
It has the sysfs bits in again. The reason for that is that we want to
expose more information by that, and remove the /proc/drbd with the
next evolutionary step. -- In case this is a show stopper, let me
remove the sysfs bits.
Here is the git-pull-request test:
(The patch subjects removed to make the mail more digestible)
The following changes since commit ccae7868b0c5697508a541c531cf96b361d62c1c:
drbd: log request sector offset and size for IO errors (2012-10-30 08:39:18 +0100)
are available in the git repository at:
git://git.drbd.org/linux-drbd.git for-jens_drbd-8.4.2
for you to fetch changes up to e877f7fdc0b1052b0e881e61f9290268eb21aa2f:
drbd: use copy_highpage (2012-11-09 12:06:44 +0100)
----------------------------------------------------------------
Akinobu Mita (1):
drbd: use copy_highpage
Andreas Gruenbacher (210):
drbd: Get rid of req_validator_fn typedef
[...]
David Howells (1):
DRBD: Fix comparison always false warning due to long/long long compare
Jing Wang (1):
drbd: check return of kmalloc in receive_uuids
Lars Ellenberg (138):
drbd: simplify condition in drbd_may_do_local_read()
[...]
Philipp Marek (1):
drbd: pass some more information to userspace.
Philipp Reisner (234):
idr: idr_for_each_entry() macro
[...]
drivers/block/drbd/Makefile | 2 +
drivers/block/drbd/drbd_actlog.c | 689 +++----
drivers/block/drbd/drbd_bitmap.c | 227 ++-
drivers/block/drbd/drbd_int.h | 1399 ++++++-------
drivers/block/drbd/drbd_interval.c | 207 ++
drivers/block/drbd/drbd_interval.h | 40 +
drivers/block/drbd/drbd_main.c | 3918 ++++++++++++++----------------------
drivers/block/drbd/drbd_nl.c | 3391 ++++++++++++++++++-------------
drivers/block/drbd/drbd_nla.c | 55 +
drivers/block/drbd/drbd_nla.h | 8 +
drivers/block/drbd/drbd_proc.c | 33 +-
drivers/block/drbd/drbd_receiver.c | 3883 ++++++++++++++++++++---------------
drivers/block/drbd/drbd_req.c | 1569 +++++++--------
drivers/block/drbd/drbd_req.h | 187 +-
drivers/block/drbd/drbd_state.c | 1857 +++++++++++++++++
drivers/block/drbd/drbd_state.h | 161 ++
drivers/block/drbd/drbd_strings.c | 1 +
drivers/block/drbd/drbd_sysfs.c | 86 +
drivers/block/drbd/drbd_worker.c | 1168 ++++++-----
drivers/block/drbd/drbd_wrappers.h | 11 +-
include/linux/drbd.h | 81 +-
include/linux/drbd_genl.h | 378 ++++
include/linux/drbd_genl_api.h | 55 +
include/linux/drbd_limits.h | 90 +-
include/linux/drbd_nl.h | 164 --
include/linux/drbd_tag_magic.h | 84 -
include/linux/genl_magic_func.h | 422 ++++
include/linux/genl_magic_struct.h | 277 +++
include/linux/idr.h | 11 +
include/linux/lru_cache.h | 67 +-
lib/lru_cache.c | 359 ++--
31 files changed, 12238 insertions(+), 8642 deletions(-)
create mode 100644 drivers/block/drbd/drbd_interval.c
create mode 100644 drivers/block/drbd/drbd_interval.h
create mode 100644 drivers/block/drbd/drbd_nla.c
create mode 100644 drivers/block/drbd/drbd_nla.h
create mode 100644 drivers/block/drbd/drbd_state.c
create mode 100644 drivers/block/drbd/drbd_state.h
create mode 100644 drivers/block/drbd/drbd_sysfs.c
create mode 100644 include/linux/drbd_genl.h
create mode 100644 include/linux/drbd_genl_api.h
delete mode 100644 include/linux/drbd_nl.h
delete mode 100644 include/linux/drbd_tag_magic.h
create mode 100644 include/linux/genl_magic_func.h
create mode 100644 include/linux/genl_magic_struct.h
Best regards,
Philipp
On 2012-11-09 12:36, Philipp Reisner wrote:
> Hi Jens,
>
> Please pull drbd-8.4.2 into your for-3.8/drivers branch.
>
> The most noticeable change is the support for multiple replicated
> volumes in a single DRBD connection.
>
> This release requires new drbd userland tools >= 8.4.0 (available
> since July 2011).
> 8.4.x is network protocol compatible with all previous releaes.
> This release brings a new meta-data format. Forward (8.3 -> 8.4)
> conversion happens complete seamless. Backward conversion
> is done by a single command (drbdadm apply-al res).
>
> The "recent changes" chapter of our user's guide describes all
> changes in detail:
> http://www.drbd.org/users-guide/ap-recent-changes.html
>
> Changelog
>
> 8.4.2 (api:genl1/proto:86-100)
> --------
> * Go into inconsistent disk state with on-io-error=pass-on policy
> * Timeouts for requests processing on the peer (previously that
> worked only if the data socket was congested)
> * Conflicting write detection is now based on an interval tree,
> removed the hash-tables (necessary for the unlimited BIO sizes)
> * Support for multiple volumes (minors, block devices) per connection;
> up to 65536 volumes per connection supported
> * Reduced IO latencies during some state changes (esp. start resync)
> * New on disk format for the AL: double capacity; 4k aligned IO; same space
> * Multiple AL changes in a single transaction (precondition for
> unlimited BIO sizes)
> * DRBD no longer imposes any limit on BIO sizes
> * Removed DRBD's limits on the number of minor devices
> * DRBD's minors can now be removed (not only unconfigured)
> * Switched the user space interface form connector to generic netlink
> * The wire-protocol is now a regular connection option, which can be
> changed while the device is online
> * IO freezing/thawing is done on connection (all volumes) level
> * fencing is done on connection (all volumes) level
> * Enforce application of activity log after primary crash in user space
> * New default values (compared to drbd-8.3) for: minor-count, ko-count, al-extents,
> c-plan-ahead, c-fill-target, c-min-rate, use-rle, on-io-error
> * Optional load balancing for read requests: new keyword "read-balance"
> * New option 'al-updates no' to disable writing transactions into the
> activity log. It is use full if you prefer a full sync after a primary
> crash, for improved performance of a spread out random write work load
> * Expose the data generation identifies via sysfs
>
>
> Jens, regarding the code:
>
> It has the sysfs bits in again. The reason for that is that we want to
> expose more information by that, and remove the /proc/drbd with the
> next evolutionary step. -- In case this is a show stopper, let me
> remove the sysfs bits.
The exact same sysfs bits I complained about last time? If yes, then I
don't understand why you haven't changed yet. Or why you are trying to
push the same bits again that got rejected last time.
> Here is the git-pull-request test:
> (The patch subjects removed to make the mail more digestible)
Please don't do that, it basically makes the pull request useless! A few
hundred extra lines is not an issue.
--
Jens Axboe
[...]
> > It has the sysfs bits in again. The reason for that is that we want to
> > expose more information by that, and remove the /proc/drbd with the
> > next evolutionary step. -- In case this is a show stopper, let me
> > remove the sysfs bits.
>
> The exact same sysfs bits I complained about last time? If yes, then I
> don't understand why you haven't changed yet. Or why you are trying to
> push the same bits again that got rejected last time.
>
I had the impression it was rejected because I submitted the pull request
too late to you. In the sense of, it might go in, if it gets submitted
for inclusion before the merge window opens...
Apparently my impression was wrong. You will get an updated pull-request
with the sysfs bits removed
> > Here is the git-pull-request test:
> > (The patch subjects removed to make the mail more digestible)
>
> Please don't do that, it basically makes the pull request useless! A few
> hundred extra lines is not an issue.
Ok.
I intend to send the updated pull-request later today.
On 2012-11-09 13:42, Philipp Reisner wrote:
> [...]
>>> It has the sysfs bits in again. The reason for that is that we want to
>>> expose more information by that, and remove the /proc/drbd with the
>>> next evolutionary step. -- In case this is a show stopper, let me
>>> remove the sysfs bits.
>>
>> The exact same sysfs bits I complained about last time? If yes, then I
>> don't understand why you haven't changed yet. Or why you are trying to
>> push the same bits again that got rejected last time.
>>
>
> I had the impression it was rejected because I submitted the pull request
> too late to you. In the sense of, it might go in, if it gets submitted
> for inclusion before the merge window opens...
> Apparently my impression was wrong. You will get an updated pull-request
> with the sysfs bits removed
It was late, but that was a different issue. I just re-read the Oct 3rd
emails on this, and I definitely did recommend that you looked at
debugfs or similar for this. We're not putting another top-queue level
directory in for a block device that is specific to drbd.
--
Jens Axboe
Jens, here it is without the sysfs stuff
The following changes since commit ccae7868b0c5697508a541c531cf96b361d62c1c:
drbd: log request sector offset and size for IO errors (2012-10-30 08:39:18 +0100)
are available in the git repository at:
git://git.drbd.org/linux-drbd.git for-jens_drbd-8.4.2
for you to fetch changes up to f1d6a328bbe63b528721a25251ad8f5f1e997804:
drbd: use copy_highpage (2012-11-09 14:22:26 +0100)
----------------------------------------------------------------
Akinobu Mita (1):
drbd: use copy_highpage
Andreas Gruenbacher (210):
drbd: Get rid of req_validator_fn typedef
drbd: Remove superfluous declaration
drbd: Consistently use block_id == ID_SYNCER for checksum based resync and online verify
drbd: Get rid of BE_DRBD_MAGIC and BE_DRBD_MAGIC_BIG
drbd: Endianness convert the constants instead of the variables
drbd: Magic reserved block_id value cleanup
drbd: Move drbd_free_tl_hash() to drbd_main()
drbd: Update outdated comment
drbd: Request lookup code cleanup (1)
drbd: Request lookup code cleanup (2)
drbd: Request lookup code cleanup (3)
drbd: Request lookup code cleanup (4)
drbd: Add interval tree data structure
drbd: Put sector and size in struct drbd_request into struct drbd_interval
drbd: Use interval tree for overlapping write request detection
drbd: Add read_requests tree
drbd: Use the read and write request trees for request lookups
drbd: Put sector and size in struct drbd_epoch_entry into struct drbd_interval
drbd: Use interval tree for overlapping epoch entry detection
drbd: Remove the unused hash tables
drbd: Convert all constants in enum drbd_req_event to upper case
drbd: Convert all constants in enum drbd_thread_state to upper case
drbd: Replace the ERR_IF macro with an assert-like macro
drbd: Remove some useless paranoia code
drbd: Inline function overlaps() is now unused
drbd: Interval tree bugfix
drbd: Move cmdname() out of drbd_int.h
drbd: Rename "enum drbd_packets" to "enum drbd_packet"
drbd: Remove redundant initialization
drbd: Initialize the sequence number sent over the network even when not used
drbd: Move sequence number logic into drbd_receiver.c and simplify it
drbd: Move some functions to where they are used
drbd: struct drbd_request: Introduce a new collision flag
drbd: Remove redundant check from drbd_contains_interval()
drbd: Allow to wait for the completion of an epoch entry as well
drbd: _req_conflicts(): Get rid of the epoch_entries tree
drbd: Remove unnecessary reference counting left-over
drbd: Defer new writes when detecting conflicting writes
drbd: Make the peer_seq updating code more obvious
drbd: Improve the drbd_find_overlap() documentation
drbd: Remove unused variable in struct drbd_conf
drbd: Rename struct drbd_epoch_entry to struct drbd_peer_request
drbd: Clean up some left-overs
drbd: Update some comments
drbd: Local variable renames: e -> peer_req
drbd: Rename drbd_submit_ee -> drbd_submit_peer_request
drbd: Rename drbd_endio_{pri,sec} -> drbd_{,peer_}request_endio
drbd: Iterate over all overlapping intervals in a tree
drbd: Remove obsolete comment
drbd: Use the IS_ALIGNED() macro in some more places
drbd: Use container_of() instead of casting
drbd: Concurrent write detection fix
drbd: Replace atomic_add_return with atomic_inc_return
drbd: Use ping-timeout when waiting for missing ack packets
drbd: Improve how conflicting writes are handled
drbd: Remove redundant check
drbd: Get rid of P_MAX_CMD
drbd: Replace get_asender_cmd() with its implementation
drbd: Remove left-over function prototypes
drbd: drbd_send(): Return a "real" error code if we have no socket
drbd: drbd_get_data_sock(): Return 0 upon success and an error code otherwise
drbd: Add drbd_send_all(): Send an entire buffer
drbd: conn_send_cmd2(): Return 0 upon success and an error code otherwise
drbd: _conn_send_cmd(): Return 0 upon success and an error code otherwise
drbd: _drbd_send_cmd(): Return 0 upon success and an error code otherwise
drbd: conn_send_cmd(): Return 0 upon success and an error code otherwise
drbd: Get rid of USE_DATA_SOCKET and USE_META_SOCKET
drbd: drbd_send_cmd(): Return 0 upon success and an error code otherwise
drbd: drbd_send_sync_param(): Return 0 upon success and an error code otherwise
drbd: drbd_send_state(): Return 0 upon success and an error code otherwise
drbd: drbd_send_handshake(): Return 0 upon success and an error code otherwise
drbd: drbd_send_protocol(): Return 0 upon success and an error code otherwise
drbd: drbd_send_uuids() and its variants: Return 0 upon success and an error code otherwise
drbd: drbd_gen_and_send_sync_uuid(): Return void: the result is never used
drbd: drbd_send_sizes(): Return 0 upon success and an error code otherwise
drbd: _conn_send_state_req(): Return 0 upon success and an error code otherwise
drbd: conn_send_state_req(): Return 0 upon success and an error code otherwise
drbd: drbd_send_state_req(): Return 0 upon success and an error code otherwise
drbd: drbd_send_sr_reply(): Return void: the result is never used
drbd: drbd_send_b_ack(): Return void: the result is never used
drbd: _drbd_send_ack(): Return 0 upon success and an error code otherwise
drbd: drbd_send_ack(): Return 0 upon success and an error code otherwise
drbd: drbd_send_ack_{dp,rp}(): Return void: the result is never used
drbd: drbd_send_ack_ex(): Return 0 upon success and an error code otherwise
drbd: drbd_send_ov_request(): Return 0 upon success and an error code otherwise
drbd: drbd_send_drequest(): Return 0 upon success and an error code otherwise
drbd: drbd_send_drequest_csum(): Return 0 upon success and an error code otherwise
drbd: drbd_send_oos(): Return 0 upon success and an error code otherwise
drbd: _drbd_no_send_page(): Return 0 upon success and an error code otherwise
drbd: _drbd_send_page(): Return 0 upon success and an error code otherwise
drbd: _drbd_send_zc_ee(): Return 0 upon success and an error code otherwise
drbd: drbd_send_block(): Return 0 upon success and an error code otherwise
drbd: _drbd_send_bio(), _drbd_send_zc_bio(): Return 0 upon success and an error code otherwise
drbd: drbd_send_dblock(): Return 0 upon success and an error code otherwise
drbd: drbd_send_short_cmd(): Return 0 upon success and an error code otherwise
drbd: Temporarily change the return type of all worker callbacks
drbd: Make all worker callbacks return 0 upon success and an error code otherwise
drbd: drbd_process_done_ee(): Return 0 upon success and an error code otherwise
drbd: decode_header(): Return 0 upon success and an error code otherwise
drbd: drbd_recv_header(): Return 0 upon success and an error code otherwise
drbd: drbd_drain_block(): Return 0 upon success and an error code otherwise
drbd: recv_dless_read(): Return 0 upon success and an error code otherwise
drbd: recv_resync_read(): Return 0 upon success and an error code otherwise
drbd: send_bitmap_rle_or_plain(): Error handling cleanup
drbd: Add drbd_recv_all(): Receive an entire buffer
drbd: Make all command handlers return 0 upon success and an error code otherwise
drbd: drbd_bm_read() never returns a positive value through drbd_bitmap_io()
drbd: _drbd_md_sync_page_io(): Return 0 upon success and an error code otherwise
drbd: drbd_md_sync_page_io(): Return 0 upon success and an error code otherwise
drbd: Remove duplicate initialization
drbd: Remove unnecessary assertion
drbd: drbd_may_do_local_read(): Use bool/true/false
drbd: Rename various functions from *_oos_* to *_out_of_sync_* for clarity
drbd: Get rid of typedef drbd_work_cb
drbd: Introduce and use drbd_recv_all_warn()
drbd: drbd_connected(): Return an error code upon failure.
drbd: Always use the same protocol version for the same peer
drbd: Move drbd_send_ping() and drbd_send_ping_ack() to drbd_main.c
drbd: Make _drbd_send_bitmap() static
drbd: Rename the DCBP_* functions to dcbp_* and move them to where they are used
drbd: Preallocate one page per drbd_socket as a receive buffer
drbd: receive_bitmap(): Use the pre-allocated receive buffer
drbd: Preallocate one page per drbd_socket as a send buffer
drbd: _drbd_send_bitmap(): Use the pre-allocated send buffer
drbd: A small cleanup in drbdd()
drbd: Remove useless error messages
drbd: Pass struct packet_info down to the receive functions
drbd: Map from (connection, volume number) to device in the receive handlers
drbd: Pass struct packet_info down to the asender receive functions
drbd: Map from (connection, volume number) to device in the asender handlers
drbd: drbd_connect(): Initialize struct drbd_socket before sending anything
drbd: _conn_send_cmd(), _drbd_send_cmd(): Pass a struct drbd_socket instead of a plain socket
drbd: Change how the initial packets are called
drbd: Change how the "handshake" packets are called
drbd: drbd_send_ping(), drbd_send_ping(): Return 0 upon success and an error code otherwise
drbd: Introduce new primitives for sending commands
drbd: Introduce drbd_header_size()
drbd: Replace and remove old primitives
drbd: Remove now-unused int_dig_out buffer
drbd: Remove some fixed header size assumptions
drbd: Remove headers from on-the-wire data structures (struct p_*)
drbd: Introduce protocol version 100 headers
drbd: Removed outdated comments and code that envisioned VNRs in header 95
drbd: validate_req_change_req_state(): Return 0 upon success and an error code otherwise
drbd: Make all asynchronous command handlers return 0 upon success and an error code otherwise
drbd: drbd_init_ee() no longer exists
drbd: Rename drbd_alloc_ee() to drbd_alloc_peer_req()
drbd: Rename drbd_free_ee() and variants to *_peer_req()
drbd: Rename drbd_release_ee() to drbd_free_peer_reqs()
drbd: Rename reclaim_net_ee(), drbd_process_done_ee(), drbd_process_done_ee(), tconn_process_done_ee() to *_peer_reqs
drbd: Rename drbd_{ ee -> peer_req }_has_active_page
drbd: Make drbd_wait_ee_list_empty() and _drbd_wait_ee_list_empty() static
drbd: Rename drbd_pp_first_pages_or_try_alloc() to __drbd_alloc_pages()
drbd: Rename drbd_pp_alloc() to drbd_alloc_pages() and make it non-static
drbd: Rename drbd_pp_free() to drbd_free_pages()
drbd: Wrong use of RCU in receive_protocol()
drbd: Rename integrity_w_tfm -> integrity_tfm
drbd: Rename integrity_r_tfm -> peer_integrity_tfm
drbd: Remove obsolete drbd_crypto_is_hash()
drbd: Introduce a "lockless" variant of drbd_send_protocoll()
drbd: Allow to change data-integrity-alg on the fly
drbd: Remove left-over unused define
drbd: Generate the drbd_set_*_defaults() functions from drbd_genl.h
drbd: Convert boolean flags on netlink from NLA_FLAG to NLA_U8
drbd: Turn no-disk-flushes into disk-flushes={yes|no}
drbd: Turn no-disk-drain into disk-drain={yes|no}
drbd: Turn no-md-flushes into md-flushes={yes|no}
drbd: Turn no-tcp-cork into tcp-cork={yes|no}
drbd: Make broadcast events return NO_ERROR
drbd: Also define the default values of boolean flags in a single place
drbd: Rename the want_lose field/flag to discard_my_data
drbd: Refer to resync-rate consistently throughout the code
drbd: Refer to connect-int consistently throughout the code
drbd: Fix the upper limit of resync-after
drbd: Convert resync-after into a signed netlink field
drbd: Rename DISK_SIZE_SECT -> DISK_SIZE
drbd: Use more generic constant names
drbd: Output signed / unsigned netlink fields correctly
drbd: Remove unused GENLA_F_MAY_IGNORE flag
drbd: Make drbd's use of netlink attribute flags less confusing
drbd: drbd_nla_check_mandatory(): Need to remove the DRBD_GENLA_F_MANDATORY flag first
drbd: drbd_adm_prepare(): Pass through error codes
drbd: Don't use empty nested netlink attributes
drbd: Use DRBD_MINOR_COUNT_DEF in one more place
drbd: Use the terminology suggested by the command names in the source code and messages
drbd: Also need to check for DRBD_GENLA_F_MANDATORY flags before nla_find_nested()
drbd: Split off netlink mandatory attribute handling into separate file
drbd: Rename DRBD_ADM_NEED_{CONN -> RESOURCE}
drbd: Convert the generic netlink interface to accept connection endpoints
drbd: Allow to pass resource options to the new-resource command
drbd: Remove dead code
drbd: Rename --dry-run to --tentative
drbd: Improve the "unexpected packet" error messages
drbd: Define scale factors in a single place
drbd: Fix the maximum accepted minor device number
drbd: Allow to create devices with a minor number > minor_count
drbd: Print memory address in hex instead of decimal in error message
drbd: receive_protocol(): Give variables more easily searchable names
drbd: receive_protocol(): Make the program flow less confusing
drbd: Be consistent in reporting incompatibilities in P_PROTOCOL settings
drbd: receive_protocol(): We cannot change our own data-integrity-alg setting here
drbd: Turn tl_apply() into tl_abort_disk_io()
drbd: Fix the data-integrity-alg setting
drbd: Get rid of MR_{READ,WRITE}_SHIFT
drbd: Update some outdated comments to match the code
drbd: Improve error reporting in drbd_md_sync_page_io()
drbd: Remove leftover prototype
drbd: drbd_bm_ALe_set_all(): Remove unused function
drbd: Stop using NLA_PUT*().
drbd: Don't unregister socket state_change callback from within the callback
David Howells (1):
DRBD: Fix comparison always false warning due to long/long long compare
Jing Wang (1):
drbd: check return of kmalloc in receive_uuids
Lars Ellenberg (135):
drbd: simplify condition in drbd_may_do_local_read()
drbd: fix typo in comment
drbd: in drbd_suspend_al, set AL_SUSPENDED before unlocking the activity log
drbd: use clear_bit_unlock() where appropriate
lru_cache.h: fix comments referring to ts_ instead of lc_
lru_cache: consolidate lc_get and lc_try_get
drbd: allow to select specific bitmap pages for writeout
lru_cache: allow multiple changes per transaction
drbd: new on-disk activity log transaction format
drbd: silence some log messages on bitmap IO
drbd: fix a wrong likely(), updated comments
drbd: remove unused define
drbd: default to detach on-io-error
drbd: only wakeup if something changed in update_peer_seq
drbd: add page pool to be used for meta data IO
drbd: use the newly introduced page pool for bitmap IO
drbd: introduce a bio_set to allocate housekeeping bios from
drbd: fix drbd_delete_device: remove vnr from volumes; idr_remove(); synchronize_rcu(); before cleanup
drbd: get rid of drbd_bcast_ee, it is of no use anymore
drbd: prepare the transition from connector to genetlink
drbd: switch configuration interface from connector to genetlink
drbd: allow holes in minor and volume id allocation
drbd: remove now unused connector related files
drbd: drbd_adm_get_status needs to show some more detail
drbd: simplify conn_all_vols_unconf, make it bool
drbd: Allow a Diskless Secondary volume to be removed
drbd: new-connection and new-minor succeed, if the object already exists
drbd: bail out if a config requrest is over-determined, and not matching
drbd: add forgotten spin_unlock
drbd: introduce in-kernel "down" command
drbd: fix conn_reconfig_start without conn_reconfig_done in drbd_adm_attach
drbd: distribute former syncer_conf settings to disk, connection, and resource level
drbd: introduce the "initialized" activity log transaction type
drbd: preparation commit, pass drbd_interval to drbd_al_begin/complete_io
drbd: prepare to activate two activity log extents at once
drbd: get rid of bio_split, allow bios of "arbitrary" size
drbd: improvements to activate/deactivate multiple activity log extents
drbd: allow status dump request all volumes of a specific resource
drbd: fix setsockopt for user mode linux
drbd: cmdname() enum to string convertion was missing a few constants
drbd: move comment about stopping the receiver thread to where it belongs
drbd: remove useless kobject_uevent from drbd_adm_connect
drbd: fix various disconnecting races
drbd: on reconfiguration requests, mind the SET_DEFAULTS flag
drbd: rename variable ndc to new_disk_conf
drbd: rename variable sc to res_opts
drbd: fix copy/paste error in comment
drbd: fix race when forcefully disconnecting
drbd: fix thread stop deadlock
drbd: fix schedule in atomic
drbd: Introduce __s32_field in the genetlink macro magic
drbd: skip spurious wait_event in drbd_al_begin_io
drbd: cosmetic: fix accidental division instead of modulo when pretty printing
drbd: spelling fix: too small
drbd: allow ping-timeout of up to 30 seconds
drbd: downgraded error printk to info
drbd: on attach, enforce clean meta data
drbd: report net config even for resources without a single volume
drbd: detach must not try to abort non-local requests
drbd: fix connect failure with all default net-options
drbd: fix bitmap writeout after aborted resync
drbd: add missing rcu locks around recently introduced idr_for_each
drbd: explicitly clear unused dp_flags in drbd_send_block
drbd: Fix a potential race that could case data inconsistency
drbd: Fix a potential write ordering issue on SyncTarget nodes
drbd: Fixed detach
drbd: fix potential data corruption and protocol error
drbd: remove now unused seq_num member from struct drbd_request
drbd: missing wakeup after drbd_rs_del_all
drbd: fix memleak in error path in bm_rw and drbd_bm_write_range
drbd: remove some very outdated comments
drbd: remove unused static helper function
drbd: don't pretend that barrier_nr == 0 was special
drbd: fix potential deadlock during "restart" of conflicting writes
drbd: make OOS_HANDED_TO_NETWORK its own case
drbd: fix READ_RETRY_REMOTE_CANCELED to not complete if device is suspended
drbd: fix WRITE_ACKED_BY_PEER_AND_SIS to not set RQ_NET_DONE
drbd: conflicting writes: make wake_up of waiting peer_requests explicit
drbd: factor out master_bio completion and drbd_request destruction paths
drbd: move put_ldev from __req_mod() to the endio callback
drbd: simplify retry path of failed READ requests
drbd: complete_conflicting_writes() should not care about connections
drbd: ignore volume number for drbd barrier packet exchange
drbd: fix resend/resubmit of frozen IO
drbd: fix race between drbdadm invalidate/verify and finishing resync
drbd: allow bitmap to change during writeout from resync_finished
drbd: bm_page_async_io: properly initialize page->private
drbd: grammar fix in log message
drbd: fix access of unallocated pages and kernel panic
drbd: fix local read error hung forever
drbd: fix wrong assert in completion/retry path of failed local reads
drbd: rename drbd_restart_write to drbd_restart_request
drbd: transfer log epoch numbers are now per resource
drbd: allow to dequeue batches of work at a time
drbd: move the drbd_work_queue from drbd_socket to drbd_connection
drbd: remove struct drbd_tl_epoch objects (barrier works)
drbd: better separate WRITE and READ code paths in drbd_make_request
drbd: __drbd_make_request() is now void
drbd: introduce completion_ref and kref to struct drbd_request
drbd: base completion and destruction of requests on ref counts
drbd: __req_mod: make DISCARD_WRITE and independend case
drbd: allow read requests to be retried after force-detach
drbd: take error path in drbd_adm_down if interrupted by signal
drbd: fix null pointer dereference with on-congestion policy when diskless
drbd: cleanup, remove two unused global flags
drbd: differentiate between normal and forced detach
drbd: report congestion if we are waiting for some userland callback
drbd: reset congestion information before reporting it in /proc/drbd
drbd: do not reset rs_pending_cnt too early
drbd: call local-io-error handler early
drbd: flush drbd work queue before invalidate/invalidate remote
drbd: introduce stop-sector to online verify
drbd: announce FLUSH/FUA capability to upper layers
drbd: fix drbd wire compatibility for empty flushes
drbd: Fix a potential issue with the DISCARD_CONCURRENT flag
drbd: only start a new epoch, if the current epoch contains writes
drbd: NEG_ACK does not imply a barrier-ack
drbd: cleanup, drop unused struct
drbd: disambiguation, s/P_DISCARD_WRITE/P_SUPERSEDED/
drbd: disambiguation, s/DISCARD_CONCURRENT/RESOLVE_CONFLICTS/
drbd: disambiguation, s/ERR_DISCARD/ERR_DISCARD_IMPOSSIBLE/
drbd: properly call drbd_rs_cancel_all() in drbd_disconnected()
drbd: don't send out P_BARRIER with stale information
drbd: temporarily suspend io in drbd_adm_disk_opts
drbd: mutex_unlock "... must no be used in interrupt context"
drbd: dequeue single work items in wait_for_work()
drbd: fix potential list_add corruption
drbd: differentiate early and later "postponing" of requests
drbd: use list_move_tail instead of list_del/list_add_tail
drbd: fix potential deadlock during bitmap (re-)allocation
drbd: a few more GFP_KERNEL -> GFP_NOIO
drbd: wait for meta data IO completion even with failed disk, unless force-detached
drbd: always write bitmap on detach
drbd: log request sector offset and size for IO errors
drbd: if the replication link breaks during handshake, keep retrying
Philipp Marek (1):
drbd: pass some more information to userspace.
Philipp Reisner (232):
idr: idr_for_each_entry() macro
drbd: Minimal struct drbd_tconn
drbd: moved net_conf from mdev to tconn
drbd: moved net_cont and net_cnt_wait from mdev to tconn
drbd: moved data and meta from mdev to tconn
drbd: moved receiver, worker and asender from mdev to tconn
drbd: moved agreed_pro_version, last_received and ko_count to tconn
drbd: moved req_lock and transfer log from mdev to tconn
drbd: moved crypto transformations and friends from mdev to tconn
drbd: Made drbd_flush_workqueue() to take a tconn instead of an mdev
drbd: Preparing to use p_header96 for all packets
drbd: Replaced all p_header80 with a generic p_header
drbd: Use new header layout
drbd: Implemented receiving of new style packets on meta socket
drbd: Do not access tconn after it was freed
drbd: Moved the state functions into its own source file
drbd: Moved the thread name into the data structure
drbd: Eliminated the user of drbd_task_to_thread()
drbd: Moved code
drbd: Do no sleep long in drbd_start_resync
drbd: Revert "Make sure we dont send state if a cluster wide state change is in progress"
drbd: Moving state related macros to drbd_state.h
drbd: conn_printk() a dev_printk() alike for drbd's connections
drbd: Converted drbd_try_connect() from mdev to tconn
drbd: Converted drbd_wait_for_connect() from mdev to tconn
drbd: Started to separated connection flags (tconn) from block device flags (mdev)
drbd: Moved DISCARD_CONCURRENT to the per connection (tconn) flags
drbd: Moved SEND_PING to the per connection (tconn) flags
drbd: Moved SIGNAL_ASENDER to the per connection (tconn) flags
drbd: Converted wake_asender() and request_ping() from mdev to tconn
drbd: Converted helper functions for drbd_send() to tconn
drbd: Converted drbd_send() from mdev to tconn
drbd: Converted drbd_send_fp() from mdev to tconn
drbd: Removed unused mdev argument from drbd_recv_short() and drbd_socket_okay()
drbd: Converted drbd_recv_fp() from mdev to tconn
drbd: Converted drbd_send_handshake() from mdev to tconn
drbd: Converted drbd_recv() from mdev to tconn
drbd: struct packet_info to hold information of decoded packets
drbd: Converted decode_header() from mdev to tconn
drbd: Converted drbd_recv_header() from mdev to tconn
drbd: Converted drbd_do_handshake() from mdev to tconn
drbd: Converted drbd_(get|put)_data_sock() and drbd_send_cmd2() to tconn
drbd: Converted drbd_do_auth() from mdev to tconn
drbd: Converted drbd_send_protocol() from mdev to tconn
drbd: Use and idr data structure to map volume numbers to mdev pointers
drbd: Converted drbd_connect() from mdev to tconn
drbd: Converted drbd_calc_cpu_mask() and drbd_thread_current_set_cpu() from mdev to tconn
drbd: Converted drbdd() from mdev to tconn
drbd: Converted drbd_free_sock() and drbd_disconnect() from mdev to tconn
drbd: Moved the mdev member into drbd_work (from drbd_request and drbd_peer_request)
drbd: Consolidated the setup of the thread name into the framework
drbd: Converted drbdd_init() from mdev to tconn
drbd: Converted drbd_asender() from mdev to tconn
drbd: Converted drbd_worker() from mdev to tconn
drbd: drbd_thread has now a pointer to a tconn instead of to a mdev
drbd: Moved some initializing code into drbd_new_tconn()
drbd: Generalized the work callbacks
drbd: Converted drbd_send_ping() and related functions from mdev to tconn
drbd: Extracted after_conn_state_ch() out of after_state_ch()
drbd: Renamed is_valid_state_transition() to is_valid_soft_transition()
drbd: Extracted is_valid_transition() out of sanitize_state()
drbd: Extracted is_valid_conn_transition() out of is_valid_transition()
drbd: Removed the os parameter form sanitize_state()
drbd: Code de-duplication; new function apply_mask_val()
drbd: Killed volume0; last step of multi-volume-enablement
drbd: Removed drbd_state_lock() and drbd_state_unlock()
drbd: Introduced tconn->cstate_mutex
drbd: Implemented conn_send_state_req()
drbd: Global_state_lock not necessary here...
drbd: Implemented receiving of P_CONN_ST_CHG_REPLY
drbd: implemented receiving of P_CONN_ST_CHG_REQ
drbd: Implemented connection wide state changes
drbd: Reworked the unconfiguring and thread stopping code
drbd: Removed the mdev parameter from the ..to_tags() and ...from_tags() functions
drbd: Improved the dec_*() macros
drbd: Converted the transfer log from mdev to tconn
drbd: Preparing the connector interface to operator on connections
drbd: Converted drbd_nl_(net_conf|disconnect)() from mdev to tconn
drbd: Implemented new commands to create/delete connections/minors
drbd: Replaced the minor_table array by an idr
drbd: Ignore packets for non existing volumes
drbd: process_done_ee() has to handle unconfigured devices now
drbd: New minors have to intherit the connection state form their connection
drbd: Allow two diskless minors to be connected
drbd: Do not modify the connection state with something else that conn_request_state()
drbd: Moved CONN_DRY_RUN to the per connection (tconn) flags
drbd: Allow packet handler functions that take a connection
drbd: Converted receive_protocol() from mdev to tconn
drbd: Separate connection state changes from minor dev state changes #1
drbd: Allow packet handler functions that take a connection (meta connection)
drbd: Converted got_Ping() and got_PingAck() from mdev to tconn
drbd: Separate connection state changes from minor dev state changes #2
drbd: Allow volumes to become primary only on one side
drbd: conn_khelper() for user mode callbacks for connections
drbd: Fixed conn_lowest_minor
drbd: Converted drbd_try_outdate_peer() from mdev to tconn
drbd: Use the idr_for_each_entry() iterator instead of idr_for_each()
drbd: Fixed logging of old connection state
drbd: Print common state changes of all volumes as connection state changes
drbd: Allow to disconnect if one volume is diskless
drbd: Do not segfault if a sync dependency reaches a diskless device
drbd: Introduced drbd_read_state()
drbd: Renamed id_susp(union drbd_state s) to drbd_suspended(struct drbd_conf *)
drbd: Moved susp, susp_nod and susp_fen to the connection object
drbd: Introduced a new type union drbd_dev_state
drbd: Renamed nms to ns_max
drbd: Calculate and provide ns_min to the w_after_conn_state_ch() work
drbd: Implemented conn_lowest_conn()
drbd: Implemented conn_lowest_disk()
drbd: Implemented IO thawing for multiple volumes
drbd: Improved logging of state changes
drbd: Use tconn in request_timer_fn()
drbd: drbd_delete_device() takes a struct drbd_conf * now
drbd: Inlined drbd_free_mdev(); it got called only from one place
drbd: rcu_read_[un]lock() for all idr accesses that do not sleep
drbd: Converted drbd_cfg_mutex into drbd_cfg_rwsem
drbd: protect all idr accesses that might sleep with drbd_cfg_rwsem
drbd: Runtime changeable wire protocol
drbd: Check consistency of net options when the get changed online
drbd: Allow online change of replication protocol only with agreed_pv >= 100
drbd: rcu_read_lock() and rcu_dereference() for tconn->net_conf
drbd: Proper locking for updates to net_conf under RCU
drbd: Considering that the two_primaries config flag can change
drbd: Eliminated drbd_free_resoruces() it is superseeded by conn_free_crypto()
drbd: Basic refcounting for drbd_tconn
drbd: Take a reference on tconn when finding a tconn by name
drbd: Removed the OBJECT_DYING and the CONFIG_PENDING bits
drbd: Consider all crypto options in connect and in net-options
drbd: Removed dead code
drbd: Renamed the net_conf_update mutex to conf_update
drbd: drbd_dew_dev_size() gets the user requests disk_size as argument
drbd: Split drbd_alter_sa() into drbd_sync_after_valid() and drbd_sync_after_changed()
drbd: Renamed (old|new)_conf into (old|new)_net_conf in receive_SyncParam
drbd: RCU for disk_conf
drbd: Made the fifo object a self contained object (preparing for RCU)
drbd: Enforce limits of disk_conf members; centralized these checks
drbd: RCU for rs_plan_s
drbd: Refcounting for mdev objects
drbd: Use RCU for the drbd_tconns list
drbd: Removing drbd_cfg_rwsem
drbd: Refuse to change network options online when...
drbd: protect updates to integrits_tfm by tconn->data->mutex
drbd: Made cmp_after_sb() more generic into convert_after_sb()
drbd: Allocation of int_dig_in and int_dig_vv was missing
drbd: Receiving part for the PROTOCOL_UPDATE packet
drbd: Send PROTOCOL_UPDATE packets when appropriate
drbd: Lower log priority for an event that is definitely not an error
drbd: Fixes from the 8.3 development branch
drbd: Improve compatibility with drbd's older than 8.3.7
drbd: Fixed state transitions in case reading meta data failes
drbd: detach from frozen backing device
drbd: Fixed removal of volumes/devices from connected resources
drbd: Eliminated the "notified peer" messages
drbd: Correctly handle resources without volumes
drbd: Changed some defaults
drbd: Dropped wrong clause to generate new current UUIDs
drbd: Consider that the no-data-condition could be in connected state
drbd: Do not mod_timer() with a past time
drbd: Missing assignment of mdev before drbd_queue_work()
drbd: Fixed w_restart_disk_io() to handle non active AL-extents
drbd: Bugfix for the connection behavior
drbd: Do not display bogus log lines for pdsk in case pdsk < D_UNKNOWN
drbd: The minor_count module parameter is only a hint nowadays
drbd: Cleanup all epoch objects upon connection loss
drbd: Remove dead code
drbd: Remove unused code
drbd: Consider the discard-my-data flag for all volumes [bugz 359]
drbd: fix "stalled" empty resync
drbd: Silenced compiler warnings
drbd: Fixes from the drbd-8.3 branch
drbd: Fixes from the drbd-8.3 branch
drbd: Fixed an obvious copy-n-paste mistake
drbd: Move the CREATE_BARRIER flag from connection to device
drbd: Move write_ordering from mdev to tconn
drbd: Prepare epochs per connection
drbd: Move list of epochs from mdev to tconn
drbd: Switch drbd_may_finish_epoch() from mdev to tconn
drbd: Fix the WO=drain implementation for multiple volumes
drbd: fix potential spinlock deadlock
drbd: Get rid of "ASSERTION FAILED: tconn->current_epoch->list not empty"
drbd: Load balancing of read requests
drbd: Load balancing method: striping
drbd: Do not call generic_make_request() while holding req_lock
drbd: fix race between disconnect and receive_state
drbd: Do not send state packets while lower than C_CONNECTED cstate
drbd: Consider that read requests could be NEG_ACKEDed
drbd: Log failures of connection state changes
drbd: Restore late assigning of tconn->data.sock and meta.sock
drbd: restart loop in drbd_make_request() [prepare for Linux-3.2]
drbd: Fixed compat issue with disconnecting 8.4 from a primary 8.3
drbd: Reinstate disabling AL updates with invalidate-remote
drbd: Consider the disk-timeout also for meta-data IO operations
drbd: If disk timeout expires fail only the affected volume
drbd: Fix module refcount leak in drbd_accept()
drbd: add missing part_round_stats to _drbd_start_io_acct
drbd: Consider that bio->bi_bdev might be modified below DRBD
drbd: Fixed processing of disk-barrier, disk-flushes and disk-drain
drbd: Delay/reject other state changes while establishing a connection
drbd: Ensure that data_size is not 0 before using data_size-1 as index
drbd: fix spelling, remove boring development log message
drbd: use bitmap_parse instead of __bitmap_parse
drbd: Move the call to listen() out of drbd_accept()
drbd: Remove drbd_accept() and use kernel_accept() instead
drbd: New disk option al-updates
drbd: pull prepare_listen_socket() out of drbd_wait_for_connect()
drbd: Keep the listening socket open while trying to connect to the peer
drbd: Remove redundant and wrong test for NULL simplification in conn_connect()
drbd: Try to connec to peer only once per cycle
drbd: More random to the connect logic
drbd: Finish requests that completed while IO was frozen
drbd: Fix completion of requests while the device is suspended
drbd: Write all pages of the bitmap after an online resize
drbd: Protect accesses to the uuid set with a spinlock
drbd: Avoid NetworkFailure state during disconnect
drbd: Remove dead code
drbd: Fix a race condition that can lead to a BUG()
drbd: Fix IO resuming after connection was established while executing the fence handler
drbd: Improve readability of IO resuming after freeze due to no data access
drbd: Do not check aspects that are not subject to change in _conn_requests_state()
drbd: Fix the way the STATE_SENT bit is cleared
drbd: Imporve the error reporting of failed conn state changes
drbd: Fix postponed requests
drbd: Call drbd_md_sync() explicitly after a state change on the connection
drbd: Fix postponed requests
drbd: Remove duplicate code
drbd: Fix comparison of is_valid_transition()'s return code
drbd: panic on delayed completion of aborted requests
drbd: Fix clearing of MDF_AL_DISABLED
drbd: fix regression: potential NULL pointer dereference
drbd: don't try to clear bits once the disk has failed
drbd: Broadcast sync progress no more often than once per second
Merge branch 'drbd-8.4_ed6' into for-3.8-drivers-drbd-8.4_ed6
drivers/block/drbd/Makefile | 2 +
drivers/block/drbd/drbd_actlog.c | 689 +++----
drivers/block/drbd/drbd_bitmap.c | 227 ++-
drivers/block/drbd/drbd_int.h | 1398 ++++++-------
drivers/block/drbd/drbd_interval.c | 207 ++
drivers/block/drbd/drbd_interval.h | 40 +
drivers/block/drbd/drbd_main.c | 3883 ++++++++++++++----------------------
drivers/block/drbd/drbd_nl.c | 3378 ++++++++++++++++++-------------
drivers/block/drbd/drbd_nla.c | 55 +
drivers/block/drbd/drbd_nla.h | 8 +
drivers/block/drbd/drbd_proc.c | 33 +-
drivers/block/drbd/drbd_receiver.c | 3883 +++++++++++++++++++++---------------
drivers/block/drbd/drbd_req.c | 1569 +++++++--------
drivers/block/drbd/drbd_req.h | 187 +-
drivers/block/drbd/drbd_state.c | 1857 +++++++++++++++++
drivers/block/drbd/drbd_state.h | 161 ++
drivers/block/drbd/drbd_strings.c | 1 +
drivers/block/drbd/drbd_worker.c | 1168 ++++++-----
drivers/block/drbd/drbd_wrappers.h | 11 +-
include/linux/drbd.h | 81 +-
include/linux/drbd_genl.h | 378 ++++
include/linux/drbd_genl_api.h | 55 +
include/linux/drbd_limits.h | 90 +-
include/linux/drbd_nl.h | 164 --
include/linux/drbd_tag_magic.h | 84 -
include/linux/genl_magic_func.h | 422 ++++
include/linux/genl_magic_struct.h | 277 +++
include/linux/idr.h | 11 +
include/linux/lru_cache.h | 67 +-
lib/lru_cache.c | 359 ++--
30 files changed, 12115 insertions(+), 8630 deletions(-)
create mode 100644 drivers/block/drbd/drbd_interval.c
create mode 100644 drivers/block/drbd/drbd_interval.h
create mode 100644 drivers/block/drbd/drbd_nla.c
create mode 100644 drivers/block/drbd/drbd_nla.h
create mode 100644 drivers/block/drbd/drbd_state.c
create mode 100644 drivers/block/drbd/drbd_state.h
create mode 100644 include/linux/drbd_genl.h
create mode 100644 include/linux/drbd_genl_api.h
delete mode 100644 include/linux/drbd_nl.h
delete mode 100644 include/linux/drbd_tag_magic.h
create mode 100644 include/linux/genl_magic_func.h
create mode 100644 include/linux/genl_magic_struct.h
On 2012-11-09 14:33, Philipp Reisner wrote:
> Jens, here it is without the sysfs stuff
Thanks, pulled into for-3.8/drivers
--
Jens Axboe
On 2012-11-09 15:18, Jens Axboe wrote:
> On 2012-11-09 14:33, Philipp Reisner wrote:
>> Jens, here it is without the sysfs stuff
>
> Thanks, pulled into for-3.8/drivers
I didn't say anything, but I've been fuming a bit the last few series of
merge windows. You need to stop these insanely massive pull requests.
I've been large since this is "just a driver", but it can't continue. We
should have reached stability a long time ago. Your pull requests
contain a shit load of items, are you guys paying per commit? Look at
these:
drbd: Request lookup code cleanup (1)
drbd: Request lookup code cleanup (2)
drbd: Request lookup code cleanup (3)
drbd: Request lookup code cleanup (4)
or
drbd: conn_send_cmd2(): Return 0 upon success and an error code otherwise
drbd: _conn_send_cmd(): Return 0 upon success and an error code otherwise
drbd: _drbd_send_cmd(): Return 0 upon success and an error code otherwise
drbd: conn_send_cmd(): Return 0 upon success and an error code otherwise
along with FIFTY or so more of these. WTF is this?
drbd: Converted helper functions for drbd_send() to tconn
drbd: Converted drbd_send() from mdev to tconn
drbd: Converted drbd_send_fp() from mdev to tconn
...
I don't think I need to go on. So from now on, to get items into the
kernel, what you will do is:
- Stop doing insane commits like the above. It just doesn't make sense.
- Send pull requests in a timely fashion. No more of this "lets collect
ALL the things" then send it off. Collect small bug fixes, send those
off. Develop some feature or make some changes, send that off. Etc.
The fact that your initial pull request had to MASK all these commits
should have rung big bells in your head. It's a clear sign of a huge
problem in your development model. If you can't clean this up, then
it's not going in.
--
Jens Axboe
Am Freitag, 9. November 2012, 15:50:24 schrieb Jens Axboe:
> On 2012-11-09 15:18, Jens Axboe wrote:
> > On 2012-11-09 14:33, Philipp Reisner wrote:
> >> Jens, here it is without the sysfs stuff
> >
> > Thanks, pulled into for-3.8/drivers
>
> I didn't say anything, but I've been fuming a bit the last few series of
> merge windows. You need to stop these insanely massive pull requests.
> I've been large since this is "just a driver", but it can't continue. We
> should have reached stability a long time ago. Your pull requests
> contain a shit load of items, are you guys paying per commit? Look at
> these:
>
> drbd: Request lookup code cleanup (1)
> drbd: Request lookup code cleanup (2)
> drbd: Request lookup code cleanup (3)
> drbd: Request lookup code cleanup (4)
>
We are living there in the belief that we should break up big
changes in review able chunks....
> or
>
> drbd: conn_send_cmd2(): Return 0 upon success and an error code
> otherwise drbd: _conn_send_cmd(): Return 0 upon success and an error code
> otherwise drbd: _drbd_send_cmd(): Return 0 upon success and an error code
> otherwise drbd: conn_send_cmd(): Return 0 upon success and an error code
> otherwise
>
Function by function gets converted to the "return 0 upon success" call
semantics.
Do you prefer that all of that should be done in a single commit?
> along with FIFTY or so more of these. WTF is this?
>
> drbd: Converted helper functions for drbd_send() to tconn
> drbd: Converted drbd_send() from mdev to tconn
> drbd: Converted drbd_send_fp() from mdev to tconn
>
Should it instead be in a single commit?
>
> I don't think I need to go on. So from now on, to get items into the
> kernel, what you will do is:
>
> - Stop doing insane commits like the above. It just doesn't make sense.
>
> - Send pull requests in a timely fashion. No more of this "lets collect
> ALL the things" then send it off. Collect small bug fixes, send those
> off. Develop some feature or make some changes, send that off. Etc.
That works well for individual features, and we have been doing that
for the last two Years.
But at this time we changed the object model. In the old code we had
a single kind of DRBD-in-kernel-object: a resource
Now we have two kinds: resources and volumes.
8.3: a resource had a single implicit volume
8.4: a resource might contain multiple volumes, each volume belongs to
a single resource.
In the next ~12 month you will get only small features/updates etc... for
the 8.4 code base.
> The fact that your initial pull request had to MASK all these commits
> should have rung big bells in your head. It's a clear sign of a huge
> problem in your development model. If you can't clean this up, then
> it's not going in.
Fundamental changes in our object model require such huge change sets.
Jens, we will not stop where we are today. We plan to introduce a new
object: a connection. (The ability to mirror for one machine to *multiple*
receivers.)
Is it a better fit to introduce it then as a new driver?
E.g. called it "drbd9".
Should it use a new major number?
Best,
Phil
On 2012-11-09 16:19, Philipp Reisner wrote:
> Am Freitag, 9. November 2012, 15:50:24 schrieb Jens Axboe:
>> On 2012-11-09 15:18, Jens Axboe wrote:
>>> On 2012-11-09 14:33, Philipp Reisner wrote:
>>>> Jens, here it is without the sysfs stuff
>>>
>>> Thanks, pulled into for-3.8/drivers
>>
>> I didn't say anything, but I've been fuming a bit the last few series of
>> merge windows. You need to stop these insanely massive pull requests.
>> I've been large since this is "just a driver", but it can't continue. We
>> should have reached stability a long time ago. Your pull requests
>> contain a shit load of items, are you guys paying per commit? Look at
>> these:
>>
>> drbd: Request lookup code cleanup (1)
>> drbd: Request lookup code cleanup (2)
>> drbd: Request lookup code cleanup (3)
>> drbd: Request lookup code cleanup (4)
>>
>
> We are living there in the belief that we should break up big
> changes in review able chunks....
Don't be silly. Of course things should be broken up. But they should be
broken up in logical steps needed to convert from state X to Y. What
does reqeust code cleanup (1) mean? Nothing. Lets take a look at the
commit message:
Move _ar_id_to_req() to drbd_receiver.c and mark it non-inline. Remove
the leading underscores from _ar_id_to_req() and _ack_id_to_req(). Mark
ar_hash_slot() inline.
which is exactly the wrong way as well. A commit message should explain
why a given change is done. The above describes how it's done. We can
all read the code and see that.
In other words, you need to school people on what a commit is, and what
a series of commits is.
>> or
>>
>> drbd: conn_send_cmd2(): Return 0 upon success and an error code
>> otherwise drbd: _conn_send_cmd(): Return 0 upon success and an error code
>> otherwise drbd: _drbd_send_cmd(): Return 0 upon success and an error code
>> otherwise drbd: conn_send_cmd(): Return 0 upon success and an error code
>> otherwise
>>
>
> Function by function gets converted to the "return 0 upon success" call
> semantics.
> Do you prefer that all of that should be done in a single commit?
For this case, yes indeed. Then it'd make more sense to describe why
that change was made. For this particular one, it's fairly easy to guess
why the change was needed.
>> along with FIFTY or so more of these. WTF is this?
>>
>> drbd: Converted helper functions for drbd_send() to tconn
>> drbd: Converted drbd_send() from mdev to tconn
>> drbd: Converted drbd_send_fp() from mdev to tconn
>>
>
> Should it instead be in a single commit?
>
>>
>> I don't think I need to go on. So from now on, to get items into the
>> kernel, what you will do is:
>>
>> - Stop doing insane commits like the above. It just doesn't make sense.
>>
>> - Send pull requests in a timely fashion. No more of this "lets collect
>> ALL the things" then send it off. Collect small bug fixes, send those
>> off. Develop some feature or make some changes, send that off. Etc.
>
> That works well for individual features, and we have been doing that
> for the last two Years.
>
> But at this time we changed the object model. In the old code we had
> a single kind of DRBD-in-kernel-object: a resource
> Now we have two kinds: resources and volumes.
>
> 8.3: a resource had a single implicit volume
>
> 8.4: a resource might contain multiple volumes, each volume belongs to
> a single resource.
>
> In the next ~12 month you will get only small features/updates etc... for
> the 8.4 code base.
If you had just structured it a bit better, then 8.3 -> 8.4 transition
would look a lot less painful. As it stands right now, it's just a huge
mess of little changes.
But I'm happy that you think that the next 12 months will have only
gradual changes for bug fixes and features.
>> The fact that your initial pull request had to MASK all these commits
>> should have rung big bells in your head. It's a clear sign of a huge
>> problem in your development model. If you can't clean this up, then
>> it's not going in.
>
> Fundamental changes in our object model require such huge change sets.
> Jens, we will not stop where we are today. We plan to introduce a new
> object: a connection. (The ability to mirror for one machine to *multiple*
> receivers.)
If you think I'm asking you to stop developing, then you are completely
missing the message. I'm asking you to CHANGE how you develop, or at
least present, these changes. Tons of little changes reak of "hey lets
do this in this function. then lets also do this in that function". A
bit more structure would go a long way. What I usually do is that I may
develop from the first model, then refactor and structure everything to
look a lot nicer. A big part of being able to review a patch series is
understanding where it's going. Step 1 does this, step 2 does that, etc.
If you have 500 changes of "change return code from void to err/0",
well... It may be that you have nice structure underneath it and it's
just hidden behind this mess. If so, it'll be easier for you to make it
better.
> Is it a better fit to introduce it then as a new driver?
> E.g. called it "drbd9".
> Should it use a new major number?
No, I definitely don't want another "here's drbd9" submission. We've
already done that for v8, and it's not been smooth sailing.
Incrementally move things over, make it a more open and logical
development model.
--
Jens Axboe