Total BKL removal, finally
Arnd Bergmann
Acked-by Alan Cox Acked-by Ingo Molnar Acked-by Thomas Gleixner Acked-by David S. Miller Acked-by David S. Miller Acked-by David S. Miller Tested-by Andrew Hendry Tested-by Nick Bowler Acked-by Andi Kleen Acked-by Frederic Weisbecker Acked-by Russell KingHi Linus,
Here are the missing pieces of the BKL removal, aside from the bits that are in the staging and drm trees. You can pull the trees in any order, there should be no interdependencies.
There was one conflict between this tree and the nfs tree a few days ago, because an #include <linux/smp_lock.h> was accidentally added back there, but the file removed here. If that happens with any other tree, the fix should be to just remove the offending #include. All patches that are in linux-next today are fine in this regard.
Thanks,
Arnd
8<-------------------------------------------------------------------
The following changes since commit f5412be599602124d2bdd49947b231dd77c0bf99:
Linux 2.6.38-rc6 (2011-02-21 17:25:52 -0800)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/arnd/bkl.git config
Arnd Bergmann (9):
adfs: remove the big kernel lock
tracing: don't trace the BKL
drivers: remove extraneous includes of smp_lock.h
hpfs: remove the BKL
ufs: remove the BKL
x25: remove the BKL
appletalk: remove the BKL
ipx: remove the BKL
BKL: That's all, folks
Matt Fleming (1):
fs/locks.c: Remove stale FIXME left over from BKL conversion
8<-------------------------------------------------------------------
commit 4ba8216cd90560bc402f52076f64d8546e8aefcb
Author: Arnd Bergmann <[email protected]> Date: Tue Jan 25 22:52:22 2011 +0100
BKL: That's all, folks
This removes the implementation of the big kernel lock, at last. A lot of people have worked on this in the past, I so the credit for this patch should be with everyone who participated in the hunt.
The names on the Cc list are the people that were the most active in this, according to the recorded git history, in alphabetical order.
include/linux/hardirq.h | 9 +--- include/linux/smp_lock.h | 65 ---------------------- init/Kconfig | 5 -- kernel/sched.c | 9 +--- lib/Kconfig.debug | 9 --- lib/Makefile | 1 - lib/kernel_lock.c | 136 ---------------------------------------------- 7 files changed, 2 insertions(+), 232 deletions(-)
commit ae7eb8979ccfa5e9e888101b9c940f20bd0f4115
Author: Matt Fleming <[email protected]> Date: Sun Feb 27 13:58:00 2011 +0000
fs/locks.c: Remove stale FIXME left over from BKL conversion
The comment is no longer true as (now that the BKL conversion is finished) a spinlock _is_ now used to protect file_lock_list, blocked_list and inode->i_flock.
fs/locks.c | 1 - 1 files changed, 0 insertions(+), 1 deletions(-)
commit b0d0d915d1d1a0fe486849f3e41333c66df620c4
Author: Arnd Bergmann <[email protected]> Date: Tue Jan 25 21:49:56 2011 +0100
ipx: remove the BKL
This replaces all instances of lock_kernel in the IPX code with lock_sock. As far as I can tell, this is safe to do, because there is no global state that needs to be locked in IPX, and the code does not recursively take the lock or sleep indefinitely while holding it.
Compile-tested only.
net/ipx/Kconfig | 1 - net/ipx/af_ipx.c | 52 ++++++++++++++++++++-------------------------------- 2 files changed, 20 insertions(+), 33 deletions(-)
commit 60d9f461a20ba59219fdcdc30cbf8e3a4ad3f625
Author: Arnd Bergmann <[email protected]> Date: Sun Jan 23 00:21:11 2011 +0100
appletalk: remove the BKL
This changes appletalk to use lock_sock instead of lock_kernel for serialization. I tried to make sure that we don't hold the socket lock during sleeping functions, but I did not try to prove whether the locks are necessary in the first place.
Compile-tested only.
drivers/net/appletalk/Kconfig | 1 - net/appletalk/ddp.c | 40 ++++++++++++++++------------------------ 2 files changed, 16 insertions(+), 25 deletions(-)
commit 77b2283604bdd7053494a97b0e2fee97148206c6
Author: Arnd Bergmann <[email protected]> Date: Sat Jan 22 23:44:59 2011 +0100
x25: remove the BKL
This replaces all instances of lock_kernel in x25 with lock_sock, taking care to release the socket lock around sleeping functions (sock_alloc_send_skb and skb_recv_datagram). It is not clear whether this is a correct solution, but it seem to be what other protocols do in the same situation.
Includes a fix suggested by Eric Dumazet.
net/x25/Kconfig | 1 - net/x25/af_x25.c | 58 ++++++++++++++++------------------------------------ net/x25/x25_out.c | 7 ++++- 3 files changed, 23 insertions(+), 43 deletions(-)
commit 788257d6101d986ac8f2741aaa35974af47f574c
Author: Arnd Bergmann <[email protected]> Date: Mon Jan 24 10:14:12 2011 +0100
ufs: remove the BKL
This introduces a new per-superblock mutex in UFS to replace the big kernel lock. I have been careful to avoid nested calls to lock_ufs and to get the lock order right with respect to other mutexes, in particular lock_super.
I did not make any attempt to prove that the big kernel lock is not needed in a particular place in the code, which is very possible.
The mutex has a significant performance impact, so it is only used on SMP or PREEMPT configurations.
As Nick Piggin noticed, any allocation inside of the lock may end up deadlocking when we get to ufs_getfrag_block in the reclaim task, so we now use GFP_NOFS.
fs/ufs/Kconfig | 1 - fs/ufs/inode.c | 78 ++++++++++++++-------------------------------------- fs/ufs/namei.c | 35 +++++++++++------------ fs/ufs/super.c | 64 +++++++++++++++++++++++++------------------ fs/ufs/truncate.c | 5 +-- fs/ufs/ufs.h | 6 +++- fs/ufs/util.c | 2 +- 7 files changed, 83 insertions(+), 108 deletions(-)
commit 9a311b96c3065f362e3348cb5d7af1a57ca6bff9
Author: Arnd Bergmann <[email protected]> Date: Sat Jan 22 20:26:12 2011 +0100
hpfs: remove the BKL
This removes the BKL in hpfs in a rather awful way, by making the code only work on uniprocessor systems without kernel preemption, as suggested by Andi Kleen.
The HPFS code probably has close to zero remaining users on current kernels, all archeological uses of the file system can probably be done with the significant restrictions.
The hpfs_lock/hpfs_unlock functions are left in the code, sincen Mikulas has indicated that he is still interested in fixing it in a better way.
fs/hpfs/Kconfig | 2 +- fs/hpfs/dir.c | 23 +++++++++++------------ fs/hpfs/file.c | 9 ++++----- fs/hpfs/hpfs_fn.h | 22 ++++++++++++++++++++++ fs/hpfs/inode.c | 9 ++++----- fs/hpfs/namei.c | 49 ++++++++++++++++++++++++------------------------- fs/hpfs/super.c | 23 +++++++++-------------- 7 files changed, 75 insertions(+), 62 deletions(-)
commit 5edc341313a188d94cde7ef87ac31647cea8601a
Author: Arnd Bergmann <[email protected]> Date: Tue Jan 25 22:08:05 2011 +0100
drivers: remove extraneous includes of smp_lock.h
These were missed the last time I cleaned this up globally, because of code moving around or new code getting merged.
drivers/scsi/megaraid/megaraid_sas_fp.c | 1 - drivers/scsi/megaraid/megaraid_sas_fusion.c | 1 - drivers/staging/easycap/easycap_ioctl.c | 1 - drivers/target/target_core_device.c | 1 - drivers/target/target_core_fabric_lib.c | 1 - drivers/target/target_core_file.c | 1 - drivers/target/target_core_hba.c | 1 - drivers/target/target_core_iblock.c | 1 - drivers/target/target_core_pscsi.c | 1 - drivers/target/target_core_rd.c | 1 - drivers/target/target_core_tpg.c | 1 - drivers/target/target_core_transport.c | 1 - drivers/tty/n_hdlc.c | 1 - drivers/tty/n_r3964.c | 1 - drivers/tty/pty.c | 1 - drivers/tty/tty_io.c | 1 - drivers/tty/tty_ldisc.c | 2 -- drivers/tty/vt/selection.c | 1 - drivers/tty/vt/vc_screen.c | 1 - drivers/tty/vt/vt.c | 1 - drivers/tty/vt/vt_ioctl.c | 1 - 21 files changed, 0 insertions(+), 22 deletions(-)
commit f51b452bed4ae5c20e1f8a790e4ed8663d909a40
Author: Arnd Bergmann <[email protected]> Date: Tue Jan 25 21:54:50 2011 +0100
tracing: don't trace the BKL
No reason to trace it when the last user is gone.
include/trace/events/bkl.h | 61 -------------------------------------------- lib/kernel_lock.c | 7 ----- 2 files changed, 0 insertions(+), 68 deletions(-)
commit 4688a066ecf60086ea82f68edb3b036b567d2c08
Author: Arnd Bergmann <[email protected]> Date: Sat Jan 22 20:05:05 2011 +0100
adfs: remove the big kernel lock
According to Russell King, adfs was written to not require the big kernel lock, and all inode updates are done under adfs_dir_lock.
All other metadata in adfs is read-only and does not require locking. The use of the BKL is the result of various pushdowns from the VFS operations.
fs/adfs/Kconfig | 1 - fs/adfs/dir.c | 6 ------ fs/adfs/inode.c | 6 ------ fs/adfs/super.c | 13 +------------ 4 files changed, 1 insertions(+), 25 deletions(-)
Quoted 5 lines · Arnd Bergmann
commit
4ba8216cd90560bc402f52076f64d8546e8aefcbAuthor: Arnd Bergmann <[email protected]> Date: Tue Jan 25 22:52:22 2011 +0100BKL: That's all, folks
All I can say is "Yay!"
Nice work Arnd.
-- Steve
Quoted 10 lines
This removes the implementation of the big kernel lock, at last. A lot of people have worked on this in the past, I so the credit for this patch should be with everyone who participated in the hunt.
The names on the Cc list are the people that were the most active in this, according to the recorded git history, in alphabetical order.