forked from Minki/linux
8ebc423238
This patch is an attempt to remove the Bkl based locking scheme from reiserfs and is intended. It is a bit inspired from an old attempt by Peter Zijlstra: http://lkml.indiana.edu/hypermail/linux/kernel/0704.2/2174.html The bkl is heavily used in this filesystem to prevent from concurrent write accesses on the filesystem. Reiserfs makes a deep use of the specific properties of the Bkl: - It can be acqquired recursively by a same task - It is released on the schedule() calls and reacquired when schedule() returns The two properties above are a roadmap for the reiserfs write locking so it's very hard to simply replace it with a common mutex. - We need a recursive-able locking unless we want to restructure several blocks of the code. - We need to identify the sites where the bkl was implictly relaxed (schedule, wait, sync, etc...) so that we can in turn release and reacquire our new lock explicitly. Such implicit releases of the lock are often required to let other resources producer/consumer do their job or we can suffer unexpected starvations or deadlocks. So the new lock that replaces the bkl here is a per superblock mutex with a specific property: it can be acquired recursively by a same task, like the bkl. For such purpose, we integrate a lock owner and a lock depth field on the superblock information structure. The first axis on this patch is to turn reiserfs_write_(un)lock() function into a wrapper to manage this mutex. Also some explicit calls to lock_kernel() have been converted to reiserfs_write_lock() helpers. The second axis is to find the important blocking sites (schedule...(), wait_on_buffer(), sync_dirty_buffer(), etc...) and then apply an explicit release of the write lock on these locations before blocking. Then we can safely wait for those who can give us resources or those who need some. Typically this is a fight between the current writer, the reiserfs workqueue (aka the async commiter) and the pdflush threads. The third axis is a consequence of the second. The write lock is usually on top of a lock dependency chain which can include the journal lock, the flush lock or the commit lock. So it's dangerous to release and trying to reacquire the write lock while we still hold other locks. This is fine with the bkl: T1 T2 lock_kernel() mutex_lock(A) unlock_kernel() // do something lock_kernel() mutex_lock(A) -> already locked by T1 schedule() (and then unlock_kernel()) lock_kernel() mutex_unlock(A) .... This is not fine with a mutex: T1 T2 mutex_lock(write) mutex_lock(A) mutex_unlock(write) // do something mutex_lock(write) mutex_lock(A) -> already locked by T1 schedule() mutex_lock(write) -> already locked by T2 deadlock The solution in this patch is to provide a helper which releases the write lock and sleep a bit if we can't lock a mutex that depend on it. It's another simulation of the bkl behaviour. The last axis is to locate the fs callbacks that are called with the bkl held, according to Documentation/filesystem/Locking. Those are: - reiserfs_remount - reiserfs_fill_super - reiserfs_put_super Reiserfs didn't need to explicitly lock because of the context of these callbacks. But now we must take care of that with the new locking. After this patch, reiserfs suffers from a slight performance regression (for now). On UP, a high volume write with dd reports an average of 27 MB/s instead of 30 MB/s without the patch applied. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Reviewed-by: Ingo Molnar <mingo@elte.hu> Cc: Jeff Mahoney <jeffm@suse.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Bron Gondwana <brong@fastmail.fm> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> LKML-Reference: <1239070789-13354-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> |
||
|---|---|---|
| .. | ||
| bitmap.c | ||
| dir.c | ||
| do_balan.c | ||
| file.c | ||
| fix_node.c | ||
| hashes.c | ||
| ibalance.c | ||
| inode.c | ||
| ioctl.c | ||
| item_ops.c | ||
| journal.c | ||
| Kconfig | ||
| lbalance.c | ||
| lock.c | ||
| Makefile | ||
| namei.c | ||
| objectid.c | ||
| prints.c | ||
| procfs.c | ||
| README | ||
| resize.c | ||
| stree.c | ||
| super.c | ||
| tail_conversion.c | ||
| xattr_acl.c | ||
| xattr_security.c | ||
| xattr_trusted.c | ||
| xattr_user.c | ||
| xattr.c | ||
[LICENSING] ReiserFS is hereby licensed under the GNU General Public License version 2. Source code files that contain the phrase "licensing governed by reiserfs/README" are "governed files" throughout this file. Governed files are licensed under the GPL. The portions of them owned by Hans Reiser, or authorized to be licensed by him, have been in the past, and likely will be in the future, licensed to other parties under other licenses. If you add your code to governed files, and don't want it to be owned by Hans Reiser, put your copyright label on that code so the poor blight and his customers can keep things straight. All portions of governed files not labeled otherwise are owned by Hans Reiser, and by adding your code to it, widely distributing it to others or sending us a patch, and leaving the sentence in stating that licensing is governed by the statement in this file, you accept this. It will be a kindness if you identify whether Hans Reiser is allowed to license code labeled as owned by you on your behalf other than under the GPL, because he wants to know if it is okay to do so and put a check in the mail to you (for non-trivial improvements) when he makes his next sale. He makes no guarantees as to the amount if any, though he feels motivated to motivate contributors, and you can surely discuss this with him before or after contributing. You have the right to decline to allow him to license your code contribution other than under the GPL. Further licensing options are available for commercial and/or other interests directly from Hans Reiser: hans@reiser.to. If you interpret the GPL as not allowing those additional licensing options, you read it wrongly, and Richard Stallman agrees with me, when carefully read you can see that those restrictions on additional terms do not apply to the owner of the copyright, and my interpretation of this shall govern for this license. Finally, nothing in this license shall be interpreted to allow you to fail to fairly credit me, or to remove my credits, without my permission, unless you are an end user not redistributing to others. If you have doubts about how to properly do that, or about what is fair, ask. (Last I spoke with him Richard was contemplating how best to address the fair crediting issue in the next GPL version.) [END LICENSING] Reiserfs is a file system based on balanced tree algorithms, which is described at http://devlinux.com/namesys. Stop reading here. Go there, then return. Send bug reports to yura@namesys.botik.ru. mkreiserfs and other utilities are in reiserfs/utils, or wherever your Linux provider put them. There is some disagreement about how useful it is for users to get their fsck and mkreiserfs out of sync with the version of reiserfs that is in their kernel, with many important distributors wanting them out of sync.:-) Please try to remember to recompile and reinstall fsck and mkreiserfs with every update of reiserfs, this is a common source of confusion. Note that some of the utilities cannot be compiled without accessing the balancing code which is in the kernel code, and relocating the utilities may require you to specify where that code can be found. Yes, if you update your reiserfs kernel module you do have to recompile your kernel, most of the time. The errors you get will be quite cryptic if your forget to do so. Real users, as opposed to folks who want to hack and then understand what went wrong, will want REISERFS_CHECK off. Hideous Commercial Pitch: Spread your development costs across other OS vendors. Select from the best in the world, not the best in your building, by buying from third party OS component suppliers. Leverage the software component development power of the internet. Be the most aggressive in taking advantage of the commercial possibilities of decentralized internet development, and add value through your branded integration that you sell as an operating system. Let your competitors be the ones to compete against the entire internet by themselves. Be hip, get with the new economic trend, before your competitors do. Send email to hans@reiser.to. To understand the code, after reading the website, start reading the code by reading reiserfs_fs.h first. Hans Reiser was the project initiator, primary architect, source of all funding for the first 5.5 years, and one of the programmers. He owns the copyright. Vladimir Saveljev was one of the programmers, and he worked long hours writing the cleanest code. He always made the effort to be the best he could be, and to make his code the best that it could be. What resulted was quite remarkable. I don't think that money can ever motivate someone to work the way he did, he is one of the most selfless men I know. Yura helps with benchmarking, coding hashes, and block pre-allocation code. Anatoly Pinchuk is a former member of our team who worked closely with Vladimir throughout the project's development. He wrote a quite substantial portion of the total code. He realized that there was a space problem with packing tails of files for files larger than a node that start on a node aligned boundary (there are reasons to want to node align files), and he invented and implemented indirect items and unformatted nodes as the solution. Konstantin Shvachko, with the help of the Russian version of a VC, tried to put me in a position where I was forced into giving control of the project to him. (Fortunately, as the person paying the money for all salaries from my dayjob I owned all copyrights, and you can't really force takeovers of sole proprietorships.) This was something curious, because he never really understood the value of our project, why we should do what we do, or why innovation was possible in general, but he was sure that he ought to be controlling it. Every innovation had to be forced past him while he was with us. He added two years to the time required to complete reiserfs, and was a net loss for me. Mikhail Gilula was a brilliant innovator who also left in a destructive way that erased the value of his contributions, and that he was shown much generosity just makes it more painful. Grigory Zaigralin was an extremely effective system administrator for our group. Igor Krasheninnikov was wonderful at hardware procurement, repair, and network installation. Jeremy Fitzhardinge wrote the teahash.c code, and he gives credit to a textbook he got the algorithm from in the code. Note that his analysis of how we could use the hashing code in making 32 bit NFS cookies work was probably more important than the actual algorithm. Colin Plumb also contributed to it. Chris Mason dived right into our code, and in just a few months produced the journaling code that dramatically increased the value of ReiserFS. He is just an amazing programmer. Igor Zagorovsky is writing much of the new item handler and extent code for our next major release. Alexander Zarochentcev (sometimes known as zam, or sasha), wrote the resizer, and is hard at work on implementing allocate on flush. SGI implemented allocate on flush before us for XFS, and generously took the time to convince me we should do it also. They are great people, and a great company. Yuri Shevchuk and Nikita Danilov are doing squid cache optimization. Vitaly Fertman is doing fsck. Jeff Mahoney, of SuSE, contributed a few cleanup fixes, most notably the endian safe patches which allow ReiserFS to run on any platform supported by the Linux kernel. SuSE, IntegratedLinux.com, Ecila, MP3.com, bigstorage.com, and the Alpha PC Company made it possible for me to not have a day job anymore, and to dramatically increase our staffing. Ecila funded hypertext feature development, MP3.com funded journaling, SuSE funded core development, IntegratedLinux.com funded squid web cache appliances, bigstorage.com funded HSM, and the alpha PC company funded the alpha port. Many of these tasks were helped by sponsors other than the ones just named. SuSE has helped in much more than just funding....