view fs/jffs2/README.Locking @ 912:dd42cdb0ab89

[IA64] Build blktap2 driver by default in x86 builds.

add CONFIG_XEN_BLKDEV_TAP2=y to buildconfigs/linux-defconfig_xen_ia64.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
author Isaku Yamahata <yamahata@valinux.co.jp>
date Mon Jun 29 12:09:16 2009 +0900 (2009-06-29)
parents 831230e53067
line source
1 $Id: README.Locking,v 1.12 2005/04/13 13:22:35 dwmw2 Exp $
4 ---------------------------
6 At least theoretically, JFFS2 does not require the Big Kernel Lock
7 (BKL), which was always helpfully obtained for it by Linux 2.4 VFS
8 code. It has its own locking, as described below.
10 This document attempts to describe the existing locking rules for
11 JFFS2. It is not expected to remain perfectly up to date, but ought to
12 be fairly close.
15 alloc_sem
16 ---------
18 The alloc_sem is a per-filesystem semaphore, used primarily to ensure
19 contiguous allocation of space on the medium. It is automatically
20 obtained during space allocations (jffs2_reserve_space()) and freed
21 upon write completion (jffs2_complete_reservation()). Note that
22 the garbage collector will obtain this right at the beginning of
23 jffs2_garbage_collect_pass() and release it at the end, thereby
24 preventing any other write activity on the file system during a
25 garbage collect pass.
27 When writing new nodes, the alloc_sem must be held until the new nodes
28 have been properly linked into the data structures for the inode to
29 which they belong. This is for the benefit of NAND flash - adding new
30 nodes to an inode may obsolete old ones, and by holding the alloc_sem
31 until this happens we ensure that any data in the write-buffer at the
32 time this happens are part of the new node, not just something that
33 was written afterwards. Hence, we can ensure the newly-obsoleted nodes
34 don't actually get erased until the write-buffer has been flushed to
35 the medium.
37 With the introduction of NAND flash support and the write-buffer,
38 the alloc_sem is also used to protect the wbuf-related members of the
39 jffs2_sb_info structure. Atomically reading the wbuf_len member to see
40 if the wbuf is currently holding any data is permitted, though.
42 Ordering constraints: See f->sem.
45 File Semaphore f->sem
46 ---------------------
48 This is the JFFS2-internal equivalent of the inode semaphore i->i_sem.
49 It protects the contents of the jffs2_inode_info private inode data,
50 including the linked list of node fragments (but see the notes below on
51 erase_completion_lock), etc.
53 The reason that the i_sem itself isn't used for this purpose is to
54 avoid deadlocks with garbage collection -- the VFS will lock the i_sem
55 before calling a function which may need to allocate space. The
56 allocation may trigger garbage-collection, which may need to move a
57 node belonging to the inode which was locked in the first place by the
58 VFS. If the garbage collection code were to attempt to lock the i_sem
59 of the inode from which it's garbage-collecting a physical node, this
60 lead to deadlock, unless we played games with unlocking the i_sem
61 before calling the space allocation functions.
63 Instead of playing such games, we just have an extra internal
64 semaphore, which is obtained by the garbage collection code and also
65 by the normal file system code _after_ allocation of space.
67 Ordering constraints:
69 1. Never attempt to allocate space or lock alloc_sem with
70 any f->sem held.
71 2. Never attempt to lock two file semaphores in one thread.
72 No ordering rules have been made for doing so.
75 erase_completion_lock spinlock
76 ------------------------------
78 This is used to serialise access to the eraseblock lists, to the
79 per-eraseblock lists of physical jffs2_raw_node_ref structures, and
80 (NB) the per-inode list of physical nodes. The latter is a special
81 case - see below.
83 As the MTD API no longer permits erase-completion callback functions
84 to be called from bottom-half (timer) context (on the basis that nobody
85 ever actually implemented such a thing), it's now sufficient to use
86 a simple spin_lock() rather than spin_lock_bh().
88 Note that the per-inode list of physical nodes (f->nodes) is a special
89 case. Any changes to _valid_ nodes (i.e. ->flash_offset & 1 == 0) in
90 the list are protected by the file semaphore f->sem. But the erase
91 code may remove _obsolete_ nodes from the list while holding only the
92 erase_completion_lock. So you can walk the list only while holding the
93 erase_completion_lock, and can drop the lock temporarily mid-walk as
94 long as the pointer you're holding is to a _valid_ node, not an
95 obsolete one.
97 The erase_completion_lock is also used to protect the c->gc_task
98 pointer when the garbage collection thread exits. The code to kill the
99 GC thread locks it, sends the signal, then unlocks it - while the GC
100 thread itself locks it, zeroes c->gc_task, then unlocks on the exit path.
103 inocache_lock spinlock
104 ----------------------
106 This spinlock protects the hashed list (c->inocache_list) of the
107 in-core jffs2_inode_cache objects (each inode in JFFS2 has the
108 correspondent jffs2_inode_cache object). So, the inocache_lock
109 has to be locked while walking the c->inocache_list hash buckets.
111 This spinlock also covers allocation of new inode numbers, which is
112 currently just '++->highest_ino++', but might one day get more complicated
113 if we need to deal with wrapping after 4 milliard inode numbers are used.
115 Note, the f->sem guarantees that the correspondent jffs2_inode_cache
116 will not be removed. So, it is allowed to access it without locking
117 the inocache_lock spinlock.
119 Ordering constraints:
121 If both erase_completion_lock and inocache_lock are needed, the
122 c->erase_completion has to be acquired first.
125 erase_free_sem
126 --------------
128 This semaphore is only used by the erase code which frees obsolete
129 node references and the jffs2_garbage_collect_deletion_dirent()
130 function. The latter function on NAND flash must read _obsolete_ nodes
131 to determine whether the 'deletion dirent' under consideration can be
132 discarded or whether it is still required to show that an inode has
133 been unlinked. Because reading from the flash may sleep, the
134 erase_completion_lock cannot be held, so an alternative, more
135 heavyweight lock was required to prevent the erase code from freeing
136 the jffs2_raw_node_ref structures in question while the garbage
137 collection code is looking at them.
139 Suggestions for alternative solutions to this problem would be welcomed.
142 wbuf_sem
143 --------
145 This read/write semaphore protects against concurrent access to the
146 write-behind buffer ('wbuf') used for flash chips where we must write
147 in blocks. It protects both the contents of the wbuf and the metadata
148 which indicates which flash region (if any) is currently covered by
149 the buffer.
151 Ordering constraints:
152 Lock wbuf_sem last, after the alloc_sem or and f->sem.
155 c->xattr_sem
156 ------------
158 This read/write semaphore protects against concurrent access to the
159 xattr related objects which include stuff in superblock and ic->xref.
160 In read-only path, write-semaphore is too much exclusion. It's enough
161 by read-semaphore. But you must hold write-semaphore when updating,
162 creating or deleting any xattr related object.
164 Once xattr_sem released, there would be no assurance for the existence
165 of those objects. Thus, a series of processes is often required to retry,
166 when updating such a object is necessary under holding read semaphore.
167 For example, do_jffs2_getxattr() holds read-semaphore to scan xref and
168 xdatum at first. But it retries this process with holding write-semaphore
169 after release read-semaphore, if it's necessary to load name/value pair
170 from medium.
172 Ordering constraints:
173 Lock xattr_sem last, after the alloc_sem.