diff options
Diffstat (limited to 'static/netbsd/man9/wapbl.9 3.html')
| -rw-r--r-- | static/netbsd/man9/wapbl.9 3.html | 424 |
1 files changed, 0 insertions, 424 deletions
diff --git a/static/netbsd/man9/wapbl.9 3.html b/static/netbsd/man9/wapbl.9 3.html deleted file mode 100644 index 241ea757..00000000 --- a/static/netbsd/man9/wapbl.9 3.html +++ /dev/null @@ -1,424 +0,0 @@ -<table class="head"> - <tr> - <td class="head-ltitle">WAPBL(9)</td> - <td class="head-vol">Kernel Developer's Manual</td> - <td class="head-rtitle">WAPBL(9)</td> - </tr> -</table> -<div class="manual-text"> -<section class="Sh"> -<h1 class="Sh" id="NAME"><a class="permalink" href="#NAME">NAME</a></h1> -<p class="Pp"><code class="Nm">WAPBL</code>, - <code class="Nm">wapbl_start</code>, <code class="Nm">wapbl_stop</code>, - <code class="Nm">wapbl_begin</code>, <code class="Nm">wapbl_end</code>, - <code class="Nm">wapbl_flush</code>, <code class="Nm">wapbl_discard</code>, - <code class="Nm">wapbl_add_buf</code>, - <code class="Nm">wapbl_remove_buf</code>, - <code class="Nm">wapbl_resize_buf</code>, - <code class="Nm">wapbl_register_inode</code>, - <code class="Nm">wapbl_unregister_inode</code>, - <code class="Nm">wapbl_register_deallocation</code>, - <code class="Nm">wapbl_jlock_assert</code>, - <code class="Nm">wapbl_junlock_assert</code> — - <span class="Nd">write-ahead physical block logging for file - systems</span></p> -</section> -<section class="Sh"> -<h1 class="Sh" id="SYNOPSIS"><a class="permalink" href="#SYNOPSIS">SYNOPSIS</a></h1> -<p class="Pp"><code class="In">#include - <<a class="In">sys/wapbl.h</a>></code></p> -<p class="Pp"><var class="Vt">typedef void (*wapbl_flush_fn_t)(struct mount *, - daddr_t *, int *, int)</var>;</p> -<p class="Pp"><var class="Ft">int</var> - <br/> - <code class="Fn">wapbl_start</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl **wlp</var>, <var class="Fa" style="white-space: nowrap;">struct mount - *mp</var>, <var class="Fa" style="white-space: nowrap;">struct vnode - *devvp</var>, <var class="Fa" style="white-space: nowrap;">daddr_t - off</var>, <var class="Fa" style="white-space: nowrap;">size_t count</var>, - <var class="Fa" style="white-space: nowrap;">size_t blksize</var>, - <var class="Fa" style="white-space: nowrap;">struct wapbl_replay *wr</var>, - <var class="Fa" style="white-space: nowrap;">wapbl_flush_fn_t flushfn</var>, - <var class="Fa" style="white-space: nowrap;">wapbl_flush_fn_t - flushabortfn</var>);</p> -<p class="Pp"><var class="Ft">int</var> - <br/> - <code class="Fn">wapbl_stop</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">int - force</var>);</p> -<p class="Pp"><var class="Ft">int</var> - <br/> - <code class="Fn">wapbl_begin</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">const char - *file</var>, <var class="Fa" style="white-space: nowrap;">int - line</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_end</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>);</p> -<p class="Pp"><var class="Ft">int</var> - <br/> - <code class="Fn">wapbl_flush</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">int - wait</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_discard</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_add_buf</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">struct buf - *bp</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_remove_buf</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">struct buf - *bp</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_resize_buf</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">struct buf - *bp</var>, <var class="Fa" style="white-space: nowrap;">long oldsz</var>, - <var class="Fa" style="white-space: nowrap;">long oldcnt</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_register_inode</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">ino_t - ino</var>, <var class="Fa" style="white-space: nowrap;">mode_t - mode</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_unregister_inode</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">ino_t - ino</var>, <var class="Fa" style="white-space: nowrap;">mode_t - mode</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_register_deallocation</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>, <var class="Fa" style="white-space: nowrap;">daddr_t - blk</var>, <var class="Fa" style="white-space: nowrap;">int len</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_jlock_assert</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>);</p> -<p class="Pp"><var class="Ft">void</var> - <br/> - <code class="Fn">wapbl_junlock_assert</code>(<var class="Fa" style="white-space: nowrap;">struct - wapbl *wl</var>);</p> -</section> -<section class="Sh"> -<h1 class="Sh" id="DESCRIPTION"><a class="permalink" href="#DESCRIPTION">DESCRIPTION</a></h1> -<p class="Pp"><code class="Nm">WAPBL</code>, or - <a class="permalink" href="#write-ahead"><i class="Em" id="write-ahead">write-ahead - physical block logging</i></a>, is an abstraction for file systems to write - physical blocks in the <a class="Xr">buffercache(9)</a> to a bounded-size - log first before their real destinations on disk. The name means:</p> -<div class="Bd-indent"> -<dl class="Bl-tag"> - <dt>logging</dt> - <dd>batches of writes are issued atomically via a log</dd> - <dt>physical block</dt> - <dd>only physical blocks, not logical file system operations, are stored in - the log</dd> - <dt>write-ahead</dt> - <dd>before writing a block to disk, its new content, rather than its old - content for roll-back, is recorded in the log</dd> -</dl> -</div> -<p class="Pp" id="transactions">When a file system using - <code class="Nm">WAPBL</code> issues writes (as in - <a class="Xr">bwrite(9)</a> or <a class="Xr">bdwrite(9)</a>), they are - grouped in batches called - <a class="permalink" href="#transactions"><i class="Em">transactions</i></a> - in memory, which are serialized to be consistent with program order before - <code class="Nm">WAPBL</code> submits them to disk atomically.</p> -<p class="Pp">Thus, within a transaction, after one write, another write need - not wait for disk I/O, and if the system is interrupted, e.g. by a crash or - by power failure, either both writes will appear on disk, or neither - will.</p> -<p class="Pp" id="log">When a transaction is full, it is written to a circular - buffer on disk called the - <a class="permalink" href="#log"><i class="Em">log</i></a>. When the - transaction has been written to disk, every write in the transaction is - submitted to disk asynchronously. Finally, the file system may issue new - writes via <code class="Nm">WAPBL</code> once enough writes submitted to - disk have completed.</p> -<p class="Pp">After interruption, such as a crash or power failure, some writes - issued by the file system may not have completed. However, the log is - written consistently with program order and before file system writes are - submitted to disk. Hence a consistent program-order view of the file system - can be attained by resubmitting the writes that were successfully stored in - the log using <a class="Xr">wapbl_replay(9)</a>. This may not be the same - state just before interruption — writes in transactions that did not - reach the disk will be excluded.</p> -<p class="Pp" id="wapbl_start">For a file system to use - <code class="Nm">WAPBL</code>, its <a class="Xr">VFS_MOUNT(9)</a> method - should first replay any journal on disk using - <a class="Xr">wapbl_replay(9)</a>, and then, if the mount is read/write, - initialize <code class="Nm">WAPBL</code> for the mount by calling - <a class="permalink" href="#wapbl_start"><code class="Fn">wapbl_start</code></a>(). - The <a class="Xr">VFS_UNMOUNT(9)</a> method should call - <a class="permalink" href="#wapbl_stop"><code class="Fn" id="wapbl_stop">wapbl_stop</code></a>().</p> -<p class="Pp" id="wapbl_begin">Before issuing any - <a class="Xr">buffercache(9)</a> writes, the file system must acquire a - shared lock on the current <code class="Nm">WAPBL</code> transaction with - <a class="permalink" href="#wapbl_begin"><code class="Fn">wapbl_begin</code></a>(), - which may sleep until there is room in the transaction for new writes. After - issuing the writes, the file system must release its shared lock on the - transaction with <code class="Fn">wapbl_end</code>(). Either all writes - issued between <code class="Fn">wapbl_begin</code>() and - <code class="Fn">wapbl_end</code>() will complete, or none of them will.</p> -<p class="Pp" id="exclusive">File systems may also witness an - <a class="permalink" href="#exclusive"><i class="Em">exclusive</i></a> lock - on the current transaction when <code class="Nm">WAPBL</code> is flushing - the transaction to disk, or aborting a flush, and invokes a file system's - callback. File systems can assert that the transaction is locked with - <a class="permalink" href="#wapbl_jlock_assert"><code class="Fn" id="wapbl_jlock_assert">wapbl_jlock_assert</code></a>(), - or not - <a class="permalink" href="#exclusively"><i class="Em" id="exclusively">exclusively</i></a> - locked, with <code class="Fn">wapbl_junlock_assert</code>().</p> -<p class="Pp" id="wapbl_register_inode">If a file system requires multiple - transactions to initialize an inode, and needs to destroy partially - initialized inodes during replay, it can register them by - <var class="Vt">ino_t</var> inode number before initialization with - <a class="permalink" href="#wapbl_register_inode"><code class="Fn">wapbl_register_inode</code></a>() - and unregister them with - <a class="permalink" href="#wapbl_unregister_inode"><code class="Fn" id="wapbl_unregister_inode">wapbl_unregister_inode</code></a>() - once initialization is complete. <code class="Nm">WAPBL</code> does not - actually concern itself whether the objects identified by - <var class="Vt">ino_t</var> values are ‘inodes’ or - ‘quaggas’ or anything else — file systems may use this - to list any objects keyed by <var class="Vt">ino_t</var> value in the - log.</p> -<p class="Pp">When a file system frees resources on disk and issues writes to - reflect the fact, it cannot then reuse the resources until the writes have - reached the disk. However, as far as the <a class="Xr">buffercache(9)</a> is - concerned, as soon as the file system issues the writes, they will appear to - have been written. So the file system must not attempt to reuse the resource - until the current <code class="Nm">WAPBL</code> transaction has been flushed - to disk.</p> -<p class="Pp" id="wapbl_register_deallocation">The file system can defer freeing - a resource by calling - <a class="permalink" href="#wapbl_register_deallocation"><code class="Fn">wapbl_register_deallocation</code></a>() - to record the disk address of the resource and length in bytes of the - resource. Then, when <code class="Nm">WAPBL</code> next flushes the - transaction to disk, it will pass an array of the disk addresses and lengths - in bytes to a file-system-supplied callback. (Again, - <code class="Nm">WAPBL</code> does not care whether the ‘disk - address’ or ‘length in bytes’ is actually that; it will - pass along <var class="Vt">daddr_t</var> and <var class="Vt">int</var> - values.)</p> -</section> -<section class="Sh"> -<h1 class="Sh" id="FUNCTIONS"><a class="permalink" href="#FUNCTIONS">FUNCTIONS</a></h1> -<dl class="Bl-tag"> - <dt><code class="Fn">wapbl_start</code>(<var class="Fa">wlp</var>, - <var class="Fa">mp</var>, <var class="Fa">devvp</var>, - <var class="Fa">off</var>, <var class="Fa">count</var>, - <var class="Fa">blksize</var>, <var class="Fa">wr</var>, - <var class="Fa">flushfn</var>, <var class="Fa">flushabortfn</var>)</dt> - <dd>Start using <code class="Nm">WAPBL</code> for the file system mounted at - <var class="Fa">mp</var>, storing a log of <var class="Fa">count</var> - disk sectors at disk address <var class="Fa">off</var> on the block device - <var class="Fa">devvp</var> writing blocks in units of - <var class="Fa">blksize</var> bytes. On success, stores an opaque - <var class="Vt">struct wapbl *</var> cookie in - <code class="Li">*</code><var class="Fa">wlp</var> for use with the other - <code class="Nm">WAPBL</code> routines and returns zero. On failure, - returns an error number. - <p class="Pp" id="wapbl_start~2">If the file system had replayed the log - with <a class="Xr">wapbl_replay(9)</a>, then <var class="Fa">wr</var> - must be the <var class="Vt">struct wapbl_replay *</var> cookie used to - replay it, and - <a class="permalink" href="#wapbl_start~2"><code class="Fn">wapbl_start</code></a>() - will register any inodes that were in the log as if with - <code class="Fn">wapbl_register_inode</code>(); otherwise - <var class="Fa">wr</var> must be <code class="Dv">NULL</code>.</p> - <p class="Pp" id="wapbl_start~3"><var class="Fa">flushfn</var> is a callback - that <code class="Nm">WAPBL</code> will invoke as - <var class="Fa">flushfn</var> (<var class="Fa">mp</var>, - <var class="Fa">deallocblks</var>, <var class="Fa">dealloclens</var>, - <var class="Fa">dealloccnt</var>) just before it flushes a transaction - to disk, with the an exclusive lock held on the transaction, where - <var class="Fa">mp</var> is the mount point passed to - <a class="permalink" href="#wapbl_start~3"><code class="Fn">wapbl_start</code></a>(), - <var class="Fa">deallocblks</var> is an array of - <var class="Fa">dealloccnt</var> disk addresses, and - <var class="Fa">dealloclens</var> is an array of - <var class="Fa">dealloccnt</var> lengths, corresponding to the addresses - and lengths the file system passed to - <code class="Fn">wapbl_register_deallocation</code>(). If flushing the - transaction to disk fails, <code class="Nm">WAPBL</code> will call - <var class="Fa">flushabortfn</var> with the same arguments to undo any - effects that <var class="Fa">flushfn</var> had.</p> - </dd> - <dt><code class="Fn">wapbl_stop</code>(<var class="Fa">wl</var>, - <var class="Fa">force</var>)</dt> - <dd>Flush the current transaction to disk and stop using - <code class="Nm">WAPBL</code>. If flushing the transaction fails and - <var class="Fa">force</var> is zero, return error. If flushing the - transaction fails and <var class="Fa">force</var> is nonzero, discard the - transaction, permanently losing any writes in it. If flushing the - transaction is successful or if <var class="Fa">force</var> is nonzero, - free memory associated with <var class="Fa">wl</var> and return zero.</dd> - <dt><code class="Fn">wapbl_begin</code>(<var class="Fa">wl</var>, - <var class="Fa">file</var>, <var class="Fa">line</var>)</dt> - <dd>Wait for space in the current transaction for new writes, flushing it if - necessary, and acquire a shared lock on it. - <p class="Pp">The lock is not exclusive: other threads may acquire shared - locks on the transaction too. The lock is not recursive: a thread may - not acquire it again without calling <var class="Fa">wapbl_end</var> - first.</p> - <p class="Pp">May sleep.</p> - <p class="Pp"><var class="Fa">file</var> and <var class="Fa">line</var> are - the file name and line number of the caller for debugging purposes.</p> - </dd> - <dt id="wapbl_end"><a class="permalink" href="#wapbl_end"><code class="Fn">wapbl_end</code></a>(<var class="Fa">wl</var>)</dt> - <dd>Release a shared lock on the transaction acquired with - <code class="Fn">wapbl_begin</code>().</dd> - <dt id="wapbl_flush"><a class="permalink" href="#wapbl_flush"><code class="Fn">wapbl_flush</code></a>(<var class="Fa">wl</var>, - <var class="Fa">wait</var>)</dt> - <dd>Flush the current transaction to disk. If <var class="Fa">wait</var> is - nonzero, wait for all writes in the current transaction to complete. - <p class="Pp">The current transaction must not be locked.</p> - </dd> - <dt id="wapbl_discard"><a class="permalink" href="#wapbl_discard"><code class="Fn">wapbl_discard</code></a>(<var class="Fa">wl</var>)</dt> - <dd>Discard the current transaction, permanently losing any writes in it. - <p class="Pp">The current transaction must not be locked.</p> - </dd> - <dt id="wapbl_add_buf"><a class="permalink" href="#wapbl_add_buf"><code class="Fn">wapbl_add_buf</code></a>(<var class="Fa">wl</var>, - <var class="Fa">bp</var>)</dt> - <dd>Add the buffer <var class="Fa">bp</var> to the current transaction, which - must be locked, because someone has asked to write it. - <p class="Pp">This is meant to be called from within - <a class="Xr">buffercache(9)</a>, not by file systems directly.</p> - </dd> - <dt id="wapbl_remove_buf"><a class="permalink" href="#wapbl_remove_buf"><code class="Fn">wapbl_remove_buf</code></a>(<var class="Fa">wl</var>, - <var class="Fa">bp</var>)</dt> - <dd>Remove the buffer <var class="Fa">bp</var>, which must have been added - using <var class="Fa">wapbl_add_buf</var>, from the current transaction, - which must be locked, because it has been invalidated (or XXX ???). - <p class="Pp">This is meant to be called from within - <a class="Xr">buffercache(9)</a>, not by file systems directly.</p> - </dd> - <dt id="wapbl_resize_buf"><a class="permalink" href="#wapbl_resize_buf"><code class="Fn">wapbl_resize_buf</code></a>(<var class="Fa">wl</var>, - <var class="Fa">bp</var>, <var class="Fa">oldsz</var>, - <var class="Fa">oldcnt</var>)</dt> - <dd>Note that the buffer <var class="Fa">bp</var>, which must have been added - using <var class="Fa">wapbl_add_buf</var>, has changed size, where - <var class="Fa">oldsz</var> is the previous allocated size in bytes and - <var class="Fa">oldcnt</var> is the previous number of valid bytes in - <var class="Fa">bp</var>. - <p class="Pp">This is meant to be called from within - <a class="Xr">buffercache(9)</a>, not by file systems directly.</p> - </dd> - <dt id="wapbl_register_inode~2"><a class="permalink" href="#wapbl_register_inode~2"><code class="Fn">wapbl_register_inode</code></a>(<var class="Fa">wl</var>, - <var class="Fa">ino</var>, <var class="Fa">mode</var>)</dt> - <dd>Register <var class="Fa">ino</var> with the mode - <var class="Fa">mode</var> as commencing initialization.</dd> - <dt id="wapbl_unregister_inode~2"><a class="permalink" href="#wapbl_unregister_inode~2"><code class="Fn">wapbl_unregister_inode</code></a>(<var class="Fa">wl</var>, - <var class="Fa">ino</var>, <var class="Fa">mode</var>)</dt> - <dd>Unregister <var class="Fa">ino</var>, which must have previously been - registered with <var class="Fa">wapbl_register_inode</var> using the same - <var class="Fa">mode</var>, now that its initialization has - completed.</dd> - <dt><code class="Fn">wapbl_register_deallocation</code>(<var class="Fa">wl</var>, - <var class="Fa">blk</var>, <var class="Fa">len</var>)</dt> - <dd>Register <var class="Fa">len</var> bytes at the disk address - <var class="Fa">blk</var> as ready for deallocation, so that they will be - passed to the <var class="Fa">flushfn</var> that was given to - <code class="Fn">wapbl_start</code>().</dd> - <dt><code class="Fn">wapbl_jlock_assert</code>(<var class="Fa">wl</var>)</dt> - <dd>Assert that the current transaction is locked. - <p class="Pp" id="any">Note that it might not be locked by the current - thread: this assertion passes if - <a class="permalink" href="#any"><i class="Em">any</i></a> thread has it - locked.</p> - </dd> - <dt id="wapbl_junlock_assert"><a class="permalink" href="#wapbl_junlock_assert"><code class="Fn">wapbl_junlock_assert</code></a>(<var class="Fa">wl</var>)</dt> - <dd>Assert that the current transaction is not exclusively locked by the - current thread. - <p class="Pp" id="wapbl_start~4">Users of <code class="Nm">WAPBL</code> - observe exclusive locks only in the <var class="Fa">flushfn</var> and - <var class="Fa">flushabortfn</var> callbacks to - <a class="permalink" href="#wapbl_start~4"><code class="Fn">wapbl_start</code></a>(). - Outside of such contexts, the transaction is never exclusively locked, - even between <code class="Fn">wapbl_begin</code>() and - <code class="Fn">wapbl_end</code>().</p> - <p class="Pp" id="wapbl_begin~2">There is no way to assert that the current - transaction is not locked at all — i.e., that the caller may - acquire a shared lock on the transaction with - <a class="permalink" href="#wapbl_begin~2"><code class="Fn">wapbl_begin</code></a>() - without danger of deadlock.</p> - </dd> -</dl> -</section> -<section class="Sh"> -<h1 class="Sh" id="CODE_REFERENCES"><a class="permalink" href="#CODE_REFERENCES">CODE - REFERENCES</a></h1> -<p class="Pp">The <code class="Nm">WAPBL</code> subsystem is implemented in - <span class="Pa">sys/kern/vfs_wapbl.c</span>, with hooks in - <span class="Pa">sys/kern/vfs_bio.c</span>.</p> -</section> -<section class="Sh"> -<h1 class="Sh" id="SEE_ALSO"><a class="permalink" href="#SEE_ALSO">SEE - ALSO</a></h1> -<p class="Pp"><a class="Xr">buffercache(9)</a>, <a class="Xr">vfsops(9)</a>, - <a class="Xr">wapbl_replay(9)</a></p> -</section> -<section class="Sh"> -<h1 class="Sh" id="BUGS"><a class="permalink" href="#BUGS">BUGS</a></h1> -<p class="Pp"><code class="Nm">WAPBL</code> works only for file system metadata - managed via the <a class="Xr">buffercache(9)</a>, and provides no way to log - writes via the page cache, as in <a class="Xr">VOP_GETPAGES(9)</a>, - <a class="Xr">VOP_PUTPAGES(9)</a>, and <a class="Xr">ubc_uiomove(9)</a>, - which is normally used for file data.</p> -<p class="Pp">Not only is <code class="Nm">WAPBL</code> unable to log writes via - the page cache, it is also unable to defer <a class="Xr">buffercache(9)</a> - writes until cached pages have been written. This manifests as the - well-known garbage-data-appended-after-crash bug in FFS: when appending to a - file, the pages containing new data may not reach the disk before the inode - update reporting its new size. After a crash, the inode update will be on - disk, but the new data will not be — instead, whatever garbage data - in the free space will appear to have been appended to the file. - <code class="Nm">WAPBL</code> exacerbates the problem by increasing the - throughput of metadata writes, because it can issue many metadata writes - asynchronously that FFS without <code class="Nm">WAPBL</code> would need to - issue synchronously in order for <a class="Xr">fsck(8)</a> to work.</p> -<p class="Pp">The criteria for when the transaction must be flushed to disk - before <code class="Fn">wapbl_begin</code>() returns are heuristic, i.e. - wrong. There is no way for a file system to communicate to - <code class="Fn">wapbl_begin</code>() how many buffers, inodes, and - deallocations it will issue via <code class="Nm">WAPBL</code> in the - transaction.</p> -<p class="Pp"><code class="Nm">WAPBL</code> mainly supports write-ahead, and has - only limited support for rolling back operations, in the form of - <code class="Fn">wapbl_register_inode</code>() and - <code class="Fn">wapbl_unregister_inode</code>(). Consequently, for example, - large writes appending to a file, which requires multiple disk block - allocations and an inode update, must occur in a single transaction — - there is no way to roll back the disk block allocations if the write fails - in the middle, e.g. because of a fault in the middle of the user buffer.</p> -<p class="Pp"><code class="Fn">wapbl_jlock_assert</code>() does not guarantee - that the current thread has the current transaction locked. - <code class="Fn">wapbl_junlock_assert</code>() does not guarantee that the - current thread does not have the current transaction locked at all.</p> -<p class="Pp">There is only one <code class="Nm">WAPBL</code> transaction for - each file system at any given time, and only one - <code class="Nm">WAPBL</code> log on disk. Consequently, all writes are - serialized. Extending <code class="Nm">WAPBL</code> to support multiple logs - per file system, partitioned according to an appropriate scheme, is left as - an exercise for the reader.</p> -<p class="Pp">There is no reason for <code class="Nm">WAPBL</code> to require - its own hooks in <a class="Xr">buffercache(9)</a>.</p> -<p class="Pp">The on-disk format used by <code class="Nm">WAPBL</code> is - undocumented.</p> -</section> -</div> -<table class="foot"> - <tr> - <td class="foot-date">March 26, 2015</td> - <td class="foot-os">NetBSD 10.1</td> - </tr> -</table> |