--- /dev/null
+#LyX 1.3 created this file. For more info see http://www.lyx.org/
+\lyxformat 221
+\textclass article
+\begin_preamble
+\usepackage{listings}
+\usepackage[usenames]{color}
+\usepackage{courier}
+
+%\usepackage{pt}
+%\usepackage{units}
+%\usepackage{coz}
+
+%\usepackage{epsf}
+%\usepackage{lncsexample}
+%\conttrue
+%\usepackage{times}
+%\usepackage{top}
+%\usepackage{graphicx}
+\newcommand{\lst}[3] {
+ \noindent\vspace{-1mm}
+ \definecolor{cKeyword}{rgb}{0.8,0.1,0.1}
+ \definecolor{cComment}{rgb}{0.2,0.5,0.7}
+ \definecolor{cString}{rgb}{0.2,0.7,0.2}
+ \lstinputlisting[caption={#2},
+ label={#1},
+ showstringspaces=false,
+ numbers=left,
+ stepnumber=1,
+ frame=lines,
+ language=C,
+ extendedchars=true,
+ basicstyle=\small\tt,
+ numberstyle=\tiny,
+ keywordstyle=\color{cKeyword},
+ commentstyle=\color{cComment},
+ stringstyle=\color{cString},
+ directivestyle=\color{magenta},
+ emph={1, 2, 3, 4, 5, 6, 7, 8, 9, 0, NULL, lustre, CFS},
+ emphstyle=\color{blue},
+ breaklines=true,
+ #3]{#1}\vspace{0.3mm}
+}
+\end_preamble
+\language english
+\inputencoding koi8-r
+\fontscheme times
+\graphics default
+\paperfontsize default
+\spacing single
+\papersize Default
+\paperpackage a4
+\use_geometry 1
+\use_amsmath 0
+\use_natbib 0
+\use_numerical_citations 0
+\paperorientation portrait
+\leftmargin 4cm
+\topmargin 2cm
+\rightmargin 2cm
+\bottommargin 2cm
+\secnumdepth 3
+\tocdepth 3
+\paragraph_separation skip
+\defskip medskip
+\quotes_language swedish
+\quotes_times 2
+\papercolumns 1
+\papersides 1
+\paperpagestyle default
+
+\layout Title
+
+SMFS Detailed Level Design
+\layout Author
+
+Mike Pershin
+\layout Date
+
+21 February 2005
+\layout Standard
+\pagebreak_bottom \noindent
+
+\begin_inset LatexCommand \tableofcontents{}
+
+\end_inset
+
+
+\layout Section
+
+
+\noun on
+Functional Specification
+\layout Subsection
+
+SMFS operations
+\layout Subsubsection
+
+Initialization and data structures
+\layout Standard
+
+SMFS implement all needed methods as filesystem but uses other filesystem
+ as store instead of block device.
+ This way demands that SMFS should care about this backstore FS like VFS.
+ SMFS defines several info structures for each filesystem object that contain
+ all needed information about backstore related stuff and etc.
+\layout Subsubsection
+
+Plugin handling
+\layout Standard
+
+SMFS has plugin API which allows to register/deregister plugins and call
+ their functions.
+ Plugins are organized in linked list.
+\layout Standard
+
+Plugin initialization starts when smfs is mounting by someone with options
+ that contain information about plugin needed.
+ Deactivated process starts when SMFS unmount occur.
+ It possible to use ioctl interface or procfs (sysfs later) for plugin managemen
+t.
+ Each plugin will receives notification from SMFS about filesystem operations.
+ Plugins should be aware of delays, transaction handling and consistency.
+\layout Standard
+
+SMFS contains information about plugin loaded with special flags.
+ There are system-wide flags and object-related to indicate exclusions.
+\layout Subsubsection
+
+Locking and plugins
+\layout Standard
+
+Every hook is called under some conditions in term of resource availability.
+ SMFS should provide that all pre- and post-operation stuff should be done
+ together with fs operation in backstore fs.
+ Plugins should be aware about these conditions and acts properly.
+\layout Subsubsection*
+
+
+\emph on
+Inode_operations
+\layout Standard
+
+locking rules: all may block, none have BKL.
+ Protection provided through using i_sem semaphore in inode.
+\layout Standard
+
+
+\begin_inset Tabular
+<lyxtabular version="3" rows="17" columns="3">
+<features>
+<column alignment="center" valignment="top" leftline="true" width="0">
+<column alignment="center" valignment="top" leftline="true" width="0">
+<column alignment="center" valignment="top" leftline="true" rightline="true" width="0">
+<row topline="true" bottomline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+i_sem
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+Notes
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+lookup
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+create
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+link
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+both
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+mknod
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+symlink
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+mkdir
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+unlink
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+both
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+rmdir
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+both
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+rename
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+all
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+truncate
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+setattr
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+getattr
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+no
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+setxattr
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+getxattr
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+no
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+listxattr
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+no
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+<row topline="true" bottomline="true">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+removexattr
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+yes
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\layout Standard
+
+\end_inset
+</cell>
+</row>
+</lyxtabular>
+
+\end_inset
+
+
+\layout Itemize
+
+Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_sem on victim.
+\layout Itemize
+
+cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
+\layout Itemize
+
+->truncate() is never called directly - it's a callback, not a method.
+ It's called by vmtruncate() - library function normally used by ->setattr().
+ Locking information above applies to that call (i.e.
+ is inherited from ->setattr() - vmtruncate() is used when ATTR_SIZE had
+ been passed).
+
+\layout Standard
+
+All inode operations may block so plugin can also block.
+ All critical inode operations is already protected by semaphore i_sem.
+\layout Subsubsection*
+
+
+\emph on
+File operations
+\layout Standard
+
+Locking rules: file operations may block and no BKL here.
+
+\layout Standard
+
+Some hooks placed around file operations so protection needed to provide
+ sync execution of pre- and post-hooks with backfs operation.
+
+\layout Subsubsection*
+
+
+\emph on
+Superblock operations
+\layout Standard
+
+Locking rules: All may block, only
+\emph on
+put_super
+\emph default
+is under BKL.
+
+\layout Standard
+
+No hooks here, so no actions are needed.
+\layout Subsubsection*
+
+File operations
+\layout Standard
+
+locking rules: All may block.
+
+\layout Standard
+
+None is under BKL except
+\emph on
+ioctl
+\layout Comment
+
+Taken from ..../linux/Documentation/filesystems/Locking
+\layout Subsubsection
+
+Upcall handling
+\layout Standard
+
+There is upcall API in SMFS.
+ Upcalls are organized in linked list and called one by one where needed.
+
+\layout Subsubsection
+
+Fsfilt operations
+\layout Standard
+
+LVFS has fsfilt operations for SMFS.
+ SMFS store copy of fsfilt operations from backstore filesystem, gets fsfilt
+ operations for itself and pass each call to backstore filesysem.
+
+\layout Comment
+
+should I give more info?
+\layout Subsubsection
+
+Transactions handling
+\layout Standard
+
+All hooks should be in the same transaction with backstore fs operation.
+ To provide this SMFS starts transaction before first hook and commit after
+ last one.
+ SMFS care about transaction where it is needed, but plugins can add some
+ stuff to it.
+ To solve this SMFS pass transaction handler to plugin, so it can do transaction
+ in right way.
+
+\layout Comment
+
+details?
+\layout Subsection
+
+Backstore filesystem
+\layout Subsubsection
+
+Initialization
+\layout Enumerate
+
+Backstore FS is mounted by SMFS while initialization, if mount is failed
+ SMFS initialization is also failed (maybe in that case we need to mount
+ backstore FS directly instead of SMFS?)
+\layout Enumerate
+
+SMFS creates own superblock using backstore one and fsfilter operations
+\layout Subsubsection
+
+Superblock operations
+\layout Standard
+
+SMFS have to initialize own inode structure reading inode with same number
+ from backstore FS.
+ Operations with such inode will be redirected to backstore FS if they are
+ invoked.
+ SMFS will store backstore FS inode to avoid reading inode every time.
+
+\layout Subsubsection
+
+Inode operations
+\layout Standard
+
+SMFS creates own inode operations for each inode.
+ When some operations in invoked SMFS calls real filesystem method to complete
+ it.
+ There are additional actions for making SMFS inode and real inode consistent.
+\layout Subsubsection
+
+File operations
+\layout Standard
+
+SMFS creates own filp, duplicate backstore FS filp and dentry.
+ They are used in file operations following by backstore operation call.
+
+\layout Section
+
+
+\noun on
+Use cases
+\layout Subsection
+
+Backstore filesystem
+\layout Standard
+
+Next options is passed to smfs and descript backstore FS completely:
+\layout Standard
+
+smfs dev=/mnt type=ext3
+\layout Subsection
+
+SMFS plugins/upcalls
+\layout Subsubsection
+
+SMFS plugin activation
+\layout Standard
+
+Now all plugins are compiled in SMFS module and plugins are setted up via
+ options when SMFS is mounting.
+ Options should be passed as mountfsoptions like this:
+\emph on
+...
+ [kml|cache|snap...]
+\layout Subsubsection
+
+SMFS plugin registration
+\layout Standard
+
+Plugins should registers with SMFS:
+\layout LyX-Code
+
+smfs_register_plugin(parameters);
+\layout Standard
+
+Parameters are struct smfs_plugin filled with valid data:
+\layout Itemize
+
+type of plugin - to distinguish it from anothers,
+\layout Itemize
+
+pre_op function - function that will be called before fs operation,
+\layout Itemize
+
+post_op function - function that will be called after fs operation,
+\layout Itemize
+
+helper function - helper function.
+ See Plugin API,
+\layout Itemize
+
+any private data - data which will be return to plugin in each call.
+\layout Subsubsection
+
+SMFS hook invokation
+\layout Standard
+
+SMFS place special wrapper in own filesystem operations to call plugins
+ hook:
+\layout LyX-Code
+
+/* this is wrapper that calls all hooks walking through the list */
+\layout LyX-Code
+
+SMFS_HOOK(opcode, parameters);
+\layout LyX-Code
+
+\layout Standard
+
+Hooks are placed in SMFS methods before and after calling backstore FS operation
+s:
+\layout LyX-Code
+
+/* this is how to SMFS uses hooks */
+\layout LyX-Code
+
+smfs_some_op()
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct inode * backfs_inode = I2CI(inode);
+\layout LyX-Code
+
+ struct smfs_file_info *sfi;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ SMFS_HOOK(hook_opcode, ...);
+\layout LyX-Code
+
+ backfs_inode->i_fop->some_op(sfi->c_file, ...);
+\layout LyX-Code
+
+ SMFS_HOOK(hook_opcode, ...);
+\layout LyX-Code
+
+}
+\layout Subsubsection
+
+SMFS upcall usage
+\layout Standard
+
+Upcalls can be placed in any place in SMFS, where side modules wants to
+ take control from SMFS.
+ Module should:
+\layout Enumerate
+
+prepares function to handle upcall events,
+\layout Enumerate
+
+registers upcall using upcall API,
+\layout Enumerate
+
+receives upcalls and handles they,
+\layout Enumerate
+
+deregisters upcall when smfs is unmounted.
+\layout Section
+
+
+\noun on
+Logic
+\noun default
+
+\noun on
+Specification
+\layout Subsection
+
+SMFS operations
+\layout Subsubsection
+
+Initialization and data structures
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../include/linux/lustre_smfs.h}{Superblock info}{firstline=87,lastline=107}
+\end_inset
+
+
+\layout LyX-Code
+
+struct smfs_inode_info {
+\layout LyX-Code
+
+ /* this first part of struct should be
+\layout LyX-Code
+
+ the same as in mds_info_info */
+\layout LyX-Code
+
+ struct lustre_id smi_id;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ /* smfs part.
+ */
+\layout LyX-Code
+
+ struct inode * backfs_inode;
+\layout LyX-Code
+
+ __u32 smi_flags; //plugins pre_inode flags
+\layout LyX-Code
+
+ struct list_head plist; /* list of plugins inode info */
+\layout LyX-Code
+
+};
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../include/linux/lustre_smfs.h}{File info}{firstline=113,lastline=116}
+\end_inset
+
+
+\layout Subsubsection
+
+Fsfilt operations
+\layout Standard
+
+Fsfilt operations for SMFS just call the same operations in backstore FS.
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../lvfs/fsfilt_smfs.c}{Fsfilt redirection example}{firstline=361,lastline=378}
+\end_inset
+
+
+\layout Subsubsection
+
+Transactions handling
+\layout Standard
+
+Each operations with hooks pass transaction handler to hook.
+ Hook can use it to decide does it need start new transaction or not.
+\layout Standard
+
+If there is no handler, hook can create own transaction:
+\layout LyX-Code
+
+hook_func_in_plugin()
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ ...
+\layout LyX-Code
+
+ if (!handle) {
+\layout LyX-Code
+
+ handle = smfs_trans_start(inode, KML_CACHE_NOOP, NULL);
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ ...
+\layout LyX-Code
+
+}
+\layout Standard
+
+It is important that plugin should also commit transaction in the same call.
+
+\layout Standard
+
+If transaction begun before hook call it should know size of transaction,
+ so we should provide a way to calculate this size through all plugins involved.
+ Each plugin can register upcall with related type that will return extra
+ size needed by that plugin.
+ SMFS will walk through upcalls and gets total extra size for transaction:
+
+\layout LyX-Code
+
+fsfilt_smfs_start(struct inode * inode, int op, ...)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ ...
+\layout LyX-Code
+
+ SMFS_TRANS_EXTRA_SIZE(inode, op); /* walking through list of plugins
+ */
+\layout LyX-Code
+
+ handle = cache_fsfilt->fs_start(cache_inode, op, ...);
+\layout LyX-Code
+
+ ...
+\layout LyX-Code
+
+}
+\layout Comment
+
+still have no good idea how to pass calculated size instead op without changes
+ in fsfilt_ext3/any_other_fs code.
+ New parameter in fs_start?
+\layout Subsection
+
+Backstore filesystem
+\layout Subsubsection
+
+Initialization/deinitialization
+\layout Standard
+
+Backstore FS is mounted by SMFS while SMFS initialization.
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/smfs_lib.c}{Mount/Umount}{firstline=163,lastline=215}
+\end_inset
+
+
+\layout Subsubsection
+
+Inode\SpecialChar ~
+info init/clear
+\layout Standard
+
+For each new inode SMFS creates private structures and getting real inode
+ from backstore FS.
+ SMFS is only user of backfs inode, so it is enough to get it here and put
+ it down in clear_inode_info.
+ Some plugins can have as SMFS-wide info as per-inode info.
+ Therefore they can also participate in init_inode_info/clear_inode_info
+ methods.
+ Here are helpers for this.
+\layout LyX-Code
+
+static void smfs_init_inode_info(struct inode *inode, void *opaque) {
+\layout LyX-Code
+
+ struct inode *cache_inode = NULL;
+\layout LyX-Code
+
+ struct smfs_iget_args *sargs;
+\layout LyX-Code
+
+ unsigned long ino;
+\layout LyX-Code
+
+\layout LyX-Code
+
+ sargs = (struct smfs_iget_args *)opaque;
+\layout LyX-Code
+
+ /* getting backing fs inode.
+ */
+\layout LyX-Code
+
+ ino = sargs ? sargs->s_ino : inode->i_ino;
+\layout LyX-Code
+
+ cache_inode = iget(S2CSB(inode->i_sb), ino);
+\layout LyX-Code
+
+ OBD_ALLOC(inode->u.generic_ip, sizeof(struct smfs_inode_info));
+\layout LyX-Code
+
+ I2CI(inode) = cache_inode;
+\layout LyX-Code
+
+ post_smfs_inode(inode, cache_inode);
+\layout LyX-Code
+
+ sm_set_inode_ops(cache_inode, inode);
+\layout LyX-Code
+
+ if (sargs && sargs->s_inode)
+\layout LyX-Code
+
+ I2SMI(inode)->smi_flags = I2SMI(sargs->s_inode)->smi_flags;
+\layout LyX-Code
+
+ SMFS_PLUGIN_HELPER(PL_INIT_INODE, inode);
+\layout LyX-Code
+
+}
+\layout LyX-Code
+
+\layout LyX-Code
+
+static void smfs_clear_inode_info(struct inode *inode) {
+\layout LyX-Code
+
+ if (I2SMI(inode)) {
+\layout LyX-Code
+
+ struct inode *cache_inode = I2CI(inode);
+\layout LyX-Code
+
+ if (cache_inode != cache_inode->i_sb->s_root->d_inode)
+\layout LyX-Code
+
+ iput(cache_inode);
+\layout LyX-Code
+
+ SMFS_PLUGIN_HELPER(PL_CLEAR_INODE, inode);
+\layout LyX-Code
+
+ OBD_FREE(inode->u.generic_ip, sizeof(struct smfs_inode_info));
+\layout LyX-Code
+
+ inode->u.generic_ip = NULL;
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+}
+\layout LyX-Code
+
+\layout Subsubsection
+
+Getting inode
+\layout Comment
+
+test_inode and set_inode are called under spinlock, so context-switching
+ is not allowed here.
+ Inode will be read after iget5_locked call
+\layout LyX-Code
+
+int smfs_test_inode(struct inode * inode, void * opaque)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct smfs_iget_args * sargs = opaque;
+\layout LyX-Code
+
+ struct smfs_up_message message;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ message.inode = inode;
+\layout LyX-Code
+
+ message.param = opaque;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ if (sargs && (inode->i_ino == sargs->s_ino)) {
+\layout LyX-Code
+
+ /* some module can add extra checks here */
+\layout LyX-Code
+
+ if (SMFS_UPCALL(SMFS_UP_TEST_INODE, (void*)&message))
+\layout LyX-Code
+
+ return 1;
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ return 0;
+\layout LyX-Code
+
+}
+\layout LyX-Code
+
+\layout LyX-Code
+
+int smfs_set_inode(struct inode *inode, void *opaque)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct smfs_up_message message;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ message.inode = inode;
+\layout LyX-Code
+
+ message.param = opaque;
+\layout LyX-Code
+
+\layout LyX-Code
+
+ /* someone can wants to do action here */
+\layout LyX-Code
+
+ SMFS_UPCALL(SMFS_UP_SET_INODE, (void*)&message);
+\layout LyX-Code
+
+ return 0;
+\layout LyX-Code
+
+}
+\layout LyX-Code
+
+\layout LyX-Code
+
+struct inode * smfs_iget(struct super_block * sb, ino_t hash,
+\layout LyX-Code
+
+ struct smfs_iget_args * sargs)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct inode *inode;
+\layout LyX-Code
+
+ inode = iget5_locked(sb, hash, smfs_test_inode,
+\layout LyX-Code
+
+ smfs_set_inode, sargs);
+\layout LyX-Code
+
+ if (inode) {
+\layout LyX-Code
+
+ if (inode->i_state & I_NEW) {
+\layout LyX-Code
+
+ smfs_init_inode_info(inode, sargs);
+\layout LyX-Code
+
+ unlock_new_inode(inode);
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ inode->i_ino = hash;
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ return inode;
+\layout LyX-Code
+
+}
+\layout Subsubsection
+
+Superblock operations
+\layout Paragraph
+
+read\SpecialChar ~
+inode2()
+\layout LyX-Code
+
+smfs_read_inode2 (struct inode * inode, void * opaque)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ smfs_init_inode_info(inode, opaque);
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+dirty\SpecialChar ~
+inode()/write\SpecialChar ~
+inode()
+\layout LyX-Code
+
+smfs_dirty/write_inode(struct inode * inode)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ backfs_inode = I2CI(inode);
+\layout LyX-Code
+
+ backfs_sb->s_op->dirty/write_inode(backfs_inode);
+\layout LyX-Code
+
+ duplicate_inode(inode, backfs_inode);
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+put\SpecialChar ~
+inode(inode)
+\layout LyX-Code
+
+smfs_put_inode(struct inode * inode)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ return;
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+delete\SpecialChar ~
+inode()
+\layout LyX-Code
+
+smfs_delete_inode(struct inode * inode) {
+\layout LyX-Code
+
+ clear_inode(inode);
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+clear\SpecialChar ~
+inode()
+\layout LyX-Code
+
+smfs_clear_inode(struct inode * inode) {
+\layout LyX-Code
+
+ smfs_clear_inode_info(inode);
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+put\SpecialChar ~
+super(super)
+\layout LyX-Code
+
+smfs_cleanup_hooks();
+\layout LyX-Code
+
+smfs_umount_cache(smfs_super_info);
+\layout LyX-Code
+
+smfs_cleanup_smb(sb);
+\layout Paragraph
+
+write\SpecialChar ~
+super()/write\SpecialChar ~
+super\SpecialChar ~
+lockfs()/unlockfs()/statfs()/remountfs()
+\layout LyX-Code
+
+backfs_sb = S2CSB(sb);
+\layout LyX-Code
+
+backfs_sb->s_op->...;
+\layout LyX-Code
+
+duplicate-sb(sb, backfs_sb);
+\layout Subsubsection
+
+Inode operations
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Create()}{firstline=44,lastline=107}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Lookup()}{firstline=109,lastline=172}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Link()}{firstline=174,lastline=231}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Unlink()}{firstline=233,lastline=276}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Symlink()}{firstline=278,lastline=327}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Mkdir()}{firstline=329,lastline=377}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Rmdir()}{firstline=379,lastline=424}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Mknod()}{firstline=426,lastline=485}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/dir.c}{Rename()}{firstline=487,lastline=557}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/file.c}{Setxattr()}{firstline=442,lastline=467}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/file.c}{Getxattr()}{firstline=469,lastline=494}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/file.c}{Listxattr()}{firstline=496,lastline=519}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/file.c}{Removexattr()}{firstline=521,lastline=544}
+\end_inset
+
+
+\layout Standard
+
+
+\begin_inset ERT
+status Open
+
+\layout Standard
+
+\backslash
+lst{../../smfs/file.c}{Truncate()}{firstline=385,lastline=404}
+\end_inset
+
+
+\layout Subsubsection
+
+File operations
+\layout Paragraph
+
+Common case (write/read/llseek/mmap/ioctl/readdir)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct inode * backfs_inode;
+\layout LyX-Code
+
+ struct smfs_file_info *sfi;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ backfs_inode = I2CI(file->f_dentry->d_inode);
+\layout LyX-Code
+
+ sfi = F2SMFI(file);
+\layout LyX-Code
+
+ pre_smfs_inode(file->f_dentry->d_inode, backfs_inode);
+\layout LyX-Code
+
+ SMFS_HOOK(..., hook_opcode , ..., PRE_HOOK, ...);
+\layout LyX-Code
+
+ if (backfs_inode->i_fop->...)
+\layout LyX-Code
+
+ backfs_inode->i_fop->...(sfi->c_file, ...);
+\layout LyX-Code
+
+ SMFS_HOOK(..., hook_opcode, ..., POST_HOOK, ...);
+\layout LyX-Code
+
+ post_smfs_inode(file->f_dentry->d_inode, backfs_inode); /* duplicate_ino
+de */
+\layout LyX-Code
+
+ duplicate_file(file, sfi->c_file);
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+fsync()
+\layout LyX-Code
+
+int smfs_fsync(struct file *file, struct dentry *dentry, int datasync)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct smfs_file_info *sfi = NULL;
+\layout LyX-Code
+
+ struct dentry *backfs_dentry = NULL;
+\layout LyX-Code
+
+ struct file *backfs_file = NULL;
+\layout LyX-Code
+
+ struct inode *backfs_inode = NULL;
+\layout LyX-Code
+
+\layout LyX-Code
+
+ backfs_inode = I2CI(dentry->d_inode);
+\layout LyX-Code
+
+ backfs_dentry = pre_smfs_dentry(NULL, backfs_inode, dentry);
+\layout LyX-Code
+
+ if (file) {
+\layout LyX-Code
+
+ sfi = F2SMFI(file);
+\layout LyX-Code
+
+ backfs_file = sfi->c_file;
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ pre_smfs_inode(dentry->d_inode, backfs_inode);
+\layout LyX-Code
+
+ if (backfs_inode->i_fop->fsync) {
+\layout LyX-Code
+
+ rc = backfs_inode->i_fop->fsync(backfs_file, backfs_dentry,
+ datasync);
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ post_smfs_inode(dentry->d_inode, backfs_inode);
+\layout LyX-Code
+
+ duplicate_file(file, backfs_file);
+\layout LyX-Code
+
+ post_smfs_dentry(backfs_dentry);
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+open()
+\layout LyX-Code
+
+int smfs_open(struct inode *inode, struct file *filp)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct inode *backfs_inode = NULL;
+\layout LyX-Code
+
+
+\layout LyX-Code
+
+ backfs_inode = I2CI(inode);
+\layout LyX-Code
+
+ smfs_init_cache_file(inode, filp);
+\layout LyX-Code
+
+ if (backfs_inode->i_fop->open)
+\layout LyX-Code
+
+ rc = cache_inode->i_fop->open(backfs_inode, F2CF(filp));
+\layout LyX-Code
+
+ duplicate_file(filp, F2CF(filp));
+\layout LyX-Code
+
+}
+\layout Paragraph
+
+release()
+\layout LyX-Code
+
+int smfs_release(struct inode *inode, struct file *filp)
+\layout LyX-Code
+
+{
+\layout LyX-Code
+
+ struct inode *backfs_inode = NULL;
+\layout LyX-Code
+
+ struct file *backfs_file = NULL;
+\layout LyX-Code
+
+ struct smfs_file_info *sfi = NULL;
+\layout LyX-Code
+
+\layout LyX-Code
+
+ backfs_inode = I2CI(inode);
+\layout LyX-Code
+
+ if (filp) {
+\layout LyX-Code
+
+ sfi = F2SMFI(filp);
+\layout LyX-Code
+
+ backfs_file = sfi->c_file;
+\layout LyX-Code
+
+ }
+\layout LyX-Code
+
+ if (backfs_inode->i_fop->release)
+\layout LyX-Code
+
+ backfs_inode->i_fop->release(backfs_inode, backfs_file);
+\layout LyX-Code
+
+ post_smfs_inode(inode, backfs_inode);
+\layout LyX-Code
+
+ smfs_cleanup_cache_file(filp);
+\layout LyX-Code
+
+}
+\layout Section
+
+State Specification
+\layout Subsection
+
+Transactions
+\layout Standard
+
+There are several situations are possible with transactions.
+ SMFS doesn't know will plugin do transaction or not.
+
+\layout Subsubsection
+
+Backstore FS will do transaction
+\layout Standard
+
+In that case plugins actions should be in same transaction with backstore
+ FS.
+ Next actions will be done:
+\layout Enumerate
+
+SMFS gets extra size for current operation from all plugins.
+\layout Enumerate
+
+If extra size is not null then some plugins will participate in transaction.
+ SMFS starts transaction with calculated size.
+\layout Enumerate
+
+SMFS call pre_hook.
+\layout Enumerate
+
+SMFS call backstore FS operation.
+\layout Enumerate
+
+SMFS call post_hook.
+\layout Enumerate
+
+SMFS commit transaction.
+\layout Subsubsection
+
+Backstore FS operation is not journaled
+\layout Standard
+
+In that case plugins may don't care about external conditions and do transaction
+ by own:
+\layout Enumerate
+
+SMFS call pre_hook or post_hook
+\layout Enumerate
+
+Plugin starts transaction.
+
+\layout Enumerate
+
+Plugin do what it should.
+\layout Enumerate
+
+Plugin commit transaction.
+\layout Standard
+
+Transaction MUST be commited where it was started.
+ It is not allowed to start it in pre_hook and commit in post_hook.
+\layout Subsubsection
+
+Special case: write operation
+\layout Standard
+
+Write operation cannot be encapsulated in SMFS transaction because really
+ this operation starts transactions for each prepare_write/commit_write
+ so current transaction can be easily overflowed.
+ At the moment SMFS do nothing with write, but plugin like KML do log for
+ write operation.
+ Therefore in case of crush we will not be able to restore KML log synced
+ with FS operations.
+\layout Subsection
+
+Locking
+\layout Subsubsection
+
+Inode operations
+\layout Enumerate
+
+VFS calls SMFS operation and takes i_sem.
+\layout Enumerate
+
+SMFS does pre_hook operation, calls backstore FS operation and then - post_hook.
+\layout Standard
+
+All operations are protected by i_sem in SMFS inode.
+ But there are possible situations when plugin hook sequence will be like
+ this:
+\layout Itemize
+
+pre_hook(inode1)
+\layout Itemize
+
+pre_hook(inode2)
+\layout Itemize
+
+post_hooks(...)
+\layout Standard
+
+This is possible because inode1 may block and operation for inode2 will
+ starts.
+ Plugin should be aware about that and use additional protection if it is
+ needed.
+
+\layout Subsubsection
+
+File operations
+\layout Standard
+
+File operations are not protected right now.
+ SMFS can also use i_sem to protect these operations, but previous words
+ about operations order are make sense here also.
+\the_end
git://git.whamcloud.com / fs / lustre-release.git / commitdiff