Stackfs: generic stackable filesystem helper functions
Subsystems (2)
PATCH 1/1 Stackfs: generic stackable filesystem helper functions
The following patches introduce stackfs_copy_* functions. These functions copy inode attributes (such as {a,c,m}time, mode, etc.) from one inode to another.
While, intended for stackable filesystems any portion of the kernel wishing to copy inode attributes can use them.
This series consists of two patches, the first introduces the stackfs functions, and the second makes eCryptfs a user.
4 files changed, 85 insertions(+), 59 deletions(-) fs/ecryptfs/file.c | 3 + fs/ecryptfs/inode.c | 71 +++++++++------------------------------------- fs/ecryptfs/main.c | 5 +-- include/linux/stack_fs.h | 65 ++++++++++++++++++++++++++++++++++++++++++
PATCH 1/1 Stackfs: Introduce stackfs_copy_{attr,inode}_*
From: Josef "Jeff" Sipek <[email protected]>
The following patch introduces several stackfs_copy_* functions which allow stackable filesystems (such as eCryptfs and Unionfs) to easily copy over (currently only) inode attributes. This prevents code duplication and allows for code reuse.
---
1 file changed, 65 insertions(+)
include/linux/stack_fs.h | 65 ++++++++++++++++++++++++++++++++++++++++++++++
include/linux/stack_fs.h new file +65 -0
@@ -0,0 +1,65 @@ +#ifndef _LINUX_STACK_FS_H +#define _LINUX_STACK_FS_H + +/* This file defines generic functions used primarily by stackable +* filesystems +*/ + +static inline void stackfs_copy_inode_size (struct inode *dst, +const struct inode *src) +{ +i_size_write (dst,i_size_read ((struct inode *)src)); + dst->i_blocks = src->i_blocks; +} + +static inline void stackfs_copy_attr_atime (struct inode *dest, +const struct inode *src) +{ + dest->i_atime = src->i_atime; +} + +static inline void stackfs_copy_attr_times (struct inode *dest, +const struct inode *src) +{ + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; +} + +static inline void stackfs_copy_attr_timesizes (struct inode *dest, +const struct inode *src) +{ + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; +stackfs_copy_inode_size (dest, src); +} + +static inline void __stackfs_copy_attr_all (struct inode *dest, +const struct inode *src, +int (*get_nlinks)(struct inode *)) +{ +if (!get_nlinks) + dest->i_nlink = src->i_nlink; +else + dest->i_nlink =get_nlinks (dest); + + dest->i_mode = src->i_mode; + dest->i_uid = src->i_uid; + dest->i_gid = src->i_gid; + dest->i_rdev = src->i_rdev; + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; + dest->i_blkbits = src->i_blkbits; + dest->i_flags = src->i_flags; +} + +static inline void stackfs_copy_attr_all (struct inode *dest, +const struct inode *src) +{ +__stackfs_copy_attr_all (dest, src,NULL ); +} + +#endif /* _LINUX_STACK_FS_H */ +
Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
include/linux/stack_fs.h line 11
Quoted 4 lines · Josef Jeff Sipek
include/linux/stack_fs.h new file +4
-0+static inline void stackfs_copy_inode_size (struct inode *dst, +const struct inode *src) +{ +i_size_write (dst,i_size_read ((struct inode *)src));
Instead of casting, I'd change the signature of i_size_read.
[PATCH] Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
Quoted 8 lines · Josef Jeff Sipek
Review
Quoted 4 lines
(quoted diff) +4
-0+static inline void stackfs_copy_inode_size(struct inode *dst, + const struct inode *src) +{ + i_size_write(dst, i_size_read((struct inode *)src));Instead of casting, I'd change the signature of
i_size_read.
Change the signature of i_size_read(), IMINOR() and IMAJOR() because they, or the functions they call, will never modify the argument. Compile-tested on x86 allmodconfig.
include/linux/fs.h +3 -3
@@ -633,7 +633,7 @@ enum inode_i_mutex_lock_class* cmpxchg8b without the need of the lock prefix). For SMP compiles * and 64bit archs it makes no difference if preempt is enabled or not. */ -static inline loff_t i_size_read (struct inode *inode) +static inline loff_t i_size_read (const struct inode *inode) {#if BITS_PER_LONG ==32 && defined(CONFIG_SMP )loff_t i_size;
@@ -672,12 +672,12 @@ static inline void i_size_write(struct i#endif } -static inline unsigned iminor (struct inode *inode) +static inline unsigned iminor (const struct inode *inode) {return MINOR (inode->i_rdev); } -static inline unsigned imajor (struct inode *inode) +static inline unsigned imajor (const struct inode *inode) {return MAJOR (inode->i_rdev); }
#<EOF>
-`J'
Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
In reply to JosefJeff SipekQuoted 6 lines · On Fri, 13 Oct 2006 07:18:49 -0400 Josef "Jeff" Sipek <[email protected]>…
From: Josef "Jeff" Sipek <[email protected]>
The following patch introduces several stackfs_copy_* functions which allow stackable filesystems (such as eCryptfs and Unionfs) to easily copy over (currently only) inode attributes. This prevents code duplication and allows for code reuse.
Fair enough.
Quoted 1 line
include/linux/stack_fs.h| 65 ++++++++++++++++++++++++++++++++++++++++++++++
The name stack_fs implies that there's a filesystem called stackfs. Only there isn't. I wonder if we can choose a better name for all of this. Maybe fs_stack_*?
Quoted 9 lines
include/linux/stack_fs.h new file +3
-0@@ -0,0 +1,65 @@ +#ifndef _LINUX_STACK_FS_H +#define _LINUX_STACK_FS_H +
include/linux/stack_fs.h line 13
Quoted 10 lines
include/linux/stack_fs.h new file +10
-0+/* This file defines generic functions used primarily by stackable +* filesystems +*/ + +static inline void stackfs_copy_inode_size (struct inode *dst, +const struct inode *src) +{ +i_size_write (dst,i_size_read ((struct inode *)src)); + dst->i_blocks = src->i_blocks; +}
What are the locking requirements for these functions? Presumably the caller must hold i_mutex on at least the source inode, and perhaps the destination one?
If i_mutex is held, i_size_read() isn't needed.
If i_mutex is held, i_size_write() isn't needed either.
So please document the locking requirements via source comments and then see if this can be simplified.
If this function stays as it is, it's too big to inline.
Quoted 21 lines
include/linux/stack_fs.h new file +21
-0+static inline void stackfs_copy_attr_atime (struct inode *dest, +const struct inode *src) +{ + dest->i_atime = src->i_atime; +} + +static inline void stackfs_copy_attr_times (struct inode *dest, +const struct inode *src) +{ + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; +} + +static inline void stackfs_copy_attr_timesizes (struct inode *dest, +const struct inode *src) +{ + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime; + dest->i_ctime = src->i_ctime; +stackfs_copy_inode_size (dest, src);
include/linux/stack_fs.h line 45
Quoted 10 lines
include/linux/stack_fs.h new file +10
-0+} + +static inline void __stackfs_copy_attr_all (struct inode *dest, +const struct inode *src, +int (*get_nlinks)(struct inode *)) +{ +if (!get_nlinks) + dest->i_nlink = src->i_nlink; +else + dest->i_nlink =get_nlinks (dest);
I cannot find a get_nlinks() in 2.6.19-rc2?
Quoted 6 lines
include/linux/stack_fs.h new file +6
-0+ dest->i_mode = src->i_mode; + dest->i_uid = src->i_uid; + dest->i_gid = src->i_gid; + dest->i_rdev = src->i_rdev; + dest->i_atime = src->i_atime; + dest->i_mtime = src->i_mtime;
include/linux/stack_fs.h line 62
Quoted 10 lines
include/linux/stack_fs.h new file +9
-0+ dest->i_ctime = src->i_ctime; + dest->i_blkbits = src->i_blkbits; + dest->i_flags = src->i_flags; +} +static inline void stackfs_copy_attr_all (struct inode *dest, +const struct inode *src) +{ +__stackfs_copy_attr_all (dest, src,NULL ); +}
Many of these functions are too large to be inlined. Suggest they be placed in fs/fs-stack.c (or whatever we call it).
The functions themselves seem a bit arbitrary. stackfs_copy_attr_timesizes() copy the three timestamps and the size. Is there actually any methodical reason for that, or is it simply some sequence which happens to have been observed in ecryptfs?
And please - if I asked these questions when reviewing the patch, others will ask them when reading the code two years from now. So please treat my questions as "gosh, I should have put a comment in there".
Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
Quoted 7 lines · Andrew Morton
Quoted 1 line · On Fri, 13 Oct 2006 07:18:49 -0400 Josef "Jeff" Sipek <[email protected]>…
include/linux/stack_fs.h| 65 ++++++++++++++++++++++++++++++++++++++++++++++The name
stack_fsimplies that there's a filesystem called stackfs. Only there isn't. I wonder if we can choose a better name for all of this. Maybe fs_stack_*?
fs_stack_* sounds good to me.
Quoted 10 lines
What are the locking requirements for these functions? Presumably the caller must hold
i_mutexon at least the source inode, and perhaps the destination one?If
i_mutexis held,i_size_read()isn't needed.If
i_mutexis held,i_size_write()isn't needed either.So please document the locking requirements via source comments and then see if this can be simplified.
Fair enough. I'll submit an updated version.
Quoted 10 lines
include/linux/stack_fs.h line 45
Quoted 8 lines
include/linux/stack_fs.h new file +8
-0+static inline void __stackfs_copy_attr_all (struct inode *dest, +const struct inode *src, +int (*get_nlinks)(struct inode *)) +{ +if (!get_nlinks) + dest->i_nlink = src->i_nlink; +else + dest->i_nlink =get_nlinks (dest);I cannot find a
get_nlinks()in 2.6.19-rc2?
It is the last argument to the function. Perhaps the function name is deceiving.
Quoted 2 lines
Many of these functions are too large to be inlined. Suggest they be placed in
fs/fs-stack.c(or whatever we call it).
Ack. As a rule of thumb, for functions like these (laundry list of assignments), what's a good threshold?
Quoted 4 lines
The functions themselves seem a bit arbitrary.
stackfs_copy_attr_timesizes()copy the three timestamps and the size. Is there actually any methodical reason for that, or is it simply some sequence which happens to have been observed in ecryptfs?
These functions come from the FiST templates [1]. Some of these can definitely removed, or split up.
Quoted 3 lines
And please - if I asked these questions when reviewing the patch, others will ask them when reading the code two years from now. So please treat my questions as "gosh, I should have put a comment in there".
That's exactly how I look at it.
Josef "Jeff" Sipek.
Signature · 2 lines
Bad pun of the week: The formula 1 control computer suffered from a race condition
Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
On Fri, Oct 13, 2006 at 04:07:05PM -0400, Josef Sipek wrote:
...
Quoted 2 lines
These functions come from the FiST templates [1]. Some of these can definitely removed, or split up.
Forgot to say what [1] was...
[1] http://www.filesystems.org/
Josef "Jeff" Sipek.
Signature · 5 lines
It used to be said [...] that AIX looks like one space alien discovered Unix, and described it to another different space alien who then implemented AIX. But their universal translators were broken and they'd had to gesture a lot. - Paul Tomblin
Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
In reply to Josef SipekQuoted 11 lines · On Fri, 13 Oct 2006 16:07:05 -0400 Josef Sipek <[email protected]> wrote:
Quoted 10 lines
include/linux/stack_fs.h line 45
Quoted 8 lines
include/linux/stack_fs.h new file +8
-0+static inline void __stackfs_copy_attr_all (struct inode *dest, +const struct inode *src, +int (*get_nlinks)(struct inode *)) +{ +if (!get_nlinks) + dest->i_nlink = src->i_nlink; +else + dest->i_nlink =get_nlinks (dest);I cannot find a
get_nlinks()in 2.6.19-rc2?
Quoted 2 lines
It is the last argument to the function. Perhaps the function name is deceiving.
doh.
That's why us old farts like to see
dest->i_nlink = (*get_nlinks)(dest);
Re: [PATCH 1 of 2] Stackfs: Introduce stackfs_copy_{attr,inode}_*
In reply to Josef SipekQuoted 3 lines
Quoted 3 lines
Many of these functions are too large to be inlined. Suggest they be placed in
fs/fs-stack.c(or whatever we call it).
fs/stack.c would probably be enough -- fs/fs-stack.c is like include/linux/reiserfs_fs.h
Quoted 2 lines
Ack. As a rule of thumb, for functions like these (laundry list of assignments), what's a good threshold?
3 or 4 I guess. Might want to take a look at other static-inline functions.
-`J'
PATCH 1/1 eCryptfs: Use stackfs's generic copy inode attr functions
In reply to JosefJeff SipekFrom: Josef "Jeff" Sipek <[email protected]>
Replace eCryptfs specific code & calls with the more generic stackfs equivalents and remove the eCryptfs specific functions.
---
3 files changed, 20 insertions(+), 59 deletions(-) fs/ecryptfs/file.c | 3 +- fs/ecryptfs/inode.c | 71 ++++++++++----------------------------------------- fs/ecryptfs/main.c | 5 ++-
fs/ecryptfs/file.c +2 -1
@@ -30,6 +30,7 @@#include <linux/security.h> #include <linux/smp_lock.h> #include <linux/compat.h> +#include <linux/stack_fs.h> #include "ecryptfs_kernel.h" /**
@@ -192,7 +193,7 @@ retry: goto retry; file->f_pos = lower_file->f_pos;if (rc >=0 ) -ecryptfs_copy_attr_atime (inode, lower_file->f_dentry->d_inode); +stackfs_copy_attr_atime (inode, lower_file->f_dentry->d_inode);return rc; }
fs/ecryptfs/inode.c +15 -56
150 lines not on this page · raw message
fs/ecryptfs/main.c +3 -2
@@ -35,6 +35,7 @@#include <linux/pagemap.h> #include <linux/key.h> #include <linux/parser.h> +#include <linux/stack_fs.h> #include "ecryptfs_kernel.h" /**
@@ -115,10 +116,10 @@ int ecryptfs_interpose(struct dentry *lod_add (dentry, inode); elsed_instantiate (dentry ,inode ); -ecryptfs_copy_attr_all (inode, lower_inode); +stackfs_copy_attr_all (inode, lower_inode);/* This size will be overwritten for real files w/ headers and * other metadata */ -ecryptfs_copy_inode_size (inode, lower_inode); +stackfs_copy_inode_size (inode, lower_inode); out:return rc; }
Hi Jeff,
Quoted 3 lines · Josef Jeff Sipek
The following patches introduce stackfs_copy_* functions. These functions copy inode attributes (such as {a,c,m}time, mode, etc.) from one inode to another.
Looks ok to me. Thanks!