kernel_optimize_test/fs/gfs2/ops_file.c

812 lines
19 KiB
C
Raw Normal View History

/*
* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
* Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
*
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
* of the GNU General Public License version 2.
*/
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/completion.h>
#include <linux/buffer_head.h>
#include <linux/pagemap.h>
#include <linux/uio.h>
#include <linux/blkdev.h>
#include <linux/mm.h>
#include <linux/smp_lock.h>
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 19:50:51 +08:00
#include <linux/fs.h>
#include <linux/gfs2_ondisk.h>
#include <linux/ext2_fs.h>
#include <linux/crc32.h>
#include <linux/iflags.h>
#include <asm/uaccess.h>
#include "gfs2.h"
#include "lm_interface.h"
#include "incore.h"
#include "bmap.h"
#include "dir.h"
#include "glock.h"
#include "glops.h"
#include "inode.h"
#include "lm.h"
#include "log.h"
#include "meta_io.h"
#include "ops_file.h"
#include "ops_vm.h"
#include "quota.h"
#include "rgrp.h"
#include "trans.h"
#include "util.h"
#include "eaops.h"
/* "bad" is for NFS support */
struct filldir_bad_entry {
char *fbe_name;
unsigned int fbe_length;
u64 fbe_offset;
struct gfs2_inum fbe_inum;
unsigned int fbe_type;
};
struct filldir_bad {
struct gfs2_sbd *fdb_sbd;
struct filldir_bad_entry *fdb_entry;
unsigned int fdb_entry_num;
unsigned int fdb_entry_off;
char *fdb_name;
unsigned int fdb_name_size;
unsigned int fdb_name_off;
};
/* For regular, non-NFS */
struct filldir_reg {
struct gfs2_sbd *fdr_sbd;
int fdr_prefetch;
filldir_t fdr_filldir;
void *fdr_opaque;
};
/*
* Most fields left uninitialised to catch anybody who tries to
* use them. f_flags set to prevent file_accessed() from touching
* any other part of this. Its use is purely as a flag so that we
* know (in readpage()) whether or not do to locking.
*/
struct file gfs2_internal_file_sentinal = {
.f_flags = O_NOATIME|O_RDONLY,
};
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 19:50:51 +08:00
static int gfs2_read_actor(read_descriptor_t *desc, struct page *page,
unsigned long offset, unsigned long size)
{
char *kaddr;
unsigned long count = desc->count;
if (size > count)
size = count;
kaddr = kmap(page);
memcpy(desc->arg.buf, kaddr + offset, size);
kunmap(page);
desc->count = count - size;
desc->written += size;
desc->arg.buf += size;
return size;
}
int gfs2_internal_read(struct gfs2_inode *ip, struct file_ra_state *ra_state,
char *buf, loff_t *pos, unsigned size)
{
struct inode *inode = &ip->i_inode;
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 19:50:51 +08:00
read_descriptor_t desc;
desc.written = 0;
desc.arg.buf = buf;
desc.count = size;
desc.error = 0;
do_generic_mapping_read(inode->i_mapping, ra_state,
&gfs2_internal_file_sentinal, pos, &desc,
gfs2_read_actor);
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 19:50:51 +08:00
return desc.written ? desc.written : desc.error;
}
/**
* gfs2_llseek - seek to a location in a file
* @file: the file
* @offset: the offset
* @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END)
*
* SEEK_END requires the glock for the file because it references the
* file's size.
*
* Returns: The new offset, or errno
*/
static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin)
{
struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
struct gfs2_holder i_gh;
loff_t error;
if (origin == 2) {
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
&i_gh);
if (!error) {
error = remote_llseek(file, offset, origin);
gfs2_glock_dq_uninit(&i_gh);
}
} else
error = remote_llseek(file, offset, origin);
return error;
}
/**
* filldir_reg_func - Report a directory entry to the caller of gfs2_dir_read()
* @opaque: opaque data used by the function
* @name: the name of the directory entry
* @length: the length of the name
* @offset: the entry's offset in the directory
* @inum: the inode number the entry points to
* @type: the type of inode the entry points to
*
* Returns: 0 on success, 1 if buffer full
*/
static int filldir_reg_func(void *opaque, const char *name, unsigned int length,
u64 offset, struct gfs2_inum *inum,
unsigned int type)
{
struct filldir_reg *fdr = (struct filldir_reg *)opaque;
struct gfs2_sbd *sdp = fdr->fdr_sbd;
int error;
error = fdr->fdr_filldir(fdr->fdr_opaque, name, length, offset,
inum->no_addr, type);
if (error)
return 1;
if (fdr->fdr_prefetch && !(length == 1 && *name == '.')) {
gfs2_glock_prefetch_num(sdp,
inum->no_addr, &gfs2_inode_glops,
LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY);
gfs2_glock_prefetch_num(sdp,
inum->no_addr, &gfs2_iopen_glops,
LM_ST_SHARED, LM_FLAG_TRY);
}
return 0;
}
/**
* readdir_reg - Read directory entries from a directory
* @file: The directory to read from
* @dirent: Buffer for dirents
* @filldir: Function used to do the copying
*
* Returns: errno
*/
static int readdir_reg(struct file *file, void *dirent, filldir_t filldir)
{
struct inode *dir = file->f_mapping->host;
struct gfs2_inode *dip = GFS2_I(dir);
struct filldir_reg fdr;
struct gfs2_holder d_gh;
u64 offset = file->f_pos;
int error;
fdr.fdr_sbd = GFS2_SB(dir);
fdr.fdr_prefetch = 1;
fdr.fdr_filldir = filldir;
fdr.fdr_opaque = dirent;
gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh);
error = gfs2_glock_nq_atime(&d_gh);
if (error) {
gfs2_holder_uninit(&d_gh);
return error;
}
error = gfs2_dir_read(dir, &offset, &fdr, filldir_reg_func);
gfs2_glock_dq_uninit(&d_gh);
file->f_pos = offset;
return error;
}
/**
* filldir_bad_func - Report a directory entry to the caller of gfs2_dir_read()
* @opaque: opaque data used by the function
* @name: the name of the directory entry
* @length: the length of the name
* @offset: the entry's offset in the directory
* @inum: the inode number the entry points to
* @type: the type of inode the entry points to
*
* For supporting NFS.
*
* Returns: 0 on success, 1 if buffer full
*/
static int filldir_bad_func(void *opaque, const char *name, unsigned int length,
u64 offset, struct gfs2_inum *inum,
unsigned int type)
{
struct filldir_bad *fdb = (struct filldir_bad *)opaque;
struct gfs2_sbd *sdp = fdb->fdb_sbd;
struct filldir_bad_entry *fbe;
if (fdb->fdb_entry_off == fdb->fdb_entry_num ||
fdb->fdb_name_off + length > fdb->fdb_name_size)
return 1;
fbe = &fdb->fdb_entry[fdb->fdb_entry_off];
fbe->fbe_name = fdb->fdb_name + fdb->fdb_name_off;
memcpy(fbe->fbe_name, name, length);
fbe->fbe_length = length;
fbe->fbe_offset = offset;
fbe->fbe_inum = *inum;
fbe->fbe_type = type;
fdb->fdb_entry_off++;
fdb->fdb_name_off += length;
if (!(length == 1 && *name == '.')) {
gfs2_glock_prefetch_num(sdp,
inum->no_addr, &gfs2_inode_glops,
LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY);
gfs2_glock_prefetch_num(sdp,
inum->no_addr, &gfs2_iopen_glops,
LM_ST_SHARED, LM_FLAG_TRY);
}
return 0;
}
/**
* readdir_bad - Read directory entries from a directory
* @file: The directory to read from
* @dirent: Buffer for dirents
* @filldir: Function used to do the copying
*
* For supporting NFS.
*
* Returns: errno
*/
static int readdir_bad(struct file *file, void *dirent, filldir_t filldir)
{
struct inode *dir = file->f_mapping->host;
struct gfs2_inode *dip = GFS2_I(dir);
struct gfs2_sbd *sdp = GFS2_SB(dir);
struct filldir_reg fdr;
unsigned int entries, size;
struct filldir_bad *fdb;
struct gfs2_holder d_gh;
u64 offset = file->f_pos;
unsigned int x;
struct filldir_bad_entry *fbe;
int error;
entries = gfs2_tune_get(sdp, gt_entries_per_readdir);
size = sizeof(struct filldir_bad) +
entries * (sizeof(struct filldir_bad_entry) + GFS2_FAST_NAME_SIZE);
fdb = kzalloc(size, GFP_KERNEL);
if (!fdb)
return -ENOMEM;
fdb->fdb_sbd = sdp;
fdb->fdb_entry = (struct filldir_bad_entry *)(fdb + 1);
fdb->fdb_entry_num = entries;
fdb->fdb_name = ((char *)fdb) + sizeof(struct filldir_bad) +
entries * sizeof(struct filldir_bad_entry);
fdb->fdb_name_size = entries * GFS2_FAST_NAME_SIZE;
gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh);
error = gfs2_glock_nq_atime(&d_gh);
if (error) {
gfs2_holder_uninit(&d_gh);
goto out;
}
error = gfs2_dir_read(dir, &offset, fdb, filldir_bad_func);
gfs2_glock_dq_uninit(&d_gh);
fdr.fdr_sbd = sdp;
fdr.fdr_prefetch = 0;
fdr.fdr_filldir = filldir;
fdr.fdr_opaque = dirent;
for (x = 0; x < fdb->fdb_entry_off; x++) {
fbe = &fdb->fdb_entry[x];
error = filldir_reg_func(&fdr,
fbe->fbe_name, fbe->fbe_length,
fbe->fbe_offset,
&fbe->fbe_inum, fbe->fbe_type);
if (error) {
file->f_pos = fbe->fbe_offset;
error = 0;
goto out;
}
}
file->f_pos = offset;
out:
kfree(fdb);
return error;
}
/**
* gfs2_readdir - Read directory entries from a directory
* @file: The directory to read from
* @dirent: Buffer for dirents
* @filldir: Function used to do the copying
*
* Returns: errno
*/
static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir)
{
int error;
if (strcmp(current->comm, "nfsd") != 0)
error = readdir_reg(file, dirent, filldir);
else
error = readdir_bad(file, dirent, filldir);
return error;
}
static const u32 iflags_to_gfs2[32] = {
[iflag_Sync] = GFS2_DIF_SYNC,
[iflag_Immutable] = GFS2_DIF_IMMUTABLE,
[iflag_Append] = GFS2_DIF_APPENDONLY,
[iflag_NoAtime] = GFS2_DIF_NOATIME,
[iflag_Index] = GFS2_DIF_EXHASH,
[iflag_JournalData] = GFS2_DIF_JDATA,
[iflag_DirectIO] = GFS2_DIF_DIRECTIO,
};
static const u32 gfs2_to_iflags[32] = {
[gfs2fl_Sync] = IFLAG_SYNC,
[gfs2fl_Immutable] = IFLAG_IMMUTABLE,
[gfs2fl_AppendOnly] = IFLAG_APPEND,
[gfs2fl_NoAtime] = IFLAG_NOATIME,
[gfs2fl_ExHash] = IFLAG_INDEX,
[gfs2fl_Jdata] = IFLAG_JOURNAL_DATA,
[gfs2fl_Directio] = IFLAG_DIRECTIO,
[gfs2fl_InheritDirectio] = IFLAG_DIRECTIO,
[gfs2fl_InheritJdata] = IFLAG_JOURNAL_DATA,
};
static int gfs2_get_flags(struct file *filp, u32 __user *ptr)
{
struct inode *inode = filp->f_dentry->d_inode;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_holder gh;
int error;
u32 iflags;
gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh);
error = gfs2_glock_nq_m_atime(1, &gh);
if (error)
return error;
iflags = iflags_cvt(gfs2_to_iflags, ip->i_di.di_flags);
if (put_user(iflags, ptr))
error = -EFAULT;
gfs2_glock_dq_m(1, &gh);
gfs2_holder_uninit(&gh);
return error;
}
/* Flags that can be set by user space */
#define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \
GFS2_DIF_DIRECTIO| \
GFS2_DIF_IMMUTABLE| \
GFS2_DIF_APPENDONLY| \
GFS2_DIF_NOATIME| \
GFS2_DIF_SYNC| \
GFS2_DIF_SYSTEM| \
GFS2_DIF_INHERIT_DIRECTIO| \
GFS2_DIF_INHERIT_JDATA)
/**
* gfs2_set_flags - set flags on an inode
* @inode: The inode
* @flags: The flags to set
* @mask: Indicates which flags are valid
*
*/
static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask)
{
struct inode *inode = filp->f_dentry->d_inode;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_sbd *sdp = GFS2_SB(inode);
struct buffer_head *bh;
struct gfs2_holder gh;
int error;
u32 new_flags, flags;
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
if (error)
return error;
flags = ip->i_di.di_flags;
new_flags = (flags & ~mask) | (reqflags & mask);
if ((new_flags ^ flags) == 0)
goto out;
if (S_ISDIR(inode->i_mode)) {
if ((new_flags ^ flags) & GFS2_DIF_JDATA)
new_flags ^= (GFS2_DIF_JDATA|GFS2_DIF_INHERIT_JDATA);
if ((new_flags ^ flags) & GFS2_DIF_DIRECTIO)
new_flags ^= (GFS2_DIF_DIRECTIO|GFS2_DIF_INHERIT_DIRECTIO);
}
error = -EINVAL;
if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET)
goto out;
error = -EPERM;
if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE))
goto out;
if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY))
goto out;
if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) &&
!capable(CAP_LINUX_IMMUTABLE))
goto out;
if (!IS_IMMUTABLE(inode)) {
error = permission(inode, MAY_WRITE, NULL);
if (error)
goto out;
}
error = gfs2_trans_begin(sdp, RES_DINODE, 0);
if (error)
goto out;
error = gfs2_meta_inode_buffer(ip, &bh);
if (error)
goto out_trans_end;
gfs2_trans_add_bh(ip->i_gl, bh, 1);
ip->i_di.di_flags = new_flags;
gfs2_dinode_out(&ip->i_di, bh->b_data);
brelse(bh);
out_trans_end:
gfs2_trans_end(sdp);
out:
gfs2_glock_dq_uninit(&gh);
return error;
}
static int gfs2_set_flags(struct file *filp, u32 __user *ptr)
{
u32 iflags, gfsflags;
if (get_user(iflags, ptr))
return -EFAULT;
gfsflags = iflags_cvt(iflags_to_gfs2, iflags);
return do_gfs2_set_flags(filp, gfsflags, ~0);
}
static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
switch(cmd) {
case IFLAGS_GET_IOC:
return gfs2_get_flags(filp, (u32 __user *)arg);
case IFLAGS_SET_IOC:
return gfs2_set_flags(filp, (u32 __user *)arg);
}
return -ENOTTY;
}
/**
* gfs2_mmap -
* @file: The file to map
* @vma: The VMA which described the mapping
*
* Returns: 0 or error code
*/
static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
{
struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
struct gfs2_holder i_gh;
int error;
gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
error = gfs2_glock_nq_atime(&i_gh);
if (error) {
gfs2_holder_uninit(&i_gh);
return error;
}
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 19:50:51 +08:00
/* This is VM_MAYWRITE instead of VM_WRITE because a call
to mprotect() can turn on VM_WRITE later. */
if ((vma->vm_flags & (VM_MAYSHARE | VM_MAYWRITE)) ==
(VM_MAYSHARE | VM_MAYWRITE))
vma->vm_ops = &gfs2_vm_ops_sharewrite;
else
vma->vm_ops = &gfs2_vm_ops_private;
gfs2_glock_dq_uninit(&i_gh);
return error;
}
/**
* gfs2_open - open a file
* @inode: the inode to open
* @file: the struct file for this opening
*
* Returns: errno
*/
static int gfs2_open(struct inode *inode, struct file *file)
{
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_holder i_gh;
struct gfs2_file *fp;
int error;
fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
if (!fp)
return -ENOMEM;
mutex_init(&fp->f_fl_mutex);
gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
file->private_data = fp;
if (S_ISREG(ip->i_di.di_mode)) {
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
&i_gh);
if (error)
goto fail;
if (!(file->f_flags & O_LARGEFILE) &&
ip->i_di.di_size > MAX_NON_LFS) {
error = -EFBIG;
goto fail_gunlock;
}
/* Listen to the Direct I/O flag */
if (ip->i_di.di_flags & GFS2_DIF_DIRECTIO)
file->f_flags |= O_DIRECT;
gfs2_glock_dq_uninit(&i_gh);
}
return 0;
fail_gunlock:
gfs2_glock_dq_uninit(&i_gh);
fail:
file->private_data = NULL;
kfree(fp);
return error;
}
/**
* gfs2_close - called to close a struct file
* @inode: the inode the struct file belongs to
* @file: the struct file being closed
*
* Returns: errno
*/
static int gfs2_close(struct inode *inode, struct file *file)
{
struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
struct gfs2_file *fp;
fp = file->private_data;
file->private_data = NULL;
if (gfs2_assert_warn(sdp, fp))
return -EIO;
kfree(fp);
return 0;
}
/**
* gfs2_fsync - sync the dirty data for a file (across the cluster)
* @file: the file that points to the dentry (we ignore this)
* @dentry: the dentry that points to the inode to sync
*
* Returns: errno
*/
static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
{
struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
return 0;
}
/**
* gfs2_lock - acquire/release a posix lock on a file
* @file: the file pointer
* @cmd: either modify or retrieve lock state, possibly wait
* @fl: type and range of lock
*
* Returns: errno
*/
static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
{
struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host);
struct lm_lockname name =
{ .ln_number = ip->i_num.no_addr,
.ln_type = LM_TYPE_PLOCK };
if (!(fl->fl_flags & FL_POSIX))
return -ENOLCK;
if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID)
return -ENOLCK;
if (sdp->sd_args.ar_localflocks) {
if (IS_GETLK(cmd)) {
struct file_lock tmp;
int ret;
ret = posix_test_lock(file, fl, &tmp);
fl->fl_type = F_UNLCK;
if (ret)
memcpy(fl, &tmp, sizeof(struct file_lock));
return 0;
} else {
return posix_lock_file_wait(file, fl);
}
}
if (IS_GETLK(cmd))
return gfs2_lm_plock_get(sdp, &name, file, fl);
else if (fl->fl_type == F_UNLCK)
return gfs2_lm_punlock(sdp, &name, file, fl);
else
return gfs2_lm_plock(sdp, &name, file, cmd, fl);
}
static int do_flock(struct file *file, int cmd, struct file_lock *fl)
{
struct gfs2_file *fp = file->private_data;
struct gfs2_holder *fl_gh = &fp->f_fl_gh;
struct gfs2_inode *ip = GFS2_I(file->f_dentry->d_inode);
struct gfs2_glock *gl;
unsigned int state;
int flags;
int error = 0;
state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED;
flags = ((IS_SETLKW(cmd)) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE;
mutex_lock(&fp->f_fl_mutex);
gl = fl_gh->gh_gl;
if (gl) {
if (fl_gh->gh_state == state)
goto out;
gfs2_glock_hold(gl);
flock_lock_file_wait(file,
&(struct file_lock){.fl_type = F_UNLCK});
gfs2_glock_dq_uninit(fl_gh);
} else {
error = gfs2_glock_get(GFS2_SB(&ip->i_inode),
ip->i_num.no_addr, &gfs2_flock_glops,
CREATE, &gl);
if (error)
goto out;
}
gfs2_holder_init(gl, state, flags, fl_gh);
gfs2_glock_put(gl);
error = gfs2_glock_nq(fl_gh);
if (error) {
gfs2_holder_uninit(fl_gh);
if (error == GLR_TRYFAILED)
error = -EAGAIN;
} else {
error = flock_lock_file_wait(file, fl);
gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
}
out:
mutex_unlock(&fp->f_fl_mutex);
return error;
}
static void do_unflock(struct file *file, struct file_lock *fl)
{
struct gfs2_file *fp = file->private_data;
struct gfs2_holder *fl_gh = &fp->f_fl_gh;
mutex_lock(&fp->f_fl_mutex);
flock_lock_file_wait(file, fl);
if (fl_gh->gh_gl)
gfs2_glock_dq_uninit(fl_gh);
mutex_unlock(&fp->f_fl_mutex);
}
/**
* gfs2_flock - acquire/release a flock lock on a file
* @file: the file pointer
* @cmd: either modify or retrieve lock state, possibly wait
* @fl: type and range of lock
*
* Returns: errno
*/
static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl)
{
struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host);
if (!(fl->fl_flags & FL_FLOCK))
return -ENOLCK;
if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID)
return -ENOLCK;
if (sdp->sd_args.ar_localflocks)
return flock_lock_file_wait(file, fl);
if (fl->fl_type == F_UNLCK) {
do_unflock(file, fl);
return 0;
} else
return do_flock(file, cmd, fl);
}
const struct file_operations gfs2_file_fops = {
.llseek = gfs2_llseek,
.read = generic_file_read,
.readv = generic_file_readv,
.aio_read = generic_file_aio_read,
[GFS2] Make journaled data files identical to normal files on disk This is a very large patch, with a few still to be resolved issues so you might want to check out the previous head of the tree since this is known to be unstable. Fixes for the various bugs will be forthcoming shortly. This patch removes the special data format which has been used up till now for journaled data files. Directories still retain the old format so that they will remain on disk compatible with earlier releases. As a result you can now do the following with journaled data files: 1) mmap them 2) export them over NFS 3) convert to/from normal files whenever you want to (the zero length restriction is gone) In addition the level at which GFS' locking is done has changed for all files (since they all now use the page cache) such that the locking is done at the page cache level rather than the level of the fs operations. This should mean that things like loopback mounts and other things which touch the page cache directly should now work. Current known issues: 1. There is a lock mode inversion problem related to the resource group hold function which needs to be resolved. 2. Any significant amount of I/O causes an oops with an offset of hex 320 (NULL pointer dereference) which appears to be related to a journaled data buffer appearing on a list where it shouldn't be. 3. Direct I/O writes are disabled for the time being (will reappear later) 4. There is probably a deadlock between the page lock and GFS' locks under certain combinations of mmap and fs operation I/O. 5. Issue relating to ref counting on internally used inodes causes a hang on umount (discovered before this patch, and not fixed by it) 6. One part of the directory metadata is different from GFS1 and will need to be resolved before next release. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2006-02-08 19:50:51 +08:00
.write = generic_file_write,
.writev = generic_file_writev,
.aio_write = generic_file_aio_write,
.unlocked_ioctl = gfs2_ioctl,
.mmap = gfs2_mmap,
.open = gfs2_open,
.release = gfs2_close,
.fsync = gfs2_fsync,
.lock = gfs2_lock,
.sendfile = generic_file_sendfile,
.flock = gfs2_flock,
.splice_read = generic_file_splice_read,
.splice_write = generic_file_splice_write,
};
const struct file_operations gfs2_dir_fops = {
.readdir = gfs2_readdir,
.unlocked_ioctl = gfs2_ioctl,
.open = gfs2_open,
.release = gfs2_close,
.fsync = gfs2_fsync,
.lock = gfs2_lock,
.flock = gfs2_flock,
};