cifs: fix "Busy inodes after umount" issues (try #2)
cover cifs: fix "Busy inodes after umount" issues (try #2)
This is the second attempt at a patchset to fix a recent outbreak of "Busy inodes after umount" issues with cifs. The approach is basically the same.
The first patch in the original set has also been dropped since Al pointed out that the existing behavior is more clear. The second and third patches have also been merged since they touch the same areas of code. Finally a few bugs in the original set have been fixed as well.
With this set, I've not been able to reproduce any "busy inodes after umount" issues, but it would be nice to have some extra eyeballs and testing on this set before it goes in.
Jeff Layton (2):
cifs: move cifs_new_fileinfo call out of cifs_posix_open
cifs: pass instantiated filp back after open call
fs/cifs/cifsproto.h | 1 - fs/cifs/dir.c | 76 +++++++++++++++++++++++++++++--------------------- fs/cifs/file.c | 59 +++++++++------------------------------- 3 files changed, 57 insertions(+), 79 deletions(-)
PATCH 1/2 cifs: move cifs_new_fileinfo call out of cifs_posix_open
Having cifs_posix_open call cifs_new_fileinfo is problematic and inconsistent with how "regular" opens work. It's also buggy as cifs_reopen_file calls this function on a reconnect, which creates a new struct cifsFileInfo that just gets leaked.
Push it out into the callers. This also allows us to get rid of the "mnt" arg to cifs_posix_open.
Finally, in the event that a cifsFileInfo isn't or can't be created, we always want to close the filehandle out on the server as the client won't have a record of the filehandle and can't actually use it. Make sure that CIFSSMBClose is called in those cases.
--- fs/cifs/cifsproto.h | 1 - fs/cifs/dir.c | 43 ++++++++++++++++++------------------------- fs/cifs/file.c | 17 ++++++++++++----- 3 files changed, 30 insertions(+), 31 deletions(-)
fs/cifs/cifsproto.h +0 -1
@@ -106,7 +106,6 @@ extern struct cifsFileInfo *cifs_new_fileinfo(struct inode *newinode, __u16 fileHandle, struct file *file, struct vfsmount *mnt, unsigned int oflags);extern int cifs_posix_open (char *full_path,struct inode **pinode, -struct vfsmount *mnt,struct super_block *sb,int mode,int oflags,__u32 *poplock,__u16 *pnetfid,int xid);
fs/cifs/dir.c +18 -25
111 lines not on this page · raw message
fs/cifs/file.c +12 -5
@@ -299,8 +299,7 @@ int cifs_open(struct inode *inode, struct file *file)int oflags = (int )cifs_posix_convert_flags (file->f_flags); oflags |=SMB_O_CREAT ;/* can not refresh inode info since size could be stale */ - rc =cifs_posix_open (full_path, &inode, file->f_path.mnt, - inode->i_sb, + rc =cifs_posix_open (full_path, &inode, inode->i_sb, cifs_sb->mnt_file_mode/* ignored */ , oflags, &oplock, &netfid, xid);if (rc ==0 ) {
@@ -308,7 +307,16 @@ int cifs_open(struct inode *inode, struct file *file)/* no need for special case handling of setting mode on read only files needed here */ - pCifsFile =cifs_fill_filedata (file); + pCifsFile =cifs_new_fileinfo (inode, netfid, file, + file->f_path.mnt, + oflags); +if (pCifsFile ==NULL ) { +CIFSSMBClose (xid, tcon, netfid); + rc = -ENOMEM ; + goto out; + } + file->private_data = pCifsFile; +cifs_posix_open_inode_helper (inode, file, pCifsInode, oplock, netfid); goto out;
@@ -513,8 +521,7 @@ reopen_error_exit: le64_to_cpu(tcon->fsUnixInfo.Capability))) {int oflags = (int )cifs_posix_convert_flags (file->f_flags);/* can not refresh inode info since size could be stale */ - rc =cifs_posix_open (full_path,NULL , file->f_path.mnt, - inode->i_sb, + rc =cifs_posix_open (full_path,NULL , inode->i_sb, cifs_sb->mnt_file_mode/* ignored */ , oflags, &oplock, &netfid, xid);if (rc ==0 ) {
Signature · 1 line
1.6.6.1
PATCH 2/2 cifs: pass instantiated filp back after open call
In reply to Jeff LaytonThe current scheme of sticking open files on a list and assuming that cifs_open will scoop them off of it is broken and leads to "Busy inodes after umount..." errors at unmount time.
The problem is that there is no guarantee that cifs_open will always be called after a ->lookup or ->create operation. If there are permissions or other problems, then it's quite likely that it *won't* be called.
Fix this by fully instantiating the filp whenever the file is created and pass that filp back to the VFS. If there is a problem, the VFS can clean up the references.
--- fs/cifs/dir.c | 35 +++++++++++++++++++++++++++-------- fs/cifs/file.c | 44 ++------------------------------------------ 2 files changed, 29 insertions(+), 50 deletions(-)
fs/cifs/dir.c +27 -8
@@ -25,6 +25,7 @@#include <linux/slab.h> #include <linux/namei.h> #include <linux/mount.h> +#include <linux/file.h> #include "cifsfs.h" #include "cifspdu.h" #include "cifsglob.h"
@@ -184,6 +185,8 @@ cifs_new_fileinfo(struct inode *newinode, __u16 fileHandle, }write_unlock (&GlobalSMBSeslock); + file->private_data = pCifsFile; +return pCifsFile; }
@@ -463,14 +466,22 @@ cifs_create_set_dentry:if (newinode && nd && (nd->flags &LOOKUP_OPEN )) {struct cifsFileInfo *pfile_info; -/* -* cifs_fill_filedata() takes care of setting cifsFileInfo -* pointer to file->private_data. -*/ - pfile_info =cifs_new_fileinfo (newinode, fileHandle,NULL , +struct file *filp; + + filp =lookup_instantiate_filp (nd, direntry, generic_file_open); +if (IS_ERR (filp)) { + rc =PTR_ERR (filp); +CIFSSMBClose (xid, tcon, fileHandle); + goto cifs_create_out; + } + + pfile_info =cifs_new_fileinfo (newinode, fileHandle, filp, nd->path.mnt, oflags); -if (pfile_info ==NULL ) +if (pfile_info ==NULL ) { +fput (filp); +CIFSSMBClose (xid, tcon, fileHandle); rc = -ENOMEM ; + } }else {CIFSSMBClose (xid, tcon, fileHandle); }
@@ -717,15 +728,23 @@ cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry, direntry->d_op = &cifs_dentry_ops;d_add (direntry, newInode);if (posix_open) { - cfile =cifs_new_fileinfo (newInode, fileHandle,NULL , + filp =lookup_instantiate_filp (nd, direntry, + generic_file_open); +if (IS_ERR (filp)) { + rc =PTR_ERR (filp); +CIFSSMBClose (xid, pTcon, fileHandle); + goto lookup_out; + } + + cfile =cifs_new_fileinfo (newInode, fileHandle, filp, nd->path.mnt, nd->intent.open.flags);if (cfile ==NULL ) { +fput (filp);CIFSSMBClose (xid, pTcon, fileHandle); rc = -ENOMEM ; goto lookup_out; } - filp =lookup_instantiate_filp (nd, direntry,NULL ); }/* since paths are not looked up by component - the parent directories are presumed to be good here */
fs/cifs/file.c +2 -42
74 lines not on this page · raw message
Signature · 1 line
1.6.6.1