iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[CIFS] Reduce chance of list corruption in find_writable_file

Browse Source 
When find_writable_file is racing with close and the session
to the server goes down, Shaggy noticed that there was a
chance that an open file in the list of files off the inode
could have been freed by close since cifs_reconnect can
block (the spinlock thus not held). This means that
we have to start over at the beginning of the list in some
cases.

There is a 2nd change that needs to be made later
(pointed out by Jeremy Allison and Shaggy) in order to
prevent cifs_close ever freeing the cifs per file info
when a write is pending.  Although we delay close from
freeing this memory for sufficiently long for all known
cases, ultimately on a very, very slow write
overlapping a close pending we need to allow close to return
(without freeing the cifs file info) and defer freeing the
memory to be the responsibility of the (sloooow) write
thread (presumably have to look at every place wrtPending
is decremented - and add a flag for deferred free for
after wrtPending goes to zero).

Acked-by: Shaggy <shaggy@us.ibm.com>
Acked-by: Shirish Pargaonkar <shirishp@us.ibm.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
This commit is contained in:
Steve French 2007-10-02 01:11:08 +00:00
parent 4084973dba
commit 9b22b0b726
1 changed files with 39 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
fs/cifs/file.c
View File

@ -1042,6 +1042,7 @@ struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode)
}
read_lock(&GlobalSMBSeslock);
refind_writable:
list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
if (open_file->closePend)
continue;
@ -1049,26 +1050,49 @@ struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode)
((open_file->pfile->f_flags & O_RDWR) ||
(open_file->pfile->f_flags & O_WRONLY))) {
atomic_inc(&open_file->wrtPending);
if (!open_file->invalidHandle) {
/* found a good writable file */
read_unlock(&GlobalSMBSeslock);
if (open_file->invalidHandle) {
return open_file;
}
read_unlock(&GlobalSMBSeslock);
/* Had to unlock since following call can block */
rc = cifs_reopen_file(open_file->pfile, FALSE);
/* if it fails, try another handle - might be */
/* dangerous to hold up writepages with retry */
if (rc) {
if (!rc) {
if (!open_file->closePend)
return open_file;
else { /* start over in case this was deleted */
/* since the list could be modified */
read_lock(&GlobalSMBSeslock);
atomic_dec(&open_file->wrtPending);
goto refind_writable;
}
}
/* if it fails, try another handle if possible -
(we can not do this if closePending since
loop could be modified - in which case we
have to start at the beginning of the list
again. Note that it would be bad
to hold up writepages here (rather than
in caller) with continuous retries */
cFYI(1, ("wp failed on reopen file"));
read_lock(&GlobalSMBSeslock);
/* can not use this handle, no write
pending on this one after all */
atomic_dec(&open_file->wrtPending);
continue;
}
}
if (open_file->closePend) {
read_lock(&GlobalSMBSeslock);
atomic_dec(&open_file->wrtPending);
continue;
}
return open_file;
if (open_file->closePend) /* list could have changed */
goto refind_writable;
/* else we simply continue to the next entry. Thus
we do not loop on reopen errors. If we
can not reopen the file, for example if we
reconnected to a server with another client
racing to delete or lock the file we would not
make progress if we restarted before the beginning
of the loop here. */
}
}
read_unlock(&GlobalSMBSeslock);