linux-stable/fs/cifs/dir.c

877 lines
22 KiB
C
Raw Permalink Normal View History

// SPDX-License-Identifier: LGPL-2.1
/*
*
* vfs operations that deal with dentries
*
* Copyright (C) International Business Machines Corp., 2002,2009
* Author(s): Steve French (sfrench@us.ibm.com)
*
*/
#include <linux/fs.h>
#include <linux/stat.h>
#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"
#include "cifsproto.h"
#include "cifs_debug.h"
#include "cifs_fs_sb.h"
#include "cifs_unicode.h"
#include "fs_context.h"
#include "cifs_ioctl.h"
cifs: Support fscache indexing rewrite Change the cifs filesystem to take account of the changes to fscache's indexing rewrite and reenable caching in cifs. The following changes have been made: (1) The fscache_netfs struct is no more, and there's no need to register the filesystem as a whole. (2) The session cookie is now an fscache_volume cookie, allocated with fscache_acquire_volume(). That takes three parameters: a string representing the "volume" in the index, a string naming the cache to use (or NULL) and a u64 that conveys coherency metadata for the volume. For cifs, I've made it render the volume name string as: "cifs,<ipaddress>,<sharename>" where the sharename has '/' characters replaced with ';'. This probably needs rethinking a bit as the total name could exceed the maximum filename component length. Further, the coherency data is currently just set to 0. It needs something else doing with it - I wonder if it would suffice simply to sum the resource_id, vol_create_time and vol_serial_number or maybe hash them. (3) The fscache_cookie_def is no more and needed information is passed directly to fscache_acquire_cookie(). The cache no longer calls back into the filesystem, but rather metadata changes are indicated at other times. fscache_acquire_cookie() is passed the same keying and coherency information as before. (4) The functions to set/reset cookies are removed and fscache_use_cookie() and fscache_unuse_cookie() are used instead. fscache_use_cookie() is passed a flag to indicate if the cookie is opened for writing. fscache_unuse_cookie() is passed updates for the metadata if we changed it (ie. if the file was opened for writing). These are called when the file is opened or closed. (5) cifs_setattr_*() are made to call fscache_resize() to change the size of the cache object. (6) The functions to read and write data are stubbed out pending a conversion to use netfslib. Changes ======= ver #8: - Abstract cache invalidation into a helper function. - Fix some checkpatch warnings[3]. ver #7: - Removed the accidentally added-back call to get the super cookie in cifs_root_iget(). - Fixed the right call to cifs_fscache_get_super_cookie() to take account of the "-o fsc" mount flag. ver #6: - Moved the change of gfpflags_allow_blocking() to current_is_kswapd() for cifs here. - Fixed one of the error paths in cifs_atomic_open() to jump around the call to use the cookie. - Fixed an additional successful return in the middle of cifs_open() to use the cookie on the way out. - Only get a volume cookie (and thus inode cookies) when "-o fsc" is supplied to mount. ver #5: - Fixed a couple of bits of cookie handling[2]: - The cookie should be released in cifs_evict_inode(), not cifsFileInfo_put_final(). The cookie needs to persist beyond file closure so that writepages will be able to write to it. - fscache_use_cookie() needs to be called in cifs_atomic_open() as it is for cifs_open(). ver #4: - Fixed the use of sizeof with memset. - tcon->vol_create_time is __le64 so doesn't need cpu_to_le64(). ver #3: - Canonicalise the cifs coherency data to make the cache portable. - Set volume coherency data. ver #2: - Use gfpflags_allow_blocking() rather than using flag directly. - Upgraded to -rc4 to allow for upstream changes[1]. - fscache_acquire_volume() now returns errors. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Jeff Layton <jlayton@kernel.org> cc: Steve French <smfrench@gmail.com> cc: Shyam Prasad N <nspmangalore@gmail.com> cc: linux-cifs@vger.kernel.org cc: linux-cachefs@redhat.com Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=23b55d673d7527b093cd97b7c217c82e70cd1af0 [1] Link: https://lore.kernel.org/r/3419813.1641592362@warthog.procyon.org.uk/ [2] Link: https://lore.kernel.org/r/CAH2r5muTanw9pJqzAHd01d9A8keeChkzGsCEH6=0rHutVLAF-A@mail.gmail.com/ [3] Link: https://lore.kernel.org/r/163819671009.215744.11230627184193298714.stgit@warthog.procyon.org.uk/ # v1 Link: https://lore.kernel.org/r/163906982979.143852.10672081929614953210.stgit@warthog.procyon.org.uk/ # v2 Link: https://lore.kernel.org/r/163967187187.1823006.247415138444991444.stgit@warthog.procyon.org.uk/ # v3 Link: https://lore.kernel.org/r/164021579335.640689.2681324337038770579.stgit@warthog.procyon.org.uk/ # v4 Link: https://lore.kernel.org/r/3462849.1641593783@warthog.procyon.org.uk/ # v5 Link: https://lore.kernel.org/r/1318953.1642024578@warthog.procyon.org.uk/ # v6 Signed-off-by: Steve French <stfrench@microsoft.com>
2020-11-17 15:56:59 +00:00
#include "fscache.h"
static void
renew_parental_timestamps(struct dentry *direntry)
{
/* BB check if there is a way to get the kernel to do this or if we
really need this */
do {
cifs_set_time(direntry, jiffies);
direntry = direntry->d_parent;
} while (!IS_ROOT(direntry));
}
char *
cifs_build_path_to_root(struct smb3_fs_context *ctx, struct cifs_sb_info *cifs_sb,
struct cifs_tcon *tcon, int add_treename)
{
int pplen = ctx->prepath ? strlen(ctx->prepath) + 1 : 0;
int dfsplen;
char *full_path = NULL;
/* if no prefix path, simply set path to the root of share to "" */
if (pplen == 0) {
full_path = kzalloc(1, GFP_KERNEL);
return full_path;
}
if (add_treename)
dfsplen = strnlen(tcon->tree_name, MAX_TREE_SIZE + 1);
else
dfsplen = 0;
full_path = kmalloc(dfsplen + pplen + 1, GFP_KERNEL);
if (full_path == NULL)
return full_path;
if (dfsplen)
memcpy(full_path, tcon->tree_name, dfsplen);
full_path[dfsplen] = CIFS_DIR_SEP(cifs_sb);
memcpy(full_path + dfsplen + 1, ctx->prepath, pplen);
convert_delimiter(full_path, CIFS_DIR_SEP(cifs_sb));
return full_path;
}
/* Note: caller must free return buffer */
const char *
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
build_path_from_dentry(struct dentry *direntry, void *page)
{
struct cifs_sb_info *cifs_sb = CIFS_SB(direntry->d_sb);
struct cifs_tcon *tcon = cifs_sb_master_tcon(cifs_sb);
bool prefix = tcon->Flags & SMB_SHARE_IS_IN_DFS;
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
return build_path_from_dentry_optional_prefix(direntry, page,
prefix);
}
char *__build_path_from_dentry_optional_prefix(struct dentry *direntry, void *page,
const char *tree, int tree_len,
bool prefix)
{
int dfsplen;
int pplen = 0;
struct cifs_sb_info *cifs_sb = CIFS_SB(direntry->d_sb);
char dirsep = CIFS_DIR_SEP(cifs_sb);
char *s;
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
if (unlikely(!page))
return ERR_PTR(-ENOMEM);
if (prefix)
dfsplen = strnlen(tree, tree_len + 1);
else
dfsplen = 0;
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_USE_PREFIX_PATH)
pplen = cifs_sb->prepath ? strlen(cifs_sb->prepath) + 1 : 0;
s = dentry_path_raw(direntry, page, PATH_MAX);
if (IS_ERR(s))
return s;
if (!s[1]) // for root we want "", not "/"
s++;
if (s < (char *)page + pplen + dfsplen)
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
return ERR_PTR(-ENAMETOOLONG);
if (pplen) {
cifs_dbg(FYI, "using cifs_sb prepath <%s>\n", cifs_sb->prepath);
s -= pplen;
memcpy(s + 1, cifs_sb->prepath, pplen - 1);
*s = '/';
}
if (dirsep != '/') {
/* BB test paths to Windows with '/' in the midst of prepath */
char *p;
for (p = s; *p; p++)
if (*p == '/')
*p = dirsep;
}
if (dfsplen) {
s -= dfsplen;
memcpy(s, tree, dfsplen);
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_POSIX_PATHS) {
int i;
for (i = 0; i < dfsplen; i++) {
if (s[i] == '\\')
s[i] = '/';
}
}
}
return s;
}
char *build_path_from_dentry_optional_prefix(struct dentry *direntry, void *page,
bool prefix)
{
struct cifs_sb_info *cifs_sb = CIFS_SB(direntry->d_sb);
struct cifs_tcon *tcon = cifs_sb_master_tcon(cifs_sb);
return __build_path_from_dentry_optional_prefix(direntry, page, tcon->tree_name,
MAX_TREE_SIZE, prefix);
}
/*
* Don't allow path components longer than the server max.
* Don't allow the separator character in a path component.
* The VFS will not allow "/", but "\" is allowed by posix.
*/
static int
check_name(struct dentry *direntry, struct cifs_tcon *tcon)
{
struct cifs_sb_info *cifs_sb = CIFS_SB(direntry->d_sb);
int i;
if (unlikely(tcon->fsAttrInfo.MaxPathNameComponentLength &&
direntry->d_name.len >
le32_to_cpu(tcon->fsAttrInfo.MaxPathNameComponentLength)))
return -ENAMETOOLONG;
if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_POSIX_PATHS)) {
for (i = 0; i < direntry->d_name.len; i++) {
if (direntry->d_name.name[i] == '\\') {
cifs_dbg(FYI, "Invalid file name\n");
return -EINVAL;
}
}
}
return 0;
}
/* Inode operations in similar order to how they appear in Linux file fs.h */
static int cifs_do_create(struct inode *inode, struct dentry *direntry, unsigned int xid,
struct tcon_link *tlink, unsigned int oflags, umode_t mode, __u32 *oplock,
struct cifs_fid *fid, struct cifs_open_info_data *buf)
{
int rc = -ENOENT;
int create_options = CREATE_NOT_DIR;
int desired_access;
struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
struct cifs_tcon *tcon = tlink_tcon(tlink);
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
const char *full_path;
void *page = alloc_dentry_path();
struct inode *newinode = NULL;
int disposition;
struct TCP_Server_Info *server = tcon->ses->server;
struct cifs_open_parms oparms;
*oplock = 0;
if (tcon->ses->server->oplocks)
*oplock = REQ_OPLOCK;
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
full_path = build_path_from_dentry(direntry, page);
if (IS_ERR(full_path)) {
free_dentry_path(page);
return PTR_ERR(full_path);
}
#ifdef CONFIG_CIFS_ALLOW_INSECURE_LEGACY
if (tcon->unix_ext && cap_unix(tcon->ses) && !tcon->broken_posix_open &&
(CIFS_UNIX_POSIX_PATH_OPS_CAP &
le64_to_cpu(tcon->fsUnixInfo.Capability))) {
rc = cifs_posix_open(full_path, &newinode, inode->i_sb, mode,
oflags, oplock, &fid->netfid, xid);
switch (rc) {
case 0:
if (newinode == NULL) {
/* query inode info */
goto cifs_create_get_file_info;
}
if (S_ISDIR(newinode->i_mode)) {
CIFSSMBClose(xid, tcon, fid->netfid);
iput(newinode);
rc = -EISDIR;
goto out;
}
if (!S_ISREG(newinode->i_mode)) {
/*
* The server may allow us to open things like
* FIFOs, but the client isn't set up to deal
* with that. If it's not a regular file, just
* close it and proceed as if it were a normal
* lookup.
*/
CIFSSMBClose(xid, tcon, fid->netfid);
goto cifs_create_get_file_info;
}
/* success, no need to query */
goto cifs_create_set_dentry;
case -ENOENT:
goto cifs_create_get_file_info;
case -EIO:
case -EINVAL:
/*
* EIO could indicate that (posix open) operation is not
* supported, despite what server claimed in capability
* negotiation.
*
* POSIX open in samba versions 3.3.1 and earlier could
* incorrectly fail with invalid parameter.
*/
tcon->broken_posix_open = true;
break;
case -EREMOTE:
case -EOPNOTSUPP:
/*
* EREMOTE indicates DFS junction, which is not handled
* in posix open. If either that or op not supported
* returned, follow the normal lookup.
*/
break;
default:
goto out;
}
/*
* fallthrough to retry, using older open call, this is case
* where server does not support this SMB level, and falsely
* claims capability (also get here for DFS case which should be
* rare for path not covered on files)
*/
}
#endif /* CONFIG_CIFS_ALLOW_INSECURE_LEGACY */
desired_access = 0;
if (OPEN_FMODE(oflags) & FMODE_READ)
desired_access |= GENERIC_READ; /* is this too little? */
if (OPEN_FMODE(oflags) & FMODE_WRITE)
desired_access |= GENERIC_WRITE;
disposition = FILE_OVERWRITE_IF;
if ((oflags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
disposition = FILE_CREATE;
else if ((oflags & (O_CREAT | O_TRUNC)) == (O_CREAT | O_TRUNC))
disposition = FILE_OVERWRITE_IF;
else if ((oflags & O_CREAT) == O_CREAT)
disposition = FILE_OPEN_IF;
else
cifs_dbg(FYI, "Create flag not set in create function\n");
/*
* BB add processing to set equivalent of mode - e.g. via CreateX with
* ACLs
*/
if (!server->ops->open) {
rc = -ENOSYS;
goto out;
}
/*
* if we're not using unix extensions, see if we need to set
* ATTR_READONLY on the create call
*/
if (!tcon->unix_ext && (mode & S_IWUGO) == 0)
create_options |= CREATE_OPTION_READONLY;
oparms = (struct cifs_open_parms) {
.tcon = tcon,
.cifs_sb = cifs_sb,
.desired_access = desired_access,
.create_options = cifs_create_options(cifs_sb, create_options),
.disposition = disposition,
.path = full_path,
.fid = fid,
.mode = mode,
};
rc = server->ops->open(xid, &oparms, oplock, buf);
if (rc) {
cifs_dbg(FYI, "cifs_create returned 0x%x\n", rc);
goto out;
}
#ifdef CONFIG_CIFS_ALLOW_INSECURE_LEGACY
/*
* If Open reported that we actually created a file then we now have to
* set the mode if possible.
*/
if ((tcon->unix_ext) && (*oplock & CIFS_CREATE_ACTION)) {
struct cifs_unix_set_info_args args = {
.mode = mode,
.ctime = NO_CHANGE_64,
.atime = NO_CHANGE_64,
.mtime = NO_CHANGE_64,
.device = 0,
};
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) {
args.uid = current_fsuid();
if (inode->i_mode & S_ISGID)
args.gid = inode->i_gid;
else
args.gid = current_fsgid();
} else {
args.uid = INVALID_UID; /* no change */
args.gid = INVALID_GID; /* no change */
}
CIFSSMBUnixSetFileInfo(xid, tcon, &args, fid->netfid,
current->tgid);
} else {
/*
* BB implement mode setting via Windows security
* descriptors e.g.
*/
/* CIFSSMBWinSetPerms(xid,tcon,path,mode,-1,-1,nls);*/
/* Could set r/o dos attribute if mode & 0222 == 0 */
}
cifs_create_get_file_info:
/* server might mask mode so we have to query for it */
if (tcon->unix_ext)
rc = cifs_get_inode_info_unix(&newinode, full_path, inode->i_sb,
xid);
else {
#else
{
#endif /* CONFIG_CIFS_ALLOW_INSECURE_LEGACY */
/* TODO: Add support for calling POSIX query info here, but passing in fid */
rc = cifs_get_inode_info(&newinode, full_path, buf, inode->i_sb, xid, fid);
if (newinode) {
if (server->ops->set_lease_key)
server->ops->set_lease_key(newinode, fid);
if ((*oplock & CIFS_CREATE_ACTION) && S_ISREG(newinode->i_mode)) {
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DYNPERM)
newinode->i_mode = mode;
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) {
newinode->i_uid = current_fsuid();
if (inode->i_mode & S_ISGID)
newinode->i_gid = inode->i_gid;
else
newinode->i_gid = current_fsgid();
}
}
}
}
#ifdef CONFIG_CIFS_ALLOW_INSECURE_LEGACY
cifs_create_set_dentry:
#endif /* CONFIG_CIFS_ALLOW_INSECURE_LEGACY */
if (rc != 0) {
cifs_dbg(FYI, "Create worked, get_inode_info failed rc = %d\n",
rc);
goto out_err;
}
if (newinode)
if (S_ISDIR(newinode->i_mode)) {
rc = -EISDIR;
goto out_err;
}
d_drop(direntry);
d_add(direntry, newinode);
out:
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
free_dentry_path(page);
return rc;
out_err:
if (server->ops->close)
server->ops->close(xid, tcon, fid);
if (newinode)
iput(newinode);
goto out;
}
int
cifs_atomic_open(struct inode *inode, struct dentry *direntry,
struct file *file, unsigned oflags, umode_t mode)
{
int rc;
unsigned int xid;
struct tcon_link *tlink;
struct cifs_tcon *tcon;
struct TCP_Server_Info *server;
struct cifs_fid fid = {};
struct cifs_pending_open open;
__u32 oplock;
struct cifsFileInfo *file_info;
struct cifs_open_info_data buf = {};
if (unlikely(cifs_forced_shutdown(CIFS_SB(inode->i_sb))))
return -EIO;
/*
* Posix open is only called (at lookup time) for file create now. For
* opens (rather than creates), because we do not know if it is a file
* or directory yet, and current Samba no longer allows us to do posix
* open on dirs, we could end up wasting an open call on what turns out
* to be a dir. For file opens, we wait to call posix open till
* cifs_open. It could be added to atomic_open in the future but the
* performance tradeoff of the extra network request when EISDIR or
* EACCES is returned would have to be weighed against the 50% reduction
* in network traffic in the other paths.
*/
if (!(oflags & O_CREAT)) {
struct dentry *res;
/*
* Check for hashed negative dentry. We have already revalidated
* the dentry and it is fine. No need to perform another lookup.
*/
if (!d_in_lookup(direntry))
return -ENOENT;
res = cifs_lookup(inode, direntry, 0);
if (IS_ERR(res))
return PTR_ERR(res);
return finish_no_open(file, res);
}
xid = get_xid();
cifs_dbg(FYI, "parent inode = 0x%p name is: %pd and dentry = 0x%p\n",
inode, direntry, direntry);
tlink = cifs_sb_tlink(CIFS_SB(inode->i_sb));
if (IS_ERR(tlink)) {
rc = PTR_ERR(tlink);
goto out_free_xid;
}
tcon = tlink_tcon(tlink);
rc = check_name(direntry, tcon);
if (rc)
goto out;
server = tcon->ses->server;
if (server->ops->new_lease_key)
server->ops->new_lease_key(&fid);
cifs_add_pending_open(&fid, tlink, &open);
rc = cifs_do_create(inode, direntry, xid, tlink, oflags, mode,
&oplock, &fid, &buf);
if (rc) {
cifs_del_pending_open(&open);
goto out;
}
if ((oflags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
file->f_mode |= FMODE_CREATED;
rc = finish_open(file, direntry, generic_file_open);
if (rc) {
if (server->ops->close)
server->ops->close(xid, tcon, &fid);
cifs_del_pending_open(&open);
goto out;
}
if (file->f_flags & O_DIRECT &&
CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO) {
if (CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_NO_BRL)
file->f_op = &cifs_file_direct_nobrl_ops;
else
file->f_op = &cifs_file_direct_ops;
}
file_info = cifs_new_fileinfo(&fid, file, tlink, oplock, buf.symlink_target);
if (file_info == NULL) {
if (server->ops->close)
server->ops->close(xid, tcon, &fid);
cifs_del_pending_open(&open);
rc = -ENOMEM;
cifs: Support fscache indexing rewrite Change the cifs filesystem to take account of the changes to fscache's indexing rewrite and reenable caching in cifs. The following changes have been made: (1) The fscache_netfs struct is no more, and there's no need to register the filesystem as a whole. (2) The session cookie is now an fscache_volume cookie, allocated with fscache_acquire_volume(). That takes three parameters: a string representing the "volume" in the index, a string naming the cache to use (or NULL) and a u64 that conveys coherency metadata for the volume. For cifs, I've made it render the volume name string as: "cifs,<ipaddress>,<sharename>" where the sharename has '/' characters replaced with ';'. This probably needs rethinking a bit as the total name could exceed the maximum filename component length. Further, the coherency data is currently just set to 0. It needs something else doing with it - I wonder if it would suffice simply to sum the resource_id, vol_create_time and vol_serial_number or maybe hash them. (3) The fscache_cookie_def is no more and needed information is passed directly to fscache_acquire_cookie(). The cache no longer calls back into the filesystem, but rather metadata changes are indicated at other times. fscache_acquire_cookie() is passed the same keying and coherency information as before. (4) The functions to set/reset cookies are removed and fscache_use_cookie() and fscache_unuse_cookie() are used instead. fscache_use_cookie() is passed a flag to indicate if the cookie is opened for writing. fscache_unuse_cookie() is passed updates for the metadata if we changed it (ie. if the file was opened for writing). These are called when the file is opened or closed. (5) cifs_setattr_*() are made to call fscache_resize() to change the size of the cache object. (6) The functions to read and write data are stubbed out pending a conversion to use netfslib. Changes ======= ver #8: - Abstract cache invalidation into a helper function. - Fix some checkpatch warnings[3]. ver #7: - Removed the accidentally added-back call to get the super cookie in cifs_root_iget(). - Fixed the right call to cifs_fscache_get_super_cookie() to take account of the "-o fsc" mount flag. ver #6: - Moved the change of gfpflags_allow_blocking() to current_is_kswapd() for cifs here. - Fixed one of the error paths in cifs_atomic_open() to jump around the call to use the cookie. - Fixed an additional successful return in the middle of cifs_open() to use the cookie on the way out. - Only get a volume cookie (and thus inode cookies) when "-o fsc" is supplied to mount. ver #5: - Fixed a couple of bits of cookie handling[2]: - The cookie should be released in cifs_evict_inode(), not cifsFileInfo_put_final(). The cookie needs to persist beyond file closure so that writepages will be able to write to it. - fscache_use_cookie() needs to be called in cifs_atomic_open() as it is for cifs_open(). ver #4: - Fixed the use of sizeof with memset. - tcon->vol_create_time is __le64 so doesn't need cpu_to_le64(). ver #3: - Canonicalise the cifs coherency data to make the cache portable. - Set volume coherency data. ver #2: - Use gfpflags_allow_blocking() rather than using flag directly. - Upgraded to -rc4 to allow for upstream changes[1]. - fscache_acquire_volume() now returns errors. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Jeff Layton <jlayton@kernel.org> cc: Steve French <smfrench@gmail.com> cc: Shyam Prasad N <nspmangalore@gmail.com> cc: linux-cifs@vger.kernel.org cc: linux-cachefs@redhat.com Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=23b55d673d7527b093cd97b7c217c82e70cd1af0 [1] Link: https://lore.kernel.org/r/3419813.1641592362@warthog.procyon.org.uk/ [2] Link: https://lore.kernel.org/r/CAH2r5muTanw9pJqzAHd01d9A8keeChkzGsCEH6=0rHutVLAF-A@mail.gmail.com/ [3] Link: https://lore.kernel.org/r/163819671009.215744.11230627184193298714.stgit@warthog.procyon.org.uk/ # v1 Link: https://lore.kernel.org/r/163906982979.143852.10672081929614953210.stgit@warthog.procyon.org.uk/ # v2 Link: https://lore.kernel.org/r/163967187187.1823006.247415138444991444.stgit@warthog.procyon.org.uk/ # v3 Link: https://lore.kernel.org/r/164021579335.640689.2681324337038770579.stgit@warthog.procyon.org.uk/ # v4 Link: https://lore.kernel.org/r/3462849.1641593783@warthog.procyon.org.uk/ # v5 Link: https://lore.kernel.org/r/1318953.1642024578@warthog.procyon.org.uk/ # v6 Signed-off-by: Steve French <stfrench@microsoft.com>
2020-11-17 15:56:59 +00:00
goto out;
}
cifs: Support fscache indexing rewrite Change the cifs filesystem to take account of the changes to fscache's indexing rewrite and reenable caching in cifs. The following changes have been made: (1) The fscache_netfs struct is no more, and there's no need to register the filesystem as a whole. (2) The session cookie is now an fscache_volume cookie, allocated with fscache_acquire_volume(). That takes three parameters: a string representing the "volume" in the index, a string naming the cache to use (or NULL) and a u64 that conveys coherency metadata for the volume. For cifs, I've made it render the volume name string as: "cifs,<ipaddress>,<sharename>" where the sharename has '/' characters replaced with ';'. This probably needs rethinking a bit as the total name could exceed the maximum filename component length. Further, the coherency data is currently just set to 0. It needs something else doing with it - I wonder if it would suffice simply to sum the resource_id, vol_create_time and vol_serial_number or maybe hash them. (3) The fscache_cookie_def is no more and needed information is passed directly to fscache_acquire_cookie(). The cache no longer calls back into the filesystem, but rather metadata changes are indicated at other times. fscache_acquire_cookie() is passed the same keying and coherency information as before. (4) The functions to set/reset cookies are removed and fscache_use_cookie() and fscache_unuse_cookie() are used instead. fscache_use_cookie() is passed a flag to indicate if the cookie is opened for writing. fscache_unuse_cookie() is passed updates for the metadata if we changed it (ie. if the file was opened for writing). These are called when the file is opened or closed. (5) cifs_setattr_*() are made to call fscache_resize() to change the size of the cache object. (6) The functions to read and write data are stubbed out pending a conversion to use netfslib. Changes ======= ver #8: - Abstract cache invalidation into a helper function. - Fix some checkpatch warnings[3]. ver #7: - Removed the accidentally added-back call to get the super cookie in cifs_root_iget(). - Fixed the right call to cifs_fscache_get_super_cookie() to take account of the "-o fsc" mount flag. ver #6: - Moved the change of gfpflags_allow_blocking() to current_is_kswapd() for cifs here. - Fixed one of the error paths in cifs_atomic_open() to jump around the call to use the cookie. - Fixed an additional successful return in the middle of cifs_open() to use the cookie on the way out. - Only get a volume cookie (and thus inode cookies) when "-o fsc" is supplied to mount. ver #5: - Fixed a couple of bits of cookie handling[2]: - The cookie should be released in cifs_evict_inode(), not cifsFileInfo_put_final(). The cookie needs to persist beyond file closure so that writepages will be able to write to it. - fscache_use_cookie() needs to be called in cifs_atomic_open() as it is for cifs_open(). ver #4: - Fixed the use of sizeof with memset. - tcon->vol_create_time is __le64 so doesn't need cpu_to_le64(). ver #3: - Canonicalise the cifs coherency data to make the cache portable. - Set volume coherency data. ver #2: - Use gfpflags_allow_blocking() rather than using flag directly. - Upgraded to -rc4 to allow for upstream changes[1]. - fscache_acquire_volume() now returns errors. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Jeff Layton <jlayton@kernel.org> cc: Steve French <smfrench@gmail.com> cc: Shyam Prasad N <nspmangalore@gmail.com> cc: linux-cifs@vger.kernel.org cc: linux-cachefs@redhat.com Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=23b55d673d7527b093cd97b7c217c82e70cd1af0 [1] Link: https://lore.kernel.org/r/3419813.1641592362@warthog.procyon.org.uk/ [2] Link: https://lore.kernel.org/r/CAH2r5muTanw9pJqzAHd01d9A8keeChkzGsCEH6=0rHutVLAF-A@mail.gmail.com/ [3] Link: https://lore.kernel.org/r/163819671009.215744.11230627184193298714.stgit@warthog.procyon.org.uk/ # v1 Link: https://lore.kernel.org/r/163906982979.143852.10672081929614953210.stgit@warthog.procyon.org.uk/ # v2 Link: https://lore.kernel.org/r/163967187187.1823006.247415138444991444.stgit@warthog.procyon.org.uk/ # v3 Link: https://lore.kernel.org/r/164021579335.640689.2681324337038770579.stgit@warthog.procyon.org.uk/ # v4 Link: https://lore.kernel.org/r/3462849.1641593783@warthog.procyon.org.uk/ # v5 Link: https://lore.kernel.org/r/1318953.1642024578@warthog.procyon.org.uk/ # v6 Signed-off-by: Steve French <stfrench@microsoft.com>
2020-11-17 15:56:59 +00:00
fscache_use_cookie(cifs_inode_cookie(file_inode(file)),
file->f_mode & FMODE_WRITE);
out:
cifs_put_tlink(tlink);
out_free_xid:
free_xid(xid);
cifs_free_open_info(&buf);
return rc;
}
int cifs_create(struct user_namespace *mnt_userns, struct inode *inode,
struct dentry *direntry, umode_t mode, bool excl)
{
int rc;
unsigned int xid = get_xid();
/*
* BB below access is probably too much for mknod to request
* but we have to do query and setpathinfo so requesting
* less could fail (unless we want to request getatr and setatr
* permissions (only). At least for POSIX we do not have to
* request so much.
*/
unsigned oflags = O_EXCL | O_CREAT | O_RDWR;
struct tcon_link *tlink;
struct cifs_tcon *tcon;
struct TCP_Server_Info *server;
struct cifs_fid fid;
__u32 oplock;
struct cifs_open_info_data buf = {};
cifs_dbg(FYI, "cifs_create parent inode = 0x%p name is: %pd and dentry = 0x%p\n",
inode, direntry, direntry);
if (unlikely(cifs_forced_shutdown(CIFS_SB(inode->i_sb)))) {
rc = -EIO;
goto out_free_xid;
}
tlink = cifs_sb_tlink(CIFS_SB(inode->i_sb));
rc = PTR_ERR(tlink);
if (IS_ERR(tlink))
goto out_free_xid;
tcon = tlink_tcon(tlink);
server = tcon->ses->server;
if (server->ops->new_lease_key)
server->ops->new_lease_key(&fid);
rc = cifs_do_create(inode, direntry, xid, tlink, oflags, mode, &oplock, &fid, &buf);
if (!rc && server->ops->close)
server->ops->close(xid, tcon, &fid);
cifs_free_open_info(&buf);
cifs_put_tlink(tlink);
out_free_xid:
free_xid(xid);
return rc;
}
int cifs_mknod(struct user_namespace *mnt_userns, struct inode *inode,
struct dentry *direntry, umode_t mode, dev_t device_number)
{
int rc = -EPERM;
unsigned int xid;
struct cifs_sb_info *cifs_sb;
struct tcon_link *tlink;
struct cifs_tcon *tcon;
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
const char *full_path;
void *page;
if (!old_valid_dev(device_number))
return -EINVAL;
cifs_sb = CIFS_SB(inode->i_sb);
if (unlikely(cifs_forced_shutdown(cifs_sb)))
return -EIO;
tlink = cifs_sb_tlink(cifs_sb);
if (IS_ERR(tlink))
return PTR_ERR(tlink);
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
page = alloc_dentry_path();
tcon = tlink_tcon(tlink);
xid = get_xid();
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
full_path = build_path_from_dentry(direntry, page);
if (IS_ERR(full_path)) {
rc = PTR_ERR(full_path);
goto mknod_out;
}
rc = tcon->ses->server->ops->make_node(xid, inode, direntry, tcon,
full_path, mode,
device_number);
mknod_out:
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
free_dentry_path(page);
free_xid(xid);
cifs_put_tlink(tlink);
return rc;
}
struct dentry *
cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
unsigned int flags)
{
unsigned int xid;
int rc = 0; /* to get around spurious gcc warning, set to zero here */
struct cifs_sb_info *cifs_sb;
struct tcon_link *tlink;
struct cifs_tcon *pTcon;
struct inode *newInode = NULL;
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
const char *full_path;
void *page;
cifs: retry lookup and readdir when EAGAIN is returned. According to the investigation performed by Jacob Shivers at Red Hat, cifs_lookup and cifs_readdir leak EAGAIN when the user session is deleted on the server. Fix this issue by implementing a retry with limits, as is implemented in cifs_revalidate_dentry_attr. Reproducer based on the work by Jacob Shivers: ~~~ $ cat readdir-cifs-test.sh #!/bin/bash # Install and configure powershell and sshd on the windows # server as descibed in # https://docs.microsoft.com/en-us/windows-server/administration/openssh/openssh_overview # This script uses expect(1) USER=dude SERVER=192.168.0.2 RPATH=root PASS='password' function debug_funcs { for line in $@ ; do echo "func $line +p" > /sys/kernel/debug/dynamic_debug/control done } function setup { echo 1 > /proc/fs/cifs/cifsFYI debug_funcs wait_for_compound_request \ smb2_query_dir_first cifs_readdir \ compound_send_recv cifs_reconnect_tcon \ generic_ip_connect cifs_reconnect \ smb2_reconnect_server smb2_reconnect \ cifs_readv_from_socket cifs_readv_receive tcpdump -i eth0 -w cifs.pcap host 192.168.2.182 & sleep 5 dmesg -C } function test_call { if [[ $1 == 1 ]] ; then tracer="strace -tt -f -s 4096 -o trace-$(date -Iseconds).txt" fi # Change the command here to anything appropriate $tracer ls $2 > /dev/null res=$? if [[ $1 == 1 ]] ; then if [[ $res == 0 ]] ; then 1>&2 echo success else 1>&2 echo "failure ($res)" fi fi } mountpoint /mnt > /dev/null || mount -t cifs -o username=$USER,pass=$PASS //$SERVER/$RPATH /mnt test_call 0 /mnt/ /usr/bin/expect << EOF set timeout 60 spawn ssh $USER@$SERVER expect "yes/no" { send "yes\r" expect "*?assword" { send "$PASS\r" } } "*?assword" { send "$PASS\r" } expect ">" { send "powershell close-smbsession -force\r" } expect ">" { send "exit\r" } expect eof EOF sysctl -w vm.drop_caches=2 > /dev/null sysctl -w vm.drop_caches=2 > /dev/null setup test_call 1 /mnt/ ~~~ Signed-off-by: Thiago Rafael Becker <trbecker@gmail.com> Acked-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-06-15 16:42:56 +00:00
int retry_count = 0;
xid = get_xid();
cifs_dbg(FYI, "parent inode = 0x%p name is: %pd and dentry = 0x%p\n",
parent_dir_inode, direntry, direntry);
/* check whether path exists */
cifs_sb = CIFS_SB(parent_dir_inode->i_sb);
tlink = cifs_sb_tlink(cifs_sb);
if (IS_ERR(tlink)) {
free_xid(xid);
return ERR_CAST(tlink);
}
pTcon = tlink_tcon(tlink);
rc = check_name(direntry, pTcon);
if (unlikely(rc)) {
cifs_put_tlink(tlink);
free_xid(xid);
return ERR_PTR(rc);
}
/* can not grab the rename sem here since it would
deadlock in the cases (beginning of sys_rename itself)
in which we already have the sb rename sem */
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
page = alloc_dentry_path();
full_path = build_path_from_dentry(direntry, page);
if (IS_ERR(full_path)) {
cifs_put_tlink(tlink);
free_xid(xid);
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
free_dentry_path(page);
return ERR_CAST(full_path);
}
if (d_really_is_positive(direntry)) {
cifs_dbg(FYI, "non-NULL inode in lookup\n");
} else {
cifs_dbg(FYI, "NULL inode in lookup\n");
}
cifs_dbg(FYI, "Full path: %s inode = 0x%p\n",
full_path, d_inode(direntry));
cifs: retry lookup and readdir when EAGAIN is returned. According to the investigation performed by Jacob Shivers at Red Hat, cifs_lookup and cifs_readdir leak EAGAIN when the user session is deleted on the server. Fix this issue by implementing a retry with limits, as is implemented in cifs_revalidate_dentry_attr. Reproducer based on the work by Jacob Shivers: ~~~ $ cat readdir-cifs-test.sh #!/bin/bash # Install and configure powershell and sshd on the windows # server as descibed in # https://docs.microsoft.com/en-us/windows-server/administration/openssh/openssh_overview # This script uses expect(1) USER=dude SERVER=192.168.0.2 RPATH=root PASS='password' function debug_funcs { for line in $@ ; do echo "func $line +p" > /sys/kernel/debug/dynamic_debug/control done } function setup { echo 1 > /proc/fs/cifs/cifsFYI debug_funcs wait_for_compound_request \ smb2_query_dir_first cifs_readdir \ compound_send_recv cifs_reconnect_tcon \ generic_ip_connect cifs_reconnect \ smb2_reconnect_server smb2_reconnect \ cifs_readv_from_socket cifs_readv_receive tcpdump -i eth0 -w cifs.pcap host 192.168.2.182 & sleep 5 dmesg -C } function test_call { if [[ $1 == 1 ]] ; then tracer="strace -tt -f -s 4096 -o trace-$(date -Iseconds).txt" fi # Change the command here to anything appropriate $tracer ls $2 > /dev/null res=$? if [[ $1 == 1 ]] ; then if [[ $res == 0 ]] ; then 1>&2 echo success else 1>&2 echo "failure ($res)" fi fi } mountpoint /mnt > /dev/null || mount -t cifs -o username=$USER,pass=$PASS //$SERVER/$RPATH /mnt test_call 0 /mnt/ /usr/bin/expect << EOF set timeout 60 spawn ssh $USER@$SERVER expect "yes/no" { send "yes\r" expect "*?assword" { send "$PASS\r" } } "*?assword" { send "$PASS\r" } expect ">" { send "powershell close-smbsession -force\r" } expect ">" { send "exit\r" } expect eof EOF sysctl -w vm.drop_caches=2 > /dev/null sysctl -w vm.drop_caches=2 > /dev/null setup test_call 1 /mnt/ ~~~ Signed-off-by: Thiago Rafael Becker <trbecker@gmail.com> Acked-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-06-15 16:42:56 +00:00
again:
if (pTcon->posix_extensions)
rc = smb311_posix_get_inode_info(&newInode, full_path, parent_dir_inode->i_sb, xid);
else if (pTcon->unix_ext) {
rc = cifs_get_inode_info_unix(&newInode, full_path,
parent_dir_inode->i_sb, xid);
} else {
rc = cifs_get_inode_info(&newInode, full_path, NULL,
parent_dir_inode->i_sb, xid, NULL);
}
if (rc == 0) {
/* since paths are not looked up by component - the parent
directories are presumed to be good here */
renew_parental_timestamps(direntry);
cifs: retry lookup and readdir when EAGAIN is returned. According to the investigation performed by Jacob Shivers at Red Hat, cifs_lookup and cifs_readdir leak EAGAIN when the user session is deleted on the server. Fix this issue by implementing a retry with limits, as is implemented in cifs_revalidate_dentry_attr. Reproducer based on the work by Jacob Shivers: ~~~ $ cat readdir-cifs-test.sh #!/bin/bash # Install and configure powershell and sshd on the windows # server as descibed in # https://docs.microsoft.com/en-us/windows-server/administration/openssh/openssh_overview # This script uses expect(1) USER=dude SERVER=192.168.0.2 RPATH=root PASS='password' function debug_funcs { for line in $@ ; do echo "func $line +p" > /sys/kernel/debug/dynamic_debug/control done } function setup { echo 1 > /proc/fs/cifs/cifsFYI debug_funcs wait_for_compound_request \ smb2_query_dir_first cifs_readdir \ compound_send_recv cifs_reconnect_tcon \ generic_ip_connect cifs_reconnect \ smb2_reconnect_server smb2_reconnect \ cifs_readv_from_socket cifs_readv_receive tcpdump -i eth0 -w cifs.pcap host 192.168.2.182 & sleep 5 dmesg -C } function test_call { if [[ $1 == 1 ]] ; then tracer="strace -tt -f -s 4096 -o trace-$(date -Iseconds).txt" fi # Change the command here to anything appropriate $tracer ls $2 > /dev/null res=$? if [[ $1 == 1 ]] ; then if [[ $res == 0 ]] ; then 1>&2 echo success else 1>&2 echo "failure ($res)" fi fi } mountpoint /mnt > /dev/null || mount -t cifs -o username=$USER,pass=$PASS //$SERVER/$RPATH /mnt test_call 0 /mnt/ /usr/bin/expect << EOF set timeout 60 spawn ssh $USER@$SERVER expect "yes/no" { send "yes\r" expect "*?assword" { send "$PASS\r" } } "*?assword" { send "$PASS\r" } expect ">" { send "powershell close-smbsession -force\r" } expect ">" { send "exit\r" } expect eof EOF sysctl -w vm.drop_caches=2 > /dev/null sysctl -w vm.drop_caches=2 > /dev/null setup test_call 1 /mnt/ ~~~ Signed-off-by: Thiago Rafael Becker <trbecker@gmail.com> Acked-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-06-15 16:42:56 +00:00
} else if (rc == -EAGAIN && retry_count++ < 10) {
goto again;
} else if (rc == -ENOENT) {
cifs_set_time(direntry, jiffies);
newInode = NULL;
} else {
if (rc != -EACCES) {
cifs_dbg(FYI, "Unexpected lookup error %d\n", rc);
/* We special case check for Access Denied - since that
is a common return code */
}
newInode = ERR_PTR(rc);
}
cifs: allocate buffer in the caller of build_path_from_dentry() build_path_from_dentry() open-codes dentry_path_raw(). The reason we can't use dentry_path_raw() in there (and postprocess the result as needed) is that the callers of build_path_from_dentry() expect that the object to be freed on cleanup and the string to be used are at the same address. That's painful, since the path is naturally built end-to-beginning - we start at the leaf and go through the ancestors, accumulating the pathname. Life would be easier if we left the buffer allocation to callers. It wouldn't be exact-sized buffer, but none of the callers keep the result for long - it's always freed before the caller returns. So there's no need to do exact-sized allocation; better use __getname()/__putname(), same as we do for pathname arguments of syscalls. What's more, there's no need to do allocation under spinlocks, so GFP_ATOMIC is not needed. Next patch will replace the open-coded dentry_path_raw() (in build_path_from_dentry_optional_prefix()) with calling the real thing. This patch only introduces wrappers for allocating/freeing the buffers and switches to new calling conventions: build_path_from_dentry(dentry, buf) expects buf to be address of a page-sized object or NULL, return value is a pathname built inside that buffer on success, ERR_PTR(-ENOMEM) if buf is NULL and ERR_PTR(-ENAMETOOLONG) if the pathname won't fit into page. Note that we don't need to check for failure when allocating the buffer in the caller - build_path_from_dentry() will do the right thing. Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Steve French <stfrench@microsoft.com>
2021-03-05 22:36:04 +00:00
free_dentry_path(page);
cifs_put_tlink(tlink);
free_xid(xid);
return d_splice_alias(newInode, direntry);
}
static int
cifs_d_revalidate(struct dentry *direntry, unsigned int flags)
{
struct inode *inode;
int rc;
if (flags & LOOKUP_RCU)
return -ECHILD;
if (d_really_is_positive(direntry)) {
inode = d_inode(direntry);
if ((flags & LOOKUP_REVAL) && !CIFS_CACHE_READ(CIFS_I(inode)))
CIFS_I(inode)->time = 0; /* force reval */
rc = cifs_revalidate_dentry(direntry);
if (rc) {
cifs_dbg(FYI, "cifs_revalidate_dentry failed with rc=%d", rc);
switch (rc) {
case -ENOENT:
case -ESTALE:
/*
* Those errors mean the dentry is invalid
* (file was deleted or recreated)
*/
return 0;
default:
/*
* Otherwise some unexpected error happened
* report it as-is to VFS layer
*/
return rc;
}
}
else {
/*
* If the inode wasn't known to be a dfs entry when
* the dentry was instantiated, such as when created
* via ->readdir(), it needs to be set now since the
* attributes will have been updated by
* cifs_revalidate_dentry().
*/
if (IS_AUTOMOUNT(inode) &&
!(direntry->d_flags & DCACHE_NEED_AUTOMOUNT)) {
spin_lock(&direntry->d_lock);
direntry->d_flags |= DCACHE_NEED_AUTOMOUNT;
spin_unlock(&direntry->d_lock);
}
return 1;
}
}
/*
* This may be nfsd (or something), anyway, we can't see the
* intent of this. So, since this can be for creation, drop it.
*/
if (!flags)
return 0;
/*
* Drop the negative dentry, in order to make sure to use the
* case sensitive name which is specified by user if this is
* for creation.
*/
if (flags & (LOOKUP_CREATE | LOOKUP_RENAME_TARGET))
return 0;
if (time_after(jiffies, cifs_get_time(direntry) + HZ) || !lookupCacheEnabled)
return 0;
return 1;
}
/* static int cifs_d_delete(struct dentry *direntry)
{
int rc = 0;
cifs_dbg(FYI, "In cifs d_delete, name = %pd\n", direntry);
return rc;
} */
const struct dentry_operations cifs_dentry_ops = {
.d_revalidate = cifs_d_revalidate,
.d_automount = cifs_dfs_d_automount,
/* d_delete: cifs_d_delete, */ /* not needed except for debugging */
};
static int cifs_ci_hash(const struct dentry *dentry, struct qstr *q)
{
struct nls_table *codepage = CIFS_SB(dentry->d_sb)->local_nls;
unsigned long hash;
wchar_t c;
int i, charlen;
hash = init_name_hash(dentry);
for (i = 0; i < q->len; i += charlen) {
charlen = codepage->char2uni(&q->name[i], q->len - i, &c);
/* error out if we can't convert the character */
if (unlikely(charlen < 0))
return charlen;
hash = partial_name_hash(cifs_toupper(c), hash);
}
q->hash = end_name_hash(hash);
return 0;
}
static int cifs_ci_compare(const struct dentry *dentry,
unsigned int len, const char *str, const struct qstr *name)
{
struct nls_table *codepage = CIFS_SB(dentry->d_sb)->local_nls;
wchar_t c1, c2;
int i, l1, l2;
/*
* We make the assumption here that uppercase characters in the local
* codepage are always the same length as their lowercase counterparts.
*
* If that's ever not the case, then this will fail to match it.
*/
if (name->len != len)
return 1;
for (i = 0; i < len; i += l1) {
/* Convert characters in both strings to UTF-16. */
l1 = codepage->char2uni(&str[i], len - i, &c1);
l2 = codepage->char2uni(&name->name[i], name->len - i, &c2);
/*
* If we can't convert either character, just declare it to
* be 1 byte long and compare the original byte.
*/
if (unlikely(l1 < 0 && l2 < 0)) {
if (str[i] != name->name[i])
return 1;
l1 = 1;
continue;
}
/*
* Here, we again ass|u|me that upper/lowercase versions of
* a character are the same length in the local NLS.
*/
if (l1 != l2)
return 1;
/* Now compare uppercase versions of these characters */
if (cifs_toupper(c1) != cifs_toupper(c2))
return 1;
}
return 0;
}
const struct dentry_operations cifs_ci_dentry_ops = {
.d_revalidate = cifs_d_revalidate,
.d_hash = cifs_ci_hash,
.d_compare = cifs_ci_compare,
.d_automount = cifs_dfs_d_automount,
};