shinys000114/linux
1
0
Fork 0
mirror of https://github.com/hardkernel/linux.git synced 2026-06-01 16:57:18 +09:00
Code Issues Packages Projects Releases Wiki Activity

lustre/llite: simplify dentry revalidate

Browse Source 
Lustre client dentry validation is protected by LDLM lock, so
any time a dentry is found, it's valid and no need to revalidate
from MDS, and even it does, there is race that it may be
invalidated after revalidation is finished.

Signed-off-by: Lai Siyao <lai.siyao@intel.com>
Reviewed-on: http://review.whamcloud.com/7475
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-3544
Reviewed-by: Peng Tao <bergwolf@gmail.com>
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Signed-off-by: Oleg Drokin <oleg.drokin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This commit is contained in:
Lai Siyao
2014-02-28 21:16:38 -05:00
committed by Greg Kroah-Hartman
parent 63d42578d5
commit f236f69b48
7 changed files with 45 additions and 313 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
drivers/staging/lustre/lustre/include/lustre/lustre_idl.h
View File

@@ -2112,7 +2112,7 @@ extern void lustre_swab_generic_32s (__u32 *val);
#define DISP_LOOKUP_POS 0x00000008
#define DISP_OPEN_CREATE 0x00000010
#define DISP_OPEN_OPEN 0x00000020
#define DISP_ENQ_COMPLETE 0x00400000
#define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
#define DISP_ENQ_OPEN_REF 0x00800000
#define DISP_ENQ_CREATE_REF 0x01000000
#define DISP_OPEN_LOCK 0x02000000

290
drivers/staging/lustre/lustre/llite/dcache.c
View File

@@ -241,9 +241,6 @@ void ll_intent_release(struct lookup_intent *it)
ptlrpc_req_finished(it->d.lustre.it_data); /* ll_file_open */
if (it_disposition(it, DISP_ENQ_CREATE_REF)) /* create rec */
ptlrpc_req_finished(it->d.lustre.it_data);
if (it_disposition(it, DISP_ENQ_COMPLETE)) /* saved req from revalidate
* to lookup */
ptlrpc_req_finished(it->d.lustre.it_data);
it->d.lustre.it_disposition = 0;
it->d.lustre.it_data = NULL;
@@ -328,262 +325,32 @@ void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft)
}
int ll_revalidate_it(struct dentry *de, int lookup_flags,
struct lookup_intent *it)
static int ll_revalidate_dentry(struct dentry *dentry,
unsigned int lookup_flags)
{
struct md_op_data *op_data;
struct ptlrpc_request *req = NULL;
struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
struct obd_export *exp;
struct inode *parent = de->d_parent->d_inode;
int rc;
struct inode *dir = dentry->d_parent->d_inode;
CDEBUG(D_VFSTRACE, "VFS Op:name=%s,intent=%s\n", de->d_name.name,
LL_IT2STR(it));
/*
* if open&create is set, talk to MDS to make sure file is created if
* necessary, because we can't do this in ->open() later since that's
* called on an inode. return 0 here to let lookup to handle this.
*/
if ((lookup_flags & (LOOKUP_OPEN | LOOKUP_CREATE)) ==
(LOOKUP_OPEN | LOOKUP_CREATE))
return 0;
LASSERT(de != de->d_sb->s_root);
if (de->d_inode == NULL) {
__u64 ibits;
/* We can only use negative dentries if this is stat or lookup,
for opens and stuff we do need to query server. */
/* If there is IT_CREAT in intent op set, then we must throw
away this negative dentry and actually do the request to
kernel to create whatever needs to be created (if possible)*/
if (it && (it->it_op & IT_CREAT))
return 0;
if (d_lustre_invalid(de))
return 0;
ibits = MDS_INODELOCK_UPDATE;
rc = ll_have_md_lock(parent, &ibits, LCK_MINMODE);
GOTO(out_sa, rc);
}
/* Never execute intents for mount points.
* Attributes will be fixed up in ll_inode_revalidate_it */
if (d_mountpoint(de))
GOTO(out_sa, rc = 1);
exp = ll_i2mdexp(de->d_inode);
OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_REVALIDATE_PAUSE, 5);
ll_frob_intent(&it, &lookup_it);
LASSERT(it);
if (it->it_op == IT_LOOKUP && !d_lustre_invalid(de))
if (lookup_flags & (LOOKUP_PARENT | LOOKUP_OPEN | LOOKUP_CREATE))
return 1;
if (it->it_op == IT_OPEN) {
struct inode *inode = de->d_inode;
struct ll_inode_info *lli = ll_i2info(inode);
struct obd_client_handle **och_p;
__u64 ibits;
if (d_need_statahead(dir, dentry) <= 0)
return 1;
/*
* We used to check for MDS_INODELOCK_OPEN here, but in fact
* just having LOOKUP lock is enough to justify inode is the
* same. And if inode is the same and we have suitable
* openhandle, then there is no point in doing another OPEN RPC
* just to throw away newly received openhandle. There are no
* security implications too, if file owner or access mode is
* change, LOOKUP lock is revoked.
*/
if (lookup_flags & LOOKUP_RCU)
return -ECHILD;
if (it->it_flags & FMODE_WRITE)
och_p = &lli->lli_mds_write_och;
else if (it->it_flags & FMODE_EXEC)
och_p = &lli->lli_mds_exec_och;
else
och_p = &lli->lli_mds_read_och;
/* Check for the proper lock. */
ibits = MDS_INODELOCK_LOOKUP;
if (!ll_have_md_lock(inode, &ibits, LCK_MINMODE))
goto do_lock;
mutex_lock(&lli->lli_och_mutex);
if (*och_p) { /* Everything is open already, do nothing */
/* Originally it was idea to do not let them steal our
* open handle from under us by (*och_usecount)++ here.
* But in case we have the handle, but we cannot use it
* due to later checks (e.g. O_CREAT|O_EXCL flags set),
* nobody would decrement counter increased here. So we
* just hope the lock won't be invalidated in between.
* But if it would be, we'll reopen the open request to
* MDS later during file open path.
*/
mutex_unlock(&lli->lli_och_mutex);
return 1;
}
mutex_unlock(&lli->lli_och_mutex);
}
if (it->it_op == IT_GETATTR) {
rc = ll_statahead_enter(parent, &de, 0);
if (rc == 1)
goto mark;
else if (rc != -EAGAIN && rc != 0)
GOTO(out, rc = 0);
}
do_lock:
op_data = ll_prep_md_op_data(NULL, parent, de->d_inode,
de->d_name.name, de->d_name.len,
0, LUSTRE_OPC_ANY, NULL);
if (IS_ERR(op_data))
return PTR_ERR(op_data);
if (!IS_POSIXACL(parent) || !exp_connect_umask(exp))
it->it_create_mode &= ~current_umask();
it->it_create_mode |= M_CHECK_STALE;
rc = md_intent_lock(exp, op_data, NULL, 0, it,
lookup_flags,
&req, ll_md_blocking_ast, 0);
it->it_create_mode &= ~M_CHECK_STALE;
ll_finish_md_op_data(op_data);
/* If req is NULL, then md_intent_lock only tried to do a lock match;
* if all was well, it will return 1 if it found locks, 0 otherwise. */
if (req == NULL && rc >= 0) {
if (!rc)
goto do_lookup;
GOTO(out, rc);
}
if (rc < 0) {
if (rc != -ESTALE) {
CDEBUG(D_INFO, "ll_intent_lock: rc %d : it->it_status "
"%d\n", rc, it->d.lustre.it_status);
}
GOTO(out, rc = 0);
}
revalidate_finish:
rc = ll_revalidate_it_finish(req, it, de);
if (rc != 0) {
if (rc != -ESTALE && rc != -ENOENT)
ll_intent_release(it);
GOTO(out, rc = 0);
}
if ((it->it_op & IT_OPEN) && de->d_inode &&
!S_ISREG(de->d_inode->i_mode) &&
!S_ISDIR(de->d_inode->i_mode)) {
ll_release_openhandle(de, it);
}
rc = 1;
out:
/* We do not free request as it may be reused during following lookup
* (see comment in mdc/mdc_locks.c::mdc_intent_lock()), request will
* be freed in ll_lookup_it or in ll_intent_release. But if
* request was not completed, we need to free it. (bug 5154, 9903) */
if (req != NULL && !it_disposition(it, DISP_ENQ_COMPLETE))
ptlrpc_req_finished(req);
if (rc == 0) {
/* mdt may grant layout lock for the newly created file, so
* release the lock to avoid leaking */
ll_intent_drop_lock(it);
ll_invalidate_aliases(de->d_inode);
} else {
__u64 bits = 0;
__u64 matched_bits = 0;
CDEBUG(D_DENTRY, "revalidated dentry %.*s (%p) parent %p "
"inode %p refc %d\n", de->d_name.len,
de->d_name.name, de, de->d_parent, de->d_inode,
d_count(de));
ll_set_lock_data(exp, de->d_inode, it, &bits);
/* Note: We have to match both LOOKUP and PERM lock
* here to make sure the dentry is valid and no one
* changing the permission.
* But if the client connects < 2.4 server, which will
* only grant LOOKUP lock, so we can only Match LOOKUP
* lock for old server */
if (exp_connect_flags(ll_i2mdexp(de->d_inode)) &&
OBD_CONNECT_LVB_TYPE)
matched_bits =
MDS_INODELOCK_LOOKUP | MDS_INODELOCK_PERM;
else
matched_bits = MDS_INODELOCK_LOOKUP;
if (((bits & matched_bits) == matched_bits) &&
d_lustre_invalid(de))
d_lustre_revalidate(de);
ll_lookup_finish_locks(it, de);
}
mark:
if (it != NULL && it->it_op == IT_GETATTR && rc > 0)
ll_statahead_mark(parent, de);
return rc;
/*
* This part is here to combat evil-evil race in real_lookup on 2.6
* kernels. The race details are: We enter do_lookup() looking for some
* name, there is nothing in dcache for this name yet and d_lookup()
* returns NULL. We proceed to real_lookup(), and while we do this,
* another process does open on the same file we looking up (most simple
* reproducer), open succeeds and the dentry is added. Now back to
* us. In real_lookup() we do d_lookup() again and suddenly find the
* dentry, so we call d_revalidate on it, but there is no lock, so
* without this code we would return 0, but unpatched real_lookup just
* returns -ENOENT in such a case instead of retrying the lookup. Once
* this is dealt with in real_lookup(), all of this ugly mess can go and
* we can just check locks in ->d_revalidate without doing any RPCs
* ever.
*/
do_lookup:
if (it != &lookup_it) {
/* MDS_INODELOCK_UPDATE needed for IT_GETATTR case. */
if (it->it_op == IT_GETATTR)
lookup_it.it_op = IT_GETATTR;
ll_lookup_finish_locks(it, de);
it = &lookup_it;
}
/* Do real lookup here. */
op_data = ll_prep_md_op_data(NULL, parent, NULL, de->d_name.name,
de->d_name.len, 0, (it->it_op & IT_CREAT ?
LUSTRE_OPC_CREATE :
LUSTRE_OPC_ANY), NULL);
if (IS_ERR(op_data))
return PTR_ERR(op_data);
rc = md_intent_lock(exp, op_data, NULL, 0, it, 0, &req,
ll_md_blocking_ast, 0);
if (rc >= 0) {
struct mdt_body *mdt_body;
struct lu_fid fid = {.f_seq = 0, .f_oid = 0, .f_ver = 0};
mdt_body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
if (de->d_inode)
fid = *ll_inode2fid(de->d_inode);
/* see if we got same inode, if not - return error */
if (lu_fid_eq(&fid, &mdt_body->fid1)) {
ll_finish_md_op_data(op_data);
op_data = NULL;
goto revalidate_finish;
}
ll_intent_release(it);
}
ll_finish_md_op_data(op_data);
GOTO(out, rc = 0);
out_sa:
/*
* For rc == 1 case, should not return directly to prevent losing
* statahead windows; for rc == 0 case, the "lookup" will be done later.
*/
if (it != NULL && it->it_op == IT_GETATTR && rc == 1)
ll_statahead_enter(parent, &de, 1);
goto mark;
do_statahead_enter(dir, &dentry, dentry->d_inode == NULL);
ll_statahead_mark(dir, dentry);
return 1;
}
/*
@@ -591,24 +358,13 @@ out_sa:
*/
int ll_revalidate_nd(struct dentry *dentry, unsigned int flags)
{
struct inode *parent = dentry->d_parent->d_inode;
int unplug = 0;
int rc;
CDEBUG(D_VFSTRACE, "VFS Op:name=%s,flags=%u\n",
CDEBUG(D_VFSTRACE, "VFS Op:name=%s, flags=%u\n",
dentry->d_name.name, flags);
if (!(flags & (LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE)) &&
ll_need_statahead(parent, dentry) > 0) {
if (flags & LOOKUP_RCU)
return -ECHILD;
if (dentry->d_inode == NULL)
unplug = 1;
do_statahead_enter(parent, &dentry, unplug);
ll_statahead_mark(parent, dentry);
}
return 1;
rc = ll_revalidate_dentry(dentry, flags);
return rc;
}

8
drivers/staging/lustre/lustre/llite/file.c
View File

@@ -446,8 +446,7 @@ static int ll_intent_file_open(struct file *file, void *lmm,
itp, NULL);
out:
ptlrpc_req_finished(itp->d.lustre.it_data);
it_clear_disposition(itp, DISP_ENQ_COMPLETE);
ptlrpc_req_finished(req);
ll_intent_drop_lock(itp);
return rc;
@@ -815,10 +814,7 @@ struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file,
* doesn't deal with openhandle, so normal openhandle will be leaked. */
LDLM_FL_NO_LRU | LDLM_FL_EXCL);
ll_finish_md_op_data(op_data);
if (req != NULL) {
ptlrpc_req_finished(req);
it_clear_disposition(&it, DISP_ENQ_COMPLETE);
}
ptlrpc_req_finished(req);
if (rc < 0)
GOTO(out_release_it, rc);

4
drivers/staging/lustre/lustre/llite/llite_internal.h
View File

@@ -1309,7 +1309,7 @@ ll_statahead_mark(struct inode *dir, struct dentry *dentry)
}
static inline int
ll_need_statahead(struct inode *dir, struct dentry *dentryp)
d_need_statahead(struct inode *dir, struct dentry *dentryp)
{
struct ll_inode_info *lli;
struct ll_dentry_data *ldd;
@@ -1354,7 +1354,7 @@ ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
{
int ret;
ret = ll_need_statahead(dir, *dentryp);
ret = d_need_statahead(dir, *dentryp);
if (ret <= 0)
return ret;

1
drivers/staging/lustre/lustre/lmv/lmv_intent.c
View File

@@ -119,7 +119,6 @@ static int lmv_intent_remote(struct obd_export *exp, void *lmm,
CDEBUG(D_INODE, "REMOTE_INTENT with fid="DFID" -> mds #%d\n",
PFID(&body->fid1), tgt->ltd_idx);
it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
rc = md_intent_lock(tgt->ltd_exp, op_data, lmm, lmmsize, it,
flags, &req, cb_blocking, extra_lock_flags);
if (rc)

1
drivers/staging/lustre/lustre/lmv/lmv_obd.c
View File

@@ -1744,7 +1744,6 @@ lmv_enqueue_remote(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
it->d.lustre.it_data = NULL;
fid1 = body->fid1;
it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
ptlrpc_req_finished(req);
tgt = lmv_find_target(lmv, &fid1);

52
drivers/staging/lustre/lustre/mdc/mdc_locks.c
View File

@@ -968,7 +968,6 @@ static int mdc_finish_intent_lock(struct obd_export *exp,
if (fid_is_sane(&op_data->op_fid2) &&
it->it_create_mode & M_CHECK_STALE &&
it->it_op != IT_GETATTR) {
it_set_disposition(it, DISP_ENQ_COMPLETE);
/* Also: did we find the same inode? */
/* sever can return one of two fids:
@@ -1139,6 +1138,12 @@ int mdc_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
struct ldlm_enqueue_info einfo = {
.ei_type = LDLM_IBITS,
.ei_mode = it_to_lock_mode(it),
.ei_cb_bl = cb_blocking,
.ei_cb_cp = ldlm_completion_ast,
};
struct lustre_handle lockh;
int rc = 0;
@@ -1164,42 +1169,19 @@ int mdc_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
return rc;
}
/* lookup_it may be called only after revalidate_it has run, because
* revalidate_it cannot return errors, only zero. Returning zero causes
* this call to lookup, which *can* return an error.
*
* We only want to execute the request associated with the intent one
* time, however, so don't send the request again. Instead, skip past
* this and use the request from revalidate. In this case, revalidate
* never dropped its reference, so the refcounts are all OK */
if (!it_disposition(it, DISP_ENQ_COMPLETE)) {
struct ldlm_enqueue_info einfo = {
.ei_type = LDLM_IBITS,
.ei_mode = it_to_lock_mode(it),
.ei_cb_bl = cb_blocking,
.ei_cb_cp = ldlm_completion_ast,
};
/* For case if upper layer did not alloc fid, do it now. */
if (!fid_is_sane(&op_data->op_fid2) && it->it_op & IT_CREAT) {
rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
if (rc < 0) {
CERROR("Can't alloc new fid, rc %d\n", rc);
return rc;
}
}
rc = mdc_enqueue(exp, &einfo, it, op_data, &lockh,
lmm, lmmsize, NULL, extra_lock_flags);
if (rc < 0)
/* For case if upper layer did not alloc fid, do it now. */
if (!fid_is_sane(&op_data->op_fid2) && it->it_op & IT_CREAT) {
rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
if (rc < 0) {
CERROR("Can't alloc new fid, rc %d\n", rc);
return rc;
} else if (!fid_is_sane(&op_data->op_fid2) ||
!(it->it_create_mode & M_CHECK_STALE)) {
/* DISP_ENQ_COMPLETE set means there is extra reference on
* request referenced from this intent, saved for subsequent
* lookup. This path is executed when we proceed to this
* lookup, so we clear DISP_ENQ_COMPLETE */
it_clear_disposition(it, DISP_ENQ_COMPLETE);
}
}
rc = mdc_enqueue(exp, &einfo, it, op_data, &lockh, lmm, lmmsize, NULL,
extra_lock_flags);
if (rc < 0)
return rc;
*reqp = it->d.lustre.it_data;
rc = mdc_finish_intent_lock(exp, *reqp, op_data, it, &lockh);
return rc;