Merge tag 'gfs2-for-6.13' of git://git.kernel.org/pub/scm/linux/kerne…

…l/git/gfs2/linux-gfs2

Pull gfs2 updates from Andreas Gruenbacher:

 - Fix the code that cleans up left-over unlinked files.

   Various fixes and minor improvements in deleting files cached or held
   open remotely.

 - Simplify the use of dlm's DLM_LKF_QUECVT flag.

 - A few other minor cleanups.

* tag 'gfs2-for-6.13' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2: (21 commits)
  gfs2: Prevent inode creation race
  gfs2: Only defer deletes when we have an iopen glock
  gfs2: Simplify DLM_LKF_QUECVT use
  gfs2: gfs2_evict_inode clarification
  gfs2: Make gfs2_inode_refresh static
  gfs2: Use get_random_u32 in gfs2_orlov_skip
  gfs2: Randomize GLF_VERIFY_DELETE work delay
  gfs2: Use mod_delayed_work in gfs2_queue_try_to_evict
  gfs2: Update to the evict / remote delete documentation
  gfs2: Call gfs2_queue_verify_delete from gfs2_evict_inode
  gfs2: Clean up delete work processing
  gfs2: Minor delete_work_func cleanup
  gfs2: Return enum evict_behavior from gfs2_upgrade_iopen_glock
  gfs2: Rename dinode_demise to evict_behavior
  gfs2: Rename GIF_{DEFERRED -> DEFER}_DELETE
  gfs2: Faster gfs2_upgrade_iopen_glock wakeups
  KMSAN: uninit-value in inode_go_dump (5)
  gfs2: Fix unlinked inode cleanup
  gfs2: Allow immediate GLF_VERIFY_DELETE work
  gfs2: Initialize gl_no_formal_ino earlier
  ...

fs/gfs2/glock.c

@@ -35,6 +35,7 @@
 #include <linux/rhashtable.h>
 #include <linux/pid_namespace.h>
 #include <linux/file.h>
+#include <linux/random.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -562,11 +563,11 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state)
 	gl->gl_tchange = jiffies;
 }
 
-static void gfs2_set_demote(struct gfs2_glock *gl)
+static void gfs2_set_demote(int nr, struct gfs2_glock *gl)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 
-	set_bit(GLF_DEMOTE, &gl->gl_flags);
+	set_bit(nr, &gl->gl_flags);
 	smp_mb();
 	wake_up(&sdp->sd_async_glock_wait);
 }
@@ -958,37 +959,45 @@ static void gfs2_glock_poke(struct gfs2_glock *gl)
 	gfs2_holder_uninit(&gh);
 }
 
-static bool gfs2_try_evict(struct gfs2_glock *gl)
+static void gfs2_try_evict(struct gfs2_glock *gl)
 {
 	struct gfs2_inode *ip;
-	bool evicted = false;
 
 	/*
 	 * If there is contention on the iopen glock and we have an inode, try
-	 * to grab and release the inode so that it can be evicted.  This will
-	 * allow the remote node to go ahead and delete the inode without us
-	 * having to do it, which will avoid rgrp glock thrashing.
+	 * to grab and release the inode so that it can be evicted.  The
+	 * GIF_DEFER_DELETE flag indicates to gfs2_evict_inode() that the inode
+	 * should not be deleted locally.  This will allow the remote node to
+	 * go ahead and delete the inode without us having to do it, which will
+	 * avoid rgrp glock thrashing.
 	 *
 	 * The remote node is likely still holding the corresponding inode
 	 * glock, so it will run before we get to verify that the delete has
-	 * happened below.
+	 * happened below.  (Verification is triggered by the call to
+	 * gfs2_queue_verify_delete() in gfs2_evict_inode().)
 	 */
 	spin_lock(&gl->gl_lockref.lock);
 	ip = gl->gl_object;
 	if (ip && !igrab(&ip->i_inode))
 		ip = NULL;
 	spin_unlock(&gl->gl_lockref.lock);
 	if (ip) {
-		gl->gl_no_formal_ino = ip->i_no_formal_ino;
-		set_bit(GIF_DEFERRED_DELETE, &ip->i_flags);
+		wait_on_inode(&ip->i_inode);
+		if (is_bad_inode(&ip->i_inode)) {
+			iput(&ip->i_inode);
+			ip = NULL;
+		}
+	}
+	if (ip) {
+		set_bit(GIF_DEFER_DELETE, &ip->i_flags);
 		d_prune_aliases(&ip->i_inode);
 		iput(&ip->i_inode);
 
 		/* If the inode was evicted, gl->gl_object will now be NULL. */
 		spin_lock(&gl->gl_lockref.lock);
 		ip = gl->gl_object;
 		if (ip) {
-			clear_bit(GIF_DEFERRED_DELETE, &ip->i_flags);
+			clear_bit(GIF_DEFER_DELETE, &ip->i_flags);
 			if (!igrab(&ip->i_inode))
 				ip = NULL;
 		}
@@ -997,9 +1006,7 @@ static bool gfs2_try_evict(struct gfs2_glock *gl)
 			gfs2_glock_poke(ip->i_gl);
 			iput(&ip->i_inode);
 		}
-		evicted = !ip;
 	}
-	return evicted;
 }
 
 bool gfs2_queue_try_to_evict(struct gfs2_glock *gl)
@@ -1008,70 +1015,47 @@ bool gfs2_queue_try_to_evict(struct gfs2_glock *gl)
 
 	if (test_and_set_bit(GLF_TRY_TO_EVICT, &gl->gl_flags))
 		return false;
-	return queue_delayed_work(sdp->sd_delete_wq,
-				  &gl->gl_delete, 0);
+	return !mod_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, 0);
 }
 
-static bool gfs2_queue_verify_evict(struct gfs2_glock *gl)
+bool gfs2_queue_verify_delete(struct gfs2_glock *gl, bool later)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
+	unsigned long delay;
 
-	if (test_and_set_bit(GLF_VERIFY_EVICT, &gl->gl_flags))
+	if (test_and_set_bit(GLF_VERIFY_DELETE, &gl->gl_flags))
 		return false;
-	return queue_delayed_work(sdp->sd_delete_wq,
-				  &gl->gl_delete, 5 * HZ);
+	delay = later ? HZ + get_random_long() % (HZ * 9) : 0;
+	return queue_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, delay);
 }
 
 static void delete_work_func(struct work_struct *work)
 {
 	struct delayed_work *dwork = to_delayed_work(work);
 	struct gfs2_glock *gl = container_of(dwork, struct gfs2_glock, gl_delete);
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
-	struct inode *inode;
-	u64 no_addr = gl->gl_name.ln_number;
+	bool verify_delete = test_and_clear_bit(GLF_VERIFY_DELETE, &gl->gl_flags);
 
-	if (test_and_clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags)) {
-		/*
-		 * If we can evict the inode, give the remote node trying to
-		 * delete the inode some time before verifying that the delete
-		 * has happened.  Otherwise, if we cause contention on the inode glock
-		 * immediately, the remote node will think that we still have
-		 * the inode in use, and so it will give up waiting.
-		 *
-		 * If we can't evict the inode, signal to the remote node that
-		 * the inode is still in use.  We'll later try to delete the
-		 * inode locally in gfs2_evict_inode.
-		 *
-		 * FIXME: We only need to verify that the remote node has
-		 * deleted the inode because nodes before this remote delete
-		 * rework won't cooperate.  At a later time, when we no longer
-		 * care about compatibility with such nodes, we can skip this
-		 * step entirely.
-		 */
-		if (gfs2_try_evict(gl)) {
-			if (test_bit(SDF_KILL, &sdp->sd_flags))
-				goto out;
-			if (gfs2_queue_verify_evict(gl))
-				return;
-		}
-		goto out;
-	}
+	if (test_and_clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags))
+		gfs2_try_evict(gl);
+
+	if (verify_delete) {
+		u64 no_addr = gl->gl_name.ln_number;
+		struct inode *inode;
 
-	if (test_and_clear_bit(GLF_VERIFY_EVICT, &gl->gl_flags)) {
 		inode = gfs2_lookup_by_inum(sdp, no_addr, gl->gl_no_formal_ino,
 					    GFS2_BLKST_UNLINKED);
 		if (IS_ERR(inode)) {
 			if (PTR_ERR(inode) == -EAGAIN &&
 			    !test_bit(SDF_KILL, &sdp->sd_flags) &&
-			    gfs2_queue_verify_evict(gl))
+			    gfs2_queue_verify_delete(gl, true))
 				return;
 		} else {
 			d_prune_aliases(inode);
 			iput(inode);
 		}
 	}
 
-out:
 	gfs2_glock_put(gl);
 }
 
@@ -1100,7 +1084,7 @@ static void glock_work_func(struct work_struct *work)
 
 		if (!delay) {
 			clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags);
-			gfs2_set_demote(gl);
+			gfs2_set_demote(GLF_DEMOTE, gl);
 		}
 	}
 	run_queue(gl, 0);
@@ -1442,10 +1426,7 @@ int gfs2_glock_async_wait(unsigned int num_gh, struct gfs2_holder *ghs)
 static void request_demote(struct gfs2_glock *gl, unsigned int state,
 			   unsigned long delay, bool remote)
 {
-	if (delay)
-		set_bit(GLF_PENDING_DEMOTE, &gl->gl_flags);
-	else
-		gfs2_set_demote(gl);
+	gfs2_set_demote(delay ? GLF_PENDING_DEMOTE : GLF_DEMOTE, gl);
 	if (gl->gl_demote_state == LM_ST_EXCLUSIVE) {
 		gl->gl_demote_state = state;
 		gl->gl_demote_time = jiffies;
@@ -1635,12 +1616,6 @@ int gfs2_glock_poll(struct gfs2_holder *gh)
 	return test_bit(HIF_WAIT, &gh->gh_iflags) ? 0 : 1;
 }
 
-static inline bool needs_demote(struct gfs2_glock *gl)
-{
-	return (test_bit(GLF_DEMOTE, &gl->gl_flags) ||
-		test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags));
-}
-
 static void __gfs2_glock_dq(struct gfs2_holder *gh)
 {
 	struct gfs2_glock *gl = gh->gh_gl;
@@ -1649,8 +1624,8 @@ static void __gfs2_glock_dq(struct gfs2_holder *gh)
 
 	/*
 	 * This holder should not be cached, so mark it for demote.
-	 * Note: this should be done before the check for needs_demote
-	 * below.
+	 * Note: this should be done before the glock_needs_demote
+	 * check below.
 	 */
 	if (gh->gh_flags & GL_NOCACHE)
 		request_demote(gl, LM_ST_UNLOCKED, 0, false);
@@ -1663,7 +1638,7 @@ static void __gfs2_glock_dq(struct gfs2_holder *gh)
 	 * If there hasn't been a demote request we are done.
 	 * (Let the remaining holders, if any, keep holding it.)
 	 */
-	if (!needs_demote(gl)) {
+	if (!glock_needs_demote(gl)) {
 		if (list_empty(&gl->gl_holders))
 			fast_path = 1;
 	}
@@ -2117,7 +2092,7 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
 void gfs2_cancel_delete_work(struct gfs2_glock *gl)
 {
 	clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags);
-	clear_bit(GLF_VERIFY_EVICT, &gl->gl_flags);
+	clear_bit(GLF_VERIFY_DELETE, &gl->gl_flags);
 	if (cancel_delayed_work(&gl->gl_delete))
 		gfs2_glock_put(gl);
 }
@@ -2370,7 +2345,7 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl)
 		*p++ = 'N';
 	if (test_bit(GLF_TRY_TO_EVICT, gflags))
 		*p++ = 'e';
-	if (test_bit(GLF_VERIFY_EVICT, gflags))
+	if (test_bit(GLF_VERIFY_DELETE, gflags))
 		*p++ = 'E';
 	*p = 0;
 	return buf;

fs/gfs2/glock.h

@@ -245,6 +245,7 @@ static inline int gfs2_glock_nq_init(struct gfs2_glock *gl,
 void gfs2_glock_cb(struct gfs2_glock *gl, unsigned int state);
 void gfs2_glock_complete(struct gfs2_glock *gl, int ret);
 bool gfs2_queue_try_to_evict(struct gfs2_glock *gl);
+bool gfs2_queue_verify_delete(struct gfs2_glock *gl, bool later);
 void gfs2_cancel_delete_work(struct gfs2_glock *gl);
 void gfs2_flush_delete_work(struct gfs2_sbd *sdp);
 void gfs2_gl_hash_clear(struct gfs2_sbd *sdp);
@@ -284,4 +285,10 @@ static inline bool gfs2_holder_queued(struct gfs2_holder *gh)
 void gfs2_inode_remember_delete(struct gfs2_glock *gl, u64 generation);
 bool gfs2_inode_already_deleted(struct gfs2_glock *gl, u64 generation);
 
+static inline bool glock_needs_demote(struct gfs2_glock *gl)
+{
+	return (test_bit(GLF_DEMOTE, &gl->gl_flags) ||
+		test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags));
+}
+
 #endif /* __GLOCK_DOT_H__ */

fs/gfs2/glops.c

@@ -470,7 +470,7 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
  * Returns: errno
  */
 
-int gfs2_inode_refresh(struct gfs2_inode *ip)
+static int gfs2_inode_refresh(struct gfs2_inode *ip)
 {
 	struct buffer_head *dibh;
 	int error;
@@ -494,11 +494,18 @@ int gfs2_inode_refresh(struct gfs2_inode *ip)
 static int inode_go_instantiate(struct gfs2_glock *gl)
 {
 	struct gfs2_inode *ip = gl->gl_object;
+	struct gfs2_glock *io_gl;
+	int error;
 
 	if (!ip) /* no inode to populate - read it in later */
 		return 0;
 
-	return gfs2_inode_refresh(ip);
+	error = gfs2_inode_refresh(ip);
+	if (error)
+		return error;
+	io_gl = ip->i_iopen_gh.gh_gl;
+	io_gl->gl_no_formal_ino = ip->i_no_formal_ino;
+	return 0;
 }
 
 static int inode_go_held(struct gfs2_holder *gh)

fs/gfs2/incore.h

@@ -329,7 +329,7 @@ enum {
 	GLF_BLOCKING			= 15,
 	GLF_UNLOCKED			= 16, /* Wait for glock to be unlocked */
 	GLF_TRY_TO_EVICT		= 17, /* iopen glocks only */
-	GLF_VERIFY_EVICT		= 18, /* iopen glocks only */
+	GLF_VERIFY_DELETE		= 18, /* iopen glocks only */
 };
 
 struct gfs2_glock {
@@ -376,7 +376,7 @@ enum {
 	GIF_SW_PAGED		= 3,
 	GIF_FREE_VFS_INODE      = 5,
 	GIF_GLOP_PENDING	= 6,
-	GIF_DEFERRED_DELETE	= 7,
+	GIF_DEFER_DELETE	= 7,
 };
 
 struct gfs2_inode {

fs/gfs2/inode.c

@@ -750,6 +750,7 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	if (error)
 		goto fail_free_inode;
 	gfs2_cancel_delete_work(io_gl);
+	io_gl->gl_no_formal_ino = ip->i_no_formal_ino;
 
 retry:
 	error = insert_inode_locked4(inode, ip->i_no_addr, iget_test, &ip->i_no_addr);

fs/gfs2/inode.h

@@ -93,8 +93,6 @@ struct inode *gfs2_lookup_by_inum(struct gfs2_sbd *sdp, u64 no_addr,
 				  u64 no_formal_ino,
 				  unsigned int blktype);
 
-int gfs2_inode_refresh(struct gfs2_inode *ip);
-
 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
 			   int is_root);
 int gfs2_permission(struct mnt_idmap *idmap,

fs/gfs2/lock_dlm.c

@@ -224,8 +224,21 @@ static int make_mode(struct gfs2_sbd *sdp, const unsigned int lmstate)
 	return -1;
 }
 
+/* Taken from fs/dlm/lock.c. */
+
+static bool middle_conversion(int cur, int req)
+{
+	return (cur == DLM_LOCK_PR && req == DLM_LOCK_CW) ||
+	       (cur == DLM_LOCK_CW && req == DLM_LOCK_PR);
+}
+
+static bool down_conversion(int cur, int req)
+{
+	return !middle_conversion(cur, req) && req < cur;
+}
+
 static u32 make_flags(struct gfs2_glock *gl, const unsigned int gfs_flags,
-		      const int req)
+		      const int cur, const int req)
 {
 	u32 lkf = 0;
 
@@ -251,7 +264,14 @@ static u32 make_flags(struct gfs2_glock *gl, const unsigned int gfs_flags,
 
 	if (!test_bit(GLF_INITIAL, &gl->gl_flags)) {
 		lkf |= DLM_LKF_CONVERT;
-		if (test_bit(GLF_BLOCKING, &gl->gl_flags))
+
+		/*
+		 * The DLM_LKF_QUECVT flag needs to be set for "first come,
+		 * first served" semantics, but it must only be set for
+		 * "upward" lock conversions or else DLM will reject the
+		 * request as invalid.
+		 */
+		if (!down_conversion(cur, req))
 			lkf |= DLM_LKF_QUECVT;
 	}
 
@@ -271,13 +291,14 @@ static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state,
 		     unsigned int flags)
 {
 	struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct;
-	int req;
+	int cur, req;
 	u32 lkf;
 	char strname[GDLM_STRNAME_BYTES] = "";
 	int error;
 
+	cur = make_mode(gl->gl_name.ln_sbd, gl->gl_state);
 	req = make_mode(gl->gl_name.ln_sbd, req_state);
-	lkf = make_flags(gl, flags, req);
+	lkf = make_flags(gl, flags, cur, req);
 	gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT);
 	gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT);
 	if (test_bit(GLF_INITIAL, &gl->gl_flags)) {