Re: [PATCH v5 10/21] nfsd: add notification handlers for dir events

On Mon, 2026-06-08 at 16:40 -0400, Chuck Lever wrote:
> 
> On Fri, May 22, 2026, at 3:42 PM, Jeff Layton wrote:
> > Add the necessary parts to accept a fsnotify callback for directory
> > change event and create a CB_NOTIFY request for it. When a dir nfsd_file
> > is created set a handle_event callback to handle the notification.
> > 
> > Use that to allocate a nfsd_notify_event object and then hand off a
> > reference to each delegation's CB_NOTIFY. If anything fails along the
> > way, recall any affected delegations.
> > 
> > Signed-off-by: Jeff Layton <[email protected]>
> 
> There are some significant-looking sashiko review findings which I did
> not follow up on.
> 

I plan to go over Sashiko's findings after I go through your responses.

> 
> > diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
> > index ea3e7deb06fa..1964a213f80e 100644
> > --- a/fs/nfsd/nfs4callback.c
> > +++ b/fs/nfsd/nfs4callback.c
> > @@ -870,21 +870,30 @@ static void nfs4_xdr_enc_cb_notify(struct 
> > rpc_rqst *req,
> >  				   const void *data)
> >  {
> >  	const struct nfsd4_callback *cb = data;
> > +	struct nfsd4_cb_notify *ncn = container_of(cb, struct 
> > nfsd4_cb_notify, ncn_cb);
> > +	struct nfs4_delegation *dp = container_of(ncn, struct 
> > nfs4_delegation, dl_cb_notify);
> >  	struct nfs4_cb_compound_hdr hdr = {
> >  		.ident = 0,
> >  		.minorversion = cb->cb_clp->cl_minorversion,
> >  	};
> > -	struct CB_NOTIFY4args args = { };
> > +	struct CB_NOTIFY4args args;
> > +	__be32 *p;
> > 
> >  	WARN_ON_ONCE(hdr.minorversion == 0);
> > 
> >  	encode_cb_compound4args(xdr, &hdr);
> >  	encode_cb_sequence4args(xdr, cb, &hdr);
> > 
> > -	/*
> > -	 * FIXME: get stateid and fh from delegation. Inline the cna_changes
> > -	 * buffer, and zero it.
> > -	 */
> > +	p = xdr_reserve_space(xdr, 4);
> > +	*p = cpu_to_be32(OP_CB_NOTIFY);
> > +
> > +	args.cna_stateid.seqid = dp->dl_stid.sc_stateid.si_generation;
> > +	memcpy(&args.cna_stateid.other, &dp->dl_stid.sc_stateid.si_opaque,
> > +	       ARRAY_SIZE(args.cna_stateid.other));
> > +	args.cna_fh.len = dp->dl_stid.sc_file->fi_fhandle.fh_size;
> > +	args.cna_fh.data = dp->dl_stid.sc_file->fi_fhandle.fh_raw;
> > +	args.cna_changes.count = ncn->ncn_nf_cnt;
> > +	args.cna_changes.element = ncn->ncn_nf;
> >  	WARN_ON_ONCE(!xdrgen_encode_CB_NOTIFY4args(xdr, &args));
> > 
> >  	hdr.nops++;
> 
> I want to avoid the need to use xdrgen to encode the CB_NOTIFY arguments.
> How about this:
> 
> +       struct nfsd4_cb_notify *ncn = container_of(cb, struct nfsd4_cb_notify, ncn_cb);
> +       struct nfs4_delegation *dp = container_of(ncn, struct nfs4_delegation, dl_cb_notify);
> 
>    ...
> 
> +       encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
> +       encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
> +       xdr_stream_encode_u32(xdr, ncn->ncn_nf_cnt);
> +       for (u32 i = 0; i < ncn->ncn_nf_cnt; i++)
> +               (void)xdrgen_encode_notify4(xdr, &ncn->ncn_nf[i]);
> 
> And then add a "pragma public notify4;" in nfs4_1.x .
> 

For those following along, Chuck and I had a private discussion and I
think we're going to keep this calling xdrgen_encode_CB_NOTIFY4args()
for now. I am dropping the WARN_ON_ONCE though.

> 
> > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> > index b0652c755b3b..20477144475b 100644
> > --- a/fs/nfsd/nfs4state.c
> > +++ b/fs/nfsd/nfs4state.c
> 
> > @@ -3461,19 +3462,131 @@ nfsd4_cb_getattr_release(struct nfsd4_callback *cb)
> >  	nfs4_put_stid(&dp->dl_stid);
> >  }
> > 
> > +static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
> > +{
> > +	bool queued;
> > +
> > +	if (test_and_set_bit(NFSD4_CALLBACK_RUNNING, &dp->dl_recall.cb_flags))
> > +		return;
> > +
> > +	/*
> > +	 * We're assuming the state code never drops its reference
> > +	 * without first removing the lease.  Since we're in this lease
> > +	 * callback (and since the lease code is serialized by the
> > +	 * flc_lock) we know the server hasn't removed the lease yet, and
> > +	 * we know it's safe to take a reference.
> > +	 */
> > +	refcount_inc(&dp->dl_stid.sc_count);
> > +	queued = nfsd4_run_cb(&dp->dl_recall);
> > +	WARN_ON_ONCE(!queued);
> > +	if (!queued)
> > +		refcount_dec(&dp->dl_stid.sc_count);
> > +}
> > +
> > +static bool
> > +nfsd4_cb_notify_prepare(struct nfsd4_callback *cb)
> > +{
> > +	struct nfsd4_cb_notify *ncn = container_of(cb, struct 
> > nfsd4_cb_notify, ncn_cb);
> > +	struct nfs4_delegation *dp = container_of(ncn, struct 
> > nfs4_delegation, dl_cb_notify);
> > +	struct nfsd_notify_event *events[NOTIFY4_EVENT_QUEUE_SIZE];
> > +	struct xdr_buf xdr = { .buflen = PAGE_SIZE * NOTIFY4_PAGE_ARRAY_SIZE,
> > +			       .pages  = ncn->ncn_pages };
> > +	struct xdr_stream stream;
> > +	struct nfsd_file *nf;
> > +	int count, i;
> > +	bool error = false;
> > +
> > +	xdr_init_encode_pages(&stream, &xdr);
> > +
> > +	spin_lock(&ncn->ncn_lock);
> > +	count = ncn->ncn_evt_cnt;
> > +
> > +	/* spurious queueing? */
> > +	if (count == 0) {
> > +		spin_unlock(&ncn->ncn_lock);
> > +		return false;
> > +	}
> > +
> > +	/* we can't keep up! */
> > +	if (count > NOTIFY4_EVENT_QUEUE_SIZE) {
> > +		spin_unlock(&ncn->ncn_lock);
> > +		goto out_recall;
> > +	}
> > +
> > +	memcpy(events, ncn->ncn_evt, sizeof(*events) * count);
> > +	ncn->ncn_evt_cnt = 0;
> > +	spin_unlock(&ncn->ncn_lock);
> > +
> > +	rcu_read_lock();
> > +	nf = 
> > nfsd_file_get(rcu_dereference(dp->dl_stid.sc_file->fi_deleg_file));
> > +	rcu_read_unlock();
> > +	if (!nf) {
> > +		for (i = 0; i < count; ++i)
> > +			nfsd_notify_event_put(events[i]);
> > +		goto out_recall;
> > +	}
> > +
> > +	for (i = 0; i < count; ++i) {
> > +		struct nfsd_notify_event *nne = events[i];
> > +
> > +		if (!error) {
> > +			u32 *maskp = (u32 *)xdr_reserve_space(&stream, sizeof(*maskp));
> > +			u8 *p;
> > +
> > +			if (!maskp) {
> > +				error = true;
> > +				goto put_event;
> > +			}
> > +
> > +			p = nfsd4_encode_notify_event(&stream, nne, dp, nf, maskp);
> > +			if (!p) {
> > +				pr_notice("Could not generate CB_NOTIFY from fsnotify mask 0x%x\n",
> > +					  nne->ne_mask);
> > +				error = true;
> > +				goto put_event;
> > +			}
> > +
> > +			ncn->ncn_nf[i].notify_mask.count = 1;
> > +			ncn->ncn_nf[i].notify_mask.element = maskp;
> > +			ncn->ncn_nf[i].notify_vals.data = p;
> > +			ncn->ncn_nf[i].notify_vals.len = (u8 *)stream.p - p;
> > +		}
> > +put_event:
> > +		nfsd_notify_event_put(nne);
> > +	}
> > +	if (!error) {
> > +		ncn->ncn_nf_cnt = count;
> > +		nfsd_file_put(nf);
> > +		return true;
> > +	}
> > +	nfsd_file_put(nf);
> > +out_recall:
> > +	nfsd_break_one_deleg(dp);
> > +	return false;
> > +}
> > +
> >  static int
> >  nfsd4_cb_notify_done(struct nfsd4_callback *cb,
> >  				struct rpc_task *task)
> >  {
> > +	struct nfsd4_cb_notify *ncn = container_of(cb, struct 
> > nfsd4_cb_notify, ncn_cb);
> > +	struct nfs4_delegation *dp = container_of(ncn, struct 
> > nfs4_delegation, dl_cb_notify);
> > +
> >  	switch (task->tk_status) {
> >  	case -NFS4ERR_DELAY:
> >  		rpc_delay(task, 2 * HZ);
> >  		return 0;
> >  	default:
> > +		/* For any other hard error, recall the deleg */
> > +		nfsd_break_one_deleg(dp);
> > +		fallthrough;
> > +	case 0:
> >  		return 1;
> >  	}
> >  }
> > 
> > +static void nfsd4_run_cb_notify(struct nfsd4_cb_notify *ncn);
> > +
> >  static void
> >  nfsd4_cb_notify_release(struct nfsd4_callback *cb)
> >  {
> > @@ -3482,6 +3595,9 @@ nfsd4_cb_notify_release(struct nfsd4_callback *cb)
> >  	struct nfs4_delegation *dp =
> >  			container_of(ncn, struct nfs4_delegation, dl_cb_notify);
> > 
> > +	/* Drain events that arrived while this callback was in flight */
> > +	if (ncn->ncn_evt_cnt > 0)
> > +		nfsd4_run_cb_notify(ncn);
> 
> The above check needs to be serialized with modification of
> ncn_evt_cnt:
>
> +       bool pending;
>  
> +       /* Drain events that arrived while this callback was in flight */
> +       spin_lock(&ncn->ncn_lock);
> +       pending = ncn->ncn_evt_cnt > 0;
> +       spin_unlock(&ncn->ncn_lock);
> +       if (pending)
> +               nfsd4_run_cb_notify(ncn);
> 

I need to ponder this. Does this matter?

NFSD4_CALLBACK_RUNNING is now clear, which should be observed by
another task queueing a new event. READ_ONCE() seems like it should be
sufficient here. I'll run it by Claude.


> 
> >  	nfs4_put_stid(&dp->dl_stid);
> >  }
> > 
> 
> > @@ -9858,3 +9954,133 @@ void nfsd_update_cmtime_attr(struct file *f, 
> > unsigned int flags)
> >  				      MINOR(inode->i_sb->s_dev),
> >  				      inode->i_ino, ret);
> >  }
> > +
> > +static void
> > +nfsd4_run_cb_notify(struct nfsd4_cb_notify *ncn)
> > +{
> > +	struct nfs4_delegation *dp = container_of(ncn, struct 
> > nfs4_delegation, dl_cb_notify);
> > +
> > +	if (test_and_set_bit(NFSD4_CALLBACK_RUNNING, &ncn->ncn_cb.cb_flags))
> > +		return;
> > +
> > +	if (!refcount_inc_not_zero(&dp->dl_stid.sc_count))
> > +		clear_bit(NFSD4_CALLBACK_RUNNING, &ncn->ncn_cb.cb_flags);
> > +	else
> > +		nfsd4_run_cb(&ncn->ncn_cb);
> > +}
> > +
> > +static struct nfsd_notify_event *
> > +alloc_nfsd_notify_event(u32 mask, const struct qstr *q, struct dentry 
> > *dentry,
> > +			struct inode *target)
> > +{
> > +	struct nfsd_notify_event *ne;
> > +
> > +	ne = kmalloc(sizeof(*ne) + q->len + 1, GFP_NOFS);
> > +	if (!ne)
> > +		return NULL;
> > +
> > +	memcpy(&ne->ne_name, q->name, q->len);
> > +	refcount_set(&ne->ne_ref, 1);
> > +	ne->ne_mask = mask;
> > +	ne->ne_name[q->len] = '\0';
> > +	ne->ne_namelen = q->len;
> > +	ne->ne_dentry = dget(dentry);
> > +	ne->ne_target = target;
> > +	if (ne->ne_target)
> > +		ihold(ne->ne_target);
> > +	return ne;
> > +}
> > +
> > +static bool
> > +should_notify_deleg(u32 mask, struct file_lease *fl)
> > +{
> > +	/* Don't notify the client generating the event */
> > +	if (nfsd_breaker_owns_lease(fl))
> > +		return false;
> > +
> > +	/* Skip if this event wasn't ignored by the lease */
> > +	if ((mask & FS_DELETE) && !(fl->c.flc_flags & FL_IGN_DIR_DELETE))
> > +		return false;
> > +	if ((mask & FS_CREATE) && !(fl->c.flc_flags & FL_IGN_DIR_CREATE))
> > +		return false;
> > +	if ((mask & FS_RENAME) && !(fl->c.flc_flags & FL_IGN_DIR_RENAME))
> > +		return false;
> > +
> > +	return true;
> > +}
> > +
> > +static void
> > +nfsd_recall_all_dir_delegs(const struct inode *dir)
> > +{
> > +	struct file_lock_context *ctx = locks_inode_context(dir);
> > +	struct file_lock_core *flc;
> > +
> > +	spin_lock(&ctx->flc_lock);
> > +	list_for_each_entry(flc, &ctx->flc_lease, flc_list) {
> > +		struct file_lease *fl = container_of(flc, struct file_lease, c);
> > +
> > +		if (fl->fl_lmops == &nfsd_lease_mng_ops)
> > +			nfsd_break_deleg_cb(fl);
> > +	}
> > +	spin_unlock(&ctx->flc_lock);
> > +}
> > +
> > +int
> > +nfsd_handle_dir_event(u32 mask, const struct inode *dir, const void 
> > *data,
> > +		      int data_type, const struct qstr *name)
> > +{
> > +	struct dentry *dentry = fsnotify_data_dentry(data, data_type);
> > +	struct inode *target = fsnotify_data_rename_target(data, data_type);
> > +	struct file_lock_context *ctx;
> > +	struct file_lock_core *flc;
> > +	struct nfsd_notify_event *evt;
> > +
> > +	/* Normalize cross-dir rename events to create/delete */
> > +	if (mask & FS_MOVED_FROM) {
> > +		mask &= ~FS_MOVED_FROM;
> > +		mask |= FS_DELETE;
> > +	}
> > +	if (mask & FS_MOVED_TO) {
> > +		mask &= ~FS_MOVED_TO;
> > +		mask |= FS_CREATE;
> > +	}
> > +
> 
> I inserted an extra check here for rename notifications:
> 
> +       /*
> +        * FS_RENAME fires on the source directory even for a cross-dir
> +        * rename, where the moved entry now lives under a different
> +        * parent. NOTIFY4_RENAME_ENTRY describes an in-place rename, so
> +        * reporting it here would advertise a name absent from this
> +        * directory.
> +        */
> +       if ((mask & FS_RENAME) && dentry && d_inode(dentry->d_parent) != dir)
> +               mask &= ~FS_RENAME;
> 

Thanks. I'll add that in.

> 
> > +	/* Don't do anything if this is not an expected event */
> > +	if (!(mask & (FS_CREATE|FS_DELETE|FS_RENAME)))
> > +		return 0;
> > +
> > +	ctx = locks_inode_context(dir);
> > +	if (!ctx || list_empty(&ctx->flc_lease))
> > +		return 0;
> > +
> > +	evt = alloc_nfsd_notify_event(mask, name, dentry, target);
> > +	if (!evt) {
> > +		nfsd_recall_all_dir_delegs(dir);
> > +		return 0;
> > +	}
> > +
> > +	spin_lock(&ctx->flc_lock);
> > +	list_for_each_entry(flc, &ctx->flc_lease, flc_list) {
> > +		struct file_lease *fl = container_of(flc, struct file_lease, c);
> > +		struct nfs4_delegation *dp = flc->flc_owner;
> > +		struct nfsd4_cb_notify *ncn = &dp->dl_cb_notify;
> > +
> 
> I added:
> 
> +               if (fl->fl_lmops != &nfsd_lease_mng_ops)
> +                       continue;
> 
> Otherwise the loop treats every lease on the inode as an nfsd delegation
> unconditionally.
> 

This is not necessary. should_notify_deleg() calls
nfsd_breaker_owns_lease(), which already checks this before doing
anything else.

> 
> > +		if (!should_notify_deleg(mask, fl))
> > +			continue;
> > +
> > +		spin_lock(&ncn->ncn_lock);
> > +		if (ncn->ncn_evt_cnt >= NOTIFY4_EVENT_QUEUE_SIZE) {
> > +			/* We're generating notifications too fast. Recall. */
> > +			spin_unlock(&ncn->ncn_lock);
> > +			nfsd_break_deleg_cb(fl);
> > +			continue;
> > +		}
> > +		ncn->ncn_evt[ncn->ncn_evt_cnt++] = nfsd_notify_event_get(evt);
> > +		spin_unlock(&ncn->ncn_lock);
> > +
> > +		nfsd4_run_cb_notify(ncn);
> > +	}
> > +	spin_unlock(&ctx->flc_lock);
> > +	nfsd_notify_event_put(evt);
> > +	return 0;
> > +}
> > diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> > index e17488a911f7..31df04675713 100644
> > --- a/fs/nfsd/nfs4xdr.c
> > +++ b/fs/nfsd/nfs4xdr.c
> > @@ -4172,6 +4172,127 @@ nfsd4_encode_fattr4(struct svc_rqst *rqstp, 
> > struct xdr_stream *xdr,
> >  	goto out;
> >  }
> > 
> > +static bool
> > +nfsd4_setup_notify_entry4(struct notify_entry4 *ne, struct xdr_stream 
> > *xdr,
> > +			  struct dentry *dentry, struct nfs4_delegation *dp,
> > +			  struct nfsd_file *nf, char *name, u32 namelen)
> > +{
> > +	uint32_t *attrmask;
> > +
> > +	/* Reserve space for attrmask */
> > +	attrmask = xdr_reserve_space(xdr, 3 * sizeof(uint32_t));
> > +	if (!attrmask)
> > +		return false;
> > +
> > +	ne->ne_file.data = name;
> > +	ne->ne_file.len = namelen;
> > +	ne->ne_attrs.attrmask.element = attrmask;
> > +
> > +	attrmask[0] = 0;
> > +	attrmask[1] = 0;
> > +	attrmask[2] = 0;
> > +	ne->ne_attrs.attr_vals.data = NULL;
> > +	ne->ne_attrs.attr_vals.len = 0;
> > +	ne->ne_attrs.attrmask.count = 1;
> > +	return true;
> > +}
> > +
> > +/**
> > + * nfsd4_encode_notify_event - encode a notify
> > + * @xdr: stream to which to encode the fattr4
> > + * @nne: nfsd_notify_event to encode
> > + * @dp: delegation where the event occurred
> > + * @nf: nfsd_file on which event occurred
> > + * @notify_mask: pointer to word where notification mask should be set
> > + *
> > + * Encode @nne into @xdr. Returns a pointer to the start of the event, 
> > or NULL if
> > + * the event couldn't be encoded. The appropriate bit in the 
> > notify_mask will also
> > + * be set on success.
> > + */
> 
> Nit: Let's use the usual kdoc style to describe the return value.
> 

Ok, will fix.

> + * Encode @nne into @xdr. The matching bit in @notify_mask is set on
> + * success.
> + *
> + * Return: pointer to the start of the encoded event, or NULL if the
> + * event could not be encoded.
> + */
> 
> 
> > +u8 *nfsd4_encode_notify_event(struct xdr_stream *xdr, struct 
> > nfsd_notify_event *nne,
> > +			      struct nfs4_delegation *dp, struct nfsd_file *nf,
> > +			      u32 *notify_mask)
> > +{
> > +	u8 *p = NULL;
> > +
> > +	*notify_mask = 0;
> > +
> > +	if (nne->ne_mask & FS_DELETE) {
> > +		struct notify_remove4 nr = { };
> > +
> > +		if (!nfsd4_setup_notify_entry4(&nr.nrm_old_entry, xdr, 
> > nne->ne_dentry, dp,
> > +					       nf, nne->ne_name, nne->ne_namelen))
> > +			goto out_err;
> > +		p = (u8 *)xdr->p;
> > +		if (!xdrgen_encode_notify_remove4(xdr, &nr))
> > +			goto out_err;
> > +		*notify_mask |= BIT(NOTIFY4_REMOVE_ENTRY);
> > +	} else if (nne->ne_mask & FS_CREATE) {
> > +		struct notify_add4 na = { };
> > +		struct notify_remove4 old = { };
> > +
> > +		if (!nfsd4_setup_notify_entry4(&na.nad_new_entry, xdr, 
> > nne->ne_dentry, dp,
> > +					       nf, nne->ne_name, nne->ne_namelen))
> > +			goto out_err;
> > +
> > +		/* If a file was overwritten, report it in nad_old_entry */
> > +		if (nne->ne_target) {
> > +			if (!nfsd4_setup_notify_entry4(&old.nrm_old_entry, xdr,
> > +						       NULL, dp, nf,
> > +						       nne->ne_name, nne->ne_namelen))
> > +				goto out_err;
> > +			na.nad_old_entry.count = 1;
> > +			na.nad_old_entry.element = &old;
> > +		}
> > +
> > +		p = (u8 *)xdr->p;
> > +		if (!xdrgen_encode_notify_add4(xdr, &na))
> > +			goto out_err;
> > +
> > +		*notify_mask |= BIT(NOTIFY4_ADD_ENTRY);
> > +	} else if (nne->ne_mask & FS_RENAME) {
> > +		struct notify_rename4 nr = { };
> > +		struct notify_remove4 old = { };
> > +		struct name_snapshot n;
> > +		bool ret;
> > +
> > +		/* Don't send any attributes in the old_entry since they're the same 
> > in new */
> > +		if (!nfsd4_setup_notify_entry4(&nr.nrn_old_entry.nrm_old_entry, xdr,
> > +					       NULL, dp, nf, nne->ne_name,
> > +					       nne->ne_namelen))
> > +			goto out_err;
> > +
> > +		take_dentry_name_snapshot(&n, nne->ne_dentry);
> > +		ret = nfsd4_setup_notify_entry4(&nr.nrn_new_entry.nad_new_entry, xdr,
> > +					       nne->ne_dentry, dp, nf, (char *)n.name.name,
> > +					       n.name.len);
> > +
> > +		/* If a file was overwritten, report it in nad_old_entry */
> > +		if (ret && nne->ne_target) {
> > +			ret = nfsd4_setup_notify_entry4(&old.nrm_old_entry, xdr,
> > +							NULL, dp, nf,
> > +							(char *)n.name.name, n.name.len);
> > +			if (ret) {
> > +				nr.nrn_new_entry.nad_old_entry.count = 1;
> > +				nr.nrn_new_entry.nad_old_entry.element = &old;
> > +			}
> > +		}
> > +
> > +		if (ret) {
> > +			p = (u8 *)xdr->p;
> > +			ret = xdrgen_encode_notify_rename4(xdr, &nr);
> > +		}
> > +		release_dentry_name_snapshot(&n);
> > +		if (!ret)
> > +			goto out_err;
> > +		*notify_mask |= BIT(NOTIFY4_RENAME_ENTRY);
> > +	}
> > +	return p;
> > +out_err:
> > +	pr_warn("nfsd: unable to marshal notify_rename4 to xdr stream\n");
> 
> Nit: The warning needs to match the semantics of nfsd4_encode_notify_event().
> How about:
> 
> +       pr_warn("nfsd: unable to marshal notify event to xdr stream\n");
> 

Sounds good.

> 
> > +	return NULL;
> > +}
> > +
> 

-- 
Jeff Layton <[email protected]>