Skip to contentSkip to conversation

nfs4_show_superblock considered harmful :-)

DISCUSSION NeilBrown 2 participants 11 messages May 29, 2020 – Jul 17, 2020

Subsystems (2)
lore
Share & tools
lore

NeilBrown

author
#0 lore raw

Hi,
I've received a report of a 5.3 kernel crashing in nfs4_show_superblock().
I was part way through preparing a patch when I concluded that the problem wasn't as straight forward as I thought.

In the crash, the 'struct file *' passed to nfs4_show_superblock() was NULL.
This file was acquired from find_any_file(), and every other caller of find_any_file() checks that the returned value is not NULL (though one BUGs if it is NULL - another WARNs).
But nfs4_show_open() and nfs4_show_lock() don't.
Maybe they should. I didn't double check, but I suspect they don't hold enough locks to ensure that the files don't get removed.

Then I noticed that nfs4_show_deleg() accesses fi_deleg_file without checking if it is NULL - Should it take fi_lock and make sure it is not NULL - and get a counted reference?
And maybe nfs4_show_layout() has the same problem?

I could probably have worked my way through fixing all of these, but then I discovered that these things are now 'struct nfsd_file *' rather than 'struct file *' and that the helpful documentation says:

* Note that this object doesn't
* hold a reference to the inode by itself, so the nf_inode pointer should
* never be dereferenced, only used for comparison.

and yet nfs4_show_superblock() contains:

	struct inode *inode = f->nf_inode;

	seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
					MAJOR(inode->i_sb->s_dev),
					 MINOR(inode->i_sb->s_dev),
					 inode->i_ino);

do you see my problem?

Is this really safe and the doco wrong? (I note that the use of nf_inode in nfsd_file_mark_find_or_create() looks wrong, but is actually safe). Or should we check if nf_file is non-NULL and use that?

In short:
- We should check find_any_file() return value - correct?
- Do we need extra locking to stabilize fi_deleg_file?
- ditto for ->ls_file
- how can nfs4_show_superblock safely get s_dev and i_ino from a nfsd_file?

Thanks,

NeilBrown

#1 lore raw
Quoted 5 lines · NeilBrown

I've received a report of a 5.3 kernel crashing in nfs4_show_superblock().
I was part way through preparing a patch when I concluded that the problem wasn't as straight forward as I thought.

Quoted 8 lines

In the crash, the 'struct file *' passed to nfs4_show_superblock() was NULL.
This file was acquired from find_any_file(), and every other caller of find_any_file() checks that the returned value is not NULL (though one BUGs if it is NULL - another WARNs).
But nfs4_show_open() and nfs4_show_lock() don't.
Maybe they should. I didn't double check, but I suspect they don't hold enough locks to ensure that the files don't get removed.

I think the only lock held is cl_lock, acquired in states_start.

We're starting here with an nfs4_stid that was found in the cl_stateids idr.

A struct nfs4_stid is freed by nfs4_put_stid(), which removes it from that idr under cl_lock before freeing the nfs4_stid and anything it points to.

I think that was the theory....

One possible problem is downgrades, like nfs4_stateid_downgrade.

I'll keep mulling it over, thanks.

--b.

Quoted 40 lines

Then I noticed that nfs4_show_deleg() accesses fi_deleg_file without checking if it is NULL - Should it take fi_lock and make sure it is not NULL - and get a counted reference?
And maybe nfs4_show_layout() has the same problem?

I could probably have worked my way through fixing all of these, but then I discovered that these things are now 'struct nfsd_file *' rather than 'struct file *' and that the helpful documentation says:

* Note that this object doesn't
* hold a reference to the inode by itself, so the nf_inode pointer should
* never be dereferenced, only used for comparison.

and yet nfs4_show_superblock() contains:

	struct inode *inode = f->nf_inode;

	seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
					MAJOR(inode->i_sb->s_dev),
					 MINOR(inode->i_sb->s_dev),
					 inode->i_ino);

do you see my problem?

Is this really safe and the doco wrong? (I note that the use of nf_inode in nfsd_file_mark_find_or_create() looks wrong, but is actually safe). Or should we check if nf_file is non-NULL and use that?

In short:
- We should check find_any_file() return value - correct?
- Do we need extra locking to stabilize fi_deleg_file?
- ditto for ->ls_file
- how can nfs4_show_superblock safely get s_dev and i_ino from a nfsd_file?

Thanks,

NeilBrown

author
#2 lore raw
Quoted 24 lines · J. Bruce Fields
Quoted 5 lines · NeilBrown

I've received a report of a 5.3 kernel crashing in nfs4_show_superblock().
I was part way through preparing a patch when I concluded that the problem wasn't as straight forward as I thought.

Quoted 8 lines

In the crash, the 'struct file *' passed to nfs4_show_superblock() was NULL.
This file was acquired from find_any_file(), and every other caller of find_any_file() checks that the returned value is not NULL (though one BUGs if it is NULL - another WARNs).
But nfs4_show_open() and nfs4_show_lock() don't.
Maybe they should. I didn't double check, but I suspect they don't hold enough locks to ensure that the files don't get removed.

I think the only lock held is cl_lock, acquired in states_start.

We're starting here with an nfs4_stid that was found in the cl_stateids idr.

A struct nfs4_stid is freed by nfs4_put_stid(), which removes it from that idr under cl_lock before freeing the nfs4_stid and anything it points to.

Quoted 5 lines

I think that was the theory....

One possible problem is downgrades, like nfs4_stateid_downgrade.

I'll keep mulling it over, thanks.

I had another look at code and maybe move_to_close_lru() is the problem. It can clear remove the files and clear sc_file without taking cl_lock. So some protection is needed against that.

I think that only applies to nfs4_show_open() - not show_lock etc. But I wonder it is might be best to include some extra protection for each different case, just in case some future code change allow sc_file to become NULL before the state is detached.

I'd feel more comforatable about nfs4_show_superblock() if it ignored nf_inode and just used nf_file - it is isn't NULL. It looks like it can never be set from non-NULL to NULL.

Thanks,
NeilBrown

#3 lore raw
Quoted 26 lines · NeilBrown
Quoted 24 lines · J. Bruce Fields
Quoted 5 lines · NeilBrown

I've received a report of a 5.3 kernel crashing in nfs4_show_superblock().
I was part way through preparing a patch when I concluded that the problem wasn't as straight forward as I thought.

Quoted 8 lines

In the crash, the 'struct file *' passed to nfs4_show_superblock() was NULL.
This file was acquired from find_any_file(), and every other caller of find_any_file() checks that the returned value is not NULL (though one BUGs if it is NULL - another WARNs).
But nfs4_show_open() and nfs4_show_lock() don't.
Maybe they should. I didn't double check, but I suspect they don't hold enough locks to ensure that the files don't get removed.

I think the only lock held is cl_lock, acquired in states_start.

We're starting here with an nfs4_stid that was found in the cl_stateids idr.

A struct nfs4_stid is freed by nfs4_put_stid(), which removes it from that idr under cl_lock before freeing the nfs4_stid and anything it points to.

Quoted 6 lines
Quoted 5 lines

I think that was the theory....

One possible problem is downgrades, like nfs4_stateid_downgrade.

I'll keep mulling it over, thanks.

Oops, I neglected this a while....

Quoted 4 lines

I had another look at code and maybe move_to_close_lru() is the problem. It can clear remove the files and clear sc_file without taking cl_lock. So some protection is needed against that.

Quoted 8 lines

I think that only applies to nfs4_show_open() - not show_lock etc. But I wonder it is might be best to include some extra protection for each different case, just in case some future code change allow sc_file to become NULL before the state is detached.

I'd feel more comforatable about nfs4_show_superblock() if it ignored nf_inode and just used nf_file - it is isn't NULL. It looks like it can never be set from non-NULL to NULL.

But then that means we've always got a reference on the inode, doesn't it? So I still don't understand the nf_inode comment.

So maybe the NULL checks are mainly all we need.

Also it looks to me like ls_file lasts as long as the layout stateid, so maybe it's OK.

--b.

commit 4eef57aa4fc0

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

fs/nfsd/nfs4state.c +17 -1
index ab5c8857ae5a..08b8376c74d7 100644
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+
 static atomic_long_t num_delegations;
 unsigned long max_delegations;
 
@@ -2444,6 +2454,8 @@ static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2481,6 +2493,8 @@ static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
author
#4 lore raw
Quoted 45 lines · J. Bruce Fields
Quoted 26 lines · NeilBrown
Quoted 24 lines · J. Bruce Fields
Quoted 5 lines · NeilBrown

I've received a report of a 5.3 kernel crashing in nfs4_show_superblock().
I was part way through preparing a patch when I concluded that the problem wasn't as straight forward as I thought.

Quoted 8 lines

In the crash, the 'struct file *' passed to nfs4_show_superblock() was NULL.
This file was acquired from find_any_file(), and every other caller of find_any_file() checks that the returned value is not NULL (though one BUGs if it is NULL - another WARNs).
But nfs4_show_open() and nfs4_show_lock() don't.
Maybe they should. I didn't double check, but I suspect they don't hold enough locks to ensure that the files don't get removed.

I think the only lock held is cl_lock, acquired in states_start.

We're starting here with an nfs4_stid that was found in the cl_stateids idr.

A struct nfs4_stid is freed by nfs4_put_stid(), which removes it from that idr under cl_lock before freeing the nfs4_stid and anything it points to.

Quoted 6 lines
Quoted 5 lines

I think that was the theory....

One possible problem is downgrades, like nfs4_stateid_downgrade.

I'll keep mulling it over, thanks.

Oops, I neglected this a while....

Quoted 9 lines

I had another look at code and maybe move_to_close_lru() is the problem. It can clear remove the files and clear sc_file without taking cl_lock. So some protection is needed against that.

I think that only applies to nfs4_show_open() - not show_lock etc. But I wonder it is might be best to include some extra protection for each different case, just in case some future code change allow sc_file to become NULL before the state is detached.

Quoted 6 lines
Quoted 3 lines

I'd feel more comforatable about nfs4_show_superblock() if it ignored nf_inode and just used nf_file - it is isn't NULL. It looks like it can never be set from non-NULL to NULL.

But then that means we've always got a reference on the inode, doesn't it? So I still don't understand the nf_inode comment.

My main problem with nf_inode is the comment

/*
* A representation of a file that has been opened by knfsd. These are hashed
* in the hashtable by inode pointer value. Note that this object doesn't
* hold a reference to the inode by itself, so the nf_inode pointer should
* never be dereferenced, only used for comparison.
*/

That comment is incompatible with the code in nfsd_file_mark_find_or_create() and with the code in nfs4_show_superblock().

Quoted 57 lines

So maybe the NULL checks are mainly all we need.

Also it looks to me like ls_file lasts as long as the layout stateid, so maybe it's OK.

--b.

commit 4eef57aa4fc0

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

fs/nfsd/nfs4state.c +14 -0
index ab5c8857ae5a..08b8376c74d7 100644
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+
 static atomic_long_t num_delegations;
 unsigned long max_delegations;
 
@@ -2444,6 +2454,8 @@ static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2481,6 +2493,8 @@ static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);

fs/nfsd/nfs4state.c static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st) · line 2535

Quoted 10 lines
fs/nfsd/nfs4state.c +3 -1
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);

You'll need to add nfsd_file_put(file) toward the end of this function. Otherwise, I think this patch is a step in the right direction.

Thanks,
NeilBrown

#5 lore raw
Quoted 63 lines · NeilBrown
Quoted 45 lines · J. Bruce Fields
Quoted 26 lines · NeilBrown
Quoted 24 lines · J. Bruce Fields
Quoted 5 lines · NeilBrown

I've received a report of a 5.3 kernel crashing in nfs4_show_superblock().
I was part way through preparing a patch when I concluded that the problem wasn't as straight forward as I thought.

Quoted 8 lines

In the crash, the 'struct file *' passed to nfs4_show_superblock() was NULL.
This file was acquired from find_any_file(), and every other caller of find_any_file() checks that the returned value is not NULL (though one BUGs if it is NULL - another WARNs).
But nfs4_show_open() and nfs4_show_lock() don't.
Maybe they should. I didn't double check, but I suspect they don't hold enough locks to ensure that the files don't get removed.

I think the only lock held is cl_lock, acquired in states_start.

We're starting here with an nfs4_stid that was found in the cl_stateids idr.

A struct nfs4_stid is freed by nfs4_put_stid(), which removes it from that idr under cl_lock before freeing the nfs4_stid and anything it points to.

Quoted 6 lines
Quoted 5 lines

I think that was the theory....

One possible problem is downgrades, like nfs4_stateid_downgrade.

I'll keep mulling it over, thanks.

Oops, I neglected this a while....

Quoted 9 lines

I had another look at code and maybe move_to_close_lru() is the problem. It can clear remove the files and clear sc_file without taking cl_lock. So some protection is needed against that.

I think that only applies to nfs4_show_open() - not show_lock etc. But I wonder it is might be best to include some extra protection for each different case, just in case some future code change allow sc_file to become NULL before the state is detached.

Quoted 6 lines
Quoted 3 lines

I'd feel more comforatable about nfs4_show_superblock() if it ignored nf_inode and just used nf_file - it is isn't NULL. It looks like it can never be set from non-NULL to NULL.

But then that means we've always got a reference on the inode, doesn't it? So I still don't understand the nf_inode comment.

My main problem with nf_inode is the comment

/*
* A representation of a file that has been opened by knfsd. These are hashed
* in the hashtable by inode pointer value. Note that this object doesn't
* hold a reference to the inode by itself, so the nf_inode pointer should
* never be dereferenced, only used for comparison.
*/

Quoted 3 lines

That comment is incompatible with the code in nfsd_file_mark_find_or_create() and with the code in nfs4_show_superblock().

Yeah, understood. I'm inclined to think the comment's just wrong, but not sure enough to be comfortable deleting it yet....

--b.

Quoted 74 lines
Quoted 58 lines

So maybe the NULL checks are mainly all we need.

Also it looks to me like ls_file lasts as long as the layout stateid, so maybe it's OK.

--b.

commit 4eef57aa4fc0

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

fs/nfsd/nfs4state.c +14 -0
index ab5c8857ae5a..08b8376c74d7 100644
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+
 static atomic_long_t num_delegations;
 unsigned long max_delegations;
 
@@ -2444,6 +2454,8 @@ static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2481,6 +2493,8 @@ static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);

fs/nfsd/nfs4state.c static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st) · line 2535

Quoted 10 lines
fs/nfsd/nfs4state.c +3 -1
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);

You'll need to add nfsd_file_put(file) toward the end of this function. Otherwise, I think this patch is a step in the right direction.

Thanks,
NeilBrown

#6 lore raw
In reply to J. Bruce Fields
Quoted 27 lines · J. Bruce Fields

commit 4eef57aa4fc0

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

fs/nfsd/nfs4state.c +10 -0
index ab5c8857ae5a..08b8376c74d7 100644
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+

...

fs/nfsd/nfs4state.c static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st) · line 2535

Quoted 11 lines
fs/nfsd/nfs4state.c +3 -1
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);

Oops, I added a "get" without a corresponding "put".

--b.

commit 8d2edfe45286

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

We hold the cl_lock here, and that's enough to keep stateid's from going away, but it's not enough to prevent the files they point to from going away. Take fi_lock and a reference and check for NULL, as we do in other code.

Fixes 78599c42ae3c ("nfsd4: add file to display list of client's opens")
fs/nfsd/nfs4state.c +18 -1
index ab5c8857ae5a..9d45117c8c18 100644
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+
 static atomic_long_t num_delegations;
 unsigned long max_delegations;
 
@@ -2444,6 +2454,8 @@ static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2481,6 +2493,8 @@ static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2529,6 +2545,7 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 	seq_printf(s, ", ");
 	nfs4_show_fname(s, file);
 	seq_printf(s, " }\n");
+	nfsd_file_put(file);
 
 	return 0;
 }
author
#7 lore raw
Quoted 65 lines · J. Bruce Fields
Quoted 27 lines · J. Bruce Fields

commit 4eef57aa4fc0

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

fs/nfsd/nfs4state.c +10 -0
index ab5c8857ae5a..08b8376c74d7 100644
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+

...

fs/nfsd/nfs4state.c static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st) · line 2535

Quoted 11 lines
fs/nfsd/nfs4state.c +3 -1
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);

Oops, I added a "get" without a corresponding "put".

--b.

commit 8d2edfe45286

Author: J. Bruce Fields <[email protected]>
Date:   Wed Jul 15 13:31:36 2020 -0400

nfsd4: fix NULL dereference in nfsd/clients display code

We hold the cl_lock here, and that's enough to keep stateid's from going away, but it's not enough to prevent the files they point to from going away. Take fi_lock and a reference and check for NULL, as we do in other code.

Fixes 78599c42ae3c ("nfsd4: add file to display list of client's opens")
fs/nfsd/nfs4state.c +0 -0
index ab5c8857ae5a..9d45117c8c18 100644

fs/nfsd/nfs4state.c find_any_file(struct nfs4_file *f) · line 515

Quoted 9 lines
fs/nfsd/nfs4state.c +6 -0
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);

A test on f->fi_deleg_file being non-NULL would make this look safer. It would also make the subsequent test on the return value appear sane.

Thanks,
NeilBrown

Quoted 44 lines
fs/nfsd/nfs4state.c +12 -1
+	spin_unlock(&f->fi_lock);
+	return ret;
+}
+
 static atomic_long_t num_delegations;
 unsigned long max_delegations;
 
@@ -2444,6 +2454,8 @@ static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2481,6 +2493,8 @@ static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
 	oo = ols->st_stateowner;
 	nf = st->sc_file;
 	file = find_any_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2513,7 +2527,9 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 
 	ds = delegstateid(st);
 	nf = st->sc_file;
-	file = nf->fi_deleg_file;
+	file = find_deleg_file(nf);
+	if (!file)
+		return 0;
 
 	seq_printf(s, "- ");
 	nfs4_show_stateid(s, &st->sc_stateid);
@@ -2529,6 +2545,7 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 	seq_printf(s, ", ");
 	nfs4_show_fname(s, file);
 	seq_printf(s, " }\n");
+	nfsd_file_put(file);
 
 	return 0;
 }
#8 lore raw
Quoted 7 lines · NeilBrown
Quoted 6 lines · J. Bruce Fields
fs/nfsd/nfs4state.c +0 -0
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
Quoted 9 lines

fs/nfsd/nfs4state.c find_any_file(struct nfs4_file *f) · line 515

Quoted 6 lines
fs/nfsd/nfs4state.c +6 -0
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);

A test on f->fi_deleg_file being non-NULL would make this look safer. It would also make the subsequent test on the return value appear sane.

Yes, thanks!-b.

fs/nfsd/nfs4state.c +3 -2
index c2a2e56c896d..6e8811e7c134 100644
@@ -509,10 +509,11 @@ find_any_file(struct nfs4_file *f)
 
 static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
 {
-	struct nfsd_file *ret;
+	struct nfsd_file *ret = NULL;
 
 	spin_lock(&f->fi_lock);
-	ret = nfsd_file_get(f->fi_deleg_file);
+	if (f->fi_deleg_file)
+		ret = nfsd_file_get(f->fi_deleg_file);
 	spin_unlock(&f->fi_lock);
 	return ret;
 }
author Reviewed-by
#9 lore raw
Quoted 28 lines · J. Bruce Fields
Quoted 7 lines · NeilBrown
Quoted 6 lines · J. Bruce Fields
fs/nfsd/nfs4state.c +0 -0
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
Quoted 9 lines

fs/nfsd/nfs4state.c find_any_file(struct nfs4_file *f) · line 515

Quoted 6 lines
fs/nfsd/nfs4state.c +6 -0
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);

A test on f->fi_deleg_file being non-NULL would make this look safer. It would also make the subsequent test on the return value appear sane.

Yes, thanks!-b.

fs/nfsd/nfs4state.c +0 -0
index c2a2e56c896d..6e8811e7c134 100644
@@ -509,10 +509,11 @@ find_any_file(struct nfs4_file *f)
 
 static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
 {

fs/nfsd/nfs4state.c find_any_file(struct nfs4_file *f) · line 519

Quoted 10 lines
fs/nfsd/nfs4state.c +3 -2
-	struct nfsd_file *ret;
+	struct nfsd_file *ret = NULL;
 
 	spin_lock(&f->fi_lock);
-	ret = nfsd_file_get(f->fi_deleg_file);
+	if (f->fi_deleg_file)
+		ret = nfsd_file_get(f->fi_deleg_file);
 	spin_unlock(&f->fi_lock);
 	return ret;
 }

for the whole patch.
Thanks,
NeilBrown

#10 lore raw
Quoted 43 lines · NeilBrown
Quoted 28 lines · J. Bruce Fields
Quoted 7 lines · NeilBrown
Quoted 6 lines · J. Bruce Fields
fs/nfsd/nfs4state.c +0 -0
@@ -507,6 +507,16 @@ find_any_file(struct nfs4_file *f)
 	return ret;
 }
 
Quoted 9 lines

fs/nfsd/nfs4state.c find_any_file(struct nfs4_file *f) · line 515

Quoted 6 lines
fs/nfsd/nfs4state.c +6 -0
+static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
+{
+	struct nfsd_file *ret;
+
+	spin_lock(&f->fi_lock);
+	ret = nfsd_file_get(f->fi_deleg_file);

A test on f->fi_deleg_file being non-NULL would make this look safer. It would also make the subsequent test on the return value appear sane.

Yes, thanks!-b.

fs/nfsd/nfs4state.c +0 -0
index c2a2e56c896d..6e8811e7c134 100644
@@ -509,10 +509,11 @@ find_any_file(struct nfs4_file *f)
 
 static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
 {

fs/nfsd/nfs4state.c find_any_file(struct nfs4_file *f) · line 519

Quoted 10 lines
fs/nfsd/nfs4state.c +3 -2
-	struct nfsd_file *ret;
+	struct nfsd_file *ret = NULL;
 
 	spin_lock(&f->fi_lock);
-	ret = nfsd_file_get(f->fi_deleg_file);
+	if (f->fi_deleg_file)
+		ret = nfsd_file_get(f->fi_deleg_file);
 	spin_unlock(&f->fi_lock);
 	return ret;
 }
Quoted 1 line

for the whole patch.

Thanks!--b.