From: Bob Peterson <rpeterso@redhat.com>
To: cluster-devel.redhat.com
Subject: [Cluster-devel] [PATCH 1/4] GFS2: Introduce helper for clearing gl_object
Date: Tue, 18 Jul 2017 13:23:02 -0500 [thread overview]
Message-ID: <20170718182305.27798-2-rpeterso@redhat.com> (raw)
In-Reply-To: <20170718182305.27798-1-rpeterso@redhat.com>
This patch introduces a new helper function in glock.h that
clears gl_object, with an added integrity check. An additional
integrity check has been added to glock_set_object, plus comments.
This is step 1 in a series to ensure gl_object integrity.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
fs/gfs2/glock.h | 34 ++++++++++++++++++++++++++++++++++
fs/gfs2/inode.c | 4 ++--
fs/gfs2/super.c | 4 ++--
3 files changed, 38 insertions(+), 4 deletions(-)
diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h
index 9ad4a6ac6c84..526d2123f758 100644
--- a/fs/gfs2/glock.h
+++ b/fs/gfs2/glock.h
@@ -13,6 +13,7 @@
#include <linux/sched.h>
#include <linux/parser.h>
#include "incore.h"
+#include "util.h"
/* Options for hostdata parser */
@@ -257,11 +258,44 @@ static inline bool gfs2_holder_initialized(struct gfs2_holder *gh)
return gh->gh_gl;
}
+/**
+ * glock_set_object - set the gl_object field of a glock
+ * @gl: the glock
+ * @object: the object
+ */
static inline void glock_set_object(struct gfs2_glock *gl, void *object)
{
spin_lock(&gl->gl_lockref.lock);
+ if (gfs2_assert_warn(gl->gl_name.ln_sbd, gl->gl_object == NULL))
+ gfs2_dump_glock(NULL, gl);
gl->gl_object = object;
spin_unlock(&gl->gl_lockref.lock);
}
+/**
+ * glock_clear_object - clear the gl_object field of a glock
+ * @gl: the glock
+ * @object: the object
+ *
+ * I'd love to similarly add this:
+ * else if (gfs2_assert_warn(gl->gl_sbd, gl->gl_object == object))
+ * gfs2_dump_glock(NULL, gl);
+ * Unfortunately, that's not possible because as soon as gfs2_delete_inode
+ * frees the block in the rgrp, another process can reassign it for an I_NEW
+ * inode in gfs2_create_inode because that calls new_inode, not gfs2_iget.
+ * That means gfs2_delete_inode may subsequently try to call this function
+ * for a glock that's already pointing to a brand new inode. If we clear the
+ * new inode's gl_object, we'll introduce metadata corruption. Function
+ * gfs2_delete_inode calls clear_inode which calls gfs2_clear_inode which also
+ * tries to clear gl_object, so it's more than just gfs2_delete_inode.
+ *
+ */
+static inline void glock_clear_object(struct gfs2_glock *gl, void *object)
+{
+ spin_lock(&gl->gl_lockref.lock);
+ if (gl->gl_object == object)
+ gl->gl_object = NULL;
+ spin_unlock(&gl->gl_lockref.lock);
+}
+
#endif /* __GLOCK_DOT_H__ */
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index acca501f8110..608e4bf60754 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -202,14 +202,14 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
fail_refresh:
ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
- glock_set_object(ip->i_iopen_gh.gh_gl, NULL);
+ glock_clear_object(ip->i_iopen_gh.gh_gl, ip);
gfs2_glock_dq_uninit(&ip->i_iopen_gh);
fail_put:
if (io_gl)
gfs2_glock_put(io_gl);
if (gfs2_holder_initialized(&i_gh))
gfs2_glock_dq_uninit(&i_gh);
- glock_set_object(ip->i_gl, NULL);
+ glock_clear_object(ip->i_gl, ip);
fail:
iget_failed(inode);
return ERR_PTR(error);
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index fdedec379b78..5fdc54158ff6 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1640,13 +1640,13 @@ static void gfs2_evict_inode(struct inode *inode)
gfs2_ordered_del_inode(ip);
clear_inode(inode);
gfs2_dir_hash_inval(ip);
- glock_set_object(ip->i_gl, NULL);
+ glock_clear_object(ip->i_gl, ip);
wait_on_bit_io(&ip->i_flags, GIF_GLOP_PENDING, TASK_UNINTERRUPTIBLE);
gfs2_glock_add_to_lru(ip->i_gl);
gfs2_glock_put(ip->i_gl);
ip->i_gl = NULL;
if (gfs2_holder_initialized(&ip->i_iopen_gh)) {
- glock_set_object(ip->i_iopen_gh.gh_gl, NULL);
+ glock_clear_object(ip->i_iopen_gh.gh_gl, ip);
ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
gfs2_glock_dq_uninit(&ip->i_iopen_gh);
}
--
2.13.3
next prev parent reply other threads:[~2017-07-18 18:23 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-07-18 18:23 [Cluster-devel] [PATCH 0/4] GFS2: Enforce gl_object integrity Bob Peterson
2017-07-18 18:23 ` Bob Peterson [this message]
2017-08-30 11:15 ` [Cluster-devel] [PATCH 1/4] GFS2: Introduce helper for clearing gl_object Andreas Gruenbacher
2017-08-30 13:18 ` Bob Peterson
2017-07-18 18:23 ` [Cluster-devel] [PATCH 2/4] GFS2: Set gl_object in inode lookup only after block type check Bob Peterson
2017-07-18 18:53 ` Bob Peterson
2017-07-18 18:23 ` [Cluster-devel] [PATCH 3/4] GFS2: Clear gl_object if gfs2_create_inode fails Bob Peterson
2017-07-18 18:23 ` [Cluster-devel] [PATCH 4/4] GFS2: Clear gl_object when deleting an inode in gfs2_delete_inode Bob Peterson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170718182305.27798-2-rpeterso@redhat.com \
--to=rpeterso@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).