From: Jeff Layton <jlayton@kernel.org>
To: xiubli@redhat.com
Cc: idryomov@gmail.com, pdonnell@redhat.com, ceph-devel@vger.kernel.org
Subject: Re: [PATCH 1/3] ceph: remove the capsnaps when removing the caps
Date: Mon, 23 Aug 2021 10:58:27 -0400 [thread overview]
Message-ID: <be0e28cf34dfcf65b1772e621557ecc276d46b95.camel@kernel.org> (raw)
In-Reply-To: <20210818080603.195722-2-xiubli@redhat.com>
On Wed, 2021-08-18 at 16:06 +0800, xiubli@redhat.com wrote:
> From: Xiubo Li <xiubli@redhat.com>
>
> The capsnaps will ihold the inodes when queuing to flush, so when
> force umounting it will close the sessions first and if the MDSes
> respond very fast and the session connections are closed just
> before killing the superblock, which will flush the msgr queue,
> then the flush capsnap callback won't ever be called, which will
> lead the memory leak bug for the ceph_inode_info.
>
> URL: https://tracker.ceph.com/issues/52295
> Signed-off-by: Xiubo Li <xiubli@redhat.com>
> ---
> fs/ceph/caps.c | 47 +++++++++++++++++++++++++++++---------------
> fs/ceph/mds_client.c | 23 +++++++++++++++++++++-
> fs/ceph/super.h | 3 +++
> 3 files changed, 56 insertions(+), 17 deletions(-)
>
> diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
> index e239f06babbc..7def99fbdca6 100644
> --- a/fs/ceph/caps.c
> +++ b/fs/ceph/caps.c
> @@ -3663,6 +3663,34 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
> iput(inode);
> }
>
> +/*
> + * Caller hold s_mutex and i_ceph_lock.
> + */
> +void ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
> + bool *wake_ci, bool *wake_mdsc)
> +{
> + struct ceph_inode_info *ci = ceph_inode(inode);
> + struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
> + bool ret;
> +
> + dout("removing capsnap %p, inode %p ci %p\n", capsnap, inode, ci);
> +
> + WARN_ON(capsnap->dirty_pages || capsnap->writing);
Can we make this a WARN_ON_ONCE too?
> + list_del(&capsnap->ci_item);
> + ret = __detach_cap_flush_from_ci(ci, &capsnap->cap_flush);
> + if (wake_ci)
> + *wake_ci = ret;
> +
> + spin_lock(&mdsc->cap_dirty_lock);
> + if (list_empty(&ci->i_cap_flush_list))
> + list_del_init(&ci->i_flushing_item);
> +
> + ret = __detach_cap_flush_from_mdsc(mdsc, &capsnap->cap_flush);
> + if (wake_mdsc)
> + *wake_mdsc = ret;
> + spin_unlock(&mdsc->cap_dirty_lock);
> +}
> +
> /*
> * Handle FLUSHSNAP_ACK. MDS has flushed snap data to disk and we can
> * throw away our cap_snap.
> @@ -3700,23 +3728,10 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
> capsnap, capsnap->follows);
> }
> }
> - if (flushed) {
> - WARN_ON(capsnap->dirty_pages || capsnap->writing);
> - dout(" removing %p cap_snap %p follows %lld\n",
> - inode, capsnap, follows);
> - list_del(&capsnap->ci_item);
> - wake_ci |= __detach_cap_flush_from_ci(ci, &capsnap->cap_flush);
> -
> - spin_lock(&mdsc->cap_dirty_lock);
> -
> - if (list_empty(&ci->i_cap_flush_list))
> - list_del_init(&ci->i_flushing_item);
> -
> - wake_mdsc |= __detach_cap_flush_from_mdsc(mdsc,
> - &capsnap->cap_flush);
> - spin_unlock(&mdsc->cap_dirty_lock);
> - }
> + if (flushed)
> + ceph_remove_capsnap(inode, capsnap, &wake_ci, &wake_mdsc);
> spin_unlock(&ci->i_ceph_lock);
> +
> if (flushed) {
> ceph_put_snap_context(capsnap->context);
> ceph_put_cap_snap(capsnap);
> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> index fa4c0fe294c1..a632e1c7cef2 100644
> --- a/fs/ceph/mds_client.c
> +++ b/fs/ceph/mds_client.c
> @@ -1604,10 +1604,30 @@ int ceph_iterate_session_caps(struct ceph_mds_session *session,
> return ret;
> }
>
> +static void remove_capsnaps(struct ceph_mds_client *mdsc, struct inode *inode)
> +{
> + struct ceph_inode_info *ci = ceph_inode(inode);
> + struct ceph_cap_snap *capsnap;
> +
> + dout("removing capsnaps, ci is %p, inode is %p\n", ci, inode);
> +
> + while (!list_empty(&ci->i_cap_snaps)) {
> + capsnap = list_first_entry(&ci->i_cap_snaps,
> + struct ceph_cap_snap, ci_item);
> + ceph_remove_capsnap(inode, capsnap, NULL, NULL);
> + ceph_put_snap_context(capsnap->context);
> + ceph_put_cap_snap(capsnap);
> + iput(inode);
> + }
> + wake_up_all(&ci->i_cap_wq);
> + wake_up_all(&mdsc->cap_flushing_wq);
> +}
> +
> static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
> void *arg)
> {
> struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
> + struct ceph_mds_client *mdsc = fsc->mdsc;
> struct ceph_inode_info *ci = ceph_inode(inode);
> LIST_HEAD(to_remove);
> bool dirty_dropped = false;
> @@ -1619,7 +1639,6 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
> __ceph_remove_cap(cap, false);
> if (!ci->i_auth_cap) {
> struct ceph_cap_flush *cf;
> - struct ceph_mds_client *mdsc = fsc->mdsc;
>
> if (READ_ONCE(fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN) {
> if (inode->i_data.nrpages > 0)
> @@ -1684,6 +1703,8 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
> ci->i_prealloc_cap_flush = NULL;
> }
> }
> + if (!list_empty(&ci->i_cap_snaps))
> + remove_capsnaps(mdsc, inode);
> spin_unlock(&ci->i_ceph_lock);
> while (!list_empty(&to_remove)) {
> struct ceph_cap_flush *cf;
> diff --git a/fs/ceph/super.h b/fs/ceph/super.h
> index 0bc36cf4c683..51ec17d12b26 100644
> --- a/fs/ceph/super.h
> +++ b/fs/ceph/super.h
> @@ -1168,6 +1168,9 @@ extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
> int had);
> extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
> struct ceph_snap_context *snapc);
> +extern void ceph_remove_capsnap(struct inode *inode,
> + struct ceph_cap_snap *capsnap,
> + bool *wake_ci, bool *wake_mdsc);
> extern void ceph_flush_snaps(struct ceph_inode_info *ci,
> struct ceph_mds_session **psession);
> extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
--
Jeff Layton <jlayton@kernel.org>
next prev parent reply other threads:[~2021-08-23 14:58 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-18 8:06 [PATCH 0/3] ceph: remove the capsnaps when removing the caps xiubli
2021-08-18 8:06 ` [PATCH 1/3] " xiubli
2021-08-23 13:47 ` Jeff Layton
2021-08-24 1:04 ` Xiubo Li
2021-08-23 14:58 ` Jeff Layton [this message]
2021-08-24 1:05 ` Xiubo Li
2021-08-18 8:06 ` [PATCH 2/3] ceph: don't WARN if we're force umounting xiubli
2021-08-23 13:49 ` Jeff Layton
2021-08-24 1:07 ` Xiubo Li
2021-08-18 8:06 ` [PATCH 3/3] ceph: don't WARN if we're iterate removing the session caps xiubli
2021-08-23 13:59 ` Jeff Layton
2021-08-24 1:43 ` Xiubo Li
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=be0e28cf34dfcf65b1772e621557ecc276d46b95.camel@kernel.org \
--to=jlayton@kernel.org \
--cc=ceph-devel@vger.kernel.org \
--cc=idryomov@gmail.com \
--cc=pdonnell@redhat.com \
--cc=xiubli@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).