From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id E642D8BE0 for ; Tue, 28 Mar 2023 14:49:20 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 68230C433D2; Tue, 28 Mar 2023 14:49:20 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1680014960; bh=FMlprNX8hPo/gJqoM7YfxZkBpN0TAI3lenpNYjq9jNw=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=PoE5NW3X65Ac0XErVG3UKxcj8sVqmWRCgSFg+wTwv0eK2HmvSIOXQf4c01mOTjNc2 5tWTd9VRXruo/JAT6ulq5onL0H//BLKqnDYZ7OuB2MpAFTDqEfniGn5e4vDWGmTCfI TOSUpLWitjxk77CmSun0EgB9lSwgCpHsKYQVPZj0= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Shyam Prasad N , "Paulo Alcantara (SUSE)" , Steve French Subject: [PATCH 6.2 116/240] smb3: fix unusable share after force unmount failure Date: Tue, 28 Mar 2023 16:41:19 +0200 Message-Id: <20230328142624.634920893@linuxfoundation.org> X-Mailer: git-send-email 2.40.0 In-Reply-To: <20230328142619.643313678@linuxfoundation.org> References: <20230328142619.643313678@linuxfoundation.org> User-Agent: quilt/0.67 Precedence: bulk X-Mailing-List: patches@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Steve French commit 491eafce1a51c457701351a4bf40733799745314 upstream. If user does forced unmount ("umount -f") while files are still open on the share (as was seen in a Kubernetes example running on SMB3.1.1 mount) then we were marking the share as "TID_EXITING" in umount_begin() which caused all subsequent operations (except write) to fail ... but unfortunately when umount_begin() is called we do not know yet that there are open files or active references on the share that would prevent unmount from succeeding. Kubernetes had example when they were doing umount -f when files were open which caused the share to become unusable until the files were closed (and the umount retried). Fix this so that TID_EXITING is not set until we are about to send the tree disconnect (not at the beginning of forced umounts in umount_begin) so that if "umount -f" fails (due to open files or references) the mount is still usable. Cc: stable@vger.kernel.org Reviewed-by: Shyam Prasad N Reviewed-by: Paulo Alcantara (SUSE) Signed-off-by: Steve French Signed-off-by: Greg Kroah-Hartman --- fs/cifs/cifsfs.c | 9 ++++++--- fs/cifs/cifssmb.c | 6 ++---- fs/cifs/connect.c | 1 + fs/cifs/smb2pdu.c | 8 ++------ 4 files changed, 11 insertions(+), 13 deletions(-) --- a/fs/cifs/cifsfs.c +++ b/fs/cifs/cifsfs.c @@ -730,13 +730,16 @@ static void cifs_umount_begin(struct sup spin_lock(&tcon->tc_lock); if ((tcon->tc_count > 1) || (tcon->status == TID_EXITING)) { /* we have other mounts to same share or we have - already tried to force umount this and woken up + already tried to umount this and woken up all waiting network requests, nothing to do */ spin_unlock(&tcon->tc_lock); spin_unlock(&cifs_tcp_ses_lock); return; - } else if (tcon->tc_count == 1) - tcon->status = TID_EXITING; + } + /* + * can not set tcon->status to TID_EXITING yet since we don't know if umount -f will + * fail later (e.g. due to open files). TID_EXITING will be set just before tdis req sent + */ spin_unlock(&tcon->tc_lock); spin_unlock(&cifs_tcp_ses_lock); --- a/fs/cifs/cifssmb.c +++ b/fs/cifs/cifssmb.c @@ -85,13 +85,11 @@ cifs_reconnect_tcon(struct cifs_tcon *tc /* * only tree disconnect, open, and write, (and ulogoff which does not - * have tcon) are allowed as we start force umount + * have tcon) are allowed as we start umount */ spin_lock(&tcon->tc_lock); if (tcon->status == TID_EXITING) { - if (smb_command != SMB_COM_WRITE_ANDX && - smb_command != SMB_COM_OPEN_ANDX && - smb_command != SMB_COM_TREE_DISCONNECT) { + if (smb_command != SMB_COM_TREE_DISCONNECT) { spin_unlock(&tcon->tc_lock); cifs_dbg(FYI, "can not send cmd %d while umounting\n", smb_command); --- a/fs/cifs/connect.c +++ b/fs/cifs/connect.c @@ -2363,6 +2363,7 @@ cifs_put_tcon(struct cifs_tcon *tcon) WARN_ON(tcon->tc_count < 0); list_del_init(&tcon->tcon_list); + tcon->status = TID_EXITING; spin_unlock(&tcon->tc_lock); spin_unlock(&cifs_tcp_ses_lock); --- a/fs/cifs/smb2pdu.c +++ b/fs/cifs/smb2pdu.c @@ -225,13 +225,9 @@ smb2_reconnect(__le16 smb2_command, stru spin_lock(&tcon->tc_lock); if (tcon->status == TID_EXITING) { /* - * only tree disconnect, open, and write, - * (and ulogoff which does not have tcon) - * are allowed as we start force umount. + * only tree disconnect allowed when disconnecting ... */ - if ((smb2_command != SMB2_WRITE) && - (smb2_command != SMB2_CREATE) && - (smb2_command != SMB2_TREE_DISCONNECT)) { + if (smb2_command != SMB2_TREE_DISCONNECT) { spin_unlock(&tcon->tc_lock); cifs_dbg(FYI, "can not send cmd %d while umounting\n", smb2_command);