From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
alan@lxorguk.ukuu.org.uk, Roland Dreier <roland@purestorage.com>,
Andy Grover <agrover@redhat.com>, Hannes Reinecke <hare@suse.de>,
Christoph Hellwig <hch@lst.de>,
Nicholas Bellinger <nab@linux-iscsi.org>
Subject: [ 09/57] iscsi-target: Fix missed wakeup race in TX thread
Date: Wed, 14 Nov 2012 20:11:17 -0800 [thread overview]
Message-ID: <20121115040933.785501464@linuxfoundation.org> (raw)
In-Reply-To: <20121115040933.223998671@linuxfoundation.org>
3.4-stable review patch. If anyone has any objections, please let me know.
------------------
From: Roland Dreier <roland@purestorage.com>
commit d5627acba9ae584cf4928af19f7ddf5f6837de32 upstream.
The sleeping code in iscsi_target_tx_thread() is susceptible to the classic
missed wakeup race:
- TX thread finishes handle_immediate_queue() and handle_response_queue(),
thinks both queues are empty.
- Another thread adds a queue entry and does wake_up_process(), which does
nothing because the TX thread is still awake.
- TX thread does schedule_timeout() and sleeps forever.
In practice this can kill an iSCSI connection if for example an initiator
does single-threaded writes and the target misses the wakeup window when
queueing an R2T; in this case the connection will be stuck until the
initiator loses patience and does some task management operation (or kills
the connection entirely).
Fix this by converting to wait_event_interruptible(), which does not
suffer from this sort of race.
Signed-off-by: Roland Dreier <roland@purestorage.com>
Cc: Andy Grover <agrover@redhat.com>
Cc: Hannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
drivers/target/iscsi/iscsi_target.c | 4 +++-
drivers/target/iscsi/iscsi_target_core.h | 1 +
drivers/target/iscsi/iscsi_target_login.c | 1 +
drivers/target/iscsi/iscsi_target_util.c | 22 ++++++++++++++++++++--
drivers/target/iscsi/iscsi_target_util.h | 1 +
5 files changed, 26 insertions(+), 3 deletions(-)
--- a/drivers/target/iscsi/iscsi_target.c
+++ b/drivers/target/iscsi/iscsi_target.c
@@ -3514,7 +3514,9 @@ restart:
*/
iscsit_thread_check_cpumask(conn, current, 1);
- schedule_timeout_interruptible(MAX_SCHEDULE_TIMEOUT);
+ wait_event_interruptible(conn->queues_wq,
+ !iscsit_conn_all_queues_empty(conn) ||
+ ts->status == ISCSI_THREAD_SET_RESET);
if ((ts->status == ISCSI_THREAD_SET_RESET) ||
signal_pending(current))
--- a/drivers/target/iscsi/iscsi_target_core.h
+++ b/drivers/target/iscsi/iscsi_target_core.h
@@ -491,6 +491,7 @@ struct iscsi_tmr_req {
};
struct iscsi_conn {
+ wait_queue_head_t queues_wq;
/* Authentication Successful for this connection */
u8 auth_complete;
/* State connection is currently in */
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -45,6 +45,7 @@ extern spinlock_t sess_idr_lock;
static int iscsi_login_init_conn(struct iscsi_conn *conn)
{
+ init_waitqueue_head(&conn->queues_wq);
INIT_LIST_HEAD(&conn->conn_list);
INIT_LIST_HEAD(&conn->conn_cmd_list);
INIT_LIST_HEAD(&conn->immed_queue_list);
--- a/drivers/target/iscsi/iscsi_target_util.c
+++ b/drivers/target/iscsi/iscsi_target_util.c
@@ -656,7 +656,7 @@ void iscsit_add_cmd_to_immediate_queue(
atomic_set(&conn->check_immediate_queue, 1);
spin_unlock_bh(&conn->immed_queue_lock);
- wake_up_process(conn->thread_set->tx_thread);
+ wake_up(&conn->queues_wq);
}
struct iscsi_queue_req *iscsit_get_cmd_from_immediate_queue(struct iscsi_conn *conn)
@@ -730,7 +730,7 @@ void iscsit_add_cmd_to_response_queue(
atomic_inc(&cmd->response_queue_count);
spin_unlock_bh(&conn->response_queue_lock);
- wake_up_process(conn->thread_set->tx_thread);
+ wake_up(&conn->queues_wq);
}
struct iscsi_queue_req *iscsit_get_cmd_from_response_queue(struct iscsi_conn *conn)
@@ -784,6 +784,24 @@ static void iscsit_remove_cmd_from_respo
}
}
+bool iscsit_conn_all_queues_empty(struct iscsi_conn *conn)
+{
+ bool empty;
+
+ spin_lock_bh(&conn->immed_queue_lock);
+ empty = list_empty(&conn->immed_queue_list);
+ spin_unlock_bh(&conn->immed_queue_lock);
+
+ if (!empty)
+ return empty;
+
+ spin_lock_bh(&conn->response_queue_lock);
+ empty = list_empty(&conn->response_queue_list);
+ spin_unlock_bh(&conn->response_queue_lock);
+
+ return empty;
+}
+
void iscsit_free_queue_reqs_for_conn(struct iscsi_conn *conn)
{
struct iscsi_queue_req *qr, *qr_tmp;
--- a/drivers/target/iscsi/iscsi_target_util.h
+++ b/drivers/target/iscsi/iscsi_target_util.h
@@ -28,6 +28,7 @@ extern struct iscsi_queue_req *iscsit_ge
extern void iscsit_add_cmd_to_response_queue(struct iscsi_cmd *, struct iscsi_conn *, u8);
extern struct iscsi_queue_req *iscsit_get_cmd_from_response_queue(struct iscsi_conn *);
extern void iscsit_remove_cmd_from_tx_queues(struct iscsi_cmd *, struct iscsi_conn *);
+extern bool iscsit_conn_all_queues_empty(struct iscsi_conn *);
extern void iscsit_free_queue_reqs_for_conn(struct iscsi_conn *);
extern void iscsit_release_cmd(struct iscsi_cmd *);
extern void iscsit_free_cmd(struct iscsi_cmd *);
next prev parent reply other threads:[~2012-11-15 4:26 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-11-15 4:11 [ 00/57] 3.4.19-stable review Greg Kroah-Hartman
2012-11-15 4:11 ` [ 01/57] xen/gntdev: dont leak memory from IOCTL_GNTDEV_MAP_GRANT_REF Greg Kroah-Hartman
2012-11-15 4:11 ` [ 02/57] xen/mmu: Use Xen specific TLB flush instead of the generic one Greg Kroah-Hartman
2012-11-15 4:11 ` [ 03/57] Input: tsc40 - remove wrong announcement of pressure support Greg Kroah-Hartman
2012-11-15 4:11 ` [ 04/57] ath9k: fix stale pointers potentially causing access to freed skbs Greg Kroah-Hartman
2012-11-15 4:11 ` [ 05/57] ath9k: Test for TID only in BlockAcks while checking tx status Greg Kroah-Hartman
2012-11-15 4:11 ` [ 06/57] rt2800: validate step value for temperature compensation Greg Kroah-Hartman
2012-11-15 4:11 ` [ 07/57] target: Dont return success from module_init() if setup fails Greg Kroah-Hartman
2012-11-15 4:11 ` [ 08/57] target: Avoid integer overflow in se_dev_align_max_sectors() Greg Kroah-Hartman
2012-11-15 4:11 ` Greg Kroah-Hartman [this message]
2012-11-15 4:11 ` [ 10/57] target: Fix incorrect usage of nested IRQ spinlocks in ABORT_TASK path Greg Kroah-Hartman
2012-11-15 4:11 ` [ 11/57] cfg80211: fix antenna gain handling Greg Kroah-Hartman
2012-11-15 4:11 ` [ 12/57] wireless: drop invalid mesh address extension frames Greg Kroah-Hartman
2012-11-15 4:11 ` [ 13/57] mac80211: use blacklist for duplicate IE check Greg Kroah-Hartman
2012-11-15 4:11 ` [ 14/57] mac80211: Only process mesh config header on frames that RA_MATCH Greg Kroah-Hartman
2012-11-15 4:11 ` [ 15/57] mac80211: dont inspect Sequence Control field on control frames Greg Kroah-Hartman
2012-11-15 4:11 ` [ 16/57] DRM/Radeon: Fix Load Detection on legacy primary DAC Greg Kroah-Hartman
2012-11-15 4:11 ` [ 17/57] drm/udl: fix stride issues scanning out stride != width*bpp Greg Kroah-Hartman
2012-11-15 4:11 ` [ 18/57] mac80211: check management frame header length Greg Kroah-Hartman
2012-11-15 4:11 ` [ 19/57] mac80211: verify that skb data is present Greg Kroah-Hartman
2012-11-15 4:11 ` [ 20/57] mac80211: make sure data is accessible in EAPOL check Greg Kroah-Hartman
2012-11-15 4:11 ` [ 21/57] mac80211: fix SSID copy on IBSS JOIN Greg Kroah-Hartman
2012-11-15 4:11 ` [ 22/57] nfsv3: Make v3 mounts fail with ETIMEDOUTs instead EIO on mountd timeouts Greg Kroah-Hartman
2012-11-15 4:11 ` [ 23/57] nfs: Show original device name verbatim in /proc/*/mount{s,info} Greg Kroah-Hartman
2012-11-15 4:11 ` [ 24/57] NFSv4: nfs4_locku_done must release the sequence id Greg Kroah-Hartman
2012-11-15 4:11 ` [ 25/57] NFSv4.1: We must release the sequence id when we fail to get a session slot Greg Kroah-Hartman
2012-11-15 4:11 ` [ 26/57] nfsd: add get_uint for u32s Greg Kroah-Hartman
2012-11-15 4:11 ` [ 27/57] NFS: fix bug in legacy DNS resolver Greg Kroah-Hartman
2012-11-15 4:11 ` [ 28/57] NFS: Fix Oopses in nfs_lookup_revalidate and nfs4_lookup_revalidate Greg Kroah-Hartman
2012-11-15 4:11 ` [ 29/57] drm: restore open_count if drm_setup fails Greg Kroah-Hartman
2012-11-15 4:11 ` [ 30/57] hwmon: (w83627ehf) Force initial bank selection Greg Kroah-Hartman
2012-11-15 4:11 ` [ 31/57] ALSA: PCM: Fix some races at disconnection Greg Kroah-Hartman
2012-11-15 4:11 ` [ 32/57] ALSA: usb-audio: Fix " Greg Kroah-Hartman
2012-11-15 4:11 ` [ 33/57] ALSA: usb-audio: Use rwsem for disconnect protection Greg Kroah-Hartman
2012-11-15 4:11 ` [ 34/57] ALSA: usb-audio: Fix races at disconnection in mixer_quirks.c Greg Kroah-Hartman
2012-11-15 4:11 ` [ 35/57] ALSA: Add a reference counter to card instance Greg Kroah-Hartman
2012-11-15 4:11 ` [ 36/57] ALSA: Avoid endless sleep after disconnect Greg Kroah-Hartman
2012-11-15 4:11 ` [ 37/57] sctp: fix call to SCTP_CMD_PROCESS_SACK in sctp_cmd_interpreter() Greg Kroah-Hartman
2012-11-15 4:11 ` [ 38/57] netlink: use kfree_rcu() in netlink_release() Greg Kroah-Hartman
2012-11-15 4:11 ` [ 39/57] tcp: fix FIONREAD/SIOCINQ Greg Kroah-Hartman
2012-11-15 4:11 ` [ 40/57] ipv6: Set default hoplimit as zero Greg Kroah-Hartman
2012-11-15 4:11 ` [ 41/57] net: usb: Fix memory leak on Tx data path Greg Kroah-Hartman
2012-11-15 4:11 ` [ 42/57] net: fix divide by zero in tcp algorithm illinois Greg Kroah-Hartman
2012-11-15 4:11 ` [ 43/57] drivers/net/ethernet/nxp/lpc_eth.c: Call mdiobus_unregister before mdiobus_free Greg Kroah-Hartman
2012-11-15 4:11 ` [ 44/57] l2tp: fix oops in l2tp_eth_create() error path Greg Kroah-Hartman
2012-11-15 4:11 ` [ 45/57] net: inet_diag -- Return error code if protocol handler is missed Greg Kroah-Hartman
2012-11-15 4:11 ` [ 46/57] af-packet: fix oops when socket is not present Greg Kroah-Hartman
2012-11-15 4:11 ` [ 47/57] ipv6: send unsolicited neighbour advertisements to all-nodes Greg Kroah-Hartman
2012-11-15 4:11 ` [ 48/57] futex: Handle futex_pi OWNER_DIED take over correctly Greg Kroah-Hartman
2012-11-15 4:11 ` [ 49/57] mmc: sdhci: fix NULL dereference in sdhci_request() tuning Greg Kroah-Hartman
2012-11-15 4:11 ` [ 50/57] drm/vmwgfx: Fix hibernation device reset Greg Kroah-Hartman
2012-11-15 4:11 ` [ 51/57] drm/vmwgfx: Fix a case where the code would BUG when trying to pin GMR memory Greg Kroah-Hartman
2012-11-15 4:12 ` [ 52/57] drm/radeon/cayman: add some missing regs to the VM reg checker Greg Kroah-Hartman
2012-11-15 4:12 ` [ 53/57] drm/radeon/si: " Greg Kroah-Hartman
2012-11-15 4:12 ` [ 54/57] drm/i915: fixup infoframe support for sdvo Greg Kroah-Hartman
2012-11-15 4:12 ` [ 55/57] drm/i915: clear the entire sdvo infoframe buffer Greg Kroah-Hartman
2012-11-15 4:12 ` [ 56/57] USB: mos7840: remove unused variable Greg Kroah-Hartman
2012-11-15 4:12 ` [ 57/57] xfs: fix reading of wrapped log data Greg Kroah-Hartman
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=20121115040933.785501464@linuxfoundation.org \
--to=gregkh@linuxfoundation.org \
--cc=agrover@redhat.com \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=hare@suse.de \
--cc=hch@lst.de \
--cc=linux-kernel@vger.kernel.org \
--cc=nab@linux-iscsi.org \
--cc=roland@purestorage.com \
--cc=stable@vger.kernel.org \
/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).