From: Mike Christie <michael.christie@oracle.com>
To: lduncan@suse.com, cleech@redhat.com, njavali@marvell.com,
mrangankar@marvell.com, GR-QLogic-Storage-Upstream@marvell.com,
martin.petersen@oracle.com, linux-scsi@vger.kernel.org,
jejb@linux.ibm.com
Cc: Mike Christie <michael.christie@oracle.com>
Subject: [PATCH v2 13/28] scsi: iscsi: Fix conn use after free during resets
Date: Tue, 25 May 2021 13:18:06 -0500 [thread overview]
Message-ID: <20210525181821.7617-14-michael.christie@oracle.com> (raw)
In-Reply-To: <20210525181821.7617-1-michael.christie@oracle.com>
If we haven't done a unbind target call we can race where
iscsi_conn_teardown wakes up the EH thread and then frees the conn while
those threads are still accessing the conn ehwait.
We can only do one TMF per session so this just moves the TMF fields from
the conn to the session. We can then rely on the
iscsi_session_teardown->iscsi_remove_session->__iscsi_unbind_session call
to remove the target and it's devices, and know after that point there is
no device or scsi-ml callout trying to access the session.
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
---
drivers/scsi/libiscsi.c | 115 +++++++++++++++++++---------------------
include/scsi/libiscsi.h | 11 ++--
2 files changed, 60 insertions(+), 66 deletions(-)
diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c
index b7445d9e99d6..94abb093098d 100644
--- a/drivers/scsi/libiscsi.c
+++ b/drivers/scsi/libiscsi.c
@@ -230,11 +230,11 @@ static int iscsi_prep_ecdb_ahs(struct iscsi_task *task)
*/
static int iscsi_check_tmf_restrictions(struct iscsi_task *task, int opcode)
{
- struct iscsi_conn *conn = task->conn;
- struct iscsi_tm *tmf = &conn->tmhdr;
+ struct iscsi_session *session = task->conn->session;
+ struct iscsi_tm *tmf = &session->tmhdr;
u64 hdr_lun;
- if (conn->tmf_state == TMF_INITIAL)
+ if (session->tmf_state == TMF_INITIAL)
return 0;
if ((tmf->opcode & ISCSI_OPCODE_MASK) != ISCSI_OP_SCSI_TMFUNC)
@@ -254,24 +254,19 @@ static int iscsi_check_tmf_restrictions(struct iscsi_task *task, int opcode)
* Fail all SCSI cmd PDUs
*/
if (opcode != ISCSI_OP_SCSI_DATA_OUT) {
- iscsi_conn_printk(KERN_INFO, conn,
- "task [op %x itt "
- "0x%x/0x%x] "
- "rejected.\n",
- opcode, task->itt,
- task->hdr_itt);
+ iscsi_session_printk(KERN_INFO, session,
+ "task [op %x itt 0x%x/0x%x] rejected.\n",
+ opcode, task->itt, task->hdr_itt);
return -EACCES;
}
/*
* And also all data-out PDUs in response to R2T
* if fast_abort is set.
*/
- if (conn->session->fast_abort) {
- iscsi_conn_printk(KERN_INFO, conn,
- "task [op %x itt "
- "0x%x/0x%x] fast abort.\n",
- opcode, task->itt,
- task->hdr_itt);
+ if (session->fast_abort) {
+ iscsi_session_printk(KERN_INFO, session,
+ "task [op %x itt 0x%x/0x%x] fast abort.\n",
+ opcode, task->itt, task->hdr_itt);
return -EACCES;
}
break;
@@ -284,7 +279,7 @@ static int iscsi_check_tmf_restrictions(struct iscsi_task *task, int opcode)
*/
if (opcode == ISCSI_OP_SCSI_DATA_OUT &&
task->hdr_itt == tmf->rtt) {
- ISCSI_DBG_SESSION(conn->session,
+ ISCSI_DBG_SESSION(session,
"Preventing task %x/%x from sending "
"data-out due to abort task in "
"progress\n", task->itt,
@@ -936,20 +931,21 @@ iscsi_data_in_rsp(struct iscsi_conn *conn, struct iscsi_hdr *hdr,
static void iscsi_tmf_rsp(struct iscsi_conn *conn, struct iscsi_hdr *hdr)
{
struct iscsi_tm_rsp *tmf = (struct iscsi_tm_rsp *)hdr;
+ struct iscsi_session *session = conn->session;
conn->exp_statsn = be32_to_cpu(hdr->statsn) + 1;
conn->tmfrsp_pdus_cnt++;
- if (conn->tmf_state != TMF_QUEUED)
+ if (session->tmf_state != TMF_QUEUED)
return;
if (tmf->response == ISCSI_TMF_RSP_COMPLETE)
- conn->tmf_state = TMF_SUCCESS;
+ session->tmf_state = TMF_SUCCESS;
else if (tmf->response == ISCSI_TMF_RSP_NO_TASK)
- conn->tmf_state = TMF_NOT_FOUND;
+ session->tmf_state = TMF_NOT_FOUND;
else
- conn->tmf_state = TMF_FAILED;
- wake_up(&conn->ehwait);
+ session->tmf_state = TMF_FAILED;
+ wake_up(&session->ehwait);
}
static int iscsi_send_nopout(struct iscsi_conn *conn, struct iscsi_nopin *rhdr)
@@ -1826,15 +1822,14 @@ EXPORT_SYMBOL_GPL(iscsi_target_alloc);
static void iscsi_tmf_timedout(struct timer_list *t)
{
- struct iscsi_conn *conn = from_timer(conn, t, tmf_timer);
- struct iscsi_session *session = conn->session;
+ struct iscsi_session *session = from_timer(session, t, tmf_timer);
spin_lock(&session->frwd_lock);
- if (conn->tmf_state == TMF_QUEUED) {
- conn->tmf_state = TMF_TIMEDOUT;
+ if (session->tmf_state == TMF_QUEUED) {
+ session->tmf_state = TMF_TIMEDOUT;
ISCSI_DBG_EH(session, "tmf timedout\n");
/* unblock eh_abort() */
- wake_up(&conn->ehwait);
+ wake_up(&session->ehwait);
}
spin_unlock(&session->frwd_lock);
}
@@ -1857,8 +1852,8 @@ static int iscsi_exec_task_mgmt_fn(struct iscsi_conn *conn,
return -EPERM;
}
conn->tmfcmd_pdus_cnt++;
- conn->tmf_timer.expires = timeout * HZ + jiffies;
- add_timer(&conn->tmf_timer);
+ session->tmf_timer.expires = timeout * HZ + jiffies;
+ add_timer(&session->tmf_timer);
ISCSI_DBG_EH(session, "tmf set timeout\n");
spin_unlock_bh(&session->frwd_lock);
@@ -1872,12 +1867,12 @@ static int iscsi_exec_task_mgmt_fn(struct iscsi_conn *conn,
* 3) session is terminated or restarted or userspace has
* given up on recovery
*/
- wait_event_interruptible(conn->ehwait, age != session->age ||
+ wait_event_interruptible(session->ehwait, age != session->age ||
session->state != ISCSI_STATE_LOGGED_IN ||
- conn->tmf_state != TMF_QUEUED);
+ session->tmf_state != TMF_QUEUED);
if (signal_pending(current))
flush_signals(current);
- del_timer_sync(&conn->tmf_timer);
+ del_timer_sync(&session->tmf_timer);
mutex_lock(&session->eh_mutex);
spin_lock_bh(&session->frwd_lock);
@@ -2308,17 +2303,17 @@ int iscsi_eh_abort(struct scsi_cmnd *sc)
}
/* only have one tmf outstanding at a time */
- if (conn->tmf_state != TMF_INITIAL)
+ if (session->tmf_state != TMF_INITIAL)
goto failed;
- conn->tmf_state = TMF_QUEUED;
+ session->tmf_state = TMF_QUEUED;
- hdr = &conn->tmhdr;
+ hdr = &session->tmhdr;
iscsi_prep_abort_task_pdu(task, hdr);
if (iscsi_exec_task_mgmt_fn(conn, hdr, age, session->abort_timeout))
goto failed;
- switch (conn->tmf_state) {
+ switch (session->tmf_state) {
case TMF_SUCCESS:
spin_unlock_bh(&session->frwd_lock);
/*
@@ -2333,7 +2328,7 @@ int iscsi_eh_abort(struct scsi_cmnd *sc)
*/
spin_lock_bh(&session->frwd_lock);
fail_scsi_task(task, DID_ABORT);
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
memset(hdr, 0, sizeof(*hdr));
spin_unlock_bh(&session->frwd_lock);
iscsi_start_tx(conn);
@@ -2344,7 +2339,7 @@ int iscsi_eh_abort(struct scsi_cmnd *sc)
goto failed_unlocked;
case TMF_NOT_FOUND:
if (!sc->SCp.ptr) {
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
memset(hdr, 0, sizeof(*hdr));
/* task completed before tmf abort response */
ISCSI_DBG_EH(session, "sc completed while abort in "
@@ -2353,7 +2348,7 @@ int iscsi_eh_abort(struct scsi_cmnd *sc)
}
fallthrough;
default:
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
goto failed;
}
@@ -2414,11 +2409,11 @@ int iscsi_eh_device_reset(struct scsi_cmnd *sc)
conn = session->leadconn;
/* only have one tmf outstanding at a time */
- if (conn->tmf_state != TMF_INITIAL)
+ if (session->tmf_state != TMF_INITIAL)
goto unlock;
- conn->tmf_state = TMF_QUEUED;
+ session->tmf_state = TMF_QUEUED;
- hdr = &conn->tmhdr;
+ hdr = &session->tmhdr;
iscsi_prep_lun_reset_pdu(sc, hdr);
if (iscsi_exec_task_mgmt_fn(conn, hdr, session->age,
@@ -2427,7 +2422,7 @@ int iscsi_eh_device_reset(struct scsi_cmnd *sc)
goto unlock;
}
- switch (conn->tmf_state) {
+ switch (session->tmf_state) {
case TMF_SUCCESS:
break;
case TMF_TIMEDOUT:
@@ -2435,7 +2430,7 @@ int iscsi_eh_device_reset(struct scsi_cmnd *sc)
iscsi_conn_failure(conn, ISCSI_ERR_SCSI_EH_SESSION_RST);
goto done;
default:
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
goto unlock;
}
@@ -2447,7 +2442,7 @@ int iscsi_eh_device_reset(struct scsi_cmnd *sc)
spin_lock_bh(&session->frwd_lock);
memset(hdr, 0, sizeof(*hdr));
fail_scsi_tasks(conn, sc->device->lun, DID_ERROR);
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
spin_unlock_bh(&session->frwd_lock);
iscsi_start_tx(conn);
@@ -2470,8 +2465,7 @@ void iscsi_session_recovery_timedout(struct iscsi_cls_session *cls_session)
spin_lock_bh(&session->frwd_lock);
if (session->state != ISCSI_STATE_LOGGED_IN) {
session->state = ISCSI_STATE_RECOVERY_FAILED;
- if (session->leadconn)
- wake_up(&session->leadconn->ehwait);
+ wake_up(&session->ehwait);
}
spin_unlock_bh(&session->frwd_lock);
}
@@ -2516,7 +2510,7 @@ int iscsi_eh_session_reset(struct scsi_cmnd *sc)
iscsi_put_conn(conn->cls_conn);
ISCSI_DBG_EH(session, "wait for relogin\n");
- wait_event_interruptible(conn->ehwait,
+ wait_event_interruptible(session->ehwait,
session->state == ISCSI_STATE_TERMINATE ||
session->state == ISCSI_STATE_LOGGED_IN ||
session->state == ISCSI_STATE_RECOVERY_FAILED);
@@ -2577,11 +2571,11 @@ static int iscsi_eh_target_reset(struct scsi_cmnd *sc)
conn = session->leadconn;
/* only have one tmf outstanding at a time */
- if (conn->tmf_state != TMF_INITIAL)
+ if (session->tmf_state != TMF_INITIAL)
goto unlock;
- conn->tmf_state = TMF_QUEUED;
+ session->tmf_state = TMF_QUEUED;
- hdr = &conn->tmhdr;
+ hdr = &session->tmhdr;
iscsi_prep_tgt_reset_pdu(sc, hdr);
if (iscsi_exec_task_mgmt_fn(conn, hdr, session->age,
@@ -2590,7 +2584,7 @@ static int iscsi_eh_target_reset(struct scsi_cmnd *sc)
goto unlock;
}
- switch (conn->tmf_state) {
+ switch (session->tmf_state) {
case TMF_SUCCESS:
break;
case TMF_TIMEDOUT:
@@ -2598,7 +2592,7 @@ static int iscsi_eh_target_reset(struct scsi_cmnd *sc)
iscsi_conn_failure(conn, ISCSI_ERR_SCSI_EH_SESSION_RST);
goto done;
default:
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
goto unlock;
}
@@ -2610,7 +2604,7 @@ static int iscsi_eh_target_reset(struct scsi_cmnd *sc)
spin_lock_bh(&session->frwd_lock);
memset(hdr, 0, sizeof(*hdr));
fail_scsi_tasks(conn, -1, DID_ERROR);
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
spin_unlock_bh(&session->frwd_lock);
iscsi_start_tx(conn);
@@ -2940,7 +2934,10 @@ iscsi_session_setup(struct iscsi_transport *iscsit, struct Scsi_Host *shost,
session->tt = iscsit;
session->dd_data = cls_session->dd_data + sizeof(*session);
+ session->tmf_state = TMF_INITIAL;
+ timer_setup(&session->tmf_timer, iscsi_tmf_timedout, 0);
mutex_init(&session->eh_mutex);
+
spin_lock_init(&session->frwd_lock);
spin_lock_init(&session->back_lock);
@@ -3044,7 +3041,6 @@ iscsi_conn_setup(struct iscsi_cls_session *cls_session, int dd_size,
conn->c_stage = ISCSI_CONN_INITIAL_STAGE;
conn->id = conn_idx;
conn->exp_statsn = 0;
- conn->tmf_state = TMF_INITIAL;
timer_setup(&conn->transport_timer, iscsi_check_transport_timeouts, 0);
@@ -3069,8 +3065,7 @@ iscsi_conn_setup(struct iscsi_cls_session *cls_session, int dd_size,
goto login_task_data_alloc_fail;
conn->login_task->data = conn->data = data;
- timer_setup(&conn->tmf_timer, iscsi_tmf_timedout, 0);
- init_waitqueue_head(&conn->ehwait);
+ init_waitqueue_head(&session->ehwait);
return cls_conn;
@@ -3105,7 +3100,7 @@ void iscsi_conn_teardown(struct iscsi_cls_conn *cls_conn)
* leading connection? then give up on recovery.
*/
session->state = ISCSI_STATE_TERMINATE;
- wake_up(&conn->ehwait);
+ wake_up(&session->ehwait);
}
spin_unlock_bh(&session->frwd_lock);
@@ -3180,7 +3175,7 @@ int iscsi_conn_start(struct iscsi_cls_conn *cls_conn)
* commands after successful recovery
*/
conn->stop_stage = 0;
- conn->tmf_state = TMF_INITIAL;
+ session->tmf_state = TMF_INITIAL;
session->age++;
if (session->age == 16)
session->age = 0;
@@ -3194,7 +3189,7 @@ int iscsi_conn_start(struct iscsi_cls_conn *cls_conn)
spin_unlock_bh(&session->frwd_lock);
iscsi_unblock_session(session->cls_session);
- wake_up(&conn->ehwait);
+ wake_up(&session->ehwait);
return 0;
}
EXPORT_SYMBOL_GPL(iscsi_conn_start);
@@ -3288,7 +3283,7 @@ void iscsi_conn_stop(struct iscsi_cls_conn *cls_conn, int flag)
spin_lock_bh(&session->frwd_lock);
fail_scsi_tasks(conn, -1, DID_TRANSPORT_DISRUPTED);
fail_mgmt_tasks(session, conn);
- memset(&conn->tmhdr, 0, sizeof(conn->tmhdr));
+ memset(&session->tmhdr, 0, sizeof(session->tmhdr));
spin_unlock_bh(&session->frwd_lock);
mutex_unlock(&session->eh_mutex);
}
diff --git a/include/scsi/libiscsi.h b/include/scsi/libiscsi.h
index 13d413a0b8b6..9d7908265afe 100644
--- a/include/scsi/libiscsi.h
+++ b/include/scsi/libiscsi.h
@@ -202,12 +202,6 @@ struct iscsi_conn {
unsigned long suspend_tx; /* suspend Tx */
unsigned long suspend_rx; /* suspend Rx */
- /* abort */
- wait_queue_head_t ehwait; /* used in eh_abort() */
- struct iscsi_tm tmhdr;
- struct timer_list tmf_timer;
- int tmf_state; /* see TMF_INITIAL, etc.*/
-
/* negotiated params */
unsigned max_recv_dlength; /* initiator_max_recv_dsl*/
unsigned max_xmit_dlength; /* target_max_recv_dsl */
@@ -277,6 +271,11 @@ struct iscsi_session {
* and recv lock.
*/
struct mutex eh_mutex;
+ /* abort */
+ wait_queue_head_t ehwait; /* used in eh_abort() */
+ struct iscsi_tm tmhdr;
+ struct timer_list tmf_timer;
+ int tmf_state; /* see TMF_INITIAL, etc.*/
/* iSCSI session-wide sequencing */
uint32_t cmdsn;
--
2.25.1
next prev parent reply other threads:[~2021-05-25 18:19 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-25 18:17 iSCSI error handler fixes v2 Mike Christie
2021-05-25 18:17 ` [PATCH v2 01/28] scsi: iscsi: Add task completion helper Mike Christie
2021-06-08 3:05 ` Martin K. Petersen
2021-05-25 18:17 ` [PATCH v2 02/28] scsi: iscsi: Stop queueing during ep_disconnect Mike Christie
2021-05-25 18:17 ` [PATCH v2 03/28] scsi: iscsi: Drop suspend calls from ep_disconnect Mike Christie
2021-05-25 18:17 ` [PATCH v2 04/28] scsi: iscsi: Force immediate failure during shutdown Mike Christie
2021-05-25 18:17 ` [PATCH v2 05/28] scsi: iscsi: Use system_unbound_wq for destroy_work Mike Christie
2021-05-25 18:17 ` [PATCH v2 06/28] scsi: iscsi: Rel ref after iscsi_lookup_endpoint Mike Christie
2021-05-25 18:18 ` [PATCH v2 07/28] scsi: iscsi: Fix in-kernel conn failure handling Mike Christie
2021-05-25 18:18 ` [PATCH v2 08/28] scsi: iscsi_tcp: Set no linger Mike Christie
2021-05-25 18:18 ` [PATCH v2 09/28] scsi: iscsi_tcp: Start socket shutdown during conn stop Mike Christie
2021-05-25 18:18 ` [PATCH v2 10/28] scsi: iscsi: Add iscsi_cls_conn refcount helpers Mike Christie
2021-05-25 18:18 ` [PATCH v2 11/28] scsi: iscsi: Have abort handler get ref to conn Mike Christie
2021-05-25 18:18 ` [PATCH v2 12/28] scsi: iscsi: Get ref to conn during reset handling Mike Christie
2021-05-25 18:18 ` Mike Christie [this message]
2021-05-25 18:18 ` [PATCH v2 14/28] scsi: iscsi: Fix shost->max_id use Mike Christie
2021-05-25 18:18 ` [PATCH v2 15/28] scsi: iscsi: Fix completion check during abort races Mike Christie
2021-05-25 18:18 ` [PATCH v2 16/28] scsi: iscsi: Flush block work before unblock Mike Christie
2021-05-25 18:18 ` [PATCH v2 17/28] scsi: iscsi: Hold task ref during TMF timeout handling Mike Christie
2021-05-25 18:18 ` [PATCH v2 18/28] scsi: iscsi: Move pool freeing Mike Christie
2021-05-25 18:18 ` [PATCH v2 19/28] scsi: qedi: Fix null ref during abort handling Mike Christie
2021-05-25 18:18 ` [PATCH v2 20/28] scsi: qedi: Fix race during abort timeouts Mike Christie
2021-05-25 18:18 ` [PATCH v2 21/28] scsi: qedi: Fix use after free during abort cleanup Mike Christie
2021-05-25 18:18 ` [PATCH v2 22/28] scsi: qedi: Fix TMF tid allocation Mike Christie
2021-05-25 18:18 ` [PATCH v2 23/28] scsi: qedi: Use GFP_NOIO for TMF allocation Mike Christie
2021-05-25 18:18 ` [PATCH v2 24/28] scsi: qedi: Fix TMF session block/unblock use Mike Christie
2021-05-25 18:18 ` [PATCH v2 25/28] scsi: qedi: Fix cleanup " Mike Christie
2021-05-25 18:18 ` [PATCH v2 26/28] scsi: qedi: Pass send_iscsi_tmf task to abort Mike Christie
2021-05-25 18:18 ` [PATCH v2 27/28] scsi: qedi: Complete TMF works before disconnect Mike Christie
2021-05-25 18:18 ` [PATCH v2 28/28] scsi: qedi: Wake up if cmd_cleanup_req is set Mike Christie
2021-06-02 5:28 ` iSCSI error handler fixes v2 Martin K. Petersen
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=20210525181821.7617-14-michael.christie@oracle.com \
--to=michael.christie@oracle.com \
--cc=GR-QLogic-Storage-Upstream@marvell.com \
--cc=cleech@redhat.com \
--cc=jejb@linux.ibm.com \
--cc=lduncan@suse.com \
--cc=linux-scsi@vger.kernel.org \
--cc=martin.petersen@oracle.com \
--cc=mrangankar@marvell.com \
--cc=njavali@marvell.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).