Ensure FCoE target interrupts work
Subsystems (3)
cover Ensure FCoE target interrupts work
From: Lee Duncan <[email protected]>
Commit 1a1975551943 "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock" changed locking for fnic/FCoE, but it did so by disabling interrupts where they weren't disabled before, and this caused FCoE targets to go offline. Reverting that patch fixed the issue.
But to handle the problem originally addressed by the commit, instead of modifying the locking, move the work to be done into a work queue.
Lee Duncan (1):
Revert "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"
Hannes Reinecke (1):
fnic: move fnic_fnic_flush_tx() to a work queue
drivers/scsi/fcoe/fcoe_ctlr.c | 20 ++++++++------------ drivers/scsi/fnic/fnic.h | 3 ++- drivers/scsi/fnic/fnic_fcs.c | 3 ++- drivers/scsi/fnic/fnic_main.c | 1 + drivers/scsi/fnic/fnic_scsi.c | 4 ++-- 5 files changed, 15 insertions(+), 16 deletions(-)
Signature · 1 line
2.43.0
PATCH 1/2 Revert "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"
From: Lee Duncan <[email protected]>
This reverts commit 1a1975551943f681772720f639ff42fbaa746212
This commit causes interrupts to be lost for FCoE devices, since it changed sping locks from "bh" to "irqsave".
Instead, a work queue should be used, and will be addressed in a separate patch.
--- drivers/scsi/fcoe/fcoe_ctlr.c | 20 ++++++++------------ 1 file changed, 8 insertions(+), 12 deletions(-)
drivers/scsi/fcoe/fcoe_ctlr.c +8 -12
@@ -319,17 +319,16 @@ static void fcoe_ctlr_announce(struct fcoe_ctlr *fip) {struct fcoe_fcf *sel;struct fcoe_fcf *fcf; -unsigned long flags;mutex_lock (&fip->ctlr_mutex); -spin_lock_irqsave (&fip->ctlr_lock, flags); +spin_lock_bh (&fip->ctlr_lock);kfree_skb (fip->flogi_req); fip->flogi_req =NULL ;list_for_each_entry (fcf, &fip->fcfs, list) fcf->flogi_sent =0 ; -spin_unlock_irqrestore (&fip->ctlr_lock, flags); +spin_unlock_bh (&fip->ctlr_lock); sel = fip->sel_fcf;if (sel &ðer_addr_equal (sel->fcf_mac, fip->dest_addr))
@@ -700,7 +699,6 @@ int fcoe_ctlr_els_send(struct fcoe_ctlr *fip, struct fc_lport *lport, {struct fc_frame *fp;struct fc_frame_header *fh; -unsigned long flags;u16 old_xid;u8 op;u8 mac[ETH_ALEN ];
@@ -734,11 +732,11 @@ int fcoe_ctlr_els_send(struct fcoe_ctlr *fip, struct fc_lport *lport, op =FIP_DT_FLOGI ;if (fip->mode ==FIP_MODE_VN2VN )break ; -spin_lock_irqsave (&fip->ctlr_lock, flags); +spin_lock_bh (&fip->ctlr_lock);kfree_skb (fip->flogi_req); fip->flogi_req = skb; fip->flogi_req_send =1 ; -spin_unlock_irqrestore (&fip->ctlr_lock, flags); +spin_unlock_bh (&fip->ctlr_lock);schedule_work (&fip->timer_work);return -EINPROGRESS ;case ELS_FDISC :
@@ -1707,11 +1705,10 @@ static int fcoe_ctlr_flogi_send_locked(struct fcoe_ctlr *fip)static int fcoe_ctlr_flogi_retry (struct fcoe_ctlr *fip) {struct fcoe_fcf *fcf; -unsigned long flags;int error;mutex_lock (&fip->ctlr_mutex); -spin_lock_irqsave (&fip->ctlr_lock, flags); +spin_lock_bh (&fip->ctlr_lock);LIBFCOE_FIP_DBG (fip,"re-sending FLOGI - reselect\n" ); fcf =fcoe_ctlr_select (fip);if (!fcf || fcf->flogi_sent) {
@@ -1722,7 +1719,7 @@ static int fcoe_ctlr_flogi_retry(struct fcoe_ctlr *fip)fcoe_ctlr_solicit (fip,NULL ); error =fcoe_ctlr_flogi_send_locked (fip); } -spin_unlock_irqrestore (&fip->ctlr_lock, flags); +spin_unlock_bh (&fip->ctlr_lock);mutex_unlock (&fip->ctlr_mutex);return error; }
@@ -1739,9 +1736,8 @@ static int fcoe_ctlr_flogi_retry(struct fcoe_ctlr *fip)static void fcoe_ctlr_flogi_send (struct fcoe_ctlr *fip) {struct fcoe_fcf *fcf; -unsigned long flags; -spin_lock_irqsave (&fip->ctlr_lock, flags); +spin_lock_bh (&fip->ctlr_lock); fcf = fip->sel_fcf;if (!fcf || !fip->flogi_req_send) goto unlock;
@@ -1768,7 +1764,7 @@ static void fcoe_ctlr_flogi_send(struct fcoe_ctlr *fip) }else /* XXX */ LIBFCOE_FIP_DBG (fip,"No FCF selected - defer send\n" ); unlock: -spin_unlock_irqrestore (&fip->ctlr_lock, flags); +spin_unlock_bh (&fip->ctlr_lock); }/**
Signature · 1 line
2.43.0
Re: [PATCH 1/2] Revert "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"
Quoted 7 lines · Lee Duncan
From: Lee Duncan <[email protected]>
This reverts commit
1a1975551943f681772720f639ff42fbaa746212This commit causes interrupts to be lost for FCoE devices, since it changed sping locks from "bh" to "irqsave".
Quoted 9 lines
Instead, a work queue should be used, and will be addressed in a separate patch.
Fixes 1a1975551943f681772720f639ff42fbaa746212Signed-off-by Lee Duncan <[email protected]>--- drivers/scsi/fcoe/fcoe_ctlr.c | 20 ++++++++------------ 1 file changed, 8 insertions(+), 12 deletions(-)
Cheers,
Hannes
Signature · 5 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), GF: Ivo Totev, Andrew McDonald, Werner Knoblich
PATCH 2/2 fnic: move fnic_fnic_flush_tx() to a work queue
In reply to Lee DuncanFrom: Hannes Reinecke <[email protected]>
Rather than call 'fnic_flush_tx()' from interrupt context we should be moving it onto a work queue to avoid any locking issues.
--- drivers/scsi/fnic/fnic.h | 3 ++- drivers/scsi/fnic/fnic_fcs.c | 3 ++- drivers/scsi/fnic/fnic_main.c | 1 + drivers/scsi/fnic/fnic_scsi.c | 4 ++-- 4 files changed, 7 insertions(+), 4 deletions(-)
drivers/scsi/fnic/fnic.h +2 -1
@@ -305,6 +305,7 @@ struct fnic {unsigned int copy_wq_base;struct work_struct link_work;struct work_struct frame_work; +struct work_struct flush_work;struct sk_buff_head frame_queue;struct sk_buff_head tx_queue;
@@ -363,7 +364,7 @@ void fnic_handle_event(struct work_struct *work);int fnic_rq_cmpl_handler (struct fnic *fnic,int );int fnic_alloc_rq_frame (struct vnic_rq *rq);void fnic_free_rq_buf (struct vnic_rq *rq,struct vnic_rq_buf *buf); -void fnic_flush_tx (struct fnic *); +void fnic_flush_tx (struct work_struct *);void fnic_eth_send (struct fcoe_ctlr *,struct sk_buff *skb);void fnic_set_port_id (struct fc_lport *,u32 ,struct fc_frame *);void fnic_update_mac (struct fc_lport *,u8 *new);
drivers/scsi/fnic/fnic_fcs.c +2 -1
@@ -1190,8 +1190,9 @@ int fnic_send(struct fc_lport *lp, struct fc_frame *fp)* * Called without fnic_lock held. */ -void fnic_flush_tx (struct fnic *fnic) +void fnic_flush_tx (struct work_struct *work) { + struct fnic *fnic = container_of(work, struct fnic, flush_work);struct sk_buff *skb;struct fc_frame *fp;
drivers/scsi/fnic/fnic_main.c +1 -0
@@ -830,6 +830,7 @@ static int fnic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)spin_lock_init (&fnic->vlans_lock);INIT_WORK (&fnic->fip_frame_work, fnic_handle_fip_frame);INIT_WORK (&fnic->event_work, fnic_handle_event); +INIT_WORK (&fnic->flush_work, fnic_flush_tx);skb_queue_head_init (&fnic->fip_frame_queue);INIT_LIST_HEAD (&fnic->evlist);INIT_LIST_HEAD (&fnic->vlans);
drivers/scsi/fnic/fnic_scsi.c +2 -2
@@ -680,7 +680,7 @@ static int fnic_fcpio_fw_reset_cmpl_handler(struct fnic *fnic,spin_unlock_irqrestore (&fnic->fnic_lock, flags); -fnic_flush_tx (fnic); +queue_work (fnic_event_queue, &fnic->flush_work); reset_cmpl_handler_end:fnic_clear_state_flags (fnic,FNIC_FLAGS_FWRESET );
@@ -736,7 +736,7 @@ static int fnic_fcpio_flogi_reg_cmpl_handler(struct fnic *fnic, }spin_unlock_irqrestore (&fnic->fnic_lock, flags); -fnic_flush_tx (fnic); +queue_work (fnic_event_queue, &fnic->flush_work);queue_work (fnic_event_queue, &fnic->frame_work); }else {spin_unlock_irqrestore (&fnic->fnic_lock, flags);
Signature · 1 line
2.43.0
kernel test robot
build botRe: [PATCH 2/2] fnic: move fnic_fnic_flush_tx() to a work queue
kernel test robot noticed the following build warnings:
Hi Lee,
kernel test robot noticed the following build warnings:
[auto build test WARNING on jejb-scsi/for-next]
[also build test WARNING on mkp-scsi/for-next linus/master v6.8-rc2 next-20240131] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Lee-Duncan/Revert-scsi-fcoe-Fix-potential-deadlock-on-fip-ctlr_lock/20240131-004656 base: https://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi.git for-next patch link: https://lore.kernel.org/r/9c51ef07a04413fb2f2bd20f1534f96e004e4e59.1706632031.git.lduncan%40suse.com patch subject: [PATCH 2/2] fnic: move fnic_fnic_flush_tx() to a work queue config: x86_64-kexec (https://download.01.org/0day-ci/archive/20240131/[email protected]/config) compiler: gcc-12 (Debian 12.2.0-14) 12.2.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240131/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All warnings (new ones prefixed by >>):
>> drivers/scsi/fnic/fnic_fcs.c:1194: warning: Function parameter or struct member 'work' not described in 'fnic_flush_tx' >> drivers/scsi/fnic/fnic_fcs.c:1194: warning: Excess function parameter 'fnic' description in 'fnic_flush_tx'
vim +1194 drivers/scsi/fnic/fnic_fcs.c
5df6d737dd4b0fe Abhijeet Joglekar 2009-04-17 1182
78112e5558064cb Joe Eykholt 2009-11-03 1183 /**
78112e5558064cb Joe Eykholt 2009-11-03 1184 * fnic_flush_tx() - send queued frames.
78112e5558064cb Joe Eykholt 2009-11-03 1185 * @fnic: fnic device
78112e5558064cb Joe Eykholt 2009-11-03 1186 *
78112e5558064cb Joe Eykholt 2009-11-03 1187 * Send frames that were waiting to go out in FC or Ethernet mode.
78112e5558064cb Joe Eykholt 2009-11-03 1188 * Whenever changing modes we purge queued frames, so these frames should
78112e5558064cb Joe Eykholt 2009-11-03 1189 * be queued for the stable mode that we're in, either FC or Ethernet.
78112e5558064cb Joe Eykholt 2009-11-03 1190 *
78112e5558064cb Joe Eykholt 2009-11-03 1191 * Called without fnic_lock held.
78112e5558064cb Joe Eykholt 2009-11-03 1192 */
7ea34b1ffb4e1aa Hannes Reinecke 2024-01-30 1193 void fnic_flush_tx(struct work_struct *work)
78112e5558064cb Joe Eykholt 2009-11-03 @1194 {
7ea34b1ffb4e1aa Hannes Reinecke 2024-01-30 1195 struct fnic *fnic = container_of(work, struct fnic, flush_work);
78112e5558064cb Joe Eykholt 2009-11-03 1196 struct sk_buff *skb;
78112e5558064cb Joe Eykholt 2009-11-03 1197 struct fc_frame *fp;
5df6d737dd4b0fe Abhijeet Joglekar 2009-04-17 1198
d9e9ab56b687da0 Brian Uchino 2010-04-09 1199 while ((skb = skb_dequeue(&fnic->tx_queue))) {
78112e5558064cb Joe Eykholt 2009-11-03 1200 fp = (struct fc_frame *)skb;
78112e5558064cb Joe Eykholt 2009-11-03 1201 fnic_send_frame(fnic, fp);
78112e5558064cb Joe Eykholt 2009-11-03 1202 }
78112e5558064cb Joe Eykholt 2009-11-03 1203 }
5df6d737dd4b0fe Abhijeet Joglekar 2009-04-17 1204 Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
Hannes Reinecke
maintainerRe: [PATCH 2/2] fnic: move fnic_fnic_flush_tx() to a work queue
In reply to Lee DuncanQuoted 5 lines · Lee Duncan
From: Hannes Reinecke <[email protected]>
Rather than call '
fnic_flush_tx()' from interrupt context we should be moving it onto a work queue to avoid any locking issues.
Quoted 9 lines
Signed-off-by Hannes Reinecke <[email protected]>Signed-off-by Lee Duncan <[email protected]>--- drivers/scsi/fnic/fnic.h | 3 ++- drivers/scsi/fnic/fnic_fcs.c | 3 ++- drivers/scsi/fnic/fnic_main.c | 1 + drivers/scsi/fnic/fnic_scsi.c | 4 ++-- 4 files changed, 7 insertions(+), 4 deletions(-)
Can you please add the 'fixes' tag to this one, too? Just the first patch doesn't fix anything, one really would need to apply both to get the issue fixed.
And, of course, fix the kbuild robot warning :-)
Cheers,
Hannes
Signature · 5 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), GF: Ivo Totev, Andrew McDonald, Werner Knoblich