qla2xxx driver bug fixes
Subsystems (2)
Nilesh Javali
maintainerv2 cover qla2xxx driver bug fixes
Martin,
Please apply the qla2xxx driver bug fixes to the scsi tree at your earliest convenience.
v2:
Fixed multi-line comment format
Created new helper routine qla_pci_disconnected()
Added new commit "qla2xxx: Do logout even if fabric scan retries got exhausted"
Added Reviewed-by, Tested-by tags
Thanks,
Nilesh
Arun Easi (3):
qla2xxx: Fix IOPS drop seen in some adapters
qla2xxx: Add H:C:T info in the log message for fc ports
qla2xxx: Fix crash in qla2xxx_mqueuecommand
Nilesh Javali (1):
qla2xxx: Update version to 10.02.00.106-k
Quinn Tran (8):
qla2xxx: fix stuck session
qla2xxx: consolidate zio threshold setting for both fcp & nvme
qla2xxx: Fix use after free in bsg
qla2xxx: fix RISC RESET completion polling
qla2xxx: fix crash in PCIe error handling
qla2xxx: fix mailbox recovery during PCIe error
qla2xxx: include AER debug mask to default
qla2xxx: Do logout even if fabric scan retries got exhausted
drivers/scsi/qla2xxx/qla_bsg.c | 3 +- drivers/scsi/qla2xxx/qla_dbg.c | 16 ++- drivers/scsi/qla2xxx/qla_dbg.h | 2 +- drivers/scsi/qla2xxx/qla_def.h | 11 +- drivers/scsi/qla2xxx/qla_gbl.h | 3 + drivers/scsi/qla2xxx/qla_gs.c | 4 + drivers/scsi/qla2xxx/qla_init.c | 115 ++++++++++++---- drivers/scsi/qla2xxx/qla_inline.h | 46 +++++++ drivers/scsi/qla2xxx/qla_iocb.c | 79 +++++++++-- drivers/scsi/qla2xxx/qla_isr.c | 9 +- drivers/scsi/qla2xxx/qla_mbx.c | 38 +++-- drivers/scsi/qla2xxx/qla_nvme.c | 10 +- drivers/scsi/qla2xxx/qla_os.c | 214 ++++++++++++++++------------- drivers/scsi/qla2xxx/qla_target.c | 6 +- drivers/scsi/qla2xxx/qla_version.h | 4 +- 15 files changed, 405 insertions(+), 155 deletions(-)
base-commit: f749d8b7a9896bc6e5ffe104cc64345037e0b152
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 1/12 qla2xxx: Fix IOPS drop seen in some adapters
From: Arun Easi <[email protected]>
Removing the response queue processing in the send path is showing IOPS drop. Add back the process_response_queue() call in the send path.
--- drivers/scsi/qla2xxx/qla_iocb.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
drivers/scsi/qla2xxx/qla_iocb.c +19 -0
@@ -1600,12 +1600,14 @@ qla24xx_start_scsi(srb_t *sp)uint16_t req_cnt;uint16_t tot_dsds;struct req_que *req =NULL ; +struct rsp_que *rsp;struct scsi_cmnd *cmd =GET_CMD_SP (sp);struct scsi_qla_host *vha = sp->vha;struct qla_hw_data *ha = vha->hw;/* Setup device pointers. */ req = vha->req; + rsp = req->rsp;/* So we know we haven't pci_map'ed anything yet */ tot_dsds =0 ;
@@ -1707,6 +1709,11 @@ qla24xx_start_scsi(srb_t *sp)/* Set chip new ring index. */ wrt_reg_dword (req->req_q_in, req->ring_index); +/* Manage unprocessed RIO/ZIO commands in response queue. */ +if (vha->flags.process_response_queue && + rsp->ring_ptr->signature !=RESPONSE_PROCESSED ) +qla24xx_process_response_queue (vha, rsp); +spin_unlock_irqrestore (&ha->hardware_lock, flags);return QLA_SUCCESS ;
@@ -1897,6 +1904,11 @@ qla24xx_dif_start_scsi(srb_t *sp)/* Set chip new ring index. */ wrt_reg_dword (req->req_q_in, req->ring_index); +/* Manage unprocessed RIO/ZIO commands in response queue. */ +if (vha->flags.process_response_queue && + rsp->ring_ptr->signature !=RESPONSE_PROCESSED ) +qla24xx_process_response_queue (vha, rsp); +spin_unlock_irqrestore (&ha->hardware_lock, flags);return QLA_SUCCESS ;
@@ -1931,6 +1943,7 @@ qla2xxx_start_scsi_mq(srb_t *sp)uint16_t req_cnt;uint16_t tot_dsds;struct req_que *req =NULL ; +struct rsp_que *rsp;struct scsi_cmnd *cmd =GET_CMD_SP (sp);struct scsi_qla_host *vha = sp->fcport->vha;struct qla_hw_data *ha = vha->hw;
@@ -1941,6 +1954,7 @@ qla2xxx_start_scsi_mq(srb_t *sp)/* Setup qpair pointers */ req = qpair->req; + rsp = qpair->rsp;/* So we know we haven't pci_map'ed anything yet */ tot_dsds =0 ;
@@ -2041,6 +2055,11 @@ qla2xxx_start_scsi_mq(srb_t *sp)/* Set chip new ring index. */ wrt_reg_dword (req->req_q_in, req->ring_index); +/* Manage unprocessed RIO/ZIO commands in response queue. */ +if (vha->flags.process_response_queue && + rsp->ring_ptr->signature !=RESPONSE_PROCESSED ) +qla24xx_process_response_queue (vha, rsp); +spin_unlock_irqrestore (&qpair->qp_lock, flags);return QLA_SUCCESS ;
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 2/12 qla2xxx: Add H:C:T info in the log message for fc ports
In reply to Nilesh JavaliFrom: Arun Easi <[email protected]>
The host:channel:scsi_target_id information is helpful in matching an fc port with a scsi device, so add it. For initiator fc ports, a -1 would be displayed for "target" part.
--- drivers/scsi/qla2xxx/qla_init.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
drivers/scsi/qla2xxx/qla_init.c +5 -4
@@ -5512,13 +5512,14 @@ qla2x00_reg_remote_port(scsi_qla_host_t *vha, fc_port_t *fcport)if (fcport->port_type &FCT_NVME_DISCOVERY ) rport_ids.roles |=FC_PORT_ROLE_NVME_DISCOVERY ; +fc_remote_port_rolechg (rport, rport_ids.roles); +ql_dbg (ql_dbg_disc, vha,0x20ee , -"%s %8phN. rport %p is %s mode\n" , - __func__, fcport->port_name, rport, +"%s: %8phN. rport %ld:0:%d (%p) is %s mode\n" , + __func__, fcport->port_name, vha->host_no, + rport->scsi_target_id, rport, (fcport->port_type ==FCT_TARGET ) ?"tgt" : ((fcport->port_type &FCT_NVME ) ?"nvme" :"ini" )); - -fc_remote_port_rolechg (rport, rport_ids.roles); }/*
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 3/12 qla2xxx: fix stuck session
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
Session was stuck due to explicit logout to target was timed out. The target was in an unresponsive state. This timeout induced an error to the GNL command from moving forward.
--- drivers/scsi/qla2xxx/qla_init.c | 1 + drivers/scsi/qla2xxx/qla_target.c | 6 +++++- 2 files changed, 6 insertions(+), 1 deletion(-)
drivers/scsi/qla2xxx/qla_init.c +1 -0
@@ -718,6 +718,7 @@ static void qla24xx_handle_gnl_done_event(scsi_qla_host_t *vha,ql_dbg (ql_dbg_disc, vha,0x20e0 ,"%s %8phC login gen changed\n" , __func__, fcport->port_name); +set_bit (RELOGIN_NEEDED , &vha->dpc_flags);return ; }
drivers/scsi/qla2xxx/qla_target.c +5 -1
@@ -1029,7 +1029,11 @@ void qlt_free_session_done(struct work_struct *work) }msleep (100 ); cnt++; -if (cnt >200 ) +/* +* wait for logout to complete before advance. Otherwise, +* straddling logout can interfere with relogin. +*/ +if (cnt >230 )break ; }
Signature · 1 line
2.19.0.rc0
Re: [PATCH v2 03/12] qla2xxx: fix stuck session
Quoted 35 lines · Nilesh Javali
From: Quinn Tran <[email protected]>
Session was stuck due to explicit logout to target was timed out. The target was in an unresponsive state. This timeout induced an error to the GNL command from moving forward.
Signed-off-by Quinn Tran <[email protected]>Signed-off-by Nilesh Javali <[email protected]>Tested-by Laurence Oberman <[email protected]>---
drivers/scsi/qla2xxx/qla_init.c| 1 +drivers/scsi/qla2xxx/qla_target.c| 6 +++++-2 files changed, 6 insertions(+), 1 deletion(-)drivers/scsi/qla2xxx/qla_init.c +1
-0@@ -718,6 +718,7 @@ static void qla24xx_handle_gnl_done_event(scsi_qla_host_t *vha,ql_dbg (ql_dbg_disc, vha,0x20e0 ,"%s %8phC login gen changed\n" , __func__, fcport->port_name); +set_bit (RELOGIN_NEEDED , &vha->dpc_flags);return ; }drivers/scsi/qla2xxx/qla_target.c +0
-0
drivers/scsi/qla2xxx/qla_target.c void qlt_free_session_done(struct work_struct *work) · line 1035
Quoted 8 lines
drivers/scsi/qla2xxx/qla_target.c +4
-1@@ -1029,7 +1029,11 @@ void qlt_free_session_done(struct work_struct *work) }msleep (100 ); cnt++; -if (cnt >200 ) +/* +* wait for logout to complete before advance. Otherwise, +* straddling logout can interfere with relogin. +*/
How about this ??
/*
* Driver timeout is set to 22 Sec, update count value to loop
* enough time for log-out to complete before advancing. Otherwise,
* straddling log-out can interfere with re-login attempt. */
Quoted 8 lines
drivers/scsi/qla2xxx/qla_target.c +1
-0+if (cnt >230 )break ; }Signature · 2 lines
-- 2.19.0.rc0
--
Himanshu Madhani Oracle Linux Engineering
Nilesh Javali
maintainerPATCH v2 4/12 qla2xxx: consolidate zio threshold setting for both fcp & nvme
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
consolidate zio threshold setting for both fcp & nvme to prevent one protocol from clobbering the setting of the other protocol.
--- drivers/scsi/qla2xxx/qla_def.h | 1 - drivers/scsi/qla2xxx/qla_os.c | 34 ++++++++++++++-------------------- 2 files changed, 14 insertions(+), 21 deletions(-)
drivers/scsi/qla2xxx/qla_def.h +0 -1
@@ -4727,7 +4727,6 @@ typedef struct scsi_qla_host {#define FX00_CRITEMP_RECOVERY 25#define FX00_HOST_INFO_RESEND 26#define QPAIR_ONLINE_CHECK_NEEDED 27 -#define SET_NVME_ZIO_THRESHOLD_NEEDED 28#define DETECT_SFP_CHANGE 29#define N2N_LOGIN_NEEDED 30#define IOCB_WORK_ACTIVE 31
drivers/scsi/qla2xxx/qla_os.c +14 -20
@@ -6969,28 +6969,23 @@ qla2x00_do_dpc(void *data)mutex_unlock (&ha->mq_lock); } -if (test_and_clear_bit (SET_NVME_ZIO_THRESHOLD_NEEDED , - &base_vha->dpc_flags)) { +if (test_and_clear_bit (SET_ZIO_THRESHOLD_NEEDED , + &base_vha->dpc_flags)) { +u16 threshold = ha->nvme_last_rptd_aen + ha->last_zio_threshold; + +if (threshold > ha->orig_fw_xcb_count) + threshold = ha->orig_fw_xcb_count; +ql_log (ql_log_info, base_vha,0xffffff , -"nvme: SET ZIO Activity exchange threshold to %d.\n" , - ha->nvme_last_rptd_aen); -if (qla27xx_set_zio_threshold (base_vha, - ha->nvme_last_rptd_aen)) { +"SET ZIO Activity exchange threshold to %d.\n" , + threshold); +if (qla27xx_set_zio_threshold (base_vha, threshold)) {ql_log (ql_log_info, base_vha,0xffffff , -"nvme: Unable to SET ZIO Activity exchange threshold to %d.\n" , - ha->nvme_last_rptd_aen); +"Unable to SET ZIO Activity exchange threshold to %d.\n" , + threshold); } } -if (test_and_clear_bit (SET_ZIO_THRESHOLD_NEEDED , - &base_vha->dpc_flags)) { -ql_log (ql_log_info, base_vha,0xffffff , -"SET ZIO Activity exchange threshold to %d.\n" , - ha->last_zio_threshold); -qla27xx_set_zio_threshold (base_vha, - ha->last_zio_threshold); - } -if (!IS_QLAFX00 (ha))qla2x00_do_dpc_all_vps (base_vha);
@@ -7218,14 +7213,13 @@ qla2x00_timer(struct timer_list *t) index =atomic_read (&ha->nvme_active_aen_cnt);if (!vha->vp_idx && (index != ha->nvme_last_rptd_aen) && - (index >=DEFAULT_ZIO_THRESHOLD ) && ha->zio_mode ==QLA_ZIO_MODE_6 && !ha->flags.host_shutting_down) { + ha->nvme_last_rptd_aen =atomic_read (&ha->nvme_active_aen_cnt);ql_log (ql_log_info, vha,0x3002 ,"nvme: Sched: Set ZIO exchange threshold to %d.\n" , ha->nvme_last_rptd_aen); - ha->nvme_last_rptd_aen =atomic_read (&ha->nvme_active_aen_cnt); -set_bit (SET_NVME_ZIO_THRESHOLD_NEEDED , &vha->dpc_flags); +set_bit (SET_ZIO_THRESHOLD_NEEDED , &vha->dpc_flags); start_dpc++; }
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 5/12 qla2xxx: Fix use after free in bsg
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
On bsg command completion, bsg_job_done was called while qla driver continue to access the bsg_job buffer. The bsg_job_done can free up resources and reuse by other task, qla continue access of the same resource can read garbage data.
localhost kernel: BUG: KASAN: use-after-free in sg_next+0x64/0x80 localhost kernel: Read of size 8 at addr ffff8883228a3330 by task swapper/26/0 localhost kernel:
localhost kernel: CPU: 26 PID: 0 Comm: swapper/26 Kdump:
loaded Tainted: G OE --------- - - 4.18.0-193.el8.x86_64+debug #1
localhost kernel: Hardware name: HP ProLiant DL360
Gen9/ProLiant DL360 Gen9, BIOS P89 08/12/2016
localhost kernel: Call Trace:
localhost kernel: <IRQ>
localhost kernel: dump_stack+0x9a/0xf0
localhost kernel: print_address_description.cold.3+0x9/0x23b localhost kernel: kasan_report.cold.4+0x65/0x95
localhost kernel: debug_dma_unmap_sg.part.12+0x10d/0x2d0 localhost kernel: qla2x00_bsg_sp_free+0xaf6/0x1010 [qla2xxx]
--- drivers/scsi/qla2xxx/qla_bsg.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/scsi/qla2xxx/qla_bsg.c +2 -1
@@ -25,10 +25,11 @@ void qla2x00_bsg_job_done(srb_t *sp, int res)struct bsg_job *bsg_job = sp->u.bsg_job;struct fc_bsg_reply *bsg_reply = bsg_job->reply; + sp->free (sp); + bsg_reply->result = res;bsg_job_done (bsg_job, bsg_reply->result, bsg_reply->reply_payload_rcv_len); - sp->free (sp); }void qla2x00_bsg_sp_free (srb_t *sp)
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 6/12 qla2xxx: Fix crash in qla2xxx_mqueuecommand
In reply to Nilesh JavaliFrom: Arun Easi <[email protected]>
RIP: 0010:kmem_cache_free+0xfa/0x1b0
Call Trace:
qla2xxx_mqueuecommand+0x2b5/0x2c0 [qla2xxx]
scsi_queue_rq+0x5e2/0xa40
__blk_mq_try_issue_directly+0x128/0x1d0
blk_mq_request_issue_directly+0x4e/0xb0Fix incorrect call to free srb in qla2xxx_mqueuecommand, as srb is now allocated by upper layers. This fixes smatch warning of srb unintended free.
--- drivers/scsi/qla2xxx/qla_os.c | 7 ------- 1 file changed, 7 deletions(-)
drivers/scsi/qla2xxx/qla_os.c +0 -7
@@ -1013,8 +1013,6 @@ qla2xxx_mqueuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd,if (rval !=QLA_SUCCESS ) {ql_dbg (ql_dbg_io + ql_dbg_verbose, vha,0x3078 ,"Start scsi failed rval=%d for cmd=%p.\n" , rval, cmd); -if (rval ==QLA_INTERFACE_ERROR ) - goto qc24_free_sp_fail_command; goto qc24_host_busy_free_sp; }
@@ -1026,11 +1024,6 @@ qla2xxx_mqueuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd, qc24_target_busy:return SCSI_MLQUEUE_TARGET_BUSY ; -qc24_free_sp_fail_command: - sp->free (sp); -CMD_SP (cmd) =NULL ; -qla2xxx_rel_qpair_sp (sp->qpair, sp); - qc24_fail_command: cmd->scsi_done (cmd);
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 7/12 qla2xxx: fix RISC RESET completion polling
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
After risc reset, the poll time for risc reset completion is too short. Fix the completion polling time.
--- drivers/scsi/qla2xxx/qla_init.c | 65 ++++++++++++++++++++++++++++++--- 1 file changed, 59 insertions(+), 6 deletions(-)
drivers/scsi/qla2xxx/qla_init.c +59 -6
@@ -2767,6 +2767,49 @@ qla81xx_reset_mpi(scsi_qla_host_t *vha)return qla81xx_write_mpi_register (vha, mb); } +static int +qla_chk_risc_recovery (scsi_qla_host_t *vha) +{ +struct qla_hw_data *ha = vha->hw; +struct device_reg_24xx __iomem *reg = &ha->iobase->isp24; +__le16 __iomem *mbptr = ®->mailbox0; +int i; +u16 mb[32 ]; +int rc =QLA_SUCCESS ; + +if (!IS_QLA27XX (ha) && !IS_QLA28XX (ha)) +return rc; + +/* this check is only valid after RISC reset */ + mb[0 ] =rd_reg_word (mbptr); + mbptr++; +if (mb[0 ] ==0xf ) { + rc =QLA_FUNCTION_FAILED ; + +for (i =1 ; i <32 ; i++) { + mb[i] =rd_reg_word (mbptr); + mbptr++; + } + +ql_log (ql_log_warn, vha,0x1015 , +"RISC reset failed. mb[0-7] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n" , + mb[0 ], mb[1 ], mb[2 ], mb[3 ], mb[4 ], mb[5 ], mb[6 ], mb[7 ]); +ql_log (ql_log_warn, vha,0x1015 , +"RISC reset failed. mb[8-15] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n" , + mb[8 ], mb[9 ], mb[10 ], mb[11 ], mb[12 ], mb[13 ], mb[14 ], + mb[15 ]); +ql_log (ql_log_warn, vha,0x1015 , +"RISC reset failed. mb[16-23] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n" , + mb[16 ], mb[17 ], mb[18 ], mb[19 ], mb[20 ], mb[21 ], mb[22 ], + mb[23 ]); +ql_log (ql_log_warn, vha,0x1015 , +"RISC reset failed. mb[24-31] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n" , + mb[24 ], mb[25 ], mb[26 ], mb[27 ], mb[28 ], mb[29 ], mb[30 ], + mb[31 ]); + } +return rc; +} +/** * qla24xx_reset_risc() - Perform full reset of ISP24xx RISC. * @vha: HA context
@@ -2783,6 +2826,7 @@ qla24xx_reset_risc(scsi_qla_host_t *vha)uint16_t wd;static int abts_cnt;/* ISP abort retry counts */ int rval =QLA_SUCCESS ; +int print =1 ;spin_lock_irqsave (&ha->hardware_lock, flags);
@@ -2871,17 +2915,26 @@ qla24xx_reset_risc(scsi_qla_host_t *vha)rd_reg_dword (®->hccr);wrt_reg_dword (®->hccr,HCCRX_CLR_RISC_RESET ); +mdelay (10 );rd_reg_dword (®->hccr); -rd_reg_word (®->mailbox0); -for (cnt =60 ;rd_reg_word (®->mailbox0) !=0 && - rval ==QLA_SUCCESS ; cnt--) { + wd =rd_reg_word (®->mailbox0); +for (cnt =300 ; wd !=0 && rval ==QLA_SUCCESS ; cnt--) {barrier (); -if (cnt) -udelay (5 ); -else +if (cnt) { +mdelay (1 ); +if (print &&qla_chk_risc_recovery (vha)) + print =0 ; + + wd =rd_reg_word (®->mailbox0); + }else { rval =QLA_FUNCTION_TIMEOUT ; + +ql_log (ql_log_warn, vha,0x015e , +"RISC reset timeout\n" ); + } } +if (rval ==QLA_SUCCESS )set_bit (RISC_RDY_AFT_RESET , &ha->fw_dump_cap_flags);
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 8/12 qla2xxx: fix crash in PCIe error handling
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
BUG: unable to handle kernel NULL pointer dereference at (null) IP: qla2x00_abort_isp+0x21/0x6b0 [qla2xxx] PGD 0 P4D 0 Oops: 0000 [#1] SMP PTI CPU: 0 PID: 1715 Comm: kworker/0:2 Tainted: GOE 4.12.14-122.37-default #1 SLE12-SP5 Hardware name: HPE Superdome Flex/Superdome Flex, BIOS Bundle:3.30.100 SFW:IP147.007.004.017.000.2009211957 09/21/2020 Workqueue: events aer_recover_work_func task: ffff9e399c14ca80 task.stack: ffffc1c58e4ac000 RIP: 0010:qla2x00_abort_isp+0x21/0x6b0 [qla2xxx] RSP: 0018:ffffc1c58e4afd50 EFLAGS: 00010282 RAX: 0000000000000000 RBX: ffff9e419cdef480 RCX: 0000000000000000 RDX: ffff9e399c14ca80 RSI: 0000000000000246 RDI: ffff9e419bbc27b8 RBP: ffff9e419bbc27b8 R08: 0000000000000004 R09: 00000000a0440000 R10: 0000000000000000 R11: ffff9e399416d1a0 R12: ffff9e419cdef000 R13: ffff9e3a7cfae800 R14: ffff9e3a7cfae800 R15: 00000000000000c0 FS: 0000000000000000(0000) GS:ffff9e39a0000000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000006cd00a005 CR4: 00000000007606f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: qla2xxx_pci_slot_reset+0x141/0x160 [qla2xxx] report_slot_reset+0x41/0x80 ? merge_result.part.4+0x30/0x30 pci_walk_bus+0x70/0x90 pcie_do_recovery+0x1db/0x2e0 aer_recover_work_func+0xc2/0xf0 process_one_work+0x14c/0x390
Disable board_disable logic where driver resources are freed while OS is in the process of recovering the adapter.
--- drivers/scsi/qla2xxx/qla_dbg.c | 16 ++- drivers/scsi/qla2xxx/qla_def.h | 10 ++ drivers/scsi/qla2xxx/qla_gbl.h | 3 + drivers/scsi/qla2xxx/qla_init.c | 40 ++++--- drivers/scsi/qla2xxx/qla_inline.h | 46 ++++++++ drivers/scsi/qla2xxx/qla_iocb.c | 60 +++++++++-- drivers/scsi/qla2xxx/qla_isr.c | 9 +- drivers/scsi/qla2xxx/qla_mbx.c | 3 +- drivers/scsi/qla2xxx/qla_nvme.c | 10 +- drivers/scsi/qla2xxx/qla_os.c | 173 ++++++++++++++++++------------ 10 files changed, 265 insertions(+), 105 deletions(-)
drivers/scsi/qla2xxx/qla_dbg.c +15 -1
@@ -113,8 +113,13 @@ qla27xx_dump_mpi_ram(struct qla_hw_data *ha, uint32_t addr, uint32_t *ram,uint32_t stat;ulong i, j, timer =6000000 ;int rval =QLA_FUNCTION_FAILED ; +scsi_qla_host_t *vha =pci_get_drvdata (ha->pdev);clear_bit (MBX_INTERRUPT , &ha->mbx_cmd_flags); + +if (qla_pci_disconnected (vha, reg)) +return rval; +for (i =0 ; i < ram_dwords; i += dwords, addr += dwords) {if (i + dwords > ram_dwords) dwords = ram_dwords - i;
@@ -138,6 +143,9 @@ qla27xx_dump_mpi_ram(struct qla_hw_data *ha, uint32_t addr, uint32_t *ram,while (timer--) {udelay (5 ); +if (qla_pci_disconnected (vha, reg)) +return rval; + stat =rd_reg_dword (®->host_status);/* Check for pending interrupts. */ if (!(stat &HSRX_RISC_INT ))
@@ -192,9 +200,13 @@ qla24xx_dump_ram(struct qla_hw_data *ha, uint32_t addr, __be32 *ram,uint32_t dwords =qla2x00_gid_list_size (ha) /4 ;uint32_t stat;ulong i, j, timer =6000000 ; +scsi_qla_host_t *vha =pci_get_drvdata (ha->pdev);clear_bit (MBX_INTERRUPT , &ha->mbx_cmd_flags); +if (qla_pci_disconnected (vha, reg)) +return rval; +for (i =0 ; i < ram_dwords; i += dwords, addr += dwords) {if (i + dwords > ram_dwords) dwords = ram_dwords - i;
@@ -216,8 +228,10 @@ qla24xx_dump_ram(struct qla_hw_data *ha, uint32_t addr, __be32 *ram, ha->flags.mbox_int =0 ;while (timer--) {udelay (5 ); - stat =rd_reg_dword (®->host_status); +if (qla_pci_disconnected (vha, reg)) +return rval; + stat =rd_reg_dword (®->host_status);/* Check for pending interrupts. */ if (!(stat &HSRX_RISC_INT ))continue ;
drivers/scsi/qla2xxx/qla_def.h +10 -0
@@ -396,6 +396,7 @@ typedef union { } b; } port_id_t;#define INVALID_PORT_ID 0xFFFFFF +#define ISP_REG16_DISCONNECT 0xFFFFstatic inline le_id_t be_id_to_le (be_id_t id) {
@@ -3857,6 +3858,13 @@ struct qla_hw_data_stat {u32 num_mpi_reset; }; +/* refer to pcie_do_recovery reference */ +typedef enum { +QLA_PCI_RESUME , +QLA_PCI_ERR_DETECTED , +QLA_PCI_MMIO_ENABLED , +QLA_PCI_SLOT_RESET , +}pci_error_state_t ;/* * Qlogic host adapter specific data structure. */
@@ -4607,6 +4615,7 @@ struct qla_hw_data {#define DEFAULT_ZIO_THRESHOLD 5struct qla_hw_data_stat stat; +pci_error_state_t pci_error_state; };struct active_regions {
@@ -4727,6 +4736,7 @@ typedef struct scsi_qla_host {#define FX00_CRITEMP_RECOVERY 25#define FX00_HOST_INFO_RESEND 26#define QPAIR_ONLINE_CHECK_NEEDED 27 +#define DO_EEH_RECOVERY 28#define DETECT_SFP_CHANGE 29#define N2N_LOGIN_NEEDED 30#define IOCB_WORK_ACTIVE 31
drivers/scsi/qla2xxx/qla_gbl.h +3 -0
@@ -224,6 +224,7 @@ extern int qla2x00_post_uevent_work(struct scsi_qla_host *, u32);extern int qla2x00_post_uevent_work (struct scsi_qla_host *,u32 );extern void qla2x00_disable_board_on_pci_error (struct work_struct *); +extern void qla_eeh_work (struct work_struct *);extern void qla2x00_sp_compl (srb_t *sp,int );extern void qla2xxx_qpair_sp_free_dma (srb_t *sp);extern void qla2xxx_qpair_sp_compl (srb_t *sp,int );
@@ -235,6 +236,8 @@ int qla24xx_post_relogin_work(struct scsi_qla_host *vha);void qla2x00_wait_for_sess_deletion (scsi_qla_host_t *);void qla24xx_process_purex_rdp (struct scsi_qla_host *vha,struct purex_item *pkt); +void qla_pci_set_eeh_busy (struct scsi_qla_host *); +void qla_schedule_eeh_work (struct scsi_qla_host *);/* * Global Functions in qla_mid.c source file.
drivers/scsi/qla2xxx/qla_init.c +26 -14
@@ -6932,22 +6932,18 @@ qla2x00_abort_isp_cleanup(scsi_qla_host_t *vha) }spin_unlock_irqrestore (&ha->vport_slock, flags); -if (!ha->flags.eeh_busy) { -/* Make sure for ISP 82XX IO DMA is complete */ -if (IS_P3P_TYPE (ha)) { -qla82xx_chip_reset_cleanup (vha); -ql_log (ql_log_info, vha,0x00b4 , -"Done chip reset cleanup.\n" ); - -/* Done waiting for pending commands. -* Reset the online flag. -*/ - vha->flags.online =0 ; - } +/* Make sure for ISP 82XX IO DMA is complete */ +if (IS_P3P_TYPE (ha)) { +qla82xx_chip_reset_cleanup (vha); +ql_log (ql_log_info, vha,0x00b4 , +"Done chip reset cleanup.\n" ); -/* Requeue all commands in outstanding command list. */ -qla2x00_abort_all_cmds (vha,DID_RESET <<16 ); +/* Done waiting for pending commands. Reset online flag */ + vha->flags.online =0 ; } + +/* Requeue all commands in outstanding command list. */ +qla2x00_abort_all_cmds (vha,DID_RESET <<16 );/* memory barrier */ wmb (); }
@@ -6978,6 +6974,12 @@ qla2x00_abort_isp(scsi_qla_host_t *vha)if (vha->hw->flags.port_isolated)return status; +if (qla2x00_isp_reg_stat (ha)) { +ql_log (ql_log_info, vha,0x803f , +"ISP Abort - ISP reg disconnect, exiting.\n" ); +return status; + } +if (test_and_clear_bit (ISP_ABORT_TO_ROM , &vha->dpc_flags)) { ha->flags.chip_reset_done =1 ; vha->flags.online =1 ;
@@ -7017,8 +7019,18 @@ qla2x00_abort_isp(scsi_qla_host_t *vha) ha->isp_ops->get_flash_version (vha, req->ring); +if (qla2x00_isp_reg_stat (ha)) { +ql_log (ql_log_info, vha,0x803f , +"ISP Abort - ISP reg disconnect pre nvram config, exiting.\n" ); +return status; + } ha->isp_ops->nvram_config (vha); +if (qla2x00_isp_reg_stat (ha)) { +ql_log (ql_log_info, vha,0x803f , +"ISP Abort - ISP reg disconnect post nvmram config, exiting.\n" ); +return status; + }if (!qla2x00_restart_isp (vha)) {clear_bit (RESET_MARKER_NEEDED , &vha->dpc_flags);
drivers/scsi/qla2xxx/qla_inline.h +46 -0
49 lines not on this page · raw message
drivers/scsi/qla2xxx/qla_iocb.c +50 -10
119 lines not on this page · raw message
drivers/scsi/qla2xxx/qla_isr.c +1 -8
@@ -270,12 +270,7 @@ qla2x00_check_reg32_for_disconnect(scsi_qla_host_t *vha, uint32_t reg)if (!test_and_set_bit (PFLG_DISCONNECTED , &vha->pci_flags) && !test_bit (PFLG_DRIVER_REMOVING , &vha->pci_flags) && !test_bit (PFLG_DRIVER_PROBING , &vha->pci_flags)) { -/* -* Schedule this (only once) on the default system -* workqueue so that all the adapter workqueues and the -* DPC thread can be shutdown cleanly. -*/ -schedule_work (&vha->hw->board_disable); +qla_schedule_eeh_work (vha); }return true; }else
@@ -1657,8 +1652,6 @@ qla2x00_async_event(scsi_qla_host_t *vha, struct rsp_que *rsp, uint16_t *mb)case MBA_TEMPERATURE_ALERT :ql_dbg (ql_dbg_async, vha,0x505e ,"TEMPERATURE ALERT: %04x %04x %04x\n" , mb[1 ], mb[2 ], mb[3 ]); -if (mb[1 ] ==0x12 ) -schedule_work (&ha->board_disable);break ;case MBA_TRANS_INSERT :
drivers/scsi/qla2xxx/qla_mbx.c +2 -1
@@ -161,7 +161,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)/* check if ISP abort is active and return cmd with timeout */ if ((test_bit (ABORT_ISP_ACTIVE , &base_vha->dpc_flags) ||test_bit (ISP_ABORT_RETRY , &base_vha->dpc_flags) || -test_bit (ISP_ABORT_NEEDED , &base_vha->dpc_flags)) && +test_bit (ISP_ABORT_NEEDED , &base_vha->dpc_flags) || + ha->flags.eeh_busy) && !is_rom_cmd (mcp->mb[0 ])) { ql_log(ql_log_info, vha, 0x1005, "Cmd 0x%x aborted with timeout since ISP Abort is pending\n",
drivers/scsi/qla2xxx/qla_nvme.c +8 -2
@@ -398,8 +398,13 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp) } req_cnt =qla24xx_calc_iocbs (vha, tot_dsds);if (req->cnt < (req_cnt +2 )) { - cnt =IS_SHADOW_REG_CAPABLE (ha) ? *req->out_ptr : -rd_reg_dword_relaxed (req->req_q_out); +if (IS_SHADOW_REG_CAPABLE (ha)) { + cnt = *req->out_ptr; + }else { + cnt =rd_reg_dword_relaxed (req->req_q_out); +if (qla2x00_check_reg16_for_disconnect (vha, cnt)) + goto queuing_error; + }if (req->ring_index < cnt) req->cnt = cnt - req->ring_index;
@@ -536,6 +541,7 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp) queuing_error:spin_unlock_irqrestore (&qpair->qp_lock, flags); +return rval; }
drivers/scsi/qla2xxx/qla_os.c +104 -69
307 lines not on this page · raw message
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 9/12 qla2xxx: fix mailbox recovery during PCIe error
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
For the mailbox thread that encounter PCIe error, pause that thread until PCIe link reset/recovery completed to prevent the thread from possibly unmapping any type of DMA resource that might be in progress at the same time.
--- drivers/scsi/qla2xxx/qla_mbx.c | 39 ++++++++++++++++++++++++++-------- 1 file changed, 30 insertions(+), 9 deletions(-)
drivers/scsi/qla2xxx/qla_mbx.c +30 -9
@@ -102,7 +102,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)int rval, i;unsigned long flags =0 ;device_reg_t *reg; -uint8_t abort_active; +uint8_t abort_active, eeh_delay;uint8_t io_lock_on;uint16_t command =0 ;uint16_t *iptr;
@@ -136,7 +136,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) "PCI error, exiting.\n");return QLA_FUNCTION_TIMEOUT ; } - + eeh_delay =0 ; reg = ha->iobase; io_lock_on = base_vha->flags.init_done;
@@ -159,11 +159,10 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) }/* check if ISP abort is active and return cmd with timeout */ -if ((test_bit (ABORT_ISP_ACTIVE , &base_vha->dpc_flags) || -test_bit (ISP_ABORT_RETRY , &base_vha->dpc_flags) || -test_bit (ISP_ABORT_NEEDED , &base_vha->dpc_flags) || - ha->flags.eeh_busy) && - !is_rom_cmd (mcp->mb[0 ])) { +if (((test_bit (ABORT_ISP_ACTIVE , &base_vha->dpc_flags) || +test_bit (ISP_ABORT_RETRY , &base_vha->dpc_flags) || +test_bit (ISP_ABORT_NEEDED , &base_vha->dpc_flags)) && + !is_rom_cmd (mcp->mb[0 ])) || ha->flags.eeh_busy) {ql_log (ql_log_info, vha,0x1005 ,"Cmd 0x%x aborted with timeout since ISP Abort is pending\n" , mcp->mb[0 ]);
@@ -186,7 +185,11 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)return QLA_FUNCTION_TIMEOUT ; }atomic_dec (&ha->num_pend_mbx_stage1); -if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) { +if (ha->flags.purge_mbox || chip_reset != ha->chip_reset || + ha->flags.eeh_busy) { +ql_log (ql_log_warn, vha,0xd035 , +"Error detected: purge[%d] eeh[%d] cmd=0x%x, Exiting.\n" , + ha->flags.purge_mbox, ha->flags.eeh_busy, mcp->mb[0 ]); rval =QLA_ABORTED ; goto premature_exit; }
@@ -266,6 +269,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)if (!wait_for_completion_timeout (&ha->mbx_intr_comp, mcp->tov *HZ )) {if (chip_reset != ha->chip_reset) { + eeh_delay = ha->flags.eeh_busy ?1 :0 ; +spin_lock_irqsave (&ha->hardware_lock, flags); ha->flags.mbox_busy =0 ; spin_unlock_irqrestore(&ha->hardware_lock,
@@ -283,6 +288,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) }else if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) { + eeh_delay = ha->flags.eeh_busy ?1 :0 ; +spin_lock_irqsave (&ha->hardware_lock, flags); ha->flags.mbox_busy =0 ;spin_unlock_irqrestore (&ha->hardware_lock, flags);
@@ -324,6 +331,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)while (!ha->flags.mbox_int) {if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) { + eeh_delay = ha->flags.eeh_busy ?1 :0 ; +spin_lock_irqsave (&ha->hardware_lock, flags); ha->flags.mbox_busy =0 ; spin_unlock_irqrestore(&ha->hardware_lock,
@@ -532,7 +541,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)clear_bit (ISP_ABORT_NEEDED , &vha->dpc_flags);/* Allow next mbx cmd to come in. */ complete (&ha->mbx_cmd_comp); -if (ha->isp_ops->abort_isp (vha)) { +if (ha->isp_ops->abort_isp (vha) && + !ha->flags.eeh_busy) {/* Failed. retry later. */ set_bit (ISP_ABORT_NEEDED , &vha->dpc_flags);
@@ -585,6 +595,17 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)ql_dbg (ql_dbg_mbx, base_vha,0x1021 ,"Done %s.\n" , __func__); } + i =500 ; +while (i && eeh_delay && (ha->pci_error_state <QLA_PCI_SLOT_RESET )) { +/* +* The caller of this mailbox encounter pci error. +* Hold the thread until PCIE link reset complete to make +* sure caller does not unmap dma while recovery is +* in progress. +*/ +msleep (1 ); + i--; + }return rval; }
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 10/12 qla2xxx: include AER debug mask to default
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
Use PCIe AER debug mask as default.
--- drivers/scsi/qla2xxx/qla_dbg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/qla2xxx/qla_dbg.h +1 -1
@@ -308,7 +308,7 @@ struct qla2xxx_fw_dump { };#define QL_MSGHDR "qla2xxx" -#define QL_DBG_DEFAULT1_MASK 0x1e400000 +#define QL_DBG_DEFAULT1_MASK 0x1e600000#define ql_log_fatal 0/* display fatal errors */ #define ql_log_warn 1/* display critical errors */
Signature · 1 line
2.19.0.rc0
Nilesh Javali
maintainerPATCH v2 11/12 qla2xxx: Do logout even if fabric scan retries got exhausted
In reply to Nilesh JavaliFrom: Quinn Tran <[email protected]>
Perform logout of all remote ports so that all IOs with driver are requeued with mid-layer for retry.
--- drivers/scsi/qla2xxx/qla_gs.c | 4 ++++ 1 file changed, 4 insertions(+)
drivers/scsi/qla2xxx/qla_gs.c +4 -0
@@ -3443,6 +3443,10 @@ void qla24xx_async_gnnft_done(scsi_qla_host_t *vha, srb_t *sp)list_for_each_entry (fcport, &vha->vp_fcports, list) {if ((fcport->flags &FCF_FABRIC_DEVICE ) !=0 ) { fcport->scan_state =QLA_FCPORT_SCAN ; +if (fcport->loop_id ==FC_NO_LOOP_ID ) + fcport->logout_on_delete =0 ; +else + fcport->logout_on_delete =1 ; } } goto login_logout;
Signature · 1 line
2.19.0.rc0
Re: [PATCH v2 11/12] qla2xxx: Do logout even if fabric scan retries got exhausted
Quoted 34 lines · Nilesh Javali
From: Quinn Tran <[email protected]>
Perform logout of all remote ports so that all IOs with driver are requeued with mid-layer for retry.
Signed-off-by Saurav Kashyap <[email protected]>Signed-off-by Nilesh Javali <[email protected]>Signed-off-by Quinn Tran <[email protected]>---
drivers/scsi/qla2xxx/qla_gs.c| 4 ++++1 file changed, 4 insertions(+)drivers/scsi/qla2xxx/qla_gs.c +4
-0@@ -3443,6 +3443,10 @@ void qla24xx_async_gnnft_done(scsi_qla_host_t *vha, srb_t *sp)list_for_each_entry (fcport, &vha->vp_fcports, list) {if ((fcport->flags &FCF_FABRIC_DEVICE ) !=0 ) { fcport->scan_state =QLA_FCPORT_SCAN ; +if (fcport->loop_id ==FC_NO_LOOP_ID ) + fcport->logout_on_delete =0 ; +else + fcport->logout_on_delete =1 ; } } goto login_logout;Signature · 1 line
2.19.0.rc0
Looks Good.
--
Himanshu Madhani Oracle Linux Engineering
Nilesh Javali
maintainerPATCH v2 12/12 qla2xxx: Update version to 10.02.00.106-k
In reply to Nilesh Javali--- drivers/scsi/qla2xxx/qla_version.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/scsi/qla2xxx/qla_version.h +2 -2
@@ -6,9 +6,9 @@/* * Driver version */ -#define QLA2XXX_VERSION "10.02.00.105-k" +#define QLA2XXX_VERSION "10.02.00.106-k"#define QLA_DRIVER_MAJOR_VER 10#define QLA_DRIVER_MINOR_VER 2#define QLA_DRIVER_PATCH_VER 0 -#define QLA_DRIVER_BETA_VER 105 +#define QLA_DRIVER_BETA_VER 106
Signature · 1 line
2.19.0.rc0
Martin K. Petersen
AppliedNilesh,
Quoted 2 lines
Please apply the qla2xxx driver bug fixes to the scsi tree at your earliest convenience.
Applied to 5.13/scsi-staging, thanks!
Signature · 1 line
Martin K. Petersen Oracle Linux Engineering
Martin K. Petersen
AppliedQuoted 5 lines · Nilesh Javali
Martin,
Please apply the qla2xxx driver bug fixes to the scsi tree at your earliest convenience.
Quoted 8 lines
v2:
Fixed multi-line comment format
Created new helper routineqla_pci_disconnected()
Added new commit "qla2xxx: Do logout even if fabric scan retries got exhausted"
Added Reviewed-by, Tested-by tags[...]
Applied to 5.13/scsi-queue, thanks!
[01/12] qla2xxx: Fix IOPS drop seen in some adapters
https://git.kernel.org/mkp/scsi/c/bcafad6c2d52 [02/12] qla2xxx: Add H:C:T info in the log message for fc ports https://git.kernel.org/mkp/scsi/c/a63f4c454149 [03/12] qla2xxx: fix stuck session https://git.kernel.org/mkp/scsi/c/c358a3d92b32 [04/12] qla2xxx: consolidate zio threshold setting for both fcp & nvme https://git.kernel.org/mkp/scsi/c/5777fef788a5 [05/12] qla2xxx: Fix use after free in bsg https://git.kernel.org/mkp/scsi/c/2ce35c0821af [06/12] qla2xxx: Fix crash in qla2xxx_mqueuecommand https://git.kernel.org/mkp/scsi/c/6641df81ab79 [07/12] qla2xxx: fix RISC RESET completion polling https://git.kernel.org/mkp/scsi/c/610d027b1e63 [08/12] qla2xxx: fix crash in PCIe error handling https://git.kernel.org/mkp/scsi/c/f7a0ed479e66 [09/12] qla2xxx: fix mailbox recovery during PCIe error https://git.kernel.org/mkp/scsi/c/daafc8d33ff6 [10/12] qla2xxx: include AER debug mask to default https://git.kernel.org/mkp/scsi/c/1cbcc531d01f [11/12] qla2xxx: Do logout even if fabric scan retries got exhausted https://git.kernel.org/mkp/scsi/c/022a2d211ce0 [12/12] qla2xxx: Update version to 10.02.00.106-k https://git.kernel.org/mkp/scsi/c/10d91a15f26e
Signature · 1 line
Martin K. Petersen Oracle Linux Engineering