[PATCH 00/11] megaraid_sas: driver updates to 07.713.01.00-rc1

[PATCH 00/11] megaraid_sas: driver updates to 07.713.01.00-rc1

[Anand Lodnoor]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=y, Size: 1136 bytes --]

This patchset contains few enhancements and fixes in megaraid_sas driver.

Anand Lodnoor (11):
  megaraid_sas: Add transition_to_ready retry logic in resume path
  megaraid_sas: Set no_write_same only for Virtual Disk
  megaraid_sas: Update queue_depth of SAS and NVMe devices
  megaraid_sas: Don’t kill already dead adapter
  megaraid_sas: Do not kill HBA if JBOD Seqence map or RAID map is
    disabled
  megaraid_sas: Do not set HBA Operational if FW is not in operational
    state
  megaraid_sas: Re-Define enum DCMD_RETURN_STATUS
  megaraid_sas: Do not initiate OCR if controller is not in ready state
  megaraid_sas: Return pended IOCTLs after 3 retries
  megaraid_sas: Use Block layer API to check SCSI device in-flight IO
    requests
  megaraid_sas: Update driver version to 07.713.01.00-rc1

 drivers/scsi/megaraid/megaraid_sas.h        |  17 ++--
 drivers/scsi/megaraid/megaraid_sas_base.c   |  99 ++++++++++++++------
 drivers/scsi/megaraid/megaraid_sas_fusion.c | 134 +++++++++++++++++++++-------
 drivers/scsi/megaraid/megaraid_sas_fusion.h |  18 +++-
 4 files changed, 201 insertions(+), 67 deletions(-)

-- 
1.8.3.1

[PATCH 01/11] megaraid_sas: Add transition_to_ready retry logic in resume path

[Anand Lodnoor]

Signed-off-by: Chandrakanth Patil <chandrakanth.patil@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_base.c | 34 +++++++++++++++++++++++++++++--
 1 file changed, 32 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index a4bc814..b5f0221 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -7593,6 +7593,7 @@ static void megasas_shutdown_controller(struct megasas_instance *instance,
 	struct Scsi_Host *host;
 	struct megasas_instance *instance;
 	int irq_flags = PCI_IRQ_LEGACY;
+	u32 status_reg;
 
 	instance = pci_get_drvdata(pdev);
 
@@ -7620,9 +7621,38 @@ static void megasas_shutdown_controller(struct megasas_instance *instance,
 	/*
 	 * We expect the FW state to be READY
 	 */
-	if (megasas_transition_to_ready(instance, 0))
-		goto fail_ready_state;
 
+	if (megasas_transition_to_ready(instance, 0)) {
+		dev_info(&instance->pdev->dev,
+			 "Failed to transition controller to ready from %s!\n",
+			 __func__);
+		if (instance->adapter_type != MFI_SERIES) {
+			status_reg =
+			instance->instancet->read_fw_status_reg(instance);
+			if (status_reg & MFI_RESET_ADAPTER) {
+				if (megasas_adp_reset_wait_for_ready
+					(instance, true, 0) == FAILED)
+					goto fail_ready_state;
+				} else {
+					goto fail_ready_state;
+				}
+		} else {
+			atomic_set(&instance->fw_reset_no_pci_access, 1);
+			instance->instancet->adp_reset
+				(instance, instance->reg_set);
+			atomic_set(&instance->fw_reset_no_pci_access, 0);
+
+			/*waitting for about 30 second before retry*/
+			ssleep(30);
+
+			if (megasas_transition_to_ready(instance, 0))
+				goto fail_ready_state;
+		}
+
+		dev_info(&instance->pdev->dev,
+			 "FW restarted successfully from %s!\n",
+			 __func__);
+	}
 	if (megasas_set_dma_mask(instance))
 		goto fail_set_dma_mask;
 
-- 
1.8.3.1

Re: [PATCH 01/11] megaraid_sas: Add transition_to_ready retry logic in resume path

[Martin K. Petersen]

Hi Anand,

First of all, you need a better commit descriptions with a rationale for
each change. Several of the patches in the posted series have vague
one-liners.


In addition, this hunk seems odd:

> +				if (megasas_adp_reset_wait_for_ready
> +					(instance, true, 0) == FAILED)
> +					goto fail_ready_state;
> +				} else {
> +					goto fail_ready_state;
> +				}


Couple of typos. And please add a space after /* and before */:

> +			/*waitting for about 30 second before retry*/

-- 
Martin K. Petersen	Oracle Linux Engineering

[PATCH 02/11] megaraid_sas: Set no_write_same only for Virtual Disk

[Anand Lodnoor]

Set no_write_same only for Virtual Disk, "no_write_same" flag should not
be set for EPDs (Encapsulated PDs).

Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_base.c   |  5 ++++-
 drivers/scsi/megaraid/megaraid_sas_fusion.h | 17 ++++++++++++++---
 2 files changed, 18 insertions(+), 4 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index b5f0221..3cd269f 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -1887,6 +1887,10 @@ void megasas_set_dynamic_target_properties(struct scsi_device *sdev,
 
 		mr_device_priv_data->is_tm_capable =
 			raid->capability.tmCapable;
+
+		if (!raid->flags.isEPD)
+			sdev->no_write_same = 1;
+
 	} else if (instance->use_seqnum_jbod_fp) {
 		pd_index = (sdev->channel * MEGASAS_MAX_DEV_PER_CHANNEL) +
 			sdev->id;
@@ -3416,7 +3420,6 @@ static int megasas_reset_target(struct scsi_cmnd *scmd)
 	.bios_param = megasas_bios_param,
 	.change_queue_depth = scsi_change_queue_depth,
 	.max_segment_size = 0xffffffff,
-	.no_write_same = 1,
 };
 
 /**
diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.h b/drivers/scsi/megaraid/megaraid_sas_fusion.h
index c013c80..8358b68 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.h
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.h
@@ -864,9 +864,20 @@ struct MR_LD_RAID {
 	u8	regTypeReqOnRead;
 	__le16     seqNum;
 
-	struct {
-		u32 ldSyncRequired:1;
-		u32 reserved:31;
+struct {
+#ifndef MFI_BIG_ENDIAN
+	u32 ldSyncRequired:1;
+	u32 regTypeReqOnReadIsValid:1;
+	u32 isEPD:1;
+	u32 enableSLDOnAllRWIOs:1;
+	u32 reserved:28;
+#else
+	u32 reserved:28;
+	u32 enableSLDOnAllRWIOs:1;
+	u32 isEPD:1;
+	u32 regTypeReqOnReadIsValid:1;
+	u32 ldSyncRequired:1;
+#endif
 	} flags;
 
 	u8	LUN[8]; /* 0x24 8 byte LUN field used for SCSI IO's */
-- 
1.8.3.1

[PATCH 03/11] megaraid_sas: Update queue_depth of SAS and NVMe devices

[Anand Lodnoor]

Signed-off-by: Chandrakanth Patil <chandrakanth.patil@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas.h b/drivers/scsi/megaraid/megaraid_sas.h
index bd81840..ddfbe6f 100644
--- a/drivers/scsi/megaraid/megaraid_sas.h
+++ b/drivers/scsi/megaraid/megaraid_sas.h
@@ -2233,9 +2233,9 @@ enum MR_PD_TYPE {
 
 /* JBOD Queue depth definitions */
 #define MEGASAS_SATA_QD	32
-#define MEGASAS_SAS_QD	64
+#define MEGASAS_SAS_QD 256
 #define MEGASAS_DEFAULT_PD_QD	64
-#define MEGASAS_NVME_QD		32
+#define MEGASAS_NVME_QD        64
 
 #define MR_DEFAULT_NVME_PAGE_SIZE	4096
 #define MR_DEFAULT_NVME_PAGE_SHIFT	12
-- 
1.8.3.1

Re: [PATCH 03/11] megaraid_sas: Update queue_depth of SAS and NVMe devices

[Martin K. Petersen]

Anand,

>  /* JBOD Queue depth definitions */
>  #define MEGASAS_SATA_QD	32
> -#define MEGASAS_SAS_QD	64
> +#define MEGASAS_SAS_QD 256
>  #define MEGASAS_DEFAULT_PD_QD	64
> -#define MEGASAS_NVME_QD		32
> +#define MEGASAS_NVME_QD        64
>  
>  #define MR_DEFAULT_NVME_PAGE_SIZE	4096
>  #define MR_DEFAULT_NVME_PAGE_SHIFT	12

Please justify this change.

-- 
Martin K. Petersen	Oracle Linux Engineering

=?y?q?=5BPATCH=2004/11=5D=20megaraid=5Fsas=3A=20Don=E2=80=99t=20kill=20already=20dead=20adapter?=

[Anand Lodnoor]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=y, Size: 1083 bytes --]

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_base.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index 3cd269f..87bdcd6 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -2154,6 +2154,13 @@ static void megasas_complete_outstanding_ioctls(struct megasas_instance *instanc
 
 void megaraid_sas_kill_hba(struct megasas_instance *instance)
 {
+	/* If adapter is already declared dead, do not issue kill HBA again */
+	if (atomic_read(&instance->adprecovery) == MEGASAS_HW_CRITICAL_ERROR) {
+		dev_warn(&instance->pdev->dev,
+			 "Adapter already dead, skipping kill HBA\n");
+		return;
+	}
+
 	/* Set critical error to block I/O & ioctls in case caller didn't */
 	atomic_set(&instance->adprecovery, MEGASAS_HW_CRITICAL_ERROR);
 	/* Wait 1 second to ensure IO or ioctls in build have posted */
-- 
1.8.3.1

Re: .PATCH.04/11.megaraid.sas.Don.t.kill.already.dead.adapter

[Martin K. Petersen]

Anand,

This patch got mangled in transit.

-- 
Martin K. Petersen	Oracle Linux Engineering

[PATCH 05/11] megaraid_sas: Do not kill HBA if JBOD Seqence map or RAID map is disabled

[Anand Lodnoor]

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_fusion.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index e301458..ef20234 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -1312,7 +1312,9 @@ static int megasas_create_sg_sense_fusion(struct megasas_instance *instance)
 	}
 
 	if (ret == DCMD_TIMEOUT)
-		megaraid_sas_kill_hba(instance);
+		dev_warn(&instance->pdev->dev,
+			 "%s DCMD timed out, continue without JBOD sequence map\n",
+			 __func__);
 
 	if (ret == DCMD_SUCCESS)
 		instance->pd_seq_map_id++;
@@ -1394,7 +1396,9 @@ static int megasas_create_sg_sense_fusion(struct megasas_instance *instance)
 		ret = megasas_issue_polled(instance, cmd);
 
 	if (ret == DCMD_TIMEOUT)
-		megaraid_sas_kill_hba(instance);
+		dev_warn(&instance->pdev->dev,
+			 "%s DCMD timed out, RAID map is disabled\n",
+			 __func__);
 
 	megasas_return_cmd(instance, cmd);
 
-- 
1.8.3.1

[PATCH 06/11] megaraid_sas: Do not set HBA Operational if FW is not in operational state

[Anand Lodnoor]

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_fusion.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index ef20234..6860fd2 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -4991,6 +4991,15 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int reason)
 				megasas_set_dynamic_target_properties(sdev, is_target_prop);
 			}
 
+			status_reg = instance->instancet->read_fw_status_reg
+					(instance);
+			abs_state = status_reg & MFI_STATE_MASK;
+			if (abs_state != MFI_STATE_OPERATIONAL) {
+				dev_info(&instance->pdev->dev,
+					 "Adapter is not OPERATIONAL, state 0x%x for scsi:%d\n",
+					 abs_state, instance->host->host_no);
+				goto out;
+			}
 			atomic_set(&instance->adprecovery, MEGASAS_HBA_OPERATIONAL);
 
 			dev_info(&instance->pdev->dev,
-- 
1.8.3.1

[PATCH 07/11] megaraid_sas: Re-Define enum DCMD_RETURN_STATUS

[Anand Lodnoor]

DCMD_INIT is introduced to indicate the initial status,
which was earlier set to MFI status and DCMD_BUSY indicates the
resource busy or locked.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas.h      |  9 +++---
 drivers/scsi/megaraid/megaraid_sas_base.c | 50 ++++++++++++++++++-------------
 2 files changed, 34 insertions(+), 25 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas.h b/drivers/scsi/megaraid/megaraid_sas.h
index ddfbe6f..25afa81 100644
--- a/drivers/scsi/megaraid/megaraid_sas.h
+++ b/drivers/scsi/megaraid/megaraid_sas.h
@@ -2640,10 +2640,11 @@ enum MEGASAS_OCR_CAUSE {
 };
 
 enum DCMD_RETURN_STATUS {
-	DCMD_SUCCESS		= 0,
-	DCMD_TIMEOUT		= 1,
-	DCMD_FAILED		= 2,
-	DCMD_NOT_FIRED		= 3,
+	DCMD_SUCCESS    = 0x00,
+	DCMD_TIMEOUT    = 0x01,
+	DCMD_FAILED     = 0x02,
+	DCMD_BUSY       = 0x03,
+	DCMD_INIT       = 0xff,
 };
 
 u8
diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index 87bdcd6..8bc4076 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -1099,7 +1099,7 @@ struct megasas_cmd *megasas_get_cmd(struct megasas_instance
 	if (atomic_read(&instance->adprecovery) == MEGASAS_HW_CRITICAL_ERROR) {
 		dev_err(&instance->pdev->dev, "Failed from %s %d\n",
 			__func__, __LINE__);
-		return DCMD_NOT_FIRED;
+		return DCMD_INIT;
 	}
 
 	instance->instancet->issue_dcmd(instance, cmd);
@@ -1123,19 +1123,19 @@ struct megasas_cmd *megasas_get_cmd(struct megasas_instance
 			  struct megasas_cmd *cmd, int timeout)
 {
 	int ret = 0;
-	cmd->cmd_status_drv = MFI_STAT_INVALID_STATUS;
+	cmd->cmd_status_drv = DCMD_INIT;
 
 	if (atomic_read(&instance->adprecovery) == MEGASAS_HW_CRITICAL_ERROR) {
 		dev_err(&instance->pdev->dev, "Failed from %s %d\n",
 			__func__, __LINE__);
-		return DCMD_NOT_FIRED;
+		return DCMD_INIT;
 	}
 
 	instance->instancet->issue_dcmd(instance, cmd);
 
 	if (timeout) {
 		ret = wait_event_timeout(instance->int_cmd_wait_q,
-				cmd->cmd_status_drv != MFI_STAT_INVALID_STATUS, timeout * HZ);
+		cmd->cmd_status_drv != DCMD_INIT, timeout * HZ);
 		if (!ret) {
 			dev_err(&instance->pdev->dev,
 				"DCMD(opcode: 0x%x) is timed out, func:%s\n",
@@ -1144,10 +1144,9 @@ struct megasas_cmd *megasas_get_cmd(struct megasas_instance
 		}
 	} else
 		wait_event(instance->int_cmd_wait_q,
-				cmd->cmd_status_drv != MFI_STAT_INVALID_STATUS);
+				cmd->cmd_status_drv != DCMD_INIT);
 
-	return (cmd->cmd_status_drv == MFI_STAT_OK) ?
-		DCMD_SUCCESS : DCMD_FAILED;
+	return cmd->cmd_status_drv;
 }
 
 /**
@@ -1190,19 +1189,19 @@ struct megasas_cmd *megasas_get_cmd(struct megasas_instance
 		cpu_to_le32(upper_32_bits(cmd_to_abort->frame_phys_addr));
 
 	cmd->sync_cmd = 1;
-	cmd->cmd_status_drv = MFI_STAT_INVALID_STATUS;
+	cmd->cmd_status_drv = DCMD_INIT;
 
 	if (atomic_read(&instance->adprecovery) == MEGASAS_HW_CRITICAL_ERROR) {
 		dev_err(&instance->pdev->dev, "Failed from %s %d\n",
 			__func__, __LINE__);
-		return DCMD_NOT_FIRED;
+		return DCMD_INIT;
 	}
 
 	instance->instancet->issue_dcmd(instance, cmd);
 
 	if (timeout) {
 		ret = wait_event_timeout(instance->abort_cmd_wait_q,
-				cmd->cmd_status_drv != MFI_STAT_INVALID_STATUS, timeout * HZ);
+		cmd->cmd_status_drv != DCMD_INIT, timeout * HZ);
 		if (!ret) {
 			opcode = cmd_to_abort->frame->dcmd.opcode;
 			dev_err(&instance->pdev->dev,
@@ -1212,13 +1211,12 @@ struct megasas_cmd *megasas_get_cmd(struct megasas_instance
 		}
 	} else
 		wait_event(instance->abort_cmd_wait_q,
-				cmd->cmd_status_drv != MFI_STAT_INVALID_STATUS);
+		cmd->cmd_status_drv != DCMD_INIT);
 
 	cmd->sync_cmd = 0;
 
 	megasas_return_cmd(instance, cmd);
-	return (cmd->cmd_status_drv == MFI_STAT_OK) ?
-		DCMD_SUCCESS : DCMD_FAILED;
+	return cmd->cmd_status_drv;
 }
 
 /**
@@ -2737,7 +2735,7 @@ static int megasas_wait_for_outstanding(struct megasas_instance *instance)
 						"reset queue\n",
 						reset_cmd);
 
-				reset_cmd->cmd_status_drv = MFI_STAT_INVALID_STATUS;
+				reset_cmd->cmd_status_drv = DCMD_INIT;
 				instance->instancet->fire_cmd(instance,
 						reset_cmd->frame_phys_addr,
 						0, instance->reg_set);
@@ -3442,7 +3440,11 @@ static int megasas_reset_target(struct scsi_cmnd *scmd)
 megasas_complete_int_cmd(struct megasas_instance *instance,
 			 struct megasas_cmd *cmd)
 {
-	cmd->cmd_status_drv = cmd->frame->io.cmd_status;
+	if (cmd->cmd_status_drv == DCMD_INIT)
+		cmd->cmd_status_drv =
+		(cmd->frame->io.cmd_status == MFI_STAT_OK) ?
+		DCMD_SUCCESS : DCMD_FAILED;
+
 	wake_up(&instance->int_cmd_wait_q);
 }
 
@@ -3461,7 +3463,7 @@ static int megasas_reset_target(struct scsi_cmnd *scmd)
 {
 	if (cmd->sync_cmd) {
 		cmd->sync_cmd = 0;
-		cmd->cmd_status_drv = 0;
+		cmd->cmd_status_drv = DCMD_SUCCESS;
 		wake_up(&instance->abort_cmd_wait_q);
 	}
 }
@@ -3737,7 +3739,7 @@ static int megasas_reset_target(struct scsi_cmnd *scmd)
 			dev_notice(&instance->pdev->dev, "%p synchronous cmd"
 						"on the internal reset queue,"
 						"issue it again.\n", cmd);
-			cmd->cmd_status_drv = MFI_STAT_INVALID_STATUS;
+			cmd->cmd_status_drv = DCMD_INIT;
 			instance->instancet->fire_cmd(instance,
 							cmd->frame_phys_addr,
 							0, instance->reg_set);
@@ -8076,6 +8078,7 @@ static int megasas_set_crash_dump_params_ioctl(struct megasas_cmd *cmd)
 	dma_addr_t sense_handle;
 	unsigned long *sense_ptr;
 	u32 opcode = 0;
+	int ret = DCMD_SUCCESS;
 
 	memset(kbuff_arr, 0, sizeof(kbuff_arr));
 
@@ -8216,13 +8219,18 @@ static int megasas_set_crash_dump_params_ioctl(struct megasas_cmd *cmd)
 	 * cmd to the SCSI mid-layer
 	 */
 	cmd->sync_cmd = 1;
-	if (megasas_issue_blocked_cmd(instance, cmd, 0) == DCMD_NOT_FIRED) {
+
+	ret = megasas_issue_blocked_cmd(instance, cmd, 0);
+	switch (ret) {
+	case DCMD_INIT:
+	case DCMD_BUSY:
 		cmd->sync_cmd = 0;
 		dev_err(&instance->pdev->dev,
 			"return -EBUSY from %s %d cmd 0x%x opcode 0x%x cmd->cmd_status_drv 0x%x\n",
-			__func__, __LINE__, cmd->frame->hdr.cmd, opcode,
-			cmd->cmd_status_drv);
-		return -EBUSY;
+			 __func__, __LINE__, cmd->frame->hdr.cmd, opcode,
+			 cmd->cmd_status_drv);
+			error = -EBUSY;
+			goto out;
 	}
 
 	cmd->sync_cmd = 0;
-- 
1.8.3.1

[PATCH 08/11] megaraid_sas: Do not initiate OCR if controller is not in ready state

[Anand Lodnoor]

Driver can initiate OCR if DCMD command is timeout, but there is a
deadlock if the driver attempts to invoke another OCR  before the
mutex lock is released from the previous session of OCR.

This patch takes care of the above scenario using new flag
MEGASAS_FUSION_OCR_NOT_POSSIBLE to indicate if OCR is possible or not.

Cc: stable@vger.kernel.org
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_base.c   | 3 ++-
 drivers/scsi/megaraid/megaraid_sas_fusion.c | 3 ++-
 drivers/scsi/megaraid/megaraid_sas_fusion.h | 1 +
 3 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index 8bc4076..da47c8b 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -4404,7 +4404,8 @@ int megasas_alloc_cmds(struct megasas_instance *instance)
 	if (instance->adapter_type == MFI_SERIES)
 		return KILL_ADAPTER;
 	else if (instance->unload ||
-			test_bit(MEGASAS_FUSION_IN_RESET, &instance->reset_flags))
+			test_bit(MEGASAS_FUSION_OCR_NOT_POSSIBLE,
+				 &instance->reset_flags))
 		return IGNORE_TIMEOUT;
 	else
 		return INITIATE_OCR;
diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index 6860fd2..8b6cc1b 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -4851,6 +4851,7 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int reason)
 	if (instance->requestorId && !instance->skip_heartbeat_timer_del)
 		del_timer_sync(&instance->sriov_heartbeat_timer);
 	set_bit(MEGASAS_FUSION_IN_RESET, &instance->reset_flags);
+	set_bit(MEGASAS_FUSION_OCR_NOT_POSSIBLE, &instance->reset_flags);
 	atomic_set(&instance->adprecovery, MEGASAS_ADPRESET_SM_POLLING);
 	instance->instancet->disable_intr(instance);
 	megasas_sync_irqs((unsigned long)instance);
@@ -5059,7 +5060,7 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int reason)
 	instance->skip_heartbeat_timer_del = 1;
 	retval = FAILED;
 out:
-	clear_bit(MEGASAS_FUSION_IN_RESET, &instance->reset_flags);
+	clear_bit(MEGASAS_FUSION_OCR_NOT_POSSIBLE, &instance->reset_flags);
 	mutex_unlock(&instance->reset_mutex);
 	return retval;
 }
diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.h b/drivers/scsi/megaraid/megaraid_sas_fusion.h
index 8358b68..d57ecc7 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.h
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.h
@@ -89,6 +89,7 @@ enum MR_RAID_FLAGS_IO_SUB_TYPE {
 
 #define MEGASAS_FP_CMD_LEN	16
 #define MEGASAS_FUSION_IN_RESET 0
+#define MEGASAS_FUSION_OCR_NOT_POSSIBLE 1
 #define RAID_1_PEER_CMDS 2
 #define JBOD_MAPS_COUNT	2
 #define MEGASAS_REDUCE_QD_COUNT 64
-- 
1.8.3.1

Re: [PATCH 08/11] megaraid_sas: Do not initiate OCR if controller is not in ready state

[Sumit Saxena]

On Sat, Jan 4, 2020 at 11:01 AM Sasha Levin <sashal@kernel.org> wrote:
>
> Hi,
>
> [This is an automated email]
>
> This commit has been processed because it contains a -stable tag.
> The stable tag indicates that it's relevant for the following trees: all
>
> The bot has tested the following trees: v5.4.7, v5.3.18, v4.19.92, v4.14.161, v4.9.207, v4.4.207.
>
> v5.4.7: Build OK!
> v5.3.18: Build OK!
> v4.19.92: Failed to apply! Possible dependencies:
>     Unable to calculate
>
> v4.14.161: Failed to apply! Possible dependencies:
>     Unable to calculate
>
> v4.9.207: Failed to apply! Possible dependencies:
>     45f4f2eb3da3 ("scsi: megaraid_sas: Add new pci device Ids for SAS3.5 Generic Megaraid Controllers")
>     69c337c0f8d7 ("scsi: megaraid_sas: SAS3.5 Generic Megaraid Controllers Fast Path for RAID 1/10 Writes")
>     a73b0a4b5d17 ("scsi: megaraid_sas: Change RAID_1_10_RMW_CMDS to RAID_1_PEER_CMDS and set value to 2")
>     d0fc91d67c59 ("scsi: megaraid_sas: Send SYNCHRONIZE_CACHE for VD to firmware")
>     fdd84e2514b0 ("scsi: megaraid_sas: SAS3.5 Generic Megaraid Controllers Stream Detection and IO Coalescing")
>
> v4.4.207: Failed to apply! Possible dependencies:
>     179ac14291a0 ("megaraid_sas: Reply Descriptor Post Queue (RDPQ) support")
>     18365b138508 ("megaraid_sas: Task management support")
>     2c048351c8e3 ("megaraid_sas: Syncing request flags macro names with firmware")
>     308ec459bc19 ("megaraid_sas: Dual queue depth support")
>     69c337c0f8d7 ("scsi: megaraid_sas: SAS3.5 Generic Megaraid Controllers Fast Path for RAID 1/10 Writes")
>     6d40afbc7d13 ("megaraid_sas: MFI IO timeout handling")
>     8a01a41d8647 ("megaraid_sas: Make adprecovery variable atomic")
>     8f05024cd3db ("megaraid_sas: Fastpath region lock bypass")
>     a73b0a4b5d17 ("scsi: megaraid_sas: Change RAID_1_10_RMW_CMDS to RAID_1_PEER_CMDS and set value to 2")
>
>
> NOTE: The patch will not be queued to stable trees until it is upstream.
>
> How should we proceed with this patch?
Hi Sasha,

Please pick this patch for stable trees where it cleanly gets applied
after it is upstream.
We will backport this patch for stable trees where it does not get
applied cleanly.

Thanks,
Sumit
>
> --
> Thanks,
> Sasha

[PATCH 09/11] megaraid_sas: Return pended IOCTLs after 3 retries

[Anand Lodnoor]

IOCTL/s causing firmware fault may end up in failed controller resets and
finally killing the adapter.

This patch fixes this problem as stated as below:
In OCR sequence, driver will attempt refiring pended IOCTLs upto two times.
If first two attempts fail, then in third attempt driver will return pended
IOCTLs with EBUSY status to application. These changes are done to ensure
if any of pended IOCTLs is causing firmware fault and resulting into OCR
failure, then in last attempt of OCR  driver will refrain firing it to
firmware and saving adapter from being killed due to faulty IOCTL.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Chandrakanth Patil <chandrakanth.patil@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_fusion.c | 58 +++++++++++++++++++++++++++--
 1 file changed, 55 insertions(+), 3 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index 8b6cc1b..0bdd477 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -4223,7 +4223,8 @@ void  megasas_reset_reply_desc(struct megasas_instance *instance)
  * megasas_refire_mgmt_cmd :	Re-fire management commands
  * @instance:				Controller's soft instance
 */
-static void megasas_refire_mgmt_cmd(struct megasas_instance *instance)
+void megasas_refire_mgmt_cmd(struct megasas_instance *instance,
+			     bool return_ioctl)
 {
 	int j;
 	struct megasas_cmd_fusion *cmd_fusion;
@@ -4287,6 +4288,16 @@ static void megasas_refire_mgmt_cmd(struct megasas_instance *instance)
 			break;
 		}
 
+		if (return_ioctl && cmd_mfi->sync_cmd &&
+		    cmd_mfi->frame->hdr.cmd != MFI_CMD_ABORT) {
+			dev_err(&instance->pdev->dev,
+				"return -EBUSY from %s %d cmd 0x%x opcode 0x%x\n",
+				__func__, __LINE__, cmd_mfi->frame->hdr.cmd,
+				le32_to_cpu(cmd_mfi->frame->dcmd.opcode));
+			cmd_mfi->cmd_status_drv = DCMD_BUSY;
+			result = COMPLETE_CMD;
+		}
+
 		switch (result) {
 		case REFIRE_CMD:
 			megasas_fire_cmd_fusion(instance, req_desc);
@@ -4302,6 +4313,37 @@ static void megasas_refire_mgmt_cmd(struct megasas_instance *instance)
 }
 
 /*
+ * megasas_return_polled_cmds: Return polled mode commands back to the pool
+ *			       before initiating an OCR.
+ * @instance:                  Controller's soft instance
+ */
+static void
+megasas_return_polled_cmds(struct megasas_instance *instance)
+{
+	int i;
+	struct megasas_cmd_fusion *cmd_fusion;
+	struct fusion_context *fusion;
+	struct megasas_cmd *cmd_mfi;
+
+	fusion = instance->ctrl_context;
+
+	for (i = instance->max_scsi_cmds; i < instance->max_fw_cmds; i++) {
+		cmd_fusion = fusion->cmd_list[i];
+		cmd_mfi = instance->cmd_list[cmd_fusion->sync_cmd_idx];
+
+		if (cmd_mfi->flags & DRV_DCMD_POLLED_MODE) {
+			if (megasas_dbg_lvl & OCR_DEBUG)
+				dev_info(&instance->pdev->dev,
+					 "%s %d return cmd 0x%x opcode 0x%x\n",
+					 __func__, __LINE__, cmd_mfi->frame->hdr.cmd,
+					 le32_to_cpu(cmd_mfi->frame->dcmd.opcode));
+			cmd_mfi->flags &= ~DRV_DCMD_POLLED_MODE;
+			megasas_return_cmd(instance, cmd_mfi);
+		}
+	}
+}
+
+/*
  * megasas_track_scsiio : Track SCSI IOs outstanding to a SCSI device
  * @instance: per adapter struct
  * @channel: the channel assigned by the OS
@@ -4956,7 +4998,9 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int reason)
 				goto kill_hba;
 			}
 
-			megasas_refire_mgmt_cmd(instance);
+			megasas_refire_mgmt_cmd(instance,
+						(i == (MEGASAS_FUSION_MAX_RESET_TRIES - 1)
+							? 1 : 0));
 
 			/* Reset load balance info */
 			if (fusion->load_balance_info)
@@ -4964,8 +5008,16 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int reason)
 				       (sizeof(struct LD_LOAD_BALANCE_INFO) *
 				       MAX_LOGICAL_DRIVES_EXT));
 
-			if (!megasas_get_map_info(instance))
+			if (!megasas_get_map_info(instance)) {
 				megasas_sync_map_info(instance);
+			} else {
+				/*
+				 * Return pending polled mode cmds before
+				 * retrying OCR
+				 */
+				megasas_return_polled_cmds(instance);
+				continue;
+			}
 
 			megasas_setup_jbod_map(instance);
 
-- 
1.8.3.1

[PATCH 10/11] megaraid_sas: Use Block layer API to check SCSI device in-flight IO requests

[Anand Lodnoor]

Remove usage of device_busy counter from driver. Instead of
device_busy counter now driver uses 'nr_active' counter of
request_queue to get the number of inflight request for a
LUN.

Link : https://patchwork.kernel.org/patch/11249297/
Signed-off-by: Chandrakanth Patil <chandrakanth.patil@broadcom.com>
Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas_fusion.c | 56 ++++++++++++++++-------------
 1 file changed, 31 insertions(+), 25 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index 0bdd477..26b45e8 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -364,6 +364,35 @@ inline void megasas_return_cmd_fusion(struct megasas_instance *instance,
 		instance->max_fw_cmds = instance->max_fw_cmds-1;
 	}
 }
+
+static inline void
+megasas_get_msix_index(struct megasas_instance *instance,
+		       struct scsi_cmnd *scmd,
+		       struct megasas_cmd_fusion *cmd,
+		       u8 data_arms)
+{
+	int sdev_busy;
+
+	/* nr_hw_queue = 1 for MegaRAID */
+	struct blk_mq_hw_ctx *hctx =
+		scmd->device->request_queue->queue_hw_ctx[0];
+
+	sdev_busy = atomic_read(&hctx->nr_active);
+
+	if (instance->perf_mode == MR_BALANCED_PERF_MODE &&
+	    sdev_busy > (data_arms * MR_DEVICE_HIGH_IOPS_DEPTH))
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			mega_mod64((atomic64_add_return(1, &instance->high_iops_outstanding) /
+					MR_HIGH_IOPS_BATCH_COUNT), instance->low_latency_index_start);
+	else if (instance->msix_load_balance)
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			(mega_mod64(atomic64_add_return(1, &instance->total_io_count),
+				instance->msix_vectors));
+	else
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			instance->reply_map[raw_smp_processor_id()];
+}
+
 /**
  * megasas_free_cmds_fusion -	Free all the cmds in the free cmd pool
  * @instance:		Adapter soft state
@@ -2829,19 +2858,7 @@ static void megasas_stream_detect(struct megasas_instance *instance,
 			fp_possible = (io_info.fpOkForIo > 0) ? true : false;
 	}
 
-	if ((instance->perf_mode == MR_BALANCED_PERF_MODE) &&
-		atomic_read(&scp->device->device_busy) >
-		(io_info.data_arms * MR_DEVICE_HIGH_IOPS_DEPTH))
-		cmd->request_desc->SCSIIO.MSIxIndex =
-			mega_mod64((atomic64_add_return(1, &instance->high_iops_outstanding) /
-				MR_HIGH_IOPS_BATCH_COUNT), instance->low_latency_index_start);
-	else if (instance->msix_load_balance)
-		cmd->request_desc->SCSIIO.MSIxIndex =
-			(mega_mod64(atomic64_add_return(1, &instance->total_io_count),
-				    instance->msix_vectors));
-	else
-		cmd->request_desc->SCSIIO.MSIxIndex =
-			instance->reply_map[raw_smp_processor_id()];
+	megasas_get_msix_index(instance, scp, cmd, io_info.data_arms);
 
 	if (instance->adapter_type >= VENTURA_SERIES) {
 		/* FP for Optimal raid level 1.
@@ -3162,18 +3179,7 @@ static void megasas_build_ld_nonrw_fusion(struct megasas_instance *instance,
 
 	cmd->request_desc->SCSIIO.DevHandle = io_request->DevHandle;
 
-	if ((instance->perf_mode == MR_BALANCED_PERF_MODE) &&
-		atomic_read(&scmd->device->device_busy) > MR_DEVICE_HIGH_IOPS_DEPTH)
-		cmd->request_desc->SCSIIO.MSIxIndex =
-			mega_mod64((atomic64_add_return(1, &instance->high_iops_outstanding) /
-				MR_HIGH_IOPS_BATCH_COUNT), instance->low_latency_index_start);
-	else if (instance->msix_load_balance)
-		cmd->request_desc->SCSIIO.MSIxIndex =
-			(mega_mod64(atomic64_add_return(1, &instance->total_io_count),
-				    instance->msix_vectors));
-	else
-		cmd->request_desc->SCSIIO.MSIxIndex =
-			instance->reply_map[raw_smp_processor_id()];
+	megasas_get_msix_index(instance, scmd, cmd, 1);
 
 	if (!fp_possible) {
 		/* system pd firmware path */
-- 
1.8.3.1

[PATCH 11/11] megaraid_sas: Update driver version to 07.713.01.00-rc1

[Anand Lodnoor]

Signed-off-by: Anand Lodnoor <anand.lodnoor@broadcom.com>
---
 drivers/scsi/megaraid/megaraid_sas.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas.h b/drivers/scsi/megaraid/megaraid_sas.h
index 25afa81..83d8c4c 100644
--- a/drivers/scsi/megaraid/megaraid_sas.h
+++ b/drivers/scsi/megaraid/megaraid_sas.h
@@ -21,8 +21,8 @@
 /*
  * MegaRAID SAS Driver meta data
  */
-#define MEGASAS_VERSION				"07.710.50.00-rc1"
-#define MEGASAS_RELDATE				"June 28, 2019"
+#define MEGASAS_VERSION				"07.713.01.00-rc1"
+#define MEGASAS_RELDATE				"Dec 27, 2019"
 
 #define MEGASAS_MSIX_NAME_LEN			32
 
-- 
1.8.3.1