Commit a1f6e021 authored by nickcheng(鄭守謙's avatar nickcheng(鄭守謙 Committed by James Bottomley
Browse files

[SCSI] areca: improve driver stability and compatibility 



Description:

    1. Implement PCI-Express error recovery function and AER
       capability, especially thanks to Yanmin Zhang's openhanded help
       about AER

    2. Implement the selection of ARCMSR_MAX_XFER_SECTORS_B=4096 if
       firmware version is latter than 1.42

    3. Add arcmsr_done4_abort_postqueue in arcmsr_iop_reset function
       to improve the stability as hot-unplug/plug

    4. Modify the ISR, arcmsr_interrupt routine, to prevent the
       inconsistency with sg_mod driver if application directly calls
       the arcmsr driver w/o passing through scsi midlayer

Signed-off-by: default avatarNick Cheng <nick.cheng@areca.com.tw>
[jejb: unused variable removal]
Signed-off-by: default avatarJames Bottomley <James.Bottomley@SteelEye.com>
parent 727eead6

drivers/scsi/arcmsr/arcmsr.h

+2 −2
Original line number Diff line number Diff line
@@ -48,9 +48,10 @@ struct class_device_attribute; 


#define ARCMSR_MAX_OUTSTANDING_CMD 						256

#define ARCMSR_MAX_FREECCB_NUM							288

#define ARCMSR_DRIVER_VERSION				"Driver Version 1.20.00.13"

#define ARCMSR_DRIVER_VERSION				"Driver Version 1.20.00.14"

#define ARCMSR_SCSI_INITIATOR_ID						255

#define ARCMSR_MAX_XFER_SECTORS							512

#define ARCMSR_MAX_XFER_SECTORS_B                                              4096

#define ARCMSR_MAX_TARGETID							 17

#define ARCMSR_MAX_TARGETLUN							  8

#define ARCMSR_MAX_CMD_PERLUN				 ARCMSR_MAX_OUTSTANDING_CMD
@@ -469,4 +470,3 @@ extern void arcmsr_post_Qbuffer(struct AdapterControlBlock *acb); 
extern struct class_device_attribute *arcmsr_host_attrs[];

extern int arcmsr_alloc_sysfs_attr(struct AdapterControlBlock *acb);

void arcmsr_free_sysfs_attr(struct AdapterControlBlock *acb);

drivers/scsi/arcmsr/arcmsr_hba.c

+423 −33
Original line number Diff line number Diff line
@@ -57,6 +57,7 @@ 
#include <linux/dma-mapping.h>

#include <linux/timer.h>

#include <linux/pci.h>

#include <linux/aer.h>

#include <asm/dma.h>

#include <asm/io.h>

#include <asm/system.h>
@@ -71,7 +72,7 @@ 
#include "arcmsr.h"



MODULE_AUTHOR("Erich Chen <erich@areca.com.tw>");

MODULE_DESCRIPTION("ARECA (ARC11xx/12xx) SATA RAID HOST Adapter");

MODULE_DESCRIPTION("ARECA (ARC11xx/12xx/13xx/16xx) SATA/SAS RAID HOST Adapter");

MODULE_LICENSE("Dual BSD/GPL");

MODULE_VERSION(ARCMSR_DRIVER_VERSION);
@@ -93,7 +94,9 @@ static void arcmsr_flush_adapter_cache(struct AdapterControlBlock *acb); 
static uint8_t arcmsr_wait_msgint_ready(struct AdapterControlBlock *acb);

static const char *arcmsr_info(struct Scsi_Host *);

static irqreturn_t arcmsr_interrupt(struct AdapterControlBlock *acb);



static pci_ers_result_t arcmsr_pci_error_detected(struct pci_dev *pdev,

						pci_channel_state_t state);

static pci_ers_result_t arcmsr_pci_slot_reset(struct pci_dev *pdev);

static int arcmsr_adjust_disk_queue_depth(struct scsi_device *sdev, int queue_depth)

{

	if (queue_depth > ARCMSR_MAX_CMD_PERLUN)
@@ -104,7 +107,8 @@ static int arcmsr_adjust_disk_queue_depth(struct scsi_device *sdev, int queue_de 


static struct scsi_host_template arcmsr_scsi_host_template = {

	.module			= THIS_MODULE,

	.name			= "ARCMSR ARECA SATA RAID HOST Adapter" ARCMSR_DRIVER_VERSION,

	.name			= "ARCMSR ARECA SATA/SAS RAID HOST Adapter"

							ARCMSR_DRIVER_VERSION,

	.info			= arcmsr_info,

	.queuecommand		= arcmsr_queue_command,

	.eh_abort_handler	= arcmsr_abort,
@@ -119,6 +123,10 @@ static struct scsi_host_template arcmsr_scsi_host_template = { 
	.use_clustering		= ENABLE_CLUSTERING,

	.shost_attrs		= arcmsr_host_attrs,

};

static struct pci_error_handlers arcmsr_pci_error_handlers = {

	.error_detected		= arcmsr_pci_error_detected,

	.slot_reset		= arcmsr_pci_slot_reset,

};



static struct pci_device_id arcmsr_device_id_table[] = {

	{PCI_DEVICE(PCI_VENDOR_ID_ARECA, PCI_DEVICE_ID_ARECA_1110)},
@@ -144,7 +152,8 @@ static struct pci_driver arcmsr_pci_driver = { 
	.id_table		= arcmsr_device_id_table,

	.probe			= arcmsr_probe,

	.remove			= arcmsr_remove,

	.shutdown		= arcmsr_shutdown

	.shutdown		= arcmsr_shutdown,

	.err_handler		= &arcmsr_pci_error_handlers,

};



static irqreturn_t arcmsr_do_interrupt(int irq, void *dev_id)
@@ -328,6 +337,8 @@ static int arcmsr_probe(struct pci_dev *pdev, 


	arcmsr_iop_init(acb);

	pci_set_drvdata(pdev, host);

	if (strncmp(acb->firm_version, "V1.42", 5) >= 0)

		host->max_sectors= ARCMSR_MAX_XFER_SECTORS_B;



	error = scsi_add_host(host, &pdev->dev);

	if (error)
@@ -338,6 +349,7 @@ static int arcmsr_probe(struct pci_dev *pdev, 
		goto out_free_sysfs;



	scsi_scan_host(host);

	pci_enable_pcie_error_reporting(pdev);

	return 0;

 out_free_sysfs:

 out_free_irq:
@@ -762,7 +774,7 @@ static irqreturn_t arcmsr_interrupt(struct AdapterControlBlock *acb) 
				switch(ccb->arcmsr_cdb.DeviceStatus) {

				case ARCMSR_DEV_SELECT_TIMEOUT: {

						acb->devstate[id][lun] = ARECA_RAID_GONE;

						ccb->pcmd->result = DID_TIME_OUT << 16;

						ccb->pcmd->result = DID_NO_CONNECT << 16;

						arcmsr_ccb_complete(ccb, 1);

					}

					break;
@@ -1247,7 +1259,7 @@ static void arcmsr_polling_ccbdone(struct AdapterControlBlock *acb, 
			switch(ccb->arcmsr_cdb.DeviceStatus) {

			case ARCMSR_DEV_SELECT_TIMEOUT: {

					acb->devstate[id][lun] = ARECA_RAID_GONE;

					ccb->pcmd->result = DID_TIME_OUT << 16;

					ccb->pcmd->result = DID_NO_CONNECT << 16;

					arcmsr_ccb_complete(ccb, 1);

				}

				break;
@@ -1281,6 +1293,94 @@ static void arcmsr_polling_ccbdone(struct AdapterControlBlock *acb, 
		}

	}

}

static void arcmsr_done4_abort_postqueue(struct AdapterControlBlock *acb)

{

	int i = 0, found = 0;

	int id, lun;

	uint32_t flag_ccb, outbound_intstatus;

	struct MessageUnit __iomem *reg = acb->pmu;

	struct CommandControlBlock *ccb;

	/*clear and abort all outbound posted Q*/



	while (((flag_ccb = readl(&reg->outbound_queueport)) != 0xFFFFFFFF) &&

(i++ < 256)){

		ccb = (struct CommandControlBlock *)(acb->vir2phy_offset +

(flag_ccb << 5));

	if (ccb){

		if ((ccb->acb != acb)||(ccb->startdone != \

ARCMSR_CCB_START)){

				printk(KERN_NOTICE "arcmsr%d: polling get \

an illegal ccb" "command done ccb = '0x%p'""ccboutstandingcount = %d \n",

					acb->host->host_no, ccb,

					atomic_read(&acb->ccboutstandingcount));

				continue;

			}



			id = ccb->pcmd->device->id;

			lun = ccb->pcmd->device->lun;

			if (!(flag_ccb & ARCMSR_CCBREPLY_FLAG_ERROR)){

				if (acb->devstate[id][lun] == ARECA_RAID_GONE)

					acb->devstate[id][lun] = ARECA_RAID_GOOD;

				ccb->pcmd->result = DID_OK << 16;

				arcmsr_ccb_complete(ccb, 1);

			}

			else {

				switch(ccb->arcmsr_cdb.DeviceStatus) {

				case ARCMSR_DEV_SELECT_TIMEOUT: {

						acb->devstate[id][lun] = ARECA_RAID_GONE;

						ccb->pcmd->result = DID_NO_CONNECT << 16;

						arcmsr_ccb_complete(ccb, 1);

				}

				break;



				case ARCMSR_DEV_ABORTED:



				case ARCMSR_DEV_INIT_FAIL: {

						acb->devstate[id][lun] =

							ARECA_RAID_GONE;

						ccb->pcmd->result =

							DID_BAD_TARGET << 16;

				arcmsr_ccb_complete(ccb, 1);

				}

				break;



				case ARCMSR_DEV_CHECK_CONDITION: {

						acb->devstate[id][lun] =

							ARECA_RAID_GOOD;

						arcmsr_report_sense_info(ccb);

						arcmsr_ccb_complete(ccb, 1);

				}

				break;



				default:

						printk(KERN_NOTICE

						      "arcmsr%d: scsi id = %d \

							lun = %d""polling and \

							getting command error \

							done""but got unknown \

							DeviceStatus = 0x%x \n",

							acb->host->host_no, id,

					   lun, ccb->arcmsr_cdb.DeviceStatus);

						acb->devstate[id][lun] =

								ARECA_RAID_GONE;

						ccb->pcmd->result =

							DID_BAD_TARGET << 16;

						arcmsr_ccb_complete(ccb, 1);

				break;

			       }

	}

		       found = 1;

	       }

	}

	if (found){

		outbound_intstatus = readl(&reg->outbound_intstatus) & \

			acb->outbound_int_enable;

		writel(outbound_intstatus, &reg->outbound_intstatus);

		/*clear interrupt*/

	}

	return;

}





static void arcmsr_iop_init(struct AdapterControlBlock *acb)

{
@@ -1314,7 +1414,6 @@ static void arcmsr_iop_init(struct AdapterControlBlock *acb) 


static void arcmsr_iop_reset(struct AdapterControlBlock *acb)

{

	struct MessageUnit __iomem *reg = acb->pmu;

	struct CommandControlBlock *ccb;

	uint32_t intmask_org;

	int i = 0;
@@ -1327,21 +1426,17 @@ static void arcmsr_iop_reset(struct AdapterControlBlock *acb) 
		/* disable all outbound interrupt */

		intmask_org = arcmsr_disable_outbound_ints(acb);

		/* clear all outbound posted Q */

		for (i = 0; i < ARCMSR_MAX_OUTSTANDING_CMD; i++)

			readl(&reg->outbound_queueport);

		arcmsr_done4_abort_postqueue(acb);

		for (i = 0; i < ARCMSR_MAX_FREECCB_NUM; i++) {

			ccb = acb->pccb_pool[i];

			if ((ccb->startdone == ARCMSR_CCB_START) ||

				(ccb->startdone == ARCMSR_CCB_ABORTED)) {

			if (ccb->startdone == ARCMSR_CCB_START) {

				ccb->startdone = ARCMSR_CCB_ABORTED;

				ccb->pcmd->result = DID_ABORT << 16;

				arcmsr_ccb_complete(ccb, 1);

			}

		}

		/* enable all outbound interrupt */

		arcmsr_enable_outbound_ints(acb, intmask_org);

	}

	atomic_set(&acb->ccboutstandingcount, 0);



}



static int arcmsr_bus_reset(struct scsi_cmnd *cmd)
@@ -1390,7 +1485,6 @@ static int arcmsr_abort(struct scsi_cmnd *cmd) 
		"arcmsr%d: abort device command of scsi id = %d lun = %d \n",

		acb->host->host_no, cmd->device->id, cmd->device->lun);

	acb->num_aborts++;



	/*

	************************************************

	** the all interrupt service routine is locked
@@ -1451,4 +1545,300 @@ static const char *arcmsr_info(struct Scsi_Host *host) 
	return buf;

}



static pci_ers_result_t arcmsr_pci_slot_reset(struct pci_dev *pdev)

{

	struct Scsi_Host *host;

	struct AdapterControlBlock *acb;

	uint8_t bus, dev_fun;

	int error;



	error = pci_enable_device(pdev);

	if (error)

		return PCI_ERS_RESULT_DISCONNECT;

	pci_set_master(pdev);



	host = scsi_host_alloc(&arcmsr_scsi_host_template, sizeof \

(struct AdapterControlBlock));

	if (!host)

		return PCI_ERS_RESULT_DISCONNECT;

	acb = (struct AdapterControlBlock *)host->hostdata;

	memset(acb, 0, sizeof (struct AdapterControlBlock));



	error = pci_set_dma_mask(pdev, DMA_64BIT_MASK);

	if (error) {

		error = pci_set_dma_mask(pdev, DMA_32BIT_MASK);

		if (error) {

			printk(KERN_WARNING

			       "scsi%d: No suitable DMA mask available\n",

			       host->host_no);

			return PCI_ERS_RESULT_DISCONNECT;

		}

	}

	bus = pdev->bus->number;

	dev_fun = pdev->devfn;

	acb = (struct AdapterControlBlock *) host->hostdata;

	memset(acb, 0, sizeof(struct AdapterControlBlock));

	acb->pdev = pdev;

	acb->host = host;

	host->max_sectors = ARCMSR_MAX_XFER_SECTORS;

	host->max_lun = ARCMSR_MAX_TARGETLUN;

	host->max_id = ARCMSR_MAX_TARGETID;/*16:8*/

	host->max_cmd_len = 16;    /*this is issue of 64bit LBA, over 2T byte*/

	host->sg_tablesize = ARCMSR_MAX_SG_ENTRIES;

	host->can_queue = ARCMSR_MAX_FREECCB_NUM; /* max simultaneous cmds */

	host->cmd_per_lun = ARCMSR_MAX_CMD_PERLUN;

	host->this_id = ARCMSR_SCSI_INITIATOR_ID;

	host->unique_id = (bus << 8) | dev_fun;

	host->irq = pdev->irq;

	error = pci_request_regions(pdev, "arcmsr");

	if (error)

		return PCI_ERS_RESULT_DISCONNECT;



	acb->pmu = ioremap(pci_resource_start(pdev, 0),

			   pci_resource_len(pdev, 0));

	if (!acb->pmu) {

		printk(KERN_NOTICE "arcmsr%d: memory"

			" mapping region fail \n", acb->host->host_no);

		return PCI_ERS_RESULT_DISCONNECT;

	}

	acb->acb_flags |= (ACB_F_MESSAGE_WQBUFFER_CLEARED |

			   ACB_F_MESSAGE_RQBUFFER_CLEARED |

			   ACB_F_MESSAGE_WQBUFFER_READED);

	acb->acb_flags &= ~ACB_F_SCSISTOPADAPTER;

	INIT_LIST_HEAD(&acb->ccb_free_list);



	error = arcmsr_alloc_ccb_pool(acb);

	if (error)

		return PCI_ERS_RESULT_DISCONNECT;



	error = request_irq(pdev->irq, arcmsr_do_interrupt,

			IRQF_DISABLED | IRQF_SHARED, "arcmsr", acb);

	if (error)

		return PCI_ERS_RESULT_DISCONNECT;



	arcmsr_iop_init(acb);

	if (strncmp(acb->firm_version, "V1.42", 5) >= 0)

	      host->max_sectors = ARCMSR_MAX_XFER_SECTORS_B;



	pci_set_drvdata(pdev, host);



	error = scsi_add_host(host, &pdev->dev);

	if (error)

		return PCI_ERS_RESULT_DISCONNECT;



	error = arcmsr_alloc_sysfs_attr(acb);

	if (error)

		return PCI_ERS_RESULT_DISCONNECT;



	scsi_scan_host(host);

	return PCI_ERS_RESULT_RECOVERED;

}



static void arcmsr_pci_ers_need_reset_forepart(struct pci_dev *pdev)

{

	struct Scsi_Host *host = pci_get_drvdata(pdev);

	struct AdapterControlBlock *acb = (struct AdapterControlBlock *) host->hostdata;

	struct MessageUnit __iomem *reg = acb->pmu;

	struct CommandControlBlock *ccb;

	/*clear and abort all outbound posted Q*/

	int i = 0, found = 0;

	int id, lun;

	uint32_t flag_ccb, outbound_intstatus;



	while (((flag_ccb = readl(&reg->outbound_queueport)) != 0xFFFFFFFF) &&

								(i++ < 256)){

			ccb = (struct CommandControlBlock *)(acb->vir2phy_offset

							 + (flag_ccb << 5));

			if (ccb){

				if ((ccb->acb != acb)||(ccb->startdone !=

							ARCMSR_CCB_START)){

					printk(KERN_NOTICE "arcmsr%d: polling \

					get an illegal ccb"" command done ccb = '0x%p'"

					"ccboutstandingcount = %d \n",

					acb->host->host_no, ccb,

					atomic_read(&acb->ccboutstandingcount));

					continue;

				}



				id = ccb->pcmd->device->id;

				lun = ccb->pcmd->device->lun;

				if (!(flag_ccb & ARCMSR_CCBREPLY_FLAG_ERROR)) {

					if (acb->devstate[id][lun] ==

								ARECA_RAID_GONE)

						acb->devstate[id][lun] =

								ARECA_RAID_GOOD;

					ccb->pcmd->result = DID_OK << 16;

					arcmsr_ccb_complete(ccb, 1);

				}

				else {

					switch(ccb->arcmsr_cdb.DeviceStatus) {

					case ARCMSR_DEV_SELECT_TIMEOUT: {

							acb->devstate[id][lun] =

							ARECA_RAID_GONE;

							ccb->pcmd->result =

							DID_NO_CONNECT << 16;

							arcmsr_ccb_complete(ccb, 1);

					}

					break;



					case ARCMSR_DEV_ABORTED:



					case ARCMSR_DEV_INIT_FAIL: {

							acb->devstate[id][lun] =

							 ARECA_RAID_GONE;

							ccb->pcmd->result =

							DID_BAD_TARGET << 16;

							arcmsr_ccb_complete(ccb, 1);

					}

					break;



					case ARCMSR_DEV_CHECK_CONDITION: {

							acb->devstate[id][lun] =

							 ARECA_RAID_GOOD;

							arcmsr_report_sense_info(ccb);

							arcmsr_ccb_complete(ccb, 1);

					}

					break;



					default:

							printk(KERN_NOTICE

								"arcmsr%d: scsi \

								id = %d lun = %d"

								" polling and \

								getting command \

								error done"

								"but got unknown \

							DeviceStatus = 0x%x \n"

							, acb->host->host_no,

								id, lun,

						ccb->arcmsr_cdb.DeviceStatus);

							acb->devstate[id][lun] =

								ARECA_RAID_GONE;

							ccb->pcmd->result =

							DID_BAD_TARGET << 16;

							arcmsr_ccb_complete(ccb, 1);

					break;

					}

				}

				found = 1;

			}

		}

	if (found){

		outbound_intstatus = readl(&reg->outbound_intstatus) &

							acb->outbound_int_enable;

		writel(outbound_intstatus, &reg->outbound_intstatus);

		/*clear interrupt*/

		    }

	return;

}





static void arcmsr_pci_ers_disconnect_forepart(struct pci_dev *pdev)

{

	struct Scsi_Host *host = pci_get_drvdata(pdev);

	struct AdapterControlBlock *acb = (struct AdapterControlBlock *) host->hostdata;

	struct MessageUnit __iomem *reg = acb->pmu;

	struct CommandControlBlock *ccb;

	/*clear and abort all outbound posted Q*/

	int i = 0, found = 0;

	int id, lun;

	uint32_t flag_ccb, outbound_intstatus;



	while (((flag_ccb = readl(&reg->outbound_queueport)) != 0xFFFFFFFF) &&

								(i++ < 256)){

			ccb = (struct CommandControlBlock *)(acb->vir2phy_offset +

							(flag_ccb << 5));

			if (ccb){

				if ((ccb->acb != acb)||(ccb->startdone !=

							ARCMSR_CCB_START)){

					printk(KERN_NOTICE

						"arcmsr%d: polling get an illegal ccb"

						" command done ccb = '0x%p'"

						"ccboutstandingcount = %d \n",

						acb->host->host_no, ccb,

						atomic_read(&acb->ccboutstandingcount));

					continue;

			}



			id = ccb->pcmd->device->id;

			lun = ccb->pcmd->device->lun;

			if (!(flag_ccb & ARCMSR_CCBREPLY_FLAG_ERROR))	{

				if (acb->devstate[id][lun] == ARECA_RAID_GONE)

					acb->devstate[id][lun] = ARECA_RAID_GOOD;

				ccb->pcmd->result = DID_OK << 16;

				arcmsr_ccb_complete(ccb, 1);

			}

			else {

				switch(ccb->arcmsr_cdb.DeviceStatus) {

				case ARCMSR_DEV_SELECT_TIMEOUT: {

						acb->devstate[id][lun] =

								ARECA_RAID_GONE;

						ccb->pcmd->result =

							DID_NO_CONNECT << 16;

						arcmsr_ccb_complete(ccb, 1);

				}

				break;



				case ARCMSR_DEV_ABORTED:



				case ARCMSR_DEV_INIT_FAIL: {

						acb->devstate[id][lun] =

								ARECA_RAID_GONE;

						ccb->pcmd->result =

							DID_BAD_TARGET << 16;

						arcmsr_ccb_complete(ccb, 1);

				}

				break;



				case ARCMSR_DEV_CHECK_CONDITION: {

						acb->devstate[id][lun] =

								ARECA_RAID_GOOD;

						arcmsr_report_sense_info(ccb);

						arcmsr_ccb_complete(ccb, 1);

				}

				break;



				default:

						printk(KERN_NOTICE "arcmsr%d: \

							scsi id = %d lun = %d"

								" polling and \

						getting command error done"

								"but got unknown \

						 DeviceStatus = 0x%x \n"

								, acb->host->host_no,

					id, lun, ccb->arcmsr_cdb.DeviceStatus);

							acb->devstate[id][lun] =

								ARECA_RAID_GONE;

							ccb->pcmd->result =

							DID_BAD_TARGET << 16;

							arcmsr_ccb_complete(ccb, 1);

				break;

				}

			}

			found = 1;

		}

	}

	if (found){

		outbound_intstatus = readl(&reg->outbound_intstatus) &

						acb->outbound_int_enable;

		writel(outbound_intstatus, &reg->outbound_intstatus);

		/*clear interrupt*/

	}

	return;

}



static pci_ers_result_t arcmsr_pci_error_detected(struct pci_dev *pdev,

						pci_channel_state_t state)

{

	switch (state) {

	case pci_channel_io_frozen:

			arcmsr_pci_ers_need_reset_forepart(pdev);

			return PCI_ERS_RESULT_NEED_RESET;

	case pci_channel_io_perm_failure:

			arcmsr_pci_ers_disconnect_forepart(pdev);

			return PCI_ERS_RESULT_DISCONNECT;

			break;

	default:

			return PCI_ERS_RESULT_NEED_RESET;

	}

}

CRA Git | Maintained and supported by SUSTech CRA and CCSE