State of big-endian for MegaRAID SAS driver

State of big-endian for MegaRAID SAS driver

[Ben Collins]

Hi folks,

Based on some searching, back when the megaraid_sas driver was first introduced in 2005, issues were raised with the lack of support for big-endian architectures[1]. In this conversation, it was said that LSI would revisit this later, and thus the driver was included without support for big-endian architectures. Here it is 8 years later, and I'm wondering if this will ever get resolved.

The platform in question is a 32-bit powerpc system. I've already fixed one issue with regard to inaccurate assumptions about pci resource sizes (attached). Basically, it is possible for 32-bit systems to support 64-bit PCI resources (as is the case on this system). I'll submit it properly at some point, or feel free to snip it off from here. I've already fixed the aacraid SRC support on this hardware (same resource size assumptions and big-endian fixes), and I'm hoping to do the same with megaraid_sas (at least as it pertains to the gen2 card).

I notice that LSI does support a Solaris/SPARCv9 driver for mr_sas, so I know that there is at least some base knowledge and working expertise on supporting big-endian. Any chance of the Solaris driver source being tossed out to get a look a what it will take to get the Linux driver in working order.

[1]: http://lkml.indiana.edu/hypermail/linux/kernel/0506.0/1525.html

megaraid_sas: Fix assumptions about PCI resource sizes

Signed-off-by: Ben Collins <ben.c@servergy.com>

diff --git a/drivers/scsi/megaraid/megaraid_sas.h b/drivers/scsi/megaraid/megaraid_sas.h
index 04a42a5..f509d14 100644
--- a/drivers/scsi/megaraid/megaraid_sas.h
+++ b/drivers/scsi/megaraid/megaraid_sas.h
@@ -1460,7 +1460,6 @@ struct megasas_instance {
 	u32 *reply_queue;
 	dma_addr_t reply_queue_h;
 
-	unsigned long base_addr;
 	struct megasas_register_set __iomem *reg_set;
 	u32 *reply_post_host_index_addr[MR_MAX_MSIX_REG_ARRAY];
 	struct megasas_pd_list          pd_list[MEGASAS_MAX_PD];
diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index 1f0ca68..5df5522 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -3498,6 +3498,7 @@ static int megasas_init_fw(struct megasas_instance *instance)
 	u32 max_sectors_1;
 	u32 max_sectors_2;
 	u32 tmp_sectors, msix_enable, scratch_pad_2;
+	resource_size_t base_addr;
 	struct megasas_register_set __iomem *reg_set;
 	struct megasas_ctrl_info *ctrl_info;
 	unsigned long bar_list;
@@ -3506,14 +3507,14 @@ static int megasas_init_fw(struct megasas_instance *instance)
 	/* Find first memory bar */
 	bar_list = pci_select_bars(instance->pdev, IORESOURCE_MEM);
 	instance->bar = find_first_bit(&bar_list, sizeof(unsigned long));
-	instance->base_addr = pci_resource_start(instance->pdev, instance->bar);
 	if (pci_request_selected_regions(instance->pdev, instance->bar,
 					 "megasas: LSI")) {
 		printk(KERN_DEBUG "megasas: IO memory region busy!\n");
 		return -EBUSY;
 	}
 
-	instance->reg_set = ioremap_nocache(instance->base_addr, 8192);
+	base_addr = pci_resource_start(instance->pdev, instance->bar);
+	instance->reg_set = ioremap_nocache(base_addr, 8192);
 
 	if (!instance->reg_set) {
 		printk(KERN_DEBUG "megasas: Failed to map IO mem\n");


--
Ben Collins - Principal Architect
Servergy, Inc. - www.servergy.com

Re: State of big-endian for MegaRAID SAS driver

[Martin K. Petersen]

>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:

Ben> Based on some searching, back when the megaraid_sas driver was
Ben> first introduced in 2005, issues were raised with the lack of
Ben> support for big-endian architectures[1]. In this conversation, it
Ben> was said that LSI would revisit this later, and thus the driver was
Ben> included without support for big-endian architectures. Here it is 8
Ben> years later, and I'm wondering if this will ever get resolved.

I have megaraid working on big endian. PoC patch against my slightly
outdated working tree attached. Happy to forward port if there's
interest?

-- 
Martin K. Petersen	Oracle Linux Engineering

commit 906b6af65cde77a13641a476b55879424bfaae24
Author: Martin K. Petersen <martin.petersen@oracle.com>
Date:   Sat Aug 10 17:55:58 2013 -0400

    megaraid_sas: Big endian support
    
    This driver only worked on a little endian system. Make sure that all
    frames submitted to the hardware are converted to little endian.
    
    Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index a7b0876..69f6ce0 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -373,12 +373,9 @@ megasas_adp_reset_xscale(struct megasas_instance *instance, struct megasas_regis
 static int 
 megasas_check_reset_xscale(struct megasas_instance *instance, struct megasas_register_set __iomem * regs)
 {
-	u32 consumer;
-	consumer = *instance->consumer;
-
-	if ((instance->adprecovery != MEGASAS_HBA_OPERATIONAL) && (*instance->consumer == MEGASAS_ADPRESET_INPROG_SIGN)) {
+	if ((instance->adprecovery != MEGASAS_HBA_OPERATIONAL) &&
+	    (le32_to_cpu(*instance->consumer) == MEGASAS_ADPRESET_INPROG_SIGN))
 		return 1;
-	}
 
 	return 0;
 }
@@ -628,9 +625,10 @@ megasas_fire_cmd_skinny(struct megasas_instance *instance,
 {
 	unsigned long flags;
 	spin_lock_irqsave(&instance->hba_lock, flags);
-	writel(0, &(regs)->inbound_high_queue_port);
-	writel((frame_phys_addr | (frame_count<<1))|1,
-		&(regs)->inbound_low_queue_port);
+	writel(upper_32_bits(frame_phys_addr),
+	       &(regs)->inbound_high_queue_port);
+	writel((lower_32_bits(frame_phys_addr) | (frame_count<<1))|1,
+	       &(regs)->inbound_low_queue_port);
 	spin_unlock_irqrestore(&instance->hba_lock, flags);
 	/*msleep(5);*/
 }
@@ -764,6 +762,7 @@ megasas_fire_cmd_gen2(struct megasas_instance *instance,
 	spin_lock_irqsave(&instance->hba_lock, flags);
 	writel((frame_phys_addr | (frame_count<<1))|1,
 			&(regs)->inbound_queue_port);
+	readl(&regs->inbound_queue_port);
 	spin_unlock_irqrestore(&instance->hba_lock, flags);
 }
 
@@ -881,8 +880,8 @@ megasas_issue_polled(struct megasas_instance *instance, struct megasas_cmd *cmd)
 
 	struct megasas_header *frame_hdr = &cmd->frame->hdr;
 
-	frame_hdr->cmd_status = 0xFF;
-	frame_hdr->flags |= MFI_FRAME_DONT_POST_IN_REPLY_QUEUE;
+	frame_hdr->cmd_status = MFI_CMD_STATUS_POLL_MODE;
+	frame_hdr->flags |= cpu_to_le16(MFI_FRAME_DONT_POST_IN_REPLY_QUEUE);
 
 	/*
 	 * Issue the frame using inbound queue port
@@ -957,10 +956,12 @@ megasas_issue_blocked_abort_cmd(struct megasas_instance *instance,
 	 */
 	abort_fr->cmd = MFI_CMD_ABORT;
 	abort_fr->cmd_status = 0xFF;
-	abort_fr->flags = 0;
+	abort_fr->flags = cpu_to_le16(0);
 	abort_fr->abort_context = cmd_to_abort->index;
-	abort_fr->abort_mfi_phys_addr_lo = cmd_to_abort->frame_phys_addr;
-	abort_fr->abort_mfi_phys_addr_hi = 0;
+	abort_fr->abort_mfi_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd_to_abort->frame_phys_addr));
+	abort_fr->abort_mfi_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd_to_abort->frame_phys_addr));
 
 	cmd->sync_cmd = 1;
 	cmd->cmd_status = 0xFF;
@@ -1001,8 +1002,8 @@ megasas_make_sgl32(struct megasas_instance *instance, struct scsi_cmnd *scp,
 
 	if (sge_count) {
 		scsi_for_each_sg(scp, os_sgl, sge_count, i) {
-			mfi_sgl->sge32[i].length = sg_dma_len(os_sgl);
-			mfi_sgl->sge32[i].phys_addr = sg_dma_address(os_sgl);
+			mfi_sgl->sge32[i].length = cpu_to_le32(sg_dma_len(os_sgl));
+			mfi_sgl->sge32[i].phys_addr = cpu_to_le32(sg_dma_address(os_sgl));
 		}
 	}
 	return sge_count;
@@ -1030,8 +1031,8 @@ megasas_make_sgl64(struct megasas_instance *instance, struct scsi_cmnd *scp,
 
 	if (sge_count) {
 		scsi_for_each_sg(scp, os_sgl, sge_count, i) {
-			mfi_sgl->sge64[i].length = sg_dma_len(os_sgl);
-			mfi_sgl->sge64[i].phys_addr = sg_dma_address(os_sgl);
+			mfi_sgl->sge64[i].length = cpu_to_le32(sg_dma_len(os_sgl));
+			mfi_sgl->sge64[i].phys_addr = cpu_to_le64(sg_dma_address(os_sgl));
 		}
 	}
 	return sge_count;
@@ -1058,10 +1059,10 @@ megasas_make_sgl_skinny(struct megasas_instance *instance,
 
 	if (sge_count) {
 		scsi_for_each_sg(scp, os_sgl, sge_count, i) {
-			mfi_sgl->sge_skinny[i].length = sg_dma_len(os_sgl);
+			mfi_sgl->sge_skinny[i].length = cpu_to_le32(sg_dma_len(os_sgl));
 			mfi_sgl->sge_skinny[i].phys_addr =
-						sg_dma_address(os_sgl);
-			mfi_sgl->sge_skinny[i].flag = 0;
+				cpu_to_le64(sg_dma_address(os_sgl));
+			mfi_sgl->sge_skinny[i].flag = cpu_to_le32(0);
 		}
 	}
 	return sge_count;
@@ -1141,7 +1142,7 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 {
 	u32 is_logical;
 	u32 device_id;
-	u16 flags = 0;
+	u16 flags = 0, timeout = 0;
 	struct megasas_pthru_frame *pthru;
 
 	is_logical = MEGASAS_IS_LOGICAL(scp);
@@ -1155,10 +1156,6 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	else if (scp->sc_data_direction == PCI_DMA_NONE)
 		flags = MFI_FRAME_DIR_NONE;
 
-	if (instance->flag_ieee == 1) {
-		flags |= MFI_FRAME_IEEE;
-	}
-
 	/*
 	 * Prepare the DCDB frame
 	 */
@@ -1168,10 +1165,8 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	pthru->target_id = device_id;
 	pthru->lun = scp->device->lun;
 	pthru->cdb_len = scp->cmd_len;
-	pthru->timeout = 0;
 	pthru->pad_0 = 0;
-	pthru->flags = flags;
-	pthru->data_xfer_len = scsi_bufflen(scp);
+	pthru->data_xfer_len = cpu_to_le32(scsi_bufflen(scp));
 
 	memcpy(pthru->cdb, scp->cmnd, scp->cmd_len);
 
@@ -1181,36 +1176,42 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	*/
 	if (scp->device->type == TYPE_TAPE) {
 		if ((scp->request->timeout / HZ) > 0xFFFF)
-			pthru->timeout = 0xFFFF;
+			timeout = 0xFFFF;
 		else
-			pthru->timeout = scp->request->timeout / HZ;
+			timeout = scp->request->timeout / HZ;
 	}
 
 	/*
 	 * Construct SGL
 	 */
 	if (instance->flag_ieee == 1) {
-		pthru->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64 | MFI_FRAME_IEEE;
 		pthru->sge_count = megasas_make_sgl_skinny(instance, scp,
 						      &pthru->sgl);
 	} else if (IS_DMA64) {
-		pthru->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64;
 		pthru->sge_count = megasas_make_sgl64(instance, scp,
 						      &pthru->sgl);
 	} else
 		pthru->sge_count = megasas_make_sgl32(instance, scp,
 						      &pthru->sgl);
 
-       if (pthru->sge_count > instance->max_num_sge) {
+	if (pthru->sge_count > instance->max_num_sge) {
         	printk("megasas: build_dcdb error, two many SGE\n");
 	        return 0;
 	}
+
+	pthru->flags = cpu_to_le16(flags);
+	pthru->timeout = cpu_to_le16(timeout);
+
 	/*
 	 * Sense info specific
 	 */
 	pthru->sense_len = SCSI_SENSE_BUFFERSIZE;
-	pthru->sense_buf_phys_addr_hi = 0;
-	pthru->sense_buf_phys_addr_lo = cmd->sense_phys_addr;
+	pthru->sense_buf_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd->sense_phys_addr));
+	pthru->sense_buf_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd->sense_phys_addr));
 
 	/*
 	 * Compute the total number of frames this command consumes. FW uses
@@ -1410,10 +1411,6 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	else if (scp->sc_data_direction == PCI_DMA_FROMDEVICE)
 		flags = MFI_FRAME_DIR_READ;
 
-	if (instance->flag_ieee == 1) {
-		flags |= MFI_FRAME_IEEE;
-	}
-
 	/*
 	 * Prepare the Logical IO frame: 2nd bit is zero for all read cmds
 	 */
@@ -1421,10 +1418,9 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	ldio->cmd_status = 0x0;
 	ldio->scsi_status = 0x0;
 	ldio->target_id = device_id;
-	ldio->timeout = 0;
+	ldio->timeout = cpu_to_le16(0);
 	ldio->reserved_0 = 0;
 	ldio->pad_0 = 0;
-	ldio->flags = flags;
 	ldio->start_lba_hi = 0;
 	ldio->access_byte = (scp->cmd_len != 6) ? scp->cmnd[1] : 0;
 
@@ -1485,11 +1481,11 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	 * Construct SGL
 	 */
 	if (instance->flag_ieee) {
-		ldio->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64 | MFI_FRAME_IEEE;
 		ldio->sge_count = megasas_make_sgl_skinny(instance, scp,
 					      &ldio->sgl);
 	} else if (IS_DMA64) {
-		ldio->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64;
 		ldio->sge_count = megasas_make_sgl64(instance, scp, &ldio->sgl);
 	} else
 		ldio->sge_count = megasas_make_sgl32(instance, scp, &ldio->sgl);
@@ -1498,12 +1494,20 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
         	printk("megasas: build_ld_io  error, sge_count = %x\n", ldio->sge_count);
 	        return 0;
 	}
+
+	ldio->lba_count = cpu_to_le32(ldio->lba_count);
+	ldio->start_lba_lo = cpu_to_le32(ldio->start_lba_lo);
+	ldio->start_lba_hi = cpu_to_le32(ldio->start_lba_hi);
+	ldio->flags = cpu_to_le16(flags);
+
 	/*
 	 * Sense info specific
 	 */
 	ldio->sense_len = SCSI_SENSE_BUFFERSIZE;
-	ldio->sense_buf_phys_addr_hi = 0;
-	ldio->sense_buf_phys_addr_lo = cmd->sense_phys_addr;
+	ldio->sense_buf_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd->sense_phys_addr));
+	ldio->sense_buf_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd->sense_phys_addr));
 
 	/*
 	 * Compute the total number of frames this command consumes. FW uses
@@ -1580,20 +1584,20 @@ megasas_dump_pending_frames(struct megasas_instance *instance)
 			ldio = (struct megasas_io_frame *)cmd->frame;
 			mfi_sgl = &ldio->sgl;
 			sgcount = ldio->sge_count;
-			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lba lo : 0x%x, lba_hi : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no, cmd->frame_count,ldio->cmd,ldio->target_id, ldio->start_lba_lo,ldio->start_lba_hi,ldio->sense_buf_phys_addr_lo,sgcount);
+			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lba lo : 0x%x, lba_hi : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no, cmd->frame_count,ldio->cmd,ldio->target_id,le32_to_cpu(ldio->start_lba_lo),le32_to_cpu(ldio->start_lba_hi),le32_to_cpu(ldio->sense_buf_phys_addr_lo),sgcount);
 		}
 		else {
 			pthru = (struct megasas_pthru_frame *) cmd->frame;
 			mfi_sgl = &pthru->sgl;
 			sgcount = pthru->sge_count;
-			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lun : 0x%x, cdb_len : 0x%x, data xfer len : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no,cmd->frame_count,pthru->cmd,pthru->target_id,pthru->lun,pthru->cdb_len , pthru->data_xfer_len,pthru->sense_buf_phys_addr_lo,sgcount);
+			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lun : 0x%x, cdb_len : 0x%x, data xfer len : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no,cmd->frame_count,pthru->cmd,pthru->target_id,pthru->lun,pthru->cdb_len,le32_to_cpu(pthru->data_xfer_len),le32_to_cpu(pthru->sense_buf_phys_addr_lo),sgcount);
 		}
 	if(megasas_dbg_lvl & MEGASAS_DBG_LVL){
 		for (n = 0; n < sgcount; n++){
 			if (IS_DMA64)
-				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%08lx ",mfi_sgl->sge64[n].length , (unsigned long)mfi_sgl->sge64[n].phys_addr) ;
+				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%llx ",le32_to_cpu(mfi_sgl->sge64[n].length),le64_to_cpu(mfi_sgl->sge64[n].phys_addr));
 			else
-				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%x ",mfi_sgl->sge32[n].length , mfi_sgl->sge32[n].phys_addr) ;
+				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%x ",le32_to_cpu(mfi_sgl->sge32[n].length),le32_to_cpu(mfi_sgl->sge32[n].phys_addr)) ;
 			}
 		}
 		printk(KERN_ERR "\n");
@@ -2834,8 +2838,8 @@ static void megasas_complete_cmd_dpc(unsigned long instance_addr)
 
 	spin_lock_irqsave(&instance->completion_lock, flags);
 
-	producer = *instance->producer;
-	consumer = *instance->consumer;
+	producer = le32_to_cpu(*instance->producer);
+	consumer = le32_to_cpu(*instance->consumer);
 
 	while (consumer != producer) {
 		context = instance->reply_queue[consumer];
@@ -2854,7 +2858,7 @@ static void megasas_complete_cmd_dpc(unsigned long instance_addr)
 		}
 	}
 
-	*instance->consumer = producer;
+	*instance->consumer = cpu_to_le32(producer);
 
 	spin_unlock_irqrestore(&instance->completion_lock, flags);
 
@@ -2876,7 +2880,7 @@ void megasas_do_ocr(struct megasas_instance *instance)
                (instance->pdev->device == PCI_DEVICE_ID_DELL_PERC5) ||
                (instance->pdev->device == PCI_DEVICE_ID_LSI_VERDE_ZCR))
        {
-               *instance->consumer     = MEGASAS_ADPRESET_INPROG_SIGN;
+               *instance->consumer = cpu_to_le32(MEGASAS_ADPRESET_INPROG_SIGN);
        }
 
 	instance->instancet->disable_intr(instance);
@@ -3366,6 +3370,7 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 {
 	unsigned long flags;
 	int exception = 0;
+	u32 opcode;
 	struct megasas_header *hdr = &cmd->frame->hdr;
 	struct fusion_context *fusion = instance->ctrl_context;
 
@@ -3470,8 +3475,10 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 	case MFI_CMD_SMP:
 	case MFI_CMD_STP:
 	case MFI_CMD_DCMD:
+		opcode = le32_to_cpu(cmd->frame->dcmd.opcode);
+
 		/* Check for LD map update */
-		if ((cmd->frame->dcmd.opcode == MR_DCMD_LD_MAP_GET_INFO) && (cmd->frame->dcmd.mbox.b[1] == 1)) {
+		if ((opcode == MR_DCMD_LD_MAP_GET_INFO) && (cmd->frame->dcmd.mbox.b[1] == 1)) {
 			fusion->fast_path_io = 0;
 			spin_lock_irqsave(instance->host->host_lock, flags);
 			if (cmd->frame->hdr.cmd_status != 0) {
@@ -3495,8 +3502,8 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 			spin_unlock_irqrestore(instance->host->host_lock, flags);
 			break;
 		}
-		if (cmd->frame->dcmd.opcode == MR_DCMD_CTRL_EVENT_GET_INFO ||
-			cmd->frame->dcmd.opcode == MR_DCMD_CTRL_EVENT_GET) {
+		if (opcode == MR_DCMD_CTRL_EVENT_GET_INFO ||
+		    opcode == MR_DCMD_CTRL_EVENT_GET) {
 			spin_lock_irqsave(&poll_aen_lock, flags);
 			megasas_poll_wait_aen = 0;
 			spin_unlock_irqrestore(&poll_aen_lock, flags);
@@ -3505,7 +3512,7 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 		/*
 		 * See if got an event notification
 		 */
-		if (cmd->frame->dcmd.opcode == MR_DCMD_CTRL_EVENT_WAIT)
+		if (opcode == MR_DCMD_CTRL_EVENT_WAIT)
 			megasas_service_aen(instance, cmd);
 		else
 			megasas_complete_int_cmd(instance, cmd);
@@ -3761,7 +3768,7 @@ megasas_deplete_reply_queue(struct megasas_instance *instance, u8 alt_status)
                                	(instance->pdev->device == PCI_DEVICE_ID_DELL_PERC5) ||
                                	(instance->pdev->device == PCI_DEVICE_ID_LSI_VERDE_ZCR))
                                	{
-					*instance->consumer	= MEGASAS_ADPRESET_INPROG_SIGN;
+					*instance->consumer = cpu_to_le32(MEGASAS_ADPRESET_INPROG_SIGN);
                        	}
 
 
@@ -4265,6 +4272,7 @@ megasas_get_pd_list(struct megasas_instance *instance)
 	struct MR_PD_LIST *ci;
 	struct MR_PD_ADDRESS *pd_addr;
 	dma_addr_t ci_h = 0;
+	u32 pd_count;
 
 	cmd = megasas_get_cmd(instance);
 
@@ -4292,13 +4300,15 @@ megasas_get_pd_list(struct megasas_instance *instance)
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0xFF;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST);
-	dcmd->opcode = MR_DCMD_PD_LIST_QUERY;
-	dcmd->sgl.sge32[0].phys_addr = ci_h;
-	dcmd->sgl.sge32[0].length = MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST);
+	dcmd->data_xfer_len =
+		cpu_to_le32(MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_PD_LIST_QUERY);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(ci_h);
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST));
 
 	if (!megasas_issue_polled(instance, cmd)) {
 		ret = 0;
@@ -4311,18 +4321,19 @@ megasas_get_pd_list(struct megasas_instance *instance)
 	*/
 
 	pd_addr = ci->addr;
+	pd_count = le32_to_cpu(ci->count);
 
 	if ( ret == 0 &&
-		(ci->count <
+		(pd_count <
 		  (MEGASAS_MAX_PD_CHANNELS * MEGASAS_MAX_DEV_PER_CHANNEL))) {
 
 		memset(instance->pd_list, 0,
 			MEGASAS_MAX_PD * sizeof(struct megasas_pd_list));
 
-		for (pd_index = 0; pd_index < ci->count; pd_index++) {
+		for (pd_index = 0; pd_index < pd_count; pd_index++) {
 
 			instance->pd_list[pd_addr->deviceId].tid	=
-							pd_addr->deviceId;
+				le16_to_cpu(pd_addr->deviceId);
 			instance->pd_list[pd_addr->deviceId].driveType	=
 							pd_addr->scsiDevType;
 			instance->pd_list[pd_addr->deviceId].driveState	=
@@ -4357,6 +4368,7 @@ megasas_get_ld_list(struct megasas_instance *instance)
 	struct megasas_dcmd_frame *dcmd;
 	struct MR_LD_LIST *ci;
 	dma_addr_t ci_h = 0;
+	u32 ld_count;
 
 	cmd = megasas_get_cmd(instance);
 
@@ -4381,12 +4393,12 @@ megasas_get_ld_list(struct megasas_instance *instance)
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0xFF;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
-	dcmd->data_xfer_len = sizeof(struct MR_LD_LIST);
-	dcmd->opcode = MR_DCMD_LD_GET_LIST;
-	dcmd->sgl.sge32[0].phys_addr = ci_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct MR_LD_LIST);
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct MR_LD_LIST));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_LD_GET_LIST);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(ci_h);
+	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct MR_LD_LIST));
 	dcmd->pad_0  = 0;
 
 	if (!megasas_issue_polled(instance, cmd)) {
@@ -4396,15 +4408,17 @@ megasas_get_ld_list(struct megasas_instance *instance)
 		ret = -1;
 	}
 
+	ld_count = le32_to_cpu(ci->ldCount);
+
 	/*
 	* the following function will get the instance PD LIST.
 	*/
 
-	if ( (ret == 0) && (ci->ldCount <= (MAX_LOGICAL_DRIVES))){
-		instance->CurLdCount = ci->ldCount;
+	if ( (ret == 0) && (ld_count <= (MAX_LOGICAL_DRIVES))){
+		instance->CurLdCount = ld_count;
 		memset(instance->ld_ids, 0xff, MEGASAS_MAX_LD_IDS);
 
-		for (ld_index = 0; ld_index < ci->ldCount; ld_index++) {
+		for (ld_index = 0; ld_index < ld_count; ld_index++) {
 			if (ci->ldList[ld_index].state != 0) {
 				ids = ci->ldList[ld_index].ref.targetId;
 				instance->ld_ids[ids] = ci->ldList[ld_index].ref.targetId;
@@ -4465,13 +4479,13 @@ megasas_get_ctrl_info(struct megasas_instance *instance,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0xFF;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = sizeof(struct megasas_ctrl_info);
-	dcmd->opcode = MR_DCMD_CTRL_GET_INFO;
-	dcmd->sgl.sge32[0].phys_addr = ci_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct megasas_ctrl_info);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_ctrl_info));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_GET_INFO);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(ci_h);
+	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct megasas_ctrl_info));
 
 	if (!megasas_issue_polled(instance, cmd)) {
 		ret = 0;
@@ -4516,14 +4530,14 @@ static int megasas_check_cpx_support( struct megasas_instance *instance)
 
 
 	dcmd = &cmd->frame->dcmd;
-	dcmd->flags = 0;
-	dcmd->opcode = MR_DCMD_CTRL_MISC_CPX_INIT_DATA_GET;
-	dcmd->data_xfer_len  = sizeof(struct mr_cpx_init_data );
+	dcmd->flags = cpu_to_le16(0);
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_MISC_CPX_INIT_DATA_GET);
+	dcmd->data_xfer_len  = cpu_to_le32(sizeof(struct mr_cpx_init_data));
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status  = 0xff;
 	dcmd->sge_count = 1;
-	dcmd->sgl.sge32[0].phys_addr = cpx_init_data_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct mr_cpx_init_data);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(cpx_init_data_h);
+	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct mr_cpx_init_data));
 
 	retval = megasas_issue_polled ( instance, cmd );
 	
@@ -4570,9 +4584,9 @@ static int megasas_send_cpx_queue_data( struct megasas_instance *instance )
 
 	//initialize dcmd data
 	dcmd = &cmd->frame->dcmd;
-	dcmd->flags = 0;
-	dcmd->opcode = MR_DCMD_CTRL_MISC_CPX_QUEUE_DATA;
-	dcmd->data_xfer_len  = sizeof( struct mr_cpx_queue_data );
+	dcmd->flags = cpu_to_le16(0);
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_MISC_CPX_QUEUE_DATA);
+	dcmd->data_xfer_len  = cpu_to_le32(sizeof( struct mr_cpx_queue_data));
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status  = 0xff;
 	dcmd->sge_count = 1;
@@ -4683,11 +4697,15 @@ megasas_issue_init_mfi(struct megasas_instance *instance)
 	memset(initq_info, 0, sizeof(struct megasas_init_queue_info));
 	init_frame->context = context;
 
-	initq_info->reply_queue_entries = instance->max_fw_cmds + 1;
-	initq_info->reply_queue_start_phys_addr_lo = instance->reply_queue_h;
+	initq_info->reply_queue_entries =
+		cpu_to_le32(instance->max_fw_cmds + 1);
+	initq_info->reply_queue_start_phys_addr_lo =
+		cpu_to_le32(instance->reply_queue_h);
 
-	initq_info->producer_index_phys_addr_lo = instance->producer_h;
-	initq_info->consumer_index_phys_addr_lo = instance->consumer_h;
+	initq_info->producer_index_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(instance->producer_h));
+	initq_info->consumer_index_phys_addr_lo =
+		cpu_to_le32(upper_32_bits(instance->consumer_h));
 
 	init_frame->cmd = MFI_CMD_INIT;
 	init_frame->cmd_status = 0xFF;
@@ -4700,9 +4718,12 @@ megasas_issue_init_mfi(struct megasas_instance *instance)
 		init_frame->driver_ver_hi = 0;
 	}
 
-	init_frame->queue_info_new_phys_addr_lo = initq_info_h;
-
-	init_frame->data_xfer_len = sizeof(struct megasas_init_queue_info);
+	init_frame->queue_info_new_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(initq_info_h));
+	init_frame->queue_info_new_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(initq_info_h));
+	init_frame->data_xfer_len =
+		cpu_to_le32(sizeof(struct megasas_init_queue_info));
 
 	/*
 	 * disable the intr before firing the init frame to FW
@@ -4736,10 +4757,15 @@ megasas_init_adapter_mfi(struct megasas_instance *instance)
 
 	reg_set = instance->reg_set;
 
+#if defined(CONFIG_SPARC)
+	instance->max_fw_cmds = 400;
+#else  /* CONFIG_SPARC */
 	/*
 	 * Get various operational parameters from status register
 	 */
 	instance->max_fw_cmds = instance->instancet->read_fw_status_reg(reg_set) & 0x00FFFF;
+#endif	/* CONFIG_SPARC */
+
 	/*
 	 * Reduce the max supported cmds by 1. This is to ensure that the
 	 * reply_q_sz (1 more than the max cmd that driver may send)
@@ -4967,8 +4993,8 @@ static int megasas_init_fw(struct megasas_instance *instance)
 	if (ctrl_info && !megasas_get_ctrl_info(instance, ctrl_info)) {
 
 		max_sectors_1 = (1 << ctrl_info->stripe_sz_ops.min) *
-		    ctrl_info->max_strips_per_io;
-		max_sectors_2 = ctrl_info->max_request_size;
+			le16_to_cpu(ctrl_info->max_strips_per_io);
+		max_sectors_2 = le32_to_cpu(ctrl_info->max_request_size);
 
 		tmp_sectors = min_t(u32, max_sectors_1 , max_sectors_2);
 		
@@ -5102,20 +5128,25 @@ megasas_get_seq_num(struct megasas_instance *instance,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
-	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = sizeof(struct megasas_evt_log_info);
-	dcmd->opcode = MR_DCMD_CTRL_EVENT_GET_INFO;
-	dcmd->sgl.sge32[0].phys_addr = el_info_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct megasas_evt_log_info);
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
+	dcmd->pad_0 = cpu_to_le16(0);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_evt_log_info));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_EVENT_GET_INFO);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(el_info_h);
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(sizeof(struct megasas_evt_log_info));
 
 	megasas_issue_blocked_cmd(instance, cmd, 0);
 
 	/*
 	 * Copy the data back into callers buffer
 	 */
-	memcpy(eli, el_info, sizeof(struct megasas_evt_log_info));
+	eli->newest_seq_num = le32_to_cpu(el_info->newest_seq_num);
+	eli->oldest_seq_num = le32_to_cpu(el_info->oldest_seq_num);
+	eli->clear_seq_num = le32_to_cpu(el_info->clear_seq_num);
+	eli->shutdown_seq_num = le32_to_cpu(el_info->shutdown_seq_num);
+	eli->boot_seq_num = le32_to_cpu(el_info->boot_seq_num);
 
 	pci_free_consistent(instance->pdev, sizeof(struct megasas_evt_log_info),
 			    el_info, el_info_h);
@@ -5162,6 +5193,7 @@ megasas_register_aen(struct megasas_instance *instance, u32 seq_num,
 	if (instance->aen_cmd) {
 
 		prev_aen.word = instance->aen_cmd->frame->dcmd.mbox.w[1];
+		prev_aen.members.locale = le16_to_cpu(prev_aen.members.locale);
 
 		/*
 		 * A class whose enum value is smaller is inclusive of all
@@ -5219,16 +5251,17 @@ megasas_register_aen(struct megasas_instance *instance, u32 seq_num,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = sizeof(struct megasas_evt_detail);
-	dcmd->opcode = MR_DCMD_CTRL_EVENT_WAIT;
-	dcmd->mbox.w[0] = seq_num;
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_evt_detail));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_EVENT_WAIT);
+	dcmd->mbox.w[0] = cpu_to_le32(seq_num);
         instance->last_seq_num = seq_num;
-	dcmd->mbox.w[1] = curr_aen.word;
-	dcmd->sgl.sge32[0].phys_addr = (u32) instance->evt_detail_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct megasas_evt_detail);
+	dcmd->mbox.w[1] = cpu_to_le32(curr_aen.word);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(instance->evt_detail_h);
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(sizeof(struct megasas_evt_detail));
 
 	if (instance->aen_cmd != NULL) {
 		megasas_return_cmd(instance, cmd);
@@ -5408,6 +5441,14 @@ megasas_set_dma_mask(struct pci_dev *pdev)
 		if (pci_set_dma_mask(pdev, DMA_BIT_MASK(32)) != 0)
 			goto fail_set_dma_mask;
 	}
+
+	/*
+	 * Ensure that all data structures are allocated in 32-bit
+	 * memory.
+	 */
+	if (pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32)) != 0)
+		goto fail_set_dma_mask;
+
 	return 0;
 
 fail_set_dma_mask:
@@ -5724,11 +5765,11 @@ static void megasas_flush_cache(struct megasas_instance *instance)
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 0;
-	dcmd->flags = MFI_FRAME_DIR_NONE;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_NONE);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = 0;
-	dcmd->opcode = MR_DCMD_CTRL_CACHE_FLUSH;
+	dcmd->data_xfer_len = cpu_to_le32(0);
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_CACHE_FLUSH);
 	dcmd->mbox.b[0] = MR_FLUSH_CTRL_CACHE | MR_FLUSH_DISK_CACHE;
 
 	megasas_issue_blocked_cmd(instance, cmd, 30);
@@ -5768,11 +5809,11 @@ static void megasas_shutdown_controller(struct megasas_instance *instance,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 0;
-	dcmd->flags = MFI_FRAME_DIR_NONE;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_NONE);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = 0;
-	dcmd->opcode = opcode;
+	dcmd->data_xfer_len = cpu_to_le32(0);
+	dcmd->opcode = cpu_to_le32(opcode);
 
 	megasas_issue_blocked_cmd(instance, cmd, 30);
 
@@ -6277,7 +6318,7 @@ megasas_mgmt_fw_ioctl(struct megasas_instance *instance,
 		goto out;
 	}
 
-	if (cmd->frame->dcmd.opcode == MR_DCMD_CTRL_IO_METRICS_GET) {
+	if (cmd->frame->dcmd.opcode == le32_to_cpu(MR_DCMD_CTRL_IO_METRICS_GET)) {
 		ProcessPerfMetricRequest(instance , kbuff_arr[0], ioc->sgl[0].iov_len);
 	}
 
@@ -6671,8 +6712,7 @@ megasas_aen_polling(struct work_struct *work)
 	instance->ev = NULL;
 	host = instance->host;
 	if (instance->evt_detail) {
-
-		switch (instance->evt_detail->code) {
+		switch (le32_to_cpu(instance->evt_detail->code)) {
 
 		case MR_EVT_PD_INSERTED:
 			if(megasas_get_pd_list(instance) == 0) {
@@ -6824,7 +6864,7 @@ megasas_aen_polling(struct work_struct *work)
 		}		
 	}
 
-	seq_num = instance->evt_detail->seq_num + 1;
+	seq_num = le32_to_cpu(instance->evt_detail->seq_num + 1);
 
 	/* Register AEN with FW for latest sequence number plus 1 */
 	class_locale.members.reserved = 0;

Re: State of big-endian for MegaRAID SAS driver

[Martin K. Petersen]

>>>>> "Martin" == Martin K Petersen <martin.petersen@oracle.com> writes:

And against upstream...

 -- 
 Martin K. Petersen	Oracle Linux Engineering

commit 6d6db0df277b46ea538be5c70e89347436a8a114
Author: Martin K. Petersen <martin.petersen@oracle.com>
Date:   Sat Aug 10 17:55:58 2013 -0400

        megaraid_sas: Big endian support
    
        This driver only worked on a little endian system. Make sure that all
        frames submitted to the hardware are converted to little endian.
    
        Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index 1f0ca68..14c4dbb 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -374,13 +374,10 @@ static int
 megasas_check_reset_xscale(struct megasas_instance *instance,
 		struct megasas_register_set __iomem *regs)
 {
-	u32 consumer;
-	consumer = *instance->consumer;
-
 	if ((instance->adprecovery != MEGASAS_HBA_OPERATIONAL) &&
-		(*instance->consumer == MEGASAS_ADPRESET_INPROG_SIGN)) {
+	    (le32_to_cpu(*instance->consumer) == MEGASAS_ADPRESET_INPROG_SIGN))
 		return 1;
-	}
+
 	return 0;
 }
 
@@ -629,9 +626,10 @@ megasas_fire_cmd_skinny(struct megasas_instance *instance,
 {
 	unsigned long flags;
 	spin_lock_irqsave(&instance->hba_lock, flags);
-	writel(0, &(regs)->inbound_high_queue_port);
-	writel((frame_phys_addr | (frame_count<<1))|1,
-		&(regs)->inbound_low_queue_port);
+	writel(upper_32_bits(frame_phys_addr),
+	       &(regs)->inbound_high_queue_port);
+	writel((lower_32_bits(frame_phys_addr) | (frame_count<<1))|1,
+	       &(regs)->inbound_low_queue_port);
 	spin_unlock_irqrestore(&instance->hba_lock, flags);
 }
 
@@ -762,6 +760,7 @@ megasas_fire_cmd_gen2(struct megasas_instance *instance,
 	spin_lock_irqsave(&instance->hba_lock, flags);
 	writel((frame_phys_addr | (frame_count<<1))|1,
 			&(regs)->inbound_queue_port);
+	readl(&regs->inbound_queue_port);
 	spin_unlock_irqrestore(&instance->hba_lock, flags);
 }
 
@@ -879,8 +878,8 @@ megasas_issue_polled(struct megasas_instance *instance, struct megasas_cmd *cmd)
 
 	struct megasas_header *frame_hdr = &cmd->frame->hdr;
 
-	frame_hdr->cmd_status = 0xFF;
-	frame_hdr->flags |= MFI_FRAME_DONT_POST_IN_REPLY_QUEUE;
+	frame_hdr->cmd_status = MFI_CMD_STATUS_POLL_MODE;
+	frame_hdr->flags |= cpu_to_le16(MFI_FRAME_DONT_POST_IN_REPLY_QUEUE);
 
 	/*
 	 * Issue the frame using inbound queue port
@@ -944,10 +943,12 @@ megasas_issue_blocked_abort_cmd(struct megasas_instance *instance,
 	 */
 	abort_fr->cmd = MFI_CMD_ABORT;
 	abort_fr->cmd_status = 0xFF;
-	abort_fr->flags = 0;
+	abort_fr->flags = cpu_to_le16(0);
 	abort_fr->abort_context = cmd_to_abort->index;
-	abort_fr->abort_mfi_phys_addr_lo = cmd_to_abort->frame_phys_addr;
-	abort_fr->abort_mfi_phys_addr_hi = 0;
+	abort_fr->abort_mfi_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd_to_abort->frame_phys_addr));
+	abort_fr->abort_mfi_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd_to_abort->frame_phys_addr));
 
 	cmd->sync_cmd = 1;
 	cmd->cmd_status = 0xFF;
@@ -986,8 +987,8 @@ megasas_make_sgl32(struct megasas_instance *instance, struct scsi_cmnd *scp,
 
 	if (sge_count) {
 		scsi_for_each_sg(scp, os_sgl, sge_count, i) {
-			mfi_sgl->sge32[i].length = sg_dma_len(os_sgl);
-			mfi_sgl->sge32[i].phys_addr = sg_dma_address(os_sgl);
+			mfi_sgl->sge32[i].length = cpu_to_le32(sg_dma_len(os_sgl));
+			mfi_sgl->sge32[i].phys_addr = cpu_to_le32(sg_dma_address(os_sgl));
 		}
 	}
 	return sge_count;
@@ -1015,8 +1016,8 @@ megasas_make_sgl64(struct megasas_instance *instance, struct scsi_cmnd *scp,
 
 	if (sge_count) {
 		scsi_for_each_sg(scp, os_sgl, sge_count, i) {
-			mfi_sgl->sge64[i].length = sg_dma_len(os_sgl);
-			mfi_sgl->sge64[i].phys_addr = sg_dma_address(os_sgl);
+			mfi_sgl->sge64[i].length = cpu_to_le32(sg_dma_len(os_sgl));
+			mfi_sgl->sge64[i].phys_addr = cpu_to_le64(sg_dma_address(os_sgl));
 		}
 	}
 	return sge_count;
@@ -1043,10 +1044,10 @@ megasas_make_sgl_skinny(struct megasas_instance *instance,
 
 	if (sge_count) {
 		scsi_for_each_sg(scp, os_sgl, sge_count, i) {
-			mfi_sgl->sge_skinny[i].length = sg_dma_len(os_sgl);
+			mfi_sgl->sge_skinny[i].length = cpu_to_le32(sg_dma_len(os_sgl));
 			mfi_sgl->sge_skinny[i].phys_addr =
-						sg_dma_address(os_sgl);
-			mfi_sgl->sge_skinny[i].flag = 0;
+				cpu_to_le64(sg_dma_address(os_sgl));
+			mfi_sgl->sge_skinny[i].flag = cpu_to_le32(0);
 		}
 	}
 	return sge_count;
@@ -1126,7 +1127,7 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 {
 	u32 is_logical;
 	u32 device_id;
-	u16 flags = 0;
+	u16 flags = 0, timeout = 0;
 	struct megasas_pthru_frame *pthru;
 
 	is_logical = MEGASAS_IS_LOGICAL(scp);
@@ -1140,10 +1141,6 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	else if (scp->sc_data_direction == PCI_DMA_NONE)
 		flags = MFI_FRAME_DIR_NONE;
 
-	if (instance->flag_ieee == 1) {
-		flags |= MFI_FRAME_IEEE;
-	}
-
 	/*
 	 * Prepare the DCDB frame
 	 */
@@ -1153,10 +1150,8 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	pthru->target_id = device_id;
 	pthru->lun = scp->device->lun;
 	pthru->cdb_len = scp->cmd_len;
-	pthru->timeout = 0;
 	pthru->pad_0 = 0;
-	pthru->flags = flags;
-	pthru->data_xfer_len = scsi_bufflen(scp);
+	pthru->data_xfer_len = cpu_to_le32(scsi_bufflen(scp));
 
 	memcpy(pthru->cdb, scp->cmnd, scp->cmd_len);
 
@@ -1166,20 +1161,20 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	*/
 	if (scp->device->type == TYPE_TAPE) {
 		if ((scp->request->timeout / HZ) > 0xFFFF)
-			pthru->timeout = 0xFFFF;
+			timeout = 0xFFFF;
 		else
-			pthru->timeout = scp->request->timeout / HZ;
+			timeout = scp->request->timeout / HZ;
 	}
 
 	/*
 	 * Construct SGL
 	 */
 	if (instance->flag_ieee == 1) {
-		pthru->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64 | MFI_FRAME_IEEE;
 		pthru->sge_count = megasas_make_sgl_skinny(instance, scp,
 						      &pthru->sgl);
 	} else if (IS_DMA64) {
-		pthru->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64;
 		pthru->sge_count = megasas_make_sgl64(instance, scp,
 						      &pthru->sgl);
 	} else
@@ -1192,12 +1187,17 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 		return 0;
 	}
 
+	pthru->flags = cpu_to_le16(flags);
+	pthru->timeout = cpu_to_le16(timeout);
+
 	/*
 	 * Sense info specific
 	 */
 	pthru->sense_len = SCSI_SENSE_BUFFERSIZE;
-	pthru->sense_buf_phys_addr_hi = 0;
-	pthru->sense_buf_phys_addr_lo = cmd->sense_phys_addr;
+	pthru->sense_buf_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd->sense_phys_addr));
+	pthru->sense_buf_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd->sense_phys_addr));
 
 	/*
 	 * Compute the total number of frames this command consumes. FW uses
@@ -1234,10 +1234,6 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	else if (scp->sc_data_direction == PCI_DMA_FROMDEVICE)
 		flags = MFI_FRAME_DIR_READ;
 
-	if (instance->flag_ieee == 1) {
-		flags |= MFI_FRAME_IEEE;
-	}
-
 	/*
 	 * Prepare the Logical IO frame: 2nd bit is zero for all read cmds
 	 */
@@ -1245,10 +1241,9 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	ldio->cmd_status = 0x0;
 	ldio->scsi_status = 0x0;
 	ldio->target_id = device_id;
-	ldio->timeout = 0;
+	ldio->timeout = cpu_to_le16(0);
 	ldio->reserved_0 = 0;
 	ldio->pad_0 = 0;
-	ldio->flags = flags;
 	ldio->start_lba_hi = 0;
 	ldio->access_byte = (scp->cmd_len != 6) ? scp->cmnd[1] : 0;
 
@@ -1309,11 +1304,11 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	 * Construct SGL
 	 */
 	if (instance->flag_ieee) {
-		ldio->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64 | MFI_FRAME_IEEE;
 		ldio->sge_count = megasas_make_sgl_skinny(instance, scp,
 					      &ldio->sgl);
 	} else if (IS_DMA64) {
-		ldio->flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64;
 		ldio->sge_count = megasas_make_sgl64(instance, scp, &ldio->sgl);
 	} else
 		ldio->sge_count = megasas_make_sgl32(instance, scp, &ldio->sgl);
@@ -1324,12 +1319,19 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 		return 0;
 	}
 
+	ldio->lba_count = cpu_to_le32(ldio->lba_count);
+	ldio->start_lba_lo = cpu_to_le32(ldio->start_lba_lo);
+	ldio->start_lba_hi = cpu_to_le32(ldio->start_lba_hi);
+	ldio->flags = cpu_to_le16(flags);
+
 	/*
 	 * Sense info specific
 	 */
 	ldio->sense_len = SCSI_SENSE_BUFFERSIZE;
-	ldio->sense_buf_phys_addr_hi = 0;
-	ldio->sense_buf_phys_addr_lo = cmd->sense_phys_addr;
+	ldio->sense_buf_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd->sense_phys_addr));
+	ldio->sense_buf_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd->sense_phys_addr));
 
 	/*
 	 * Compute the total number of frames this command consumes. FW uses
@@ -1400,20 +1402,20 @@ megasas_dump_pending_frames(struct megasas_instance *instance)
 			ldio = (struct megasas_io_frame *)cmd->frame;
 			mfi_sgl = &ldio->sgl;
 			sgcount = ldio->sge_count;
-			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lba lo : 0x%x, lba_hi : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no, cmd->frame_count,ldio->cmd,ldio->target_id, ldio->start_lba_lo,ldio->start_lba_hi,ldio->sense_buf_phys_addr_lo,sgcount);
+			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lba lo : 0x%x, lba_hi : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no, cmd->frame_count,ldio->cmd,ldio->target_id,le32_to_cpu(ldio->start_lba_lo),le32_to_cpu(ldio->start_lba_hi),le32_to_cpu(ldio->sense_buf_phys_addr_lo),sgcount);
 		}
 		else {
 			pthru = (struct megasas_pthru_frame *) cmd->frame;
 			mfi_sgl = &pthru->sgl;
 			sgcount = pthru->sge_count;
-			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lun : 0x%x, cdb_len : 0x%x, data xfer len : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no,cmd->frame_count,pthru->cmd,pthru->target_id,pthru->lun,pthru->cdb_len , pthru->data_xfer_len,pthru->sense_buf_phys_addr_lo,sgcount);
+			printk(KERN_ERR "megasas[%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, lun : 0x%x, cdb_len : 0x%x, data xfer len : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x\n",instance->host->host_no,cmd->frame_count,pthru->cmd,pthru->target_id,pthru->lun,pthru->cdb_len,le32_to_cpu(pthru->data_xfer_len),le32_to_cpu(pthru->sense_buf_phys_addr_lo),sgcount);
 		}
 	if(megasas_dbg_lvl & MEGASAS_DBG_LVL){
 		for (n = 0; n < sgcount; n++){
 			if (IS_DMA64)
-				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%08lx ",mfi_sgl->sge64[n].length , (unsigned long)mfi_sgl->sge64[n].phys_addr) ;
+				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%llx ",le32_to_cpu(mfi_sgl->sge64[n].length),le64_to_cpu(mfi_sgl->sge64[n].phys_addr));
 			else
-				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%x ",mfi_sgl->sge32[n].length , mfi_sgl->sge32[n].phys_addr) ;
+				printk(KERN_ERR "megasas: sgl len : 0x%x, sgl addr : 0x%x ",le32_to_cpu(mfi_sgl->sge32[n].length),le32_to_cpu(mfi_sgl->sge32[n].phys_addr)) ;
 			}
 		}
 		printk(KERN_ERR "\n");
@@ -1674,8 +1676,8 @@ static void megasas_complete_cmd_dpc(unsigned long instance_addr)
 
 	spin_lock_irqsave(&instance->completion_lock, flags);
 
-	producer = *instance->producer;
-	consumer = *instance->consumer;
+	producer = le32_to_cpu(*instance->producer);
+	consumer = le32_to_cpu(*instance->consumer);
 
 	while (consumer != producer) {
 		context = instance->reply_queue[consumer];
@@ -1695,7 +1697,7 @@ static void megasas_complete_cmd_dpc(unsigned long instance_addr)
 		}
 	}
 
-	*instance->consumer = producer;
+	*instance->consumer = cpu_to_le32(producer);
 
 	spin_unlock_irqrestore(&instance->completion_lock, flags);
 
@@ -1716,7 +1718,7 @@ void megasas_do_ocr(struct megasas_instance *instance)
 	if ((instance->pdev->device == PCI_DEVICE_ID_LSI_SAS1064R) ||
 	(instance->pdev->device == PCI_DEVICE_ID_DELL_PERC5) ||
 	(instance->pdev->device == PCI_DEVICE_ID_LSI_VERDE_ZCR)) {
-		*instance->consumer     = MEGASAS_ADPRESET_INPROG_SIGN;
+		*instance->consumer = cpu_to_le32(MEGASAS_ADPRESET_INPROG_SIGN);
 	}
 	instance->instancet->disable_intr(instance);
 	instance->adprecovery   = MEGASAS_ADPRESET_SM_INFAULT;
@@ -2183,6 +2185,7 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 		     u8 alt_status)
 {
 	int exception = 0;
+	u32 opcode;
 	struct megasas_header *hdr = &cmd->frame->hdr;
 	unsigned long flags;
 	struct fusion_context *fusion = instance->ctrl_context;
@@ -2287,8 +2290,10 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 	case MFI_CMD_SMP:
 	case MFI_CMD_STP:
 	case MFI_CMD_DCMD:
+		opcode = le32_to_cpu(cmd->frame->dcmd.opcode);
+
 		/* Check for LD map update */
-		if ((cmd->frame->dcmd.opcode == MR_DCMD_LD_MAP_GET_INFO) &&
+		if ((opcode == MR_DCMD_LD_MAP_GET_INFO) &&
 		    (cmd->frame->dcmd.mbox.b[1] == 1)) {
 			fusion->fast_path_io = 0;
 			spin_lock_irqsave(instance->host->host_lock, flags);
@@ -2323,8 +2328,8 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 					       flags);
 			break;
 		}
-		if (cmd->frame->dcmd.opcode == MR_DCMD_CTRL_EVENT_GET_INFO ||
-			cmd->frame->dcmd.opcode == MR_DCMD_CTRL_EVENT_GET) {
+		if (opcode == MR_DCMD_CTRL_EVENT_GET_INFO ||
+		    opcode == MR_DCMD_CTRL_EVENT_GET) {
 			spin_lock_irqsave(&poll_aen_lock, flags);
 			megasas_poll_wait_aen = 0;
 			spin_unlock_irqrestore(&poll_aen_lock, flags);
@@ -2333,7 +2338,7 @@ megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
 		/*
 		 * See if got an event notification
 		 */
-		if (cmd->frame->dcmd.opcode == MR_DCMD_CTRL_EVENT_WAIT)
+		if (opcode == MR_DCMD_CTRL_EVENT_WAIT)
 			megasas_service_aen(instance, cmd);
 		else
 			megasas_complete_int_cmd(instance, cmd);
@@ -2606,7 +2611,7 @@ megasas_deplete_reply_queue(struct megasas_instance *instance,
 					PCI_DEVICE_ID_LSI_VERDE_ZCR)) {
 
 				*instance->consumer =
-					MEGASAS_ADPRESET_INPROG_SIGN;
+					cpu_to_le32(MEGASAS_ADPRESET_INPROG_SIGN);
 			}
 
 
@@ -3116,6 +3121,7 @@ megasas_get_pd_list(struct megasas_instance *instance)
 	struct MR_PD_LIST *ci;
 	struct MR_PD_ADDRESS *pd_addr;
 	dma_addr_t ci_h = 0;
+	u32 pd_count;
 
 	cmd = megasas_get_cmd(instance);
 
@@ -3143,13 +3149,15 @@ megasas_get_pd_list(struct megasas_instance *instance)
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0xFF;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST);
-	dcmd->opcode = MR_DCMD_PD_LIST_QUERY;
-	dcmd->sgl.sge32[0].phys_addr = ci_h;
-	dcmd->sgl.sge32[0].length = MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST);
+	dcmd->data_xfer_len =
+		cpu_to_le32(MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_PD_LIST_QUERY);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(ci_h);
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(MEGASAS_MAX_PD * sizeof(struct MR_PD_LIST));
 
 	if (!megasas_issue_polled(instance, cmd)) {
 		ret = 0;
@@ -3162,18 +3170,19 @@ megasas_get_pd_list(struct megasas_instance *instance)
 	*/
 
 	pd_addr = ci->addr;
+	pd_count = le32_to_cpu(ci->count);
 
 	if ( ret == 0 &&
-		(ci->count <
+		(pd_count <
 		  (MEGASAS_MAX_PD_CHANNELS * MEGASAS_MAX_DEV_PER_CHANNEL))) {
 
 		memset(instance->pd_list, 0,
 			MEGASAS_MAX_PD * sizeof(struct megasas_pd_list));
 
-		for (pd_index = 0; pd_index < ci->count; pd_index++) {
+		for (pd_index = 0; pd_index < pd_count; pd_index++) {
 
 			instance->pd_list[pd_addr->deviceId].tid	=
-							pd_addr->deviceId;
+				le16_to_cpu(pd_addr->deviceId);
 			instance->pd_list[pd_addr->deviceId].driveType	=
 							pd_addr->scsiDevType;
 			instance->pd_list[pd_addr->deviceId].driveState	=
@@ -3207,6 +3216,7 @@ megasas_get_ld_list(struct megasas_instance *instance)
 	struct megasas_dcmd_frame *dcmd;
 	struct MR_LD_LIST *ci;
 	dma_addr_t ci_h = 0;
+	u32 ld_count;
 
 	cmd = megasas_get_cmd(instance);
 
@@ -3233,12 +3243,12 @@ megasas_get_ld_list(struct megasas_instance *instance)
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0xFF;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
-	dcmd->data_xfer_len = sizeof(struct MR_LD_LIST);
-	dcmd->opcode = MR_DCMD_LD_GET_LIST;
-	dcmd->sgl.sge32[0].phys_addr = ci_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct MR_LD_LIST);
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct MR_LD_LIST));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_LD_GET_LIST);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(ci_h);
+	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct MR_LD_LIST));
 	dcmd->pad_0  = 0;
 
 	if (!megasas_issue_polled(instance, cmd)) {
@@ -3247,12 +3257,14 @@ megasas_get_ld_list(struct megasas_instance *instance)
 		ret = -1;
 	}
 
+	ld_count = le32_to_cpu(ci->ldCount);
+
 	/* the following function will get the instance PD LIST */
 
-	if ((ret == 0) && (ci->ldCount <= MAX_LOGICAL_DRIVES)) {
+	if ((ret == 0) && (ld_count <= MAX_LOGICAL_DRIVES)) {
 		memset(instance->ld_ids, 0xff, MEGASAS_MAX_LD_IDS);
 
-		for (ld_index = 0; ld_index < ci->ldCount; ld_index++) {
+		for (ld_index = 0; ld_index < ld_count; ld_index++) {
 			if (ci->ldList[ld_index].state != 0) {
 				ids = ci->ldList[ld_index].ref.targetId;
 				instance->ld_ids[ids] =
@@ -3313,13 +3325,13 @@ megasas_get_ctrl_info(struct megasas_instance *instance,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0xFF;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = sizeof(struct megasas_ctrl_info);
-	dcmd->opcode = MR_DCMD_CTRL_GET_INFO;
-	dcmd->sgl.sge32[0].phys_addr = ci_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct megasas_ctrl_info);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_ctrl_info));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_GET_INFO);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(ci_h);
+	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct megasas_ctrl_info));
 
 	if (!megasas_issue_polled(instance, cmd)) {
 		ret = 0;
@@ -3375,17 +3387,25 @@ megasas_issue_init_mfi(struct megasas_instance *instance)
 	memset(initq_info, 0, sizeof(struct megasas_init_queue_info));
 	init_frame->context = context;
 
-	initq_info->reply_queue_entries = instance->max_fw_cmds + 1;
-	initq_info->reply_queue_start_phys_addr_lo = instance->reply_queue_h;
+	initq_info->reply_queue_entries =
+		cpu_to_le32(instance->max_fw_cmds + 1);
+	initq_info->reply_queue_start_phys_addr_lo =
+		cpu_to_le32(instance->reply_queue_h);
 
-	initq_info->producer_index_phys_addr_lo = instance->producer_h;
-	initq_info->consumer_index_phys_addr_lo = instance->consumer_h;
+	initq_info->producer_index_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(instance->producer_h));
+	initq_info->consumer_index_phys_addr_lo =
+		cpu_to_le32(upper_32_bits(instance->consumer_h));
 
 	init_frame->cmd = MFI_CMD_INIT;
 	init_frame->cmd_status = 0xFF;
-	init_frame->queue_info_new_phys_addr_lo = initq_info_h;
 
-	init_frame->data_xfer_len = sizeof(struct megasas_init_queue_info);
+	init_frame->queue_info_new_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(initq_info_h));
+	init_frame->queue_info_new_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(initq_info_h));
+	init_frame->data_xfer_len =
+		cpu_to_le32(sizeof(struct megasas_init_queue_info));
 
 	/*
 	 * disable the intr before firing the init frame to FW
@@ -3423,6 +3443,7 @@ megasas_init_adapter_mfi(struct megasas_instance *instance)
 	 * Get various operational parameters from status register
 	 */
 	instance->max_fw_cmds = instance->instancet->read_fw_status_reg(reg_set) & 0x00FFFF;
+
 	/*
 	 * Reduce the max supported cmds by 1. This is to ensure that the
 	 * reply_q_sz (1 more than the max cmd that driver may send)
@@ -3665,8 +3686,8 @@ static int megasas_init_fw(struct megasas_instance *instance)
 	if (ctrl_info && !megasas_get_ctrl_info(instance, ctrl_info)) {
 
 		max_sectors_1 = (1 << ctrl_info->stripe_sz_ops.min) *
-		    ctrl_info->max_strips_per_io;
-		max_sectors_2 = ctrl_info->max_request_size;
+			le16_to_cpu(ctrl_info->max_strips_per_io);
+		max_sectors_2 = le32_to_cpu(ctrl_info->max_request_size);
 
 		tmp_sectors = min_t(u32, max_sectors_1 , max_sectors_2);
 
@@ -3802,20 +3823,25 @@ megasas_get_seq_num(struct megasas_instance *instance,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
-	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = sizeof(struct megasas_evt_log_info);
-	dcmd->opcode = MR_DCMD_CTRL_EVENT_GET_INFO;
-	dcmd->sgl.sge32[0].phys_addr = el_info_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct megasas_evt_log_info);
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
+	dcmd->pad_0 = cpu_to_le16(0);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_evt_log_info));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_EVENT_GET_INFO);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(el_info_h);
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(sizeof(struct megasas_evt_log_info));
 
 	megasas_issue_blocked_cmd(instance, cmd);
 
 	/*
 	 * Copy the data back into callers buffer
 	 */
-	memcpy(eli, el_info, sizeof(struct megasas_evt_log_info));
+	eli->newest_seq_num = le32_to_cpu(el_info->newest_seq_num);
+	eli->oldest_seq_num = le32_to_cpu(el_info->oldest_seq_num);
+	eli->clear_seq_num = le32_to_cpu(el_info->clear_seq_num);
+	eli->shutdown_seq_num = le32_to_cpu(el_info->shutdown_seq_num);
+	eli->boot_seq_num = le32_to_cpu(el_info->boot_seq_num);
 
 	pci_free_consistent(instance->pdev, sizeof(struct megasas_evt_log_info),
 			    el_info, el_info_h);
@@ -3862,6 +3888,7 @@ megasas_register_aen(struct megasas_instance *instance, u32 seq_num,
 	if (instance->aen_cmd) {
 
 		prev_aen.word = instance->aen_cmd->frame->dcmd.mbox.w[1];
+		prev_aen.members.locale = le16_to_cpu(prev_aen.members.locale);
 
 		/*
 		 * A class whose enum value is smaller is inclusive of all
@@ -3917,16 +3944,17 @@ megasas_register_aen(struct megasas_instance *instance, u32 seq_num,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 1;
-	dcmd->flags = MFI_FRAME_DIR_READ;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
 	instance->last_seq_num = seq_num;
-	dcmd->data_xfer_len = sizeof(struct megasas_evt_detail);
-	dcmd->opcode = MR_DCMD_CTRL_EVENT_WAIT;
-	dcmd->mbox.w[0] = seq_num;
-	dcmd->mbox.w[1] = curr_aen.word;
-	dcmd->sgl.sge32[0].phys_addr = (u32) instance->evt_detail_h;
-	dcmd->sgl.sge32[0].length = sizeof(struct megasas_evt_detail);
+	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_evt_detail));
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_EVENT_WAIT);
+	dcmd->mbox.w[0] = cpu_to_le32(seq_num);
+	dcmd->mbox.w[1] = cpu_to_le32(curr_aen.word);
+	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(instance->evt_detail_h);
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(sizeof(struct megasas_evt_detail));
 
 	if (instance->aen_cmd != NULL) {
 		megasas_return_cmd(instance, cmd);
@@ -4068,6 +4096,14 @@ megasas_set_dma_mask(struct pci_dev *pdev)
 		if (pci_set_dma_mask(pdev, DMA_BIT_MASK(32)) != 0)
 			goto fail_set_dma_mask;
 	}
+
+	/*
+	 * Ensure that all data structures are allocated in 32-bit
+	 * memory.
+	 */
+	if (pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32)) != 0)
+		goto fail_set_dma_mask;
+
 	return 0;
 
 fail_set_dma_mask:
@@ -4386,11 +4422,11 @@ static void megasas_flush_cache(struct megasas_instance *instance)
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 0;
-	dcmd->flags = MFI_FRAME_DIR_NONE;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_NONE);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = 0;
-	dcmd->opcode = MR_DCMD_CTRL_CACHE_FLUSH;
+	dcmd->data_xfer_len = cpu_to_le32(0);
+	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_CACHE_FLUSH);
 	dcmd->mbox.b[0] = MR_FLUSH_CTRL_CACHE | MR_FLUSH_DISK_CACHE;
 
 	megasas_issue_blocked_cmd(instance, cmd);
@@ -4431,11 +4467,11 @@ static void megasas_shutdown_controller(struct megasas_instance *instance,
 	dcmd->cmd = MFI_CMD_DCMD;
 	dcmd->cmd_status = 0x0;
 	dcmd->sge_count = 0;
-	dcmd->flags = MFI_FRAME_DIR_NONE;
-	dcmd->timeout = 0;
+	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_NONE);
+	dcmd->timeout = cpu_to_le16(0);
 	dcmd->pad_0 = 0;
-	dcmd->data_xfer_len = 0;
-	dcmd->opcode = opcode;
+	dcmd->data_xfer_len = cpu_to_le32(0);
+	dcmd->opcode = cpu_to_le32(opcode);
 
 	megasas_issue_blocked_cmd(instance, cmd);
 
@@ -5327,7 +5363,7 @@ megasas_aen_polling(struct work_struct *work)
 	host = instance->host;
 	if (instance->evt_detail) {
 
-		switch (instance->evt_detail->code) {
+		switch (le32_to_cpu(instance->evt_detail->code)) {
 		case MR_EVT_PD_INSERTED:
 			if (megasas_get_pd_list(instance) == 0) {
 			for (i = 0; i < MEGASAS_MAX_PD_CHANNELS; i++) {
@@ -5514,7 +5550,7 @@ megasas_aen_polling(struct work_struct *work)
 		return ;
 	}
 
-	seq_num = instance->evt_detail->seq_num + 1;
+	seq_num = le32_to_cpu(instance->evt_detail->seq_num + 1);
 
 	/* Register AEN with FW for latest sequence number plus 1 */
 	class_locale.members.reserved = 0;

Re: State of big-endian for MegaRAID SAS driver

[Martin K. Petersen]

>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:

Ben> You're the man. Forward ported this to 3.11-rc code base. Worked
Ben> like a charm.

Heh, you just beat me to it :)

Formatting differences aside there were a few glitches, however. I
outlined them below...

Btw. the reason I did cpu_to_le16(0) etc. was mostly to annotate the
endian-sensitive parameters in case somebody changes them in the future.

-- 
Martin K. Petersen	Oracle Linux Engineering

diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index 7bf9c5c..154e9e1 100644
@@ -1170,7 +1170,7 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	 * Construct SGL
 	 */
 	if (instance->flag_ieee == 1) {
-		flags |= MFI_FRAME_SGL64;
+		flags |= MFI_FRAME_SGL64 | MFI_FRAME_IEEE;
 		pthru->sge_count = megasas_make_sgl_skinny(instance, scp,
 						      &pthru->sgl);
 	} else if (IS_DMA64) {
@@ -1328,8 +1328,10 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
 	 * Sense info specific
 	 */
 	ldio->sense_len = SCSI_SENSE_BUFFERSIZE;
-	ldio->sense_buf_phys_addr_hi = 0;
-	ldio->sense_buf_phys_addr_lo = cmd->sense_phys_addr;
+	ldio->sense_buf_phys_addr_hi =
+		cpu_to_le32(upper_32_bits(cmd->sense_phys_addr));
+	ldio->sense_buf_phys_addr_lo =
+		cpu_to_le32(lower_32_bits(cmd->sense_phys_addr));
 
 	/*
 	 * Compute the total number of frames this command consumes. FW uses
@@ -3944,14 +3945,15 @@ megasas_register_aen(struct megasas_instance *instance, u32 seq_num,
 	dcmd->sge_count = 1;
 	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
 	dcmd->timeout = cpu_to_le16(0);
-	dcmd->pad_0 = cpu_to_le16(0);
+	dcmd->pad_0 = 0;
 	instance->last_seq_num = seq_num;
 	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_evt_detail));
 	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_EVENT_WAIT);
-	dcmd->mbox.w[0] = seq_num;
-	dcmd->mbox.w[1] = curr_aen.word;
+	dcmd->mbox.w[0] = cpu_to_le32(seq_num);
+	dcmd->mbox.w[1] = cpu_to_le32(curr_aen.word);
 	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(instance->evt_detail_h);
-	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct megasas_evt_detail));
+	dcmd->sgl.sge32[0].length =
+		cpu_to_le32(sizeof(struct megasas_evt_detail));
 
 	if (instance->aen_cmd != NULL) {
 		megasas_return_cmd(instance, cmd);
@@ -5358,9 +5360,9 @@ megasas_aen_polling(struct work_struct *work)
 	}
 	instance->ev = NULL;
 	host = instance->host;
-	if (le32_to_cpu(instance->evt_detail)) {
+	if (instance->evt_detail) {
 
-		switch (instance->evt_detail->code) {
+		switch (le32_to_cpu(instance->evt_detail->code)) {
 		case MR_EVT_PD_INSERTED:
 			if (megasas_get_pd_list(instance) == 0) {
 			for (i = 0; i < MEGASAS_MAX_PD_CHANNELS; i++) {

Re: State of big-endian for MegaRAID SAS driver

[Ben Collins]

On Aug 14, 2013, at 3:44 PM, "Martin K. Petersen" <martin.petersen@oracle.com> wrote:

>>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:
> 
> Ben> You're the man. Forward ported this to 3.11-rc code base. Worked
> Ben> like a charm.
> 
> Heh, you just beat me to it :)
> 
> Formatting differences aside there were a few glitches, however. I
> outlined them below...
> 
> Btw. the reason I did cpu_to_le16(0) etc. was mostly to annotate the
> endian-sensitive parameters in case somebody changes them in the future.

Right, I had over done it to make it consistent between similar code blocks. I'd like to see this (and the previous patch I sent for resource_size_t) included upstream. Considering your patch has had two success stories (I suspect yours was on SPARC, so that would even make two different platforms) and low likelihood of regression on little endian, I think it's pretty safe.

The only thing I think that could make it better is to annotate the structures in megaraid_sas.h as __le32 where applicable.

Thanks again Martin.

> -- 
> Martin K. Petersen	Oracle Linux Engineering
> 
> diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
> index 7bf9c5c..154e9e1 100644
> @@ -1170,7 +1170,7 @@ megasas_build_dcdb(struct megasas_instance *instance, struct scsi_cmnd *scp,
> 	 * Construct SGL
> 	 */
> 	if (instance->flag_ieee == 1) {
> -		flags |= MFI_FRAME_SGL64;
> +		flags |= MFI_FRAME_SGL64 | MFI_FRAME_IEEE;
> 		pthru->sge_count = megasas_make_sgl_skinny(instance, scp,
> 						      &pthru->sgl);
> 	} else if (IS_DMA64) {
> @@ -1328,8 +1328,10 @@ megasas_build_ldio(struct megasas_instance *instance, struct scsi_cmnd *scp,
> 	 * Sense info specific
> 	 */
> 	ldio->sense_len = SCSI_SENSE_BUFFERSIZE;
> -	ldio->sense_buf_phys_addr_hi = 0;
> -	ldio->sense_buf_phys_addr_lo = cmd->sense_phys_addr;
> +	ldio->sense_buf_phys_addr_hi =
> +		cpu_to_le32(upper_32_bits(cmd->sense_phys_addr));
> +	ldio->sense_buf_phys_addr_lo =
> +		cpu_to_le32(lower_32_bits(cmd->sense_phys_addr));
> 
> 	/*
> 	 * Compute the total number of frames this command consumes. FW uses
> @@ -3944,14 +3945,15 @@ megasas_register_aen(struct megasas_instance *instance, u32 seq_num,
> 	dcmd->sge_count = 1;
> 	dcmd->flags = cpu_to_le16(MFI_FRAME_DIR_READ);
> 	dcmd->timeout = cpu_to_le16(0);
> -	dcmd->pad_0 = cpu_to_le16(0);
> +	dcmd->pad_0 = 0;
> 	instance->last_seq_num = seq_num;
> 	dcmd->data_xfer_len = cpu_to_le32(sizeof(struct megasas_evt_detail));
> 	dcmd->opcode = cpu_to_le32(MR_DCMD_CTRL_EVENT_WAIT);
> -	dcmd->mbox.w[0] = seq_num;
> -	dcmd->mbox.w[1] = curr_aen.word;
> +	dcmd->mbox.w[0] = cpu_to_le32(seq_num);
> +	dcmd->mbox.w[1] = cpu_to_le32(curr_aen.word);
> 	dcmd->sgl.sge32[0].phys_addr = cpu_to_le32(instance->evt_detail_h);
> -	dcmd->sgl.sge32[0].length = cpu_to_le32(sizeof(struct megasas_evt_detail));
> +	dcmd->sgl.sge32[0].length =
> +		cpu_to_le32(sizeof(struct megasas_evt_detail));
> 
> 	if (instance->aen_cmd != NULL) {
> 		megasas_return_cmd(instance, cmd);
> @@ -5358,9 +5360,9 @@ megasas_aen_polling(struct work_struct *work)
> 	}
> 	instance->ev = NULL;
> 	host = instance->host;
> -	if (le32_to_cpu(instance->evt_detail)) {
> +	if (instance->evt_detail) {
> 
> -		switch (instance->evt_detail->code) {
> +		switch (le32_to_cpu(instance->evt_detail->code)) {
> 		case MR_EVT_PD_INSERTED:
> 			if (megasas_get_pd_list(instance) == 0) {
> 			for (i = 0; i < MEGASAS_MAX_PD_CHANNELS; i++) {
> 

--
Ben Collins - Principal Architect
Servergy, Inc. - www.servergy.com

Re: State of big-endian for MegaRAID SAS driver

[Martin K. Petersen]

>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:

Ben,

Ben> Considering your patch has had two success stories (I suspect yours
Ben> was on SPARC, so that would even make two different platforms) and
Ben> low likelihood of regression on little endian, I think it's pretty
Ben> safe.

I ordered a MegaRAID card today so I can test on x86 just to be sure.


Ben> The only thing I think that could make it better is to annotate the
Ben> structures in megaraid_sas.h as __le32 where applicable.

I have an older patch that does just that. And another one that gets rid
of the 32-bit DMA limitations on large systems. But I didn't want to
clutter the initial submission.

-- 
Martin K. Petersen	Oracle Linux Engineering

RE: State of big-endian for MegaRAID SAS driver

[Saxena, Sumit]

This patch: Big endian support for megaraid_sas driver is not complete. In few areas, we need to do changes like this patch does not have changes for Thunderbolt/invader/Fury controllers(file: megaraid_sas_fusion.c), and needs few changes in file: megaraid_sas_fp.c(while host reads some member of raidmap, it needs endianness change there also)..

LSI has megaraid_sas driver with big endian support for driver version "06.506.00.00". But to add big endian support for latest upstream megaraid_sas driver (06.600.18.00), we need to make some more changes.
We will get back with big endian support patch on top of latest upstream megaraid_sas driver.

Sumit

>-----Original Message-----
>From: Martin K. Petersen [mailto:martin.petersen@oracle.com]
>Sent: Friday, August 16, 2013 6:42 AM
>To: Ben Collins
>Cc: Martin K. Petersen; DL-MegaRAID Linux; linux-scsi;
>aradford@gmail.com
>Subject: Re: State of big-endian for MegaRAID SAS driver
>
>>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:
>
>Ben,
>
>Ben> Considering your patch has had two success stories (I suspect yours
>Ben> was on SPARC, so that would even make two different platforms) and
>Ben> low likelihood of regression on little endian, I think it's pretty
>Ben> safe.
>
>I ordered a MegaRAID card today so I can test on x86 just to be sure.
>
>
>Ben> The only thing I think that could make it better is to annotate the
>Ben> structures in megaraid_sas.h as __le32 where applicable.
>
>I have an older patch that does just that. And another one that gets rid
>of the 32-bit DMA limitations on large systems. But I didn't want to
>clutter the initial submission.
>
>--
>Martin K. Petersen	Oracle Linux Engineering

Re: State of big-endian for MegaRAID SAS driver

[Ben Collins]

On Aug 19, 2013, at 7:16 AM, "Saxena, Sumit" <Sumit.Saxena@lsi.com> wrote:
> This patch: Big endian support for megaraid_sas driver is not complete. In few areas, we need to do changes like this patch does not have changes for Thunderbolt/invader/Fury controllers(file: megaraid_sas_fusion.c), and needs few changes in file: megaraid_sas_fp.c(while host reads some member of raidmap, it needs endianness change there also)..
> 
> LSI has megaraid_sas driver with big endian support for driver version "06.506.00.00". But to add big endian support for latest upstream megaraid_sas driver (06.600.18.00), we need to make some more changes.
> We will get back with big endian support patch on top of latest upstream megaraid_sas driver.
> 

Is this work-in-progress available somewhere? What is the time frame you expect to have this supported by LSI directly? Based on the fact that big-endian support in megaraid_sas is 8 years in the making, forgive me if I'm a little skeptical of progress in the near term.

I realize that the big-endian patch from Martin isn't complete for fusion, but it is for other cards. Given that our company has a POWER based product using MegaRAID LSI SAS due on the market in a matter of months, I'd like to have stock kernels supporting it sooner rather than later.

Since this patch has zero effect on currently supported platforms (little-endian), I don't see the harm in letting it through, since it actually does work, and more importantly, works right now.

>> -----Original Message-----
>> From: Martin K. Petersen [mailto:martin.petersen@oracle.com]
>> Sent: Friday, August 16, 2013 6:42 AM
>> To: Ben Collins
>> Cc: Martin K. Petersen; DL-MegaRAID Linux; linux-scsi;
>> aradford@gmail.com
>> Subject: Re: State of big-endian for MegaRAID SAS driver
>> 
>>>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:
>> 
>> Ben,
>> 
>> Ben> Considering your patch has had two success stories (I suspect yours
>> Ben> was on SPARC, so that would even make two different platforms) and
>> Ben> low likelihood of regression on little endian, I think it's pretty
>> Ben> safe.
>> 
>> I ordered a MegaRAID card today so I can test on x86 just to be sure.
>> 
>> 
>> Ben> The only thing I think that could make it better is to annotate the
>> Ben> structures in megaraid_sas.h as __le32 where applicable.
>> 
>> I have an older patch that does just that. And another one that gets rid
>> of the 32-bit DMA limitations on large systems. But I didn't want to
>> clutter the initial submission.
>> 
>> --
>> Martin K. Petersen	Oracle Linux Engineering
> 
> 

--
Benjamin M. Collins | Principal Architect
Servergy, Inc. | Save Energy. Work Smart. ®
m (757) 243-7557 | e ben.c@servergy.com
L LinkedIn | T twitter.com/Servergy
www.servergy.com

CONFIDENTIALITY NOTICE: This communication contains privileged and/or confidential information to be maintained with the strictest confidence. It is intended solely for the use of the person or entity in which it is addressed. If you are not the intended recipient, you are STRICTLY PROHIBITED from disclosing, copying, distributing or using any of this information. If you received this communication in error, please contact the sender immediately and destroy the material in its entirety, whether electronic or hard copy.  Email may contain forward looking comments that are based on managements best-faith estimates which are subject to change. For information only. Not an offer, or solicitation of any offer to buy or sell any security, investment or other product.

--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: State of big-endian for MegaRAID SAS driver

[Saxena, Sumit]

>-----Original Message-----
>From: Ben Collins [mailto:ben.c@servergy.com]
>Sent: Monday, August 19, 2013 10:55 PM
>To: Saxena, Sumit
>Cc: Martin K. Petersen; DL-MegaRAID Linux; linux-scsi;
>aradford@gmail.com
>Subject: Re: State of big-endian for MegaRAID SAS driver
>
>On Aug 19, 2013, at 7:16 AM, "Saxena, Sumit" <Sumit.Saxena@lsi.com>
>wrote:
>> This patch: Big endian support for megaraid_sas driver is not
>complete. In few areas, we need to do changes like this patch does not
>have changes for Thunderbolt/invader/Fury controllers(file:
>megaraid_sas_fusion.c), and needs few changes in file:
>megaraid_sas_fp.c(while host reads some member of raidmap, it needs
>endianness change there also)..
>>
>> LSI has megaraid_sas driver with big endian support for driver version
>"06.506.00.00". But to add big endian support for latest upstream
>megaraid_sas driver (06.600.18.00), we need to make some more changes.
>> We will get back with big endian support patch on top of latest
>upstream megaraid_sas driver.
>>
>
>Is this work-in-progress available somewhere? What is the time frame you
>expect to have this supported by LSI directly? Based on the fact that
>big-endian support in megaraid_sas is 8 years in the making, forgive me
>if I'm a little skeptical of progress in the near term.
>
 
We are conferring with Marketing and Product Management on direction  wrt  Big endian support for megaraid_sas driver.  Our goal is to gain clear direction within two weeks.

>I realize that the big-endian patch from Martin isn't complete for
>fusion, but it is for other cards. Given that our company has a POWER
>based product using MegaRAID LSI SAS due on the market in a matter of
>months, I'd like to have stock kernels supporting it sooner rather than
>later.
>
>Since this patch has zero effect on currently supported platforms
>(little-endian), I don't see the harm in letting it through, since it
>actually does work, and more importantly, works right now.
>

Martin's patch does not have few changes needed in function "megasas_init_fw()"(inside file: megaraid_sas_base.c), so for consistency and driver's maintenance point of view- big endian support should be committed with complete set of changes, which works for all generation controllers.

>>> -----Original Message-----
>>> From: Martin K. Petersen [mailto:martin.petersen@oracle.com]
>>> Sent: Friday, August 16, 2013 6:42 AM
>>> To: Ben Collins
>>> Cc: Martin K. Petersen; DL-MegaRAID Linux; linux-scsi;
>>> aradford@gmail.com
>>> Subject: Re: State of big-endian for MegaRAID SAS driver
>>>
>>>>>>>> "Ben" == Ben Collins <ben.c@servergy.com> writes:
>>>
>>> Ben,
>>>
>>> Ben> Considering your patch has had two success stories (I suspect
>>> Ben> yours was on SPARC, so that would even make two different
>>> Ben> platforms) and low likelihood of regression on little endian, I
>>> Ben> think it's pretty safe.
>>>
>>> I ordered a MegaRAID card today so I can test on x86 just to be sure.
>>>
>>>
>>> Ben> The only thing I think that could make it better is to annotate
>>> Ben> the structures in megaraid_sas.h as __le32 where applicable.
>>>
>>> I have an older patch that does just that. And another one that gets
>>> rid of the 32-bit DMA limitations on large systems. But I didn't want
>>> to clutter the initial submission.
>>>
>>> --
>>> Martin K. Petersen	Oracle Linux Engineering
>>
>>
>
>--
>Benjamin M. Collins | Principal Architect Servergy, Inc. | Save Energy.
>Work Smart. (r) m (757) 243-7557 | e ben.c@servergy.com L LinkedIn | T
>twitter.com/Servergy www.servergy.com
>
>CONFIDENTIALITY NOTICE: This communication contains privileged and/or
>confidential information to be maintained with the strictest confidence.
>It is intended solely for the use of the person or entity in which it is
>addressed. If you are not the intended recipient, you are STRICTLY
>PROHIBITED from disclosing, copying, distributing or using any of this
>information. If you received this communication in error, please contact
>the sender immediately and destroy the material in its entirety, whether
>electronic or hard copy.  Email may contain forward looking comments
>that are based on managements best-faith estimates which are subject to
>change. For information only. Not an offer, or solicitation of any offer
>to buy or sell any security, investment or other product.
>

Re: State of big-endian for MegaRAID SAS driver

[Christoph Hellwig]

On Mon, Aug 26, 2013 at 12:51:07PM +0530, Saxena, Sumit wrote:
> We are conferring with Marketing and Product Management on direction  wrt  Big endian support for megaraid_sas driver.  Our goal is to gain clear direction within two weeks.

Sorry guys, but this is just utter bullshit.  There are obviously
correct patches out that make an important class of controllers work
with no downside at all given that code doesn't even change for little
endian compiled driver binaries.

I have to say I'm a little sad and disappointed we're even arguing this
in any form and that patch haven't been merged long ago.

RE: State of big-endian for MegaRAID SAS driver

[Maloy, Joe]

Chris,

I would appreciate you toning down your rhetoric.  Please keep it professional.

What Sumit reported is correct.  We are polling our Marketing and Product management organizations to see if we want to productize this type of support in our MR Linux drivers. We would not submit upstream changes that have not been successfully tested by our PT test org based on a GCA (general customer availability) or async release effort.    

Thanks,

Joe Maloy


  

-----Original Message-----
From: Christoph Hellwig [mailto:hch@infradead.org] 
Sent: Monday, August 26, 2013 3:06 AM
To: Saxena, Sumit
Cc: Ben Collins; Martin K. Petersen; DL-MegaRAID Linux; linux-scsi; aradford@gmail.com; Maloy, Joe
Subject: Re: State of big-endian for MegaRAID SAS driver

On Mon, Aug 26, 2013 at 12:51:07PM +0530, Saxena, Sumit wrote:
> We are conferring with Marketing and Product Management on direction  wrt  Big endian support for megaraid_sas driver.  Our goal is to gain clear direction within two weeks.

Sorry guys, but this is just utter bullshit.  There are obviously correct patches out that make an important class of controllers work with no downside at all given that code doesn't even change for little endian compiled driver binaries.

I have to say I'm a little sad and disappointed we're even arguing this in any form and that patch haven't been merged long ago.

Re: State of big-endian for MegaRAID SAS driver

[Ben Collins]

[-- Attachment #1: Type: text/plain, Size: 1963 bytes --]

On Aug 26, 2013, at 4:15 PM, "Maloy, Joe" <Joe.Maloy@lsi.com> wrote:
> Chris,
> 
> I would appreciate you toning down your rhetoric.  Please keep it professional.
> 
> What Sumit reported is correct.  We are polling our Marketing and Product management organizations to see if we want to productize this type of support in our MR Linux drivers. We would not submit upstream changes that have not been successfully tested by our PT test org based on a GCA (general customer availability) or async release effort.

Professionally speaking, however, having the megaraid_sas driver support big-endian is inline with the requirements of the driver being in the kernel tree to begin with. This support was promised 8 years back by LSI. Quite frankly, I can't imagine that LSI has to approve of this support before it can be included into the kernel proper. If LSI doesn't want to support big-endian, that's fine with me. However, if the community wants to support it in a public driver, I don't think LSIs Product Management and Marketing teams have anything useful to say about it.

a) This driver doesn't work on big-endian
b) This patch supports more architectures
c) This patch doesn't break existing architecture support (not even that which LSI wants to support)

Not including this patch does nothing useful. Including it will put megaraid_sas on the path to doing what every driver in the tree either does or attempts to do, and more important, should do.

Historically, waiting on commercial upstream to deliver what the community has already done is wasted time. Likewise, historically, allowing patches that partially implement things (in this respect, support one generation of card for megaraid_sas) leads to further support from the community, in a much faster timeframe.

--
Servergy  : http://www.servergy.com/
SwissDisk : http://www.swissdisk.com/
Ubuntu    : http://www.ubuntu.com/
My Blog   : http://ben-collins.blogspot.com/


[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 203 bytes --]

Re: State of big-endian for MegaRAID SAS driver

[Christoph Hellwig]

On Mon, Aug 26, 2013 at 02:15:42PM -0600, Maloy, Joe wrote:
> 
> Chris,
> 
> I would appreciate you toning down your rhetoric.  Please keep it professional.

If the b word is not in your business dictionary I have no idea how you
get any work done, and I really don't care if you feel offended or not
at all.

The real lack of professionalism is that LSI blocks a driver fix for no
reason at all.

Re: State of big-endian for MegaRAID SAS driver

[Christoph Hellwig]

On Mon, Aug 26, 2013 at 04:26:30PM -0400, Ben Collins wrote:
> Historically, waiting on commercial upstream to deliver what the community has already done is wasted time. Likewise, historically, allowing patches that partially implement things (in this respect, support one generation of card for megaraid_sas) leads to further support from the community, in a much faster timefram

*nod*  Also historically ignoring hardware vendor mangement chains has
always lead to better drivers.  Not saying that some vendors don't have
competent engineer (many do, mand don't), but as soon as management gets
involved in decisions about drivers all hope is lost.

RE: State of big-endian for MegaRAID SAS driver

[Maloy, Joe]

Chris and company,


There is a reason for the delay. 

We have done a code review of what has been proposed and we believe it is incomplete.  We don't think that in its current state it will serve Linux as well as we all want it to.

My team has the ability to address this enhancement fully, but the green light can't come from me alone. Hence, I am working within a time frame of two weeks to get those within LSI that need to be accounted for to be on the same page.  


Thanks,


Joe



-----Original Message-----
From: Christoph Hellwig [mailto:hch@infradead.org] 
Sent: Monday, August 26, 2013 2:42 PM
To: Maloy, Joe
Cc: Christoph Hellwig; Saxena, Sumit; Ben Collins; Martin K. Petersen; DL-MegaRAID Linux; linux-scsi; aradford@gmail.com; Maly, JoLyn
Subject: Re: State of big-endian for MegaRAID SAS driver

On Mon, Aug 26, 2013 at 02:15:42PM -0600, Maloy, Joe wrote:
> 
> Chris,
> 
> I would appreciate you toning down your rhetoric.  Please keep it professional.

If the b word is not in your business dictionary I have no idea how you get any work done, and I really don't care if you feel offended or not at all.

The real lack of professionalism is that LSI blocks a driver fix for no reason at all.

Re: State of big-endian for MegaRAID SAS driver

[Ben Collins]

[-- Attachment #1: Type: text/plain, Size: 1467 bytes --]

> We have done a code review of what has been proposed and we believe it is incomplete.  We don't think that in its current state it will serve Linux as well as we all want it to.


I fully appreciate you're intent, but I can't imagine how you can think that it wont serve Linux in it's current state when it does in fact serve Linux (and its users) in it's current state.

What doesn't serve Linux (and its users) is blocking patches in the hopes of you taking two weeks to get two non-technical departments to approve you and/or some other engineer(s) to do what you consider a better job. You cannot guarantee that this will even be "approved" and you can't guarantee that the end will result will come at any reasonable point in time after that arbitrary waiting period.

The delay serves no purpose. LSIs solution to this problem with only be a superset of this patch. This patch will be unchanged from your solution because there is absolutely no other way to implement this. It's not like there is a technical discussion about how the patch works. It's plain and simple.

So I say include this patch. If you get some sort of ok by your bobbling heads to complete the work, then you can even tell them "it's partly done already." No manager can argue with less work (aka investment).

--
Servergy  : http://www.servergy.com/
SwissDisk : http://www.swissdisk.com/
Ubuntu    : http://www.ubuntu.com/
My Blog   : http://ben-collins.blogspot.com/


[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 203 bytes --]

Re: State of big-endian for MegaRAID SAS driver

[Martin K. Petersen]

>>>>> "Sumit" == Saxena, Sumit <Sumit.Saxena@lsi.com> writes:

>> I realize that the big-endian patch from Martin isn't complete for
>> fusion, but it is for other cards. 

Correct. I had no interest in Fury/Invader and no access to those
boards.


Sumit> Martin's patch does not have few changes needed in function
Sumit> "megasas_init_fw()"(inside file: megaraid_sas_base.c), 

Patches welcome!

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: State of big-endian for MegaRAID SAS driver

[Martin K. Petersen]

>>>>> "Sumit" == Saxena, Sumit <Sumit.Saxena@lsi.com> writes:

Sumit> LSI has megaraid_sas driver with big endian support for driver
Sumit> version "06.506.00.00". But to add big endian support for latest
Sumit> upstream megaraid_sas driver (06.600.18.00), we need to make some
Sumit> more changes. 

There are a few very important observations to make from that statement:

1. If you have big endian support patches already why aren't they
   upstream?

2. Why do you have diverging driver branches? An older version has big
   endian support but somehow that support is missing from a later
   version of the driver...

-- 
Martin K. Petersen	Oracle Linux Engineering