2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
*
|
|
|
|
* Linux MegaRAID device driver
|
|
|
|
*
|
2005-11-18 02:13:31 +08:00
|
|
|
* Copyright (c) 2002 LSI Logic Corporation.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License
|
|
|
|
* as published by the Free Software Foundation; either version
|
|
|
|
* 2 of the License, or (at your option) any later version.
|
|
|
|
*
|
|
|
|
* Copyright (c) 2002 Red Hat, Inc. All rights reserved.
|
|
|
|
* - fixes
|
|
|
|
* - speed-ups (list handling fixes, issued_list, optimizations.)
|
|
|
|
* - lots of cleanups.
|
|
|
|
*
|
|
|
|
* Copyright (c) 2003 Christoph Hellwig <hch@lst.de>
|
|
|
|
* - new-style, hotplug-aware pci probing and scsi registration
|
|
|
|
*
|
2005-11-18 02:13:31 +08:00
|
|
|
* Version : v2.00.4 Mon Nov 14 14:02:43 EST 2005 - Seokmann Ju
|
|
|
|
* <Seokmann.Ju@lsil.com>
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
|
|
|
* Description: Linux device driver for LSI Logic MegaRAID controller
|
|
|
|
*
|
|
|
|
* Supported controllers: MegaRAID 418, 428, 438, 466, 762, 467, 471, 490, 493
|
|
|
|
* 518, 520, 531, 532
|
|
|
|
*
|
|
|
|
* This driver is supported by LSI Logic, with assistance from Red Hat, Dell,
|
|
|
|
* and others. Please send updates to the mailing list
|
|
|
|
* linux-scsi@vger.kernel.org .
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <asm/uaccess.h>
|
|
|
|
#include <asm/io.h>
|
2005-06-19 19:42:05 +08:00
|
|
|
#include <linux/completion.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <linux/delay.h>
|
|
|
|
#include <linux/proc_fs.h>
|
|
|
|
#include <linux/reboot.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/list.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/pci.h>
|
|
|
|
#include <linux/init.h>
|
2006-03-28 17:56:48 +08:00
|
|
|
#include <linux/dma-mapping.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <scsi/scsicam.h>
|
|
|
|
|
|
|
|
#include "scsi.h"
|
|
|
|
#include <scsi/scsi_host.h>
|
|
|
|
|
|
|
|
#include "megaraid.h"
|
|
|
|
|
2005-11-18 02:13:31 +08:00
|
|
|
#define MEGARAID_MODULE_VERSION "2.00.4"
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-11-18 02:13:31 +08:00
|
|
|
MODULE_AUTHOR ("sju@lsil.com");
|
|
|
|
MODULE_DESCRIPTION ("LSI Logic MegaRAID legacy driver");
|
2005-04-17 06:20:36 +08:00
|
|
|
MODULE_LICENSE ("GPL");
|
|
|
|
MODULE_VERSION(MEGARAID_MODULE_VERSION);
|
|
|
|
|
|
|
|
static unsigned int max_cmd_per_lun = DEF_CMD_PER_LUN;
|
|
|
|
module_param(max_cmd_per_lun, uint, 0);
|
|
|
|
MODULE_PARM_DESC(max_cmd_per_lun, "Maximum number of commands which can be issued to a single LUN (default=DEF_CMD_PER_LUN=63)");
|
|
|
|
|
|
|
|
static unsigned short int max_sectors_per_io = MAX_SECTORS_PER_IO;
|
|
|
|
module_param(max_sectors_per_io, ushort, 0);
|
|
|
|
MODULE_PARM_DESC(max_sectors_per_io, "Maximum number of sectors per I/O request (default=MAX_SECTORS_PER_IO=128)");
|
|
|
|
|
|
|
|
|
|
|
|
static unsigned short int max_mbox_busy_wait = MBOX_BUSY_WAIT;
|
|
|
|
module_param(max_mbox_busy_wait, ushort, 0);
|
|
|
|
MODULE_PARM_DESC(max_mbox_busy_wait, "Maximum wait for mailbox in microseconds if busy (default=MBOX_BUSY_WAIT=10)");
|
|
|
|
|
2006-12-04 09:49:23 +08:00
|
|
|
#define RDINDOOR(adapter) readl((adapter)->mmio_base + 0x20)
|
|
|
|
#define RDOUTDOOR(adapter) readl((adapter)->mmio_base + 0x2C)
|
|
|
|
#define WRINDOOR(adapter,value) writel(value, (adapter)->mmio_base + 0x20)
|
|
|
|
#define WROUTDOOR(adapter,value) writel(value, (adapter)->mmio_base + 0x2C)
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Global variables
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int hba_count;
|
|
|
|
static adapter_t *hba_soft_state[MAX_CONTROLLERS];
|
|
|
|
static struct proc_dir_entry *mega_proc_dir_entry;
|
|
|
|
|
|
|
|
/* For controller re-ordering */
|
|
|
|
static struct mega_hbas mega_hbas[MAX_CONTROLLERS];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The File Operations structure for the serial/ioctl interface of the driver
|
|
|
|
*/
|
2007-02-12 16:55:34 +08:00
|
|
|
static const struct file_operations megadev_fops = {
|
2005-04-17 06:20:36 +08:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.ioctl = megadev_ioctl,
|
|
|
|
.open = megadev_open,
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Array to structures for storing the information about the controllers. This
|
|
|
|
* information is sent to the user level applications, when they do an ioctl
|
|
|
|
* for this information.
|
|
|
|
*/
|
|
|
|
static struct mcontroller mcontroller[MAX_CONTROLLERS];
|
|
|
|
|
|
|
|
/* The current driver version */
|
|
|
|
static u32 driver_ver = 0x02000000;
|
|
|
|
|
|
|
|
/* major number used by the device for character interface */
|
|
|
|
static int major;
|
|
|
|
|
|
|
|
#define IS_RAID_CH(hba, ch) (((hba)->mega_ch_class >> (ch)) & 0x01)
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Debug variable to print some diagnostic messages
|
|
|
|
*/
|
|
|
|
static int trace_level;
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_setup_mailbox()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Allocates a 8 byte aligned memory for the handshake mailbox.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_setup_mailbox(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned long align;
|
|
|
|
|
|
|
|
adapter->una_mbox64 = pci_alloc_consistent(adapter->dev,
|
|
|
|
sizeof(mbox64_t), &adapter->una_mbox64_dma);
|
|
|
|
|
|
|
|
if( !adapter->una_mbox64 ) return -1;
|
|
|
|
|
|
|
|
adapter->mbox = &adapter->una_mbox64->mbox;
|
|
|
|
|
|
|
|
adapter->mbox = (mbox_t *)((((unsigned long) adapter->mbox) + 15) &
|
|
|
|
(~0UL ^ 0xFUL));
|
|
|
|
|
|
|
|
adapter->mbox64 = (mbox64_t *)(((unsigned long)adapter->mbox) - 8);
|
|
|
|
|
|
|
|
align = ((void *)adapter->mbox) - ((void *)&adapter->una_mbox64->mbox);
|
|
|
|
|
|
|
|
adapter->mbox_dma = adapter->una_mbox64_dma + 8 + align;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Register the mailbox if the controller is an io-mapped controller
|
|
|
|
*/
|
|
|
|
if( adapter->flag & BOARD_IOMAP ) {
|
|
|
|
|
|
|
|
outb_p(adapter->mbox_dma & 0xFF,
|
|
|
|
adapter->host->io_port + MBOX_PORT0);
|
|
|
|
|
|
|
|
outb_p((adapter->mbox_dma >> 8) & 0xFF,
|
|
|
|
adapter->host->io_port + MBOX_PORT1);
|
|
|
|
|
|
|
|
outb_p((adapter->mbox_dma >> 16) & 0xFF,
|
|
|
|
adapter->host->io_port + MBOX_PORT2);
|
|
|
|
|
|
|
|
outb_p((adapter->mbox_dma >> 24) & 0xFF,
|
|
|
|
adapter->host->io_port + MBOX_PORT3);
|
|
|
|
|
|
|
|
outb_p(ENABLE_MBOX_BYTE,
|
|
|
|
adapter->host->io_port + ENABLE_MBOX_REGION);
|
|
|
|
|
|
|
|
irq_ack(adapter);
|
|
|
|
|
|
|
|
irq_enable(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mega_query_adapter()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Issue the adapter inquiry commands to the controller and find out
|
|
|
|
* information and parameter about the devices attached
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_query_adapter(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
dma_addr_t prod_info_dma_handle;
|
|
|
|
mega_inquiry3 *inquiry3;
|
|
|
|
u8 raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
int retval;
|
|
|
|
|
|
|
|
/* Initialize adapter inquiry mailbox */
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset((void *)adapter->mega_buffer, 0, MEGA_BUFFER_SIZE);
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to issue Inquiry3 command
|
|
|
|
* if not succeeded, then issue MEGA_MBOXCMD_ADAPTERINQ command and
|
|
|
|
* update enquiry3 structure
|
|
|
|
*/
|
|
|
|
mbox->m_out.xferaddr = (u32)adapter->buf_dma_handle;
|
|
|
|
|
|
|
|
inquiry3 = (mega_inquiry3 *)adapter->mega_buffer;
|
|
|
|
|
|
|
|
raw_mbox[0] = FC_NEW_CONFIG; /* i.e. mbox->cmd=0xA1 */
|
|
|
|
raw_mbox[2] = NC_SUBOP_ENQUIRY3; /* i.e. 0x0F */
|
|
|
|
raw_mbox[3] = ENQ3_GET_SOLICITED_FULL; /* i.e. 0x02 */
|
|
|
|
|
|
|
|
/* Issue a blocking command to the card */
|
|
|
|
if ((retval = issue_scb_block(adapter, raw_mbox))) {
|
|
|
|
/* the adapter does not support 40ld */
|
|
|
|
|
|
|
|
mraid_ext_inquiry *ext_inq;
|
|
|
|
mraid_inquiry *inq;
|
|
|
|
dma_addr_t dma_handle;
|
|
|
|
|
|
|
|
ext_inq = pci_alloc_consistent(adapter->dev,
|
|
|
|
sizeof(mraid_ext_inquiry), &dma_handle);
|
|
|
|
|
|
|
|
if( ext_inq == NULL ) return -1;
|
|
|
|
|
|
|
|
inq = &ext_inq->raid_inq;
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = (u32)dma_handle;
|
|
|
|
|
|
|
|
/*issue old 0x04 command to adapter */
|
|
|
|
mbox->m_out.cmd = MEGA_MBOXCMD_ADPEXTINQ;
|
|
|
|
|
|
|
|
issue_scb_block(adapter, raw_mbox);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* update Enquiry3 and ProductInfo structures with
|
|
|
|
* mraid_inquiry structure
|
|
|
|
*/
|
|
|
|
mega_8_to_40ld(inq, inquiry3,
|
|
|
|
(mega_product_info *)&adapter->product_info);
|
|
|
|
|
|
|
|
pci_free_consistent(adapter->dev, sizeof(mraid_ext_inquiry),
|
|
|
|
ext_inq, dma_handle);
|
|
|
|
|
|
|
|
} else { /*adapter supports 40ld */
|
|
|
|
adapter->flag |= BOARD_40LD;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get product_info, which is static information and will be
|
|
|
|
* unchanged
|
|
|
|
*/
|
|
|
|
prod_info_dma_handle = pci_map_single(adapter->dev, (void *)
|
|
|
|
&adapter->product_info,
|
|
|
|
sizeof(mega_product_info), PCI_DMA_FROMDEVICE);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = prod_info_dma_handle;
|
|
|
|
|
|
|
|
raw_mbox[0] = FC_NEW_CONFIG; /* i.e. mbox->cmd=0xA1 */
|
|
|
|
raw_mbox[2] = NC_SUBOP_PRODUCT_INFO; /* i.e. 0x0E */
|
|
|
|
|
|
|
|
if ((retval = issue_scb_block(adapter, raw_mbox)))
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: Product_info cmd failed with error: %d\n",
|
|
|
|
retval);
|
|
|
|
|
|
|
|
pci_unmap_single(adapter->dev, prod_info_dma_handle,
|
|
|
|
sizeof(mega_product_info), PCI_DMA_FROMDEVICE);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* kernel scans the channels from 0 to <= max_channel
|
|
|
|
*/
|
|
|
|
adapter->host->max_channel =
|
|
|
|
adapter->product_info.nchannels + NVIRT_CHAN -1;
|
|
|
|
|
|
|
|
adapter->host->max_id = 16; /* max targets per channel */
|
|
|
|
|
|
|
|
adapter->host->max_lun = 7; /* Upto 7 luns for non disk devices */
|
|
|
|
|
|
|
|
adapter->host->cmd_per_lun = max_cmd_per_lun;
|
|
|
|
|
|
|
|
adapter->numldrv = inquiry3->num_ldrv;
|
|
|
|
|
|
|
|
adapter->max_cmds = adapter->product_info.max_commands;
|
|
|
|
|
|
|
|
if(adapter->max_cmds > MAX_COMMANDS)
|
|
|
|
adapter->max_cmds = MAX_COMMANDS;
|
|
|
|
|
|
|
|
adapter->host->can_queue = adapter->max_cmds - 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get the maximum number of scatter-gather elements supported by this
|
|
|
|
* firmware
|
|
|
|
*/
|
|
|
|
mega_get_max_sgl(adapter);
|
|
|
|
|
|
|
|
adapter->host->sg_tablesize = adapter->sglen;
|
|
|
|
|
|
|
|
|
|
|
|
/* use HP firmware and bios version encoding */
|
|
|
|
if (adapter->product_info.subsysvid == HP_SUBSYS_VID) {
|
|
|
|
sprintf (adapter->fw_version, "%c%d%d.%d%d",
|
|
|
|
adapter->product_info.fw_version[2],
|
|
|
|
adapter->product_info.fw_version[1] >> 8,
|
|
|
|
adapter->product_info.fw_version[1] & 0x0f,
|
|
|
|
adapter->product_info.fw_version[0] >> 8,
|
|
|
|
adapter->product_info.fw_version[0] & 0x0f);
|
|
|
|
sprintf (adapter->bios_version, "%c%d%d.%d%d",
|
|
|
|
adapter->product_info.bios_version[2],
|
|
|
|
adapter->product_info.bios_version[1] >> 8,
|
|
|
|
adapter->product_info.bios_version[1] & 0x0f,
|
|
|
|
adapter->product_info.bios_version[0] >> 8,
|
|
|
|
adapter->product_info.bios_version[0] & 0x0f);
|
|
|
|
} else {
|
|
|
|
memcpy(adapter->fw_version,
|
|
|
|
(char *)adapter->product_info.fw_version, 4);
|
|
|
|
adapter->fw_version[4] = 0;
|
|
|
|
|
|
|
|
memcpy(adapter->bios_version,
|
|
|
|
(char *)adapter->product_info.bios_version, 4);
|
|
|
|
|
|
|
|
adapter->bios_version[4] = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
printk(KERN_NOTICE "megaraid: [%s:%s] detected %d logical drives.\n",
|
|
|
|
adapter->fw_version, adapter->bios_version, adapter->numldrv);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do we support extended (>10 bytes) cdbs
|
|
|
|
*/
|
|
|
|
adapter->support_ext_cdb = mega_support_ext_cdb(adapter);
|
|
|
|
if (adapter->support_ext_cdb)
|
|
|
|
printk(KERN_NOTICE "megaraid: supports extended CDBs.\n");
|
|
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_runpendq()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Runs through the list of pending requests.
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
mega_runpendq(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
if(!list_empty(&adapter->pending_list))
|
|
|
|
__mega_runpendq(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* megaraid_queue()
|
|
|
|
* @scmd - Issue this scsi command
|
|
|
|
* @done - the callback hook into the scsi mid-layer
|
|
|
|
*
|
|
|
|
* The command queuing entry point for the mid-layer.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
megaraid_queue(Scsi_Cmnd *scmd, void (*done)(Scsi_Cmnd *))
|
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
scb_t *scb;
|
|
|
|
int busy=0;
|
2005-11-01 03:12:07 +08:00
|
|
|
unsigned long flags;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
adapter = (adapter_t *)scmd->device->host->hostdata;
|
|
|
|
|
|
|
|
scmd->scsi_done = done;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate and build a SCB request
|
|
|
|
* busy flag will be set if mega_build_cmd() command could not
|
|
|
|
* allocate scb. We will return non-zero status in that case.
|
|
|
|
* NOTE: scb can be null even though certain commands completed
|
|
|
|
* successfully, e.g., MODE_SENSE and TEST_UNIT_READY, we would
|
|
|
|
* return 0 in that case.
|
|
|
|
*/
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
spin_lock_irqsave(&adapter->lock, flags);
|
2005-04-17 06:20:36 +08:00
|
|
|
scb = mega_build_cmd(adapter, scmd, &busy);
|
2005-11-30 04:36:16 +08:00
|
|
|
if (!scb)
|
|
|
|
goto out;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-11-30 04:36:16 +08:00
|
|
|
scb->state |= SCB_PENDQ;
|
|
|
|
list_add_tail(&scb->list, &adapter->pending_list);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-11-30 04:36:16 +08:00
|
|
|
/*
|
|
|
|
* Check if the HBA is in quiescent state, e.g., during a
|
|
|
|
* delete logical drive opertion. If it is, don't run
|
|
|
|
* the pending_list.
|
|
|
|
*/
|
|
|
|
if (atomic_read(&adapter->quiescent) == 0)
|
|
|
|
mega_runpendq(adapter);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-11-30 04:36:16 +08:00
|
|
|
busy = 0;
|
|
|
|
out:
|
|
|
|
spin_unlock_irqrestore(&adapter->lock, flags);
|
2005-04-17 06:20:36 +08:00
|
|
|
return busy;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_allocate_scb()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @cmd - scsi command from the mid-layer
|
|
|
|
*
|
|
|
|
* Allocate a SCB structure. This is the central structure for controller
|
|
|
|
* commands.
|
|
|
|
*/
|
|
|
|
static inline scb_t *
|
|
|
|
mega_allocate_scb(adapter_t *adapter, Scsi_Cmnd *cmd)
|
|
|
|
{
|
|
|
|
struct list_head *head = &adapter->free_list;
|
|
|
|
scb_t *scb;
|
|
|
|
|
|
|
|
/* Unlink command from Free List */
|
|
|
|
if( !list_empty(head) ) {
|
|
|
|
|
|
|
|
scb = list_entry(head->next, scb_t, list);
|
|
|
|
|
|
|
|
list_del_init(head->next);
|
|
|
|
|
|
|
|
scb->state = SCB_ACTIVE;
|
|
|
|
scb->cmd = cmd;
|
|
|
|
scb->dma_type = MEGA_DMA_TYPE_NONE;
|
|
|
|
|
|
|
|
return scb;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_get_ldrv_num()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @cmd - scsi mid layer command
|
|
|
|
* @channel - channel on the controller
|
|
|
|
*
|
|
|
|
* Calculate the logical drive number based on the information in scsi command
|
|
|
|
* and the channel number.
|
|
|
|
*/
|
|
|
|
static inline int
|
|
|
|
mega_get_ldrv_num(adapter_t *adapter, Scsi_Cmnd *cmd, int channel)
|
|
|
|
{
|
|
|
|
int tgt;
|
|
|
|
int ldrv_num;
|
|
|
|
|
|
|
|
tgt = cmd->device->id;
|
|
|
|
|
|
|
|
if ( tgt > adapter->this_id )
|
|
|
|
tgt--; /* we do not get inquires for initiator id */
|
|
|
|
|
|
|
|
ldrv_num = (channel * 15) + tgt;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we have a logical drive with boot enabled, project it first
|
|
|
|
*/
|
|
|
|
if( adapter->boot_ldrv_enabled ) {
|
|
|
|
if( ldrv_num == 0 ) {
|
|
|
|
ldrv_num = adapter->boot_ldrv;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if( ldrv_num <= adapter->boot_ldrv ) {
|
|
|
|
ldrv_num--;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If "delete logical drive" feature is enabled on this controller.
|
|
|
|
* Do only if at least one delete logical drive operation was done.
|
|
|
|
*
|
|
|
|
* Also, after logical drive deletion, instead of logical drive number,
|
|
|
|
* the value returned should be 0x80+logical drive id.
|
|
|
|
*
|
|
|
|
* These is valid only for IO commands.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (adapter->support_random_del && adapter->read_ldidmap )
|
|
|
|
switch (cmd->cmnd[0]) {
|
|
|
|
case READ_6: /* fall through */
|
|
|
|
case WRITE_6: /* fall through */
|
|
|
|
case READ_10: /* fall through */
|
|
|
|
case WRITE_10:
|
|
|
|
ldrv_num += 0x80;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ldrv_num;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_build_cmd()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @cmd - Prepare using this scsi command
|
|
|
|
* @busy - busy flag if no resources
|
|
|
|
*
|
|
|
|
* Prepares a command and scatter gather list for the controller. This routine
|
|
|
|
* also finds out if the commands is intended for a logical drive or a
|
|
|
|
* physical device and prepares the controller command accordingly.
|
|
|
|
*
|
|
|
|
* We also re-order the logical drives and physical devices based on their
|
|
|
|
* boot settings.
|
|
|
|
*/
|
|
|
|
static scb_t *
|
|
|
|
mega_build_cmd(adapter_t *adapter, Scsi_Cmnd *cmd, int *busy)
|
|
|
|
{
|
|
|
|
mega_ext_passthru *epthru;
|
|
|
|
mega_passthru *pthru;
|
|
|
|
scb_t *scb;
|
|
|
|
mbox_t *mbox;
|
|
|
|
long seg;
|
|
|
|
char islogical;
|
|
|
|
int max_ldrv_num;
|
|
|
|
int channel = 0;
|
|
|
|
int target = 0;
|
|
|
|
int ldrv_num = 0; /* logical drive number */
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* filter the internal and ioctl commands
|
|
|
|
*/
|
|
|
|
if((cmd->cmnd[0] == MEGA_INTERNAL_CMD)) {
|
2006-06-03 19:21:13 +08:00
|
|
|
return cmd->request_buffer;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We know what channels our logical drives are on - mega_find_card()
|
|
|
|
*/
|
|
|
|
islogical = adapter->logdrv_chan[cmd->device->channel];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The theory: If physical drive is chosen for boot, all the physical
|
|
|
|
* devices are exported before the logical drives, otherwise physical
|
|
|
|
* devices are pushed after logical drives, in which case - Kernel sees
|
|
|
|
* the physical devices on virtual channel which is obviously converted
|
|
|
|
* to actual channel on the HBA.
|
|
|
|
*/
|
|
|
|
if( adapter->boot_pdrv_enabled ) {
|
|
|
|
if( islogical ) {
|
|
|
|
/* logical channel */
|
|
|
|
channel = cmd->device->channel -
|
|
|
|
adapter->product_info.nchannels;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* this is physical channel */
|
|
|
|
channel = cmd->device->channel;
|
|
|
|
target = cmd->device->id;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* boot from a physical disk, that disk needs to be
|
|
|
|
* exposed first IF both the channels are SCSI, then
|
|
|
|
* booting from the second channel is not allowed.
|
|
|
|
*/
|
|
|
|
if( target == 0 ) {
|
|
|
|
target = adapter->boot_pdrv_tgt;
|
|
|
|
}
|
|
|
|
else if( target == adapter->boot_pdrv_tgt ) {
|
|
|
|
target = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if( islogical ) {
|
|
|
|
/* this is the logical channel */
|
|
|
|
channel = cmd->device->channel;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* physical channel */
|
|
|
|
channel = cmd->device->channel - NVIRT_CHAN;
|
|
|
|
target = cmd->device->id;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
if(islogical) {
|
|
|
|
|
|
|
|
/* have just LUN 0 for each target on virtual channels */
|
|
|
|
if (cmd->device->lun) {
|
|
|
|
cmd->result = (DID_BAD_TARGET << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
ldrv_num = mega_get_ldrv_num(adapter, cmd, channel);
|
|
|
|
|
|
|
|
|
|
|
|
max_ldrv_num = (adapter->flag & BOARD_40LD) ?
|
|
|
|
MAX_LOGICAL_DRIVES_40LD : MAX_LOGICAL_DRIVES_8LD;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* max_ldrv_num increases by 0x80 if some logical drive was
|
|
|
|
* deleted.
|
|
|
|
*/
|
|
|
|
if(adapter->read_ldidmap)
|
|
|
|
max_ldrv_num += 0x80;
|
|
|
|
|
|
|
|
if(ldrv_num > max_ldrv_num ) {
|
|
|
|
cmd->result = (DID_BAD_TARGET << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if( cmd->device->lun > 7) {
|
|
|
|
/*
|
|
|
|
* Do not support lun >7 for physically accessed
|
|
|
|
* devices
|
|
|
|
*/
|
|
|
|
cmd->result = (DID_BAD_TARGET << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
*
|
|
|
|
* Logical drive commands
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
if(islogical) {
|
|
|
|
switch (cmd->cmnd[0]) {
|
|
|
|
case TEST_UNIT_READY:
|
|
|
|
#if MEGA_HAVE_CLUSTERING
|
|
|
|
/*
|
|
|
|
* Do we support clustering and is the support enabled
|
|
|
|
* If no, return success always
|
|
|
|
*/
|
|
|
|
if( !adapter->has_cluster ) {
|
|
|
|
cmd->result = (DID_OK << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(!(scb = mega_allocate_scb(adapter, cmd))) {
|
|
|
|
*busy = 1;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
scb->raw_mbox[0] = MEGA_CLUSTER_CMD;
|
|
|
|
scb->raw_mbox[2] = MEGA_RESERVATION_STATUS;
|
|
|
|
scb->raw_mbox[3] = ldrv_num;
|
|
|
|
|
|
|
|
scb->dma_direction = PCI_DMA_NONE;
|
|
|
|
|
|
|
|
return scb;
|
|
|
|
#else
|
|
|
|
cmd->result = (DID_OK << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
#endif
|
|
|
|
|
2005-10-01 22:38:05 +08:00
|
|
|
case MODE_SENSE: {
|
|
|
|
char *buf;
|
|
|
|
|
|
|
|
if (cmd->use_sg) {
|
|
|
|
struct scatterlist *sg;
|
|
|
|
|
|
|
|
sg = (struct scatterlist *)cmd->request_buffer;
|
|
|
|
buf = kmap_atomic(sg->page, KM_IRQ0) +
|
|
|
|
sg->offset;
|
|
|
|
} else
|
|
|
|
buf = cmd->request_buffer;
|
2005-12-08 06:46:57 +08:00
|
|
|
memset(buf, 0, cmd->cmnd[4]);
|
2005-10-01 22:38:05 +08:00
|
|
|
if (cmd->use_sg) {
|
|
|
|
struct scatterlist *sg;
|
|
|
|
|
|
|
|
sg = (struct scatterlist *)cmd->request_buffer;
|
|
|
|
kunmap_atomic(buf - sg->offset, KM_IRQ0);
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
cmd->result = (DID_OK << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
2005-10-01 22:38:05 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
case READ_CAPACITY:
|
|
|
|
case INQUIRY:
|
|
|
|
|
|
|
|
if(!(adapter->flag & (1L << cmd->device->channel))) {
|
|
|
|
|
|
|
|
printk(KERN_NOTICE
|
|
|
|
"scsi%d: scanning scsi channel %d ",
|
|
|
|
adapter->host->host_no,
|
|
|
|
cmd->device->channel);
|
|
|
|
printk("for logical drives.\n");
|
|
|
|
|
|
|
|
adapter->flag |= (1L << cmd->device->channel);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Allocate a SCB and initialize passthru */
|
|
|
|
if(!(scb = mega_allocate_scb(adapter, cmd))) {
|
|
|
|
*busy = 1;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
pthru = scb->pthru;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)scb->raw_mbox;
|
|
|
|
memset(mbox, 0, sizeof(scb->raw_mbox));
|
|
|
|
memset(pthru, 0, sizeof(mega_passthru));
|
|
|
|
|
|
|
|
pthru->timeout = 0;
|
|
|
|
pthru->ars = 1;
|
|
|
|
pthru->reqsenselen = 14;
|
|
|
|
pthru->islogical = 1;
|
|
|
|
pthru->logdrv = ldrv_num;
|
|
|
|
pthru->cdblen = cmd->cmd_len;
|
|
|
|
memcpy(pthru->cdb, cmd->cmnd, cmd->cmd_len);
|
|
|
|
|
|
|
|
if( adapter->has_64bit_addr ) {
|
|
|
|
mbox->m_out.cmd = MEGA_MBOXCMD_PASSTHRU64;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
mbox->m_out.cmd = MEGA_MBOXCMD_PASSTHRU;
|
|
|
|
}
|
|
|
|
|
|
|
|
scb->dma_direction = PCI_DMA_FROMDEVICE;
|
|
|
|
|
|
|
|
pthru->numsgelements = mega_build_sglist(adapter, scb,
|
|
|
|
&pthru->dataxferaddr, &pthru->dataxferlen);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = scb->pthru_dma_addr;
|
|
|
|
|
|
|
|
return scb;
|
|
|
|
|
|
|
|
case READ_6:
|
|
|
|
case WRITE_6:
|
|
|
|
case READ_10:
|
|
|
|
case WRITE_10:
|
|
|
|
case READ_12:
|
|
|
|
case WRITE_12:
|
|
|
|
|
|
|
|
/* Allocate a SCB and initialize mailbox */
|
|
|
|
if(!(scb = mega_allocate_scb(adapter, cmd))) {
|
|
|
|
*busy = 1;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
mbox = (mbox_t *)scb->raw_mbox;
|
|
|
|
|
|
|
|
memset(mbox, 0, sizeof(scb->raw_mbox));
|
|
|
|
mbox->m_out.logdrv = ldrv_num;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* A little hack: 2nd bit is zero for all scsi read
|
|
|
|
* commands and is set for all scsi write commands
|
|
|
|
*/
|
|
|
|
if( adapter->has_64bit_addr ) {
|
|
|
|
mbox->m_out.cmd = (*cmd->cmnd & 0x02) ?
|
|
|
|
MEGA_MBOXCMD_LWRITE64:
|
|
|
|
MEGA_MBOXCMD_LREAD64 ;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
mbox->m_out.cmd = (*cmd->cmnd & 0x02) ?
|
|
|
|
MEGA_MBOXCMD_LWRITE:
|
|
|
|
MEGA_MBOXCMD_LREAD ;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 6-byte READ(0x08) or WRITE(0x0A) cdb
|
|
|
|
*/
|
|
|
|
if( cmd->cmd_len == 6 ) {
|
|
|
|
mbox->m_out.numsectors = (u32) cmd->cmnd[4];
|
|
|
|
mbox->m_out.lba =
|
|
|
|
((u32)cmd->cmnd[1] << 16) |
|
|
|
|
((u32)cmd->cmnd[2] << 8) |
|
|
|
|
(u32)cmd->cmnd[3];
|
|
|
|
|
|
|
|
mbox->m_out.lba &= 0x1FFFFF;
|
|
|
|
|
|
|
|
#if MEGA_HAVE_STATS
|
|
|
|
/*
|
|
|
|
* Take modulo 0x80, since the logical drive
|
|
|
|
* number increases by 0x80 when a logical
|
|
|
|
* drive was deleted
|
|
|
|
*/
|
|
|
|
if (*cmd->cmnd == READ_6) {
|
|
|
|
adapter->nreads[ldrv_num%0x80]++;
|
|
|
|
adapter->nreadblocks[ldrv_num%0x80] +=
|
|
|
|
mbox->m_out.numsectors;
|
|
|
|
} else {
|
|
|
|
adapter->nwrites[ldrv_num%0x80]++;
|
|
|
|
adapter->nwriteblocks[ldrv_num%0x80] +=
|
|
|
|
mbox->m_out.numsectors;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 10-byte READ(0x28) or WRITE(0x2A) cdb
|
|
|
|
*/
|
|
|
|
if( cmd->cmd_len == 10 ) {
|
|
|
|
mbox->m_out.numsectors =
|
|
|
|
(u32)cmd->cmnd[8] |
|
|
|
|
((u32)cmd->cmnd[7] << 8);
|
|
|
|
mbox->m_out.lba =
|
|
|
|
((u32)cmd->cmnd[2] << 24) |
|
|
|
|
((u32)cmd->cmnd[3] << 16) |
|
|
|
|
((u32)cmd->cmnd[4] << 8) |
|
|
|
|
(u32)cmd->cmnd[5];
|
|
|
|
|
|
|
|
#if MEGA_HAVE_STATS
|
|
|
|
if (*cmd->cmnd == READ_10) {
|
|
|
|
adapter->nreads[ldrv_num%0x80]++;
|
|
|
|
adapter->nreadblocks[ldrv_num%0x80] +=
|
|
|
|
mbox->m_out.numsectors;
|
|
|
|
} else {
|
|
|
|
adapter->nwrites[ldrv_num%0x80]++;
|
|
|
|
adapter->nwriteblocks[ldrv_num%0x80] +=
|
|
|
|
mbox->m_out.numsectors;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 12-byte READ(0xA8) or WRITE(0xAA) cdb
|
|
|
|
*/
|
|
|
|
if( cmd->cmd_len == 12 ) {
|
|
|
|
mbox->m_out.lba =
|
|
|
|
((u32)cmd->cmnd[2] << 24) |
|
|
|
|
((u32)cmd->cmnd[3] << 16) |
|
|
|
|
((u32)cmd->cmnd[4] << 8) |
|
|
|
|
(u32)cmd->cmnd[5];
|
|
|
|
|
|
|
|
mbox->m_out.numsectors =
|
|
|
|
((u32)cmd->cmnd[6] << 24) |
|
|
|
|
((u32)cmd->cmnd[7] << 16) |
|
|
|
|
((u32)cmd->cmnd[8] << 8) |
|
|
|
|
(u32)cmd->cmnd[9];
|
|
|
|
|
|
|
|
#if MEGA_HAVE_STATS
|
|
|
|
if (*cmd->cmnd == READ_12) {
|
|
|
|
adapter->nreads[ldrv_num%0x80]++;
|
|
|
|
adapter->nreadblocks[ldrv_num%0x80] +=
|
|
|
|
mbox->m_out.numsectors;
|
|
|
|
} else {
|
|
|
|
adapter->nwrites[ldrv_num%0x80]++;
|
|
|
|
adapter->nwriteblocks[ldrv_num%0x80] +=
|
|
|
|
mbox->m_out.numsectors;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If it is a read command
|
|
|
|
*/
|
|
|
|
if( (*cmd->cmnd & 0x0F) == 0x08 ) {
|
|
|
|
scb->dma_direction = PCI_DMA_FROMDEVICE;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
scb->dma_direction = PCI_DMA_TODEVICE;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Calculate Scatter-Gather info */
|
|
|
|
mbox->m_out.numsgelements = mega_build_sglist(adapter, scb,
|
|
|
|
(u32 *)&mbox->m_out.xferaddr, (u32 *)&seg);
|
|
|
|
|
|
|
|
return scb;
|
|
|
|
|
|
|
|
#if MEGA_HAVE_CLUSTERING
|
|
|
|
case RESERVE: /* Fall through */
|
|
|
|
case RELEASE:
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do we support clustering and is the support enabled
|
|
|
|
*/
|
|
|
|
if( ! adapter->has_cluster ) {
|
|
|
|
|
|
|
|
cmd->result = (DID_BAD_TARGET << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Allocate a SCB and initialize mailbox */
|
|
|
|
if(!(scb = mega_allocate_scb(adapter, cmd))) {
|
|
|
|
*busy = 1;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
scb->raw_mbox[0] = MEGA_CLUSTER_CMD;
|
|
|
|
scb->raw_mbox[2] = ( *cmd->cmnd == RESERVE ) ?
|
|
|
|
MEGA_RESERVE_LD : MEGA_RELEASE_LD;
|
|
|
|
|
|
|
|
scb->raw_mbox[3] = ldrv_num;
|
|
|
|
|
|
|
|
scb->dma_direction = PCI_DMA_NONE;
|
|
|
|
|
|
|
|
return scb;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
default:
|
|
|
|
cmd->result = (DID_BAD_TARGET << 16);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Passthru drive commands
|
|
|
|
*/
|
|
|
|
else {
|
|
|
|
/* Allocate a SCB and initialize passthru */
|
|
|
|
if(!(scb = mega_allocate_scb(adapter, cmd))) {
|
|
|
|
*busy = 1;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
mbox = (mbox_t *)scb->raw_mbox;
|
|
|
|
memset(mbox, 0, sizeof(scb->raw_mbox));
|
|
|
|
|
|
|
|
if( adapter->support_ext_cdb ) {
|
|
|
|
|
|
|
|
epthru = mega_prepare_extpassthru(adapter, scb, cmd,
|
|
|
|
channel, target);
|
|
|
|
|
|
|
|
mbox->m_out.cmd = MEGA_MBOXCMD_EXTPTHRU;
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = scb->epthru_dma_addr;
|
|
|
|
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
|
|
|
|
pthru = mega_prepare_passthru(adapter, scb, cmd,
|
|
|
|
channel, target);
|
|
|
|
|
|
|
|
/* Initialize mailbox */
|
|
|
|
if( adapter->has_64bit_addr ) {
|
|
|
|
mbox->m_out.cmd = MEGA_MBOXCMD_PASSTHRU64;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
mbox->m_out.cmd = MEGA_MBOXCMD_PASSTHRU;
|
|
|
|
}
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = scb->pthru_dma_addr;
|
|
|
|
|
|
|
|
}
|
|
|
|
return scb;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_prepare_passthru()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @scb - our scsi control block
|
|
|
|
* @cmd - scsi command from the mid-layer
|
|
|
|
* @channel - actual channel on the controller
|
|
|
|
* @target - actual id on the controller.
|
|
|
|
*
|
|
|
|
* prepare a command for the scsi physical devices.
|
|
|
|
*/
|
|
|
|
static mega_passthru *
|
|
|
|
mega_prepare_passthru(adapter_t *adapter, scb_t *scb, Scsi_Cmnd *cmd,
|
|
|
|
int channel, int target)
|
|
|
|
{
|
|
|
|
mega_passthru *pthru;
|
|
|
|
|
|
|
|
pthru = scb->pthru;
|
|
|
|
memset(pthru, 0, sizeof (mega_passthru));
|
|
|
|
|
|
|
|
/* 0=6sec/1=60sec/2=10min/3=3hrs */
|
|
|
|
pthru->timeout = 2;
|
|
|
|
|
|
|
|
pthru->ars = 1;
|
|
|
|
pthru->reqsenselen = 14;
|
|
|
|
pthru->islogical = 0;
|
|
|
|
|
|
|
|
pthru->channel = (adapter->flag & BOARD_40LD) ? 0 : channel;
|
|
|
|
|
|
|
|
pthru->target = (adapter->flag & BOARD_40LD) ?
|
|
|
|
(channel << 4) | target : target;
|
|
|
|
|
|
|
|
pthru->cdblen = cmd->cmd_len;
|
|
|
|
pthru->logdrv = cmd->device->lun;
|
|
|
|
|
|
|
|
memcpy(pthru->cdb, cmd->cmnd, cmd->cmd_len);
|
|
|
|
|
|
|
|
/* Not sure about the direction */
|
|
|
|
scb->dma_direction = PCI_DMA_BIDIRECTIONAL;
|
|
|
|
|
|
|
|
/* Special Code for Handling READ_CAPA/ INQ using bounce buffers */
|
|
|
|
switch (cmd->cmnd[0]) {
|
|
|
|
case INQUIRY:
|
|
|
|
case READ_CAPACITY:
|
|
|
|
if(!(adapter->flag & (1L << cmd->device->channel))) {
|
|
|
|
|
|
|
|
printk(KERN_NOTICE
|
|
|
|
"scsi%d: scanning scsi channel %d [P%d] ",
|
|
|
|
adapter->host->host_no,
|
|
|
|
cmd->device->channel, channel);
|
|
|
|
printk("for physical devices.\n");
|
|
|
|
|
|
|
|
adapter->flag |= (1L << cmd->device->channel);
|
|
|
|
}
|
|
|
|
/* Fall through */
|
|
|
|
default:
|
|
|
|
pthru->numsgelements = mega_build_sglist(adapter, scb,
|
|
|
|
&pthru->dataxferaddr, &pthru->dataxferlen);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return pthru;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_prepare_extpassthru()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @scb - our scsi control block
|
|
|
|
* @cmd - scsi command from the mid-layer
|
|
|
|
* @channel - actual channel on the controller
|
|
|
|
* @target - actual id on the controller.
|
|
|
|
*
|
|
|
|
* prepare a command for the scsi physical devices. This rountine prepares
|
|
|
|
* commands for devices which can take extended CDBs (>10 bytes)
|
|
|
|
*/
|
|
|
|
static mega_ext_passthru *
|
|
|
|
mega_prepare_extpassthru(adapter_t *adapter, scb_t *scb, Scsi_Cmnd *cmd,
|
|
|
|
int channel, int target)
|
|
|
|
{
|
|
|
|
mega_ext_passthru *epthru;
|
|
|
|
|
|
|
|
epthru = scb->epthru;
|
|
|
|
memset(epthru, 0, sizeof(mega_ext_passthru));
|
|
|
|
|
|
|
|
/* 0=6sec/1=60sec/2=10min/3=3hrs */
|
|
|
|
epthru->timeout = 2;
|
|
|
|
|
|
|
|
epthru->ars = 1;
|
|
|
|
epthru->reqsenselen = 14;
|
|
|
|
epthru->islogical = 0;
|
|
|
|
|
|
|
|
epthru->channel = (adapter->flag & BOARD_40LD) ? 0 : channel;
|
|
|
|
epthru->target = (adapter->flag & BOARD_40LD) ?
|
|
|
|
(channel << 4) | target : target;
|
|
|
|
|
|
|
|
epthru->cdblen = cmd->cmd_len;
|
|
|
|
epthru->logdrv = cmd->device->lun;
|
|
|
|
|
|
|
|
memcpy(epthru->cdb, cmd->cmnd, cmd->cmd_len);
|
|
|
|
|
|
|
|
/* Not sure about the direction */
|
|
|
|
scb->dma_direction = PCI_DMA_BIDIRECTIONAL;
|
|
|
|
|
|
|
|
switch(cmd->cmnd[0]) {
|
|
|
|
case INQUIRY:
|
|
|
|
case READ_CAPACITY:
|
|
|
|
if(!(adapter->flag & (1L << cmd->device->channel))) {
|
|
|
|
|
|
|
|
printk(KERN_NOTICE
|
|
|
|
"scsi%d: scanning scsi channel %d [P%d] ",
|
|
|
|
adapter->host->host_no,
|
|
|
|
cmd->device->channel, channel);
|
|
|
|
printk("for physical devices.\n");
|
|
|
|
|
|
|
|
adapter->flag |= (1L << cmd->device->channel);
|
|
|
|
}
|
|
|
|
/* Fall through */
|
|
|
|
default:
|
|
|
|
epthru->numsgelements = mega_build_sglist(adapter, scb,
|
|
|
|
&epthru->dataxferaddr, &epthru->dataxferlen);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return epthru;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
__mega_runpendq(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
scb_t *scb;
|
|
|
|
struct list_head *pos, *next;
|
|
|
|
|
|
|
|
/* Issue any pending commands to the card */
|
|
|
|
list_for_each_safe(pos, next, &adapter->pending_list) {
|
|
|
|
|
|
|
|
scb = list_entry(pos, scb_t, list);
|
|
|
|
|
|
|
|
if( !(scb->state & SCB_ISSUED) ) {
|
|
|
|
|
|
|
|
if( issue_scb(adapter, scb) != 0 )
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* issue_scb()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @scb - scsi control block
|
|
|
|
*
|
|
|
|
* Post a command to the card if the mailbox is available, otherwise return
|
|
|
|
* busy. We also take the scb from the pending list if the mailbox is
|
|
|
|
* available.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
issue_scb(adapter_t *adapter, scb_t *scb)
|
|
|
|
{
|
|
|
|
volatile mbox64_t *mbox64 = adapter->mbox64;
|
|
|
|
volatile mbox_t *mbox = adapter->mbox;
|
|
|
|
unsigned int i = 0;
|
|
|
|
|
|
|
|
if(unlikely(mbox->m_in.busy)) {
|
|
|
|
do {
|
|
|
|
udelay(1);
|
|
|
|
i++;
|
|
|
|
} while( mbox->m_in.busy && (i < max_mbox_busy_wait) );
|
|
|
|
|
|
|
|
if(mbox->m_in.busy) return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy mailbox data into host structure */
|
|
|
|
memcpy((char *)&mbox->m_out, (char *)scb->raw_mbox,
|
|
|
|
sizeof(struct mbox_out));
|
|
|
|
|
|
|
|
mbox->m_out.cmdid = scb->idx; /* Set cmdid */
|
|
|
|
mbox->m_in.busy = 1; /* Set busy */
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Increment the pending queue counter
|
|
|
|
*/
|
|
|
|
atomic_inc(&adapter->pend_cmds);
|
|
|
|
|
|
|
|
switch (mbox->m_out.cmd) {
|
|
|
|
case MEGA_MBOXCMD_LREAD64:
|
|
|
|
case MEGA_MBOXCMD_LWRITE64:
|
|
|
|
case MEGA_MBOXCMD_PASSTHRU64:
|
|
|
|
case MEGA_MBOXCMD_EXTPTHRU:
|
|
|
|
mbox64->xfer_segment_lo = mbox->m_out.xferaddr;
|
|
|
|
mbox64->xfer_segment_hi = 0;
|
|
|
|
mbox->m_out.xferaddr = 0xFFFFFFFF;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
mbox64->xfer_segment_lo = 0;
|
|
|
|
mbox64->xfer_segment_hi = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* post the command
|
|
|
|
*/
|
|
|
|
scb->state |= SCB_ISSUED;
|
|
|
|
|
|
|
|
if( likely(adapter->flag & BOARD_MEMMAP) ) {
|
|
|
|
mbox->m_in.poll = 0;
|
|
|
|
mbox->m_in.ack = 0;
|
|
|
|
WRINDOOR(adapter, adapter->mbox_dma | 0x1);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
irq_enable(adapter);
|
|
|
|
issue_command(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Wait until the controller's mailbox is available
|
|
|
|
*/
|
|
|
|
static inline int
|
|
|
|
mega_busywait_mbox (adapter_t *adapter)
|
|
|
|
{
|
|
|
|
if (adapter->mbox->m_in.busy)
|
|
|
|
return __mega_busywait_mbox(adapter);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* issue_scb_block()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @raw_mbox - the mailbox
|
|
|
|
*
|
|
|
|
* Issue a scb in synchronous and non-interrupt mode
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
issue_scb_block(adapter_t *adapter, u_char *raw_mbox)
|
|
|
|
{
|
|
|
|
volatile mbox64_t *mbox64 = adapter->mbox64;
|
|
|
|
volatile mbox_t *mbox = adapter->mbox;
|
|
|
|
u8 byte;
|
|
|
|
|
|
|
|
/* Wait until mailbox is free */
|
|
|
|
if(mega_busywait_mbox (adapter))
|
|
|
|
goto bug_blocked_mailbox;
|
|
|
|
|
|
|
|
/* Copy mailbox data into host structure */
|
|
|
|
memcpy((char *) mbox, raw_mbox, sizeof(struct mbox_out));
|
|
|
|
mbox->m_out.cmdid = 0xFE;
|
|
|
|
mbox->m_in.busy = 1;
|
|
|
|
|
|
|
|
switch (raw_mbox[0]) {
|
|
|
|
case MEGA_MBOXCMD_LREAD64:
|
|
|
|
case MEGA_MBOXCMD_LWRITE64:
|
|
|
|
case MEGA_MBOXCMD_PASSTHRU64:
|
|
|
|
case MEGA_MBOXCMD_EXTPTHRU:
|
|
|
|
mbox64->xfer_segment_lo = mbox->m_out.xferaddr;
|
|
|
|
mbox64->xfer_segment_hi = 0;
|
|
|
|
mbox->m_out.xferaddr = 0xFFFFFFFF;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
mbox64->xfer_segment_lo = 0;
|
|
|
|
mbox64->xfer_segment_hi = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( likely(adapter->flag & BOARD_MEMMAP) ) {
|
|
|
|
mbox->m_in.poll = 0;
|
|
|
|
mbox->m_in.ack = 0;
|
|
|
|
mbox->m_in.numstatus = 0xFF;
|
|
|
|
mbox->m_in.status = 0xFF;
|
|
|
|
WRINDOOR(adapter, adapter->mbox_dma | 0x1);
|
|
|
|
|
|
|
|
while((volatile u8)mbox->m_in.numstatus == 0xFF)
|
|
|
|
cpu_relax();
|
|
|
|
|
|
|
|
mbox->m_in.numstatus = 0xFF;
|
|
|
|
|
|
|
|
while( (volatile u8)mbox->m_in.poll != 0x77 )
|
|
|
|
cpu_relax();
|
|
|
|
|
|
|
|
mbox->m_in.poll = 0;
|
|
|
|
mbox->m_in.ack = 0x77;
|
|
|
|
|
|
|
|
WRINDOOR(adapter, adapter->mbox_dma | 0x2);
|
|
|
|
|
|
|
|
while(RDINDOOR(adapter) & 0x2)
|
|
|
|
cpu_relax();
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
irq_disable(adapter);
|
|
|
|
issue_command(adapter);
|
|
|
|
|
|
|
|
while (!((byte = irq_state(adapter)) & INTR_VALID))
|
|
|
|
cpu_relax();
|
|
|
|
|
|
|
|
set_irq_state(adapter, byte);
|
|
|
|
irq_enable(adapter);
|
|
|
|
irq_ack(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
return mbox->m_in.status;
|
|
|
|
|
|
|
|
bug_blocked_mailbox:
|
|
|
|
printk(KERN_WARNING "megaraid: Blocked mailbox......!!\n");
|
|
|
|
udelay (1000);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* megaraid_isr_iomapped()
|
|
|
|
* @irq - irq
|
|
|
|
* @devp - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Interrupt service routine for io-mapped controllers.
|
|
|
|
* Find out if our device is interrupting. If yes, acknowledge the interrupt
|
|
|
|
* and service the completed commands.
|
|
|
|
*/
|
|
|
|
static irqreturn_t
|
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers
Maintain a per-CPU global "struct pt_regs *" variable which can be used instead
of passing regs around manually through all ~1800 interrupt handlers in the
Linux kernel.
The regs pointer is used in few places, but it potentially costs both stack
space and code to pass it around. On the FRV arch, removing the regs parameter
from all the genirq function results in a 20% speed up of the IRQ exit path
(ie: from leaving timer_interrupt() to leaving do_IRQ()).
Where appropriate, an arch may override the generic storage facility and do
something different with the variable. On FRV, for instance, the address is
maintained in GR28 at all times inside the kernel as part of general exception
handling.
Having looked over the code, it appears that the parameter may be handed down
through up to twenty or so layers of functions. Consider a USB character
device attached to a USB hub, attached to a USB controller that posts its
interrupts through a cascaded auxiliary interrupt controller. A character
device driver may want to pass regs to the sysrq handler through the input
layer which adds another few layers of parameter passing.
I've build this code with allyesconfig for x86_64 and i386. I've runtested the
main part of the code on FRV and i386, though I can't test most of the drivers.
I've also done partial conversion for powerpc and MIPS - these at least compile
with minimal configurations.
This will affect all archs. Mostly the changes should be relatively easy.
Take do_IRQ(), store the regs pointer at the beginning, saving the old one:
struct pt_regs *old_regs = set_irq_regs(regs);
And put the old one back at the end:
set_irq_regs(old_regs);
Don't pass regs through to generic_handle_irq() or __do_IRQ().
In timer_interrupt(), this sort of change will be necessary:
- update_process_times(user_mode(regs));
- profile_tick(CPU_PROFILING, regs);
+ update_process_times(user_mode(get_irq_regs()));
+ profile_tick(CPU_PROFILING);
I'd like to move update_process_times()'s use of get_irq_regs() into itself,
except that i386, alone of the archs, uses something other than user_mode().
Some notes on the interrupt handling in the drivers:
(*) input_dev() is now gone entirely. The regs pointer is no longer stored in
the input_dev struct.
(*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does
something different depending on whether it's been supplied with a regs
pointer or not.
(*) Various IRQ handler function pointers have been moved to type
irq_handler_t.
Signed-Off-By: David Howells <dhowells@redhat.com>
(cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
|
|
|
megaraid_isr_iomapped(int irq, void *devp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
adapter_t *adapter = devp;
|
|
|
|
unsigned long flags;
|
|
|
|
u8 status;
|
|
|
|
u8 nstatus;
|
|
|
|
u8 completed[MAX_FIRMWARE_STATUS];
|
|
|
|
u8 byte;
|
|
|
|
int handled = 0;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* loop till F/W has more commands for us to complete.
|
|
|
|
*/
|
|
|
|
spin_lock_irqsave(&adapter->lock, flags);
|
|
|
|
|
|
|
|
do {
|
|
|
|
/* Check if a valid interrupt is pending */
|
|
|
|
byte = irq_state(adapter);
|
|
|
|
if( (byte & VALID_INTR_BYTE) == 0 ) {
|
|
|
|
/*
|
|
|
|
* No more pending commands
|
|
|
|
*/
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
set_irq_state(adapter, byte);
|
|
|
|
|
|
|
|
while((nstatus = (volatile u8)adapter->mbox->m_in.numstatus)
|
|
|
|
== 0xFF)
|
|
|
|
cpu_relax();
|
|
|
|
adapter->mbox->m_in.numstatus = 0xFF;
|
|
|
|
|
|
|
|
status = adapter->mbox->m_in.status;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* decrement the pending queue counter
|
|
|
|
*/
|
|
|
|
atomic_sub(nstatus, &adapter->pend_cmds);
|
|
|
|
|
|
|
|
memcpy(completed, (void *)adapter->mbox->m_in.completed,
|
|
|
|
nstatus);
|
|
|
|
|
|
|
|
/* Acknowledge interrupt */
|
|
|
|
irq_ack(adapter);
|
|
|
|
|
|
|
|
mega_cmd_done(adapter, completed, nstatus, status);
|
|
|
|
|
|
|
|
mega_rundoneq(adapter);
|
|
|
|
|
|
|
|
handled = 1;
|
|
|
|
|
|
|
|
/* Loop through any pending requests */
|
|
|
|
if(atomic_read(&adapter->quiescent) == 0) {
|
|
|
|
mega_runpendq(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
} while(1);
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&adapter->lock, flags);
|
|
|
|
|
|
|
|
return IRQ_RETVAL(handled);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* megaraid_isr_memmapped()
|
|
|
|
* @irq - irq
|
|
|
|
* @devp - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Interrupt service routine for memory-mapped controllers.
|
|
|
|
* Find out if our device is interrupting. If yes, acknowledge the interrupt
|
|
|
|
* and service the completed commands.
|
|
|
|
*/
|
|
|
|
static irqreturn_t
|
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers
Maintain a per-CPU global "struct pt_regs *" variable which can be used instead
of passing regs around manually through all ~1800 interrupt handlers in the
Linux kernel.
The regs pointer is used in few places, but it potentially costs both stack
space and code to pass it around. On the FRV arch, removing the regs parameter
from all the genirq function results in a 20% speed up of the IRQ exit path
(ie: from leaving timer_interrupt() to leaving do_IRQ()).
Where appropriate, an arch may override the generic storage facility and do
something different with the variable. On FRV, for instance, the address is
maintained in GR28 at all times inside the kernel as part of general exception
handling.
Having looked over the code, it appears that the parameter may be handed down
through up to twenty or so layers of functions. Consider a USB character
device attached to a USB hub, attached to a USB controller that posts its
interrupts through a cascaded auxiliary interrupt controller. A character
device driver may want to pass regs to the sysrq handler through the input
layer which adds another few layers of parameter passing.
I've build this code with allyesconfig for x86_64 and i386. I've runtested the
main part of the code on FRV and i386, though I can't test most of the drivers.
I've also done partial conversion for powerpc and MIPS - these at least compile
with minimal configurations.
This will affect all archs. Mostly the changes should be relatively easy.
Take do_IRQ(), store the regs pointer at the beginning, saving the old one:
struct pt_regs *old_regs = set_irq_regs(regs);
And put the old one back at the end:
set_irq_regs(old_regs);
Don't pass regs through to generic_handle_irq() or __do_IRQ().
In timer_interrupt(), this sort of change will be necessary:
- update_process_times(user_mode(regs));
- profile_tick(CPU_PROFILING, regs);
+ update_process_times(user_mode(get_irq_regs()));
+ profile_tick(CPU_PROFILING);
I'd like to move update_process_times()'s use of get_irq_regs() into itself,
except that i386, alone of the archs, uses something other than user_mode().
Some notes on the interrupt handling in the drivers:
(*) input_dev() is now gone entirely. The regs pointer is no longer stored in
the input_dev struct.
(*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does
something different depending on whether it's been supplied with a regs
pointer or not.
(*) Various IRQ handler function pointers have been moved to type
irq_handler_t.
Signed-Off-By: David Howells <dhowells@redhat.com>
(cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
|
|
|
megaraid_isr_memmapped(int irq, void *devp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
adapter_t *adapter = devp;
|
|
|
|
unsigned long flags;
|
|
|
|
u8 status;
|
|
|
|
u32 dword = 0;
|
|
|
|
u8 nstatus;
|
|
|
|
u8 completed[MAX_FIRMWARE_STATUS];
|
|
|
|
int handled = 0;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* loop till F/W has more commands for us to complete.
|
|
|
|
*/
|
|
|
|
spin_lock_irqsave(&adapter->lock, flags);
|
|
|
|
|
|
|
|
do {
|
|
|
|
/* Check if a valid interrupt is pending */
|
|
|
|
dword = RDOUTDOOR(adapter);
|
|
|
|
if(dword != 0x10001234) {
|
|
|
|
/*
|
|
|
|
* No more pending commands
|
|
|
|
*/
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
WROUTDOOR(adapter, 0x10001234);
|
|
|
|
|
|
|
|
while((nstatus = (volatile u8)adapter->mbox->m_in.numstatus)
|
|
|
|
== 0xFF) {
|
|
|
|
cpu_relax();
|
|
|
|
}
|
|
|
|
adapter->mbox->m_in.numstatus = 0xFF;
|
|
|
|
|
|
|
|
status = adapter->mbox->m_in.status;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* decrement the pending queue counter
|
|
|
|
*/
|
|
|
|
atomic_sub(nstatus, &adapter->pend_cmds);
|
|
|
|
|
|
|
|
memcpy(completed, (void *)adapter->mbox->m_in.completed,
|
|
|
|
nstatus);
|
|
|
|
|
|
|
|
/* Acknowledge interrupt */
|
|
|
|
WRINDOOR(adapter, 0x2);
|
|
|
|
|
|
|
|
handled = 1;
|
|
|
|
|
2006-12-04 09:49:23 +08:00
|
|
|
while( RDINDOOR(adapter) & 0x02 )
|
|
|
|
cpu_relax();
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
mega_cmd_done(adapter, completed, nstatus, status);
|
|
|
|
|
|
|
|
mega_rundoneq(adapter);
|
|
|
|
|
|
|
|
/* Loop through any pending requests */
|
|
|
|
if(atomic_read(&adapter->quiescent) == 0) {
|
|
|
|
mega_runpendq(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
} while(1);
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&adapter->lock, flags);
|
|
|
|
|
|
|
|
return IRQ_RETVAL(handled);
|
|
|
|
}
|
|
|
|
/**
|
|
|
|
* mega_cmd_done()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @completed - array of ids of completed commands
|
|
|
|
* @nstatus - number of completed commands
|
|
|
|
* @status - status of the last command completed
|
|
|
|
*
|
|
|
|
* Complete the comamnds and call the scsi mid-layer callback hooks.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_cmd_done(adapter_t *adapter, u8 completed[], int nstatus, int status)
|
|
|
|
{
|
|
|
|
mega_ext_passthru *epthru = NULL;
|
|
|
|
struct scatterlist *sgl;
|
|
|
|
Scsi_Cmnd *cmd = NULL;
|
|
|
|
mega_passthru *pthru = NULL;
|
|
|
|
mbox_t *mbox = NULL;
|
|
|
|
u8 c;
|
|
|
|
scb_t *scb;
|
|
|
|
int islogical;
|
|
|
|
int cmdid;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* for all the commands completed, call the mid-layer callback routine
|
|
|
|
* and free the scb.
|
|
|
|
*/
|
|
|
|
for( i = 0; i < nstatus; i++ ) {
|
|
|
|
|
|
|
|
cmdid = completed[i];
|
|
|
|
|
|
|
|
if( cmdid == CMDID_INT_CMDS ) { /* internal command */
|
|
|
|
scb = &adapter->int_scb;
|
|
|
|
cmd = scb->cmd;
|
|
|
|
mbox = (mbox_t *)scb->raw_mbox;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Internal command interface do not fire the extended
|
|
|
|
* passthru or 64-bit passthru
|
|
|
|
*/
|
|
|
|
pthru = scb->pthru;
|
|
|
|
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
scb = &adapter->scb_list[cmdid];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure f/w has completed a valid command
|
|
|
|
*/
|
|
|
|
if( !(scb->state & SCB_ISSUED) || scb->cmd == NULL ) {
|
|
|
|
printk(KERN_CRIT
|
|
|
|
"megaraid: invalid command ");
|
|
|
|
printk("Id %d, scb->state:%x, scsi cmd:%p\n",
|
|
|
|
cmdid, scb->state, scb->cmd);
|
|
|
|
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Was a abort issued for this command
|
|
|
|
*/
|
|
|
|
if( scb->state & SCB_ABORT ) {
|
|
|
|
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: aborted cmd %lx[%x] complete.\n",
|
|
|
|
scb->cmd->serial_number, scb->idx);
|
|
|
|
|
|
|
|
scb->cmd->result = (DID_ABORT << 16);
|
|
|
|
|
|
|
|
list_add_tail(SCSI_LIST(scb->cmd),
|
|
|
|
&adapter->completed_list);
|
|
|
|
|
|
|
|
mega_free_scb(adapter, scb);
|
|
|
|
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Was a reset issued for this command
|
|
|
|
*/
|
|
|
|
if( scb->state & SCB_RESET ) {
|
|
|
|
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: reset cmd %lx[%x] complete.\n",
|
|
|
|
scb->cmd->serial_number, scb->idx);
|
|
|
|
|
|
|
|
scb->cmd->result = (DID_RESET << 16);
|
|
|
|
|
|
|
|
list_add_tail(SCSI_LIST(scb->cmd),
|
|
|
|
&adapter->completed_list);
|
|
|
|
|
|
|
|
mega_free_scb (adapter, scb);
|
|
|
|
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
cmd = scb->cmd;
|
|
|
|
pthru = scb->pthru;
|
|
|
|
epthru = scb->epthru;
|
|
|
|
mbox = (mbox_t *)scb->raw_mbox;
|
|
|
|
|
|
|
|
#if MEGA_HAVE_STATS
|
|
|
|
{
|
|
|
|
|
|
|
|
int logdrv = mbox->m_out.logdrv;
|
|
|
|
|
|
|
|
islogical = adapter->logdrv_chan[cmd->channel];
|
|
|
|
/*
|
|
|
|
* Maintain an error counter for the logical drive.
|
|
|
|
* Some application like SNMP agent need such
|
|
|
|
* statistics
|
|
|
|
*/
|
|
|
|
if( status && islogical && (cmd->cmnd[0] == READ_6 ||
|
|
|
|
cmd->cmnd[0] == READ_10 ||
|
|
|
|
cmd->cmnd[0] == READ_12)) {
|
|
|
|
/*
|
|
|
|
* Logical drive number increases by 0x80 when
|
|
|
|
* a logical drive is deleted
|
|
|
|
*/
|
|
|
|
adapter->rd_errors[logdrv%0x80]++;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( status && islogical && (cmd->cmnd[0] == WRITE_6 ||
|
|
|
|
cmd->cmnd[0] == WRITE_10 ||
|
|
|
|
cmd->cmnd[0] == WRITE_12)) {
|
|
|
|
/*
|
|
|
|
* Logical drive number increases by 0x80 when
|
|
|
|
* a logical drive is deleted
|
|
|
|
*/
|
|
|
|
adapter->wr_errors[logdrv%0x80]++;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do not return the presence of hard disk on the channel so,
|
|
|
|
* inquiry sent, and returned data==hard disk or removable
|
|
|
|
* hard disk and not logical, request should return failure! -
|
|
|
|
* PJ
|
|
|
|
*/
|
|
|
|
islogical = adapter->logdrv_chan[cmd->device->channel];
|
|
|
|
if( cmd->cmnd[0] == INQUIRY && !islogical ) {
|
|
|
|
|
|
|
|
if( cmd->use_sg ) {
|
|
|
|
sgl = (struct scatterlist *)
|
|
|
|
cmd->request_buffer;
|
|
|
|
|
|
|
|
if( sgl->page ) {
|
|
|
|
c = *(unsigned char *)
|
|
|
|
page_address((&sgl[0])->page) +
|
|
|
|
(&sgl[0])->offset;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: invalid sg.\n");
|
|
|
|
c = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
c = *(u8 *)cmd->request_buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(IS_RAID_CH(adapter, cmd->device->channel) &&
|
|
|
|
((c & 0x1F ) == TYPE_DISK)) {
|
|
|
|
status = 0xF0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* clear result; otherwise, success returns corrupt value */
|
|
|
|
cmd->result = 0;
|
|
|
|
|
|
|
|
/* Convert MegaRAID status to Linux error code */
|
|
|
|
switch (status) {
|
|
|
|
case 0x00: /* SUCCESS , i.e. SCSI_STATUS_GOOD */
|
|
|
|
cmd->result |= (DID_OK << 16);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x02: /* ERROR_ABORTED, i.e.
|
|
|
|
SCSI_STATUS_CHECK_CONDITION */
|
|
|
|
|
|
|
|
/* set sense_buffer and result fields */
|
|
|
|
if( mbox->m_out.cmd == MEGA_MBOXCMD_PASSTHRU ||
|
|
|
|
mbox->m_out.cmd == MEGA_MBOXCMD_PASSTHRU64 ) {
|
|
|
|
|
|
|
|
memcpy(cmd->sense_buffer, pthru->reqsensearea,
|
|
|
|
14);
|
|
|
|
|
|
|
|
cmd->result = (DRIVER_SENSE << 24) |
|
|
|
|
(DID_OK << 16) |
|
|
|
|
(CHECK_CONDITION << 1);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if (mbox->m_out.cmd == MEGA_MBOXCMD_EXTPTHRU) {
|
|
|
|
|
|
|
|
memcpy(cmd->sense_buffer,
|
|
|
|
epthru->reqsensearea, 14);
|
|
|
|
|
|
|
|
cmd->result = (DRIVER_SENSE << 24) |
|
|
|
|
(DID_OK << 16) |
|
|
|
|
(CHECK_CONDITION << 1);
|
|
|
|
} else {
|
|
|
|
cmd->sense_buffer[0] = 0x70;
|
|
|
|
cmd->sense_buffer[2] = ABORTED_COMMAND;
|
|
|
|
cmd->result |= (CHECK_CONDITION << 1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x08: /* ERR_DEST_DRIVE_FAILED, i.e.
|
|
|
|
SCSI_STATUS_BUSY */
|
|
|
|
cmd->result |= (DID_BUS_BUSY << 16) | status;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
#if MEGA_HAVE_CLUSTERING
|
|
|
|
/*
|
|
|
|
* If TEST_UNIT_READY fails, we know
|
|
|
|
* MEGA_RESERVATION_STATUS failed
|
|
|
|
*/
|
|
|
|
if( cmd->cmnd[0] == TEST_UNIT_READY ) {
|
|
|
|
cmd->result |= (DID_ERROR << 16) |
|
|
|
|
(RESERVATION_CONFLICT << 1);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
/*
|
|
|
|
* Error code returned is 1 if Reserve or Release
|
|
|
|
* failed or the input parameter is invalid
|
|
|
|
*/
|
|
|
|
if( status == 1 &&
|
|
|
|
(cmd->cmnd[0] == RESERVE ||
|
|
|
|
cmd->cmnd[0] == RELEASE) ) {
|
|
|
|
|
|
|
|
cmd->result |= (DID_ERROR << 16) |
|
|
|
|
(RESERVATION_CONFLICT << 1);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
#endif
|
|
|
|
cmd->result |= (DID_BAD_TARGET << 16)|status;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Only free SCBs for the commands coming down from the
|
|
|
|
* mid-layer, not for which were issued internally
|
|
|
|
*
|
|
|
|
* For internal command, restore the status returned by the
|
|
|
|
* firmware so that user can interpret it.
|
|
|
|
*/
|
|
|
|
if( cmdid == CMDID_INT_CMDS ) { /* internal command */
|
|
|
|
cmd->result = status;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove the internal command from the pending list
|
|
|
|
*/
|
|
|
|
list_del_init(&scb->list);
|
|
|
|
scb->state = SCB_FREE;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
mega_free_scb(adapter, scb);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Add Scsi_Command to end of completed queue */
|
|
|
|
list_add_tail(SCSI_LIST(cmd), &adapter->completed_list);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mega_runpendq()
|
|
|
|
*
|
|
|
|
* Run through the list of completed requests and finish it
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_rundoneq (adapter_t *adapter)
|
|
|
|
{
|
|
|
|
Scsi_Cmnd *cmd;
|
|
|
|
struct list_head *pos;
|
|
|
|
|
|
|
|
list_for_each(pos, &adapter->completed_list) {
|
|
|
|
|
2005-11-01 01:31:56 +08:00
|
|
|
struct scsi_pointer* spos = (struct scsi_pointer *)pos;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
cmd = list_entry(spos, Scsi_Cmnd, SCp);
|
|
|
|
cmd->scsi_done(cmd);
|
|
|
|
}
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&adapter->completed_list);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Free a SCB structure
|
|
|
|
* Note: We assume the scsi commands associated with this scb is not free yet.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_free_scb(adapter_t *adapter, scb_t *scb)
|
|
|
|
{
|
2005-10-01 22:38:05 +08:00
|
|
|
unsigned long length;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
switch( scb->dma_type ) {
|
|
|
|
|
|
|
|
case MEGA_DMA_TYPE_NONE:
|
|
|
|
break;
|
|
|
|
|
|
|
|
case MEGA_BULK_DATA:
|
2005-10-01 22:38:05 +08:00
|
|
|
if (scb->cmd->use_sg == 0)
|
|
|
|
length = scb->cmd->request_bufflen;
|
|
|
|
else {
|
|
|
|
struct scatterlist *sgl =
|
|
|
|
(struct scatterlist *)scb->cmd->request_buffer;
|
|
|
|
length = sgl->length;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
pci_unmap_page(adapter->dev, scb->dma_h_bulkdata,
|
2005-10-01 22:38:05 +08:00
|
|
|
length, scb->dma_direction);
|
2005-04-17 06:20:36 +08:00
|
|
|
break;
|
|
|
|
|
|
|
|
case MEGA_SGLIST:
|
|
|
|
pci_unmap_sg(adapter->dev, scb->cmd->request_buffer,
|
|
|
|
scb->cmd->use_sg, scb->dma_direction);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove from the pending list
|
|
|
|
*/
|
|
|
|
list_del_init(&scb->list);
|
|
|
|
|
|
|
|
/* Link the scb back into free list */
|
|
|
|
scb->state = SCB_FREE;
|
|
|
|
scb->cmd = NULL;
|
|
|
|
|
|
|
|
list_add(&scb->list, &adapter->free_list);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
__mega_busywait_mbox (adapter_t *adapter)
|
|
|
|
{
|
|
|
|
volatile mbox_t *mbox = adapter->mbox;
|
|
|
|
long counter;
|
|
|
|
|
|
|
|
for (counter = 0; counter < 10000; counter++) {
|
|
|
|
if (!mbox->m_in.busy)
|
|
|
|
return 0;
|
|
|
|
udelay(100); yield();
|
|
|
|
}
|
|
|
|
return -1; /* give up after 1 second */
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copies data to SGLIST
|
|
|
|
* Note: For 64 bit cards, we need a minimum of one SG element for read/write
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_build_sglist(adapter_t *adapter, scb_t *scb, u32 *buf, u32 *len)
|
|
|
|
{
|
|
|
|
struct scatterlist *sgl;
|
|
|
|
struct page *page;
|
|
|
|
unsigned long offset;
|
2005-10-01 22:38:05 +08:00
|
|
|
unsigned int length;
|
2005-04-17 06:20:36 +08:00
|
|
|
Scsi_Cmnd *cmd;
|
|
|
|
int sgcnt;
|
|
|
|
int idx;
|
|
|
|
|
|
|
|
cmd = scb->cmd;
|
|
|
|
|
|
|
|
/* Scatter-gather not used */
|
2005-10-01 22:38:05 +08:00
|
|
|
if( cmd->use_sg == 0 || (cmd->use_sg == 1 &&
|
|
|
|
!adapter->has_64bit_addr)) {
|
|
|
|
|
|
|
|
if (cmd->use_sg == 0) {
|
|
|
|
page = virt_to_page(cmd->request_buffer);
|
|
|
|
offset = offset_in_page(cmd->request_buffer);
|
|
|
|
length = cmd->request_bufflen;
|
|
|
|
} else {
|
|
|
|
sgl = (struct scatterlist *)cmd->request_buffer;
|
|
|
|
page = sgl->page;
|
|
|
|
offset = sgl->offset;
|
|
|
|
length = sgl->length;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
scb->dma_h_bulkdata = pci_map_page(adapter->dev,
|
|
|
|
page, offset,
|
2005-10-01 22:38:05 +08:00
|
|
|
length,
|
2005-04-17 06:20:36 +08:00
|
|
|
scb->dma_direction);
|
|
|
|
scb->dma_type = MEGA_BULK_DATA;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We need to handle special 64-bit commands that need a
|
|
|
|
* minimum of 1 SG
|
|
|
|
*/
|
|
|
|
if( adapter->has_64bit_addr ) {
|
|
|
|
scb->sgl64[0].address = scb->dma_h_bulkdata;
|
2005-10-01 22:38:05 +08:00
|
|
|
scb->sgl64[0].length = length;
|
2005-04-17 06:20:36 +08:00
|
|
|
*buf = (u32)scb->sgl_dma_addr;
|
2005-10-01 22:38:05 +08:00
|
|
|
*len = (u32)length;
|
2005-04-17 06:20:36 +08:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
*buf = (u32)scb->dma_h_bulkdata;
|
2005-10-01 22:38:05 +08:00
|
|
|
*len = (u32)length;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
sgl = (struct scatterlist *)cmd->request_buffer;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy Scatter-Gather list info into controller structure.
|
|
|
|
*
|
|
|
|
* The number of sg elements returned must not exceed our limit
|
|
|
|
*/
|
|
|
|
sgcnt = pci_map_sg(adapter->dev, sgl, cmd->use_sg,
|
|
|
|
scb->dma_direction);
|
|
|
|
|
|
|
|
scb->dma_type = MEGA_SGLIST;
|
|
|
|
|
2006-06-23 17:06:06 +08:00
|
|
|
BUG_ON(sgcnt > adapter->sglen);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-10-01 22:38:05 +08:00
|
|
|
*len = 0;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
for( idx = 0; idx < sgcnt; idx++, sgl++ ) {
|
|
|
|
|
|
|
|
if( adapter->has_64bit_addr ) {
|
|
|
|
scb->sgl64[idx].address = sg_dma_address(sgl);
|
2005-10-01 22:38:05 +08:00
|
|
|
*len += scb->sgl64[idx].length = sg_dma_len(sgl);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
else {
|
|
|
|
scb->sgl[idx].address = sg_dma_address(sgl);
|
2005-10-01 22:38:05 +08:00
|
|
|
*len += scb->sgl[idx].length = sg_dma_len(sgl);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Reset pointer and length fields */
|
|
|
|
*buf = scb->sgl_dma_addr;
|
|
|
|
|
|
|
|
/* Return count of SG requests */
|
|
|
|
return sgcnt;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mega_8_to_40ld()
|
|
|
|
*
|
|
|
|
* takes all info in AdapterInquiry structure and puts it into ProductInfo and
|
|
|
|
* Enquiry3 structures for later use
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_8_to_40ld(mraid_inquiry *inquiry, mega_inquiry3 *enquiry3,
|
|
|
|
mega_product_info *product_info)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
product_info->max_commands = inquiry->adapter_info.max_commands;
|
|
|
|
enquiry3->rebuild_rate = inquiry->adapter_info.rebuild_rate;
|
|
|
|
product_info->nchannels = inquiry->adapter_info.nchannels;
|
|
|
|
|
|
|
|
for (i = 0; i < 4; i++) {
|
|
|
|
product_info->fw_version[i] =
|
|
|
|
inquiry->adapter_info.fw_version[i];
|
|
|
|
|
|
|
|
product_info->bios_version[i] =
|
|
|
|
inquiry->adapter_info.bios_version[i];
|
|
|
|
}
|
|
|
|
enquiry3->cache_flush_interval =
|
|
|
|
inquiry->adapter_info.cache_flush_interval;
|
|
|
|
|
|
|
|
product_info->dram_size = inquiry->adapter_info.dram_size;
|
|
|
|
|
|
|
|
enquiry3->num_ldrv = inquiry->logdrv_info.num_ldrv;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_LOGICAL_DRIVES_8LD; i++) {
|
|
|
|
enquiry3->ldrv_size[i] = inquiry->logdrv_info.ldrv_size[i];
|
|
|
|
enquiry3->ldrv_prop[i] = inquiry->logdrv_info.ldrv_prop[i];
|
|
|
|
enquiry3->ldrv_state[i] = inquiry->logdrv_info.ldrv_state[i];
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < (MAX_PHYSICAL_DRIVES); i++)
|
|
|
|
enquiry3->pdrv_state[i] = inquiry->pdrv_info.pdrv_state[i];
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
mega_free_sgl(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
scb_t *scb;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for(i = 0; i < adapter->max_cmds; i++) {
|
|
|
|
|
|
|
|
scb = &adapter->scb_list[i];
|
|
|
|
|
|
|
|
if( scb->sgl64 ) {
|
|
|
|
pci_free_consistent(adapter->dev,
|
|
|
|
sizeof(mega_sgl64) * adapter->sglen,
|
|
|
|
scb->sgl64,
|
|
|
|
scb->sgl_dma_addr);
|
|
|
|
|
|
|
|
scb->sgl64 = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( scb->pthru ) {
|
|
|
|
pci_free_consistent(adapter->dev, sizeof(mega_passthru),
|
|
|
|
scb->pthru, scb->pthru_dma_addr);
|
|
|
|
|
|
|
|
scb->pthru = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( scb->epthru ) {
|
|
|
|
pci_free_consistent(adapter->dev,
|
|
|
|
sizeof(mega_ext_passthru),
|
|
|
|
scb->epthru, scb->epthru_dma_addr);
|
|
|
|
|
|
|
|
scb->epthru = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get information about the card/driver
|
|
|
|
*/
|
|
|
|
const char *
|
|
|
|
megaraid_info(struct Scsi_Host *host)
|
|
|
|
{
|
|
|
|
static char buffer[512];
|
|
|
|
adapter_t *adapter;
|
|
|
|
|
|
|
|
adapter = (adapter_t *)host->hostdata;
|
|
|
|
|
|
|
|
sprintf (buffer,
|
|
|
|
"LSI Logic MegaRAID %s %d commands %d targs %d chans %d luns",
|
|
|
|
adapter->fw_version, adapter->product_info.max_commands,
|
|
|
|
adapter->host->max_id, adapter->host->max_channel,
|
|
|
|
adapter->host->max_lun);
|
|
|
|
return buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Abort a previous SCSI request. Only commands on the pending list can be
|
|
|
|
* aborted. All the commands issued to the F/W must complete.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
megaraid_abort(Scsi_Cmnd *cmd)
|
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
adapter = (adapter_t *)cmd->device->host->hostdata;
|
|
|
|
|
|
|
|
rval = megaraid_abort_and_reset(adapter, cmd, SCB_ABORT);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is required here to complete any completed requests
|
|
|
|
* to be communicated over to the mid layer.
|
|
|
|
*/
|
|
|
|
mega_rundoneq(adapter);
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int
|
2005-06-26 21:45:39 +08:00
|
|
|
megaraid_reset(struct scsi_cmnd *cmd)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
megacmd_t mc;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
adapter = (adapter_t *)cmd->device->host->hostdata;
|
|
|
|
|
|
|
|
#if MEGA_HAVE_CLUSTERING
|
|
|
|
mc.cmd = MEGA_CLUSTER_CMD;
|
|
|
|
mc.opcode = MEGA_RESET_RESERVATIONS;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
if( mega_internal_command(adapter, &mc, NULL) != 0 ) {
|
2005-04-17 06:20:36 +08:00
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: reservation reset failed.\n");
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
printk(KERN_INFO "megaraid: reservation reset.\n");
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-06-26 21:45:39 +08:00
|
|
|
spin_lock_irq(&adapter->lock);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
rval = megaraid_abort_and_reset(adapter, cmd, SCB_RESET);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is required here to complete any completed requests
|
|
|
|
* to be communicated over to the mid layer.
|
|
|
|
*/
|
|
|
|
mega_rundoneq(adapter);
|
2005-05-28 19:55:48 +08:00
|
|
|
spin_unlock_irq(&adapter->lock);
|
|
|
|
|
2005-06-26 21:45:39 +08:00
|
|
|
return rval;
|
2005-05-28 19:55:48 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* megaraid_abort_and_reset()
|
|
|
|
* @adapter - megaraid soft state
|
|
|
|
* @cmd - scsi command to be aborted or reset
|
|
|
|
* @aor - abort or reset flag
|
|
|
|
*
|
|
|
|
* Try to locate the scsi command in the pending queue. If found and is not
|
|
|
|
* issued to the controller, abort/reset it. Otherwise return failure
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
megaraid_abort_and_reset(adapter_t *adapter, Scsi_Cmnd *cmd, int aor)
|
|
|
|
{
|
|
|
|
struct list_head *pos, *next;
|
|
|
|
scb_t *scb;
|
|
|
|
|
|
|
|
printk(KERN_WARNING "megaraid: %s-%lx cmd=%x <c=%d t=%d l=%d>\n",
|
|
|
|
(aor == SCB_ABORT)? "ABORTING":"RESET", cmd->serial_number,
|
|
|
|
cmd->cmnd[0], cmd->device->channel,
|
|
|
|
cmd->device->id, cmd->device->lun);
|
|
|
|
|
|
|
|
if(list_empty(&adapter->pending_list))
|
|
|
|
return FALSE;
|
|
|
|
|
|
|
|
list_for_each_safe(pos, next, &adapter->pending_list) {
|
|
|
|
|
|
|
|
scb = list_entry(pos, scb_t, list);
|
|
|
|
|
|
|
|
if (scb->cmd == cmd) { /* Found command */
|
|
|
|
|
|
|
|
scb->state |= aor;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if this command has firmare owenership. If
|
|
|
|
* yes, we cannot reset this command. Whenever, f/w
|
|
|
|
* completes this command, we will return appropriate
|
|
|
|
* status from ISR.
|
|
|
|
*/
|
|
|
|
if( scb->state & SCB_ISSUED ) {
|
|
|
|
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: %s-%lx[%x], fw owner.\n",
|
|
|
|
(aor==SCB_ABORT) ? "ABORTING":"RESET",
|
|
|
|
cmd->serial_number, scb->idx);
|
|
|
|
|
|
|
|
return FALSE;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Not yet issued! Remove from the pending
|
|
|
|
* list
|
|
|
|
*/
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: %s-%lx[%x], driver owner.\n",
|
|
|
|
(aor==SCB_ABORT) ? "ABORTING":"RESET",
|
|
|
|
cmd->serial_number, scb->idx);
|
|
|
|
|
|
|
|
mega_free_scb(adapter, scb);
|
|
|
|
|
|
|
|
if( aor == SCB_ABORT ) {
|
|
|
|
cmd->result = (DID_ABORT << 16);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
cmd->result = (DID_RESET << 16);
|
|
|
|
}
|
|
|
|
|
|
|
|
list_add_tail(SCSI_LIST(cmd),
|
|
|
|
&adapter->completed_list);
|
|
|
|
|
|
|
|
return TRUE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return FALSE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int
|
|
|
|
make_local_pdev(adapter_t *adapter, struct pci_dev **pdev)
|
|
|
|
{
|
|
|
|
*pdev = kmalloc(sizeof(struct pci_dev), GFP_KERNEL);
|
|
|
|
|
|
|
|
if( *pdev == NULL ) return -1;
|
|
|
|
|
|
|
|
memcpy(*pdev, adapter->dev, sizeof(struct pci_dev));
|
|
|
|
|
2006-03-28 17:56:48 +08:00
|
|
|
if( pci_set_dma_mask(*pdev, DMA_32BIT_MASK) != 0 ) {
|
2005-04-17 06:20:36 +08:00
|
|
|
kfree(*pdev);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
free_local_pdev(struct pci_dev *pdev)
|
|
|
|
{
|
|
|
|
kfree(pdev);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_allocate_inquiry()
|
|
|
|
* @dma_handle - handle returned for dma address
|
|
|
|
* @pdev - handle to pci device
|
|
|
|
*
|
|
|
|
* allocates memory for inquiry structure
|
|
|
|
*/
|
|
|
|
static inline void *
|
|
|
|
mega_allocate_inquiry(dma_addr_t *dma_handle, struct pci_dev *pdev)
|
|
|
|
{
|
|
|
|
return pci_alloc_consistent(pdev, sizeof(mega_inquiry3), dma_handle);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
mega_free_inquiry(void *inquiry, dma_addr_t dma_handle, struct pci_dev *pdev)
|
|
|
|
{
|
|
|
|
pci_free_consistent(pdev, sizeof(mega_inquiry3), inquiry, dma_handle);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
/* Following code handles /proc fs */
|
|
|
|
|
|
|
|
#define CREATE_READ_PROC(string, func) create_proc_read_entry(string, \
|
|
|
|
S_IRUSR | S_IFREG, \
|
|
|
|
controller_proc_dir_entry, \
|
|
|
|
func, adapter)
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_create_proc_entry()
|
|
|
|
* @index - index in soft state array
|
|
|
|
* @parent - parent node for this /proc entry
|
|
|
|
*
|
|
|
|
* Creates /proc entries for our controllers.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_create_proc_entry(int index, struct proc_dir_entry *parent)
|
|
|
|
{
|
|
|
|
struct proc_dir_entry *controller_proc_dir_entry = NULL;
|
|
|
|
u8 string[64] = { 0 };
|
|
|
|
adapter_t *adapter = hba_soft_state[index];
|
|
|
|
|
|
|
|
sprintf(string, "hba%d", adapter->host->host_no);
|
|
|
|
|
|
|
|
controller_proc_dir_entry =
|
|
|
|
adapter->controller_proc_dir_entry = proc_mkdir(string, parent);
|
|
|
|
|
|
|
|
if(!controller_proc_dir_entry) {
|
|
|
|
printk(KERN_WARNING "\nmegaraid: proc_mkdir failed\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
adapter->proc_read = CREATE_READ_PROC("config", proc_read_config);
|
|
|
|
adapter->proc_stat = CREATE_READ_PROC("stat", proc_read_stat);
|
|
|
|
adapter->proc_mbox = CREATE_READ_PROC("mailbox", proc_read_mbox);
|
|
|
|
#if MEGA_HAVE_ENH_PROC
|
|
|
|
adapter->proc_rr = CREATE_READ_PROC("rebuild-rate", proc_rebuild_rate);
|
|
|
|
adapter->proc_battery = CREATE_READ_PROC("battery-status",
|
|
|
|
proc_battery);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Display each physical drive on its channel
|
|
|
|
*/
|
|
|
|
adapter->proc_pdrvstat[0] = CREATE_READ_PROC("diskdrives-ch0",
|
|
|
|
proc_pdrv_ch0);
|
|
|
|
adapter->proc_pdrvstat[1] = CREATE_READ_PROC("diskdrives-ch1",
|
|
|
|
proc_pdrv_ch1);
|
|
|
|
adapter->proc_pdrvstat[2] = CREATE_READ_PROC("diskdrives-ch2",
|
|
|
|
proc_pdrv_ch2);
|
|
|
|
adapter->proc_pdrvstat[3] = CREATE_READ_PROC("diskdrives-ch3",
|
|
|
|
proc_pdrv_ch3);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Display a set of up to 10 logical drive through each of following
|
|
|
|
* /proc entries
|
|
|
|
*/
|
|
|
|
adapter->proc_rdrvstat[0] = CREATE_READ_PROC("raiddrives-0-9",
|
|
|
|
proc_rdrv_10);
|
|
|
|
adapter->proc_rdrvstat[1] = CREATE_READ_PROC("raiddrives-10-19",
|
|
|
|
proc_rdrv_20);
|
|
|
|
adapter->proc_rdrvstat[2] = CREATE_READ_PROC("raiddrives-20-29",
|
|
|
|
proc_rdrv_30);
|
|
|
|
adapter->proc_rdrvstat[3] = CREATE_READ_PROC("raiddrives-30-39",
|
|
|
|
proc_rdrv_40);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_read_config()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display configuration information about the controller.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_read_config(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
int len = 0;
|
|
|
|
|
|
|
|
len += sprintf(page+len, "%s", MEGARAID_VERSION);
|
|
|
|
|
|
|
|
if(adapter->product_info.product_name[0])
|
|
|
|
len += sprintf(page+len, "%s\n",
|
|
|
|
adapter->product_info.product_name);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Controller Type: ");
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_MEMMAP ) {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"438/466/467/471/493/518/520/531/532\n");
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"418/428/434\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
if(adapter->flag & BOARD_40LD) {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"Controller Supports 40 Logical Drives\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
if(adapter->flag & BOARD_64BIT) {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"Controller capable of 64-bit memory addressing\n");
|
|
|
|
}
|
|
|
|
if( adapter->has_64bit_addr ) {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"Controller using 64-bit memory addressing\n");
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"Controller is not using 64-bit memory addressing\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Base = %08lx, Irq = %d, ", adapter->base,
|
|
|
|
adapter->host->irq);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Logical Drives = %d, Channels = %d\n",
|
|
|
|
adapter->numldrv, adapter->product_info.nchannels);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Version =%s:%s, DRAM = %dMb\n",
|
|
|
|
adapter->fw_version, adapter->bios_version,
|
|
|
|
adapter->product_info.dram_size);
|
|
|
|
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"Controller Queue Depth = %d, Driver Queue Depth = %d\n",
|
|
|
|
adapter->product_info.max_commands, adapter->max_cmds);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "support_ext_cdb = %d\n",
|
|
|
|
adapter->support_ext_cdb);
|
|
|
|
len += sprintf(page+len, "support_random_del = %d\n",
|
|
|
|
adapter->support_random_del);
|
|
|
|
len += sprintf(page+len, "boot_ldrv_enabled = %d\n",
|
|
|
|
adapter->boot_ldrv_enabled);
|
|
|
|
len += sprintf(page+len, "boot_ldrv = %d\n",
|
|
|
|
adapter->boot_ldrv);
|
|
|
|
len += sprintf(page+len, "boot_pdrv_enabled = %d\n",
|
|
|
|
adapter->boot_pdrv_enabled);
|
|
|
|
len += sprintf(page+len, "boot_pdrv_ch = %d\n",
|
|
|
|
adapter->boot_pdrv_ch);
|
|
|
|
len += sprintf(page+len, "boot_pdrv_tgt = %d\n",
|
|
|
|
adapter->boot_pdrv_tgt);
|
|
|
|
len += sprintf(page+len, "quiescent = %d\n",
|
|
|
|
atomic_read(&adapter->quiescent));
|
|
|
|
len += sprintf(page+len, "has_cluster = %d\n",
|
|
|
|
adapter->has_cluster);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "\nModule Parameters:\n");
|
|
|
|
len += sprintf(page+len, "max_cmd_per_lun = %d\n",
|
|
|
|
max_cmd_per_lun);
|
|
|
|
len += sprintf(page+len, "max_sectors_per_io = %d\n",
|
|
|
|
max_sectors_per_io);
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_read_stat()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Diaplay statistical information about the I/O activity.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_read_stat(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
int len;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
i = 0; /* avoid compilation warnings */
|
|
|
|
len = 0;
|
|
|
|
adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
len = sprintf(page, "Statistical Information for this controller\n");
|
|
|
|
len += sprintf(page+len, "pend_cmds = %d\n",
|
|
|
|
atomic_read(&adapter->pend_cmds));
|
|
|
|
#if MEGA_HAVE_STATS
|
|
|
|
for(i = 0; i < adapter->numldrv; i++) {
|
|
|
|
len += sprintf(page+len, "Logical Drive %d:\n", i);
|
|
|
|
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"\tReads Issued = %lu, Writes Issued = %lu\n",
|
|
|
|
adapter->nreads[i], adapter->nwrites[i]);
|
|
|
|
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"\tSectors Read = %lu, Sectors Written = %lu\n",
|
|
|
|
adapter->nreadblocks[i], adapter->nwriteblocks[i]);
|
|
|
|
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"\tRead errors = %lu, Write errors = %lu\n\n",
|
|
|
|
adapter->rd_errors[i], adapter->wr_errors[i]);
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
len += sprintf(page+len,
|
|
|
|
"IO and error counters not compiled in driver.\n");
|
|
|
|
#endif
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_read_mbox()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display mailbox information for the last command issued. This information
|
|
|
|
* is good for debugging.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_read_mbox(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
volatile mbox_t *mbox = adapter->mbox;
|
|
|
|
int len = 0;
|
|
|
|
|
|
|
|
len = sprintf(page, "Contents of Mail Box Structure\n");
|
|
|
|
len += sprintf(page+len, " Fw Command = 0x%02x\n",
|
|
|
|
mbox->m_out.cmd);
|
|
|
|
len += sprintf(page+len, " Cmd Sequence = 0x%02x\n",
|
|
|
|
mbox->m_out.cmdid);
|
|
|
|
len += sprintf(page+len, " No of Sectors= %04d\n",
|
|
|
|
mbox->m_out.numsectors);
|
|
|
|
len += sprintf(page+len, " LBA = 0x%02x\n",
|
|
|
|
mbox->m_out.lba);
|
|
|
|
len += sprintf(page+len, " DTA = 0x%08x\n",
|
|
|
|
mbox->m_out.xferaddr);
|
|
|
|
len += sprintf(page+len, " Logical Drive= 0x%02x\n",
|
|
|
|
mbox->m_out.logdrv);
|
|
|
|
len += sprintf(page+len, " No of SG Elmt= 0x%02x\n",
|
|
|
|
mbox->m_out.numsgelements);
|
|
|
|
len += sprintf(page+len, " Busy = %01x\n",
|
|
|
|
mbox->m_in.busy);
|
|
|
|
len += sprintf(page+len, " Status = 0x%02x\n",
|
|
|
|
mbox->m_in.status);
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_rebuild_rate()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display current rebuild rate
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_rebuild_rate(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
dma_addr_t dma_handle;
|
|
|
|
caddr_t inquiry;
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
int len = 0;
|
|
|
|
|
|
|
|
if( make_local_pdev(adapter, &pdev) != 0 ) {
|
|
|
|
*eof = 1;
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( (inquiry = mega_allocate_inquiry(&dma_handle, pdev)) == NULL ) {
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
*eof = 1;
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( mega_adapinq(adapter, dma_handle) != 0 ) {
|
|
|
|
|
|
|
|
len = sprintf(page, "Adapter inquiry failed.\n");
|
|
|
|
|
|
|
|
printk(KERN_WARNING "megaraid: inquiry failed.\n");
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
len = sprintf(page, "Rebuild Rate: [%d%%]\n",
|
|
|
|
((mega_inquiry3 *)inquiry)->rebuild_rate);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
len = sprintf(page, "Rebuild Rate: [%d%%]\n",
|
|
|
|
((mraid_ext_inquiry *)
|
|
|
|
inquiry)->raid_inq.adapter_info.rebuild_rate);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_battery()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display information about the battery module on the controller.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_battery(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
dma_addr_t dma_handle;
|
|
|
|
caddr_t inquiry;
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
u8 battery_status = 0;
|
|
|
|
char str[256];
|
|
|
|
int len = 0;
|
|
|
|
|
|
|
|
if( make_local_pdev(adapter, &pdev) != 0 ) {
|
|
|
|
*eof = 1;
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( (inquiry = mega_allocate_inquiry(&dma_handle, pdev)) == NULL ) {
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
*eof = 1;
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( mega_adapinq(adapter, dma_handle) != 0 ) {
|
|
|
|
|
|
|
|
len = sprintf(page, "Adapter inquiry failed.\n");
|
|
|
|
|
|
|
|
printk(KERN_WARNING "megaraid: inquiry failed.\n");
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
battery_status = ((mega_inquiry3 *)inquiry)->battery_status;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
battery_status = ((mraid_ext_inquiry *)inquiry)->
|
|
|
|
raid_inq.adapter_info.battery_status;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Decode the battery status
|
|
|
|
*/
|
|
|
|
sprintf(str, "Battery Status:[%d]", battery_status);
|
|
|
|
|
|
|
|
if(battery_status == MEGA_BATT_CHARGE_DONE)
|
|
|
|
strcat(str, " Charge Done");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_MODULE_MISSING)
|
|
|
|
strcat(str, " Module Missing");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_LOW_VOLTAGE)
|
|
|
|
strcat(str, " Low Voltage");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_TEMP_HIGH)
|
|
|
|
strcat(str, " Temperature High");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_PACK_MISSING)
|
|
|
|
strcat(str, " Pack Missing");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_CHARGE_INPROG)
|
|
|
|
strcat(str, " Charge In-progress");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_CHARGE_FAIL)
|
|
|
|
strcat(str, " Charge Fail");
|
|
|
|
|
|
|
|
if(battery_status & MEGA_BATT_CYCLES_EXCEEDED)
|
|
|
|
strcat(str, " Cycles Exceeded");
|
|
|
|
|
|
|
|
len = sprintf(page, "%s\n", str);
|
|
|
|
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_pdrv_ch0()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display information about the physical drives on physical channel 0.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_pdrv_ch0(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_pdrv(adapter, page, 0));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_pdrv_ch1()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display information about the physical drives on physical channel 1.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_pdrv_ch1(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_pdrv(adapter, page, 1));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_pdrv_ch2()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display information about the physical drives on physical channel 2.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_pdrv_ch2(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_pdrv(adapter, page, 2));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_pdrv_ch3()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display information about the physical drives on physical channel 3.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_pdrv_ch3(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_pdrv(adapter, page, 3));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_pdrv()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display information about the physical drives.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_pdrv(adapter_t *adapter, char *page, int channel)
|
|
|
|
{
|
|
|
|
dma_addr_t dma_handle;
|
|
|
|
char *scsi_inq;
|
|
|
|
dma_addr_t scsi_inq_dma_handle;
|
|
|
|
caddr_t inquiry;
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
u8 *pdrv_state;
|
|
|
|
u8 state;
|
|
|
|
int tgt;
|
|
|
|
int max_channels;
|
|
|
|
int len = 0;
|
|
|
|
char str[80];
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if( make_local_pdev(adapter, &pdev) != 0 ) {
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( (inquiry = mega_allocate_inquiry(&dma_handle, pdev)) == NULL ) {
|
|
|
|
goto free_pdev;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( mega_adapinq(adapter, dma_handle) != 0 ) {
|
|
|
|
len = sprintf(page, "Adapter inquiry failed.\n");
|
|
|
|
|
|
|
|
printk(KERN_WARNING "megaraid: inquiry failed.\n");
|
|
|
|
|
|
|
|
goto free_inquiry;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
scsi_inq = pci_alloc_consistent(pdev, 256, &scsi_inq_dma_handle);
|
|
|
|
|
|
|
|
if( scsi_inq == NULL ) {
|
|
|
|
len = sprintf(page, "memory not available for scsi inq.\n");
|
|
|
|
|
|
|
|
goto free_inquiry;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
pdrv_state = ((mega_inquiry3 *)inquiry)->pdrv_state;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
pdrv_state = ((mraid_ext_inquiry *)inquiry)->
|
|
|
|
raid_inq.pdrv_info.pdrv_state;
|
|
|
|
}
|
|
|
|
|
|
|
|
max_channels = adapter->product_info.nchannels;
|
|
|
|
|
|
|
|
if( channel >= max_channels ) {
|
|
|
|
goto free_pci;
|
|
|
|
}
|
|
|
|
|
|
|
|
for( tgt = 0; tgt <= MAX_TARGET; tgt++ ) {
|
|
|
|
|
|
|
|
i = channel*16 + tgt;
|
|
|
|
|
|
|
|
state = *(pdrv_state + i);
|
|
|
|
|
|
|
|
switch( state & 0x0F ) {
|
|
|
|
|
|
|
|
case PDRV_ONLINE:
|
|
|
|
sprintf(str,
|
|
|
|
"Channel:%2d Id:%2d State: Online",
|
|
|
|
channel, tgt);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case PDRV_FAILED:
|
|
|
|
sprintf(str,
|
|
|
|
"Channel:%2d Id:%2d State: Failed",
|
|
|
|
channel, tgt);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case PDRV_RBLD:
|
|
|
|
sprintf(str,
|
|
|
|
"Channel:%2d Id:%2d State: Rebuild",
|
|
|
|
channel, tgt);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case PDRV_HOTSPARE:
|
|
|
|
sprintf(str,
|
|
|
|
"Channel:%2d Id:%2d State: Hot spare",
|
|
|
|
channel, tgt);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
sprintf(str,
|
|
|
|
"Channel:%2d Id:%2d State: Un-configured",
|
|
|
|
channel, tgt);
|
|
|
|
break;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This interface displays inquiries for disk drives
|
|
|
|
* only. Inquries for logical drives and non-disk
|
|
|
|
* devices are available through /proc/scsi/scsi
|
|
|
|
*/
|
|
|
|
memset(scsi_inq, 0, 256);
|
|
|
|
if( mega_internal_dev_inquiry(adapter, channel, tgt,
|
|
|
|
scsi_inq_dma_handle) ||
|
|
|
|
(scsi_inq[0] & 0x1F) != TYPE_DISK ) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for overflow. We print less than 240
|
|
|
|
* characters for inquiry
|
|
|
|
*/
|
|
|
|
if( (len + 240) >= PAGE_SIZE ) break;
|
|
|
|
|
|
|
|
len += sprintf(page+len, "%s.\n", str);
|
|
|
|
|
|
|
|
len += mega_print_inquiry(page+len, scsi_inq);
|
|
|
|
}
|
|
|
|
|
|
|
|
free_pci:
|
|
|
|
pci_free_consistent(pdev, 256, scsi_inq, scsi_inq_dma_handle);
|
|
|
|
free_inquiry:
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
free_pdev:
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Display scsi inquiry
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_print_inquiry(char *page, char *scsi_inq)
|
|
|
|
{
|
|
|
|
int len = 0;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
len = sprintf(page, " Vendor: ");
|
|
|
|
for( i = 8; i < 16; i++ ) {
|
|
|
|
len += sprintf(page+len, "%c", scsi_inq[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, " Model: ");
|
|
|
|
|
|
|
|
for( i = 16; i < 32; i++ ) {
|
|
|
|
len += sprintf(page+len, "%c", scsi_inq[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, " Rev: ");
|
|
|
|
|
|
|
|
for( i = 32; i < 36; i++ ) {
|
|
|
|
len += sprintf(page+len, "%c", scsi_inq[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, "\n");
|
|
|
|
|
|
|
|
i = scsi_inq[0] & 0x1f;
|
|
|
|
|
2006-07-05 02:15:20 +08:00
|
|
|
len += sprintf(page+len, " Type: %s ", scsi_device_type(i));
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
len += sprintf(page+len,
|
|
|
|
" ANSI SCSI revision: %02x", scsi_inq[2] & 0x07);
|
|
|
|
|
|
|
|
if( (scsi_inq[2] & 0x07) == 1 && (scsi_inq[3] & 0x0f) == 1 )
|
|
|
|
len += sprintf(page+len, " CCS\n");
|
|
|
|
else
|
|
|
|
len += sprintf(page+len, "\n");
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_rdrv_10()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display real time information about the logical drives 0 through 9.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_rdrv_10(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_rdrv(adapter, page, 0, 9));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_rdrv_20()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display real time information about the logical drives 0 through 9.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_rdrv_20(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_rdrv(adapter, page, 10, 19));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_rdrv_30()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display real time information about the logical drives 0 through 9.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_rdrv_30(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_rdrv(adapter, page, 20, 29));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_rdrv_40()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @start - where the actual data has been written in page
|
|
|
|
* @offset - same meaning as the read system call
|
|
|
|
* @count - same meaning as the read system call
|
|
|
|
* @eof - set if no more data needs to be returned
|
|
|
|
* @data - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Display real time information about the logical drives 0 through 9.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_rdrv_40(char *page, char **start, off_t offset, int count, int *eof,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
adapter_t *adapter = (adapter_t *)data;
|
|
|
|
|
|
|
|
*eof = 1;
|
|
|
|
|
|
|
|
return (proc_rdrv(adapter, page, 30, 39));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* proc_rdrv()
|
|
|
|
* @page - buffer to write the data in
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @start - starting logical drive to display
|
|
|
|
* @end - ending logical drive to display
|
|
|
|
*
|
|
|
|
* We do not print the inquiry information since its already available through
|
|
|
|
* /proc/scsi/scsi interface
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
proc_rdrv(adapter_t *adapter, char *page, int start, int end )
|
|
|
|
{
|
|
|
|
dma_addr_t dma_handle;
|
|
|
|
logdrv_param *lparam;
|
|
|
|
megacmd_t mc;
|
|
|
|
char *disk_array;
|
|
|
|
dma_addr_t disk_array_dma_handle;
|
|
|
|
caddr_t inquiry;
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
u8 *rdrv_state;
|
|
|
|
int num_ldrv;
|
|
|
|
u32 array_sz;
|
|
|
|
int len = 0;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if( make_local_pdev(adapter, &pdev) != 0 ) {
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( (inquiry = mega_allocate_inquiry(&dma_handle, pdev)) == NULL ) {
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
if( mega_adapinq(adapter, dma_handle) != 0 ) {
|
|
|
|
|
|
|
|
len = sprintf(page, "Adapter inquiry failed.\n");
|
|
|
|
|
|
|
|
printk(KERN_WARNING "megaraid: inquiry failed.\n");
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(&mc, 0, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
array_sz = sizeof(disk_array_40ld);
|
|
|
|
|
|
|
|
rdrv_state = ((mega_inquiry3 *)inquiry)->ldrv_state;
|
|
|
|
|
|
|
|
num_ldrv = ((mega_inquiry3 *)inquiry)->num_ldrv;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
array_sz = sizeof(disk_array_8ld);
|
|
|
|
|
|
|
|
rdrv_state = ((mraid_ext_inquiry *)inquiry)->
|
|
|
|
raid_inq.logdrv_info.ldrv_state;
|
|
|
|
|
|
|
|
num_ldrv = ((mraid_ext_inquiry *)inquiry)->
|
|
|
|
raid_inq.logdrv_info.num_ldrv;
|
|
|
|
}
|
|
|
|
|
|
|
|
disk_array = pci_alloc_consistent(pdev, array_sz,
|
|
|
|
&disk_array_dma_handle);
|
|
|
|
|
|
|
|
if( disk_array == NULL ) {
|
|
|
|
len = sprintf(page, "memory not available.\n");
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
mc.xferaddr = (u32)disk_array_dma_handle;
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
mc.cmd = FC_NEW_CONFIG;
|
|
|
|
mc.opcode = OP_DCMD_READ_CONFIG;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
if( mega_internal_command(adapter, &mc, NULL) ) {
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
len = sprintf(page, "40LD read config failed.\n");
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
pci_free_consistent(pdev, array_sz, disk_array,
|
|
|
|
disk_array_dma_handle);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
mc.cmd = NEW_READ_CONFIG_8LD;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
if( mega_internal_command(adapter, &mc, NULL) ) {
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
mc.cmd = READ_CONFIG_8LD;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
if( mega_internal_command(adapter, &mc,
|
2005-04-17 06:20:36 +08:00
|
|
|
NULL) ){
|
|
|
|
|
|
|
|
len = sprintf(page,
|
|
|
|
"8LD read config failed.\n");
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
pci_free_consistent(pdev, array_sz,
|
|
|
|
disk_array,
|
|
|
|
disk_array_dma_handle);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
for( i = start; i < ( (end+1 < num_ldrv) ? end+1 : num_ldrv ); i++ ) {
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
lparam =
|
|
|
|
&((disk_array_40ld *)disk_array)->ldrv[i].lparam;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
lparam =
|
|
|
|
&((disk_array_8ld *)disk_array)->ldrv[i].lparam;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for overflow. We print less than 240 characters for
|
|
|
|
* information about each logical drive.
|
|
|
|
*/
|
|
|
|
if( (len + 240) >= PAGE_SIZE ) break;
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Logical drive:%2d:, ", i);
|
|
|
|
|
|
|
|
switch( rdrv_state[i] & 0x0F ) {
|
|
|
|
case RDRV_OFFLINE:
|
|
|
|
len += sprintf(page+len, "state: offline");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RDRV_DEGRADED:
|
|
|
|
len += sprintf(page+len, "state: degraded");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RDRV_OPTIMAL:
|
|
|
|
len += sprintf(page+len, "state: optimal");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RDRV_DELETED:
|
|
|
|
len += sprintf(page+len, "state: deleted");
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
len += sprintf(page+len, "state: unknown");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if check consistency or initialization is going on
|
|
|
|
* for this logical drive.
|
|
|
|
*/
|
|
|
|
if( (rdrv_state[i] & 0xF0) == 0x20 ) {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
", check-consistency in progress");
|
|
|
|
}
|
|
|
|
else if( (rdrv_state[i] & 0xF0) == 0x10 ) {
|
|
|
|
len += sprintf(page+len,
|
|
|
|
", initialization in progress");
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, "\n");
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Span depth:%3d, ",
|
|
|
|
lparam->span_depth);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "RAID level:%3d, ",
|
|
|
|
lparam->level);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Stripe size:%3d, ",
|
|
|
|
lparam->stripe_sz ? lparam->stripe_sz/2: 128);
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Row size:%3d\n",
|
|
|
|
lparam->row_size);
|
|
|
|
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Read Policy: ");
|
|
|
|
|
|
|
|
switch(lparam->read_ahead) {
|
|
|
|
|
|
|
|
case NO_READ_AHEAD:
|
|
|
|
len += sprintf(page+len, "No read ahead, ");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case READ_AHEAD:
|
|
|
|
len += sprintf(page+len, "Read ahead, ");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case ADAP_READ_AHEAD:
|
|
|
|
len += sprintf(page+len, "Adaptive, ");
|
|
|
|
break;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Write Policy: ");
|
|
|
|
|
|
|
|
switch(lparam->write_mode) {
|
|
|
|
|
|
|
|
case WRMODE_WRITE_THRU:
|
|
|
|
len += sprintf(page+len, "Write thru, ");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case WRMODE_WRITE_BACK:
|
|
|
|
len += sprintf(page+len, "Write back, ");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
len += sprintf(page+len, "Cache Policy: ");
|
|
|
|
|
|
|
|
switch(lparam->direct_io) {
|
|
|
|
|
|
|
|
case CACHED_IO:
|
|
|
|
len += sprintf(page+len, "Cached IO\n\n");
|
|
|
|
break;
|
|
|
|
|
|
|
|
case DIRECT_IO:
|
|
|
|
len += sprintf(page+len, "Direct IO\n\n");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
mega_free_inquiry(inquiry, dma_handle, pdev);
|
|
|
|
|
|
|
|
pci_free_consistent(pdev, array_sz, disk_array,
|
|
|
|
disk_array_dma_handle);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* megaraid_biosparam()
|
|
|
|
*
|
|
|
|
* Return the disk geometry for a particular disk
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
megaraid_biosparam(struct scsi_device *sdev, struct block_device *bdev,
|
|
|
|
sector_t capacity, int geom[])
|
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
unsigned char *bh;
|
|
|
|
int heads;
|
|
|
|
int sectors;
|
|
|
|
int cylinders;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
/* Get pointer to host config structure */
|
|
|
|
adapter = (adapter_t *)sdev->host->hostdata;
|
|
|
|
|
|
|
|
if (IS_RAID_CH(adapter, sdev->channel)) {
|
|
|
|
/* Default heads (64) & sectors (32) */
|
|
|
|
heads = 64;
|
|
|
|
sectors = 32;
|
|
|
|
cylinders = (ulong)capacity / (heads * sectors);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle extended translation size for logical drives
|
|
|
|
* > 1Gb
|
|
|
|
*/
|
|
|
|
if ((ulong)capacity >= 0x200000) {
|
|
|
|
heads = 255;
|
|
|
|
sectors = 63;
|
|
|
|
cylinders = (ulong)capacity / (heads * sectors);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* return result */
|
|
|
|
geom[0] = heads;
|
|
|
|
geom[1] = sectors;
|
|
|
|
geom[2] = cylinders;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
bh = scsi_bios_ptable(bdev);
|
|
|
|
|
|
|
|
if( bh ) {
|
|
|
|
rval = scsi_partsize(bh, capacity,
|
|
|
|
&geom[2], &geom[0], &geom[1]);
|
|
|
|
kfree(bh);
|
|
|
|
if( rval != -1 )
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
printk(KERN_INFO
|
|
|
|
"megaraid: invalid partition on this disk on channel %d\n",
|
|
|
|
sdev->channel);
|
|
|
|
|
|
|
|
/* Default heads (64) & sectors (32) */
|
|
|
|
heads = 64;
|
|
|
|
sectors = 32;
|
|
|
|
cylinders = (ulong)capacity / (heads * sectors);
|
|
|
|
|
|
|
|
/* Handle extended translation size for logical drives > 1Gb */
|
|
|
|
if ((ulong)capacity >= 0x200000) {
|
|
|
|
heads = 255;
|
|
|
|
sectors = 63;
|
|
|
|
cylinders = (ulong)capacity / (heads * sectors);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* return result */
|
|
|
|
geom[0] = heads;
|
|
|
|
geom[1] = sectors;
|
|
|
|
geom[2] = cylinders;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_init_scb()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Allocate memory for the various pointers in the scb structures:
|
|
|
|
* scatter-gather list pointer, passthru and extended passthru structure
|
|
|
|
* pointers.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_init_scb(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
scb_t *scb;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for( i = 0; i < adapter->max_cmds; i++ ) {
|
|
|
|
|
|
|
|
scb = &adapter->scb_list[i];
|
|
|
|
|
|
|
|
scb->sgl64 = NULL;
|
|
|
|
scb->sgl = NULL;
|
|
|
|
scb->pthru = NULL;
|
|
|
|
scb->epthru = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
for( i = 0; i < adapter->max_cmds; i++ ) {
|
|
|
|
|
|
|
|
scb = &adapter->scb_list[i];
|
|
|
|
|
|
|
|
scb->idx = i;
|
|
|
|
|
|
|
|
scb->sgl64 = pci_alloc_consistent(adapter->dev,
|
|
|
|
sizeof(mega_sgl64) * adapter->sglen,
|
|
|
|
&scb->sgl_dma_addr);
|
|
|
|
|
|
|
|
scb->sgl = (mega_sglist *)scb->sgl64;
|
|
|
|
|
|
|
|
if( !scb->sgl ) {
|
|
|
|
printk(KERN_WARNING "RAID: Can't allocate sglist.\n");
|
|
|
|
mega_free_sgl(adapter);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
scb->pthru = pci_alloc_consistent(adapter->dev,
|
|
|
|
sizeof(mega_passthru),
|
|
|
|
&scb->pthru_dma_addr);
|
|
|
|
|
|
|
|
if( !scb->pthru ) {
|
|
|
|
printk(KERN_WARNING "RAID: Can't allocate passthru.\n");
|
|
|
|
mega_free_sgl(adapter);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
scb->epthru = pci_alloc_consistent(adapter->dev,
|
|
|
|
sizeof(mega_ext_passthru),
|
|
|
|
&scb->epthru_dma_addr);
|
|
|
|
|
|
|
|
if( !scb->epthru ) {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"Can't allocate extended passthru.\n");
|
|
|
|
mega_free_sgl(adapter);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
scb->dma_type = MEGA_DMA_TYPE_NONE;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Link to free list
|
|
|
|
* lock not required since we are loading the driver, so no
|
|
|
|
* commands possible right now.
|
|
|
|
*/
|
|
|
|
scb->state = SCB_FREE;
|
|
|
|
scb->cmd = NULL;
|
|
|
|
list_add(&scb->list, &adapter->free_list);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* megadev_open()
|
|
|
|
* @inode - unused
|
|
|
|
* @filep - unused
|
|
|
|
*
|
|
|
|
* Routines for the character/ioctl interface to the driver. Find out if this
|
|
|
|
* is a valid open. If yes, increment the module use count so that it cannot
|
|
|
|
* be unloaded.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
megadev_open (struct inode *inode, struct file *filep)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Only allow superuser to access private ioctl interface
|
|
|
|
*/
|
|
|
|
if( !capable(CAP_SYS_ADMIN) ) return -EACCES;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* megadev_ioctl()
|
|
|
|
* @inode - Our device inode
|
|
|
|
* @filep - unused
|
|
|
|
* @cmd - ioctl command
|
|
|
|
* @arg - user buffer
|
|
|
|
*
|
|
|
|
* ioctl entry point for our private ioctl interface. We move the data in from
|
|
|
|
* the user space, prepare the command (if necessary, convert the old MIMD
|
|
|
|
* ioctl to new ioctl command), and issue a synchronous command to the
|
|
|
|
* controller.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
megadev_ioctl(struct inode *inode, struct file *filep, unsigned int cmd,
|
|
|
|
unsigned long arg)
|
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
nitioctl_t uioc;
|
|
|
|
int adapno;
|
|
|
|
int rval;
|
|
|
|
mega_passthru __user *upthru; /* user address for passthru */
|
|
|
|
mega_passthru *pthru; /* copy user passthru here */
|
|
|
|
dma_addr_t pthru_dma_hndl;
|
|
|
|
void *data = NULL; /* data to be transferred */
|
|
|
|
dma_addr_t data_dma_hndl; /* dma handle for data xfer area */
|
|
|
|
megacmd_t mc;
|
|
|
|
megastat_t __user *ustats;
|
|
|
|
int num_ldrv;
|
|
|
|
u32 uxferaddr = 0;
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
|
|
|
|
ustats = NULL; /* avoid compilation warnings */
|
|
|
|
num_ldrv = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure only USCSICMD are issued through this interface.
|
|
|
|
* MIMD application would still fire different command.
|
|
|
|
*/
|
|
|
|
if( (_IOC_TYPE(cmd) != MEGAIOC_MAGIC) && (cmd != USCSICMD) ) {
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check and convert a possible MIMD command to NIT command.
|
|
|
|
* mega_m_to_n() copies the data from the user space, so we do not
|
|
|
|
* have to do it here.
|
|
|
|
* NOTE: We will need some user address to copyout the data, therefore
|
|
|
|
* the inteface layer will also provide us with the required user
|
|
|
|
* addresses.
|
|
|
|
*/
|
|
|
|
memset(&uioc, 0, sizeof(nitioctl_t));
|
|
|
|
if( (rval = mega_m_to_n( (void __user *)arg, &uioc)) != 0 )
|
|
|
|
return rval;
|
|
|
|
|
|
|
|
|
|
|
|
switch( uioc.opcode ) {
|
|
|
|
|
|
|
|
case GET_DRIVER_VER:
|
|
|
|
if( put_user(driver_ver, (u32 __user *)uioc.uioc_uaddr) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
case GET_N_ADAP:
|
|
|
|
if( put_user(hba_count, (u32 __user *)uioc.uioc_uaddr) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Shucks. MIMD interface returns a positive value for number
|
|
|
|
* of adapters. TODO: Change it to return 0 when there is no
|
|
|
|
* applicatio using mimd interface.
|
|
|
|
*/
|
|
|
|
return hba_count;
|
|
|
|
|
|
|
|
case GET_ADAP_INFO:
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Which adapter
|
|
|
|
*/
|
|
|
|
if( (adapno = GETADAP(uioc.adapno)) >= hba_count )
|
|
|
|
return (-ENODEV);
|
|
|
|
|
|
|
|
if( copy_to_user(uioc.uioc_uaddr, mcontroller+adapno,
|
|
|
|
sizeof(struct mcontroller)) )
|
|
|
|
return (-EFAULT);
|
|
|
|
break;
|
|
|
|
|
|
|
|
#if MEGA_HAVE_STATS
|
|
|
|
|
|
|
|
case GET_STATS:
|
|
|
|
/*
|
|
|
|
* Which adapter
|
|
|
|
*/
|
|
|
|
if( (adapno = GETADAP(uioc.adapno)) >= hba_count )
|
|
|
|
return (-ENODEV);
|
|
|
|
|
|
|
|
adapter = hba_soft_state[adapno];
|
|
|
|
|
|
|
|
ustats = uioc.uioc_uaddr;
|
|
|
|
|
|
|
|
if( copy_from_user(&num_ldrv, &ustats->num_ldrv, sizeof(int)) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for the validity of the logical drive number
|
|
|
|
*/
|
|
|
|
if( num_ldrv >= MAX_LOGICAL_DRIVES_40LD ) return -EINVAL;
|
|
|
|
|
|
|
|
if( copy_to_user(ustats->nreads, adapter->nreads,
|
|
|
|
num_ldrv*sizeof(u32)) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( copy_to_user(ustats->nreadblocks, adapter->nreadblocks,
|
|
|
|
num_ldrv*sizeof(u32)) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( copy_to_user(ustats->nwrites, adapter->nwrites,
|
|
|
|
num_ldrv*sizeof(u32)) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( copy_to_user(ustats->nwriteblocks, adapter->nwriteblocks,
|
|
|
|
num_ldrv*sizeof(u32)) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( copy_to_user(ustats->rd_errors, adapter->rd_errors,
|
|
|
|
num_ldrv*sizeof(u32)) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( copy_to_user(ustats->wr_errors, adapter->wr_errors,
|
|
|
|
num_ldrv*sizeof(u32)) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
#endif
|
|
|
|
case MBOX_CMD:
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Which adapter
|
|
|
|
*/
|
|
|
|
if( (adapno = GETADAP(uioc.adapno)) >= hba_count )
|
|
|
|
return (-ENODEV);
|
|
|
|
|
|
|
|
adapter = hba_soft_state[adapno];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Deletion of logical drive is a special case. The adapter
|
|
|
|
* should be quiescent before this command is issued.
|
|
|
|
*/
|
|
|
|
if( uioc.uioc_rmbox[0] == FC_DEL_LOGDRV &&
|
|
|
|
uioc.uioc_rmbox[2] == OP_DEL_LOGDRV ) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do we support this feature
|
|
|
|
*/
|
|
|
|
if( !adapter->support_random_del ) {
|
|
|
|
printk(KERN_WARNING "megaraid: logdrv ");
|
|
|
|
printk("delete on non-supporting F/W.\n");
|
|
|
|
|
|
|
|
return (-EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
rval = mega_del_logdrv( adapter, uioc.uioc_rmbox[3] );
|
|
|
|
|
|
|
|
if( rval == 0 ) {
|
|
|
|
memset(&mc, 0, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
mc.status = rval;
|
|
|
|
|
|
|
|
rval = mega_n_to_m((void __user *)arg, &mc);
|
|
|
|
}
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* This interface only support the regular passthru commands.
|
|
|
|
* Reject extended passthru and 64-bit passthru
|
|
|
|
*/
|
|
|
|
if( uioc.uioc_rmbox[0] == MEGA_MBOXCMD_PASSTHRU64 ||
|
|
|
|
uioc.uioc_rmbox[0] == MEGA_MBOXCMD_EXTPTHRU ) {
|
|
|
|
|
|
|
|
printk(KERN_WARNING "megaraid: rejected passthru.\n");
|
|
|
|
|
|
|
|
return (-EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For all internal commands, the buffer must be allocated in
|
|
|
|
* <4GB address range
|
|
|
|
*/
|
|
|
|
if( make_local_pdev(adapter, &pdev) != 0 )
|
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
/* Is it a passthru command or a DCMD */
|
|
|
|
if( uioc.uioc_rmbox[0] == MEGA_MBOXCMD_PASSTHRU ) {
|
|
|
|
/* Passthru commands */
|
|
|
|
|
|
|
|
pthru = pci_alloc_consistent(pdev,
|
|
|
|
sizeof(mega_passthru),
|
|
|
|
&pthru_dma_hndl);
|
|
|
|
|
|
|
|
if( pthru == NULL ) {
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
return (-ENOMEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The user passthru structure
|
|
|
|
*/
|
|
|
|
upthru = (mega_passthru __user *)MBOX(uioc)->xferaddr;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy in the user passthru here.
|
|
|
|
*/
|
|
|
|
if( copy_from_user(pthru, upthru,
|
|
|
|
sizeof(mega_passthru)) ) {
|
|
|
|
|
|
|
|
pci_free_consistent(pdev,
|
|
|
|
sizeof(mega_passthru), pthru,
|
|
|
|
pthru_dma_hndl);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return (-EFAULT);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is there a data transfer
|
|
|
|
*/
|
|
|
|
if( pthru->dataxferlen ) {
|
|
|
|
data = pci_alloc_consistent(pdev,
|
|
|
|
pthru->dataxferlen,
|
|
|
|
&data_dma_hndl);
|
|
|
|
|
|
|
|
if( data == NULL ) {
|
|
|
|
pci_free_consistent(pdev,
|
|
|
|
sizeof(mega_passthru),
|
|
|
|
pthru,
|
|
|
|
pthru_dma_hndl);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return (-ENOMEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the user address and point the kernel
|
|
|
|
* address at just allocated memory
|
|
|
|
*/
|
|
|
|
uxferaddr = pthru->dataxferaddr;
|
|
|
|
pthru->dataxferaddr = data_dma_hndl;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is data coming down-stream
|
|
|
|
*/
|
|
|
|
if( pthru->dataxferlen && (uioc.flags & UIOC_WR) ) {
|
|
|
|
/*
|
|
|
|
* Get the user data
|
|
|
|
*/
|
|
|
|
if( copy_from_user(data, (char __user *)uxferaddr,
|
|
|
|
pthru->dataxferlen) ) {
|
|
|
|
rval = (-EFAULT);
|
|
|
|
goto freemem_and_return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(&mc, 0, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
mc.cmd = MEGA_MBOXCMD_PASSTHRU;
|
|
|
|
mc.xferaddr = (u32)pthru_dma_hndl;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Issue the command
|
|
|
|
*/
|
2005-11-01 03:12:07 +08:00
|
|
|
mega_internal_command(adapter, &mc, pthru);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
rval = mega_n_to_m((void __user *)arg, &mc);
|
|
|
|
|
|
|
|
if( rval ) goto freemem_and_return;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is data going up-stream
|
|
|
|
*/
|
|
|
|
if( pthru->dataxferlen && (uioc.flags & UIOC_RD) ) {
|
|
|
|
if( copy_to_user((char __user *)uxferaddr, data,
|
|
|
|
pthru->dataxferlen) ) {
|
|
|
|
rval = (-EFAULT);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Send the request sense data also, irrespective of
|
|
|
|
* whether the user has asked for it or not.
|
|
|
|
*/
|
2006-09-15 20:43:11 +08:00
|
|
|
if (copy_to_user(upthru->reqsensearea,
|
|
|
|
pthru->reqsensearea, 14))
|
|
|
|
rval = -EFAULT;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
freemem_and_return:
|
|
|
|
if( pthru->dataxferlen ) {
|
|
|
|
pci_free_consistent(pdev,
|
|
|
|
pthru->dataxferlen, data,
|
|
|
|
data_dma_hndl);
|
|
|
|
}
|
|
|
|
|
|
|
|
pci_free_consistent(pdev, sizeof(mega_passthru),
|
|
|
|
pthru, pthru_dma_hndl);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* DCMD commands */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is there a data transfer
|
|
|
|
*/
|
|
|
|
if( uioc.xferlen ) {
|
|
|
|
data = pci_alloc_consistent(pdev,
|
|
|
|
uioc.xferlen, &data_dma_hndl);
|
|
|
|
|
|
|
|
if( data == NULL ) {
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
return (-ENOMEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
uxferaddr = MBOX(uioc)->xferaddr;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is data coming down-stream
|
|
|
|
*/
|
|
|
|
if( uioc.xferlen && (uioc.flags & UIOC_WR) ) {
|
|
|
|
/*
|
|
|
|
* Get the user data
|
|
|
|
*/
|
|
|
|
if( copy_from_user(data, (char __user *)uxferaddr,
|
|
|
|
uioc.xferlen) ) {
|
|
|
|
|
|
|
|
pci_free_consistent(pdev,
|
|
|
|
uioc.xferlen,
|
|
|
|
data, data_dma_hndl);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return (-EFAULT);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(&mc, MBOX(uioc), sizeof(megacmd_t));
|
|
|
|
|
|
|
|
mc.xferaddr = (u32)data_dma_hndl;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Issue the command
|
|
|
|
*/
|
2005-11-01 03:12:07 +08:00
|
|
|
mega_internal_command(adapter, &mc, NULL);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
rval = mega_n_to_m((void __user *)arg, &mc);
|
|
|
|
|
|
|
|
if( rval ) {
|
|
|
|
if( uioc.xferlen ) {
|
|
|
|
pci_free_consistent(pdev,
|
|
|
|
uioc.xferlen, data,
|
|
|
|
data_dma_hndl);
|
|
|
|
}
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is data going up-stream
|
|
|
|
*/
|
|
|
|
if( uioc.xferlen && (uioc.flags & UIOC_RD) ) {
|
|
|
|
if( copy_to_user((char __user *)uxferaddr, data,
|
|
|
|
uioc.xferlen) ) {
|
|
|
|
|
|
|
|
rval = (-EFAULT);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if( uioc.xferlen ) {
|
|
|
|
pci_free_consistent(pdev,
|
|
|
|
uioc.xferlen, data,
|
|
|
|
data_dma_hndl);
|
|
|
|
}
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
default:
|
|
|
|
return (-EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_m_to_n()
|
|
|
|
* @arg - user address
|
|
|
|
* @uioc - new ioctl structure
|
|
|
|
*
|
|
|
|
* A thin layer to convert older mimd interface ioctl structure to NIT ioctl
|
|
|
|
* structure
|
|
|
|
*
|
|
|
|
* Converts the older mimd ioctl structure to newer NIT structure
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_m_to_n(void __user *arg, nitioctl_t *uioc)
|
|
|
|
{
|
|
|
|
struct uioctl_t uioc_mimd;
|
|
|
|
char signature[8] = {0};
|
|
|
|
u8 opcode;
|
|
|
|
u8 subopcode;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* check is the application conforms to NIT. We do not have to do much
|
|
|
|
* in that case.
|
|
|
|
* We exploit the fact that the signature is stored in the very
|
|
|
|
* begining of the structure.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if( copy_from_user(signature, arg, 7) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
if( memcmp(signature, "MEGANIT", 7) == 0 ) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* NOTE NOTE: The nit ioctl is still under flux because of
|
|
|
|
* change of mailbox definition, in HPE. No applications yet
|
|
|
|
* use this interface and let's not have applications use this
|
|
|
|
* interface till the new specifitions are in place.
|
|
|
|
*/
|
|
|
|
return -EINVAL;
|
|
|
|
#if 0
|
|
|
|
if( copy_from_user(uioc, arg, sizeof(nitioctl_t)) )
|
|
|
|
return (-EFAULT);
|
|
|
|
return 0;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Else assume we have mimd uioctl_t as arg. Convert to nitioctl_t
|
|
|
|
*
|
|
|
|
* Get the user ioctl structure
|
|
|
|
*/
|
|
|
|
if( copy_from_user(&uioc_mimd, arg, sizeof(struct uioctl_t)) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get the opcode and subopcode for the commands
|
|
|
|
*/
|
|
|
|
opcode = uioc_mimd.ui.fcs.opcode;
|
|
|
|
subopcode = uioc_mimd.ui.fcs.subopcode;
|
|
|
|
|
|
|
|
switch (opcode) {
|
|
|
|
case 0x82:
|
|
|
|
|
|
|
|
switch (subopcode) {
|
|
|
|
|
|
|
|
case MEGAIOC_QDRVRVER: /* Query driver version */
|
|
|
|
uioc->opcode = GET_DRIVER_VER;
|
|
|
|
uioc->uioc_uaddr = uioc_mimd.data;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case MEGAIOC_QNADAP: /* Get # of adapters */
|
|
|
|
uioc->opcode = GET_N_ADAP;
|
|
|
|
uioc->uioc_uaddr = uioc_mimd.data;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case MEGAIOC_QADAPINFO: /* Get adapter information */
|
|
|
|
uioc->opcode = GET_ADAP_INFO;
|
|
|
|
uioc->adapno = uioc_mimd.ui.fcs.adapno;
|
|
|
|
uioc->uioc_uaddr = uioc_mimd.data;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
return(-EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
|
|
case 0x81:
|
|
|
|
|
|
|
|
uioc->opcode = MBOX_CMD;
|
|
|
|
uioc->adapno = uioc_mimd.ui.fcs.adapno;
|
|
|
|
|
|
|
|
memcpy(uioc->uioc_rmbox, uioc_mimd.mbox, 18);
|
|
|
|
|
|
|
|
uioc->xferlen = uioc_mimd.ui.fcs.length;
|
|
|
|
|
|
|
|
if( uioc_mimd.outlen ) uioc->flags = UIOC_RD;
|
|
|
|
if( uioc_mimd.inlen ) uioc->flags |= UIOC_WR;
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x80:
|
|
|
|
|
|
|
|
uioc->opcode = MBOX_CMD;
|
|
|
|
uioc->adapno = uioc_mimd.ui.fcs.adapno;
|
|
|
|
|
|
|
|
memcpy(uioc->uioc_rmbox, uioc_mimd.mbox, 18);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Choose the xferlen bigger of input and output data
|
|
|
|
*/
|
|
|
|
uioc->xferlen = uioc_mimd.outlen > uioc_mimd.inlen ?
|
|
|
|
uioc_mimd.outlen : uioc_mimd.inlen;
|
|
|
|
|
|
|
|
if( uioc_mimd.outlen ) uioc->flags = UIOC_RD;
|
|
|
|
if( uioc_mimd.inlen ) uioc->flags |= UIOC_WR;
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
return (-EINVAL);
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mega_n_to_m()
|
|
|
|
* @arg - user address
|
|
|
|
* @mc - mailbox command
|
|
|
|
*
|
|
|
|
* Updates the status information to the application, depending on application
|
|
|
|
* conforms to older mimd ioctl interface or newer NIT ioctl interface
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_n_to_m(void __user *arg, megacmd_t *mc)
|
|
|
|
{
|
|
|
|
nitioctl_t __user *uiocp;
|
|
|
|
megacmd_t __user *umc;
|
|
|
|
mega_passthru __user *upthru;
|
|
|
|
struct uioctl_t __user *uioc_mimd;
|
|
|
|
char signature[8] = {0};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* check is the application conforms to NIT.
|
|
|
|
*/
|
|
|
|
if( copy_from_user(signature, arg, 7) )
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( memcmp(signature, "MEGANIT", 7) == 0 ) {
|
|
|
|
|
|
|
|
uiocp = arg;
|
|
|
|
|
|
|
|
if( put_user(mc->status, (u8 __user *)&MBOX_P(uiocp)->status) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
if( mc->cmd == MEGA_MBOXCMD_PASSTHRU ) {
|
|
|
|
|
|
|
|
umc = MBOX_P(uiocp);
|
|
|
|
|
|
|
|
if (get_user(upthru, (mega_passthru __user * __user *)&umc->xferaddr))
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
if( put_user(mc->status, (u8 __user *)&upthru->scsistatus))
|
|
|
|
return (-EFAULT);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
uioc_mimd = arg;
|
|
|
|
|
|
|
|
if( put_user(mc->status, (u8 __user *)&uioc_mimd->mbox[17]) )
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
if( mc->cmd == MEGA_MBOXCMD_PASSTHRU ) {
|
|
|
|
|
|
|
|
umc = (megacmd_t __user *)uioc_mimd->mbox;
|
|
|
|
|
|
|
|
if (get_user(upthru, (mega_passthru __user * __user *)&umc->xferaddr))
|
|
|
|
return (-EFAULT);
|
|
|
|
|
|
|
|
if( put_user(mc->status, (u8 __user *)&upthru->scsistatus) )
|
|
|
|
return (-EFAULT);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* MEGARAID 'FW' commands.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_is_bios_enabled()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* issue command to find out if the BIOS is enabled for this controller
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_is_bios_enabled(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
memset((void *)adapter->mega_buffer, 0, MEGA_BUFFER_SIZE);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = (u32)adapter->buf_dma_handle;
|
|
|
|
|
|
|
|
raw_mbox[0] = IS_BIOS_ENABLED;
|
|
|
|
raw_mbox[2] = GET_BIOS;
|
|
|
|
|
|
|
|
|
|
|
|
ret = issue_scb_block(adapter, raw_mbox);
|
|
|
|
|
|
|
|
return *(char *)adapter->mega_buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_enum_raid_scsi()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Find out what channels are RAID/SCSI. This information is used to
|
|
|
|
* differentiate the virtual channels and physical channels and to support
|
|
|
|
* ROMB feature and non-disk devices.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_enum_raid_scsi(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* issue command to find out what channels are raid/scsi
|
|
|
|
*/
|
|
|
|
raw_mbox[0] = CHNL_CLASS;
|
|
|
|
raw_mbox[2] = GET_CHNL_CLASS;
|
|
|
|
|
|
|
|
memset((void *)adapter->mega_buffer, 0, MEGA_BUFFER_SIZE);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = (u32)adapter->buf_dma_handle;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Non-ROMB firmware fail this command, so all channels
|
|
|
|
* must be shown RAID
|
|
|
|
*/
|
|
|
|
adapter->mega_ch_class = 0xFF;
|
|
|
|
|
|
|
|
if(!issue_scb_block(adapter, raw_mbox)) {
|
|
|
|
adapter->mega_ch_class = *((char *)adapter->mega_buffer);
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
for( i = 0; i < adapter->product_info.nchannels; i++ ) {
|
|
|
|
if( (adapter->mega_ch_class >> i) & 0x01 ) {
|
|
|
|
printk(KERN_INFO "megaraid: channel[%d] is raid.\n",
|
|
|
|
i);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
printk(KERN_INFO "megaraid: channel[%d] is scsi.\n",
|
|
|
|
i);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_get_boot_drv()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Find out which device is the boot device. Note, any logical drive or any
|
|
|
|
* phyical device (e.g., a CDROM) can be designated as a boot device.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_get_boot_drv(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
struct private_bios_data *prv_bios_data;
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
u16 cksum = 0;
|
|
|
|
u8 *cksum_p;
|
|
|
|
u8 boot_pdrv;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
raw_mbox[0] = BIOS_PVT_DATA;
|
|
|
|
raw_mbox[2] = GET_BIOS_PVT_DATA;
|
|
|
|
|
|
|
|
memset((void *)adapter->mega_buffer, 0, MEGA_BUFFER_SIZE);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = (u32)adapter->buf_dma_handle;
|
|
|
|
|
|
|
|
adapter->boot_ldrv_enabled = 0;
|
|
|
|
adapter->boot_ldrv = 0;
|
|
|
|
|
|
|
|
adapter->boot_pdrv_enabled = 0;
|
|
|
|
adapter->boot_pdrv_ch = 0;
|
|
|
|
adapter->boot_pdrv_tgt = 0;
|
|
|
|
|
|
|
|
if(issue_scb_block(adapter, raw_mbox) == 0) {
|
|
|
|
prv_bios_data =
|
|
|
|
(struct private_bios_data *)adapter->mega_buffer;
|
|
|
|
|
|
|
|
cksum = 0;
|
|
|
|
cksum_p = (char *)prv_bios_data;
|
|
|
|
for (i = 0; i < 14; i++ ) {
|
|
|
|
cksum += (u16)(*cksum_p++);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (prv_bios_data->cksum == (u16)(0-cksum) ) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If MSB is set, a physical drive is set as boot
|
|
|
|
* device
|
|
|
|
*/
|
|
|
|
if( prv_bios_data->boot_drv & 0x80 ) {
|
|
|
|
adapter->boot_pdrv_enabled = 1;
|
|
|
|
boot_pdrv = prv_bios_data->boot_drv & 0x7F;
|
|
|
|
adapter->boot_pdrv_ch = boot_pdrv / 16;
|
|
|
|
adapter->boot_pdrv_tgt = boot_pdrv % 16;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
adapter->boot_ldrv_enabled = 1;
|
|
|
|
adapter->boot_ldrv = prv_bios_data->boot_drv;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_support_random_del()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Find out if this controller supports random deletion and addition of
|
|
|
|
* logical drives
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_support_random_del(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* issue command
|
|
|
|
*/
|
|
|
|
raw_mbox[0] = FC_DEL_LOGDRV;
|
|
|
|
raw_mbox[2] = OP_SUP_DEL_LOGDRV;
|
|
|
|
|
|
|
|
rval = issue_scb_block(adapter, raw_mbox);
|
|
|
|
|
|
|
|
return !rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_support_ext_cdb()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Find out if this firmware support cdblen > 10
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_support_ext_cdb(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
/*
|
|
|
|
* issue command to find out if controller supports extended CDBs.
|
|
|
|
*/
|
|
|
|
raw_mbox[0] = 0xA4;
|
|
|
|
raw_mbox[2] = 0x16;
|
|
|
|
|
|
|
|
rval = issue_scb_block(adapter, raw_mbox);
|
|
|
|
|
|
|
|
return !rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_del_logdrv()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @logdrv - logical drive to be deleted
|
|
|
|
*
|
|
|
|
* Delete the specified logical drive. It is the responsibility of the user
|
|
|
|
* app to let the OS know about this operation.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_del_logdrv(adapter_t *adapter, int logdrv)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
scb_t *scb;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Stop sending commands to the controller, queue them internally.
|
|
|
|
* When deletion is complete, ISR will flush the queue.
|
|
|
|
*/
|
|
|
|
atomic_set(&adapter->quiescent, 1);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Wait till all the issued commands are complete and there are no
|
|
|
|
* commands in the pending queue
|
|
|
|
*/
|
|
|
|
while (atomic_read(&adapter->pend_cmds) > 0 ||
|
|
|
|
!list_empty(&adapter->pending_list))
|
|
|
|
msleep(1000); /* sleep for 1s */
|
|
|
|
|
|
|
|
rval = mega_do_del_logdrv(adapter, logdrv);
|
|
|
|
|
|
|
|
spin_lock_irqsave(&adapter->lock, flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If delete operation was successful, add 0x80 to the logical drive
|
|
|
|
* ids for commands in the pending queue.
|
|
|
|
*/
|
|
|
|
if (adapter->read_ldidmap) {
|
|
|
|
struct list_head *pos;
|
|
|
|
list_for_each(pos, &adapter->pending_list) {
|
|
|
|
scb = list_entry(pos, scb_t, list);
|
|
|
|
if (scb->pthru->logdrv < 0x80 )
|
|
|
|
scb->pthru->logdrv += 0x80;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
atomic_set(&adapter->quiescent, 0);
|
|
|
|
|
|
|
|
mega_runpendq(adapter);
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&adapter->lock, flags);
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
mega_do_del_logdrv(adapter_t *adapter, int logdrv)
|
|
|
|
{
|
|
|
|
megacmd_t mc;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
memset( &mc, 0, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
mc.cmd = FC_DEL_LOGDRV;
|
|
|
|
mc.opcode = OP_DEL_LOGDRV;
|
|
|
|
mc.subopcode = logdrv;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
rval = mega_internal_command(adapter, &mc, NULL);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* log this event */
|
|
|
|
if(rval) {
|
|
|
|
printk(KERN_WARNING "megaraid: Delete LD-%d failed.", logdrv);
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* After deleting first logical drive, the logical drives must be
|
|
|
|
* addressed by adding 0x80 to the logical drive id.
|
|
|
|
*/
|
|
|
|
adapter->read_ldidmap = 1;
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_get_max_sgl()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Find out the maximum number of scatter-gather elements supported by this
|
|
|
|
* version of the firmware
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_get_max_sgl(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(mbox, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
memset((void *)adapter->mega_buffer, 0, MEGA_BUFFER_SIZE);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = (u32)adapter->buf_dma_handle;
|
|
|
|
|
|
|
|
raw_mbox[0] = MAIN_MISC_OPCODE;
|
|
|
|
raw_mbox[2] = GET_MAX_SG_SUPPORT;
|
|
|
|
|
|
|
|
|
|
|
|
if( issue_scb_block(adapter, raw_mbox) ) {
|
|
|
|
/*
|
|
|
|
* f/w does not support this command. Choose the default value
|
|
|
|
*/
|
|
|
|
adapter->sglen = MIN_SGLIST;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
adapter->sglen = *((char *)adapter->mega_buffer);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure this is not more than the resources we are
|
|
|
|
* planning to allocate
|
|
|
|
*/
|
|
|
|
if ( adapter->sglen > MAX_SGLIST )
|
|
|
|
adapter->sglen = MAX_SGLIST;
|
|
|
|
}
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_support_cluster()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
*
|
|
|
|
* Find out if this firmware support cluster calls.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_support_cluster(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
unsigned char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox;
|
|
|
|
|
|
|
|
mbox = (mbox_t *)raw_mbox;
|
|
|
|
|
|
|
|
memset(mbox, 0, sizeof(raw_mbox));
|
|
|
|
|
|
|
|
memset((void *)adapter->mega_buffer, 0, MEGA_BUFFER_SIZE);
|
|
|
|
|
|
|
|
mbox->m_out.xferaddr = (u32)adapter->buf_dma_handle;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to get the initiator id. This command will succeed iff the
|
|
|
|
* clustering is available on this HBA.
|
|
|
|
*/
|
|
|
|
raw_mbox[0] = MEGA_GET_TARGET_ID;
|
|
|
|
|
|
|
|
if( issue_scb_block(adapter, raw_mbox) == 0 ) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Cluster support available. Get the initiator target id.
|
|
|
|
* Tell our id to mid-layer too.
|
|
|
|
*/
|
|
|
|
adapter->this_id = *(u32 *)adapter->mega_buffer;
|
|
|
|
adapter->host->this_id = adapter->this_id;
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_adapinq()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @dma_handle - DMA address of the buffer
|
|
|
|
*
|
|
|
|
* Issue internal comamnds while interrupts are available.
|
|
|
|
* We only issue direct mailbox commands from within the driver. ioctl()
|
|
|
|
* interface using these routines can issue passthru commands.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_adapinq(adapter_t *adapter, dma_addr_t dma_handle)
|
|
|
|
{
|
|
|
|
megacmd_t mc;
|
|
|
|
|
|
|
|
memset(&mc, 0, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
if( adapter->flag & BOARD_40LD ) {
|
|
|
|
mc.cmd = FC_NEW_CONFIG;
|
|
|
|
mc.opcode = NC_SUBOP_ENQUIRY3;
|
|
|
|
mc.subopcode = ENQ3_GET_SOLICITED_FULL;
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
mc.cmd = MEGA_MBOXCMD_ADPEXTINQ;
|
|
|
|
}
|
|
|
|
|
|
|
|
mc.xferaddr = (u32)dma_handle;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
if ( mega_internal_command(adapter, &mc, NULL) != 0 ) {
|
2005-04-17 06:20:36 +08:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/** mega_internal_dev_inquiry()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @ch - channel for this device
|
|
|
|
* @tgt - ID of this device
|
|
|
|
* @buf_dma_handle - DMA address of the buffer
|
|
|
|
*
|
|
|
|
* Issue the scsi inquiry for the specified device.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
mega_internal_dev_inquiry(adapter_t *adapter, u8 ch, u8 tgt,
|
|
|
|
dma_addr_t buf_dma_handle)
|
|
|
|
{
|
|
|
|
mega_passthru *pthru;
|
|
|
|
dma_addr_t pthru_dma_handle;
|
|
|
|
megacmd_t mc;
|
|
|
|
int rval;
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For all internal commands, the buffer must be allocated in <4GB
|
|
|
|
* address range
|
|
|
|
*/
|
|
|
|
if( make_local_pdev(adapter, &pdev) != 0 ) return -1;
|
|
|
|
|
|
|
|
pthru = pci_alloc_consistent(pdev, sizeof(mega_passthru),
|
|
|
|
&pthru_dma_handle);
|
|
|
|
|
|
|
|
if( pthru == NULL ) {
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
pthru->timeout = 2;
|
|
|
|
pthru->ars = 1;
|
|
|
|
pthru->reqsenselen = 14;
|
|
|
|
pthru->islogical = 0;
|
|
|
|
|
|
|
|
pthru->channel = (adapter->flag & BOARD_40LD) ? 0 : ch;
|
|
|
|
|
|
|
|
pthru->target = (adapter->flag & BOARD_40LD) ? (ch << 4)|tgt : tgt;
|
|
|
|
|
|
|
|
pthru->cdblen = 6;
|
|
|
|
|
|
|
|
pthru->cdb[0] = INQUIRY;
|
|
|
|
pthru->cdb[1] = 0;
|
|
|
|
pthru->cdb[2] = 0;
|
|
|
|
pthru->cdb[3] = 0;
|
|
|
|
pthru->cdb[4] = 255;
|
|
|
|
pthru->cdb[5] = 0;
|
|
|
|
|
|
|
|
|
|
|
|
pthru->dataxferaddr = (u32)buf_dma_handle;
|
|
|
|
pthru->dataxferlen = 256;
|
|
|
|
|
|
|
|
memset(&mc, 0, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
mc.cmd = MEGA_MBOXCMD_PASSTHRU;
|
|
|
|
mc.xferaddr = (u32)pthru_dma_handle;
|
|
|
|
|
2005-11-01 03:12:07 +08:00
|
|
|
rval = mega_internal_command(adapter, &mc, pthru);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
pci_free_consistent(pdev, sizeof(mega_passthru), pthru,
|
|
|
|
pthru_dma_handle);
|
|
|
|
|
|
|
|
free_local_pdev(pdev);
|
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_internal_command()
|
|
|
|
* @adapter - pointer to our soft state
|
|
|
|
* @mc - the mailbox command
|
|
|
|
* @pthru - Passthru structure for DCDB commands
|
|
|
|
*
|
|
|
|
* Issue the internal commands in interrupt mode.
|
|
|
|
* The last argument is the address of the passthru structure if the command
|
|
|
|
* to be fired is a passthru command
|
|
|
|
*
|
|
|
|
* lockscope specifies whether the caller has already acquired the lock. Of
|
|
|
|
* course, the caller must know which lock we are talking about.
|
|
|
|
*
|
|
|
|
* Note: parameter 'pthru' is null for non-passthru commands.
|
|
|
|
*/
|
|
|
|
static int
|
2005-11-01 03:12:07 +08:00
|
|
|
mega_internal_command(adapter_t *adapter, megacmd_t *mc, mega_passthru *pthru)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
Scsi_Cmnd *scmd;
|
|
|
|
struct scsi_device *sdev;
|
|
|
|
scb_t *scb;
|
|
|
|
int rval;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The internal commands share one command id and hence are
|
|
|
|
* serialized. This is so because we want to reserve maximum number of
|
|
|
|
* available command ids for the I/O commands.
|
|
|
|
*/
|
2006-01-11 20:16:10 +08:00
|
|
|
mutex_lock(&adapter->int_mtx);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
scb = &adapter->int_scb;
|
|
|
|
memset(scb, 0, sizeof(scb_t));
|
|
|
|
|
|
|
|
scmd = &adapter->int_scmd;
|
|
|
|
memset(scmd, 0, sizeof(Scsi_Cmnd));
|
|
|
|
|
|
|
|
sdev = kmalloc(sizeof(struct scsi_device), GFP_KERNEL);
|
|
|
|
memset(sdev, 0, sizeof(struct scsi_device));
|
|
|
|
scmd->device = sdev;
|
|
|
|
|
|
|
|
scmd->device->host = adapter->host;
|
2006-06-03 19:21:13 +08:00
|
|
|
scmd->request_buffer = (void *)scb;
|
2005-04-17 06:20:36 +08:00
|
|
|
scmd->cmnd[0] = MEGA_INTERNAL_CMD;
|
|
|
|
|
|
|
|
scb->state |= SCB_ACTIVE;
|
|
|
|
scb->cmd = scmd;
|
|
|
|
|
|
|
|
memcpy(scb->raw_mbox, mc, sizeof(megacmd_t));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Is it a passthru command
|
|
|
|
*/
|
|
|
|
if( mc->cmd == MEGA_MBOXCMD_PASSTHRU ) {
|
|
|
|
|
|
|
|
scb->pthru = pthru;
|
|
|
|
}
|
|
|
|
|
|
|
|
scb->idx = CMDID_INT_CMDS;
|
|
|
|
|
|
|
|
megaraid_queue(scmd, mega_internal_done);
|
|
|
|
|
2005-06-19 19:42:05 +08:00
|
|
|
wait_for_completion(&adapter->int_waitq);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
rval = scmd->result;
|
|
|
|
mc->status = scmd->result;
|
|
|
|
kfree(sdev);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Print a debug message for all failed commands. Applications can use
|
|
|
|
* this information.
|
|
|
|
*/
|
|
|
|
if( scmd->result && trace_level ) {
|
|
|
|
printk("megaraid: cmd [%x, %x, %x] status:[%x]\n",
|
|
|
|
mc->cmd, mc->opcode, mc->subopcode, scmd->result);
|
|
|
|
}
|
|
|
|
|
2006-01-11 20:16:10 +08:00
|
|
|
mutex_unlock(&adapter->int_mtx);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
return rval;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* mega_internal_done()
|
|
|
|
* @scmd - internal scsi command
|
|
|
|
*
|
|
|
|
* Callback routine for internal commands.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mega_internal_done(Scsi_Cmnd *scmd)
|
|
|
|
{
|
|
|
|
adapter_t *adapter;
|
|
|
|
|
|
|
|
adapter = (adapter_t *)scmd->device->host->hostdata;
|
|
|
|
|
2005-06-19 19:42:05 +08:00
|
|
|
complete(&adapter->int_waitq);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static struct scsi_host_template megaraid_template = {
|
|
|
|
.module = THIS_MODULE,
|
|
|
|
.name = "MegaRAID",
|
2005-11-18 02:13:31 +08:00
|
|
|
.proc_name = "megaraid_legacy",
|
2005-04-17 06:20:36 +08:00
|
|
|
.info = megaraid_info,
|
|
|
|
.queuecommand = megaraid_queue,
|
|
|
|
.bios_param = megaraid_biosparam,
|
|
|
|
.max_sectors = MAX_SECTORS_PER_IO,
|
|
|
|
.can_queue = MAX_COMMANDS,
|
|
|
|
.this_id = DEFAULT_INITIATOR_ID,
|
|
|
|
.sg_tablesize = MAX_SGLIST,
|
|
|
|
.cmd_per_lun = DEF_CMD_PER_LUN,
|
|
|
|
.use_clustering = ENABLE_CLUSTERING,
|
|
|
|
.eh_abort_handler = megaraid_abort,
|
|
|
|
.eh_device_reset_handler = megaraid_reset,
|
|
|
|
.eh_bus_reset_handler = megaraid_reset,
|
|
|
|
.eh_host_reset_handler = megaraid_reset,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __devinit
|
|
|
|
megaraid_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
|
|
|
|
{
|
|
|
|
struct Scsi_Host *host;
|
|
|
|
adapter_t *adapter;
|
|
|
|
unsigned long mega_baseport, tbase, flag = 0;
|
|
|
|
u16 subsysid, subsysvid;
|
|
|
|
u8 pci_bus, pci_dev_func;
|
|
|
|
int irq, i, j;
|
|
|
|
int error = -ENODEV;
|
|
|
|
|
|
|
|
if (pci_enable_device(pdev))
|
|
|
|
goto out;
|
|
|
|
pci_set_master(pdev);
|
|
|
|
|
|
|
|
pci_bus = pdev->bus->number;
|
|
|
|
pci_dev_func = pdev->devfn;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The megaraid3 stuff reports the ID of the Intel part which is not
|
|
|
|
* remotely specific to the megaraid
|
|
|
|
*/
|
|
|
|
if (pdev->vendor == PCI_VENDOR_ID_INTEL) {
|
|
|
|
u16 magic;
|
|
|
|
/*
|
|
|
|
* Don't fall over the Compaq management cards using the same
|
|
|
|
* PCI identifier
|
|
|
|
*/
|
|
|
|
if (pdev->subsystem_vendor == PCI_VENDOR_ID_COMPAQ &&
|
|
|
|
pdev->subsystem_device == 0xC000)
|
|
|
|
return -ENODEV;
|
|
|
|
/* Now check the magic signature byte */
|
|
|
|
pci_read_config_word(pdev, PCI_CONF_AMISIG, &magic);
|
|
|
|
if (magic != HBA_SIGNATURE_471 && magic != HBA_SIGNATURE)
|
|
|
|
return -ENODEV;
|
|
|
|
/* Ok it is probably a megaraid */
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For these vendor and device ids, signature offsets are not
|
|
|
|
* valid and 64 bit is implicit
|
|
|
|
*/
|
|
|
|
if (id->driver_data & BOARD_64BIT)
|
|
|
|
flag |= BOARD_64BIT;
|
|
|
|
else {
|
|
|
|
u32 magic64;
|
|
|
|
|
|
|
|
pci_read_config_dword(pdev, PCI_CONF_AMISIG64, &magic64);
|
|
|
|
if (magic64 == HBA_SIGNATURE_64BIT)
|
|
|
|
flag |= BOARD_64BIT;
|
|
|
|
}
|
|
|
|
|
|
|
|
subsysvid = pdev->subsystem_vendor;
|
|
|
|
subsysid = pdev->subsystem_device;
|
|
|
|
|
|
|
|
printk(KERN_NOTICE "megaraid: found 0x%4.04x:0x%4.04x:bus %d:",
|
|
|
|
id->vendor, id->device, pci_bus);
|
|
|
|
|
|
|
|
printk("slot %d:func %d\n",
|
|
|
|
PCI_SLOT(pci_dev_func), PCI_FUNC(pci_dev_func));
|
|
|
|
|
|
|
|
/* Read the base port and IRQ from PCI */
|
|
|
|
mega_baseport = pci_resource_start(pdev, 0);
|
|
|
|
irq = pdev->irq;
|
|
|
|
|
|
|
|
tbase = mega_baseport;
|
|
|
|
if (pci_resource_flags(pdev, 0) & IORESOURCE_MEM) {
|
|
|
|
flag |= BOARD_MEMMAP;
|
|
|
|
|
|
|
|
if (!request_mem_region(mega_baseport, 128, "megaraid")) {
|
|
|
|
printk(KERN_WARNING "megaraid: mem region busy!\n");
|
|
|
|
goto out_disable_device;
|
|
|
|
}
|
|
|
|
|
|
|
|
mega_baseport = (unsigned long)ioremap(mega_baseport, 128);
|
|
|
|
if (!mega_baseport) {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: could not map hba memory\n");
|
|
|
|
goto out_release_region;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
flag |= BOARD_IOMAP;
|
|
|
|
mega_baseport += 0x10;
|
|
|
|
|
|
|
|
if (!request_region(mega_baseport, 16, "megaraid"))
|
|
|
|
goto out_disable_device;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Initialize SCSI Host structure */
|
|
|
|
host = scsi_host_alloc(&megaraid_template, sizeof(adapter_t));
|
|
|
|
if (!host)
|
|
|
|
goto out_iounmap;
|
|
|
|
|
|
|
|
adapter = (adapter_t *)host->hostdata;
|
|
|
|
memset(adapter, 0, sizeof(adapter_t));
|
|
|
|
|
|
|
|
printk(KERN_NOTICE
|
|
|
|
"scsi%d:Found MegaRAID controller at 0x%lx, IRQ:%d\n",
|
|
|
|
host->host_no, mega_baseport, irq);
|
|
|
|
|
|
|
|
adapter->base = mega_baseport;
|
2006-12-04 09:49:23 +08:00
|
|
|
if (flag & BOARD_MEMMAP)
|
|
|
|
adapter->mmio_base = (void __iomem *) mega_baseport;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&adapter->free_list);
|
|
|
|
INIT_LIST_HEAD(&adapter->pending_list);
|
|
|
|
INIT_LIST_HEAD(&adapter->completed_list);
|
|
|
|
|
|
|
|
adapter->flag = flag;
|
|
|
|
spin_lock_init(&adapter->lock);
|
|
|
|
|
|
|
|
host->cmd_per_lun = max_cmd_per_lun;
|
|
|
|
host->max_sectors = max_sectors_per_io;
|
|
|
|
|
|
|
|
adapter->dev = pdev;
|
|
|
|
adapter->host = host;
|
|
|
|
|
|
|
|
adapter->host->irq = irq;
|
|
|
|
|
|
|
|
if (flag & BOARD_MEMMAP)
|
|
|
|
adapter->host->base = tbase;
|
|
|
|
else {
|
|
|
|
adapter->host->io_port = tbase;
|
|
|
|
adapter->host->n_io_port = 16;
|
|
|
|
}
|
|
|
|
|
|
|
|
adapter->host->unique_id = (pci_bus << 8) | pci_dev_func;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate buffer to issue internal commands.
|
|
|
|
*/
|
|
|
|
adapter->mega_buffer = pci_alloc_consistent(adapter->dev,
|
|
|
|
MEGA_BUFFER_SIZE, &adapter->buf_dma_handle);
|
|
|
|
if (!adapter->mega_buffer) {
|
|
|
|
printk(KERN_WARNING "megaraid: out of RAM.\n");
|
|
|
|
goto out_host_put;
|
|
|
|
}
|
|
|
|
|
|
|
|
adapter->scb_list = kmalloc(sizeof(scb_t) * MAX_COMMANDS, GFP_KERNEL);
|
|
|
|
if (!adapter->scb_list) {
|
|
|
|
printk(KERN_WARNING "megaraid: out of RAM.\n");
|
|
|
|
goto out_free_cmd_buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (request_irq(irq, (adapter->flag & BOARD_MEMMAP) ?
|
|
|
|
megaraid_isr_memmapped : megaraid_isr_iomapped,
|
2006-07-02 10:29:42 +08:00
|
|
|
IRQF_SHARED, "megaraid", adapter)) {
|
2005-04-17 06:20:36 +08:00
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: Couldn't register IRQ %d!\n", irq);
|
|
|
|
goto out_free_scb_list;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mega_setup_mailbox(adapter))
|
|
|
|
goto out_free_irq;
|
|
|
|
|
|
|
|
if (mega_query_adapter(adapter))
|
|
|
|
goto out_free_mbox;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Have checks for some buggy f/w
|
|
|
|
*/
|
|
|
|
if ((subsysid == 0x1111) && (subsysvid == 0x1111)) {
|
|
|
|
/*
|
|
|
|
* Which firmware
|
|
|
|
*/
|
|
|
|
if (!strcmp(adapter->fw_version, "3.00") ||
|
|
|
|
!strcmp(adapter->fw_version, "3.01")) {
|
|
|
|
|
|
|
|
printk( KERN_WARNING
|
|
|
|
"megaraid: Your card is a Dell PERC "
|
|
|
|
"2/SC RAID controller with "
|
|
|
|
"firmware\nmegaraid: 3.00 or 3.01. "
|
|
|
|
"This driver is known to have "
|
|
|
|
"corruption issues\nmegaraid: with "
|
|
|
|
"those firmware versions on this "
|
|
|
|
"specific card. In order\nmegaraid: "
|
|
|
|
"to protect your data, please upgrade "
|
|
|
|
"your firmware to version\nmegaraid: "
|
|
|
|
"3.10 or later, available from the "
|
|
|
|
"Dell Technical Support web\n"
|
|
|
|
"megaraid: site at\nhttp://support."
|
|
|
|
"dell.com/us/en/filelib/download/"
|
|
|
|
"index.asp?fileid=2940\n"
|
|
|
|
);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we have a HP 1M(0x60E7)/2M(0x60E8) controller with
|
|
|
|
* firmware H.01.07, H.01.08, and H.01.09 disable 64 bit
|
|
|
|
* support, since this firmware cannot handle 64 bit
|
|
|
|
* addressing
|
|
|
|
*/
|
|
|
|
if ((subsysvid == HP_SUBSYS_VID) &&
|
|
|
|
((subsysid == 0x60E7) || (subsysid == 0x60E8))) {
|
|
|
|
/*
|
|
|
|
* which firmware
|
|
|
|
*/
|
|
|
|
if (!strcmp(adapter->fw_version, "H01.07") ||
|
|
|
|
!strcmp(adapter->fw_version, "H01.08") ||
|
|
|
|
!strcmp(adapter->fw_version, "H01.09") ) {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: Firmware H.01.07, "
|
|
|
|
"H.01.08, and H.01.09 on 1M/2M "
|
|
|
|
"controllers\n"
|
|
|
|
"megaraid: do not support 64 bit "
|
|
|
|
"addressing.\nmegaraid: DISABLING "
|
|
|
|
"64 bit support.\n");
|
|
|
|
adapter->flag &= ~BOARD_64BIT;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mega_is_bios_enabled(adapter))
|
|
|
|
mega_hbas[hba_count].is_bios_enabled = 1;
|
|
|
|
mega_hbas[hba_count].hostdata_addr = adapter;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find out which channel is raid and which is scsi. This is
|
|
|
|
* for ROMB support.
|
|
|
|
*/
|
|
|
|
mega_enum_raid_scsi(adapter);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find out if a logical drive is set as the boot drive. If
|
|
|
|
* there is one, will make that as the first logical drive.
|
|
|
|
* ROMB: Do we have to boot from a physical drive. Then all
|
|
|
|
* the physical drives would appear before the logical disks.
|
|
|
|
* Else, all the physical drives would be exported to the mid
|
|
|
|
* layer after logical drives.
|
|
|
|
*/
|
|
|
|
mega_get_boot_drv(adapter);
|
|
|
|
|
|
|
|
if (adapter->boot_pdrv_enabled) {
|
|
|
|
j = adapter->product_info.nchannels;
|
|
|
|
for( i = 0; i < j; i++ )
|
|
|
|
adapter->logdrv_chan[i] = 0;
|
|
|
|
for( i = j; i < NVIRT_CHAN + j; i++ )
|
|
|
|
adapter->logdrv_chan[i] = 1;
|
|
|
|
} else {
|
|
|
|
for (i = 0; i < NVIRT_CHAN; i++)
|
|
|
|
adapter->logdrv_chan[i] = 1;
|
|
|
|
for (i = NVIRT_CHAN; i < MAX_CHANNELS+NVIRT_CHAN; i++)
|
|
|
|
adapter->logdrv_chan[i] = 0;
|
|
|
|
adapter->mega_ch_class <<= NVIRT_CHAN;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do we support random deletion and addition of logical
|
|
|
|
* drives
|
|
|
|
*/
|
|
|
|
adapter->read_ldidmap = 0; /* set it after first logdrv
|
|
|
|
delete cmd */
|
|
|
|
adapter->support_random_del = mega_support_random_del(adapter);
|
|
|
|
|
|
|
|
/* Initialize SCBs */
|
|
|
|
if (mega_init_scb(adapter))
|
|
|
|
goto out_free_mbox;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the pending commands counter
|
|
|
|
*/
|
|
|
|
atomic_set(&adapter->pend_cmds, 0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the adapter quiescent flag
|
|
|
|
*/
|
|
|
|
atomic_set(&adapter->quiescent, 0);
|
|
|
|
|
|
|
|
hba_soft_state[hba_count] = adapter;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Fill in the structure which needs to be passed back to the
|
|
|
|
* application when it does an ioctl() for controller related
|
|
|
|
* information.
|
|
|
|
*/
|
|
|
|
i = hba_count;
|
|
|
|
|
|
|
|
mcontroller[i].base = mega_baseport;
|
|
|
|
mcontroller[i].irq = irq;
|
|
|
|
mcontroller[i].numldrv = adapter->numldrv;
|
|
|
|
mcontroller[i].pcibus = pci_bus;
|
|
|
|
mcontroller[i].pcidev = id->device;
|
|
|
|
mcontroller[i].pcifun = PCI_FUNC (pci_dev_func);
|
|
|
|
mcontroller[i].pciid = -1;
|
|
|
|
mcontroller[i].pcivendor = id->vendor;
|
|
|
|
mcontroller[i].pcislot = PCI_SLOT(pci_dev_func);
|
|
|
|
mcontroller[i].uid = (pci_bus << 8) | pci_dev_func;
|
|
|
|
|
|
|
|
|
|
|
|
/* Set the Mode of addressing to 64 bit if we can */
|
|
|
|
if ((adapter->flag & BOARD_64BIT) && (sizeof(dma_addr_t) == 8)) {
|
2006-03-28 17:56:48 +08:00
|
|
|
pci_set_dma_mask(pdev, DMA_64BIT_MASK);
|
2005-04-17 06:20:36 +08:00
|
|
|
adapter->has_64bit_addr = 1;
|
|
|
|
} else {
|
2006-03-28 17:56:48 +08:00
|
|
|
pci_set_dma_mask(pdev, DMA_32BIT_MASK);
|
2005-04-17 06:20:36 +08:00
|
|
|
adapter->has_64bit_addr = 0;
|
|
|
|
}
|
|
|
|
|
2006-01-11 20:16:10 +08:00
|
|
|
mutex_init(&adapter->int_mtx);
|
2005-06-19 19:42:05 +08:00
|
|
|
init_completion(&adapter->int_waitq);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
adapter->this_id = DEFAULT_INITIATOR_ID;
|
|
|
|
adapter->host->this_id = DEFAULT_INITIATOR_ID;
|
|
|
|
|
|
|
|
#if MEGA_HAVE_CLUSTERING
|
|
|
|
/*
|
|
|
|
* Is cluster support enabled on this controller
|
|
|
|
* Note: In a cluster the HBAs ( the initiators ) will have
|
|
|
|
* different target IDs and we cannot assume it to be 7. Call
|
|
|
|
* to mega_support_cluster() will get the target ids also if
|
|
|
|
* the cluster support is available
|
|
|
|
*/
|
|
|
|
adapter->has_cluster = mega_support_cluster(adapter);
|
|
|
|
if (adapter->has_cluster) {
|
|
|
|
printk(KERN_NOTICE
|
|
|
|
"megaraid: Cluster driver, initiator id:%d\n",
|
|
|
|
adapter->this_id);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
pci_set_drvdata(pdev, host);
|
|
|
|
|
|
|
|
mega_create_proc_entry(hba_count, mega_proc_dir_entry);
|
|
|
|
|
|
|
|
error = scsi_add_host(host, &pdev->dev);
|
|
|
|
if (error)
|
|
|
|
goto out_free_mbox;
|
|
|
|
|
|
|
|
scsi_scan_host(host);
|
|
|
|
hba_count++;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_free_mbox:
|
|
|
|
pci_free_consistent(adapter->dev, sizeof(mbox64_t),
|
|
|
|
adapter->una_mbox64, adapter->una_mbox64_dma);
|
|
|
|
out_free_irq:
|
|
|
|
free_irq(adapter->host->irq, adapter);
|
|
|
|
out_free_scb_list:
|
|
|
|
kfree(adapter->scb_list);
|
|
|
|
out_free_cmd_buffer:
|
|
|
|
pci_free_consistent(adapter->dev, MEGA_BUFFER_SIZE,
|
|
|
|
adapter->mega_buffer, adapter->buf_dma_handle);
|
|
|
|
out_host_put:
|
|
|
|
scsi_host_put(host);
|
|
|
|
out_iounmap:
|
|
|
|
if (flag & BOARD_MEMMAP)
|
|
|
|
iounmap((void *)mega_baseport);
|
|
|
|
out_release_region:
|
|
|
|
if (flag & BOARD_MEMMAP)
|
|
|
|
release_mem_region(tbase, 128);
|
|
|
|
else
|
|
|
|
release_region(mega_baseport, 16);
|
|
|
|
out_disable_device:
|
|
|
|
pci_disable_device(pdev);
|
|
|
|
out:
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
__megaraid_shutdown(adapter_t *adapter)
|
|
|
|
{
|
|
|
|
u_char raw_mbox[sizeof(struct mbox_out)];
|
|
|
|
mbox_t *mbox = (mbox_t *)raw_mbox;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* Flush adapter cache */
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
raw_mbox[0] = FLUSH_ADAPTER;
|
|
|
|
|
|
|
|
free_irq(adapter->host->irq, adapter);
|
|
|
|
|
|
|
|
/* Issue a blocking (interrupts disabled) command to the card */
|
|
|
|
issue_scb_block(adapter, raw_mbox);
|
|
|
|
|
|
|
|
/* Flush disks cache */
|
|
|
|
memset(&mbox->m_out, 0, sizeof(raw_mbox));
|
|
|
|
raw_mbox[0] = FLUSH_SYSTEM;
|
|
|
|
|
|
|
|
/* Issue a blocking (interrupts disabled) command to the card */
|
|
|
|
issue_scb_block(adapter, raw_mbox);
|
|
|
|
|
|
|
|
if (atomic_read(&adapter->pend_cmds) > 0)
|
|
|
|
printk(KERN_WARNING "megaraid: pending commands!!\n");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Have a delibrate delay to make sure all the caches are
|
|
|
|
* actually flushed.
|
|
|
|
*/
|
|
|
|
for (i = 0; i <= 10; i++)
|
|
|
|
mdelay(1000);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
megaraid_remove_one(struct pci_dev *pdev)
|
|
|
|
{
|
|
|
|
struct Scsi_Host *host = pci_get_drvdata(pdev);
|
|
|
|
adapter_t *adapter = (adapter_t *)host->hostdata;
|
|
|
|
char buf[12] = { 0 };
|
|
|
|
|
|
|
|
scsi_remove_host(host);
|
|
|
|
|
|
|
|
__megaraid_shutdown(adapter);
|
|
|
|
|
|
|
|
/* Free our resources */
|
|
|
|
if (adapter->flag & BOARD_MEMMAP) {
|
|
|
|
iounmap((void *)adapter->base);
|
|
|
|
release_mem_region(adapter->host->base, 128);
|
|
|
|
} else
|
|
|
|
release_region(adapter->base, 16);
|
|
|
|
|
|
|
|
mega_free_sgl(adapter);
|
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
if (adapter->controller_proc_dir_entry) {
|
|
|
|
remove_proc_entry("stat", adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("config",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("mailbox",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
#if MEGA_HAVE_ENH_PROC
|
|
|
|
remove_proc_entry("rebuild-rate",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("battery-status",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
|
|
|
|
remove_proc_entry("diskdrives-ch0",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("diskdrives-ch1",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("diskdrives-ch2",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("diskdrives-ch3",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
|
|
|
|
remove_proc_entry("raiddrives-0-9",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("raiddrives-10-19",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("raiddrives-20-29",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
remove_proc_entry("raiddrives-30-39",
|
|
|
|
adapter->controller_proc_dir_entry);
|
|
|
|
#endif
|
|
|
|
sprintf(buf, "hba%d", adapter->host->host_no);
|
|
|
|
remove_proc_entry(buf, mega_proc_dir_entry);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
pci_free_consistent(adapter->dev, MEGA_BUFFER_SIZE,
|
|
|
|
adapter->mega_buffer, adapter->buf_dma_handle);
|
|
|
|
kfree(adapter->scb_list);
|
|
|
|
pci_free_consistent(adapter->dev, sizeof(mbox64_t),
|
|
|
|
adapter->una_mbox64, adapter->una_mbox64_dma);
|
|
|
|
|
|
|
|
scsi_host_put(host);
|
|
|
|
pci_disable_device(pdev);
|
|
|
|
|
|
|
|
hba_count--;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2005-06-24 08:35:56 +08:00
|
|
|
megaraid_shutdown(struct pci_dev *pdev)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2005-06-24 08:35:56 +08:00
|
|
|
struct Scsi_Host *host = pci_get_drvdata(pdev);
|
2005-04-17 06:20:36 +08:00
|
|
|
adapter_t *adapter = (adapter_t *)host->hostdata;
|
|
|
|
|
|
|
|
__megaraid_shutdown(adapter);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct pci_device_id megaraid_pci_tbl[] = {
|
|
|
|
{PCI_VENDOR_ID_AMI, PCI_DEVICE_ID_AMI_MEGARAID,
|
|
|
|
PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
|
|
|
|
{PCI_VENDOR_ID_AMI, PCI_DEVICE_ID_AMI_MEGARAID2,
|
|
|
|
PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
|
|
|
|
{PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_AMI_MEGARAID3,
|
|
|
|
PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
|
|
|
|
{0,}
|
|
|
|
};
|
|
|
|
MODULE_DEVICE_TABLE(pci, megaraid_pci_tbl);
|
|
|
|
|
|
|
|
static struct pci_driver megaraid_pci_driver = {
|
2006-02-10 03:32:59 +08:00
|
|
|
.name = "megaraid_legacy",
|
2005-04-17 06:20:36 +08:00
|
|
|
.id_table = megaraid_pci_tbl,
|
|
|
|
.probe = megaraid_probe_one,
|
|
|
|
.remove = __devexit_p(megaraid_remove_one),
|
2005-06-24 08:35:56 +08:00
|
|
|
.shutdown = megaraid_shutdown,
|
2005-04-17 06:20:36 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
static int __init megaraid_init(void)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if ((max_cmd_per_lun <= 0) || (max_cmd_per_lun > MAX_CMD_PER_LUN))
|
|
|
|
max_cmd_per_lun = MAX_CMD_PER_LUN;
|
|
|
|
if (max_mbox_busy_wait > MBOX_BUSY_WAIT)
|
|
|
|
max_mbox_busy_wait = MBOX_BUSY_WAIT;
|
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
mega_proc_dir_entry = proc_mkdir("megaraid", &proc_root);
|
|
|
|
if (!mega_proc_dir_entry) {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: failed to create megaraid root\n");
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
error = pci_module_init(&megaraid_pci_driver);
|
|
|
|
if (error) {
|
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
remove_proc_entry("megaraid", &proc_root);
|
|
|
|
#endif
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Register the driver as a character device, for applications
|
|
|
|
* to access it for ioctls.
|
|
|
|
* First argument (major) to register_chrdev implies a dynamic
|
|
|
|
* major number allocation.
|
|
|
|
*/
|
2005-11-18 02:13:31 +08:00
|
|
|
major = register_chrdev(0, "megadev_legacy", &megadev_fops);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (!major) {
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"megaraid: failed to register char device\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit megaraid_exit(void)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Unregister the character device interface to the driver.
|
|
|
|
*/
|
2005-11-18 02:13:31 +08:00
|
|
|
unregister_chrdev(major, "megadev_legacy");
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
pci_unregister_driver(&megaraid_pci_driver);
|
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
remove_proc_entry("megaraid", &proc_root);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
module_init(megaraid_init);
|
|
|
|
module_exit(megaraid_exit);
|
|
|
|
|
|
|
|
/* vi: set ts=8 sw=8 tw=78: */
|