Attachment 298533 Details for Bug 230717 – updated kdump patch

[patch] updated kdump patch

p00001_cciss_kdump_rhel5.2_polling.patch (text/plain), 15.88 KB, created by Mike Miller (OS Dev) on 2008-03-19 16:13:14 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Mike Miller (OS Dev)

Created: 2008-03-19 16:13:14 UTC

Size: 15.88 KB

patch

obsolete

>diff -urNp linux-2.6.18.i386-orig/drivers/block/cciss.c linux-2.6.18.i386/drivers/block/cciss.c
>--- linux-2.6.18.i386-orig/drivers/block/cciss.c	2008-03-14 13:50:14.000000000 -0500
>+++ linux-2.6.18.i386/drivers/block/cciss.c	2008-03-19 09:51:58.000000000 -0500
>@@ -190,6 +190,16 @@ static int sendcmd_withirq(__u8 cmd, int
> 
> static void fail_all_cmds(unsigned long ctlr);
> static void cciss_shutdown(struct pci_dev *);
>+static void print_cmd(CommandList_struct *);
>+static int cciss_reset_controller(struct pci_dev *, int, ctlr_info_t *);
>+
>+/* Create a thread to poll for completions if we're coming up
>+ * in a crashkernel.
>+ */
>+static int cciss_completion_thread(void *ctlr);
>+static int start_completion_thread(ctlr_info_t *h,
>+                int (*cciss_completion_thread)(void *), int *rc);
>+  
> 
> #ifdef CONFIG_PROC_FS
> static void cciss_procinit(int i);
>@@ -336,7 +346,7 @@ static int cciss_seq_show(struct seq_fil
> 		drv->raid_level = RAID_UNKNOWN;
> 	seq_printf(seq, "cciss/c%dd%d:"
> 			"\t%4u.%02uGB\tRAID %s\n",
>-			ctlr, *pos, (int)vol_sz, (int)vol_sz_frac,
>+			ctlr, (int) *pos, (int)vol_sz, (int)vol_sz_frac,
> 			raid_label[drv->raid_level]);
> 
> 	return 0;
>@@ -566,8 +576,11 @@ static int cciss_open(struct inode *inod
> 	printk(KERN_DEBUG "cciss_open %s\n", inode->i_bdev->bd_disk->disk_name);
> #endif				/* CCISS_DEBUG */
> 
>-	if (host->busy_initializing || drv->busy_configuring)
>-		return -EBUSY;
>+	if (!reset_devices) {
>+		if (host->busy_initializing || drv->busy_configuring)
>+			return -EBUSY;
>+	}
>+
> 	/*
> 	 * Root is allowed to open raw volume zero even if it's not configured
> 	 * so array config can still work. Root is also allowed to open any
>@@ -1817,7 +1830,7 @@ static int fill_cmd(CommandList_struct *
> 		case 0:	/* ABORT message */
> 			c->Request.CDBLen = 12;
> 			c->Request.Type.Attribute = ATTR_SIMPLE;
>-			c->Request.Type.Direction = XFER_WRITE;
>+			c->Request.Type.Direction = XFER_NONE;
> 			c->Request.Timeout = 0;
> 			c->Request.CDB[0] = cmd;	/* abort */
> 			c->Request.CDB[1] = 0;	/* abort a command */
>@@ -1825,18 +1838,17 @@ static int fill_cmd(CommandList_struct *
> 			memcpy(&c->Request.CDB[4], buff, 8);
> 			break;
> 		case 1:	/* RESET message */
>-			c->Request.CDBLen = 12;
>+			c->Request.CDBLen = 16;
> 			c->Request.Type.Attribute = ATTR_SIMPLE;
>-			c->Request.Type.Direction = XFER_WRITE;
>+			c->Request.Type.Direction = XFER_NONE;
> 			c->Request.Timeout = 0;
> 			memset(&c->Request.CDB[0], 0, sizeof(c->Request.CDB));
> 			c->Request.CDB[0] = cmd;	/* reset */
>-			c->Request.CDB[1] = 0x04;	/* reset a LUN */
> 			break;
> 		case 3:	/* No-Op message */
> 			c->Request.CDBLen = 1;
> 			c->Request.Type.Attribute = ATTR_SIMPLE;
>-			c->Request.Type.Direction = XFER_WRITE;
>+			c->Request.Type.Direction = XFER_NONE;
> 			c->Request.Timeout = 0;
> 			c->Request.CDB[0] = cmd;
> 			break;
>@@ -2146,11 +2158,18 @@ static int cciss_revalidate(struct gendi
> static unsigned long pollcomplete(int ctlr)
> {
> 	unsigned long done;
>-	int i;
>+	int i, timeout;
> 
>-	/* Wait (up to 20 seconds) for a command to complete */
>+	/* Wait (up to 20 seconds) for a command to complete
>+	 * under normal conditions.
>+	 * If reset_devices wait at least 60 seconds. It takes
>+	 * longer than advertised for the controller to become
>+	 * ready after a reset. I'd rather have a more dynamic
>+	 * method to determine if it's ready.
>+	 */
> 
>-	for (i = 20 * HZ; i > 0; i--) {
>+	timeout = reset_devices ? 60 * HZ : 20 * HZ;
>+	for (i = timeout; i > 0; i--) {
> 		done = hba[ctlr]->access.command_completed(hba[ctlr]);
> 		if (done == FIFO_EMPTY)
> 			schedule_timeout_uninterruptible(1);
>@@ -2181,7 +2200,7 @@ static int add_sendcmd_reject(__u8 cmd, 
> 
> 	/* If it's not the scsi tape stuff doing error handling, (abort */
> 	/* or reset) then we don't expect anything weird. */
>-	if (cmd != CCISS_RESET_MSG && cmd != CCISS_ABORT_MSG) {
>+	if (cmd != CCISS_RESET_MSG && cmd != CCISS_ABORT_MSG && !reset_devices) {
> #endif
> 		printk(KERN_WARNING "cciss cciss%d: SendCmd "
> 		       "Invalid command list address returned! (%lx)\n",
>@@ -2207,8 +2226,10 @@ static int add_sendcmd_reject(__u8 cmd, 
> 		return 1;
> 	}
> 	/* Save it for later */
>-	srl->complete[srl->ncompletions] = complete;
>-	srl->ncompletions++;
>+	if (!reset_devices) {
>+		srl->complete[srl->ncompletions] = complete;
>+		srl->ncompletions++;
>+	}
> #endif
> 	return 0;
> }
>@@ -2240,7 +2261,7 @@ static int sendcmd(__u8 cmd, int ctlr, v
> 		cmd_free(info_p, c, 1);
> 		return status;
> 	}
>-      resend_cmd1:
>+resend_cmd1:
> 	/*
> 	 * Disable interrupt
> 	 */
>@@ -2276,6 +2297,9 @@ static int sendcmd(__u8 cmd, int ctlr, v
> #endif				/* CCISS_DEBUG */
> 
> 		if (complete == 1) {
>+#ifdef CCISS_DEBUG
>+			print_cmd(c);
>+#endif
> 			printk(KERN_WARNING
> 			       "cciss cciss%d: SendCmd Timeout out, "
> 			       "No command list address returned!\n", ctlr);
>@@ -2726,13 +2750,17 @@ static irqreturn_t do_cciss_intr(int irq
> 	unsigned long flags;
> 	__u32 a, a1, a2;
> 
>-	if (interrupt_not_for_us(h))
>-		return IRQ_NONE;
>+	if (!reset_devices) {
>+		if (interrupt_not_for_us(h))
>+			return IRQ_NONE;
>+	}
>+	
> 	/*
> 	 * If there are completed commands in the completion queue,
> 	 * we had better do something about it.
> 	 */
> 	spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
>+	
> 	while (interrupt_pending(h)) {
> 		while ((a = get_next_completion(h)) != FIFO_EMPTY) {
> 			a1 = a;
>@@ -2878,7 +2906,6 @@ static void __devinit cciss_interrupt_mo
> 	    (board_id == 0x40800E11) ||
> 	    (board_id == 0x40820E11) || (board_id == 0x40830E11))
> 		goto default_int_mode;
>-
> 	if (pci_find_capability(pdev, PCI_CAP_ID_MSIX)) {
> 		err = pci_enable_msix(pdev, cciss_msix_entries, 4);
> 		if (!err) {
>@@ -2956,9 +2983,11 @@ static int cciss_pci_init(ctlr_info_t *c
> #endif				/* CCISS_DEBUG */
> 
> /* If the kernel supports MSI/MSI-X we will try to enable that functionality,
>- * else we use the IO-APIC interrupt assigned to us by system ROM.
>+ * else we use the IO-APIC interrupt assigned to us by system ROM. If we're
>+ * booting into a crashkernel we use polling mode.
>  */
>-	cciss_interrupt_mode(c, pdev, board_id);
>+	if (!reset_devices)
>+		cciss_interrupt_mode(c, pdev, board_id);
> 
> 	/*
> 	 * Memory base addr is first addr , the second points to the config
>@@ -3223,7 +3252,7 @@ static void cciss_getgeometry(int cntl_n
> 			goto geo_inq;
> 		}
> 		cciss_read_capacity(cntl_num, i, 0, &total_size, &block_size);
>-
>+		
> 		/* If read_capacity returns all F's the volume is >2TB */
> 		/* so we switch to 16-byte CDBs for all read/write ops */
> 		if(total_size == 0xFFFFFFFFULL) {
>@@ -3287,6 +3316,111 @@ static void free_hba(int i)
> }
> 
> /*
>+ * If were coming up in a crash kernel we need to reset
>+ * the controller to clear any outstanding commands and
>+ * get the card in a sane state.
>+ * We use polling mode operation while in the crash kernel
>+ * to avoid interrupt related issues.
>+ */
>+static int cciss_reset_controller(struct pci_dev *pdev, 
>+	int c, ctlr_info_t *hba)
>+{
>+
>+	int rc, i, err;
>+	char reset_buf[4];
>+	CommandList_struct *cmd;
>+	dma_addr_t cmd_dhandle;
>+	ushort subsystem_vendor_id, subsystem_device_id;
>+	__u32 board_id, scratchpad = 0;
>+	u64bit temp64;
>+	
>+	/* Set up the buffer w/reset controller message. */
>+	memset(reset_buf, RESET_CONTROLLER, 4);
>+
>+	/* Set up pci mappings so we can access registers. */
>+	hba->paddr = pci_resource_start(pdev, 0);
>+	hba->vaddr = remap_pci_mem(hba->paddr, 0x250);
>+	hba->pdev = pdev;
>+
>+	/* Find our access methods for this board. */
>+	subsystem_vendor_id = pdev->subsystem_vendor;
>+	subsystem_device_id = pdev->subsystem_device;
>+	board_id = (((__u32) (subsystem_device_id << 16) & 0xffff0000) |
>+		    subsystem_vendor_id);
>+	for (i = 0; i < ARRAY_SIZE(products); i++) {
>+		if (board_id == products[i].board_id) {
>+			hba->product_name = products[i].product_name;
>+			hba->access = *(products[i].access);
>+			hba->nr_cmds = products[i].nr_cmds;
>+			break;
>+		}
>+	}
>+
>+	/* make sure the board interrupts are off */
>+	hba->access.set_intr_mask(hba, CCISS_INTR_OFF);
>+
>+	/* Allocate a cmd & error info struct and set up ptrs. */
>+	cmd = (CommandList_struct *) pci_alloc_consistent(hba->pdev,
>+			sizeof(CommandList_struct) + sizeof(ErrorInfo_struct),
>+			&cmd_dhandle);
>+	memset(cmd, 0, sizeof(*cmd) + sizeof(*cmd->err_info));
>+	cmd->err_info = ((void *) cmd) + sizeof(*cmd);
>+	cmd->busaddr = (__u32) cmd_dhandle;
>+	temp64.val = (__u64) cmd_dhandle + sizeof(CommandList_struct);
>+	cmd->ErrDesc.Addr.lower = temp64.val32.lower;
>+	cmd->ErrDesc.Addr.upper = temp64.val32.upper;
>+	cmd->ErrDesc.Len = sizeof(ErrorInfo_struct);
>+
>+	/* Build the command. */
>+	rc = fill_cmd(cmd, CCISS_RESET_MSG, c, reset_buf, 4, 0,
>+				0, 0, NULL, TYPE_MSG);
>+	if (rc == IO_ERROR)
>+		return -EIO;
>+	else
>+		rc = 0;
>+
>+	/* Send it down. */
>+	hba->access.submit_command(hba, cmd);
>+
>+	/* Wait some time for the scratchpad to be reset. */
>+	do {
>+		mdelay(25);
>+		scratchpad = readl(hba->vaddr + SA5_SCRATCHPAD_OFFSET);
>+	} while (scratchpad == CCISS_FIRMWARE_READY);
>+
>+	/* Now wait for the scratchpad to become ready. */ 
>+	for (i = 0; i < 1200; i++) {
>+		scratchpad = readl(hba->vaddr + SA5_SCRATCHPAD_OFFSET);
>+		if (scratchpad == CCISS_FIRMWARE_READY) {
>+			break;
>+		}
>+		set_current_state(TASK_INTERRUPTIBLE);
>+		schedule_timeout(HZ / 10);	/* wait 100ms */
>+	}
>+	if (scratchpad != CCISS_FIRMWARE_READY) {
>+		printk(KERN_WARNING "cciss: Board not ready.  Timed out.\n");
>+		err = -ENODEV;
>+	}
>+
>+	/* Free the command. */
>+	pci_free_consistent(pdev, sizeof(*cmd) + sizeof(*cmd->err_info),
>+				cmd, cmd_dhandle);
>+
>+	return rc;
>+}
>+
>+static void kill_polling_thread(ctlr_info_t *h)
>+{
>+	int flags = 0;
>+	if (h->polling_thread) {
>+		printk("starting to kill the  polling thread\n");
>+		spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
>+		h->polling_thread_run = 0;
>+		spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
>+	}
>+}
>+
>+/*
>  *  This is it.  Find all the controllers and register them.  I really hate
>  *  stealing all these major device numbers.
>  *  returns the number of block devices registered.
>@@ -3305,6 +3439,16 @@ static int __devinit cciss_init_one(stru
> 
> 	hba[i]->busy_initializing = 1;
> 
>+	if (reset_devices) {
>+		rc = cciss_reset_controller(pdev, i, hba[i]);
>+		if (rc) {
>+			printk(KERN_WARNING "cciss: controller reset"
>+					" failed with %d\n", rc);
>+			free_hba(i);
>+			return -EIO;
>+		}
>+	}
>+
> 	if (cciss_pci_init(hba[i], pdev) != 0)
> 		goto clean1;
> 
>@@ -3342,11 +3486,16 @@ static int __devinit cciss_init_one(stru
> 
> 	/* make sure the board interrupts are off */
> 	hba[i]->access.set_intr_mask(hba[i], CCISS_INTR_OFF);
>-	if (request_irq(hba[i]->intr[SIMPLE_MODE_INT], do_cciss_intr,
>-			IRQF_DISABLED | IRQF_SHARED, hba[i]->devname, hba[i])) {
>-		printk(KERN_ERR "cciss: Unable to get irq %d for %s\n",
>-		       hba[i]->intr[SIMPLE_MODE_INT], hba[i]->devname);
>-		goto clean2;
>+
>+	if (!reset_devices) {
>+		printk(KERN_WARNING "cciss: requesting IRQ\n");
>+		if (request_irq(hba[i]->intr[SIMPLE_MODE_INT], do_cciss_intr,
>+				IRQF_DISABLED | IRQF_SHARED,
>+				hba[i]->devname, hba[i])) {
>+			printk(KERN_ERR "cciss: Unable to get irq %d for %s\n",
>+			       hba[i]->intr[SIMPLE_MODE_INT], hba[i]->devname);
>+			goto clean2;
>+		}
> 	}
> 
> 	printk(KERN_INFO "%s: <0x%x> at PCI %s IRQ %d%s using DAC\n",
>@@ -3370,6 +3519,7 @@ static int __devinit cciss_init_one(stru
> 		printk(KERN_ERR "cciss: out of memory");
> 		goto clean4;
> 	}
>+
> #ifdef CONFIG_CISS_SCSI_TAPE
> 	hba[i]->scsi_rejects.complete =
> 	    kmalloc(sizeof(hba[i]->scsi_rejects.complete[0]) *
>@@ -3398,8 +3548,17 @@ static int __devinit cciss_init_one(stru
> 
> 	cciss_scsi_setup(i);
> 
>-	/* Turn the interrupts on so we can service requests */
>-	hba[i]->access.set_intr_mask(hba[i], CCISS_INTR_ON);
>+	if (reset_devices) {
>+		if (start_completion_thread(hba[i],
>+				cciss_completion_thread, &rc) != 0)
>+			printk("Unable to start cciss_completion_thread\n");
>+	}
>+
>+	/* If we're not booting a crashkernel then turn interrupts
>+	 * on so we can service requests
>+	 */
>+	if (!reset_devices)
>+		hba[i]->access.set_intr_mask(hba[i], CCISS_INTR_ON);
> 
> 	cciss_procinit(i);
> 
>@@ -3459,6 +3618,7 @@ static int __devinit cciss_init_one(stru
> 			continue;
> 		blk_queue_hardsect_size(q, drv->block_size);
> 		set_capacity(disk, drv->nr_blocks);
>+		printk("cciss: calling add_disk\n");
> 		add_disk(disk);
> 		j++;
> 	} while (j <= hba[i]->highest_lun);
>@@ -3491,7 +3651,8 @@ static int __devinit cciss_init_one(stru
> 			blk_cleanup_queue(drv->queue);
> 	}
> 	pci_release_regions(pdev);
>-	pci_disable_device(pdev);
>+	printk("cciss: would have called pci_disable_device\n");
>+	/* pci_disable_device(pdev); */
> 	pci_set_drvdata(pdev, NULL);
> 	free_hba(i);
> 	return -1;
>@@ -3558,6 +3719,9 @@ static void __devexit cciss_remove_one(s
> 		}
> 	}
> 
>+	if (reset_devices)
>+		kill_polling_thread(hba[i]);
>+
> 	cciss_unregister_scsi(i);	/* unhook from SCSI subsystem */
> 
> 	cciss_shutdown(pdev);
>@@ -3608,6 +3772,35 @@ static int __init cciss_init(void)
> 	return pci_register_driver(&cciss_pci_driver);
> }
> 
>+static int start_completion_thread(ctlr_info_t *h,
>+			int (*cciss_completion_thread)(void *), int *rc) {
>+
>+	int flags = 0;
>+
>+	spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
>+	h->polling_thread_run = 1;
>+	spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
>+
>+	kernel_thread(cciss_completion_thread, h, 0);
>+	return 0;
>+}
>+
>+static int cciss_completion_thread(void *ctlr)
>+{
>+	ctlr_info_t *h = (ctlr_info_t *) ctlr;
>+	daemonize("cciss_completion");
>+
>+	h->polling_thread = current;
>+	printk("cciss%d: Completion thread starting.\n", h->ctlr);
>+
>+	while (h->polling_thread_run) {
>+		do_cciss_intr(0, h, NULL);
>+		msleep(1);
>+	}
>+
>+	return 0;
>+}
>+
> static void __exit cciss_cleanup(void)
> {
> 	int i;
>diff -urNp linux-2.6.18.i386-orig/drivers/block/cciss_cmd.h linux-2.6.18.i386/drivers/block/cciss_cmd.h
>--- linux-2.6.18.i386-orig/drivers/block/cciss_cmd.h	2008-03-14 13:50:06.000000000 -0500
>+++ linux-2.6.18.i386/drivers/block/cciss_cmd.h	2008-03-14 14:11:05.000000000 -0500
>@@ -102,6 +102,31 @@ typedef struct _InquiryData_struct
>   BYTE data_byte[36];
> } InquiryData_struct;
> 
>+/* defines for the messages supported by Smart Array */
>+#define CISS_ABORT_MSG	0x00
>+#define CISS_RESET_MSG	0x01
>+#define CISS_SCAN_MSG	0x02
>+#define CISS_NO_OP	0x03
>+
>+/* target defines for CCISS_ABORT_MSG */
>+#define TASK_ABORT	0x00	/* abort a command */
>+#define TASKSET_ABORT	0x01	/* abort all commands to a target */
>+#define CISS_CLEAR_ACA	0x02	/* clear ACA condition */
>+#define CLEAR_TASKSET	0x03	/* abort all commands to a target */
>+
>+/* target defines for CCISS_RESET_MSG */
>+#define RESET_CONTROLLER	0x00	/* reset the controller */
>+#define RESET_BUS		0x01	/* reset the bus */
>+#define RESET_TARGET		0x03	/* reset a target device */
>+#define RESET_LUN		0x04	/* reset a logical volume */
>+
>+/* target defines for CCISS_SCAN_MSG
>+ * the target is specified in the LUN field */
>+#define SCAN_ALL	0x00	/* scan all buses */
>+#define SCAN_BUS	0x01	/* scan a bus */
>+#define SCAN_TARGET	0x03	/* scan a target device */
>+#define SCAN_LUN	0x04	/* scan a logical volume */
>+
> #define CISS_REPORT_LOG 0xc2    /* Report Logical LUNs */
> #define CISS_REPORT_PHYS 0xc3   /* Report Physical LUNs */
> // Data returned
>diff -urNp linux-2.6.18.i386-orig/drivers/block/cciss.h linux-2.6.18.i386/drivers/block/cciss.h
>--- linux-2.6.18.i386-orig/drivers/block/cciss.h	2008-03-14 13:50:07.000000000 -0500
>+++ linux-2.6.18.i386/drivers/block/cciss.h	2008-03-14 14:11:05.000000000 -0500
>@@ -59,6 +59,8 @@ struct ctlr_info 
> 	__u32	board_id;
> 	void __iomem *vaddr;
> 	unsigned long paddr;
>+        struct task_struct *polling_thread;
>+        int     polling_thread_run;
> 	int 	nr_cmds; /* Number of commands allowed on this controller */
> 	CfgTable_struct __iomem *cfgtable;
> 	int	interrupts_enabled;
>diff -urNp linux-2.6.18.i386-orig/drivers/block/cciss_scsi.c linux-2.6.18.i386/drivers/block/cciss_scsi.c
>--- linux-2.6.18.i386-orig/drivers/block/cciss_scsi.c	2008-03-14 13:50:14.000000000 -0500
>+++ linux-2.6.18.i386/drivers/block/cciss_scsi.c	2008-03-14 14:11:05.000000000 -0500
>@@ -293,6 +293,8 @@ print_bytes (unsigned char *c, int len, 
> 	}
> }
> 
>+#endif
>+
> static void
> print_cmd(CommandList_struct *cp)
> {
>@@ -339,8 +341,6 @@ print_cmd(CommandList_struct *cp)
> 			
> }
> 
>-#endif
>-
> static int 
> find_bus_target_lun(int ctlr, int *bus, int *target, int *lun)
> {

Actions: View | Diff

Attachments on bug 230717: 149128 | 149208 | 292178 | 293393 | 293602 | 293604 | 294226 | 298090 | 298451 | 298533 | 298806 | 299186 | 299200 | 302760 | 302938 | 303074