[SCSI] scsi_debug: Thin provisioning support

This version fixes 64-bit modulo on 32-bit as well as inadvertent map
updates when TP was disabled.

Implement support for thin provisioning in scsi_debug.  No actual memory
de-allocation is taking place.  The intent is to emulate a thinly
provisioned storage device, not to be one.

There are four new module options:

 - unmap_granularity specifies the granularity at which to track mapped
   blocks (specified in number of logical blocks).  2048 (1 MB) is a
   realistic value for disk arrays although some may have a finer
   granularity.

 - unmap_alignment specifies the first LBA which is naturally aligned on
   an unmap_granularity boundary.

 - unmap_max_desc specifies the maximum number of ranges that can be
   unmapped using one UNMAP command.  If this is 0, only WRITE SAME is
   supported and UNMAP will cause a check condition.

 - unmap_max_blocks specifies the maximum number of blocks that can be
   unmapped using a single UNMAP command.  Default is 0xffffffff.

These parameters are reported in the new and extended block limits VPD.

If unmap_granularity is specified the device is tagged as thin
provisioning capable in READ CAPACITY(16).  A bitmap is allocated to
track whether blocks are mapped or not.  A WRITE request will cause a
block to be mapped.  So will WRITE SAME unless the UNMAP bit is set.

Blocks can be unmapped using either WRITE SAME or UNMAP.  No accounting
is done to track partial blocks.  This means that only whole blocks will
be marked free.  This is how the array people tell me their firmwares
work.

GET LBA STATUS is also supported.  This command reports whether a block
is mapped or not, and how long the adjoining mapped/unmapped extent is.

The block allocation bitmap can also be viewed from user space via:

	/sys/bus/pseudo/drivers/scsi_debug/map
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-by: Douglas Gilbert <dgilbert@interlog.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

drivers/scsi/scsi_debug.c

@@ -44,6 +44,8 @@
 
 #include <net/checksum.h>
 
+#include <asm/unaligned.h>
+
 #include <scsi/scsi.h>
 #include <scsi/scsi_cmnd.h>
 #include <scsi/scsi_device.h>
@@ -105,6 +107,10 @@ static const char * scsi_debug_version_date = "20070104";
 #define DEF_ATO 1
 #define DEF_PHYSBLK_EXP 0
 #define DEF_LOWEST_ALIGNED 0
+#define DEF_UNMAP_MAX_BLOCKS 0
+#define DEF_UNMAP_MAX_DESC 0
+#define DEF_UNMAP_GRANULARITY 0
+#define DEF_UNMAP_ALIGNMENT 0
 
 /* bit mask values for scsi_debug_opts */
 #define SCSI_DEBUG_OPT_NOISE   1
@@ -162,6 +168,10 @@ static int scsi_debug_guard = DEF_GUARD;
 static int scsi_debug_ato = DEF_ATO;
 static int scsi_debug_physblk_exp = DEF_PHYSBLK_EXP;
 static int scsi_debug_lowest_aligned = DEF_LOWEST_ALIGNED;
+static int scsi_debug_unmap_max_desc = DEF_UNMAP_MAX_DESC;
+static int scsi_debug_unmap_max_blocks = DEF_UNMAP_MAX_BLOCKS;
+static int scsi_debug_unmap_granularity = DEF_UNMAP_GRANULARITY;
+static int scsi_debug_unmap_alignment = DEF_UNMAP_ALIGNMENT;
 
 static int scsi_debug_cmnd_count = 0;
 
@@ -223,7 +233,9 @@ static struct sdebug_queued_cmd queued_arr[SCSI_DEBUG_CANQUEUE];
 
 static unsigned char * fake_storep;	/* ramdisk storage */
 static unsigned char *dif_storep;	/* protection info */
+static void *map_storep;		/* provisioning map */
 
+static unsigned long map_size;
 static int num_aborts = 0;
 static int num_dev_resets = 0;
 static int num_bus_resets = 0;
@@ -317,6 +329,7 @@ static void get_data_transfer_info(unsigned char *cmd,
 			(u32)cmd[28] << 24;
 		break;
 
+	case WRITE_SAME_16:
 	case WRITE_16:
 	case READ_16:
 		*lba = (u64)cmd[9] | (u64)cmd[8] << 8 |
@@ -335,6 +348,7 @@ static void get_data_transfer_info(unsigned char *cmd,
 		*num = (u32)cmd[9] | (u32)cmd[8] << 8 | (u32)cmd[7] << 16 |
 			(u32)cmd[6] << 24;
 		break;
+	case WRITE_SAME:
 	case WRITE_10:
 	case READ_10:
 	case XDWRITEREAD_10:
@@ -691,6 +705,29 @@ static int inquiry_evpd_b0(unsigned char * arr)
 		arr[6] = (sdebug_store_sectors >> 8) & 0xff;
 		arr[7] = sdebug_store_sectors & 0xff;
 	}
+
+	if (scsi_debug_unmap_max_desc) {
+		unsigned int blocks;
+
+		if (scsi_debug_unmap_max_blocks)
+			blocks = scsi_debug_unmap_max_blocks;
+		else
+			blocks = 0xffffffff;
+
+		put_unaligned_be32(blocks, &arr[16]);
+		put_unaligned_be32(scsi_debug_unmap_max_desc, &arr[20]);
+	}
+
+	if (scsi_debug_unmap_alignment) {
+		put_unaligned_be32(scsi_debug_unmap_alignment, &arr[28]);
+		arr[28] |= 0x80; /* UGAVALID */
+	}
+
+	if (scsi_debug_unmap_granularity) {
+		put_unaligned_be32(scsi_debug_unmap_granularity, &arr[24]);
+		return 0x3c; /* Mandatory page length for thin provisioning */
+	}
+
 	return sizeof(vpdb0_data);
 }
 
@@ -974,6 +1011,10 @@ static int resp_readcap16(struct scsi_cmnd * scp,
 	arr[11] = scsi_debug_sector_size & 0xff;
 	arr[13] = scsi_debug_physblk_exp & 0xf;
 	arr[14] = (scsi_debug_lowest_aligned >> 8) & 0x3f;
+
+	if (scsi_debug_unmap_granularity)
+		arr[14] |= 0x80; /* TPE */
+
 	arr[15] = scsi_debug_lowest_aligned & 0xff;
 
 	if (scsi_debug_dif) {
@@ -1887,6 +1928,70 @@ static int prot_verify_write(struct scsi_cmnd *SCpnt, sector_t start_sec,
 	return ret;
 }
 
+static unsigned int map_state(sector_t lba, unsigned int *num)
+{
+	unsigned int granularity, alignment, mapped;
+	sector_t block, next, end;
+
+	granularity = scsi_debug_unmap_granularity;
+	alignment = granularity - scsi_debug_unmap_alignment;
+	block = lba + alignment;
+	do_div(block, granularity);
+
+	mapped = test_bit(block, map_storep);
+
+	if (mapped)
+		next = find_next_zero_bit(map_storep, map_size, block);
+	else
+		next = find_next_bit(map_storep, map_size, block);
+
+	end = next * granularity - scsi_debug_unmap_alignment;
+	*num = end - lba;
+
+	return mapped;
+}
+
+static void map_region(sector_t lba, unsigned int len)
+{
+	unsigned int granularity, alignment;
+	sector_t end = lba + len;
+
+	granularity = scsi_debug_unmap_granularity;
+	alignment = granularity - scsi_debug_unmap_alignment;
+
+	while (lba < end) {
+		sector_t block, rem;
+
+		block = lba + alignment;
+		rem = do_div(block, granularity);
+
+		set_bit(block, map_storep);
+
+		lba += granularity - rem;
+	}
+}
+
+static void unmap_region(sector_t lba, unsigned int len)
+{
+	unsigned int granularity, alignment;
+	sector_t end = lba + len;
+
+	granularity = scsi_debug_unmap_granularity;
+	alignment = granularity - scsi_debug_unmap_alignment;
+
+	while (lba < end) {
+		sector_t block, rem;
+
+		block = lba + alignment;
+		rem = do_div(block, granularity);
+
+		if (rem == 0 && lba + granularity <= end)
+			clear_bit(block, map_storep);
+
+		lba += granularity - rem;
+	}
+}
+
 static int resp_write(struct scsi_cmnd *SCpnt, unsigned long long lba,
 		      unsigned int num, struct sdebug_dev_info *devip,
 		      u32 ei_lba)
@@ -1910,6 +2015,8 @@ static int resp_write(struct scsi_cmnd *SCpnt, unsigned long long lba,
 
 	write_lock_irqsave(&atomic_rw, iflags);
 	ret = do_device_access(SCpnt, devip, lba, num, 1);
+	if (scsi_debug_unmap_granularity)
+		map_region(lba, num);
 	write_unlock_irqrestore(&atomic_rw, iflags);
 	if (-1 == ret)
 		return (DID_ERROR << 16);
@@ -1917,7 +2024,141 @@ static int resp_write(struct scsi_cmnd *SCpnt, unsigned long long lba,
 		 (SCSI_DEBUG_OPT_NOISE & scsi_debug_opts))
 		printk(KERN_INFO "scsi_debug: write: cdb indicated=%u, "
 		       " IO sent=%d bytes\n", num * scsi_debug_sector_size, ret);
+
+	return 0;
+}
+
+static int resp_write_same(struct scsi_cmnd *scmd, unsigned long long lba,
+		      unsigned int num, struct sdebug_dev_info *devip,
+			   u32 ei_lba, unsigned int unmap)
+{
+	unsigned long iflags;
+	unsigned long long i;
+	int ret;
+
+	ret = check_device_access_params(devip, lba, num);
+	if (ret)
+		return ret;
+
+	write_lock_irqsave(&atomic_rw, iflags);
+
+	if (unmap && scsi_debug_unmap_granularity) {
+		unmap_region(lba, num);
+		goto out;
+	}
+
+	/* Else fetch one logical block */
+	ret = fetch_to_dev_buffer(scmd,
+				  fake_storep + (lba * scsi_debug_sector_size),
+				  scsi_debug_sector_size);
+
+	if (-1 == ret) {
+		write_unlock_irqrestore(&atomic_rw, iflags);
+		return (DID_ERROR << 16);
+	} else if ((ret < (num * scsi_debug_sector_size)) &&
+		 (SCSI_DEBUG_OPT_NOISE & scsi_debug_opts))
+		printk(KERN_INFO "scsi_debug: write same: cdb indicated=%u, "
+		       " IO sent=%d bytes\n", num * scsi_debug_sector_size, ret);
+
+	/* Copy first sector to remaining blocks */
+	for (i = 1 ; i < num ; i++)
+		memcpy(fake_storep + ((lba + i) * scsi_debug_sector_size),
+		       fake_storep + (lba * scsi_debug_sector_size),
+		       scsi_debug_sector_size);
+
+	if (scsi_debug_unmap_granularity)
+		map_region(lba, num);
+out:
+	write_unlock_irqrestore(&atomic_rw, iflags);
+
+	return 0;
+}
+
+struct unmap_block_desc {
+	__be64	lba;
+	__be32	blocks;
+	__be32	__reserved;
+};
+
+static int resp_unmap(struct scsi_cmnd * scmd, struct sdebug_dev_info * devip)
+{
+	unsigned char *buf;
+	struct unmap_block_desc *desc;
+	unsigned int i, payload_len, descriptors;
+	int ret;
+
+	ret = check_readiness(scmd, 1, devip);
+	if (ret)
+		return ret;
+
+	payload_len = get_unaligned_be16(&scmd->cmnd[7]);
+	BUG_ON(scsi_bufflen(scmd) != payload_len);
+
+	descriptors = (payload_len - 8) / 16;
+
+	buf = kmalloc(scsi_bufflen(scmd), GFP_ATOMIC);
+	if (!buf)
+		return check_condition_result;
+
+	scsi_sg_copy_to_buffer(scmd, buf, scsi_bufflen(scmd));
+
+	BUG_ON(get_unaligned_be16(&buf[0]) != payload_len - 2);
+	BUG_ON(get_unaligned_be16(&buf[2]) != descriptors * 16);
+
+	desc = (void *)&buf[8];
+
+	for (i = 0 ; i < descriptors ; i++) {
+		unsigned long long lba = get_unaligned_be64(&desc[i].lba);
+		unsigned int num = get_unaligned_be32(&desc[i].blocks);
+
+		ret = check_device_access_params(devip, lba, num);
+		if (ret)
+			goto out;
+
+		unmap_region(lba, num);
+	}
+
+	ret = 0;
+
+out:
+	kfree(buf);
+
+	return ret;
+}
+
+#define SDEBUG_GET_LBA_STATUS_LEN 32
+
+static int resp_get_lba_status(struct scsi_cmnd * scmd,
+			       struct sdebug_dev_info * devip)
+{
+	unsigned long long lba;
+	unsigned int alloc_len, mapped, num;
+	unsigned char arr[SDEBUG_GET_LBA_STATUS_LEN];
+	int ret;
+
+	ret = check_readiness(scmd, 1, devip);
+	if (ret)
+		return ret;
+
+	lba = get_unaligned_be64(&scmd->cmnd[2]);
+	alloc_len = get_unaligned_be32(&scmd->cmnd[10]);
+
+	if (alloc_len < 24)
 		return 0;
+
+	ret = check_device_access_params(devip, lba, 1);
+	if (ret)
+		return ret;
+
+	mapped = map_state(lba, &num);
+
+	memset(arr, 0, SDEBUG_GET_LBA_STATUS_LEN);
+	put_unaligned_be32(16, &arr[0]);	/* Parameter Data Length */
+	put_unaligned_be64(lba, &arr[8]);	/* LBA */
+	put_unaligned_be32(num, &arr[16]);	/* Number of blocks */
+	arr[20] = !mapped;			/* mapped = 0, unmapped = 1 */
+
+	return fill_from_dev_buffer(scmd, arr, SDEBUG_GET_LBA_STATUS_LEN);
 }
 
 #define SDEBUG_RLUN_ARR_SZ 256
@@ -2430,6 +2671,10 @@ module_param_named(guard, scsi_debug_guard, int, S_IRUGO);
 module_param_named(ato, scsi_debug_ato, int, S_IRUGO);
 module_param_named(physblk_exp, scsi_debug_physblk_exp, int, S_IRUGO);
 module_param_named(lowest_aligned, scsi_debug_lowest_aligned, int, S_IRUGO);
+module_param_named(unmap_max_blocks, scsi_debug_unmap_max_blocks, int, S_IRUGO);
+module_param_named(unmap_max_desc, scsi_debug_unmap_max_desc, int, S_IRUGO);
+module_param_named(unmap_granularity, scsi_debug_unmap_granularity, int, S_IRUGO);
+module_param_named(unmap_alignment, scsi_debug_unmap_alignment, int, S_IRUGO);
 
 MODULE_AUTHOR("Eric Youngdale + Douglas Gilbert");
 MODULE_DESCRIPTION("SCSI debug adapter driver");
@@ -2458,6 +2703,10 @@ MODULE_PARM_DESC(dix, "data integrity extensions mask (def=0)");
 MODULE_PARM_DESC(dif, "data integrity field type: 0-3 (def=0)");
 MODULE_PARM_DESC(guard, "protection checksum: 0=crc, 1=ip (def=0)");
 MODULE_PARM_DESC(ato, "application tag ownership: 0=disk 1=host (def=1)");
+MODULE_PARM_DESC(unmap_max_blocks, "max # of blocks can be unmapped in one cmd (def=0)");
+MODULE_PARM_DESC(unmap_max_desc, "max # of ranges that can be unmapped in one cmd (def=0)");
+MODULE_PARM_DESC(unmap_granularity, "thin provisioning granularity in blocks (def=0)");
+MODULE_PARM_DESC(unmap_alignment, "lowest aligned thin provisioning lba (def=0)");
 
 static char sdebug_info[256];
 
@@ -2816,6 +3065,23 @@ static ssize_t sdebug_ato_show(struct device_driver *ddp, char *buf)
 }
 DRIVER_ATTR(ato, S_IRUGO, sdebug_ato_show, NULL);
 
+static ssize_t sdebug_map_show(struct device_driver *ddp, char *buf)
+{
+	ssize_t count;
+
+	if (scsi_debug_unmap_granularity == 0)
+		return scnprintf(buf, PAGE_SIZE, "0-%u\n",
+				 sdebug_store_sectors);
+
+	count = bitmap_scnlistprintf(buf, PAGE_SIZE, map_storep, map_size);
+
+	buf[count++] = '\n';
+	buf[count++] = 0;
+
+	return count;
+}
+DRIVER_ATTR(map, S_IRUGO, sdebug_map_show, NULL);
+
 
 /* Note: The following function creates attribute files in the
    /sys/bus/pseudo/drivers/scsi_debug directory. The advantage of these
@@ -2847,11 +3113,13 @@ static int do_create_driverfs_files(void)
 	ret |= driver_create_file(&sdebug_driverfs_driver, &driver_attr_dif);
 	ret |= driver_create_file(&sdebug_driverfs_driver, &driver_attr_guard);
 	ret |= driver_create_file(&sdebug_driverfs_driver, &driver_attr_ato);
+	ret |= driver_create_file(&sdebug_driverfs_driver, &driver_attr_map);
 	return ret;
 }
 
 static void do_remove_driverfs_files(void)
 {
+	driver_remove_file(&sdebug_driverfs_driver, &driver_attr_map);
 	driver_remove_file(&sdebug_driverfs_driver, &driver_attr_ato);
 	driver_remove_file(&sdebug_driverfs_driver, &driver_attr_guard);
 	driver_remove_file(&sdebug_driverfs_driver, &driver_attr_dif);
@@ -2989,6 +3257,36 @@ static int __init scsi_debug_init(void)
 		memset(dif_storep, 0xff, dif_size);
 	}
 
+	if (scsi_debug_unmap_granularity) {
+		unsigned int map_bytes;
+
+		if (scsi_debug_unmap_granularity < scsi_debug_unmap_alignment) {
+			printk(KERN_ERR
+			       "%s: ERR: unmap_granularity < unmap_alignment\n",
+			       __func__);
+			return -EINVAL;
+		}
+
+		map_size = (sdebug_store_sectors / scsi_debug_unmap_granularity);
+		map_bytes = map_size >> 3;
+		map_storep = vmalloc(map_bytes);
+
+		printk(KERN_INFO "scsi_debug_init: %lu provisioning blocks\n",
+		       map_size);
+
+		if (map_storep == NULL) {
+			printk(KERN_ERR "scsi_debug_init: out of mem. (MAP)\n");
+			ret = -ENOMEM;
+			goto free_vm;
+		}
+
+		memset(map_storep, 0x0, map_bytes);
+
+		/* Map first 1KB for partition table */
+		if (scsi_debug_num_parts)
+			map_region(0, 2);
+	}
+
 	ret = device_register(&pseudo_primary);
 	if (ret < 0) {
 		printk(KERN_WARNING "scsi_debug: device_register error: %d\n",
@@ -3041,6 +3339,8 @@ static int __init scsi_debug_init(void)
 dev_unreg:
 	device_unregister(&pseudo_primary);
 free_vm:
+	if (map_storep)
+		vfree(map_storep);
 	if (dif_storep)
 		vfree(dif_storep);
 	vfree(fake_storep);
@@ -3167,6 +3467,7 @@ int scsi_debug_queuecommand(struct scsi_cmnd *SCpnt, done_funct_t done)
 	int inj_dif = 0;
 	int inj_dix = 0;
 	int delay_override = 0;
+	int unmap = 0;
 
 	scsi_set_resid(SCpnt, 0);
 	if ((SCSI_DEBUG_OPT_NOISE & scsi_debug_opts) && cmd) {
@@ -3272,13 +3573,21 @@ int scsi_debug_queuecommand(struct scsi_cmnd *SCpnt, done_funct_t done)
 		errsts = resp_readcap(SCpnt, devip);
 		break;
 	case SERVICE_ACTION_IN:
-		if (SAI_READ_CAPACITY_16 != cmd[1]) {
+		if (cmd[1] == SAI_READ_CAPACITY_16)
+			errsts = resp_readcap16(SCpnt, devip);
+		else if (cmd[1] == SAI_GET_LBA_STATUS) {
+
+			if (scsi_debug_unmap_max_desc == 0) {
+				mk_sense_buffer(devip, ILLEGAL_REQUEST,
+						INVALID_COMMAND_OPCODE, 0);
+				errsts = check_condition_result;
+			} else
+				errsts = resp_get_lba_status(SCpnt, devip);
+		} else {
 			mk_sense_buffer(devip, ILLEGAL_REQUEST,
 					INVALID_OPCODE, 0);
 			errsts = check_condition_result;
-			break;
 		}
-		errsts = resp_readcap16(SCpnt, devip);
 		break;
 	case MAINTENANCE_IN:
 		if (MI_REPORT_TARGET_PGS != cmd[1]) {
@@ -3378,6 +3687,29 @@ int scsi_debug_queuecommand(struct scsi_cmnd *SCpnt, done_funct_t done)
 			errsts = illegal_condition_result;
 		}
 		break;
+	case WRITE_SAME_16:
+		if (cmd[1] & 0x8)
+			unmap = 1;
+		/* fall through */
+	case WRITE_SAME:
+		errsts = check_readiness(SCpnt, 0, devip);
+		if (errsts)
+			break;
+		get_data_transfer_info(cmd, &lba, &num, &ei_lba);
+		errsts = resp_write_same(SCpnt, lba, num, devip, ei_lba, unmap);
+		break;
+	case UNMAP:
+		errsts = check_readiness(SCpnt, 0, devip);
+		if (errsts)
+			break;
+
+		if (scsi_debug_unmap_max_desc == 0) {
+			mk_sense_buffer(devip, ILLEGAL_REQUEST,
+					INVALID_COMMAND_OPCODE, 0);
+			errsts = check_condition_result;
+		} else
+			errsts = resp_unmap(SCpnt, devip);
+		break;
 	case MODE_SENSE:
 	case MODE_SENSE_10:
 		errsts = resp_mode_sense(SCpnt, target, devip);