static struct block_place get_degraded_block(u64 block, int total_disks, int block_size, u64 device_length)
{
struct block_place degraded_place;
u64 block_pos;
u64 empty_zone_offset;
block_pos = block;
// Let's count device number in empty zone
degraded_place.device_number = sector_div(block_pos, (total_disks - 1));
// Little fix
if (degraded_place.device_number >= DEGRADED_DISK)
degraded_place.device_number++;
// Now let's count sector number in empty_zone.
// First of all, we should count empty_zone_offset.
empty_zone_offset = device_length;
// p_blocks + e_block = 3.
sector_div(empty_zone_offset, (total_disks - 3));
// Now empty_zone_offset == one real disk capacity (including empty and parity)
sector_div(empty_zone_offset, total_disks); // Almost ready.
degraded_place.sector = empty_zone_offset + block_pos * block_size;
return degraded_place;
}
static struct ppa_addr linear_to_generic_addr(struct nvm_dev *dev,
struct ppa_addr r)
{
struct ppa_addr l;
int secs, pgs, blks, luns;
sector_t ppa = r.ppa;
l.ppa = 0;
div_u64_rem(ppa, dev->sec_per_pg, &secs);
l.g.sec = secs;
sector_div(ppa, dev->sec_per_pg);
div_u64_rem(ppa, dev->pgs_per_blk, &pgs);
l.g.pg = pgs;
sector_div(ppa, dev->pgs_per_blk);
div_u64_rem(ppa, dev->blks_per_lun, &blks);
l.g.blk = blks;
sector_div(ppa, dev->blks_per_lun);
div_u64_rem(ppa, dev->luns_per_chnl, &luns);
l.g.lun = luns;
sector_div(ppa, dev->luns_per_chnl);
l.g.ch = ppa;
return l;
}
static int null_lnvm_id(struct nvm_dev *dev, struct nvm_id *id)
{
sector_t size = gb * 1024 * 1024 * 1024ULL;
sector_t blksize;
struct nvm_id_group *grp;
id->ver_id = 0x1;
id->vmnt = 0;
id->cgrps = 1;
id->cap = 0x2;
id->dom = 0x1;
id->ppaf.blk_offset = 0;
id->ppaf.blk_len = 16;
id->ppaf.pg_offset = 16;
id->ppaf.pg_len = 16;
id->ppaf.sect_offset = 32;
id->ppaf.sect_len = 8;
id->ppaf.pln_offset = 40;
id->ppaf.pln_len = 8;
id->ppaf.lun_offset = 48;
id->ppaf.lun_len = 8;
id->ppaf.ch_offset = 56;
id->ppaf.ch_len = 8;
sector_div(size, bs); /* convert size to pages */
size >>= 8; /* concert size to pgs pr blk */
grp = &id->groups[0];
grp->mtype = 0;
grp->fmtype = 0;
grp->num_ch = 1;
grp->num_pg = 256;
blksize = size;
size >>= 16;
grp->num_lun = size + 1;
sector_div(blksize, grp->num_lun);
grp->num_blk = blksize;
grp->num_pln = 1;
grp->fpg_sz = bs;
grp->csecs = bs;
grp->trdt = 25000;
grp->trdm = 25000;
grp->tprt = 500000;
grp->tprm = 500000;
grp->tbet = 1500000;
grp->tbem = 1500000;
grp->mpos = 0x010101; /* single plane rwe */
grp->cpar = hw_queue_depth;
return 0;
}
/**
* ata_std_bios_param - generic bios head/sector/cylinder calculator used by sd.
* @sdev: SCSI device for which BIOS geometry is to be determined
* @bdev: block device associated with @sdev
* @capacity: capacity of SCSI device
* @geom: location to which geometry will be output
*
* Generic bios head/sector/cylinder calculator
* used by sd. Most BIOSes nowadays expect a XXX/255/16 (CHS)
* mapping. Some situations may arise where the disk is not
* bootable if this is not used.
*
* LOCKING:
* Defined by the SCSI layer. We don't really care.
*
* RETURNS:
* Zero.
*/
int ata_std_bios_param(struct scsi_device *sdev, struct block_device *bdev,
sector_t capacity, int geom[])
{
geom[0] = 255;
geom[1] = 63;
sector_div(capacity, 255*63);
geom[2] = capacity;
return 0;
}
int pt_getgeo(struct block_device * block_device, struct hd_geometry * hg)
{
hg->heads = 255;
hg->sectors = 63;
hg->cylinders = get_capacity(block_device->bd_disk);
sector_div(hg->cylinders, hg->heads * hg->sectors);
return 0;
}
static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
{
int i, c, err;
sector_t curr_zone_end, sectors;
struct md_rdev *smallest, *rdev1, *rdev2, *rdev, **dev;
struct strip_zone *zone;
int cnt;
char b[BDEVNAME_SIZE];
char b2[BDEVNAME_SIZE];
struct r0conf *conf = kzalloc(sizeof(*conf), GFP_KERNEL);
unsigned short blksize = 512;
*private_conf = ERR_PTR(-ENOMEM);
if (!conf)
return -ENOMEM;
rdev_for_each(rdev1, mddev) {
pr_debug("md/raid0:%s: looking at %s\n",
mdname(mddev),
bdevname(rdev1->bdev, b));
c = 0;
/* round size to chunk_size */
sectors = rdev1->sectors;
sector_div(sectors, mddev->chunk_sectors);
rdev1->sectors = sectors * mddev->chunk_sectors;
blksize = max(blksize, queue_logical_block_size(
rdev1->bdev->bd_disk->queue));
rdev_for_each(rdev2, mddev) {
pr_debug("md/raid0:%s: comparing %s(%llu)"
" with %s(%llu)\n",
mdname(mddev),
bdevname(rdev1->bdev,b),
(unsigned long long)rdev1->sectors,
bdevname(rdev2->bdev,b2),
(unsigned long long)rdev2->sectors);
if (rdev2 == rdev1) {
pr_debug("md/raid0:%s: END\n",
mdname(mddev));
break;
}
if (rdev2->sectors == rdev1->sectors) {
/*
* Not unique, don't count it as a new
* group
*/
pr_debug("md/raid0:%s: EQUAL\n",
mdname(mddev));
c = 1;
break;
}
pr_debug("md/raid0:%s: NOT EQUAL\n",
mdname(mddev));
}
static struct recover_stripe raid6_recover(struct insane_c *ctx, u64 block, int device_number) {
struct recover_stripe result;
int block_place, counter, device, total_disks, chunk_size;
u64 onotole;
total_disks = raid6_alg.ndisks;
chunk_size = ctx->chunk_size;
// place of block in current stripe
onotole = block + device_number;
block_place = sector_div(onotole ,total_disks);
// starting block
onotole = block;
device = sector_div(onotole, total_disks);
if (device != 0)
device = total_disks - device;
else
device = 0;
counter = 0;
// we should read (total_disks - 2) blocks to recover
while (counter < total_disks - 2) {
if (device != device_number) {
result.read_sector[counter] = block * chunk_size;
result.read_device[counter] = device;
counter++;
}
device++;
onotole = sector_div(device, total_disks);
device = onotole;
}
result.write_device = device_number;
result.write_sector = block * chunk_size;
result.quantity = total_disks - 2;
return result;
}
static void pblk_set_provision(struct pblk *pblk, long nr_free_blks)
{
struct nvm_tgt_dev *dev = pblk->dev;
struct nvm_geo *geo = &dev->geo;
sector_t provisioned;
pblk->over_pct = 20;
provisioned = nr_free_blks;
provisioned *= (100 - pblk->over_pct);
sector_div(provisioned, 100);
/* Internally pblk manages all free blocks, but all calculations based
* on user capacity consider only provisioned blocks
*/
pblk->rl.total_blocks = nr_free_blks;
pblk->rl.nr_secs = nr_free_blks * geo->sec_per_blk;
pblk->capacity = provisioned * geo->sec_per_blk;
atomic_set(&pblk->rl.free_blocks, nr_free_blks);
}
/*
* Check the amount of free space and suspend/resume accordingly.
*/
static int check_free_space(struct bsd_acct_struct *acct, struct file *file)
{
struct kstatfs sbuf;
int res;
int act;
sector_t resume;
sector_t suspend;
spin_lock(&acct_lock);
res = acct->active;
if (!file || !acct->needcheck)
goto out;
spin_unlock(&acct_lock);
/* May block */
if (vfs_statfs(file->f_path.dentry, &sbuf))
return res;
suspend = sbuf.f_blocks * SUSPEND;
resume = sbuf.f_blocks * RESUME;
sector_div(suspend, 100);
sector_div(resume, 100);
if (sbuf.f_bavail <= suspend)
act = -1;
else if (sbuf.f_bavail >= resume)
act = 1;
else
act = 0;
/*
* If some joker switched acct->file under us we'ld better be
* silent and _not_ touch anything.
*/
spin_lock(&acct_lock);
if (file != acct->file) {
if (act)
res = act>0;
goto out;
}
if (acct->active) {
if (act < 0) {
acct->active = 0;
printk(KERN_INFO "Process accounting paused\n");
}
} else {
if (act > 0) {
acct->active = 1;
printk(KERN_INFO "Process accounting resumed\n");
}
}
del_timer(&acct->timer);
acct->needcheck = 0;
acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
add_timer(&acct->timer);
res = acct->active;
out:
spin_unlock(&acct_lock);
return res;
}
static int null_add_dev(void)
{
struct gendisk *disk;
struct nullb *nullb;
sector_t size;
int rv;
nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
if (!nullb) {
rv = -ENOMEM;
goto out;
}
spin_lock_init(&nullb->lock);
if (queue_mode == NULL_Q_MQ && use_per_node_hctx)
submit_queues = nr_online_nodes;
rv = setup_queues(nullb);
if (rv)
goto out_free_nullb;
if (queue_mode == NULL_Q_MQ) {
nullb->tag_set.ops = &null_mq_ops;
nullb->tag_set.nr_hw_queues = submit_queues;
nullb->tag_set.queue_depth = hw_queue_depth;
nullb->tag_set.numa_node = home_node;
nullb->tag_set.cmd_size = sizeof(struct nullb_cmd);
nullb->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
nullb->tag_set.driver_data = nullb;
rv = blk_mq_alloc_tag_set(&nullb->tag_set);
if (rv)
goto out_cleanup_queues;
nullb->q = blk_mq_init_queue(&nullb->tag_set);
if (IS_ERR(nullb->q)) {
rv = -ENOMEM;
goto out_cleanup_tags;
}
} else if (queue_mode == NULL_Q_BIO) {
nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
if (!nullb->q) {
rv = -ENOMEM;
goto out_cleanup_queues;
}
blk_queue_make_request(nullb->q, null_queue_bio);
rv = init_driver_queues(nullb);
if (rv)
goto out_cleanup_blk_queue;
} else {
nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
if (!nullb->q) {
rv = -ENOMEM;
goto out_cleanup_queues;
}
blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
rv = init_driver_queues(nullb);
if (rv)
goto out_cleanup_blk_queue;
}
nullb->q->queuedata = nullb;
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, nullb->q);
disk = nullb->disk = alloc_disk_node(1, home_node);
if (!disk) {
rv = -ENOMEM;
goto out_cleanup_blk_queue;
}
mutex_lock(&lock);
list_add_tail(&nullb->list, &nullb_list);
nullb->index = nullb_indexes++;
mutex_unlock(&lock);
blk_queue_logical_block_size(nullb->q, bs);
blk_queue_physical_block_size(nullb->q, bs);
size = gb * 1024 * 1024 * 1024ULL;
sector_div(size, bs);
set_capacity(disk, size);
disk->flags |= GENHD_FL_EXT_DEVT | GENHD_FL_SUPPRESS_PARTITION_INFO;
disk->major = null_major;
disk->first_minor = nullb->index;
disk->fops = &null_fops;
disk->private_data = nullb;
disk->queue = nullb->q;
sprintf(disk->disk_name, "nullb%d", nullb->index);
add_disk(disk);
return 0;
out_cleanup_blk_queue:
blk_cleanup_queue(nullb->q);
out_cleanup_tags:
if (queue_mode == NULL_Q_MQ)
blk_mq_free_tag_set(&nullb->tag_set);
out_cleanup_queues:
cleanup_queues(nullb);
out_free_nullb:
kfree(nullb);
out:
return rv;
}
// Sector and device mapping callback
static struct parity_places algorithm_raid6(struct insane_c *ctx, u64 block, sector_t *sector, int *device_number)
{
struct parity_places parity;
u64 position;
u64 i, Y;
u64 data_block; // Data block number
u64 local_gap; // Parity blocks skipped in current lane
u64 lane;
u64 block_start, block_offset;
int block_size;
int total_disks;
block_size = ctx->chunk_size;
total_disks = raid6_alg.ndisks;
data_block = *device_number + block * total_disks;
lane = data_block;
position = sector_div(lane, total_disks - raid6_alg.p_blocks);
i = lane;
Y = sector_div(i, total_disks);
local_gap = 2;
/* Now we have "square" of blocks.
* Position is horisontal coordinate, Y - vertical coordinate
*
* We would like to see something like this (D - data block, S - syndrome).
* _
* DDDDDSS |
* DDDDSSD |
* DDDSSDD |
* DDSSDDD > Square 1
* DSSDDDD |
* SSDDDDD |
* SDDDDDS _|
= DDDDDSS |
* DDDDSSD |
* DDDSSDD |
* DDSSDDD > Square 2
* DSSDDDD |
* SSDDDDD |
* SDDDDDS _|
*
*
* Local gap - how many syndromes we should skip in current stripe.
* For example, in this masterpiece scheme local gap equals:
* 0 in the top left corner;
* 1 in the last stripe;
* 2 in all other positions.
*
* So, local_gap_scheme looks like:
*
* 0000022
* 0000222
* 0002222
* 0022222
* 0222222
* 2222222
* 1111111
*
* Default local gap equals 2.
* At first, we will decrease local gap in the last stripe (Inner clause)
* Secondly, we will decrease local gap in the top left corner (Outer clause)
*/
// If we are in last stripe in square then we skip 1 syndrom in current lane
if( Y == (total_disks - 1) )
local_gap = 1;
// If we didn't cross square diagonal then we don't skip syndromes in
// current lane
if( position + Y < (total_disks - 2) )
local_gap = 0;
// Remap block accounting all gaps
position = data_block + local_gap + (2 * lane);
// Remap device number
*device_number = sector_div(position, total_disks);
// For sequential writing: let's check number of current block
parity.last_block = false;
if (ctx->io_pattern == SEQUENTIAL)
{
if (*device_number + (2 - local_gap) == (total_disks - 1))
parity.last_block = true;
}
// Get offset in block and remap sector
block_offset = sector_div(*sector, block_size);
block_start = position * block_size;
*sector = block_start + block_offset;
// Now it's time to count, where our syndromes are.
parity.start_device = 0;
parity.start_sector = block_start;
// Parities have same sector, different devices
parity.sector_number[0] = block_start;
parity.sector_number[1] = block_start;
parity.device_number[1] = total_disks - 1 - Y;
if( Y < total_disks - 1 )
parity.device_number[0] = parity.device_number[1] - 1;
else
parity.device_number[0] = total_disks - 1;
parity.device_number[2] = -1;
return parity;
}
int nbdx_register_block_device(struct nbdx_file *nbdx_file)
{
sector_t size = nbdx_file->stbuf.st_size;
int page_size = PAGE_SIZE;
int err = 0;
pr_debug("%s called\n", __func__);
nbdx_file->major = nbdx_major;
#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 16, 0)
nbdx_mq_reg.nr_hw_queues = submit_queues;
nbdx_file->queue = blk_mq_init_queue(&nbdx_mq_reg, nbdx_file);
#else
nbdx_file->tag_set.ops = &nbdx_mq_ops;
nbdx_file->tag_set.nr_hw_queues = submit_queues;
nbdx_file->tag_set.queue_depth = NBDX_QUEUE_DEPTH;
nbdx_file->tag_set.numa_node = NUMA_NO_NODE;
nbdx_file->tag_set.cmd_size = sizeof(struct raio_io_u);
nbdx_file->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
nbdx_file->tag_set.driver_data = nbdx_file;
err = blk_mq_alloc_tag_set(&nbdx_file->tag_set);
if (err)
goto out;
nbdx_file->queue = blk_mq_init_queue(&nbdx_file->tag_set);
#endif
if (IS_ERR(nbdx_file->queue)) {
pr_err("%s: Failed to allocate blk queue ret=%ld\n",
__func__, PTR_ERR(nbdx_file->queue));
err = PTR_ERR(nbdx_file->queue);
goto blk_mq_init;
}
nbdx_file->queue->queuedata = nbdx_file;
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nbdx_file->queue);
queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, nbdx_file->queue);
nbdx_file->disk = alloc_disk_node(1, NUMA_NO_NODE);
if (!nbdx_file->disk) {
pr_err("%s: Failed to allocate disk node\n", __func__);
err = -ENOMEM;
goto alloc_disk;
}
nbdx_file->disk->major = nbdx_file->major;
nbdx_file->disk->first_minor = nbdx_file->index;
nbdx_file->disk->fops = &nbdx_ops;
nbdx_file->disk->queue = nbdx_file->queue;
nbdx_file->disk->private_data = nbdx_file;
blk_queue_logical_block_size(nbdx_file->queue, NBDX_SECT_SIZE);
blk_queue_physical_block_size(nbdx_file->queue, NBDX_SECT_SIZE);
sector_div(page_size, NBDX_SECT_SIZE);
blk_queue_max_hw_sectors(nbdx_file->queue, page_size * MAX_SGL_LEN);
sector_div(size, NBDX_SECT_SIZE);
set_capacity(nbdx_file->disk, size);
sscanf(nbdx_file->dev_name, "%s", nbdx_file->disk->disk_name);
add_disk(nbdx_file->disk);
goto out;
alloc_disk:
blk_cleanup_queue(nbdx_file->queue);
blk_mq_init:
#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0)
blk_mq_free_tag_set(&nbdx_file->tag_set);
#endif
out:
return err;
}
static dm_oblock_t to_hblock(struct smq_policy *mq, dm_oblock_t b)
{
sector_t r = from_oblock(b);
(void) sector_div(r, mq->cache_blocks_per_hotspot_block);
return to_oblock(r);
}
static int null_add_dev(void)
{
struct gendisk *disk;
struct nullb *nullb;
sector_t size;
nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
if (!nullb)
return -ENOMEM;
spin_lock_init(&nullb->lock);
if (setup_queues(nullb))
goto err;
if (queue_mode == NULL_Q_MQ) {
null_mq_reg.numa_node = home_node;
null_mq_reg.queue_depth = hw_queue_depth;
if (use_per_node_hctx) {
null_mq_reg.ops->alloc_hctx = null_alloc_hctx;
null_mq_reg.ops->free_hctx = null_free_hctx;
null_mq_reg.nr_hw_queues = nr_online_nodes;
} else {
null_mq_reg.ops->alloc_hctx = blk_mq_alloc_single_hw_queue;
null_mq_reg.ops->free_hctx = blk_mq_free_single_hw_queue;
null_mq_reg.nr_hw_queues = submit_queues;
}
nullb->q = blk_mq_init_queue(&null_mq_reg, nullb);
} else if (queue_mode == NULL_Q_BIO) {
nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
blk_queue_make_request(nullb->q, null_queue_bio);
} else {
nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
if (nullb->q)
blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
}
if (!nullb->q)
goto queue_fail;
nullb->q->queuedata = nullb;
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
disk = nullb->disk = alloc_disk_node(1, home_node);
if (!disk) {
queue_fail:
if (queue_mode == NULL_Q_MQ)
blk_mq_free_queue(nullb->q);
else
blk_cleanup_queue(nullb->q);
cleanup_queues(nullb);
err:
kfree(nullb);
return -ENOMEM;
}
mutex_lock(&lock);
list_add_tail(&nullb->list, &nullb_list);
nullb->index = nullb_indexes++;
mutex_unlock(&lock);
blk_queue_logical_block_size(nullb->q, bs);
blk_queue_physical_block_size(nullb->q, bs);
size = gb * 1024 * 1024 * 1024ULL;
sector_div(size, bs);
set_capacity(disk, size);
disk->flags |= GENHD_FL_EXT_DEVT;
disk->major = null_major;
disk->first_minor = nullb->index;
disk->fops = &null_fops;
disk->private_data = nullb;
disk->queue = nullb->q;
sprintf(disk->disk_name, "nullb%d", nullb->index);
add_disk(disk);
return 0;
}
/*
* Algorithm of RAID6E with degraded drive
*/
static struct parity_places algorithm_raid6e( struct insane_c *ctx, u64 block, sector_t *sector, int *device_number)
{
struct parity_places parity;
struct block_place degraded_place;
u64 i, Y;
u64 position;
u64 local_gap;
u64 data_block;
u64 lane;
u64 block_offset, block_start;
int block_size;
int total_disks;
sector_t device_length;
block_size = ctx->chunk_size;
total_disks = raid6e_alg.ndisks;
device_length = ctx->ti->len;
data_block = *device_number + block * total_disks;
lane = data_block;
// NORMAL SITUATION
// Everything like in RAID 6
position = sector_div(lane, total_disks - raid6e_alg.p_blocks);
i = lane;
Y = sector_div(i, total_disks);
local_gap = 2;
// If we are in last stripe in square then we skip 1 syndrome in current lane
if (Y == (total_disks - 1))
local_gap = 1;
// If we didn't cross square diagonal then we don't skip syndromes in
// current lane
if (position + Y < (total_disks - raid6e_alg.p_blocks))
local_gap = 0;
// Remap block accounting all gaps
position = data_block + local_gap + (raid6e_alg.p_blocks * lane);
// Remap device_number
*device_number = sector_div(position, total_disks);
// For sequential writing: let's check number of current block
parity.last_block = false;
if (ctx->io_pattern == SEQUENTIAL)
{
if (*device_number + (2 - local_gap) == (total_disks - 1))
parity.last_block = true;
}
// Get offset in block and remap sector
block_offset = sector_div(*sector, block_size);
block_start = position * block_size;
*sector = position * block_size + i;
// Now it's time to count, where our syndromes are
parity.start_device = 0;
parity.start_sector = block_start;
parity.sector_number[0] = block_start;
parity.sector_number[1] = block_start;
parity.device_number[1] = total_disks - 1 - Y;
if (Y < total_disks - 1)
parity.device_number[0] = parity.device_number[1] - 1;
else
parity.device_number[0] = total_disks - 1;
if (*device_number == DEGRADED_DISK) {
degraded_place = get_degraded_block(position, total_disks, block_size, device_length);
*device_number = degraded_place.device_number;
*sector = degraded_place.sector + i;
}
else if (parity.device_number[0] == DEGRADED_DISK) {
degraded_place = get_degraded_block(position, total_disks, block_size, device_length);
parity.device_number[0] = degraded_place.device_number;
parity.sector_number[0] = degraded_place.sector;
}
else if (parity.device_number[1] == DEGRADED_DISK) {
degraded_place = get_degraded_block(position, total_disks, block_size, device_length);
parity.device_number[1] = degraded_place.device_number;
parity.sector_number[1] = degraded_place.sector;
}
parity.device_number[2] = -1;
return parity;
}
/*
* Construct a striped mapping.
* <number of stripes> <chunk size> [<dev_path> <offset>]+
*/
static int vm_ctr(struct dm_target *ti, unsigned int argc, char **argv)
{
struct vm_c *vc;
sector_t width, tmp_len;
uint32_t vms;
uint32_t chunk_size;
int r;
unsigned long long i;
if (argc < 2) {
ti->error = "Not enough arguments";
return -EINVAL;
}
if (kstrtouint(argv[0], 10, &vms) || !vms) {
ti->error = "Invalid stripe count";
return -EINVAL;
}
if (kstrtouint(argv[1], 10, &chunk_size) || !chunk_size) {
ti->error = "Invalid chunk_size";
return -EINVAL;
}
width = ti->len;
if (sector_div(width, vms)) {
ti->error = "Target length not divisible by "
"number of stripes";
return -EINVAL;
}
tmp_len = width;
if (sector_div(tmp_len, chunk_size)) {
ti->error = "Target length not divisible by "
"chunk size";
return -EINVAL;
}
/*
* Do we have enough arguments for that many stripes ?
*/
if (argc != (2 + 2 * vms)) {
ti->error = "Not enough destinations "
"specified";
return -EINVAL;
}
vc = alloc_context(vms);
if (!vc) {
ti->error = "Memory allocation for striped context "
"failed";
return -ENOMEM;
}
INIT_WORK(&vc->trigger_event, trigger_event);
/* Set pointer to dm target; used in trigger_event */
vc->ti = ti;
vc->vms = vms;
vc->vm_width = width;
if (vms & (vms - 1))
vc->vms_shift = -1;
else
vc->vms_shift = __ffs(vms);
r = dm_set_target_max_io_len(ti, chunk_size);
if (r) {
kfree(vc);
return r;
}
ti->num_flush_bios = vms;
ti->num_discard_bios = vms;
ti->num_write_same_bios = vms;
vc->chunk_size = chunk_size;
if (chunk_size & (chunk_size - 1))
vc->chunk_size_shift = -1;
else
vc->chunk_size_shift = __ffs(chunk_size);
/*
* Get the stripe destinations.
*/
for (i = 0; i < vms; i++) {
argv += 2;
r = get_vm(ti, vc, i, argv);
if (r < 0) {
ti->error = "Couldn't parse stripe destination";
while (i--)
dm_put_device(ti, vc->vm[i].dev);
kfree(vc);
return r;
}
atomic_set(&(vc->vm[i].error_count), 0);
}
/*volume manager initialize*/
vc->wp = 0;//////current 0 is NVMe
//vc->wp = 1;
vc->ws = kmalloc(sizeof(unsigned long long) * vc->vms, GFP_KERNEL);
for(i = 0; i<vc->vms; i++)
vc->ws[i] = 0;
vc->gp_list = kmalloc(sizeof(char) * vc->vms, GFP_KERNEL);
vc->num_gp = 0;
vc->io_client = dm_io_client_create();
vc->gs = NULL;
vc->overload = 0;
for(i=0; i<vc->vms; i++)
vc->gp_list[i] = Clean_Weight;//0 is clean
{
unsigned long long tem, disk_size;
tem = 0;
for(i = 0; i<vms; i++){
struct block_device *cur_bdev = vc->vm[i].dev->bdev;
vc->vm[i].end_sector = i_size_read(cur_bdev->bd_inode)>>9;//unit of sector
printk("vm%llu start_sector %llu, end_sector %llu, target_offset %llu\n",
i, (unsigned long long) vc->vm[i].physical_start, (unsigned long long) vc->vm[i].end_sector, (unsigned long long)dm_target_offset(ti, vc->ws[i]));
disk_size = vc->vm[i].end_sector * 512;
do_div(disk_size, (unsigned long long) vc->vm[i].dev->bdev->bd_block_size);
tem += disk_size;
}
vc->num_entry = tem;//num entry is blk num
}
printk("num entry is %llu, node size is %lu, req mem is %llu\n", vc->num_entry, sizeof(struct flag_nodes), sizeof(struct flag_nodes) * vc->num_entry);
//flag set initialize
vc->fs = (struct flag_set *) kmalloc(sizeof(struct flag_set), GFP_KERNEL);
vc->fs->node_buf = kmem_cache_create("dirty_data_buf", sizeof(struct flag_nodes),
0, (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD), NULL);
vc->fs->table = (struct flag_nodes **)vmalloc(sizeof(struct flag_nodes*) * vc->num_entry);
for(i=0; i<vc->num_entry; i++){
//vc->fs->table[i] = NULL;//late alloc code
vc->fs->table[i] = kmem_cache_alloc(vc->fs->node_buf, GFP_KERNEL);//pre alloc start
vc->fs->table[i]->msector = -1;
vc->fs->table[i]->wp = -1;//pre alloc end
}
vc->num_map_block = 0;//vc->num_entry * sizeof(struct flag_nodes) / 4096;
//vc->ws[0] += vc->num_map_block;
vc->fs->reverse_table = vmalloc(sizeof(struct reverse_nodes*) * vc->vms);
vc->d_num = kmalloc(sizeof(unsigned long long) * vc->vms, GFP_KERNEL);
for(i=0; i<vc->vms; i++){
unsigned long long j;
unsigned long long r_table_size = (vc->vm[i].end_sector + 7);
unsigned long long phy_sect = vc->vm[i].physical_start;
do_div(phy_sect, 8);
do_div(r_table_size, 8);
printk("r_table_size = %llu\n", r_table_size);
vc->vm[i].num_dirty = r_table_size - phy_sect;
vc->d_num[i] = vc->vm[i].num_dirty;
vc->fs->reverse_table[i] = vmalloc(sizeof(struct reverse_nodes) * r_table_size);
for(j=0; j<r_table_size; j++){
vc->fs->reverse_table[i][j].index = -1;
vc->fs->reverse_table[i][j].dirty = 1;
vc->fs->reverse_table[i][j].size = -1;
}
//printk("%u's first ptr is %p, final ptr is %p\n", i, &(vc->fs->reverse_table[i][0]), &(vc->fs->reverse_table[i][j]));
}
for(i=0; i<vc->vms; i++){
unsigned int minor = atom(vc->vm[i].dev->name);
unsigned int major = atoj(vc->vm[i].dev->name);
printk("dev name is %s\t", vc->vm[i].dev->name);
if(major != 2600) vc->vm[i].main_dev = minor >> minor_shift;
else vc->vm[i].main_dev = minor - 1;
vc->vm[i].maj_dev = major;
printk("main %u, maj %u\n", vc->vm[i].main_dev, vc->vm[i].maj_dev);
}