static int __init iostash_init(void)
{
int ret = -1, i;
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
ERR("Kernel version < 2.6.28 not supported.");
return -ENOENT;
#endif
DBG("++ iostash_init() ++\n");
memset(&gctx, 0, sizeof(gctx));
do {
gctx.io_pool = KMEM_CACHE(iostash_bio, 0);
if (!gctx.io_pool) {
ERR("iostash_init: KMEM_CACHE() failed\n");
break;
}
gctx.io_client = dm_io_client_create();
if (IS_ERR(gctx.io_client)) {
ERR("iostash_init: dm_io_client() failed\n");
break;
}
gctx.sce = sce_create();
if (!gctx.sce) {
ERR("iostash_init: sce_create() failed\n");
break;
}
gctx.pdm = pdm_create(gctx.sce, poptask_read, poptask_write);
if (_init_iostash_kobjects()) {
ERR("KOBJECT INIT FAILED!");
_destroy_iostash_kobjects();
}
mutex_init(&gctx.ctl_mtx);
for (i = 0; i < IOSTASH_MAXHDD_BCKTS; ++i)
INIT_LIST_HEAD(&gctx.hddtbl.bucket[i]);
for (i = 0; i < IOSTASH_MAXSSD_BCKTS; ++i)
INIT_LIST_HEAD(&gctx.ssdtbl.bucket[i]);
ret = 0;
} while (0);
if (ret) {
_free_resource();
}
DBG("-- iostash_init() returns = %d --\n", ret);
return ret;
}
static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
unsigned int argc, char **argv,
struct dm_dev *dev)
{
enum sync sync = DEFAULTSYNC;
struct log_c *lc;
uint32_t region_size;
unsigned int region_count;
size_t bitset_size, buf_size;
int r;
char dummy;
if (argc < 1 || argc > 2) {
DMWARN("wrong number of arguments to dirty region log");
return -EINVAL;
}
if (argc > 1) {
if (!strcmp(argv[1], "sync"))
sync = FORCESYNC;
else if (!strcmp(argv[1], "nosync"))
sync = NOSYNC;
else {
DMWARN("unrecognised sync argument to "
"dirty region log: %s", argv[1]);
return -EINVAL;
}
}
if (sscanf(argv[0], "%u%c", ®ion_size, &dummy) != 1 ||
!_check_region_size(ti, region_size)) {
DMWARN("invalid region size %s", argv[0]);
return -EINVAL;
}
region_count = dm_sector_div_up(ti->len, region_size);
lc = kmalloc(sizeof(*lc), GFP_KERNEL);
if (!lc) {
DMWARN("couldn't allocate core log");
return -ENOMEM;
}
lc->ti = ti;
lc->touched_dirtied = 0;
lc->touched_cleaned = 0;
lc->flush_failed = 0;
lc->region_size = region_size;
lc->region_count = region_count;
lc->sync = sync;
/*
* Work out how many "unsigned long"s we need to hold the bitset.
*/
bitset_size = dm_round_up(region_count,
sizeof(*lc->clean_bits) << BYTE_SHIFT);
bitset_size >>= BYTE_SHIFT;
lc->bitset_uint32_count = bitset_size / sizeof(*lc->clean_bits);
/*
* Disk log?
*/
if (!dev) {
lc->clean_bits = vmalloc(bitset_size);
if (!lc->clean_bits) {
DMWARN("couldn't allocate clean bitset");
kfree(lc);
return -ENOMEM;
}
lc->disk_header = NULL;
} else {
lc->log_dev = dev;
lc->log_dev_failed = 0;
lc->log_dev_flush_failed = 0;
lc->header_location.bdev = lc->log_dev->bdev;
lc->header_location.sector = 0;
/*
* Buffer holds both header and bitset.
*/
buf_size =
dm_round_up((LOG_OFFSET << SECTOR_SHIFT) + bitset_size,
bdev_logical_block_size(lc->header_location.
bdev));
if (buf_size > i_size_read(dev->bdev->bd_inode)) {
DMWARN("log device %s too small: need %llu bytes",
dev->name, (unsigned long long)buf_size);
kfree(lc);
return -EINVAL;
}
lc->header_location.count = buf_size >> SECTOR_SHIFT;
lc->io_req.mem.type = DM_IO_VMA;
lc->io_req.notify.fn = NULL;
lc->io_req.client = dm_io_client_create();
if (IS_ERR(lc->io_req.client)) {
r = PTR_ERR(lc->io_req.client);
DMWARN("couldn't allocate disk io client");
kfree(lc);
return r;
}
lc->disk_header = vmalloc(buf_size);
if (!lc->disk_header) {
DMWARN("couldn't allocate disk log buffer");
dm_io_client_destroy(lc->io_req.client);
kfree(lc);
return -ENOMEM;
}
lc->io_req.mem.ptr.vma = lc->disk_header;
lc->clean_bits = (void *)lc->disk_header +
(LOG_OFFSET << SECTOR_SHIFT);
}
memset(lc->clean_bits, -1, bitset_size);
lc->sync_bits = vmalloc(bitset_size);
if (!lc->sync_bits) {
DMWARN("couldn't allocate sync bitset");
if (!dev)
vfree(lc->clean_bits);
else
dm_io_client_destroy(lc->io_req.client);
vfree(lc->disk_header);
kfree(lc);
return -ENOMEM;
}
memset(lc->sync_bits, (sync == NOSYNC) ? -1 : 0, bitset_size);
lc->sync_count = (sync == NOSYNC) ? region_count : 0;
lc->recovering_bits = vzalloc(bitset_size);
if (!lc->recovering_bits) {
DMWARN("couldn't allocate sync bitset");
vfree(lc->sync_bits);
if (!dev)
vfree(lc->clean_bits);
else
dm_io_client_destroy(lc->io_req.client);
vfree(lc->disk_header);
kfree(lc);
return -ENOMEM;
}
lc->sync_search = 0;
log->context = lc;
return 0;
}
/*----------------------------------------------------------------
* disk log constructor/destructor
*
* argv contains 2 - 4 arguments:
* <log_device> <region_size> [[no]sync] [block_on_error]
*--------------------------------------------------------------*/
static int disk_ctr(struct dirty_log *log, struct dm_target *ti,
unsigned int argc, char **argv)
{
int r;
size_t size, bitset_size;
struct log_c *lc;
struct dm_dev *dev;
uint32_t *clean_bits;
if (argc < 2 || argc > 4) {
DMWARN("wrong number of arguments to disk mirror log");
return -EINVAL;
}
r = dm_get_device(ti, argv[0], 0, 0 /* FIXME */,
FMODE_READ | FMODE_WRITE, &dev);
if (r)
return r;
r = core_ctr(log, ti, argc - 1, argv + 1);
if (r) {
dm_put_device(ti, dev);
return r;
}
lc = (struct log_c *) log->context;
lc->log_dev = dev;
lc->log_dev_failed = 0;
/* setup the disk header fields */
lc->header_location.bdev = lc->log_dev->bdev;
lc->header_location.sector = 0;
/* Include both the header and the bitset in one buffer. */
bitset_size = lc->bitset_uint32_count * sizeof(uint32_t);
size = dm_round_up((LOG_OFFSET << SECTOR_SHIFT) + bitset_size,
ti->limits.hardsect_size);
if (size > dev->bdev->bd_inode->i_size) {
DMWARN("log device %s too small: need %llu bytes",
dev->name, (unsigned long long)size);
r = -EINVAL;
goto bad;
}
lc->header_location.count = size >> SECTOR_SHIFT;
lc->disk_header = vmalloc(size);
if (!lc->disk_header) {
r = -ENOMEM;
goto bad;
}
/*
* Deallocate the clean_bits buffer that was allocated in core_ctr()
* and point it at the appropriate place in the disk_header buffer.
*/
clean_bits = lc->clean_bits;
lc->clean_bits = (void *)lc->disk_header + (LOG_OFFSET << SECTOR_SHIFT);
memcpy(lc->clean_bits, clean_bits, bitset_size);
vfree(clean_bits);
lc->io_req.mem.type = DM_IO_VMA;
lc->io_req.client = dm_io_client_create(dm_div_up(size, PAGE_SIZE));
if (IS_ERR(lc->io_req.client)) {
r = PTR_ERR(lc->io_req.client);
DMWARN("couldn't allocate disk io client");
vfree(lc->disk_header);
goto bad;
}
return 0;
bad:
dm_put_device(ti, lc->log_dev);
core_dtr(log);
return r;
}
/*
* 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);
}
