nullb: factor disk parameters
authorShaohua Li <shli@fb.com>
Mon, 14 Aug 2017 22:04:52 +0000 (15:04 -0700)
committerJens Axboe <axboe@kernel.dk>
Wed, 23 Aug 2017 14:54:00 +0000 (08:54 -0600)
When we switch to configfs interface, each disk could have different
configuration. To prepare for the change, we move most disk setting to a
separate data structure. The existing module parameter interface is
kept. The 'nr_devices' and 'shared_tags' don't make sense for per-disk
setting, so they are remained as global settings.

Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/block/null_blk.c

index f1d0ca02099955d35417df4b42855ee1345c4f2b..73938cde11f3d1a9ee0bfa668d92506ea4e2291a 100644 (file)
@@ -25,11 +25,29 @@ struct nullb_queue {
        unsigned long *tag_map;
        wait_queue_head_t wait;
        unsigned int queue_depth;
+       struct nullb_device *dev;
 
        struct nullb_cmd *cmds;
 };
 
+struct nullb_device {
+       struct nullb *nullb;
+
+       unsigned long size; /* device size in MB */
+       unsigned long completion_nsec; /* time in ns to complete a request */
+       unsigned int submit_queues; /* number of submission queues */
+       unsigned int home_node; /* home node for the device */
+       unsigned int queue_mode; /* block interface */
+       unsigned int blocksize; /* block size */
+       unsigned int irqmode; /* IRQ completion handler */
+       unsigned int hw_queue_depth; /* queue depth */
+       bool use_lightnvm; /* register as a LightNVM device */
+       bool blocking; /* blocking blk-mq device */
+       bool use_per_node_hctx; /* use per-node allocation for hardware context */
+};
+
 struct nullb {
+       struct nullb_device *dev;
        struct list_head list;
        unsigned int index;
        struct request_queue *q;
@@ -65,15 +83,15 @@ enum {
        NULL_Q_MQ               = 2,
 };
 
-static int submit_queues = 1;
-module_param(submit_queues, int, S_IRUGO);
+static int g_submit_queues = 1;
+module_param_named(submit_queues, g_submit_queues, int, S_IRUGO);
 MODULE_PARM_DESC(submit_queues, "Number of submission queues");
 
-static int home_node = NUMA_NO_NODE;
-module_param(home_node, int, S_IRUGO);
+static int g_home_node = NUMA_NO_NODE;
+module_param_named(home_node, g_home_node, int, S_IRUGO);
 MODULE_PARM_DESC(home_node, "Home node for the device");
 
-static int queue_mode = NULL_Q_MQ;
+static int g_queue_mode = NULL_Q_MQ;
 
 static int null_param_store_val(const char *str, int *val, int min, int max)
 {
@@ -92,7 +110,7 @@ static int null_param_store_val(const char *str, int *val, int min, int max)
 
 static int null_set_queue_mode(const char *str, const struct kernel_param *kp)
 {
-       return null_param_store_val(str, &queue_mode, NULL_Q_BIO, NULL_Q_MQ);
+       return null_param_store_val(str, &g_queue_mode, NULL_Q_BIO, NULL_Q_MQ);
 }
 
 static const struct kernel_param_ops null_queue_mode_param_ops = {
@@ -100,38 +118,38 @@ static const struct kernel_param_ops null_queue_mode_param_ops = {
        .get    = param_get_int,
 };
 
-device_param_cb(queue_mode, &null_queue_mode_param_ops, &queue_mode, S_IRUGO);
+device_param_cb(queue_mode, &null_queue_mode_param_ops, &g_queue_mode, S_IRUGO);
 MODULE_PARM_DESC(queue_mode, "Block interface to use (0=bio,1=rq,2=multiqueue)");
 
-static int gb = 250;
-module_param(gb, int, S_IRUGO);
+static int g_gb = 250;
+module_param_named(gb, g_gb, int, S_IRUGO);
 MODULE_PARM_DESC(gb, "Size in GB");
 
-static int bs = 512;
-module_param(bs, int, S_IRUGO);
+static int g_bs = 512;
+module_param_named(bs, g_bs, int, S_IRUGO);
 MODULE_PARM_DESC(bs, "Block size (in bytes)");
 
 static int nr_devices = 1;
 module_param(nr_devices, int, S_IRUGO);
 MODULE_PARM_DESC(nr_devices, "Number of devices to register");
 
-static bool use_lightnvm;
-module_param(use_lightnvm, bool, S_IRUGO);
+static bool g_use_lightnvm;
+module_param_named(use_lightnvm, g_use_lightnvm, bool, S_IRUGO);
 MODULE_PARM_DESC(use_lightnvm, "Register as a LightNVM device");
 
-static bool blocking;
-module_param(blocking, bool, S_IRUGO);
+static bool g_blocking;
+module_param_named(blocking, g_blocking, bool, S_IRUGO);
 MODULE_PARM_DESC(blocking, "Register as a blocking blk-mq driver device");
 
 static bool shared_tags;
 module_param(shared_tags, bool, S_IRUGO);
 MODULE_PARM_DESC(shared_tags, "Share tag set between devices for blk-mq");
 
-static int irqmode = NULL_IRQ_SOFTIRQ;
+static int g_irqmode = NULL_IRQ_SOFTIRQ;
 
 static int null_set_irqmode(const char *str, const struct kernel_param *kp)
 {
-       return null_param_store_val(str, &irqmode, NULL_IRQ_NONE,
+       return null_param_store_val(str, &g_irqmode, NULL_IRQ_NONE,
                                        NULL_IRQ_TIMER);
 }
 
@@ -140,21 +158,47 @@ static const struct kernel_param_ops null_irqmode_param_ops = {
        .get    = param_get_int,
 };
 
-device_param_cb(irqmode, &null_irqmode_param_ops, &irqmode, S_IRUGO);
+device_param_cb(irqmode, &null_irqmode_param_ops, &g_irqmode, S_IRUGO);
 MODULE_PARM_DESC(irqmode, "IRQ completion handler. 0-none, 1-softirq, 2-timer");
 
-static unsigned long completion_nsec = 10000;
-module_param(completion_nsec, ulong, S_IRUGO);
+static unsigned long g_completion_nsec = 10000;
+module_param_named(completion_nsec, g_completion_nsec, ulong, S_IRUGO);
 MODULE_PARM_DESC(completion_nsec, "Time in ns to complete a request in hardware. Default: 10,000ns");
 
-static int hw_queue_depth = 64;
-module_param(hw_queue_depth, int, S_IRUGO);
+static int g_hw_queue_depth = 64;
+module_param_named(hw_queue_depth, g_hw_queue_depth, int, S_IRUGO);
 MODULE_PARM_DESC(hw_queue_depth, "Queue depth for each hardware queue. Default: 64");
 
-static bool use_per_node_hctx = false;
-module_param(use_per_node_hctx, bool, S_IRUGO);
+static bool g_use_per_node_hctx;
+module_param_named(use_per_node_hctx, g_use_per_node_hctx, bool, S_IRUGO);
 MODULE_PARM_DESC(use_per_node_hctx, "Use per-node allocation for hardware context queues. Default: false");
 
+static struct nullb_device *null_alloc_dev(void)
+{
+       struct nullb_device *dev;
+
+       dev = kzalloc(sizeof(*dev), GFP_KERNEL);
+       if (!dev)
+               return NULL;
+       dev->size = g_gb * 1024;
+       dev->completion_nsec = g_completion_nsec;
+       dev->submit_queues = g_submit_queues;
+       dev->home_node = g_home_node;
+       dev->queue_mode = g_queue_mode;
+       dev->blocksize = g_bs;
+       dev->irqmode = g_irqmode;
+       dev->hw_queue_depth = g_hw_queue_depth;
+       dev->use_lightnvm = g_use_lightnvm;
+       dev->blocking = g_blocking;
+       dev->use_per_node_hctx = g_use_per_node_hctx;
+       return dev;
+}
+
+static void null_free_dev(struct nullb_device *dev)
+{
+       kfree(dev);
+}
+
 static void put_tag(struct nullb_queue *nq, unsigned int tag)
 {
        clear_bit_unlock(tag, nq->tag_map);
@@ -193,7 +237,7 @@ static struct nullb_cmd *__alloc_cmd(struct nullb_queue *nq)
                cmd = &nq->cmds[tag];
                cmd->tag = tag;
                cmd->nq = nq;
-               if (irqmode == NULL_IRQ_TIMER) {
+               if (nq->dev->irqmode == NULL_IRQ_TIMER) {
                        hrtimer_init(&cmd->timer, CLOCK_MONOTONIC,
                                     HRTIMER_MODE_REL);
                        cmd->timer.function = null_cmd_timer_expired;
@@ -229,6 +273,7 @@ static struct nullb_cmd *alloc_cmd(struct nullb_queue *nq, int can_wait)
 static void end_cmd(struct nullb_cmd *cmd)
 {
        struct request_queue *q = NULL;
+       int queue_mode = cmd->nq->dev->queue_mode;
 
        if (cmd->rq)
                q = cmd->rq->q;
@@ -267,14 +312,16 @@ static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
 
 static void null_cmd_end_timer(struct nullb_cmd *cmd)
 {
-       ktime_t kt = completion_nsec;
+       ktime_t kt = cmd->nq->dev->completion_nsec;
 
        hrtimer_start(&cmd->timer, kt, HRTIMER_MODE_REL);
 }
 
 static void null_softirq_done_fn(struct request *rq)
 {
-       if (queue_mode == NULL_Q_MQ)
+       struct nullb *nullb = rq->q->queuedata;
+
+       if (nullb->dev->queue_mode == NULL_Q_MQ)
                end_cmd(blk_mq_rq_to_pdu(rq));
        else
                end_cmd(rq->special);
@@ -283,9 +330,9 @@ static void null_softirq_done_fn(struct request *rq)
 static inline void null_handle_cmd(struct nullb_cmd *cmd)
 {
        /* Complete IO by inline, softirq or timer */
-       switch (irqmode) {
+       switch (cmd->nq->dev->irqmode) {
        case NULL_IRQ_SOFTIRQ:
-               switch (queue_mode)  {
+               switch (cmd->nq->dev->queue_mode)  {
                case NULL_Q_MQ:
                        blk_mq_complete_request(cmd->rq);
                        break;
@@ -366,15 +413,16 @@ static blk_status_t null_queue_rq(struct blk_mq_hw_ctx *hctx,
                         const struct blk_mq_queue_data *bd)
 {
        struct nullb_cmd *cmd = blk_mq_rq_to_pdu(bd->rq);
+       struct nullb_queue *nq = hctx->driver_data;
 
        might_sleep_if(hctx->flags & BLK_MQ_F_BLOCKING);
 
-       if (irqmode == NULL_IRQ_TIMER) {
+       if (nq->dev->irqmode == NULL_IRQ_TIMER) {
                hrtimer_init(&cmd->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
                cmd->timer.function = null_cmd_timer_expired;
        }
        cmd->rq = bd->rq;
-       cmd->nq = hctx->driver_data;
+       cmd->nq = nq;
 
        blk_mq_start_request(bd->rq);
 
@@ -438,7 +486,8 @@ static int null_lnvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
 
 static int null_lnvm_id(struct nvm_dev *dev, struct nvm_id *id)
 {
-       sector_t size = gb * 1024 * 1024 * 1024ULL;
+       struct nullb *nullb = dev->q->queuedata;
+       sector_t size = (sector_t)nullb->dev->size * 1024 * 1024ULL;
        sector_t blksize;
        struct nvm_id_group *grp;
 
@@ -460,7 +509,7 @@ static int null_lnvm_id(struct nvm_dev *dev, struct nvm_id *id)
        id->ppaf.ch_offset = 56;
        id->ppaf.ch_len = 8;
 
-       sector_div(size, bs); /* convert size to pages */
+       sector_div(size, nullb->dev->blocksize); /* convert size to pages */
        size >>= 8; /* concert size to pgs pr blk */
        grp = &id->grp;
        grp->mtype = 0;
@@ -474,8 +523,8 @@ static int null_lnvm_id(struct nvm_dev *dev, struct nvm_id *id)
        grp->num_blk = blksize;
        grp->num_pln = 1;
 
-       grp->fpg_sz = bs;
-       grp->csecs = bs;
+       grp->fpg_sz = nullb->dev->blocksize;
+       grp->csecs = nullb->dev->blocksize;
        grp->trdt = 25000;
        grp->trdm = 25000;
        grp->tprt = 500000;
@@ -483,7 +532,7 @@ static int null_lnvm_id(struct nvm_dev *dev, struct nvm_id *id)
        grp->tbet = 1500000;
        grp->tbem = 1500000;
        grp->mpos = 0x010101; /* single plane rwe */
-       grp->cpar = hw_queue_depth;
+       grp->cpar = nullb->dev->hw_queue_depth;
 
        return 0;
 }
@@ -568,19 +617,23 @@ static void null_nvm_unregister(struct nullb *nullb) {}
 
 static void null_del_dev(struct nullb *nullb)
 {
+       struct nullb_device *dev = nullb->dev;
+
        list_del_init(&nullb->list);
 
-       if (use_lightnvm)
+       if (dev->use_lightnvm)
                null_nvm_unregister(nullb);
        else
                del_gendisk(nullb->disk);
        blk_cleanup_queue(nullb->q);
-       if (queue_mode == NULL_Q_MQ && nullb->tag_set == &nullb->__tag_set)
+       if (dev->queue_mode == NULL_Q_MQ &&
+           nullb->tag_set == &nullb->__tag_set)
                blk_mq_free_tag_set(nullb->tag_set);
-       if (!use_lightnvm)
+       if (!dev->use_lightnvm)
                put_disk(nullb->disk);
        cleanup_queues(nullb);
        kfree(nullb);
+       dev->nullb = NULL;
 }
 
 static int null_open(struct block_device *bdev, fmode_t mode)
@@ -605,6 +658,7 @@ static void null_init_queue(struct nullb *nullb, struct nullb_queue *nq)
 
        init_waitqueue_head(&nq->wait);
        nq->queue_depth = nullb->queue_depth;
+       nq->dev = nullb->dev;
 }
 
 static void null_init_queues(struct nullb *nullb)
@@ -652,13 +706,13 @@ static int setup_commands(struct nullb_queue *nq)
 
 static int setup_queues(struct nullb *nullb)
 {
-       nullb->queues = kzalloc(submit_queues * sizeof(struct nullb_queue),
-                                                               GFP_KERNEL);
+       nullb->queues = kzalloc(nullb->dev->submit_queues *
+               sizeof(struct nullb_queue), GFP_KERNEL);
        if (!nullb->queues)
                return -ENOMEM;
 
        nullb->nr_queues = 0;
-       nullb->queue_depth = hw_queue_depth;
+       nullb->queue_depth = nullb->dev->hw_queue_depth;
 
        return 0;
 }
@@ -668,7 +722,7 @@ static int init_driver_queues(struct nullb *nullb)
        struct nullb_queue *nq;
        int i, ret = 0;
 
-       for (i = 0; i < submit_queues; i++) {
+       for (i = 0; i < nullb->dev->submit_queues; i++) {
                nq = &nullb->queues[i];
 
                null_init_queue(nullb, nq);
@@ -686,10 +740,10 @@ static int null_gendisk_register(struct nullb *nullb)
        struct gendisk *disk;
        sector_t size;
 
-       disk = nullb->disk = alloc_disk_node(1, home_node);
+       disk = nullb->disk = alloc_disk_node(1, nullb->dev->home_node);
        if (!disk)
                return -ENOMEM;
-       size = gb * 1024 * 1024 * 1024ULL;
+       size = (sector_t)nullb->dev->size * 1024 * 1024ULL;
        set_capacity(disk, size >> 9);
 
        disk->flags |= GENHD_FL_EXT_DEVT | GENHD_FL_SUPPRESS_PARTITION_INFO;
@@ -704,32 +758,36 @@ static int null_gendisk_register(struct nullb *nullb)
        return 0;
 }
 
-static int null_init_tag_set(struct blk_mq_tag_set *set)
+static int null_init_tag_set(struct nullb *nullb, struct blk_mq_tag_set *set)
 {
        set->ops = &null_mq_ops;
-       set->nr_hw_queues = submit_queues;
-       set->queue_depth = hw_queue_depth;
-       set->numa_node = home_node;
+       set->nr_hw_queues = nullb ? nullb->dev->submit_queues :
+                                               g_submit_queues;
+       set->queue_depth = nullb ? nullb->dev->hw_queue_depth :
+                                               g_hw_queue_depth;
+       set->numa_node = nullb ? nullb->dev->home_node : g_home_node;
        set->cmd_size   = sizeof(struct nullb_cmd);
        set->flags = BLK_MQ_F_SHOULD_MERGE;
        set->driver_data = NULL;
 
-       if (blocking)
+       if (nullb->dev->blocking)
                set->flags |= BLK_MQ_F_BLOCKING;
 
        return blk_mq_alloc_tag_set(set);
 }
 
-static int null_add_dev(void)
+static int null_add_dev(struct nullb_device *dev)
 {
        struct nullb *nullb;
        int rv;
 
-       nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
+       nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, dev->home_node);
        if (!nullb) {
                rv = -ENOMEM;
                goto out;
        }
+       nullb->dev = dev;
+       dev->nullb = nullb;
 
        spin_lock_init(&nullb->lock);
 
@@ -737,13 +795,13 @@ static int null_add_dev(void)
        if (rv)
                goto out_free_nullb;
 
-       if (queue_mode == NULL_Q_MQ) {
+       if (dev->queue_mode == NULL_Q_MQ) {
                if (shared_tags) {
                        nullb->tag_set = &tag_set;
                        rv = 0;
                } else {
                        nullb->tag_set = &nullb->__tag_set;
-                       rv = null_init_tag_set(nullb->tag_set);
+                       rv = null_init_tag_set(nullb, nullb->tag_set);
                }
 
                if (rv)
@@ -755,8 +813,8 @@ static int null_add_dev(void)
                        goto out_cleanup_tags;
                }
                null_init_queues(nullb);
-       } else if (queue_mode == NULL_Q_BIO) {
-               nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
+       } else if (dev->queue_mode == NULL_Q_BIO) {
+               nullb->q = blk_alloc_queue_node(GFP_KERNEL, dev->home_node);
                if (!nullb->q) {
                        rv = -ENOMEM;
                        goto out_cleanup_queues;
@@ -766,7 +824,8 @@ static int null_add_dev(void)
                if (rv)
                        goto out_cleanup_blk_queue;
        } else {
-               nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
+               nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock,
+                                               dev->home_node);
                if (!nullb->q) {
                        rv = -ENOMEM;
                        goto out_cleanup_queues;
@@ -786,12 +845,12 @@ static int null_add_dev(void)
        nullb->index = nullb_indexes++;
        mutex_unlock(&lock);
 
-       blk_queue_logical_block_size(nullb->q, bs);
-       blk_queue_physical_block_size(nullb->q, bs);
+       blk_queue_logical_block_size(nullb->q, dev->blocksize);
+       blk_queue_physical_block_size(nullb->q, dev->blocksize);
 
        sprintf(nullb->disk_name, "nullb%d", nullb->index);
 
-       if (use_lightnvm)
+       if (dev->use_lightnvm)
                rv = null_nvm_register(nullb);
        else
                rv = null_gendisk_register(nullb);
@@ -807,13 +866,14 @@ static int null_add_dev(void)
 out_cleanup_blk_queue:
        blk_cleanup_queue(nullb->q);
 out_cleanup_tags:
-       if (queue_mode == NULL_Q_MQ && nullb->tag_set == &nullb->__tag_set)
+       if (dev->queue_mode == NULL_Q_MQ && nullb->tag_set == &nullb->__tag_set)
                blk_mq_free_tag_set(nullb->tag_set);
 out_cleanup_queues:
        cleanup_queues(nullb);
 out_free_nullb:
        kfree(nullb);
 out:
+       null_free_dev(dev);
        return rv;
 }
 
@@ -822,38 +882,39 @@ static int __init null_init(void)
        int ret = 0;
        unsigned int i;
        struct nullb *nullb;
+       struct nullb_device *dev;
 
-       if (bs > PAGE_SIZE) {
+       if (g_bs > PAGE_SIZE) {
                pr_warn("null_blk: invalid block size\n");
                pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
-               bs = PAGE_SIZE;
+               g_bs = PAGE_SIZE;
        }
 
-       if (use_lightnvm && bs != 4096) {
+       if (g_use_lightnvm && g_bs != 4096) {
                pr_warn("null_blk: LightNVM only supports 4k block size\n");
                pr_warn("null_blk: defaults block size to 4k\n");
-               bs = 4096;
+               g_bs = 4096;
        }
 
-       if (use_lightnvm && queue_mode != NULL_Q_MQ) {
+       if (g_use_lightnvm && g_queue_mode != NULL_Q_MQ) {
                pr_warn("null_blk: LightNVM only supported for blk-mq\n");
                pr_warn("null_blk: defaults queue mode to blk-mq\n");
-               queue_mode = NULL_Q_MQ;
+               g_queue_mode = NULL_Q_MQ;
        }
 
-       if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
-               if (submit_queues != nr_online_nodes) {
+       if (g_queue_mode == NULL_Q_MQ && g_use_per_node_hctx) {
+               if (g_submit_queues != nr_online_nodes) {
                        pr_warn("null_blk: submit_queues param is set to %u.\n",
                                                        nr_online_nodes);
-                       submit_queues = nr_online_nodes;
+                       g_submit_queues = nr_online_nodes;
                }
-       } else if (submit_queues > nr_cpu_ids)
-               submit_queues = nr_cpu_ids;
-       else if (submit_queues <= 0)
-               submit_queues = 1;
+       } else if (g_submit_queues > nr_cpu_ids)
+               g_submit_queues = nr_cpu_ids;
+       else if (g_submit_queues <= 0)
+               g_submit_queues = 1;
 
-       if (queue_mode == NULL_Q_MQ && shared_tags) {
-               ret = null_init_tag_set(&tag_set);
+       if (g_queue_mode == NULL_Q_MQ && shared_tags) {
+               ret = null_init_tag_set(NULL, &tag_set);
                if (ret)
                        return ret;
        }
@@ -866,7 +927,7 @@ static int __init null_init(void)
                goto err_tagset;
        }
 
-       if (use_lightnvm) {
+       if (g_use_lightnvm) {
                ppa_cache = kmem_cache_create("ppa_cache", 64 * sizeof(u64),
                                                                0, 0, NULL);
                if (!ppa_cache) {
@@ -877,9 +938,14 @@ static int __init null_init(void)
        }
 
        for (i = 0; i < nr_devices; i++) {
-               ret = null_add_dev();
-               if (ret)
+               dev = null_alloc_dev();
+               if (!dev)
+                       goto err_dev;
+               ret = null_add_dev(dev);
+               if (ret) {
+                       null_free_dev(dev);
                        goto err_dev;
+               }
        }
 
        pr_info("null: module loaded\n");
@@ -888,13 +954,15 @@ static int __init null_init(void)
 err_dev:
        while (!list_empty(&nullb_list)) {
                nullb = list_entry(nullb_list.next, struct nullb, list);
+               dev = nullb->dev;
                null_del_dev(nullb);
+               null_free_dev(dev);
        }
        kmem_cache_destroy(ppa_cache);
 err_ppa:
        unregister_blkdev(null_major, "nullb");
 err_tagset:
-       if (queue_mode == NULL_Q_MQ && shared_tags)
+       if (g_queue_mode == NULL_Q_MQ && shared_tags)
                blk_mq_free_tag_set(&tag_set);
        return ret;
 }
@@ -907,12 +975,16 @@ static void __exit null_exit(void)
 
        mutex_lock(&lock);
        while (!list_empty(&nullb_list)) {
+               struct nullb_device *dev;
+
                nullb = list_entry(nullb_list.next, struct nullb, list);
+               dev = nullb->dev;
                null_del_dev(nullb);
+               null_free_dev(dev);
        }
        mutex_unlock(&lock);
 
-       if (queue_mode == NULL_Q_MQ && shared_tags)
+       if (g_queue_mode == NULL_Q_MQ && shared_tags)
                blk_mq_free_tag_set(&tag_set);
 
        kmem_cache_destroy(ppa_cache);