/*        $NetBSD: dm_target_delay.c,v 1.2 2020/01/21 16:27:53 tkusumi Exp $      */

/*
* Copyright (c) 2020 The NetBSD Foundation, Inc.
* Copyright (c) 2015 The DragonFly Project.  All rights reserved.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Tomohiro Kusumi <[email protected]>.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: dm_target_delay.c,v 1.2 2020/01/21 16:27:53 tkusumi Exp $");

#include <sys/types.h>
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/buf.h>
#include <sys/kmem.h>
#include <sys/mutex.h>
#include <sys/queue.h>
#include <sys/callout.h>
#include <sys/pool.h>
#include <sys/kthread.h>

#include "dm.h"

struct dm_delay_buf {
       TAILQ_ENTRY(dm_delay_buf) entry;
       struct buf *bp;
       int expire;
};
TAILQ_HEAD(dm_delay_buf_list, dm_delay_buf);

struct dm_delay_info {
       dm_pdev_t *pdev;
       uint64_t offset;
       int delay;
       int count;
       int enabled;
       struct dm_delay_buf_list buf_list;
       struct callout cal;
       kmutex_t buf_mtx;
       kmutex_t cal_mtx;
       kmutex_t token; /* lwkt_token in DragonFly */
};

typedef struct target_delay_config {
       struct dm_delay_info read;
       struct dm_delay_info write;
       int argc; /* either 3 or 6 */
} dm_target_delay_config_t;

static int _init(struct dm_delay_info*, char**);
static int _table(struct dm_delay_info*, char*);
static void _strategy(struct dm_delay_info*, struct buf*);
static __inline void _submit(struct dm_delay_info*, struct buf*);
static void _destroy(struct dm_delay_info*);
static void _timeout(void*);
static void _thread(void*);
static __inline void _debug(struct dm_delay_info*, const char*);

static struct pool pool;
static bool pool_initialized = false;

#ifdef DM_TARGET_MODULE
/*
* Every target can be compiled directly to dm driver or as a
* separate module this part of target is used for loading targets
* to dm driver.
* Target can be unloaded from kernel only if there are no users of
* it e.g. there are no devices which uses that target.
*/
#include <sys/kernel.h>
#include <sys/module.h>

MODULE(MODULE_CLASS_MISC, dm_target_delay, NULL);

static int
dm_target_delay_modcmd(modcmd_t cmd, void *arg)
{
       dm_target_t *dmt;
       int r;

       switch (cmd) {
       case MODULE_CMD_INIT:
               if ((dmt = dm_target_lookup("delay")) != NULL) {
                       dm_target_unbusy(dmt);
                       return EEXIST;
               }
               dmt = dm_target_alloc("delay");

               dmt->version[0] = 1;
               dmt->version[1] = 0;
               dmt->version[2] = 0;
               dmt->init = &dm_target_delay_init;
               dmt->table = &dm_target_delay_table;
               dmt->strategy = &dm_target_delay_strategy;
               dmt->sync = &dm_target_delay_sync;
               dmt->destroy = &dm_target_delay_destroy;
               //dmt->upcall = &dm_target_delay_upcall;
               dmt->secsize = &dm_target_delay_secsize;

               dm_target_delay_pool_create();
               r = dm_target_insert(dmt);

               break;

       case MODULE_CMD_FINI:
               r = dm_target_rem("delay");
               dm_target_delay_pool_destroy();
               break;

       case MODULE_CMD_STAT:
               return ENOTTY;

       default:
               return ENOTTY;
       }

       return r;
}
#endif

int
dm_target_delay_init(dm_table_entry_t *table_en, int argc, char **argv)
{
       dm_target_delay_config_t *tdc;
       int ret;

       if (argc != 3 && argc != 6) {
               printf("Delay target takes 3 or 6 args, %d given\n", argc);
               return EINVAL;
       }

       aprint_debug("Delay target init function called: argc=%d\n", argc);

       tdc = kmem_alloc(sizeof(dm_target_delay_config_t), KM_SLEEP);
       tdc->argc = argc;

       ret = _init(&tdc->read, argv);
       if (ret) {
               kmem_free(tdc, sizeof(*tdc));
               return ret;
       }

       if (argc == 6)
               argv += 3;

       ret = _init(&tdc->write, argv);
       if (ret) {
               dm_pdev_decr(tdc->read.pdev);
               kmem_free(tdc, sizeof(*tdc));
               return ret;
       }

       dm_table_add_deps(table_en, tdc->read.pdev);
       dm_table_add_deps(table_en, tdc->write.pdev);

       table_en->target_config = tdc;

       return 0;
}

static int
_init(struct dm_delay_info *di, char **argv)
{
       dm_pdev_t *dmp;
       int tmp;

       if (argv[0] == NULL)
               return EINVAL;
       if ((dmp = dm_pdev_insert(argv[0])) == NULL)
               return ENOENT;

       di->pdev = dmp;
       di->offset = atoi64(argv[1]);
       tmp = atoi64(argv[2]);
       di->delay = tmp * hz / 1000;
       di->count = 0;

       TAILQ_INIT(&di->buf_list);
       callout_init(&di->cal, 0);
       mutex_init(&di->buf_mtx, MUTEX_DEFAULT, IPL_NONE);
       mutex_init(&di->cal_mtx, MUTEX_DEFAULT, IPL_NONE);
       mutex_init(&di->token, MUTEX_DEFAULT, IPL_NONE);

       di->enabled = 1;
       if (kthread_create(PRI_NONE, 0, NULL, _thread, di, NULL,
           "dmdlthread")) {
               printf("Failed to create delay target's kthread\n");
               /* XXX cleanup ? */
               return EINVAL;
       }

       _debug(di, "init");
       return 0;
}

char *
dm_target_delay_info(void *target_config)
{
       dm_target_delay_config_t *tdc;
       char *params;

       tdc = target_config;
       KASSERT(tdc != NULL);

       aprint_debug("Delay target info function called\n");

       params = kmem_alloc(DM_MAX_PARAMS_SIZE, KM_SLEEP);
       snprintf(params, DM_MAX_PARAMS_SIZE, "%d %d", tdc->read.count,
           tdc->write.count);

       return params;
}

char *
dm_target_delay_table(void *target_config)
{
       dm_target_delay_config_t *tdc;
       char *params, *p;

       tdc = target_config;
       KASSERT(tdc != NULL);

       aprint_debug("Delay target table function called\n");

       params = kmem_alloc(DM_MAX_PARAMS_SIZE, KM_SLEEP);
       p = params;
       p += _table(&tdc->read, p);
       if (tdc->argc == 6) {
               p += snprintf(p, DM_MAX_PARAMS_SIZE, " ");
               _table(&tdc->write, p);
       }

       return params;
}

static int
_table(struct dm_delay_info *di, char *p)
{

       return snprintf(p, DM_MAX_PARAMS_SIZE, "%s %" PRIu64 " %d",
           di->pdev->udev_name, di->offset, di->delay);
}

int
dm_target_delay_strategy(dm_table_entry_t *table_en, struct buf *bp)
{
       dm_target_delay_config_t *tdc;
       struct dm_delay_info *di;

       tdc = table_en->target_config;
       KASSERT(tdc != NULL);

       if (bp->b_flags & B_READ)
               di = &tdc->read;
       else
               di = &tdc->write;

       if (di) {
               if (di->delay) {
                       _strategy(di, bp);
               } else {
                       _submit(di, bp);
               }
       } else {
               struct vnode *vnode = tdc->write.pdev->pdev_vnode;
               KASSERT(0);
               VOP_STRATEGY(vnode, bp);
       }

       return 0;
}

static void
_strategy(struct dm_delay_info *di, struct buf *bp)
{
       struct dm_delay_buf *dp;

       dp = pool_get(&pool, PR_WAITOK);
       dp->bp = bp;
       dp->expire = tick + di->delay;

       mutex_enter(&di->buf_mtx);
       di->count++;
       TAILQ_INSERT_TAIL(&di->buf_list, dp, entry);
       mutex_exit(&di->buf_mtx);

       mutex_enter(&di->cal_mtx);
       if (!callout_pending(&di->cal))
               callout_reset(&di->cal, di->delay, _timeout, di);
       mutex_exit(&di->cal_mtx);
}

static __inline void
_submit(struct dm_delay_info *di, struct buf *bp)
{

       _debug(di, "submit");

       bp->b_blkno += di->offset;
       VOP_STRATEGY(di->pdev->pdev_vnode, bp);
}

static void
_submit_queue(struct dm_delay_info *di, int submit_all)
{
       struct dm_delay_buf *dp;
       struct dm_delay_buf_list tmp_list;
       int next = -1;
       int reset = 0;

       _debug(di, "submitq");
       TAILQ_INIT(&tmp_list);

       mutex_enter(&di->buf_mtx);
       while ((dp = TAILQ_FIRST(&di->buf_list)) != NULL) {
               if (submit_all || tick > dp->expire) {
                       TAILQ_REMOVE(&di->buf_list, dp, entry);
                       TAILQ_INSERT_TAIL(&tmp_list, dp, entry);
                       di->count--;
                       continue;
               }
               if (reset == 0) {
                       reset = 1;
                       next = dp->expire;
               } else {
                       next = lmin(next, dp->expire);
               }
       }
       mutex_exit(&di->buf_mtx);

       if (reset) {
               mutex_enter(&di->cal_mtx);
               callout_reset(&di->cal, next - tick, _timeout, di);
               mutex_exit(&di->cal_mtx);
       }

       while ((dp = TAILQ_FIRST(&tmp_list)) != NULL) {
               TAILQ_REMOVE(&tmp_list, dp, entry);
               _submit(di, dp->bp);
               pool_put(&pool, dp);
       }
}

int
dm_target_delay_sync(dm_table_entry_t *table_en)
{
       dm_target_delay_config_t *tdc;
       int cmd;

       tdc = table_en->target_config;
       cmd = 1;

       return VOP_IOCTL(tdc->write.pdev->pdev_vnode, DIOCCACHESYNC, &cmd,
           FREAD | FWRITE, kauth_cred_get());
}

int
dm_target_delay_destroy(dm_table_entry_t *table_en)
{
       dm_target_delay_config_t *tdc;

       tdc = table_en->target_config;
       if (tdc == NULL)
               goto out;

       _destroy(&tdc->read);
       _destroy(&tdc->write);

       kmem_free(tdc, sizeof(*tdc));
out:
       /* Unbusy target so we can unload it */
       dm_target_unbusy(table_en->target);

       return 0;
}

static void
_destroy(struct dm_delay_info *di)
{

       _debug(di, "destroy");

       mutex_enter(&di->token);
       di->enabled = 0;

       mutex_enter(&di->cal_mtx);
       if (callout_pending(&di->cal))
               callout_halt(&di->cal, NULL);
       mutex_exit(&di->cal_mtx);

       _submit_queue(di, 1);
       mutex_exit(&di->token);
       wakeup(di);
       /* XXX race here (DragonFly uses lwkt_token instead of mutex) */
       tsleep(&di->enabled, 0, "dmdldestroy", 0);

       mutex_destroy(&di->token);
       mutex_destroy(&di->cal_mtx);
       mutex_destroy(&di->buf_mtx);

       dm_pdev_decr(di->pdev);
}

#if 0
int
dm_target_delay_upcall(dm_table_entry_t *table_en, struct buf *bp)
{

       return 0;
}
#endif

int
dm_target_delay_secsize(dm_table_entry_t *table_en, unsigned int *secsizep)
{
       dm_target_delay_config_t *tdc;
       unsigned int secsize;

       secsize = 0;

       tdc = table_en->target_config;
       if (tdc != NULL)
               secsize = tdc->write.pdev->pdev_secsize;

       *secsizep = secsize;

       return 0;
}

static void
_timeout(void *arg)
{
       struct dm_delay_info *di = arg;

       _debug(di, "timeout");
       wakeup(di);
}

static void
_thread(void *arg)
{
       struct dm_delay_info *di = arg;

       _debug(di, "thread init");
       mutex_enter(&di->token);

       while (di->enabled) {
               mutex_exit(&di->token);
               tsleep(di, 0, "dmdlthread", 0);
               _submit_queue(di, 0);
               mutex_enter(&di->token);
       }

       mutex_exit(&di->token);
       wakeup(&di->enabled);

       _debug(di, "thread exit");
       kthread_exit(0);
}

static __inline void
_debug(struct dm_delay_info *di, const char *msg)
{

       aprint_debug("%-8s: %d pdev=%s offset=%ju delay=%d count=%d\n",
           msg, di->enabled, di->pdev->name, (uintmax_t)di->offset, di->delay,
           di->count);
}

void
dm_target_delay_pool_create(void)
{

       if (!pool_initialized) {
               pool_init(&pool, sizeof(struct dm_delay_buf), 0, 0, 0,
                   "dmdlobj", &pool_allocator_nointr, IPL_NONE);
               pool_initialized = true;
               aprint_debug("Delay target pool initialized\n");
       }
}

void
dm_target_delay_pool_destroy(void)
{

       if (pool_initialized) {
               pool_destroy(&pool);
               pool_initialized = false;
               aprint_debug("Delay target pool destroyed\n");
       }
}