X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=zbd.c;h=b2d9424972da3d85e47c6fa80120987a7519a85b;hp=ba2c0401a95519696e38b7af852bd4a281823a39;hb=refs%2Fheads%2Fmaster;hpb=9fb714dae71993dda948c4757fdad6ee580099b8 diff --git a/zbd.c b/zbd.c index ba2c0401..8a092cbe 100644 --- a/zbd.c +++ b/zbd.c @@ -11,6 +11,7 @@ #include #include +#include "compiler/compiler.h" #include "os/os.h" #include "file.h" #include "fio.h" @@ -102,13 +103,12 @@ static bool zbd_zone_full(const struct fio_file *f, struct fio_zone_info *z, static void zone_lock(struct thread_data *td, const struct fio_file *f, struct fio_zone_info *z) { - struct zoned_block_device_info *zbd = f->zbd_info; - uint32_t nz = z - zbd->zone_info; - +#ifndef NDEBUG + unsigned int const nz = zbd_zone_idx(f, z); /* A thread should never lock zones outside its working area. */ assert(f->min_zone <= nz && nz < f->max_zone); - assert(z->has_wp); +#endif /* * Lock the io_u target zone. The zone will be unlocked if io_u offset @@ -128,11 +128,8 @@ static void zone_lock(struct thread_data *td, const struct fio_file *f, static inline void zone_unlock(struct fio_zone_info *z) { - int ret; - assert(z->has_wp); - ret = pthread_mutex_unlock(&z->mutex); - assert(!ret); + pthread_mutex_unlock(&z->mutex); } static inline struct fio_zone_info *zbd_get_zone(const struct fio_file *f, @@ -213,8 +210,8 @@ static int zbd_report_zones(struct thread_data *td, struct fio_file *f, ret = blkzoned_report_zones(td, f, offset, zones, nr_zones); if (ret < 0) { td_verror(td, errno, "report zones failed"); - log_err("%s: report zones from sector %"PRIu64" failed (%d).\n", - f->file_name, offset >> 9, errno); + log_err("%s: report zones from sector %"PRIu64" failed (nr_zones=%d; errno=%d).\n", + f->file_name, offset >> 9, nr_zones, errno); } else if (ret == 0) { td_verror(td, errno, "Empty zone report"); log_err("%s: report zones from sector %"PRIu64" is empty.\n", @@ -254,7 +251,7 @@ static int zbd_reset_wp(struct thread_data *td, struct fio_file *f, } /** - * zbd_reset_zone - reset the write pointer of a single zone + * __zbd_reset_zone - reset the write pointer of a single zone * @td: FIO thread data. * @f: FIO file associated with the disk for which to reset a write pointer. * @z: Zone to reset. @@ -263,8 +260,8 @@ static int zbd_reset_wp(struct thread_data *td, struct fio_file *f, * * The caller must hold z->mutex. */ -static int zbd_reset_zone(struct thread_data *td, struct fio_file *f, - struct fio_zone_info *z) +static int __zbd_reset_zone(struct thread_data *td, struct fio_file *f, + struct fio_zone_info *z) { uint64_t offset = z->start; uint64_t length = (z+1)->start - offset; @@ -304,39 +301,65 @@ static int zbd_reset_zone(struct thread_data *td, struct fio_file *f, } /** - * zbd_close_zone - Remove a zone from the open zones array. + * zbd_write_zone_put - Remove a zone from the write target zones array. * @td: FIO thread data. - * @f: FIO file associated with the disk for which to reset a write pointer. + * @f: FIO file that has the write zones array to remove. * @zone_idx: Index of the zone to remove. * * The caller must hold f->zbd_info->mutex. */ -static void zbd_close_zone(struct thread_data *td, const struct fio_file *f, - struct fio_zone_info *z) +static void zbd_write_zone_put(struct thread_data *td, const struct fio_file *f, + struct fio_zone_info *z) { - uint32_t ozi; + uint32_t zi; - if (!z->open) + if (!z->write) return; - for (ozi = 0; ozi < f->zbd_info->num_open_zones; ozi++) { - if (zbd_get_zone(f, f->zbd_info->open_zones[ozi]) == z) + for (zi = 0; zi < f->zbd_info->num_write_zones; zi++) { + if (zbd_get_zone(f, f->zbd_info->write_zones[zi]) == z) break; } - if (ozi == f->zbd_info->num_open_zones) + if (zi == f->zbd_info->num_write_zones) return; - dprint(FD_ZBD, "%s: closing zone %u\n", + dprint(FD_ZBD, "%s: removing zone %u from write zone array\n", f->file_name, zbd_zone_idx(f, z)); - memmove(f->zbd_info->open_zones + ozi, - f->zbd_info->open_zones + ozi + 1, - (ZBD_MAX_OPEN_ZONES - (ozi + 1)) * - sizeof(f->zbd_info->open_zones[0])); + memmove(f->zbd_info->write_zones + zi, + f->zbd_info->write_zones + zi + 1, + (ZBD_MAX_WRITE_ZONES - (zi + 1)) * + sizeof(f->zbd_info->write_zones[0])); + + f->zbd_info->num_write_zones--; + td->num_write_zones--; + z->write = 0; +} + +/** + * zbd_reset_zone - reset the write pointer of a single zone and remove the zone + * from the array of write zones. + * @td: FIO thread data. + * @f: FIO file associated with the disk for which to reset a write pointer. + * @z: Zone to reset. + * + * Returns 0 upon success and a negative error code upon failure. + * + * The caller must hold z->mutex. + */ +static int zbd_reset_zone(struct thread_data *td, struct fio_file *f, + struct fio_zone_info *z) +{ + int ret; + + ret = __zbd_reset_zone(td, f, z); + if (ret) + return ret; - f->zbd_info->num_open_zones--; - td->num_open_zones--; - z->open = 0; + pthread_mutex_lock(&f->zbd_info->mutex); + zbd_write_zone_put(td, f, z); + pthread_mutex_unlock(&f->zbd_info->mutex); + return 0; } /** @@ -394,7 +417,8 @@ static int zbd_reset_zones(struct thread_data *td, struct fio_file *f, const uint64_t min_bs = td->o.min_bs[DDIR_WRITE]; int res = 0; - assert(min_bs); + if (fio_unlikely(0 == min_bs)) + return 1; dprint(FD_ZBD, "%s: examining zones %u .. %u\n", f->file_name, zbd_zone_idx(f, zb), zbd_zone_idx(f, ze)); @@ -404,9 +428,6 @@ static int zbd_reset_zones(struct thread_data *td, struct fio_file *f, continue; zone_lock(td, f, z); - pthread_mutex_lock(&f->zbd_info->mutex); - zbd_close_zone(td, f, z); - pthread_mutex_unlock(&f->zbd_info->mutex); if (z->wp != z->start) { dprint(FD_ZBD, "%s: resetting zone %u\n", @@ -450,21 +471,47 @@ static int zbd_get_max_open_zones(struct thread_data *td, struct fio_file *f, } /** - * zbd_open_zone - Add a zone to the array of open zones. + * zbd_get_max_active_zones - Get the maximum number of active zones + * @td: FIO thread data + * @f: FIO file for which to get max active zones + * + * Returns max_active_zones limit value of the target file if it is available. + * Otherwise return zero, which means no limit. + */ +static unsigned int zbd_get_max_active_zones(struct thread_data *td, + struct fio_file *f) +{ + unsigned int max_active_zones; + int ret; + + if (td->io_ops && td->io_ops->get_max_active_zones) + ret = td->io_ops->get_max_active_zones(td, f, + &max_active_zones); + else + ret = blkzoned_get_max_active_zones(td, f, &max_active_zones); + if (ret < 0) { + dprint(FD_ZBD, "%s: max_active_zones is not available\n", + f->file_name); + return 0; + } + + return max_active_zones; +} + +/** + * __zbd_write_zone_get - Add a zone to the array of write zones. * @td: fio thread data. - * @f: fio file that has the open zones to add. + * @f: fio file that has the write zones array to add. * @zone_idx: Index of the zone to add. * - * Open a ZBD zone if it is not already open. Returns true if either the zone - * was already open or if the zone was successfully added to the array of open - * zones without exceeding the maximum number of open zones. Returns false if - * the zone was not already open and opening the zone would cause the zone limit - * to be exceeded. + * Do same operation as @zbd_write_zone_get, except it adds the zone at + * @zone_idx to write target zones array even when it does not have remainder + * space to write one block. */ -static bool zbd_open_zone(struct thread_data *td, const struct fio_file *f, - struct fio_zone_info *z) +static bool __zbd_write_zone_get(struct thread_data *td, + const struct fio_file *f, + struct fio_zone_info *z) { - const uint64_t min_bs = td->o.min_bs[DDIR_WRITE]; struct zoned_block_device_info *zbdi = f->zbd_info; uint32_t zone_idx = zbd_zone_idx(f, z); bool res = true; @@ -476,24 +523,24 @@ static bool zbd_open_zone(struct thread_data *td, const struct fio_file *f, * Skip full zones with data verification enabled because resetting a * zone causes data loss and hence causes verification to fail. */ - if (td->o.verify != VERIFY_NONE && zbd_zone_full(f, z, min_bs)) + if (td->o.verify != VERIFY_NONE && zbd_zone_remainder(z) == 0) return false; /* - * zbdi->max_open_zones == 0 means that there is no limit on the maximum - * number of open zones. In this case, do no track open zones in - * zbdi->open_zones array. + * zbdi->max_write_zones == 0 means that there is no limit on the + * maximum number of write target zones. In this case, do no track write + * target zones in zbdi->write_zones array. */ - if (!zbdi->max_open_zones) + if (!zbdi->max_write_zones) return true; pthread_mutex_lock(&zbdi->mutex); - if (z->open) { + if (z->write) { /* * If the zone is going to be completely filled by writes - * already in-flight, handle it as a full zone instead of an - * open zone. + * already in-flight, handle it as a full zone instead of a + * write target zone. */ if (!zbd_zone_remainder(z)) res = false; @@ -503,17 +550,17 @@ static bool zbd_open_zone(struct thread_data *td, const struct fio_file *f, res = false; /* Zero means no limit */ if (td->o.job_max_open_zones > 0 && - td->num_open_zones >= td->o.job_max_open_zones) + td->num_write_zones >= td->o.job_max_open_zones) goto out; - if (zbdi->num_open_zones >= zbdi->max_open_zones) + if (zbdi->num_write_zones >= zbdi->max_write_zones) goto out; - dprint(FD_ZBD, "%s: opening zone %u\n", + dprint(FD_ZBD, "%s: adding zone %u to write zone array\n", f->file_name, zone_idx); - zbdi->open_zones[zbdi->num_open_zones++] = zone_idx; - td->num_open_zones++; - z->open = 1; + zbdi->write_zones[zbdi->num_write_zones++] = zone_idx; + td->num_write_zones++; + z->write = 1; res = true; out: @@ -521,14 +568,40 @@ out: return res; } +/** + * zbd_write_zone_get - Add a zone to the array of write zones. + * @td: fio thread data. + * @f: fio file that has the open zones to add. + * @zone_idx: Index of the zone to add. + * + * Add a ZBD zone to write target zones array, if it is not yet added. Returns + * true if either the zone was already added or if the zone was successfully + * added to the array without exceeding the maximum number of write zones. + * Returns false if the zone was not already added and addition of the zone + * would cause the zone limit to be exceeded. + */ +static bool zbd_write_zone_get(struct thread_data *td, const struct fio_file *f, + struct fio_zone_info *z) +{ + const uint64_t min_bs = td->o.min_bs[DDIR_WRITE]; + + /* + * Skip full zones with data verification enabled because resetting a + * zone causes data loss and hence causes verification to fail. + */ + if (td->o.verify != VERIFY_NONE && zbd_zone_full(f, z, min_bs)) + return false; + + return __zbd_write_zone_get(td, f, z); +} + /* Verify whether direct I/O is used for all host-managed zoned block drives. */ static bool zbd_using_direct_io(void) { - struct thread_data *td; struct fio_file *f; - int i, j; + int j; - for_each_td(td, i) { + for_each_td(td) { if (td->o.odirect || !(td->o.td_ddir & TD_DDIR_WRITE)) continue; for_each_file(td, f, j) { @@ -536,7 +609,7 @@ static bool zbd_using_direct_io(void) f->zbd_info->model == ZBD_HOST_MANAGED) return false; } - } + } end_for_each(); return true; } @@ -600,9 +673,20 @@ static bool zbd_zone_align_file_sizes(struct thread_data *td, return false; } + if (td->o.td_ddir == TD_DDIR_READ) { + z = zbd_offset_to_zone(f, f->file_offset + f->io_size); + new_end = z->start; + if (f->file_offset + f->io_size > new_end) { + log_info("%s: rounded io_size from %"PRIu64" to %"PRIu64"\n", + f->file_name, f->io_size, + new_end - f->file_offset); + f->io_size = new_end - f->file_offset; + } + return true; + } + z = zbd_offset_to_zone(f, f->file_offset); - if ((f->file_offset != z->start) && - (td->o.td_ddir != TD_DDIR_READ)) { + if (f->file_offset != z->start) { new_offset = zbd_zone_end(z); if (new_offset >= f->file_offset + f->io_size) { log_info("%s: io_size must be at least one zone\n", @@ -618,8 +702,7 @@ static bool zbd_zone_align_file_sizes(struct thread_data *td, z = zbd_offset_to_zone(f, f->file_offset + f->io_size); new_end = z->start; - if ((td->o.td_ddir != TD_DDIR_READ) && - (f->file_offset + f->io_size != new_end)) { + if (f->file_offset + f->io_size != new_end) { if (new_end <= f->file_offset) { log_info("%s: io_size must be at least one zone\n", f->file_name); @@ -639,27 +722,25 @@ static bool zbd_zone_align_file_sizes(struct thread_data *td, */ static bool zbd_verify_sizes(void) { - struct thread_data *td; struct fio_file *f; - int i, j; + int j; - for_each_td(td, i) { + for_each_td(td) { for_each_file(td, f, j) { if (!zbd_zone_align_file_sizes(td, f)) return false; } - } + } end_for_each(); return true; } static bool zbd_verify_bs(void) { - struct thread_data *td; struct fio_file *f; - int i, j; + int j; - for_each_td(td, i) { + for_each_td(td) { if (td_trim(td) && (td->o.min_bs[DDIR_TRIM] != td->o.max_bs[DDIR_TRIM] || td->o.bssplit_nr[DDIR_TRIM])) { @@ -680,7 +761,7 @@ static bool zbd_verify_bs(void) return false; } } - } + } end_for_each(); return true; } @@ -779,7 +860,8 @@ static int parse_zone_info(struct thread_data *td, struct fio_file *f) int nr_zones, nrz; struct zbd_zone *zones, *z; struct fio_zone_info *p; - uint64_t zone_size, offset; + uint64_t zone_size, offset, capacity; + bool same_zone_cap = true; struct zoned_block_device_info *zbd_info = NULL; int i, j, ret = -ENOMEM; @@ -796,6 +878,7 @@ static int parse_zone_info(struct thread_data *td, struct fio_file *f) } zone_size = zones[0].len; + capacity = zones[0].capacity; nr_zones = (f->real_file_size + zone_size - 1) / zone_size; if (td->o.zone_size == 0) { @@ -824,6 +907,8 @@ static int parse_zone_info(struct thread_data *td, struct fio_file *f) PTHREAD_MUTEX_RECURSIVE); p->start = z->start; p->capacity = z->capacity; + if (capacity != z->capacity) + same_zone_cap = false; switch (z->cond) { case ZBD_ZONE_COND_NOT_WP: @@ -848,8 +933,9 @@ static int parse_zone_info(struct thread_data *td, struct fio_file *f) p->cond = z->cond; if (j > 0 && p->start != p[-1].start + zone_size) { - log_info("%s: invalid zone data\n", - f->file_name); + log_info("%s: invalid zone data [%d:%d]: %"PRIu64" + %"PRIu64" != %"PRIu64"\n", + f->file_name, j, i, + p[-1].start, zone_size, p->start); ret = -EINVAL; goto out; } @@ -878,6 +964,12 @@ static int parse_zone_info(struct thread_data *td, struct fio_file *f) f->zbd_info->zone_size_log2 = is_power_of_2(zone_size) ? ilog2(zone_size) : 0; f->zbd_info->nr_zones = nr_zones; + f->zbd_info->max_active_zones = zbd_get_max_active_zones(td, f); + + if (same_zone_cap) + dprint(FD_ZBD, "Zone capacity = %"PRIu64" KB\n", + capacity / 1024); + zbd_info = NULL; ret = 0; @@ -887,7 +979,7 @@ out: return ret; } -static int zbd_set_max_open_zones(struct thread_data *td, struct fio_file *f) +static int zbd_set_max_write_zones(struct thread_data *td, struct fio_file *f) { struct zoned_block_device_info *zbd = f->zbd_info; unsigned int max_open_zones; @@ -895,7 +987,7 @@ static int zbd_set_max_open_zones(struct thread_data *td, struct fio_file *f) if (zbd->model != ZBD_HOST_MANAGED || td->o.ignore_zone_limits) { /* Only host-managed devices have a max open limit */ - zbd->max_open_zones = td->o.max_open_zones; + zbd->max_write_zones = td->o.max_open_zones; goto out; } @@ -906,13 +998,13 @@ static int zbd_set_max_open_zones(struct thread_data *td, struct fio_file *f) if (!max_open_zones) { /* No device limit */ - zbd->max_open_zones = td->o.max_open_zones; + zbd->max_write_zones = td->o.max_open_zones; } else if (!td->o.max_open_zones) { /* No user limit. Set limit to device limit */ - zbd->max_open_zones = max_open_zones; + zbd->max_write_zones = max_open_zones; } else if (td->o.max_open_zones <= max_open_zones) { /* Both user limit and dev limit. User limit not too large */ - zbd->max_open_zones = td->o.max_open_zones; + zbd->max_write_zones = td->o.max_open_zones; } else { /* Both user limit and dev limit. User limit too large */ td_verror(td, EINVAL, @@ -924,15 +1016,15 @@ static int zbd_set_max_open_zones(struct thread_data *td, struct fio_file *f) out: /* Ensure that the limit is not larger than FIO's internal limit */ - if (zbd->max_open_zones > ZBD_MAX_OPEN_ZONES) { + if (zbd->max_write_zones > ZBD_MAX_WRITE_ZONES) { td_verror(td, EINVAL, "'max_open_zones' value is too large"); log_err("'max_open_zones' value is larger than %u\n", - ZBD_MAX_OPEN_ZONES); + ZBD_MAX_WRITE_ZONES); return -EINVAL; } - dprint(FD_ZBD, "%s: using max open zones limit: %"PRIu32"\n", - f->file_name, zbd->max_open_zones); + dprint(FD_ZBD, "%s: using max write zones limit: %"PRIu32"\n", + f->file_name, zbd->max_write_zones); return 0; } @@ -974,7 +1066,7 @@ static int zbd_create_zone_info(struct thread_data *td, struct fio_file *f) assert(f->zbd_info); f->zbd_info->model = zbd_model; - ret = zbd_set_max_open_zones(td, f); + ret = zbd_set_max_write_zones(td, f); if (ret) { zbd_free_zone_info(f); return ret; @@ -1009,11 +1101,10 @@ void zbd_free_zone_info(struct fio_file *f) */ static int zbd_init_zone_info(struct thread_data *td, struct fio_file *file) { - struct thread_data *td2; struct fio_file *f2; - int i, j, ret; + int j, ret; - for_each_td(td2, i) { + for_each_td(td2) { for_each_file(td2, f2, j) { if (td2 == td && f2 == file) continue; @@ -1024,7 +1115,7 @@ static int zbd_init_zone_info(struct thread_data *td, struct fio_file *file) file->zbd_info->refcount++; return 0; } - } + } end_for_each(); ret = zbd_create_zone_info(td, file); if (ret < 0) @@ -1168,7 +1259,7 @@ int zbd_setup_files(struct thread_data *td) assert(f->min_zone < f->max_zone); if (td->o.max_open_zones > 0 && - zbd->max_open_zones != td->o.max_open_zones) { + zbd->max_write_zones != td->o.max_open_zones) { log_err("Different 'max_open_zones' values\n"); return 1; } @@ -1178,34 +1269,36 @@ int zbd_setup_files(struct thread_data *td) * global max open zones limit. (As the tracking of open zones * is disabled when there is no global max open zones limit.) */ - if (td->o.job_max_open_zones && !zbd->max_open_zones) { + if (td->o.job_max_open_zones && !zbd->max_write_zones) { log_err("'job_max_open_zones' cannot be used without a global open zones limit\n"); return 1; } /* - * zbd->max_open_zones is the global limit shared for all jobs + * zbd->max_write_zones is the global limit shared for all jobs * that target the same zoned block device. Force sync the per * thread global limit with the actual global limit. (The real * per thread/job limit is stored in td->o.job_max_open_zones). */ - td->o.max_open_zones = zbd->max_open_zones; + td->o.max_open_zones = zbd->max_write_zones; for (zi = f->min_zone; zi < f->max_zone; zi++) { z = &zbd->zone_info[zi]; if (z->cond != ZBD_ZONE_COND_IMP_OPEN && - z->cond != ZBD_ZONE_COND_EXP_OPEN) + z->cond != ZBD_ZONE_COND_EXP_OPEN && + z->cond != ZBD_ZONE_COND_CLOSED) + continue; + if (!zbd->max_active_zones && + z->cond == ZBD_ZONE_COND_CLOSED) continue; - if (zbd_open_zone(td, f, z)) + if (__zbd_write_zone_get(td, f, z)) continue; /* * If the number of open zones exceeds specified limits, - * reset all extra open zones. + * error out. */ - if (zbd_reset_zone(td, f, z) < 0) { - log_err("Failed to reest zone %d\n", zi); - return 1; - } + log_err("Number of open zones exceeds max_open_zones limit\n"); + return 1; } } @@ -1268,9 +1361,6 @@ void zbd_file_reset(struct thread_data *td, struct fio_file *f) if (td->o.verify != VERIFY_NONE) { verify_data_left = td->runstate == TD_VERIFYING || td->io_hist_len || td->verify_batch; - if (td->io_hist_len && td->o.verify_backlog) - verify_data_left = - td->io_hist_len % td->o.verify_backlog; if (!verify_data_left) zbd_reset_zones(td, f, zb, ze); } @@ -1278,57 +1368,54 @@ void zbd_file_reset(struct thread_data *td, struct fio_file *f) zbd_reset_write_cnt(td, f); } -/* Return random zone index for one of the open zones. */ +/* Return random zone index for one of the write target zones. */ static uint32_t pick_random_zone_idx(const struct fio_file *f, const struct io_u *io_u) { return (io_u->offset - f->file_offset) * - f->zbd_info->num_open_zones / f->io_size; + f->zbd_info->num_write_zones / f->io_size; } static bool any_io_in_flight(void) { - struct thread_data *td; - int i; - - for_each_td(td, i) { + for_each_td(td) { if (td->io_u_in_flight) return true; - } + } end_for_each(); return false; } /* - * Modify the offset of an I/O unit that does not refer to an open zone such - * that it refers to an open zone. Close an open zone and open a new zone if - * necessary. The open zone is searched across sequential zones. + * Modify the offset of an I/O unit that does not refer to a zone such that + * in write target zones array. Add a zone to or remove a zone from the lsit if + * necessary. The write target zone is searched across sequential zones. * This algorithm can only work correctly if all write pointers are * a multiple of the fio block size. The caller must neither hold z->mutex * nor f->zbd_info->mutex. Returns with z->mutex held upon success. */ -static struct fio_zone_info *zbd_convert_to_open_zone(struct thread_data *td, - struct io_u *io_u) +static struct fio_zone_info *zbd_convert_to_write_zone(struct thread_data *td, + struct io_u *io_u) { const uint64_t min_bs = td->o.min_bs[io_u->ddir]; struct fio_file *f = io_u->file; struct zoned_block_device_info *zbdi = f->zbd_info; struct fio_zone_info *z; - unsigned int open_zone_idx = -1; + unsigned int write_zone_idx = -1; uint32_t zone_idx, new_zone_idx; int i; - bool wait_zone_close; + bool wait_zone_write; bool in_flight; bool should_retry = true; assert(is_valid_offset(f, io_u->offset)); - if (zbdi->max_open_zones || td->o.job_max_open_zones) { + if (zbdi->max_write_zones || td->o.job_max_open_zones) { /* - * This statement accesses zbdi->open_zones[] on purpose + * This statement accesses zbdi->write_zones[] on purpose * without locking. */ - zone_idx = zbdi->open_zones[pick_random_zone_idx(f, io_u)]; + zone_idx = zbdi->write_zones[pick_random_zone_idx(f, io_u)]; } else { zone_idx = zbd_offset_to_zone_idx(f, io_u->offset); } @@ -1358,34 +1445,34 @@ static struct fio_zone_info *zbd_convert_to_open_zone(struct thread_data *td, if (z->has_wp) { if (z->cond != ZBD_ZONE_COND_OFFLINE && - zbdi->max_open_zones == 0 && + zbdi->max_write_zones == 0 && td->o.job_max_open_zones == 0) goto examine_zone; - if (zbdi->num_open_zones == 0) { - dprint(FD_ZBD, "%s(%s): no zones are open\n", + if (zbdi->num_write_zones == 0) { + dprint(FD_ZBD, "%s(%s): no zone is write target\n", __func__, f->file_name); - goto open_other_zone; + goto choose_other_zone; } } /* - * List of opened zones is per-device, shared across all + * Array of write target zones is per-device, shared across all * threads. Start with quasi-random candidate zone. Ignore * zones which don't belong to thread's offset/size area. */ - open_zone_idx = pick_random_zone_idx(f, io_u); - assert(!open_zone_idx || - open_zone_idx < zbdi->num_open_zones); - tmp_idx = open_zone_idx; + write_zone_idx = pick_random_zone_idx(f, io_u); + assert(!write_zone_idx || + write_zone_idx < zbdi->num_write_zones); + tmp_idx = write_zone_idx; - for (i = 0; i < zbdi->num_open_zones; i++) { + for (i = 0; i < zbdi->num_write_zones; i++) { uint32_t tmpz; - if (tmp_idx >= zbdi->num_open_zones) + if (tmp_idx >= zbdi->num_write_zones) tmp_idx = 0; - tmpz = zbdi->open_zones[tmp_idx]; + tmpz = zbdi->write_zones[tmp_idx]; if (f->min_zone <= tmpz && tmpz < f->max_zone) { - open_zone_idx = tmp_idx; + write_zone_idx = tmp_idx; goto found_candidate_zone; } @@ -1403,7 +1490,7 @@ static struct fio_zone_info *zbd_convert_to_open_zone(struct thread_data *td, return NULL; found_candidate_zone: - new_zone_idx = zbdi->open_zones[open_zone_idx]; + new_zone_idx = zbdi->write_zones[write_zone_idx]; if (new_zone_idx == zone_idx) break; zone_idx = new_zone_idx; @@ -1422,32 +1509,32 @@ examine_zone: goto out; } -open_other_zone: - /* Check if number of open zones reaches one of limits. */ - wait_zone_close = - zbdi->num_open_zones == f->max_zone - f->min_zone || - (zbdi->max_open_zones && - zbdi->num_open_zones == zbdi->max_open_zones) || +choose_other_zone: + /* Check if number of write target zones reaches one of limits. */ + wait_zone_write = + zbdi->num_write_zones == f->max_zone - f->min_zone || + (zbdi->max_write_zones && + zbdi->num_write_zones == zbdi->max_write_zones) || (td->o.job_max_open_zones && - td->num_open_zones == td->o.job_max_open_zones); + td->num_write_zones == td->o.job_max_open_zones); pthread_mutex_unlock(&zbdi->mutex); /* Only z->mutex is held. */ /* - * When number of open zones reaches to one of limits, wait for - * zone close before opening a new zone. + * When number of write target zones reaches to one of limits, wait for + * zone write completion to one of them before trying a new zone. */ - if (wait_zone_close) { + if (wait_zone_write) { dprint(FD_ZBD, - "%s(%s): quiesce to allow open zones to close\n", + "%s(%s): quiesce to remove a zone from write target zones array\n", __func__, f->file_name); io_u_quiesce(td); } retry: - /* Zone 'z' is full, so try to open a new zone. */ + /* Zone 'z' is full, so try to choose a new zone. */ for (i = f->io_size / zbdi->zone_size; i > 0; i--) { zone_idx++; if (z->has_wp) @@ -1462,18 +1549,18 @@ retry: if (!z->has_wp) continue; zone_lock(td, f, z); - if (z->open) + if (z->write) continue; - if (zbd_open_zone(td, f, z)) + if (zbd_write_zone_get(td, f, z)) goto out; } /* Only z->mutex is held. */ - /* Check whether the write fits in any of the already opened zones. */ + /* Check whether the write fits in any of the write target zones. */ pthread_mutex_lock(&zbdi->mutex); - for (i = 0; i < zbdi->num_open_zones; i++) { - zone_idx = zbdi->open_zones[i]; + for (i = 0; i < zbdi->num_write_zones; i++) { + zone_idx = zbdi->write_zones[i]; if (zone_idx < f->min_zone || zone_idx >= f->max_zone) continue; pthread_mutex_unlock(&zbdi->mutex); @@ -1489,19 +1576,20 @@ retry: /* * When any I/O is in-flight or when all I/Os in-flight get completed, - * the I/Os might have closed zones then retry the steps to open a zone. - * Before retry, call io_u_quiesce() to complete in-flight writes. + * the I/Os might have removed zones from the write target array then + * retry the steps to choose a zone. Before retry, call io_u_quiesce() + * to complete in-flight writes. */ in_flight = any_io_in_flight(); if (in_flight || should_retry) { dprint(FD_ZBD, - "%s(%s): wait zone close and retry open zones\n", + "%s(%s): wait zone write and retry write target zone selection\n", __func__, f->file_name); + should_retry = in_flight; pthread_mutex_unlock(&zbdi->mutex); zone_unlock(z); io_u_quiesce(td); zone_lock(td, f, z); - should_retry = in_flight; goto retry; } @@ -1509,7 +1597,7 @@ retry: zone_unlock(z); - dprint(FD_ZBD, "%s(%s): did not open another zone\n", + dprint(FD_ZBD, "%s(%s): did not choose another write zone\n", __func__, f->file_name); return NULL; @@ -1579,7 +1667,8 @@ zbd_find_zone(struct thread_data *td, struct io_u *io_u, uint64_t min_bytes, * @io_u: I/O unit * @z: zone info pointer * - * If the write command made the zone full, close it. + * If the write command made the zone full, remove it from the write target + * zones array. * * The caller must hold z->mutex. */ @@ -1591,7 +1680,7 @@ static void zbd_end_zone_io(struct thread_data *td, const struct io_u *io_u, if (io_u->ddir == DDIR_WRITE && io_u->offset + io_u->buflen >= zbd_zone_capacity_end(z)) { pthread_mutex_lock(&f->zbd_info->mutex); - zbd_close_zone(td, f, z); + zbd_write_zone_put(td, f, z); pthread_mutex_unlock(&f->zbd_info->mutex); } } @@ -1663,10 +1752,9 @@ unlock: static void zbd_put_io(struct thread_data *td, const struct io_u *io_u) { const struct fio_file *f = io_u->file; - struct zoned_block_device_info *zbd_info = f->zbd_info; struct fio_zone_info *z; - assert(zbd_info); + assert(f->zbd_info); z = zbd_offset_to_zone(f, io_u->offset); assert(z->has_wp); @@ -1794,7 +1882,8 @@ enum fio_ddir zbd_adjust_ddir(struct thread_data *td, struct io_u *io_u, if (ddir != DDIR_READ || !td_rw(td)) return ddir; - if (io_u->file->last_start[DDIR_WRITE] != -1ULL || td->o.read_beyond_wp) + if (io_u->file->last_start[DDIR_WRITE] != -1ULL || + td->o.read_beyond_wp || td->o.rwmix[DDIR_WRITE] == 0) return DDIR_READ; return DDIR_WRITE; @@ -1951,7 +2040,7 @@ retry: if (zbd_zone_remainder(zb) > 0 && zbd_zone_remainder(zb) < min_bs) { pthread_mutex_lock(&f->zbd_info->mutex); - zbd_close_zone(td, f, zb); + zbd_write_zone_put(td, f, zb); pthread_mutex_unlock(&f->zbd_info->mutex); dprint(FD_ZBD, "%s: finish zone %d\n", @@ -1974,11 +2063,11 @@ retry: zone_lock(td, f, zb); } - if (!zbd_open_zone(td, f, zb)) { + if (!zbd_write_zone_get(td, f, zb)) { zone_unlock(zb); - zb = zbd_convert_to_open_zone(td, io_u); + zb = zbd_convert_to_write_zone(td, io_u); if (!zb) { - dprint(FD_IO, "%s: can't convert to open zone", + dprint(FD_IO, "%s: can't convert to write target zone", f->file_name); goto eof; } @@ -2020,7 +2109,7 @@ retry: */ io_u_quiesce(td); zb->reset_zone = 0; - if (zbd_reset_zone(td, f, zb) < 0) + if (__zbd_reset_zone(td, f, zb) < 0) goto eof; if (zb->capacity < min_bs) { @@ -2089,6 +2178,7 @@ retry: case DDIR_WAIT: case DDIR_LAST: case DDIR_INVAL: + case DDIR_TIMEOUT: goto accept; } @@ -2139,7 +2229,7 @@ char *zbd_write_status(const struct thread_stat *ts) * Return io_u_completed when reset zone succeeds. Return 0 when the target zone * does not have write pointer. On error, return negative errno. */ -int zbd_do_io_u_trim(const struct thread_data *td, struct io_u *io_u) +int zbd_do_io_u_trim(struct thread_data *td, struct io_u *io_u) { struct fio_file *f = io_u->file; struct fio_zone_info *z; @@ -2161,3 +2251,15 @@ int zbd_do_io_u_trim(const struct thread_data *td, struct io_u *io_u) return io_u_completed; } + +void zbd_log_err(const struct thread_data *td, const struct io_u *io_u) +{ + const struct fio_file *f = io_u->file; + + if (td->o.zone_mode != ZONE_MODE_ZBD) + return; + + if (io_u->error == EOVERFLOW) + log_err("%s: Exceeded max_active_zones limit. Check conditions of zones out of I/O ranges.\n", + f->file_name); +}