1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2020 Western Digital Corporation or its affiliates.
6 #include <linux/blkdev.h>
7 #include <linux/vmalloc.h>
10 int nvme_revalidate_zones(struct nvme_ns *ns)
12 struct request_queue *q = ns->queue;
15 ret = blk_revalidate_disk_zones(ns->disk, NULL);
17 blk_queue_max_zone_append_sectors(q, ns->ctrl->max_zone_append);
21 static int nvme_set_max_append(struct nvme_ctrl *ctrl)
23 struct nvme_command c = { };
24 struct nvme_id_ctrl_zns *id;
27 id = kzalloc(sizeof(*id), GFP_KERNEL);
31 c.identify.opcode = nvme_admin_identify;
32 c.identify.cns = NVME_ID_CNS_CS_CTRL;
33 c.identify.csi = NVME_CSI_ZNS;
35 status = nvme_submit_sync_cmd(ctrl->admin_q, &c, id, sizeof(*id));
42 ctrl->max_zone_append = 1 << (id->zasl + 3);
44 ctrl->max_zone_append = ctrl->max_hw_sectors;
49 int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf)
51 struct nvme_effects_log *log = ns->head->effects;
52 struct request_queue *q = ns->queue;
53 struct nvme_command c = { };
54 struct nvme_id_ns_zns *id;
57 /* Driver requires zone append support */
58 if (!(le32_to_cpu(log->iocs[nvme_cmd_zone_append]) &
59 NVME_CMD_EFFECTS_CSUPP)) {
60 dev_warn(ns->ctrl->device,
61 "append not supported for zoned namespace:%d\n",
66 /* Lazily query controller append limit for the first zoned namespace */
67 if (!ns->ctrl->max_zone_append) {
68 status = nvme_set_max_append(ns->ctrl);
73 id = kzalloc(sizeof(*id), GFP_KERNEL);
77 c.identify.opcode = nvme_admin_identify;
78 c.identify.nsid = cpu_to_le32(ns->head->ns_id);
79 c.identify.cns = NVME_ID_CNS_CS_NS;
80 c.identify.csi = NVME_CSI_ZNS;
82 status = nvme_submit_sync_cmd(ns->ctrl->admin_q, &c, id, sizeof(*id));
87 * We currently do not handle devices requiring any of the zoned
88 * operation characteristics.
91 dev_warn(ns->ctrl->device,
92 "zone operations:%x not supported for namespace:%u\n",
93 le16_to_cpu(id->zoc), ns->head->ns_id);
98 ns->zsze = nvme_lba_to_sect(ns, le64_to_cpu(id->lbafe[lbaf].zsze));
99 if (!is_power_of_2(ns->zsze)) {
100 dev_warn(ns->ctrl->device,
101 "invalid zone size:%llu for namespace:%u\n",
102 ns->zsze, ns->head->ns_id);
107 q->limits.zoned = BLK_ZONED_HM;
108 blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
109 blk_queue_max_open_zones(q, le32_to_cpu(id->mor) + 1);
110 blk_queue_max_active_zones(q, le32_to_cpu(id->mar) + 1);
116 static void *nvme_zns_alloc_report_buffer(struct nvme_ns *ns,
117 unsigned int nr_zones, size_t *buflen)
119 struct request_queue *q = ns->disk->queue;
123 const size_t min_bufsize = sizeof(struct nvme_zone_report) +
124 sizeof(struct nvme_zone_descriptor);
126 nr_zones = min_t(unsigned int, nr_zones,
127 get_capacity(ns->disk) >> ilog2(ns->zsze));
129 bufsize = sizeof(struct nvme_zone_report) +
130 nr_zones * sizeof(struct nvme_zone_descriptor);
131 bufsize = min_t(size_t, bufsize,
132 queue_max_hw_sectors(q) << SECTOR_SHIFT);
133 bufsize = min_t(size_t, bufsize, queue_max_segments(q) << PAGE_SHIFT);
135 while (bufsize >= min_bufsize) {
136 buf = __vmalloc(bufsize, GFP_KERNEL | __GFP_NORETRY);
146 static int nvme_zone_parse_entry(struct nvme_ns *ns,
147 struct nvme_zone_descriptor *entry,
148 unsigned int idx, report_zones_cb cb,
151 struct blk_zone zone = { };
153 if ((entry->zt & 0xf) != NVME_ZONE_TYPE_SEQWRITE_REQ) {
154 dev_err(ns->ctrl->device, "invalid zone type %#x\n",
159 zone.type = BLK_ZONE_TYPE_SEQWRITE_REQ;
160 zone.cond = entry->zs >> 4;
162 zone.capacity = nvme_lba_to_sect(ns, le64_to_cpu(entry->zcap));
163 zone.start = nvme_lba_to_sect(ns, le64_to_cpu(entry->zslba));
164 zone.wp = nvme_lba_to_sect(ns, le64_to_cpu(entry->wp));
166 return cb(&zone, idx, data);
169 static int nvme_ns_report_zones(struct nvme_ns *ns, sector_t sector,
170 unsigned int nr_zones, report_zones_cb cb, void *data)
172 struct nvme_zone_report *report;
173 struct nvme_command c = { };
174 int ret, zone_idx = 0;
178 report = nvme_zns_alloc_report_buffer(ns, nr_zones, &buflen);
182 c.zmr.opcode = nvme_cmd_zone_mgmt_recv;
183 c.zmr.nsid = cpu_to_le32(ns->head->ns_id);
184 c.zmr.numd = cpu_to_le32(nvme_bytes_to_numd(buflen));
185 c.zmr.zra = NVME_ZRA_ZONE_REPORT;
186 c.zmr.zrasf = NVME_ZRASF_ZONE_REPORT_ALL;
187 c.zmr.pr = NVME_REPORT_ZONE_PARTIAL;
189 sector &= ~(ns->zsze - 1);
190 while (zone_idx < nr_zones && sector < get_capacity(ns->disk)) {
191 memset(report, 0, buflen);
193 c.zmr.slba = cpu_to_le64(nvme_sect_to_lba(ns, sector));
194 ret = nvme_submit_sync_cmd(ns->queue, &c, report, buflen);
201 nz = min((unsigned int)le64_to_cpu(report->nr_zones), nr_zones);
205 for (i = 0; i < nz && zone_idx < nr_zones; i++) {
206 ret = nvme_zone_parse_entry(ns, &report->entries[i],
213 sector += ns->zsze * nz;
225 int nvme_report_zones(struct gendisk *disk, sector_t sector,
226 unsigned int nr_zones, report_zones_cb cb, void *data)
228 struct nvme_ns_head *head = NULL;
232 ns = nvme_get_ns_from_disk(disk, &head, &srcu_idx);
236 if (ns->head->ids.csi == NVME_CSI_ZNS)
237 ret = nvme_ns_report_zones(ns, sector, nr_zones, cb, data);
240 nvme_put_ns_from_disk(head, srcu_idx);
245 blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns, struct request *req,
246 struct nvme_command *c, enum nvme_zone_mgmt_action action)
248 c->zms.opcode = nvme_cmd_zone_mgmt_send;
249 c->zms.nsid = cpu_to_le32(ns->head->ns_id);
250 c->zms.slba = cpu_to_le64(nvme_sect_to_lba(ns, blk_rq_pos(req)));
253 if (req_op(req) == REQ_OP_ZONE_RESET_ALL)
254 c->zms.select_all = 1;