From: Klaus Jensen <k.jen...@samsung.com>
Convert zone reset from ad-hoc multi aio tracking to use standard QEMU
AIOCB processing.
Signed-off-by: Klaus Jensen <k.jen...@samsung.com>
---
hw/block/nvme.c | 233 +++++++++++++++++++++++++++++-------------------
1 file changed, 139 insertions(+), 94 deletions(-)
diff --git a/hw/block/nvme.c b/hw/block/nvme.c
index 773b41527c79..0d067f186ed7 100644
--- a/hw/block/nvme.c
+++ b/hw/block/nvme.c
@@ -1325,6 +1325,29 @@ static uint16_t nvme_zrm_close(NvmeNamespace *ns,
NvmeZone *zone)
}
}
+static uint16_t nvme_zrm_reset(NvmeNamespace *ns, NvmeZone *zone)
+{
+ switch (nvme_get_zone_state(zone)) {
+ case NVME_ZONE_STATE_EXPLICITLY_OPEN:
+ case NVME_ZONE_STATE_IMPLICITLY_OPEN:
+ nvme_aor_dec_open(ns);
+ /* fallthrough */
+ case NVME_ZONE_STATE_CLOSED:
+ nvme_aor_dec_active(ns);
+ /* fallthrough */
+ case NVME_ZONE_STATE_FULL:
+ zone->w_ptr = zone->d.zslba;
+ zone->d.wp = zone->w_ptr;
+ nvme_assign_zone_state(ns, zone, NVME_ZONE_STATE_EMPTY);
+ /* fallthrough */
+ case NVME_ZONE_STATE_EMPTY:
+ return NVME_SUCCESS;
+
+ default:
+ return NVME_ZONE_INVAL_TRANSITION;
+ }
+}
+
static void nvme_zrm_auto_transition_zone(NvmeNamespace *ns)
{
NvmeZone *zone;
@@ -1472,53 +1495,6 @@ static void nvme_misc_cb(void *opaque, int ret)
nvme_enqueue_req_completion(nvme_cq(req), req);
}
-struct nvme_zone_reset_ctx {
- NvmeRequest *req;
- NvmeZone *zone;
-};
-
-static void nvme_aio_zone_reset_cb(void *opaque, int ret)
-{
- struct nvme_zone_reset_ctx *ctx = opaque;
- NvmeRequest *req = ctx->req;
- NvmeNamespace *ns = req->ns;
- NvmeZone *zone = ctx->zone;
- uintptr_t *resets = (uintptr_t *)&req->opaque;
-
- g_free(ctx);
-
- trace_pci_nvme_aio_zone_reset_cb(nvme_cid(req), zone->d.zslba);
-
- if (!ret) {
- switch (nvme_get_zone_state(zone)) {
- case NVME_ZONE_STATE_EXPLICITLY_OPEN:
- case NVME_ZONE_STATE_IMPLICITLY_OPEN:
- nvme_aor_dec_open(ns);
- /* fall through */
- case NVME_ZONE_STATE_CLOSED:
- nvme_aor_dec_active(ns);
- /* fall through */
- case NVME_ZONE_STATE_FULL:
- zone->w_ptr = zone->d.zslba;
- zone->d.wp = zone->w_ptr;
- nvme_assign_zone_state(ns, zone, NVME_ZONE_STATE_EMPTY);
- /* fall through */
- default:
- break;
- }
- } else {
- nvme_aio_err(req, ret);
- }
-
- (*resets)--;
-
- if (*resets) {
- return;
- }
-
- nvme_enqueue_req_completion(nvme_cq(req), req);
-}
-
struct nvme_compare_ctx {
QEMUIOVector iov;
uint8_t *bounce;
@@ -2336,41 +2312,6 @@ static uint16_t nvme_finish_zone(NvmeNamespace *ns,
NvmeZone *zone,
return nvme_zrm_finish(ns, zone);
}
-static uint16_t nvme_reset_zone(NvmeNamespace *ns, NvmeZone *zone,
- NvmeZoneState state, NvmeRequest *req)
-{
- uintptr_t *resets = (uintptr_t *)&req->opaque;
- struct nvme_zone_reset_ctx *ctx;
-
- switch (state) {
- case NVME_ZONE_STATE_EMPTY:
- return NVME_SUCCESS;
- case NVME_ZONE_STATE_EXPLICITLY_OPEN:
- case NVME_ZONE_STATE_IMPLICITLY_OPEN:
- case NVME_ZONE_STATE_CLOSED:
- case NVME_ZONE_STATE_FULL:
- break;
- default:
- return NVME_ZONE_INVAL_TRANSITION;
- }
-
- /*
- * The zone reset aio callback needs to know the zone that is being reset
- * in order to transition the zone on completion.
- */
- ctx = g_new(struct nvme_zone_reset_ctx, 1);
- ctx->req = req;
- ctx->zone = zone;
-
- (*resets)++;
-
- blk_aio_pwrite_zeroes(ns->blkconf.blk, nvme_l2b(ns, zone->d.zslba),
- nvme_l2b(ns, ns->zone_size), BDRV_REQ_MAY_UNMAP,
- nvme_aio_zone_reset_cb, ctx);
-
- return NVME_NO_COMPLETE;
-}
-
static uint16_t nvme_offline_zone(NvmeNamespace *ns, NvmeZone *zone,
NvmeZoneState state, NvmeRequest *req)
{
@@ -2499,12 +2440,115 @@ out:
return status;
}
+typedef struct NvmeZoneResetAIOCB {
+ BlockAIOCB common;
+ BlockAIOCB *aiocb;
+ NvmeRequest *req;
+ QEMUBH *bh;
+ int ret;
+
+ bool all;
+
+ struct {
+ int idx;
+ NvmeZone *zone;
+ } curr;
+} NvmeZoneResetAIOCB;
+
+static void nvme_zone_reset_cancel(BlockAIOCB *aiocb)
+{
+ NvmeZoneResetAIOCB *iocb = container_of(aiocb, NvmeZoneResetAIOCB, common);
+ NvmeRequest *req = iocb->req;
+ NvmeNamespace *ns = req->ns;
+
+ iocb->curr.idx = ns->num_zones;
+
+ iocb->ret = -ECANCELED;
+
+ if (iocb->aiocb) {
+ blk_aio_cancel_async(iocb->aiocb);
+ iocb->aiocb = NULL;
+ }
+}
+
+static const AIOCBInfo nvme_zone_reset_aiocb_info = {
+ .aiocb_size = sizeof(NvmeZoneResetAIOCB),
+ .cancel_async = nvme_zone_reset_cancel,
+};
+
+static void nvme_zone_reset_bh(void *opaque)
+{
+ NvmeZoneResetAIOCB *iocb = opaque;
+
+ iocb->common.cb(iocb->common.opaque, iocb->ret);
+
+ qemu_bh_delete(iocb->bh);
+ iocb->bh = NULL;
+ qemu_aio_unref(iocb);
+}
+
+static void nvme_zone_reset_aio_cb(void *opaque, int ret)
+{
+ NvmeZoneResetAIOCB *iocb = opaque;
+ NvmeRequest *req = iocb->req;
+ NvmeNamespace *ns = req->ns;
+
+ if (ret < 0) {
+ iocb->ret = ret;
+ goto done;
+ }
+
+ if (iocb->curr.zone) {
+ nvme_zrm_reset(ns, iocb->curr.zone);
+
+ if (!iocb->all) {
+ goto done;
+ }
+ }
+
+ while (iocb->curr.idx < ns->num_zones) {
+ NvmeZone *zone = &ns->zone_array[iocb->curr.idx++];
+
+ switch (nvme_get_zone_state(zone)) {
+ case NVME_ZONE_STATE_EMPTY:
+ if (!iocb->all) {
+ goto done;
+ }
+
+ continue;
+
+ case NVME_ZONE_STATE_EXPLICITLY_OPEN:
+ case NVME_ZONE_STATE_IMPLICITLY_OPEN:
+ case NVME_ZONE_STATE_CLOSED:
+ case NVME_ZONE_STATE_FULL:
+ iocb->curr.zone = zone;
+ break;
+
+ default:
+ continue;
+ }
+
+ iocb->aiocb = blk_aio_pwrite_zeroes(ns->blkconf.blk,
+ nvme_l2b(ns, zone->d.zslba),
+ nvme_l2b(ns, ns->zone_size),
+ BDRV_REQ_MAY_UNMAP,
+ nvme_zone_reset_aio_cb, iocb);
+ return;
+ }
+
+done:
+ iocb->aiocb = NULL;
+ if (iocb->bh) {
+ qemu_bh_schedule(iocb->bh);
+ }
+}
+
static uint16_t nvme_zone_mgmt_send(NvmeCtrl *n, NvmeRequest *req)
{
NvmeCmd *cmd = (NvmeCmd *)&req->cmd;
NvmeNamespace *ns = req->ns;
NvmeZone *zone;
- uintptr_t *resets;
+ NvmeZoneResetAIOCB *iocb;
uint8_t *zd_ext;
uint32_t dw13 = le32_to_cpu(cmd->cdw13);
uint64_t slba = 0;
@@ -2515,7 +2559,7 @@ static uint16_t nvme_zone_mgmt_send(NvmeCtrl *n,
NvmeRequest *req)
enum NvmeZoneProcessingMask proc_mask = NVME_PROC_CURRENT_ZONE;
action = dw13 & 0xff;
- all = dw13 & 0x100;
+ all = !!(dw13 & 0x100);
req->status = NVME_SUCCESS;
@@ -2559,21 +2603,22 @@ static uint16_t nvme_zone_mgmt_send(NvmeCtrl *n,
NvmeRequest *req)
break;
case NVME_ZONE_ACTION_RESET:
- resets = (uintptr_t *)&req->opaque;
-
- if (all) {
- proc_mask = NVME_PROC_OPENED_ZONES | NVME_PROC_CLOSED_ZONES |
- NVME_PROC_FULL_ZONES;
- }
trace_pci_nvme_reset_zone(slba, zone_idx, all);
- *resets = 1;
+ iocb = blk_aio_get(&nvme_zone_reset_aiocb_info, ns->blkconf.blk,
+ nvme_misc_cb, req);
- status = nvme_do_zone_op(ns, zone, proc_mask, nvme_reset_zone, req);
+ iocb->req = req;
+ iocb->bh = qemu_bh_new(nvme_zone_reset_bh, iocb);
+ iocb->ret = 0;
+ iocb->all = all;
+ iocb->curr.idx = zone_idx;
+ iocb->curr.zone = NULL;
- (*resets)--;
+ nvme_zone_reset_aio_cb(iocb, 0);
+ req->aiocb = &iocb->common;
- return *resets ? NVME_NO_COMPLETE : req->status;
+ return NVME_NO_COMPLETE;
case NVME_ZONE_ACTION_OFFLINE:
if (all) {
--
2.30.1
